/ Check-in [ad08794d]
Login

Many hyperlinks are disabled.
Use anonymous login to enable hyperlinks.

Overview
Comment:Merge all changes associated with the version 3.6.21 release into the OS-X branch.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | apple-osx
Files: files | file ages | folders
SHA1:ad08794d7239bb804eb39c00170821732b60cb65
User & Date: drh 2009-12-07 23:53:52
Context
2009-12-16
23:46
Merge the latest changes on trunk (and especially the fix for the (xANDy)OR(z) bug) into apple-osx. check-in: 5754a3a5 user: drh tags: apple-osx
2009-12-07
23:53
Merge all changes associated with the version 3.6.21 release into the OS-X branch. check-in: ad08794d user: drh tags: apple-osx
16:39
Version 3.6.21 check-in: 1ed88e9d user: drh tags: trunk, release
2009-11-09
19:30
Fix for lock structure sharing with AFP-style locking check-in: 62f15c0a user: adam tags: apple-osx
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to Makefile.in.

171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
...
268
269
270
271
272
273
274

275
276
277
278
279
280
281
...
307
308
309
310
311
312
313
314
315
316
317
318
319

320
321
322

323
324
325
326
327
328
329
...
358
359
360
361
362
363
364

365
366
367
368
369
370
371
...
382
383
384
385
386
387
388

389
390
391
392
393
394
395
...
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
...
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
...
734
735
736
737
738
739
740



741
742
743
744
745
746
747
        memjournal.lo \
        mutex.lo mutex_noop.lo mutex_os2.lo mutex_unix.lo mutex_w32.lo \
        notify.lo opcodes.lo os.lo os_unix.lo os_win.lo os_os2.lo \
        pager.lo parse.lo pcache.lo pcache1.lo pragma.lo prepare.lo printf.lo \
        random.lo resolve.lo rowset.lo select.lo status.lo \
        table.lo tokenize.lo trigger.lo update.lo \
        util.lo vacuum.lo \
        vdbe.lo vdbeapi.lo vdbeaux.lo vdbeblob.lo vdbemem.lo \
        walker.lo where.lo utf.lo vtab.lo

# Object files for the amalgamation.
#
OBJS1 = sqlite3.lo

# Determine the real value of LIBOBJ based on the 'configure' script
................................................................................
  $(TOP)/src/vacuum.c \
  $(TOP)/src/vdbe.c \
  $(TOP)/src/vdbe.h \
  $(TOP)/src/vdbeapi.c \
  $(TOP)/src/vdbeaux.c \
  $(TOP)/src/vdbeblob.c \
  $(TOP)/src/vdbemem.c \

  $(TOP)/src/vdbeInt.h \
  $(TOP)/src/vtab.c \
  $(TOP)/src/walker.c \
  $(TOP)/src/where.c

# Generated source code files
#
................................................................................
  $(TOP)/ext/fts2/fts2_porter.c \
  $(TOP)/ext/fts2/fts2_tokenizer.h \
  $(TOP)/ext/fts2/fts2_tokenizer.c \
  $(TOP)/ext/fts2/fts2_tokenizer1.c
SRC += \
  $(TOP)/ext/fts3/fts3.c \
  $(TOP)/ext/fts3/fts3.h \
  $(TOP)/ext/fts3/fts3_expr.c \
  $(TOP)/ext/fts3/fts3_expr.h \
  $(TOP)/ext/fts3/fts3_hash.c \
  $(TOP)/ext/fts3/fts3_hash.h \
  $(TOP)/ext/fts3/fts3_icu.c \
  $(TOP)/ext/fts3/fts3_porter.c \

  $(TOP)/ext/fts3/fts3_tokenizer.h \
  $(TOP)/ext/fts3/fts3_tokenizer.c \
  $(TOP)/ext/fts3/fts3_tokenizer1.c

SRC += \
  $(TOP)/ext/icu/sqliteicu.h \
  $(TOP)/ext/icu/icu.c
SRC += \
  $(TOP)/ext/rtree/rtree.h \
  $(TOP)/ext/rtree/rtree.c
SRC += \
................................................................................
  $(TOP)/src/tokenize.c \
  $(TOP)/src/utf.c \
  $(TOP)/src/util.c \
  $(TOP)/src/vdbe.c \
  $(TOP)/src/vdbeapi.c \
  $(TOP)/src/vdbeaux.c \
  $(TOP)/src/vdbemem.c \

  $(TOP)/src/where.c \
  parse.c

# Source code to the actual test files.
#
TESTSRC = \
  $(TOP)/src/test1.c \
................................................................................
  $(TOP)/src/test_backup.c \
  $(TOP)/src/test_btree.c \
  $(TOP)/src/test_config.c \
  $(TOP)/src/test_devsym.c \
  $(TOP)/src/test_func.c \
  $(TOP)/src/test_hexio.c \
  $(TOP)/src/test_init.c \

  $(TOP)/src/test_journal.c \
  $(TOP)/src/test_malloc.c \
  $(TOP)/src/test_mutex.c \
  $(TOP)/src/test_onefile.c \
  $(TOP)/src/test_osinst.c \
  $(TOP)/src/test_pcache.c \
  $(TOP)/src/test_schema.c \
................................................................................
  $(TOP)/ext/fts1/fts1_tokenizer.h
HDR += \
  $(TOP)/ext/fts2/fts2.h \
  $(TOP)/ext/fts2/fts2_hash.h \
  $(TOP)/ext/fts2/fts2_tokenizer.h
HDR += \
  $(TOP)/ext/fts3/fts3.h \
  $(TOP)/ext/fts3/fts3_expr.h \
  $(TOP)/ext/fts3/fts3_hash.h \
  $(TOP)/ext/fts3/fts3_tokenizer.h
HDR += \
  $(TOP)/ext/rtree/rtree.h
HDR += \
  $(TOP)/ext/icu/sqliteicu.h
HDR += \
  $(TOP)/ext/sqlrr/sqlrr.h
  
# If using the amalgamation, use sqlite3.c directly to build the test
# fixture.  Otherwise link against libsqlite3.la.  (This distinction is
# necessary because the test fixture requires non-API symbols which are
# hidden when the library is built via the amalgamation).
#
TESTFIXTURE_SRC0 = $(TESTSRC2) libsqlite3.la
TESTFIXTURE_SRC1 = sqlite3.c
................................................................................

Makefile: $(TOP)/Makefile.in
	./config.status

sqlite3.pc: $(TOP)/sqlite3.pc.in
	./config.status

# Generate the file "last_change" which contains the date of change
# of the most recently modified source code file
#
last_change:	$(SRC)
	cat $(SRC) | grep '$$Id: ' | sort -k 5 | tail -1 \
          | $(NAWK) '{print $$5,$$6}' >last_change

libsqlite3.la:	$(LIBOBJ)
	$(LTLINK) -o $@ $(LIBOBJ) $(TLIBS) \
		${ALLOWRELEASE} -rpath "$(libdir)" -version-info "8:6:8"

libtclsqlite3.la:	tclsqlite.lo libsqlite3.la
	$(LTLINK) -o $@ tclsqlite.lo \
		libsqlite3.la @TCL_STUB_LIB_SPEC@ $(TLIBS) \
................................................................................

vdbeblob.lo:	$(TOP)/src/vdbeblob.c $(HDR)
	$(LTCOMPILE) $(TEMP_STORE) -c $(TOP)/src/vdbeblob.c

vdbemem.lo:	$(TOP)/src/vdbemem.c $(HDR)
	$(LTCOMPILE) $(TEMP_STORE) -c $(TOP)/src/vdbemem.c




vtab.lo:	$(TOP)/src/vtab.c $(HDR)
	$(LTCOMPILE) $(TEMP_STORE) -c $(TOP)/src/vtab.c

walker.lo:	$(TOP)/src/walker.c $(HDR)
	$(LTCOMPILE) $(TEMP_STORE) -c $(TOP)/src/walker.c

where.lo:	$(TOP)/src/where.c $(HDR)







|







 







>







 







|
|




>


|
>







 







>







 







>







 







|








|







 







<
<
<
<
<
<
<







 







>
>
>







171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
...
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
...
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
...
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
...
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
...
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
...
462
463
464
465
466
467
468







469
470
471
472
473
474
475
...
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
        memjournal.lo \
        mutex.lo mutex_noop.lo mutex_os2.lo mutex_unix.lo mutex_w32.lo \
        notify.lo opcodes.lo os.lo os_unix.lo os_win.lo os_os2.lo \
        pager.lo parse.lo pcache.lo pcache1.lo pragma.lo prepare.lo printf.lo \
        random.lo resolve.lo rowset.lo select.lo status.lo \
        table.lo tokenize.lo trigger.lo update.lo \
        util.lo vacuum.lo \
        vdbe.lo vdbeapi.lo vdbeaux.lo vdbeblob.lo vdbemem.lo vdbetrace.lo \
        walker.lo where.lo utf.lo vtab.lo

# Object files for the amalgamation.
#
OBJS1 = sqlite3.lo

# Determine the real value of LIBOBJ based on the 'configure' script
................................................................................
  $(TOP)/src/vacuum.c \
  $(TOP)/src/vdbe.c \
  $(TOP)/src/vdbe.h \
  $(TOP)/src/vdbeapi.c \
  $(TOP)/src/vdbeaux.c \
  $(TOP)/src/vdbeblob.c \
  $(TOP)/src/vdbemem.c \
  $(TOP)/src/vdbetrace.c \
  $(TOP)/src/vdbeInt.h \
  $(TOP)/src/vtab.c \
  $(TOP)/src/walker.c \
  $(TOP)/src/where.c

# Generated source code files
#
................................................................................
  $(TOP)/ext/fts2/fts2_porter.c \
  $(TOP)/ext/fts2/fts2_tokenizer.h \
  $(TOP)/ext/fts2/fts2_tokenizer.c \
  $(TOP)/ext/fts2/fts2_tokenizer1.c
SRC += \
  $(TOP)/ext/fts3/fts3.c \
  $(TOP)/ext/fts3/fts3.h \
  $(TOP)/ext/fts3/fts3Int.h \
  $(TOP)/ext/fts3/fts3_expr.c \
  $(TOP)/ext/fts3/fts3_hash.c \
  $(TOP)/ext/fts3/fts3_hash.h \
  $(TOP)/ext/fts3/fts3_icu.c \
  $(TOP)/ext/fts3/fts3_porter.c \
  $(TOP)/ext/fts3/fts3_snippet.c \
  $(TOP)/ext/fts3/fts3_tokenizer.h \
  $(TOP)/ext/fts3/fts3_tokenizer.c \
  $(TOP)/ext/fts3/fts3_tokenizer1.c \
  $(TOP)/ext/fts3/fts3_write.c
SRC += \
  $(TOP)/ext/icu/sqliteicu.h \
  $(TOP)/ext/icu/icu.c
SRC += \
  $(TOP)/ext/rtree/rtree.h \
  $(TOP)/ext/rtree/rtree.c
SRC += \
................................................................................
  $(TOP)/src/tokenize.c \
  $(TOP)/src/utf.c \
  $(TOP)/src/util.c \
  $(TOP)/src/vdbe.c \
  $(TOP)/src/vdbeapi.c \
  $(TOP)/src/vdbeaux.c \
  $(TOP)/src/vdbemem.c \
  $(TOP)/src/vdbetrace.c \
  $(TOP)/src/where.c \
  parse.c

# Source code to the actual test files.
#
TESTSRC = \
  $(TOP)/src/test1.c \
................................................................................
  $(TOP)/src/test_backup.c \
  $(TOP)/src/test_btree.c \
  $(TOP)/src/test_config.c \
  $(TOP)/src/test_devsym.c \
  $(TOP)/src/test_func.c \
  $(TOP)/src/test_hexio.c \
  $(TOP)/src/test_init.c \
  $(TOP)/src/test_intarray.c \
  $(TOP)/src/test_journal.c \
  $(TOP)/src/test_malloc.c \
  $(TOP)/src/test_mutex.c \
  $(TOP)/src/test_onefile.c \
  $(TOP)/src/test_osinst.c \
  $(TOP)/src/test_pcache.c \
  $(TOP)/src/test_schema.c \
................................................................................
  $(TOP)/ext/fts1/fts1_tokenizer.h
HDR += \
  $(TOP)/ext/fts2/fts2.h \
  $(TOP)/ext/fts2/fts2_hash.h \
  $(TOP)/ext/fts2/fts2_tokenizer.h
HDR += \
  $(TOP)/ext/fts3/fts3.h \
  $(TOP)/ext/fts3/fts3Int.h \
  $(TOP)/ext/fts3/fts3_hash.h \
  $(TOP)/ext/fts3/fts3_tokenizer.h
HDR += \
  $(TOP)/ext/rtree/rtree.h
HDR += \
  $(TOP)/ext/icu/sqliteicu.h
HDR += \
  $(TOP)/ext/sqlrr/sqlrr.h

# If using the amalgamation, use sqlite3.c directly to build the test
# fixture.  Otherwise link against libsqlite3.la.  (This distinction is
# necessary because the test fixture requires non-API symbols which are
# hidden when the library is built via the amalgamation).
#
TESTFIXTURE_SRC0 = $(TESTSRC2) libsqlite3.la
TESTFIXTURE_SRC1 = sqlite3.c
................................................................................

Makefile: $(TOP)/Makefile.in
	./config.status

sqlite3.pc: $(TOP)/sqlite3.pc.in
	./config.status








libsqlite3.la:	$(LIBOBJ)
	$(LTLINK) -o $@ $(LIBOBJ) $(TLIBS) \
		${ALLOWRELEASE} -rpath "$(libdir)" -version-info "8:6:8"

libtclsqlite3.la:	tclsqlite.lo libsqlite3.la
	$(LTLINK) -o $@ tclsqlite.lo \
		libsqlite3.la @TCL_STUB_LIB_SPEC@ $(TLIBS) \
................................................................................

vdbeblob.lo:	$(TOP)/src/vdbeblob.c $(HDR)
	$(LTCOMPILE) $(TEMP_STORE) -c $(TOP)/src/vdbeblob.c

vdbemem.lo:	$(TOP)/src/vdbemem.c $(HDR)
	$(LTCOMPILE) $(TEMP_STORE) -c $(TOP)/src/vdbemem.c

vdbetrace.lo:	$(TOP)/src/vdbetrace.c $(HDR)
	$(LTCOMPILE) $(TEMP_STORE) -c $(TOP)/src/vdbetrace.c

vtab.lo:	$(TOP)/src/vtab.c $(HDR)
	$(LTCOMPILE) $(TEMP_STORE) -c $(TOP)/src/vtab.c

walker.lo:	$(TOP)/src/walker.c $(HDR)
	$(LTCOMPILE) $(TEMP_STORE) -c $(TOP)/src/walker.c

where.lo:	$(TOP)/src/where.c $(HDR)

Changes to VERSION.

1
3.6.20
|
1
3.6.21

Changes to configure.

1
2
3
4
5
6
7
8
9
10
...
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
....
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
....
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
....
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
.....
13968
13969
13970
13971
13972
13973
13974
13975
13976
13977
13978
13979
13980
13981
13982
.....
14021
14022
14023
14024
14025
14026
14027
14028
14029
14030
14031
14032
14033
14034
14035
#! /bin/sh
# Guess values for system-dependent variables and create Makefiles.
# Generated by GNU Autoconf 2.62 for sqlite 3.6.20.
#
# Copyright (C) 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001,
# 2002, 2003, 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
# This configure script is free software; the Free Software Foundation
# gives unlimited permission to copy, distribute and modify it.
## --------------------- ##
## M4sh Initialization.  ##
................................................................................
MFLAGS=
MAKEFLAGS=
SHELL=${CONFIG_SHELL-/bin/sh}

# Identity of this package.
PACKAGE_NAME='sqlite'
PACKAGE_TARNAME='sqlite'
PACKAGE_VERSION='3.6.20'
PACKAGE_STRING='sqlite 3.6.20'
PACKAGE_BUGREPORT=''

# Factoring default headers for most tests.
ac_includes_default="\
#include <stdio.h>
#ifdef HAVE_SYS_TYPES_H
# include <sys/types.h>
................................................................................
#
# Report the --help message.
#
if test "$ac_init_help" = "long"; then
  # Omit some internal or obsolete options to make the list less imposing.
  # This message is too long to be a string in the A/UX 3.1 sh.
  cat <<_ACEOF
\`configure' configures sqlite 3.6.20 to adapt to many kinds of systems.

Usage: $0 [OPTION]... [VAR=VALUE]...

To assign environment variables (e.g., CC, CFLAGS...), specify them as
VAR=VALUE.  See below for descriptions of some of the useful variables.

Defaults for the options are specified in brackets.
................................................................................
  --build=BUILD     configure for building on BUILD [guessed]
  --host=HOST       cross-compile to build programs to run on HOST [BUILD]
_ACEOF
fi

if test -n "$ac_init_help"; then
  case $ac_init_help in
     short | recursive ) echo "Configuration of sqlite 3.6.20:";;
   esac
  cat <<\_ACEOF

Optional Features:
  --disable-option-checking  ignore unrecognized --enable/--with options
  --disable-FEATURE       do not include FEATURE (same as --enable-FEATURE=no)
  --enable-FEATURE[=ARG]  include FEATURE [ARG=yes]
................................................................................
    cd "$ac_pwd" || { ac_status=$?; break; }
  done
fi

test -n "$ac_init_help" && exit $ac_status
if $ac_init_version; then
  cat <<\_ACEOF
sqlite configure 3.6.20
generated by GNU Autoconf 2.62

Copyright (C) 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001,
2002, 2003, 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
This configure script is free software; the Free Software Foundation
gives unlimited permission to copy, distribute and modify it.
_ACEOF
  exit
fi
cat >config.log <<_ACEOF
This file contains any messages produced by compilers while
running configure, to aid debugging if configure makes a mistake.

It was created by sqlite $as_me 3.6.20, which was
generated by GNU Autoconf 2.62.  Invocation command line was

  $ $0 $@

_ACEOF
exec 5>>config.log
{
................................................................................

exec 6>&1

# Save the log message, to keep $[0] and so on meaningful, and to
# report actual input values of CONFIG_FILES etc. instead of their
# values after options handling.
ac_log="
This file was extended by sqlite $as_me 3.6.20, which was
generated by GNU Autoconf 2.62.  Invocation command line was

  CONFIG_FILES    = $CONFIG_FILES
  CONFIG_HEADERS  = $CONFIG_HEADERS
  CONFIG_LINKS    = $CONFIG_LINKS
  CONFIG_COMMANDS = $CONFIG_COMMANDS
  $ $0 $@
................................................................................
$config_commands

Report bugs to <bug-autoconf@gnu.org>."

_ACEOF
cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
ac_cs_version="\\
sqlite config.status 3.6.20
configured by $0, generated by GNU Autoconf 2.62,
  with options \\"`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`\\"

Copyright (C) 2008 Free Software Foundation, Inc.
This config.status script is free software; the Free Software Foundation
gives unlimited permission to copy, distribute and modify it."



|







 







|
|







 







|







 







|







 







|













|







 







|







 







|







1
2
3
4
5
6
7
8
9
10
...
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
....
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
....
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
....
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
.....
13968
13969
13970
13971
13972
13973
13974
13975
13976
13977
13978
13979
13980
13981
13982
.....
14021
14022
14023
14024
14025
14026
14027
14028
14029
14030
14031
14032
14033
14034
14035
#! /bin/sh
# Guess values for system-dependent variables and create Makefiles.
# Generated by GNU Autoconf 2.62 for sqlite 3.6.21.
#
# Copyright (C) 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001,
# 2002, 2003, 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
# This configure script is free software; the Free Software Foundation
# gives unlimited permission to copy, distribute and modify it.
## --------------------- ##
## M4sh Initialization.  ##
................................................................................
MFLAGS=
MAKEFLAGS=
SHELL=${CONFIG_SHELL-/bin/sh}

# Identity of this package.
PACKAGE_NAME='sqlite'
PACKAGE_TARNAME='sqlite'
PACKAGE_VERSION='3.6.21'
PACKAGE_STRING='sqlite 3.6.21'
PACKAGE_BUGREPORT=''

# Factoring default headers for most tests.
ac_includes_default="\
#include <stdio.h>
#ifdef HAVE_SYS_TYPES_H
# include <sys/types.h>
................................................................................
#
# Report the --help message.
#
if test "$ac_init_help" = "long"; then
  # Omit some internal or obsolete options to make the list less imposing.
  # This message is too long to be a string in the A/UX 3.1 sh.
  cat <<_ACEOF
\`configure' configures sqlite 3.6.21 to adapt to many kinds of systems.

Usage: $0 [OPTION]... [VAR=VALUE]...

To assign environment variables (e.g., CC, CFLAGS...), specify them as
VAR=VALUE.  See below for descriptions of some of the useful variables.

Defaults for the options are specified in brackets.
................................................................................
  --build=BUILD     configure for building on BUILD [guessed]
  --host=HOST       cross-compile to build programs to run on HOST [BUILD]
_ACEOF
fi

if test -n "$ac_init_help"; then
  case $ac_init_help in
     short | recursive ) echo "Configuration of sqlite 3.6.21:";;
   esac
  cat <<\_ACEOF

Optional Features:
  --disable-option-checking  ignore unrecognized --enable/--with options
  --disable-FEATURE       do not include FEATURE (same as --enable-FEATURE=no)
  --enable-FEATURE[=ARG]  include FEATURE [ARG=yes]
................................................................................
    cd "$ac_pwd" || { ac_status=$?; break; }
  done
fi

test -n "$ac_init_help" && exit $ac_status
if $ac_init_version; then
  cat <<\_ACEOF
sqlite configure 3.6.21
generated by GNU Autoconf 2.62

Copyright (C) 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001,
2002, 2003, 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
This configure script is free software; the Free Software Foundation
gives unlimited permission to copy, distribute and modify it.
_ACEOF
  exit
fi
cat >config.log <<_ACEOF
This file contains any messages produced by compilers while
running configure, to aid debugging if configure makes a mistake.

It was created by sqlite $as_me 3.6.21, which was
generated by GNU Autoconf 2.62.  Invocation command line was

  $ $0 $@

_ACEOF
exec 5>>config.log
{
................................................................................

exec 6>&1

# Save the log message, to keep $[0] and so on meaningful, and to
# report actual input values of CONFIG_FILES etc. instead of their
# values after options handling.
ac_log="
This file was extended by sqlite $as_me 3.6.21, which was
generated by GNU Autoconf 2.62.  Invocation command line was

  CONFIG_FILES    = $CONFIG_FILES
  CONFIG_HEADERS  = $CONFIG_HEADERS
  CONFIG_LINKS    = $CONFIG_LINKS
  CONFIG_COMMANDS = $CONFIG_COMMANDS
  $ $0 $@
................................................................................
$config_commands

Report bugs to <bug-autoconf@gnu.org>."

_ACEOF
cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
ac_cs_version="\\
sqlite config.status 3.6.21
configured by $0, generated by GNU Autoconf 2.62,
  with options \\"`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`\\"

Copyright (C) 2008 Free Software Foundation, Inc.
This config.status script is free software; the Free Software Foundation
gives unlimited permission to copy, distribute and modify it."

Changes to ext/async/sqlite3async.c.

1227
1228
1229
1230
1231
1232
1233
1234
1235

1236
1237
1238
1239
1240
1241
1242

  /* Because of the way intra-process file locking works, this backend
  ** needs to return a canonical path. The following block assumes the
  ** file-system uses unix style paths. 
  */
  if( rc==SQLITE_OK ){
    int i, j;
    int n = nPathOut;
    char *z = zPathOut;

    while( n>1 && z[n-1]=='/' ){ n--; }
    for(i=j=0; i<n; i++){
      if( z[i]=='/' ){
        if( z[i+1]=='/' ) continue;
        if( z[i+1]=='.' && i+2<n && z[i+2]=='/' ){
          i += 1;
          continue;







<

>







1227
1228
1229
1230
1231
1232
1233

1234
1235
1236
1237
1238
1239
1240
1241
1242

  /* Because of the way intra-process file locking works, this backend
  ** needs to return a canonical path. The following block assumes the
  ** file-system uses unix style paths. 
  */
  if( rc==SQLITE_OK ){
    int i, j;

    char *z = zPathOut;
    int n = strlen(z);
    while( n>1 && z[n-1]=='/' ){ n--; }
    for(i=j=0; i<n; i++){
      if( z[i]=='/' ){
        if( z[i+1]=='/' ) continue;
        if( z[i+1]=='.' && i+2<n && z[i+2]=='/' ){
          i += 1;
          continue;

Changes to ext/fts3/fts3.c.

274
275
276
277
278
279
280


281
282

283
284
285
286
287
288
289
290
291
292
293
294

295
296
297
298
299
300
301
...
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
...
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
...
376
377
378
379
380
381
382
383
384
385
386
387
388

389
390
391
392
393
394
395
396
397
398
399
400

401
402
403

404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419




420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395


1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954

1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516

2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573

2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
....
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616

2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747























































































































































































3748
3749
3750






3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761

3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777

3778



3779
3780
3781
3782
3783
3784
3785



3786
3787
3788
3789
3790





3791
3792

3793
3794
3795
3796
3797

3798
3799
3800
3801

3802




3803
3804




3805
3806
3807

3808
3809

3810
3811











3812
3813
3814
3815







3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
5336
5337
5338
5339
5340
5341
5342
5343
5344
5345
5346
5347
5348
5349
5350
5351
5352
5353
5354
5355
5356
5357
5358
5359
5360
5361
5362
5363
5364
5365
5366
5367
5368
5369
5370
5371
5372
5373
5374
5375
5376
5377
5378
5379
5380
5381
5382
5383
5384
5385
5386
5387
5388
5389
5390
5391
5392
5393
5394
5395
5396
5397
5398
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410
5411
5412
5413
5414
5415
5416
5417
5418
5419
5420
5421
5422
5423
5424
5425
5426
5427
5428
5429
5430
5431
5432
5433
5434
5435
5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453
5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466
5467
5468
5469
5470
5471
5472
5473
5474
5475
5476
5477
5478
5479
5480
5481
5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495
5496
5497
5498
5499
5500
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522
5523
5524
5525
5526
5527
5528
5529
5530
5531
5532
5533
5534
5535
5536
5537
5538
5539
5540
5541
5542
5543
5544
5545
5546
5547
5548
5549
5550
5551
5552
5553
5554
5555
5556
5557
5558
5559
5560
5561
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
5621
5622
5623
5624
5625
5626
5627
5628
5629
5630
5631
5632
5633
5634
5635
5636
5637
5638
5639
5640
5641
5642
5643
5644
5645
5646
5647
5648
5649
5650
5651
5652
5653
5654
5655
5656
5657
5658
5659
5660
5661
5662
5663
5664
5665
5666
5667
5668
5669
5670
5671
5672
5673
5674
5675
5676
5677
5678
5679
5680
5681
5682
5683
5684
5685
5686
5687
5688
5689
5690
5691
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708
5709
5710
5711
5712
5713
5714
5715
5716
5717
5718
5719
5720
5721
5722
5723
5724
5725
5726
5727
5728
5729
5730
5731
5732
5733
5734
5735
5736
5737
5738
5739
5740
5741
5742
5743
5744
5745
5746
5747
5748
5749
5750
5751
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773
5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
5784
5785
5786
5787
5788
5789
5790
5791
5792
5793
5794
5795
5796
5797
5798
5799
5800
5801
5802
5803
5804
5805
5806
5807
5808
5809
5810
5811
5812
5813
5814
5815
5816
5817
5818
5819
5820
5821
5822
5823
5824
5825
5826
5827
5828
5829
5830
5831
5832
5833
5834
5835
5836
5837
5838
5839
5840
5841
5842
5843
5844
5845
5846
5847
5848
5849
5850
5851
5852
5853
5854
5855
5856
5857
5858
5859
5860
5861
5862
5863
5864
5865
5866
5867
5868
5869
5870
5871
5872
5873
5874
5875
5876
5877
5878
5879
5880
5881
5882
5883
5884
5885
5886
5887
5888
5889
5890
5891
5892
5893
5894
5895
5896
5897
5898
5899
5900
5901
5902
5903
5904
5905
5906
5907
5908
5909
5910
5911
5912
5913
5914
5915
5916
5917
5918
5919
5920
5921
5922
5923
5924
5925
5926
5927
5928
5929
5930
5931
5932
5933
5934
5935
5936
5937
5938
5939
5940
5941
5942
5943
5944
5945
5946
5947
5948
5949
5950
5951
5952
5953
5954
5955
5956
5957
5958
5959
5960
5961
5962
5963
5964
5965
5966
5967
5968
5969
5970
5971
5972
5973
5974
5975
5976
5977
5978
5979
5980
5981
5982
5983
5984
5985
5986
5987
5988
5989
5990
5991
5992
5993
5994
5995
5996
5997
5998
5999
6000
6001
6002
6003
6004
6005
6006
6007
6008
6009
6010
6011
6012
6013
6014
6015
6016
6017
6018
6019
6020
6021
6022
6023
6024
6025
6026
6027
6028
6029
6030
6031
6032
6033
6034
6035
6036
6037
6038
6039
6040
6041
6042
6043
6044
6045
6046
6047
6048
6049
6050
6051
6052
6053
6054
6055
6056
6057
6058
6059
6060
6061
6062
6063
6064
6065
6066
6067
6068
6069
6070
6071
6072
6073
6074
6075
6076
6077
6078
6079
6080
6081
6082
6083
6084
6085
6086
6087
6088
6089
6090
6091
6092
6093
6094
6095
6096
6097
6098
6099
6100
6101
6102
6103
6104
6105
6106
6107
6108
6109
6110
6111
6112
6113
6114
6115
6116
6117
6118
6119
6120
6121
6122
6123
6124
6125
6126
6127
6128
6129
6130
6131
6132
6133
6134
6135
6136
6137
6138
6139
6140
6141
6142
6143
6144
6145
6146
6147
6148
6149
6150
6151
6152
6153
6154
6155
6156
6157
6158
6159
6160
6161
6162
6163
6164
6165
6166
6167
6168
6169
6170
6171
6172
6173
6174

6175
6176
6177
6178
6179
6180
6181
6182
6183
6184
6185
6186
6187
6188
6189
6190
6191
6192
6193
6194
6195
6196
6197
6198
6199
6200
6201
6202
6203
6204
6205
6206
6207
6208
6209
6210
6211
6212
6213
6214
6215
6216
6217
6218
6219
6220
6221
6222
6223
6224
6225
6226
6227
6228
6229
6230
6231
6232
6233
6234
6235
6236
6237
6238
6239
6240
6241
6242
6243
6244
6245
6246
6247
6248
6249
6250
6251
6252
6253
6254
6255
6256
6257
6258
6259
6260
6261
6262
6263
6264
6265
6266
6267
6268
6269
6270
6271
6272
6273
6274
6275
6276
6277
6278
6279
6280
6281
6282
6283
6284
6285
6286
6287
6288
6289
6290
6291
6292
6293
6294
6295
6296
6297
6298
6299
6300
6301
6302
6303
6304
6305
6306
6307
6308
6309
6310
6311
6312
6313
6314
6315
6316
6317
6318
6319
6320
6321
6322
6323
6324
6325
6326
6327
6328
6329
6330
6331
6332
6333
6334
6335
6336
6337
6338
6339
6340
6341
6342
6343
6344
6345
6346
6347
6348
6349
6350
6351
6352
6353
6354
6355
6356
6357
6358
6359
6360
6361
6362
6363
6364
6365
6366
6367
6368
6369
6370
6371
6372
6373
6374
6375
6376
6377
6378
6379
6380
6381
6382
6383
6384
6385
6386
6387
6388
6389
6390
6391
6392
6393
6394
6395
6396
6397
6398
6399
6400
6401
6402
6403
6404
6405
6406
6407
6408
6409
6410
6411
6412
6413
6414

6415
6416

6417
6418
6419
6420
6421
6422
6423
6424
6425
6426
6427
6428
6429
6430
6431
6432
6433
6434
6435
6436
6437
6438
6439
6440
6441
6442
6443
6444



6445
6446
6447
6448
6449
6450
6451
6452
6453
6454
6455
6456
6457
6458

6459
6460
6461
6462
6463
6464
6465
6466
6467
6468
6469
6470
6471
6472

6473
6474
6475
6476
6477
6478
6479
6480
6481
6482
6483
6484
6485
6486
6487
6488
6489
6490
6491
6492
6493
6494
6495
6496
6497
6498
6499
6500
6501
6502
6503
6504
6505
6506
6507
6508
6509
6510
6511
6512
6513
6514
6515
6516
6517
6518
6519
6520
6521
6522
6523
6524
6525
6526
6527
6528
6529
6530
6531
6532
6533
6534
6535
6536
6537
6538
6539
6540
6541
6542
6543
6544
6545
6546
6547
6548
6549
6550
6551
6552
6553
6554
6555
6556
6557
6558
6559
6560
6561
6562
6563
6564

6565
6566
6567
6568
6569
6570
6571
6572
6573
6574
6575
6576
6577
6578
6579
6580
6581
6582
6583
6584
6585
6586
6587
6588
6589
6590
6591
6592
6593
6594
6595
6596
6597
6598
6599
6600
6601
6602
6603
6604
6605
6606
6607
6608
6609
6610
6611
6612
6613
6614
6615
6616
6617
6618
6619
6620
6621
6622
6623
6624
6625
6626
6627
6628
6629
6630
6631
6632
6633
6634
6635
6636
6637
6638
6639
6640
6641
6642
6643
6644
6645
6646
6647
6648
6649
6650
6651
6652
6653
6654
6655
6656
6657
6658
6659
6660
6661
6662
6663
6664
6665
6666
6667
6668
6669
6670
6671


6672
6673
6674
6675
6676
6677
6678
6679
6680

6681
6682
6683
6684
6685
6686
6687
6688
6689
6690
6691
6692
6693


6694
6695
6696
6697
6698
6699
6700
6701
6702
6703
6704
6705
6706
6707
6708
6709
6710
6711
6712
6713
6714
6715
6716
6717
6718
6719
6720
6721
6722
6723
6724
6725
6726
6727
6728
6729
6730
6731
6732
6733
6734
6735
6736
6737
6738
6739
6740
6741
6742
6743
6744
6745
6746
6747
6748
6749
6750
6751
6752
6753
6754
6755
6756
6757
6758
6759
6760
6761
6762
6763
6764
6765
6766
6767
6768
6769
6770
6771
6772
6773
6774
6775
6776
6777
6778
6779
6780
6781
6782
6783
6784
6785
6786
6787
6788

6789
6790
6791
6792


6793
6794
6795
6796
6797
6798
6799
6800
6801
6802
6803
6804
6805
6806
6807
6808
6809
6810
6811
6812
6813
6814
6815
6816




6817
6818
6819
6820
6821
6822
6823
6824
6825
6826
6827
6828
6829
6830
6831
6832
6833
6834
6835
6836
6837
6838
6839
6840
6841
6842

6843
6844
6845
6846
6847
6848
6849
6850





6851
6852









6853
6854
6855
6856
6857
6858
6859
6860
6861
6862
6863
6864
6865
6866
6867
6868
6869
6870



6871
6872
6873
6874
6875
6876
6877
6878
6879
6880
6881
6882


6883
6884
6885
6886
6887
6888
6889
6890
....
6893
6894
6895
6896
6897
6898
6899
6900
6901
6902
6903
6904
6905
6906
6907
6908
6909
6910
6911
6912
6913
6914
6915
6916
6917
6918
6919
6920





6921
6922
6923
6924
6925
6926
6927
6928
6929
....
6935
6936
6937
6938
6939
6940
6941
6942
6943
6944
6945
6946
6947
6948
6949
6950
6951
6952
6953
6954
6955
6956
6957
6958
6959
6960
6961
6962
6963
6964
6965
6966
6967
6968
6969
6970
6971
....
6985
6986
6987
6988
6989
6990
6991
6992
6993
6994
6995
6996
6997
6998
6999
7000
7001
7002
7003
7004
....
7017
7018
7019
7020
7021
7022
7023
7024

#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)

#if defined(SQLITE_ENABLE_FTS3) && !defined(SQLITE_CORE)
# define SQLITE_CORE 1
#endif



#include <assert.h>
#include <stdlib.h>

#include <stdio.h>
#include <string.h>
#include <ctype.h>

#include "fts3.h"
#include "fts3_expr.h"
#include "fts3_hash.h"
#include "fts3_tokenizer.h"
#ifndef SQLITE_CORE 
# include "sqlite3ext.h"
  SQLITE_EXTENSION_INIT1
#endif



/* TODO(shess) MAN, this thing needs some refactoring.  At minimum, it
** would be nice to order the file better, perhaps something along the
** lines of:
**
**  - utility functions
................................................................................

#if 0
# define FTSTRACE(A)  printf A; fflush(stdout)
#else
# define FTSTRACE(A)
#endif

/* It is not safe to call isspace(), tolower(), or isalnum() on
** hi-bit-set characters.  This is the same solution used in the
** tokenizer.
*/
/* TODO(shess) The snippet-generation code should be using the
** tokenizer-generated tokens rather than doing its own local
** tokenization.
*/
/* TODO(shess) Is __isascii() a portable version of (c&0x80)==0? */
static int safe_isspace(char c){
  return (c&0x80)==0 ? isspace(c) : 0;
}
static int safe_tolower(char c){
  return (c&0x80)==0 ? tolower(c) : c;
}
static int safe_isalnum(char c){
  return (c&0x80)==0 ? isalnum(c) : 0;
}

typedef enum DocListType {
  DL_DOCIDS,              /* docids only */
  DL_POSITIONS,           /* docids + positions */
  DL_POSITIONS_OFFSETS    /* docids + positions + offsets */
} DocListType;

/*
................................................................................

enum {
  POS_END = 0,        /* end of this position list */
  POS_COLUMN,         /* followed by new column number */
  POS_BASE
};

/* MERGE_COUNT controls how often we merge segments (see comment at
** top of file).
*/
#define MERGE_COUNT 16

/* utility functions */

/* CLEAR() and SCRAMBLE() abstract memset() on a pointer to a single
** record to prevent errors of the form:
**
** my_function(SomeType *b){
**   memset(b, '\0', sizeof(b));  // sizeof(b)!=sizeof(*b)
................................................................................

#ifndef NDEBUG
#  define SCRAMBLE(b) memset(b, 0x55, sizeof(*(b)))
#else
#  define SCRAMBLE(b)
#endif

/* We may need up to VARINT_MAX bytes to store an encoded 64-bit integer. */
#define VARINT_MAX 10

/* Write a 64-bit variable-length integer to memory starting at p[0].
 * The length of data written will be between 1 and VARINT_MAX bytes.
 * The number of bytes written is returned. */

static int fts3PutVarint(char *p, sqlite_int64 v){
  unsigned char *q = (unsigned char *) p;
  sqlite_uint64 vu = v;
  do{
    *q++ = (unsigned char) ((vu & 0x7f) | 0x80);
    vu >>= 7;
  }while( vu!=0 );
  q[-1] &= 0x7f;  /* turn off high bit in final byte */
  assert( q - (unsigned char *)p <= VARINT_MAX );
  return (int) (q - (unsigned char *)p);
}


/* Read a 64-bit variable-length integer from memory starting at p[0].
 * Return the number of bytes read, or 0 on error.
 * The value is stored in *v. */

static int fts3GetVarint(const char *p, sqlite_int64 *v){
  const unsigned char *q = (const unsigned char *) p;
  sqlite_uint64 x = 0, y = 1;
  while( (*q & 0x80) == 0x80 ){
    x += y * (*q++ & 0x7f);
    y <<= 7;
    if( q - (unsigned char *)p >= VARINT_MAX ){  /* bad data */
      assert( 0 );
      return 0;
    }
  }
  x += y * (*q++);
  *v = (sqlite_int64) x;
  return (int) (q - (unsigned char *)p);
}





static int fts3GetVarint32(const char *p, int *pi){
 sqlite_int64 i;
 int ret = fts3GetVarint(p, &i);
 *pi = (int) i;
 assert( *pi==i );
 return ret;
}

/*******************************************************************/
/* DataBuffer is used to collect data into a buffer in piecemeal
** fashion.  It implements the usual distinction between amount of
** data currently stored (nData) and buffer capacity (nCapacity).
**
** dataBufferInit - create a buffer with given initial capacity.
** dataBufferReset - forget buffer's data, retaining capacity.
** dataBufferDestroy - free buffer's data.
** dataBufferSwap - swap contents of two buffers.
** dataBufferExpand - expand capacity without adding data.
** dataBufferAppend - append data.
** dataBufferAppend2 - append two pieces of data at once.
** dataBufferReplace - replace buffer's data.
*/
typedef struct DataBuffer {
  char *pData;          /* Pointer to malloc'ed buffer. */
  int nCapacity;        /* Size of pData buffer. */
  int nData;            /* End of data loaded into pData. */
} DataBuffer;

static void dataBufferInit(DataBuffer *pBuffer, int nCapacity){
  assert( nCapacity>=0 );
  pBuffer->nData = 0;
  pBuffer->nCapacity = nCapacity;
  pBuffer->pData = nCapacity==0 ? NULL : sqlite3_malloc(nCapacity);
}
static void dataBufferReset(DataBuffer *pBuffer){
  pBuffer->nData = 0;
}
static void dataBufferDestroy(DataBuffer *pBuffer){
  if( pBuffer->pData!=NULL ) sqlite3_free(pBuffer->pData);
  SCRAMBLE(pBuffer);
}
static void dataBufferSwap(DataBuffer *pBuffer1, DataBuffer *pBuffer2){
  DataBuffer tmp = *pBuffer1;
  *pBuffer1 = *pBuffer2;
  *pBuffer2 = tmp;
}
static void dataBufferExpand(DataBuffer *pBuffer, int nAddCapacity){
  assert( nAddCapacity>0 );
  /* TODO(shess) Consider expanding more aggressively.  Note that the
  ** underlying malloc implementation may take care of such things for
  ** us already.
  */
  if( pBuffer->nData+nAddCapacity>pBuffer->nCapacity ){
    pBuffer->nCapacity = pBuffer->nData+nAddCapacity;
    pBuffer->pData = sqlite3_realloc(pBuffer->pData, pBuffer->nCapacity);
  }
}
static void dataBufferAppend(DataBuffer *pBuffer,
                             const char *pSource, int nSource){
  assert( nSource>0 && pSource!=NULL );
  dataBufferExpand(pBuffer, nSource);
  memcpy(pBuffer->pData+pBuffer->nData, pSource, nSource);
  pBuffer->nData += nSource;
}
static void dataBufferAppend2(DataBuffer *pBuffer,
                              const char *pSource1, int nSource1,
                              const char *pSource2, int nSource2){
  assert( nSource1>0 && pSource1!=NULL );
  assert( nSource2>0 && pSource2!=NULL );
  dataBufferExpand(pBuffer, nSource1+nSource2);
  memcpy(pBuffer->pData+pBuffer->nData, pSource1, nSource1);
  memcpy(pBuffer->pData+pBuffer->nData+nSource1, pSource2, nSource2);
  pBuffer->nData += nSource1+nSource2;
}
static void dataBufferReplace(DataBuffer *pBuffer,
                              const char *pSource, int nSource){
  dataBufferReset(pBuffer);
  dataBufferAppend(pBuffer, pSource, nSource);
}

/* StringBuffer is a null-terminated version of DataBuffer. */
typedef struct StringBuffer {
  DataBuffer b;            /* Includes null terminator. */
} StringBuffer;

static void initStringBuffer(StringBuffer *sb){
  dataBufferInit(&sb->b, 100);
  dataBufferReplace(&sb->b, "", 1);
}
static int stringBufferLength(StringBuffer *sb){
  return sb->b.nData-1;
}
static char *stringBufferData(StringBuffer *sb){
  return sb->b.pData;
}
static void stringBufferDestroy(StringBuffer *sb){
  dataBufferDestroy(&sb->b);
}

static void nappend(StringBuffer *sb, const char *zFrom, int nFrom){
  assert( sb->b.nData>0 );
  if( nFrom>0 ){
    sb->b.nData--;
    dataBufferAppend2(&sb->b, zFrom, nFrom, "", 1);
  }
}
static void append(StringBuffer *sb, const char *zFrom){
  nappend(sb, zFrom, strlen(zFrom));
}

/* Append a list of strings separated by commas. */
static void appendList(StringBuffer *sb, int nString, char **azString){
  int i;
  for(i=0; i<nString; ++i){
    if( i>0 ) append(sb, ", ");
    append(sb, azString[i]);
  }
}

static int endsInWhiteSpace(StringBuffer *p){
  return stringBufferLength(p)>0 &&
    safe_isspace(stringBufferData(p)[stringBufferLength(p)-1]);
}

/* If the StringBuffer ends in something other than white space, add a
** single space character to the end.
*/
static void appendWhiteSpace(StringBuffer *p){
  if( stringBufferLength(p)==0 ) return;
  if( !endsInWhiteSpace(p) ) append(p, " ");
}

/* Remove white space from the end of the StringBuffer */
static void trimWhiteSpace(StringBuffer *p){
  while( endsInWhiteSpace(p) ){
    p->b.pData[--p->b.nData-1] = '\0';
  }
}

/*******************************************************************/
/* DLReader is used to read document elements from a doclist.  The
** current docid is cached, so dlrDocid() is fast.  DLReader does not
** own the doclist buffer.
**
** dlrAtEnd - true if there's no more data to read.
** dlrDocid - docid of current document.
** dlrDocData - doclist data for current document (including docid).
** dlrDocDataBytes - length of same.
** dlrAllDataBytes - length of all remaining data.
** dlrPosData - position data for current document.
** dlrPosDataLen - length of pos data for current document (incl POS_END).
** dlrStep - step to current document.
** dlrInit - initial for doclist of given type against given data.
** dlrDestroy - clean up.
**
** Expected usage is something like:
**
**   DLReader reader;
**   dlrInit(&reader, pData, nData);
**   while( !dlrAtEnd(&reader) ){
**     // calls to dlrDocid() and kin.
**     dlrStep(&reader);
**   }
**   dlrDestroy(&reader);
*/
typedef struct DLReader {
  DocListType iType;
  const char *pData;
  int nData;

  sqlite_int64 iDocid;
  int nElement;
} DLReader;

static int dlrAtEnd(DLReader *pReader){
  assert( pReader->nData>=0 );
  return pReader->nData==0;
}
static sqlite_int64 dlrDocid(DLReader *pReader){
  assert( !dlrAtEnd(pReader) );
  return pReader->iDocid;
}
static const char *dlrDocData(DLReader *pReader){
  assert( !dlrAtEnd(pReader) );
  return pReader->pData;
}
static int dlrDocDataBytes(DLReader *pReader){
  assert( !dlrAtEnd(pReader) );
  return pReader->nElement;
}
static int dlrAllDataBytes(DLReader *pReader){
  assert( !dlrAtEnd(pReader) );
  return pReader->nData;
}
/* TODO(shess) Consider adding a field to track iDocid varint length
** to make these two functions faster.  This might matter (a tiny bit)
** for queries.
*/
static const char *dlrPosData(DLReader *pReader){
  sqlite_int64 iDummy;
  int n = fts3GetVarint(pReader->pData, &iDummy);
  assert( !dlrAtEnd(pReader) );
  return pReader->pData+n;
}
static int dlrPosDataLen(DLReader *pReader){
  sqlite_int64 iDummy;
  int n = fts3GetVarint(pReader->pData, &iDummy);
  assert( !dlrAtEnd(pReader) );
  return pReader->nElement-n;
}
static void dlrStep(DLReader *pReader){
  assert( !dlrAtEnd(pReader) );

  /* Skip past current doclist element. */
  assert( pReader->nElement<=pReader->nData );
  pReader->pData += pReader->nElement;
  pReader->nData -= pReader->nElement;

  /* If there is more data, read the next doclist element. */
  if( pReader->nData!=0 ){
    sqlite_int64 iDocidDelta;
    int iDummy, n = fts3GetVarint(pReader->pData, &iDocidDelta);
    pReader->iDocid += iDocidDelta;
    if( pReader->iType>=DL_POSITIONS ){
      assert( n<pReader->nData );
      while( 1 ){
        n += fts3GetVarint32(pReader->pData+n, &iDummy);
        assert( n<=pReader->nData );
        if( iDummy==POS_END ) break;
        if( iDummy==POS_COLUMN ){
          n += fts3GetVarint32(pReader->pData+n, &iDummy);
          assert( n<pReader->nData );
        }else if( pReader->iType==DL_POSITIONS_OFFSETS ){
          n += fts3GetVarint32(pReader->pData+n, &iDummy);
          n += fts3GetVarint32(pReader->pData+n, &iDummy);
          assert( n<pReader->nData );
        }
      }
    }
    pReader->nElement = n;
    assert( pReader->nElement<=pReader->nData );
  }
}
static void dlrInit(DLReader *pReader, DocListType iType,
                    const char *pData, int nData){
  assert( pData!=NULL && nData!=0 );
  pReader->iType = iType;
  pReader->pData = pData;
  pReader->nData = nData;
  pReader->nElement = 0;
  pReader->iDocid = 0;

  /* Load the first element's data.  There must be a first element. */
  dlrStep(pReader);
}
static void dlrDestroy(DLReader *pReader){
  SCRAMBLE(pReader);
}

#ifndef NDEBUG
/* Verify that the doclist can be validly decoded.  Also returns the
** last docid found because it is convenient in other assertions for
** DLWriter.
*/
static void docListValidate(DocListType iType, const char *pData, int nData,
                            sqlite_int64 *pLastDocid){
  sqlite_int64 iPrevDocid = 0;
  assert( nData>0 );
  assert( pData!=0 );
  assert( pData+nData>pData );
  while( nData!=0 ){
    sqlite_int64 iDocidDelta;
    int n = fts3GetVarint(pData, &iDocidDelta);
    iPrevDocid += iDocidDelta;
    if( iType>DL_DOCIDS ){
      int iDummy;
      while( 1 ){
        n += fts3GetVarint32(pData+n, &iDummy);
        if( iDummy==POS_END ) break;
        if( iDummy==POS_COLUMN ){
          n += fts3GetVarint32(pData+n, &iDummy);
        }else if( iType>DL_POSITIONS ){
          n += fts3GetVarint32(pData+n, &iDummy);
          n += fts3GetVarint32(pData+n, &iDummy);
        }
        assert( n<=nData );
      }
    }
    assert( n<=nData );
    pData += n;
    nData -= n;
  }
  if( pLastDocid ) *pLastDocid = iPrevDocid;
}
#define ASSERT_VALID_DOCLIST(i, p, n, o) docListValidate(i, p, n, o)
#else
#define ASSERT_VALID_DOCLIST(i, p, n, o) assert( 1 )
#endif

/*******************************************************************/
/* DLWriter is used to write doclist data to a DataBuffer.  DLWriter
** always appends to the buffer and does not own it.
**
** dlwInit - initialize to write a given type doclistto a buffer.
** dlwDestroy - clear the writer's memory.  Does not free buffer.
** dlwAppend - append raw doclist data to buffer.
** dlwCopy - copy next doclist from reader to writer.
** dlwAdd - construct doclist element and append to buffer.
**    Only apply dlwAdd() to DL_DOCIDS doclists (else use PLWriter).
*/
typedef struct DLWriter {
  DocListType iType;
  DataBuffer *b;
  sqlite_int64 iPrevDocid;
#ifndef NDEBUG
  int has_iPrevDocid;
#endif
} DLWriter;

static void dlwInit(DLWriter *pWriter, DocListType iType, DataBuffer *b){
  pWriter->b = b;
  pWriter->iType = iType;
  pWriter->iPrevDocid = 0;
#ifndef NDEBUG
  pWriter->has_iPrevDocid = 0;
#endif
}
static void dlwDestroy(DLWriter *pWriter){
  SCRAMBLE(pWriter);
}
/* iFirstDocid is the first docid in the doclist in pData.  It is
** needed because pData may point within a larger doclist, in which
** case the first item would be delta-encoded.
**
** iLastDocid is the final docid in the doclist in pData.  It is
** needed to create the new iPrevDocid for future delta-encoding.  The
** code could decode the passed doclist to recreate iLastDocid, but
** the only current user (docListMerge) already has decoded this
** information.
*/
/* TODO(shess) This has become just a helper for docListMerge.
** Consider a refactor to make this cleaner.
*/
static void dlwAppend(DLWriter *pWriter,
                      const char *pData, int nData,
                      sqlite_int64 iFirstDocid, sqlite_int64 iLastDocid){
  sqlite_int64 iDocid = 0;
  char c[VARINT_MAX];
  int nFirstOld, nFirstNew;     /* Old and new varint len of first docid. */
#ifndef NDEBUG
  sqlite_int64 iLastDocidDelta;
#endif

  /* Recode the initial docid as delta from iPrevDocid. */
  nFirstOld = fts3GetVarint(pData, &iDocid);
  assert( nFirstOld<nData || (nFirstOld==nData && pWriter->iType==DL_DOCIDS) );
  nFirstNew = fts3PutVarint(c, iFirstDocid-pWriter->iPrevDocid);

  /* Verify that the incoming doclist is valid AND that it ends with
  ** the expected docid.  This is essential because we'll trust this
  ** docid in future delta-encoding.
  */
  ASSERT_VALID_DOCLIST(pWriter->iType, pData, nData, &iLastDocidDelta);
  assert( iLastDocid==iFirstDocid-iDocid+iLastDocidDelta );

  /* Append recoded initial docid and everything else.  Rest of docids
  ** should have been delta-encoded from previous initial docid.
  */
  if( nFirstOld<nData ){
    dataBufferAppend2(pWriter->b, c, nFirstNew,
                      pData+nFirstOld, nData-nFirstOld);
  }else{
    dataBufferAppend(pWriter->b, c, nFirstNew);
  }
  pWriter->iPrevDocid = iLastDocid;
}
static void dlwCopy(DLWriter *pWriter, DLReader *pReader){
  dlwAppend(pWriter, dlrDocData(pReader), dlrDocDataBytes(pReader),
            dlrDocid(pReader), dlrDocid(pReader));
}
static void dlwAdd(DLWriter *pWriter, sqlite_int64 iDocid){
  char c[VARINT_MAX];
  int n = fts3PutVarint(c, iDocid-pWriter->iPrevDocid);

  /* Docids must ascend. */
  assert( !pWriter->has_iPrevDocid || iDocid>pWriter->iPrevDocid );
  assert( pWriter->iType==DL_DOCIDS );

  dataBufferAppend(pWriter->b, c, n);
  pWriter->iPrevDocid = iDocid;
#ifndef NDEBUG
  pWriter->has_iPrevDocid = 1;
#endif
}

/*******************************************************************/
/* PLReader is used to read data from a document's position list.  As
** the caller steps through the list, data is cached so that varints
** only need to be decoded once.
**
** plrInit, plrDestroy - create/destroy a reader.
** plrColumn, plrPosition, plrStartOffset, plrEndOffset - accessors
** plrAtEnd - at end of stream, only call plrDestroy once true.
** plrStep - step to the next element.
*/
typedef struct PLReader {
  /* These refer to the next position's data.  nData will reach 0 when
  ** reading the last position, so plrStep() signals EOF by setting
  ** pData to NULL.
  */
  const char *pData;
  int nData;

  DocListType iType;
  int iColumn;         /* the last column read */
  int iPosition;       /* the last position read */
  int iStartOffset;    /* the last start offset read */
  int iEndOffset;      /* the last end offset read */
} PLReader;

static int plrAtEnd(PLReader *pReader){
  return pReader->pData==NULL;
}
static int plrColumn(PLReader *pReader){
  assert( !plrAtEnd(pReader) );
  return pReader->iColumn;
}
static int plrPosition(PLReader *pReader){
  assert( !plrAtEnd(pReader) );
  return pReader->iPosition;
}
static int plrStartOffset(PLReader *pReader){
  assert( !plrAtEnd(pReader) );
  return pReader->iStartOffset;
}
static int plrEndOffset(PLReader *pReader){
  assert( !plrAtEnd(pReader) );
  return pReader->iEndOffset;
}
static void plrStep(PLReader *pReader){
  int i, n;

  assert( !plrAtEnd(pReader) );

  if( pReader->nData==0 ){
    pReader->pData = NULL;
    return;
  }

  n = fts3GetVarint32(pReader->pData, &i);
  if( i==POS_COLUMN ){
    n += fts3GetVarint32(pReader->pData+n, &pReader->iColumn);
    pReader->iPosition = 0;
    pReader->iStartOffset = 0;
    n += fts3GetVarint32(pReader->pData+n, &i);
  }
  /* Should never see adjacent column changes. */
  assert( i!=POS_COLUMN );

  if( i==POS_END ){
    pReader->nData = 0;
    pReader->pData = NULL;
    return;
  }

  pReader->iPosition += i-POS_BASE;
  if( pReader->iType==DL_POSITIONS_OFFSETS ){
    n += fts3GetVarint32(pReader->pData+n, &i);
    pReader->iStartOffset += i;
    n += fts3GetVarint32(pReader->pData+n, &i);
    pReader->iEndOffset = pReader->iStartOffset+i;
  }
  assert( n<=pReader->nData );
  pReader->pData += n;
  pReader->nData -= n;
}

static void plrInit(PLReader *pReader, DLReader *pDLReader){
  pReader->pData = dlrPosData(pDLReader);
  pReader->nData = dlrPosDataLen(pDLReader);
  pReader->iType = pDLReader->iType;
  pReader->iColumn = 0;
  pReader->iPosition = 0;
  pReader->iStartOffset = 0;
  pReader->iEndOffset = 0;
  plrStep(pReader);
}
static void plrDestroy(PLReader *pReader){
  SCRAMBLE(pReader);
}

/*******************************************************************/
/* PLWriter is used in constructing a document's position list.  As a
** convenience, if iType is DL_DOCIDS, PLWriter becomes a no-op.
** PLWriter writes to the associated DLWriter's buffer.
**
** plwInit - init for writing a document's poslist.
** plwDestroy - clear a writer.
** plwAdd - append position and offset information.
** plwCopy - copy next position's data from reader to writer.
** plwTerminate - add any necessary doclist terminator.
**
** Calling plwAdd() after plwTerminate() may result in a corrupt
** doclist.
*/
/* TODO(shess) Until we've written the second item, we can cache the
** first item's information.  Then we'd have three states:
**
** - initialized with docid, no positions.
** - docid and one position.
** - docid and multiple positions.
**
** Only the last state needs to actually write to dlw->b, which would
** be an improvement in the DLCollector case.
*/
typedef struct PLWriter {
  DLWriter *dlw;

  int iColumn;    /* the last column written */
  int iPos;       /* the last position written */
  int iOffset;    /* the last start offset written */
} PLWriter;

/* TODO(shess) In the case where the parent is reading these values
** from a PLReader, we could optimize to a copy if that PLReader has
** the same type as pWriter.
*/
static void plwAdd(PLWriter *pWriter, int iColumn, int iPos,
                   int iStartOffset, int iEndOffset){
  /* Worst-case space for POS_COLUMN, iColumn, iPosDelta,
  ** iStartOffsetDelta, and iEndOffsetDelta.
  */
  char c[5*VARINT_MAX];
  int n = 0;

  /* Ban plwAdd() after plwTerminate(). */
  assert( pWriter->iPos!=-1 );

  if( pWriter->dlw->iType==DL_DOCIDS ) return;

  if( iColumn!=pWriter->iColumn ){
    n += fts3PutVarint(c+n, POS_COLUMN);
    n += fts3PutVarint(c+n, iColumn);
    pWriter->iColumn = iColumn;
    pWriter->iPos = 0;
    pWriter->iOffset = 0;
  }
  assert( iPos>=pWriter->iPos );
  n += fts3PutVarint(c+n, POS_BASE+(iPos-pWriter->iPos));
  pWriter->iPos = iPos;
  if( pWriter->dlw->iType==DL_POSITIONS_OFFSETS ){
    assert( iStartOffset>=pWriter->iOffset );
    n += fts3PutVarint(c+n, iStartOffset-pWriter->iOffset);
    pWriter->iOffset = iStartOffset;
    assert( iEndOffset>=iStartOffset );
    n += fts3PutVarint(c+n, iEndOffset-iStartOffset);
  }
  dataBufferAppend(pWriter->dlw->b, c, n);
}
static void plwCopy(PLWriter *pWriter, PLReader *pReader){
  plwAdd(pWriter, plrColumn(pReader), plrPosition(pReader),
         plrStartOffset(pReader), plrEndOffset(pReader));
}
static void plwInit(PLWriter *pWriter, DLWriter *dlw, sqlite_int64 iDocid){
  char c[VARINT_MAX];
  int n;

  pWriter->dlw = dlw;

  /* Docids must ascend. */
  assert( !pWriter->dlw->has_iPrevDocid || iDocid>pWriter->dlw->iPrevDocid );
  n = fts3PutVarint(c, iDocid-pWriter->dlw->iPrevDocid);
  dataBufferAppend(pWriter->dlw->b, c, n);
  pWriter->dlw->iPrevDocid = iDocid;
#ifndef NDEBUG
  pWriter->dlw->has_iPrevDocid = 1;
#endif

  pWriter->iColumn = 0;
  pWriter->iPos = 0;
  pWriter->iOffset = 0;
}
/* TODO(shess) Should plwDestroy() also terminate the doclist?  But
** then plwDestroy() would no longer be just a destructor, it would
** also be doing work, which isn't consistent with the overall idiom.
** Another option would be for plwAdd() to always append any necessary
** terminator, so that the output is always correct.  But that would
** add incremental work to the common case with the only benefit being
** API elegance.  Punt for now.
*/
static void plwTerminate(PLWriter *pWriter){
  if( pWriter->dlw->iType>DL_DOCIDS ){
    char c[VARINT_MAX];
    int n = fts3PutVarint(c, POS_END);
    dataBufferAppend(pWriter->dlw->b, c, n);
  }
#ifndef NDEBUG
  /* Mark as terminated for assert in plwAdd(). */
  pWriter->iPos = -1;
#endif
}
static void plwDestroy(PLWriter *pWriter){
  SCRAMBLE(pWriter);
}

/*******************************************************************/
/* DLCollector wraps PLWriter and DLWriter to provide a
** dynamically-allocated doclist area to use during tokenization.
**
** dlcNew - malloc up and initialize a collector.
** dlcDelete - destroy a collector and all contained items.
** dlcAddPos - append position and offset information.
** dlcAddDoclist - add the collected doclist to the given buffer.
** dlcNext - terminate the current document and open another.
*/
typedef struct DLCollector {
  DataBuffer b;
  DLWriter dlw;
  PLWriter plw;
} DLCollector;

/* TODO(shess) This could also be done by calling plwTerminate() and
** dataBufferAppend().  I tried that, expecting nominal performance
** differences, but it seemed to pretty reliably be worth 1% to code
** it this way.  I suspect it is the incremental malloc overhead (some
** percentage of the plwTerminate() calls will cause a realloc), so
** this might be worth revisiting if the DataBuffer implementation
** changes.
*/
static void dlcAddDoclist(DLCollector *pCollector, DataBuffer *b){
  if( pCollector->dlw.iType>DL_DOCIDS ){
    char c[VARINT_MAX];
    int n = fts3PutVarint(c, POS_END);
    dataBufferAppend2(b, pCollector->b.pData, pCollector->b.nData, c, n);
  }else{
    dataBufferAppend(b, pCollector->b.pData, pCollector->b.nData);
  }
}
static void dlcNext(DLCollector *pCollector, sqlite_int64 iDocid){
  plwTerminate(&pCollector->plw);
  plwDestroy(&pCollector->plw);
  plwInit(&pCollector->plw, &pCollector->dlw, iDocid);
}
static void dlcAddPos(DLCollector *pCollector, int iColumn, int iPos,
                      int iStartOffset, int iEndOffset){
  plwAdd(&pCollector->plw, iColumn, iPos, iStartOffset, iEndOffset);
}

static DLCollector *dlcNew(sqlite_int64 iDocid, DocListType iType){
  DLCollector *pCollector = sqlite3_malloc(sizeof(DLCollector));
  dataBufferInit(&pCollector->b, 0);
  dlwInit(&pCollector->dlw, iType, &pCollector->b);
  plwInit(&pCollector->plw, &pCollector->dlw, iDocid);
  return pCollector;
}
static void dlcDelete(DLCollector *pCollector){
  plwDestroy(&pCollector->plw);
  dlwDestroy(&pCollector->dlw);
  dataBufferDestroy(&pCollector->b);
  SCRAMBLE(pCollector);
  sqlite3_free(pCollector);
}


/* Copy the doclist data of iType in pData/nData into *out, trimming
** unnecessary data as we go.  Only columns matching iColumn are
** copied, all columns copied if iColumn is -1.  Elements with no
** matching columns are dropped.  The output is an iOutType doclist.
*/
/* NOTE(shess) This code is only valid after all doclists are merged.
** If this is run before merges, then doclist items which represent
** deletion will be trimmed, and will thus not effect a deletion
** during the merge.
*/
static void docListTrim(DocListType iType, const char *pData, int nData,
                        int iColumn, DocListType iOutType, DataBuffer *out){
  DLReader dlReader;
  DLWriter dlWriter;

  assert( iOutType<=iType );

  dlrInit(&dlReader, iType, pData, nData);
  dlwInit(&dlWriter, iOutType, out);

  while( !dlrAtEnd(&dlReader) ){
    PLReader plReader;
    PLWriter plWriter;
    int match = 0;

    plrInit(&plReader, &dlReader);

    while( !plrAtEnd(&plReader) ){
      if( iColumn==-1 || plrColumn(&plReader)==iColumn ){
        if( !match ){
          plwInit(&plWriter, &dlWriter, dlrDocid(&dlReader));
          match = 1;
        }
        plwAdd(&plWriter, plrColumn(&plReader), plrPosition(&plReader),
               plrStartOffset(&plReader), plrEndOffset(&plReader));
      }
      plrStep(&plReader);
    }
    if( match ){
      plwTerminate(&plWriter);
      plwDestroy(&plWriter);
    }

    plrDestroy(&plReader);
    dlrStep(&dlReader);
  }
  dlwDestroy(&dlWriter);
  dlrDestroy(&dlReader);
}

/* Used by docListMerge() to keep doclists in the ascending order by
** docid, then ascending order by age (so the newest comes first).
*/
typedef struct OrderedDLReader {
  DLReader *pReader;

  /* TODO(shess) If we assume that docListMerge pReaders is ordered by
  ** age (which we do), then we could use pReader comparisons to break
  ** ties.
  */
  int idx;
} OrderedDLReader;

/* Order eof to end, then by docid asc, idx desc. */
static int orderedDLReaderCmp(OrderedDLReader *r1, OrderedDLReader *r2){
  if( dlrAtEnd(r1->pReader) ){
    if( dlrAtEnd(r2->pReader) ) return 0;  /* Both atEnd(). */
    return 1;                              /* Only r1 atEnd(). */
  }
  if( dlrAtEnd(r2->pReader) ) return -1;   /* Only r2 atEnd(). */

  if( dlrDocid(r1->pReader)<dlrDocid(r2->pReader) ) return -1;
  if( dlrDocid(r1->pReader)>dlrDocid(r2->pReader) ) return 1;

  /* Descending on idx. */
  return r2->idx-r1->idx;
}

/* Bubble p[0] to appropriate place in p[1..n-1].  Assumes that
** p[1..n-1] is already sorted.
*/
/* TODO(shess) Is this frequent enough to warrant a binary search?
** Before implementing that, instrument the code to check.  In most
** current usage, I expect that p[0] will be less than p[1] a very
** high proportion of the time.
*/
static void orderedDLReaderReorder(OrderedDLReader *p, int n){
  while( n>1 && orderedDLReaderCmp(p, p+1)>0 ){
    OrderedDLReader tmp = p[0];
    p[0] = p[1];
    p[1] = tmp;
    n--;
    p++;
  }
}

/* Given an array of doclist readers, merge their doclist elements
** into out in sorted order (by docid), dropping elements from older
** readers when there is a duplicate docid.  pReaders is assumed to be
** ordered by age, oldest first.
*/
/* TODO(shess) nReaders must be <= MERGE_COUNT.  This should probably
** be fixed.
*/
static void docListMerge(DataBuffer *out,
                         DLReader *pReaders, int nReaders){
  OrderedDLReader readers[MERGE_COUNT];
  DLWriter writer;
  int i, n;
  const char *pStart = 0;
  int nStart = 0;
  sqlite_int64 iFirstDocid = 0, iLastDocid = 0;

  assert( nReaders>0 );
  if( nReaders==1 ){
    dataBufferAppend(out, dlrDocData(pReaders), dlrAllDataBytes(pReaders));
    return;
  }

  assert( nReaders<=MERGE_COUNT );
  n = 0;
  for(i=0; i<nReaders; i++){
    assert( pReaders[i].iType==pReaders[0].iType );
    readers[i].pReader = pReaders+i;
    readers[i].idx = i;
    n += dlrAllDataBytes(&pReaders[i]);
  }
  /* Conservatively size output to sum of inputs.  Output should end
  ** up strictly smaller than input.
  */
  dataBufferExpand(out, n);

  /* Get the readers into sorted order. */
  while( i-->0 ){
    orderedDLReaderReorder(readers+i, nReaders-i);
  }

  dlwInit(&writer, pReaders[0].iType, out);
  while( !dlrAtEnd(readers[0].pReader) ){
    sqlite_int64 iDocid = dlrDocid(readers[0].pReader);

    /* If this is a continuation of the current buffer to copy, extend
    ** that buffer.  memcpy() seems to be more efficient if it has a
    ** lots of data to copy.
    */
    if( dlrDocData(readers[0].pReader)==pStart+nStart ){
      nStart += dlrDocDataBytes(readers[0].pReader);
    }else{
      if( pStart!=0 ){
        dlwAppend(&writer, pStart, nStart, iFirstDocid, iLastDocid);
      }
      pStart = dlrDocData(readers[0].pReader);
      nStart = dlrDocDataBytes(readers[0].pReader);
      iFirstDocid = iDocid;
    }
    iLastDocid = iDocid;
    dlrStep(readers[0].pReader);

    /* Drop all of the older elements with the same docid. */
    for(i=1; i<nReaders &&
             !dlrAtEnd(readers[i].pReader) &&
             dlrDocid(readers[i].pReader)==iDocid; i++){
      dlrStep(readers[i].pReader);
    }

    /* Get the readers back into order. */
    while( i-->0 ){
      orderedDLReaderReorder(readers+i, nReaders-i);
    }
  }

  /* Copy over any remaining elements. */
  if( nStart>0 ) dlwAppend(&writer, pStart, nStart, iFirstDocid, iLastDocid);
  dlwDestroy(&writer);
}

/* Helper function for posListUnion().  Compares the current position
** between left and right, returning as standard C idiom of <0 if
** left<right, >0 if left>right, and 0 if left==right.  "End" always
** compares greater.
*/
static int posListCmp(PLReader *pLeft, PLReader *pRight){
  assert( pLeft->iType==pRight->iType );
  if( pLeft->iType==DL_DOCIDS ) return 0;

  if( plrAtEnd(pLeft) ) return plrAtEnd(pRight) ? 0 : 1;
  if( plrAtEnd(pRight) ) return -1;

  if( plrColumn(pLeft)<plrColumn(pRight) ) return -1;
  if( plrColumn(pLeft)>plrColumn(pRight) ) return 1;

  if( plrPosition(pLeft)<plrPosition(pRight) ) return -1;
  if( plrPosition(pLeft)>plrPosition(pRight) ) return 1;
  if( pLeft->iType==DL_POSITIONS ) return 0;

  if( plrStartOffset(pLeft)<plrStartOffset(pRight) ) return -1;
  if( plrStartOffset(pLeft)>plrStartOffset(pRight) ) return 1;

  if( plrEndOffset(pLeft)<plrEndOffset(pRight) ) return -1;
  if( plrEndOffset(pLeft)>plrEndOffset(pRight) ) return 1;

  return 0;
}

/* Write the union of position lists in pLeft and pRight to pOut.
** "Union" in this case meaning "All unique position tuples".  Should
** work with any doclist type, though both inputs and the output
** should be the same type.
*/
static void posListUnion(DLReader *pLeft, DLReader *pRight, DLWriter *pOut){
  PLReader left, right;
  PLWriter writer;

  assert( dlrDocid(pLeft)==dlrDocid(pRight) );
  assert( pLeft->iType==pRight->iType );
  assert( pLeft->iType==pOut->iType );

  plrInit(&left, pLeft);
  plrInit(&right, pRight);
  plwInit(&writer, pOut, dlrDocid(pLeft));

  while( !plrAtEnd(&left) || !plrAtEnd(&right) ){
    int c = posListCmp(&left, &right);
    if( c<0 ){
      plwCopy(&writer, &left);
      plrStep(&left);
    }else if( c>0 ){
      plwCopy(&writer, &right);
      plrStep(&right);
    }else{
      plwCopy(&writer, &left);
      plrStep(&left);
      plrStep(&right);
    }
  }

  plwTerminate(&writer);
  plwDestroy(&writer);
  plrDestroy(&left);
  plrDestroy(&right);
}

/* Write the union of doclists in pLeft and pRight to pOut.  For
** docids in common between the inputs, the union of the position
** lists is written.  Inputs and outputs are always type DL_DEFAULT.
*/
static void docListUnion(
  const char *pLeft, int nLeft,
  const char *pRight, int nRight,
  DataBuffer *pOut      /* Write the combined doclist here */
){
  DLReader left, right;
  DLWriter writer;

  if( nLeft==0 ){
    if( nRight!=0) dataBufferAppend(pOut, pRight, nRight);
    return;
  }
  if( nRight==0 ){
    dataBufferAppend(pOut, pLeft, nLeft);
    return;
  }

  dlrInit(&left, DL_DEFAULT, pLeft, nLeft);
  dlrInit(&right, DL_DEFAULT, pRight, nRight);
  dlwInit(&writer, DL_DEFAULT, pOut);

  while( !dlrAtEnd(&left) || !dlrAtEnd(&right) ){
    if( dlrAtEnd(&right) ){
      dlwCopy(&writer, &left);
      dlrStep(&left);
    }else if( dlrAtEnd(&left) ){
      dlwCopy(&writer, &right);
      dlrStep(&right);
    }else if( dlrDocid(&left)<dlrDocid(&right) ){
      dlwCopy(&writer, &left);
      dlrStep(&left);
    }else if( dlrDocid(&left)>dlrDocid(&right) ){
      dlwCopy(&writer, &right);
      dlrStep(&right);
    }else{
      posListUnion(&left, &right, &writer);
      dlrStep(&left);
      dlrStep(&right);
    }
  }

  dlrDestroy(&left);
  dlrDestroy(&right);
  dlwDestroy(&writer);
}

/* 
** This function is used as part of the implementation of phrase and
** NEAR matching.
**
** pLeft and pRight are DLReaders positioned to the same docid in
** lists of type DL_POSITION. This function writes an entry to the
** DLWriter pOut for each position in pRight that is less than
** (nNear+1) greater (but not equal to or smaller) than a position 
** in pLeft. For example, if nNear is 0, and the positions contained
** by pLeft and pRight are:
**
**    pLeft:  5 10 15 20
**    pRight: 6  9 17 21
**
** then the docid is added to pOut. If pOut is of type DL_POSITIONS,
** then a positionids "6" and "21" are also added to pOut.
**
** If boolean argument isSaveLeft is true, then positionids are copied
** from pLeft instead of pRight. In the example above, the positions "5"
** and "20" would be added instead of "6" and "21".


*/
static void posListPhraseMerge(
  DLReader *pLeft, 
  DLReader *pRight,
  int nNear,
  int isSaveLeft,
  DLWriter *pOut
){
  PLReader left, right;
  PLWriter writer;
  int match = 0;

  assert( dlrDocid(pLeft)==dlrDocid(pRight) );
  assert( pOut->iType!=DL_POSITIONS_OFFSETS );

  plrInit(&left, pLeft);
  plrInit(&right, pRight);

  while( !plrAtEnd(&left) && !plrAtEnd(&right) ){
    if( plrColumn(&left)<plrColumn(&right) ){
      plrStep(&left);
    }else if( plrColumn(&left)>plrColumn(&right) ){
      plrStep(&right);
    }else if( plrPosition(&left)>=plrPosition(&right) ){
      plrStep(&right);
    }else{
      if( (plrPosition(&right)-plrPosition(&left))<=(nNear+1) ){
        if( !match ){
          plwInit(&writer, pOut, dlrDocid(pLeft));
          match = 1;
        }
        if( !isSaveLeft ){
          plwAdd(&writer, plrColumn(&right), plrPosition(&right), 0, 0);
        }else{
          plwAdd(&writer, plrColumn(&left), plrPosition(&left), 0, 0);
        }
        plrStep(&right);
      }else{
        plrStep(&left);
      }
    }
  }

  if( match ){
    plwTerminate(&writer);
    plwDestroy(&writer);
  }

  plrDestroy(&left);
  plrDestroy(&right);
}

/*
** Compare the values pointed to by the PLReaders passed as arguments. 
** Return -1 if the value pointed to by pLeft is considered less than
** the value pointed to by pRight, +1 if it is considered greater
** than it, or 0 if it is equal. i.e.
**
**     (*pLeft - *pRight)
**
** A PLReader that is in the EOF condition is considered greater than
** any other. If neither argument is in EOF state, the return value of
** plrColumn() is used. If the plrColumn() values are equal, the
** comparison is on the basis of plrPosition().
*/
static int plrCompare(PLReader *pLeft, PLReader *pRight){
  assert(!plrAtEnd(pLeft) || !plrAtEnd(pRight));

  if( plrAtEnd(pRight) || plrAtEnd(pLeft) ){
    return (plrAtEnd(pRight) ? -1 : 1);
  }
  if( plrColumn(pLeft)!=plrColumn(pRight) ){
    return ((plrColumn(pLeft)<plrColumn(pRight)) ? -1 : 1);
  }
  if( plrPosition(pLeft)!=plrPosition(pRight) ){
    return ((plrPosition(pLeft)<plrPosition(pRight)) ? -1 : 1);
  }
  return 0;
}

/* We have two doclists with positions:  pLeft and pRight. Depending
** on the value of the nNear parameter, perform either a phrase
** intersection (if nNear==0) or a NEAR intersection (if nNear>0)
** and write the results into pOut.
**
** A phrase intersection means that two documents only match
** if pLeft.iPos+1==pRight.iPos.
**
** A NEAR intersection means that two documents only match if 
** (abs(pLeft.iPos-pRight.iPos)<nNear).
**
** If a NEAR intersection is requested, then the nPhrase argument should
** be passed the number of tokens in the two operands to the NEAR operator
** combined. For example:
**
**       Query syntax               nPhrase
**      ------------------------------------
**       "A B C" NEAR "D E"         5
**       A NEAR B                   2
**
** iType controls the type of data written to pOut.  If iType is
** DL_POSITIONS, the positions are those from pRight.
*/
static void docListPhraseMerge(
  const char *pLeft, int nLeft,
  const char *pRight, int nRight,
  int nNear,            /* 0 for a phrase merge, non-zero for a NEAR merge */
  int nPhrase,          /* Number of tokens in left+right operands to NEAR */
  DocListType iType,    /* Type of doclist to write to pOut */
  DataBuffer *pOut      /* Write the combined doclist here */
){
  DLReader left, right;
  DLWriter writer;

  if( nLeft==0 || nRight==0 ) return;

  assert( iType!=DL_POSITIONS_OFFSETS );

  dlrInit(&left, DL_POSITIONS, pLeft, nLeft);
  dlrInit(&right, DL_POSITIONS, pRight, nRight);
  dlwInit(&writer, iType, pOut);

  while( !dlrAtEnd(&left) && !dlrAtEnd(&right) ){
    if( dlrDocid(&left)<dlrDocid(&right) ){
      dlrStep(&left);
    }else if( dlrDocid(&right)<dlrDocid(&left) ){
      dlrStep(&right);
    }else{
      if( nNear==0 ){
        posListPhraseMerge(&left, &right, 0, 0, &writer);
      }else{
        /* This case occurs when two terms (simple terms or phrases) are
         * connected by a NEAR operator, span (nNear+1). i.e.
         *
         *     '"terrible company" NEAR widget'
         */
        DataBuffer one = {0, 0, 0};
        DataBuffer two = {0, 0, 0};

        DLWriter dlwriter2;
        DLReader dr1 = {0, 0, 0, 0, 0}; 
        DLReader dr2 = {0, 0, 0, 0, 0};

        dlwInit(&dlwriter2, iType, &one);
        posListPhraseMerge(&right, &left, nNear-3+nPhrase, 1, &dlwriter2);
        dlwInit(&dlwriter2, iType, &two);
        posListPhraseMerge(&left, &right, nNear-1, 0, &dlwriter2);

        if( one.nData) dlrInit(&dr1, iType, one.pData, one.nData);
        if( two.nData) dlrInit(&dr2, iType, two.pData, two.nData);

        if( !dlrAtEnd(&dr1) || !dlrAtEnd(&dr2) ){
          PLReader pr1 = {0};
          PLReader pr2 = {0};

          PLWriter plwriter;
          plwInit(&plwriter, &writer, dlrDocid(dlrAtEnd(&dr1)?&dr2:&dr1));

          if( one.nData ) plrInit(&pr1, &dr1);
          if( two.nData ) plrInit(&pr2, &dr2);
          while( !plrAtEnd(&pr1) || !plrAtEnd(&pr2) ){
            int iCompare = plrCompare(&pr1, &pr2);
            switch( iCompare ){
              case -1:
                plwCopy(&plwriter, &pr1);
                plrStep(&pr1);
                break;
              case 1:
                plwCopy(&plwriter, &pr2);
                plrStep(&pr2);
                break;
              case 0:
                plwCopy(&plwriter, &pr1);
                plrStep(&pr1);
                plrStep(&pr2);
                break;
            }
          }
          plwTerminate(&plwriter);
        }
        dataBufferDestroy(&one);
        dataBufferDestroy(&two);
      }
      dlrStep(&left);
      dlrStep(&right);
    }
  }

  dlrDestroy(&left);
  dlrDestroy(&right);
  dlwDestroy(&writer);
}

/* We have two DL_DOCIDS doclists:  pLeft and pRight.
** Write the intersection of these two doclists into pOut as a
** DL_DOCIDS doclist.
*/
static void docListAndMerge(
  const char *pLeft, int nLeft,
  const char *pRight, int nRight,
  DataBuffer *pOut      /* Write the combined doclist here */
){
  DLReader left, right;
  DLWriter writer;

  if( nLeft==0 || nRight==0 ) return;

  dlrInit(&left, DL_DOCIDS, pLeft, nLeft);
  dlrInit(&right, DL_DOCIDS, pRight, nRight);
  dlwInit(&writer, DL_DOCIDS, pOut);

  while( !dlrAtEnd(&left) && !dlrAtEnd(&right) ){
    if( dlrDocid(&left)<dlrDocid(&right) ){
      dlrStep(&left);
    }else if( dlrDocid(&right)<dlrDocid(&left) ){
      dlrStep(&right);
    }else{
      dlwAdd(&writer, dlrDocid(&left));
      dlrStep(&left);
      dlrStep(&right);
    }
  }

  dlrDestroy(&left);
  dlrDestroy(&right);
  dlwDestroy(&writer);
}

/* We have two DL_DOCIDS doclists:  pLeft and pRight.
** Write the union of these two doclists into pOut as a
** DL_DOCIDS doclist.
*/
static void docListOrMerge(
  const char *pLeft, int nLeft,
  const char *pRight, int nRight,
  DataBuffer *pOut      /* Write the combined doclist here */
){
  DLReader left, right;
  DLWriter writer;

  if( nLeft==0 ){
    if( nRight!=0 ) dataBufferAppend(pOut, pRight, nRight);
    return;
  }
  if( nRight==0 ){
    dataBufferAppend(pOut, pLeft, nLeft);
    return;
  }

  dlrInit(&left, DL_DOCIDS, pLeft, nLeft);
  dlrInit(&right, DL_DOCIDS, pRight, nRight);
  dlwInit(&writer, DL_DOCIDS, pOut);

  while( !dlrAtEnd(&left) || !dlrAtEnd(&right) ){
    if( dlrAtEnd(&right) ){
      dlwAdd(&writer, dlrDocid(&left));
      dlrStep(&left);
    }else if( dlrAtEnd(&left) ){
      dlwAdd(&writer, dlrDocid(&right));
      dlrStep(&right);
    }else if( dlrDocid(&left)<dlrDocid(&right) ){
      dlwAdd(&writer, dlrDocid(&left));
      dlrStep(&left);
    }else if( dlrDocid(&right)<dlrDocid(&left) ){
      dlwAdd(&writer, dlrDocid(&right));
      dlrStep(&right);
    }else{
      dlwAdd(&writer, dlrDocid(&left));
      dlrStep(&left);
      dlrStep(&right);
    }
  }

  dlrDestroy(&left);
  dlrDestroy(&right);
  dlwDestroy(&writer);
}

/* We have two DL_DOCIDS doclists:  pLeft and pRight.
** Write into pOut as DL_DOCIDS doclist containing all documents that
** occur in pLeft but not in pRight.
*/
static void docListExceptMerge(
  const char *pLeft, int nLeft,
  const char *pRight, int nRight,
  DataBuffer *pOut      /* Write the combined doclist here */
){
  DLReader left, right;
  DLWriter writer;

  if( nLeft==0 ) return;
  if( nRight==0 ){
    dataBufferAppend(pOut, pLeft, nLeft);
    return;
  }

  dlrInit(&left, DL_DOCIDS, pLeft, nLeft);
  dlrInit(&right, DL_DOCIDS, pRight, nRight);
  dlwInit(&writer, DL_DOCIDS, pOut);

  while( !dlrAtEnd(&left) ){
    while( !dlrAtEnd(&right) && dlrDocid(&right)<dlrDocid(&left) ){
      dlrStep(&right);
    }
    if( dlrAtEnd(&right) || dlrDocid(&left)<dlrDocid(&right) ){
      dlwAdd(&writer, dlrDocid(&left));
    }
    dlrStep(&left);
  }

  dlrDestroy(&left);
  dlrDestroy(&right);
  dlwDestroy(&writer);
}

static char *string_dup_n(const char *s, int n){
  char *str = sqlite3_malloc(n + 1);
  memcpy(str, s, n);
  str[n] = '\0';
  return str;
}

/* Duplicate a string; the caller must free() the returned string.
 * (We don't use strdup() since it is not part of the standard C library and
 * may not be available everywhere.) */
static char *string_dup(const char *s){
  return string_dup_n(s, strlen(s));
}

/* Format a string, replacing each occurrence of the % character with
 * zDb.zName.  This may be more convenient than sqlite_mprintf()
 * when one string is used repeatedly in a format string.
 * The caller must free() the returned string. */
static char *string_format(const char *zFormat,
                           const char *zDb, const char *zName){
  const char *p;
  size_t len = 0;
  size_t nDb = strlen(zDb);
  size_t nName = strlen(zName);
  size_t nFullTableName = nDb+1+nName;
  char *result;
  char *r;

  /* first compute length needed */
  for(p = zFormat ; *p ; ++p){
    len += (*p=='%' ? nFullTableName : 1);
  }
  len += 1;  /* for null terminator */

  r = result = sqlite3_malloc(len);
  for(p = zFormat; *p; ++p){
    if( *p=='%' ){
      memcpy(r, zDb, nDb);
      r += nDb;
      *r++ = '.';
      memcpy(r, zName, nName);
      r += nName;
    } else {
      *r++ = *p;
    }
  }
  *r++ = '\0';
  assert( r == result + len );
  return result;
}

static int sql_exec(sqlite3 *db, const char *zDb, const char *zName,
                    const char *zFormat){
  char *zCommand = string_format(zFormat, zDb, zName);
  int rc;
  FTSTRACE(("FTS3 sql: %s\n", zCommand));
  rc = sqlite3_exec(db, zCommand, NULL, 0, NULL);
  sqlite3_free(zCommand);
  return rc;
}

static int sql_prepare(sqlite3 *db, const char *zDb, const char *zName,
                       sqlite3_stmt **ppStmt, const char *zFormat){
  char *zCommand = string_format(zFormat, zDb, zName);
  int rc;
  FTSTRACE(("FTS3 prepare: %s\n", zCommand));
  rc = sqlite3_prepare_v2(db, zCommand, -1, ppStmt, NULL);
  sqlite3_free(zCommand);
  return rc;
}

/* end utility functions */

/* Forward reference */
typedef struct fulltext_vtab fulltext_vtab;

/*
** An instance of the following structure keeps track of generated
** matching-word offset information and snippets.
*/
typedef struct Snippet {
  int nMatch;     /* Total number of matches */
  int nAlloc;     /* Space allocated for aMatch[] */
  struct snippetMatch { /* One entry for each matching term */
    char snStatus;       /* Status flag for use while constructing snippets */
    short int iCol;      /* The column that contains the match */
    short int iTerm;     /* The index in Query.pTerms[] of the matching term */
    int iToken;          /* The index of the matching document token */
    short int nByte;     /* Number of bytes in the term */
    int iStart;          /* The offset to the first character of the term */
  } *aMatch;      /* Points to space obtained from malloc */
  char *zOffset;  /* Text rendering of aMatch[] */
  int nOffset;    /* strlen(zOffset) */
  char *zSnippet; /* Snippet text */
  int nSnippet;   /* strlen(zSnippet) */
} Snippet;


typedef enum QueryType {
  QUERY_GENERIC,   /* table scan */
  QUERY_DOCID,     /* lookup by docid */
  QUERY_FULLTEXT   /* QUERY_FULLTEXT + [i] is a full-text search for column i*/
} QueryType;

typedef enum fulltext_statement {
  CONTENT_INSERT_STMT,
  CONTENT_SELECT_STMT,
  CONTENT_UPDATE_STMT,
  CONTENT_DELETE_STMT,
  CONTENT_EXISTS_STMT,

  BLOCK_INSERT_STMT,
  BLOCK_SELECT_STMT,
  BLOCK_DELETE_STMT,
  BLOCK_DELETE_ALL_STMT,

  SEGDIR_MAX_INDEX_STMT,
  SEGDIR_SET_STMT,
  SEGDIR_SELECT_LEVEL_STMT,
  SEGDIR_SPAN_STMT,
  SEGDIR_DELETE_STMT,
  SEGDIR_SELECT_SEGMENT_STMT,
  SEGDIR_SELECT_ALL_STMT,
  SEGDIR_DELETE_ALL_STMT,
  SEGDIR_COUNT_STMT,

  MAX_STMT                     /* Always at end! */
} fulltext_statement;

/* These must exactly match the enum above. */
/* TODO(shess): Is there some risk that a statement will be used in two
** cursors at once, e.g.  if a query joins a virtual table to itself?
** If so perhaps we should move some of these to the cursor object.
*/
static const char *const fulltext_zStatement[MAX_STMT] = {
  /* CONTENT_INSERT */ NULL,  /* generated in contentInsertStatement() */
  /* CONTENT_SELECT */ NULL,  /* generated in contentSelectStatement() */
  /* CONTENT_UPDATE */ NULL,  /* generated in contentUpdateStatement() */
  /* CONTENT_DELETE */ "delete from %_content where docid = ?",
  /* CONTENT_EXISTS */ "select docid from %_content limit 1",

  /* BLOCK_INSERT */
  "insert into %_segments (blockid, block) values (null, ?)",
  /* BLOCK_SELECT */ "select block from %_segments where blockid = ?",
  /* BLOCK_DELETE */ "delete from %_segments where blockid between ? and ?",
  /* BLOCK_DELETE_ALL */ "delete from %_segments",

  /* SEGDIR_MAX_INDEX */ "select max(idx) from %_segdir where level = ?",
  /* SEGDIR_SET */ "insert into %_segdir values (?, ?, ?, ?, ?, ?)",
  /* SEGDIR_SELECT_LEVEL */
  "select start_block, leaves_end_block, root from %_segdir "
  " where level = ? order by idx",
  /* SEGDIR_SPAN */
  "select min(start_block), max(end_block) from %_segdir "
  " where level = ? and start_block <> 0",
  /* SEGDIR_DELETE */ "delete from %_segdir where level = ?",

  /* NOTE(shess): The first three results of the following two
  ** statements must match.
  */
  /* SEGDIR_SELECT_SEGMENT */
  "select start_block, leaves_end_block, root from %_segdir "
  " where level = ? and idx = ?",
  /* SEGDIR_SELECT_ALL */
  "select start_block, leaves_end_block, root from %_segdir "
  " order by level desc, idx asc",
  /* SEGDIR_DELETE_ALL */ "delete from %_segdir",
  /* SEGDIR_COUNT */ "select count(*), ifnull(max(level),0) from %_segdir",
};

/*
** A connection to a fulltext index is an instance of the following
** structure.  The xCreate and xConnect methods create an instance
** of this structure and xDestroy and xDisconnect free that instance.
** All other methods receive a pointer to the structure as one of their
** arguments.
*/
struct fulltext_vtab {
  sqlite3_vtab base;               /* Base class used by SQLite core */
  sqlite3 *db;                     /* The database connection */
  const char *zDb;                 /* logical database name */
  const char *zName;               /* virtual table name */
  int nColumn;                     /* number of columns in virtual table */
  char **azColumn;                 /* column names.  malloced */
  char **azContentColumn;          /* column names in content table; malloced */
  sqlite3_tokenizer *pTokenizer;   /* tokenizer for inserts and queries */

  /* Precompiled statements which we keep as long as the table is
  ** open.
  */
  sqlite3_stmt *pFulltextStatements[MAX_STMT];

  /* Precompiled statements used for segment merges.  We run a
  ** separate select across the leaf level of each tree being merged.
  */
  sqlite3_stmt *pLeafSelectStmts[MERGE_COUNT];
  /* The statement used to prepare pLeafSelectStmts. */
#define LEAF_SELECT \
  "select block from %_segments where blockid between ? and ? order by blockid"

  /* These buffer pending index updates during transactions.
  ** nPendingData estimates the memory size of the pending data.  It
  ** doesn't include the hash-bucket overhead, nor any malloc
  ** overhead.  When nPendingData exceeds kPendingThreshold, the
  ** buffer is flushed even before the transaction closes.
  ** pendingTerms stores the data, and is only valid when nPendingData
  ** is >=0 (nPendingData<0 means pendingTerms has not been
  ** initialized).  iPrevDocid is the last docid written, used to make
  ** certain we're inserting in sorted order.
  */
  int nPendingData;
#define kPendingThreshold (1*1024*1024)
  sqlite_int64 iPrevDocid;
  fts3Hash pendingTerms;
};

/*
** When the core wants to do a query, it create a cursor using a
** call to xOpen.  This structure is an instance of a cursor.  It
** is destroyed by xClose.
*/
typedef struct fulltext_cursor {
  sqlite3_vtab_cursor base;        /* Base class used by SQLite core */
  QueryType iCursorType;           /* Copy of sqlite3_index_info.idxNum */
  sqlite3_stmt *pStmt;             /* Prepared statement in use by the cursor */
  int eof;                         /* True if at End Of Results */
  Fts3Expr *pExpr;                 /* Parsed MATCH query string */
  Snippet snippet;                 /* Cached snippet for the current row */
  int iColumn;                     /* Column being searched */
  DataBuffer result;               /* Doclist results from fulltextQuery */
  DLReader reader;                 /* Result reader if result not empty */
} fulltext_cursor;

static fulltext_vtab *cursor_vtab(fulltext_cursor *c){
  return (fulltext_vtab *) c->base.pVtab;
}

static const sqlite3_module fts3Module;   /* forward declaration */

/* Return a dynamically generated statement of the form
 *   insert into %_content (docid, ...) values (?, ...)
 */
static const char *contentInsertStatement(fulltext_vtab *v){
  StringBuffer sb;

  int i;

  initStringBuffer(&sb);
  append(&sb, "insert into %_content (docid, ");
  appendList(&sb, v->nColumn, v->azContentColumn);
  append(&sb, ") values (?");
  for(i=0; i<v->nColumn; ++i)
    append(&sb, ", ?");
  append(&sb, ")");
  return stringBufferData(&sb);
}

/* Return a dynamically generated statement of the form
 *   select <content columns> from %_content where docid = ?
 */
static const char *contentSelectStatement(fulltext_vtab *v){
  StringBuffer sb;
  initStringBuffer(&sb);
  append(&sb, "SELECT ");
  appendList(&sb, v->nColumn, v->azContentColumn);
  append(&sb, " FROM %_content WHERE docid = ?");
  return stringBufferData(&sb);
}

/* Return a dynamically generated statement of the form
 *   update %_content set [col_0] = ?, [col_1] = ?, ...
 *                    where docid = ?
 */
static const char *contentUpdateStatement(fulltext_vtab *v){
  StringBuffer sb;
  int i;

  initStringBuffer(&sb);
  append(&sb, "update %_content set ");
  for(i=0; i<v->nColumn; ++i) {
    if( i>0 ){
      append(&sb, ", ");
    }
    append(&sb, v->azContentColumn[i]);
    append(&sb, " = ?");
  }
  append(&sb, " where docid = ?");
  return stringBufferData(&sb);
}

/* Puts a freshly-prepared statement determined by iStmt in *ppStmt.
** If the indicated statement has never been prepared, it is prepared
** and cached, otherwise the cached version is reset.
*/
static int sql_get_statement(fulltext_vtab *v, fulltext_statement iStmt,
                             sqlite3_stmt **ppStmt){
  assert( iStmt<MAX_STMT );
  if( v->pFulltextStatements[iStmt]==NULL ){
    const char *zStmt;
    int rc;
    switch( iStmt ){
      case CONTENT_INSERT_STMT:
        zStmt = contentInsertStatement(v); break;
      case CONTENT_SELECT_STMT:
        zStmt = contentSelectStatement(v); break;
      case CONTENT_UPDATE_STMT:
        zStmt = contentUpdateStatement(v); break;
      default:
        zStmt = fulltext_zStatement[iStmt];
    }
    rc = sql_prepare(v->db, v->zDb, v->zName, &v->pFulltextStatements[iStmt],
                         zStmt);
    if( zStmt != fulltext_zStatement[iStmt]) sqlite3_free((void *) zStmt);
    if( rc!=SQLITE_OK ) return rc;
  } else {
    int rc = sqlite3_reset(v->pFulltextStatements[iStmt]);
    if( rc!=SQLITE_OK ) return rc;
  }

  *ppStmt = v->pFulltextStatements[iStmt];
  return SQLITE_OK;
}

/* Like sqlite3_step(), but convert SQLITE_DONE to SQLITE_OK and
** SQLITE_ROW to SQLITE_ERROR.  Useful for statements like UPDATE,
** where we expect no results.
*/
static int sql_single_step(sqlite3_stmt *s){
  int rc = sqlite3_step(s);
  return (rc==SQLITE_DONE) ? SQLITE_OK : rc;
}

/* Like sql_get_statement(), but for special replicated LEAF_SELECT
** statements.  idx -1 is a special case for an uncached version of
** the statement (used in the optimize implementation).
*/
/* TODO(shess) Write version for generic statements and then share
** that between the cached-statement functions.
*/
static int sql_get_leaf_statement(fulltext_vtab *v, int idx,
                                  sqlite3_stmt **ppStmt){
  assert( idx>=-1 && idx<MERGE_COUNT );
  if( idx==-1 ){
    return sql_prepare(v->db, v->zDb, v->zName, ppStmt, LEAF_SELECT);
  }else if( v->pLeafSelectStmts[idx]==NULL ){
    int rc = sql_prepare(v->db, v->zDb, v->zName, &v->pLeafSelectStmts[idx],
                         LEAF_SELECT);
    if( rc!=SQLITE_OK ) return rc;
  }else{
    int rc = sqlite3_reset(v->pLeafSelectStmts[idx]);
    if( rc!=SQLITE_OK ) return rc;
  }

  *ppStmt = v->pLeafSelectStmts[idx];
  return SQLITE_OK;
}

/* insert into %_content (docid, ...) values ([docid], [pValues])
** If the docid contains SQL NULL, then a unique docid will be
** generated.
*/
static int content_insert(fulltext_vtab *v, sqlite3_value *docid,
                          sqlite3_value **pValues){
  sqlite3_stmt *s;
  int i;
  int rc = sql_get_statement(v, CONTENT_INSERT_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_value(s, 1, docid);
  if( rc!=SQLITE_OK ) return rc;

  for(i=0; i<v->nColumn; ++i){
    rc = sqlite3_bind_value(s, 2+i, pValues[i]);
    if( rc!=SQLITE_OK ) return rc;
  }

  return sql_single_step(s);
}

/* update %_content set col0 = pValues[0], col1 = pValues[1], ...
 *                  where docid = [iDocid] */
static int content_update(fulltext_vtab *v, sqlite3_value **pValues,
                          sqlite_int64 iDocid){
  sqlite3_stmt *s;
  int i;
  int rc = sql_get_statement(v, CONTENT_UPDATE_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  for(i=0; i<v->nColumn; ++i){
    rc = sqlite3_bind_value(s, 1+i, pValues[i]);
    if( rc!=SQLITE_OK ) return rc;
  }

  rc = sqlite3_bind_int64(s, 1+v->nColumn, iDocid);
  if( rc!=SQLITE_OK ) return rc;

  return sql_single_step(s);
}

static void freeStringArray(int nString, const char **pString){
  int i;

  for (i=0 ; i < nString ; ++i) {
    if( pString[i]!=NULL ) sqlite3_free((void *) pString[i]);
  }
  sqlite3_free((void *) pString);
}

/* select * from %_content where docid = [iDocid]
 * The caller must delete the returned array and all strings in it.
 * null fields will be NULL in the returned array.
 *
 * TODO: Perhaps we should return pointer/length strings here for consistency
 * with other code which uses pointer/length. */
static int content_select(fulltext_vtab *v, sqlite_int64 iDocid,
                          const char ***pValues){
  sqlite3_stmt *s;
  const char **values;
  int i;
  int rc;

  *pValues = NULL;

  rc = sql_get_statement(v, CONTENT_SELECT_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int64(s, 1, iDocid);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_step(s);
  if( rc!=SQLITE_ROW ) return rc;

  values = (const char **) sqlite3_malloc(v->nColumn * sizeof(const char *));
  for(i=0; i<v->nColumn; ++i){
    if( sqlite3_column_type(s, i)==SQLITE_NULL ){
      values[i] = NULL;
    }else{
      values[i] = string_dup((char*)sqlite3_column_text(s, i));
    }
  }

  /* We expect only one row.  We must execute another sqlite3_step()
   * to complete the iteration; otherwise the table will remain locked. */
  rc = sqlite3_step(s);
  if( rc==SQLITE_DONE ){
    *pValues = values;
    return SQLITE_OK;
  }

  freeStringArray(v->nColumn, values);
  return rc;
}

/* delete from %_content where docid = [iDocid ] */
static int content_delete(fulltext_vtab *v, sqlite_int64 iDocid){
  sqlite3_stmt *s;
  int rc = sql_get_statement(v, CONTENT_DELETE_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int64(s, 1, iDocid);
  if( rc!=SQLITE_OK ) return rc;

  return sql_single_step(s);
}

/* Returns SQLITE_ROW if any rows exist in %_content, SQLITE_DONE if
** no rows exist, and any error in case of failure.
*/
static int content_exists(fulltext_vtab *v){
  sqlite3_stmt *s;
  int rc = sql_get_statement(v, CONTENT_EXISTS_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_step(s);
  if( rc!=SQLITE_ROW ) return rc;

  /* We expect only one row.  We must execute another sqlite3_step()
   * to complete the iteration; otherwise the table will remain locked. */
  rc = sqlite3_step(s);
  if( rc==SQLITE_DONE ) return SQLITE_ROW;
  if( rc==SQLITE_ROW ) return SQLITE_ERROR;
  return rc;
}

/* insert into %_segments values ([pData])
**   returns assigned blockid in *piBlockid
*/
static int block_insert(fulltext_vtab *v, const char *pData, int nData,
                        sqlite_int64 *piBlockid){
  sqlite3_stmt *s;
  int rc = sql_get_statement(v, BLOCK_INSERT_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_blob(s, 1, pData, nData, SQLITE_STATIC);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_step(s);
  if( rc==SQLITE_ROW ) return SQLITE_ERROR;
  if( rc!=SQLITE_DONE ) return rc;

  /* blockid column is an alias for rowid. */
  *piBlockid = sqlite3_last_insert_rowid(v->db);
  return SQLITE_OK;
}

/* delete from %_segments
**   where blockid between [iStartBlockid] and [iEndBlockid]
**
** Deletes the range of blocks, inclusive, used to delete the blocks
** which form a segment.
*/
static int block_delete(fulltext_vtab *v,
                        sqlite_int64 iStartBlockid, sqlite_int64 iEndBlockid){
  sqlite3_stmt *s;
  int rc = sql_get_statement(v, BLOCK_DELETE_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int64(s, 1, iStartBlockid);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int64(s, 2, iEndBlockid);
  if( rc!=SQLITE_OK ) return rc;

  return sql_single_step(s);
}

/* Returns SQLITE_ROW with *pidx set to the maximum segment idx found
** at iLevel.  Returns SQLITE_DONE if there are no segments at
** iLevel.  Otherwise returns an error.
*/
static int segdir_max_index(fulltext_vtab *v, int iLevel, int *pidx){
  sqlite3_stmt *s;
  int rc = sql_get_statement(v, SEGDIR_MAX_INDEX_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int(s, 1, iLevel);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_step(s);
  /* Should always get at least one row due to how max() works. */
  if( rc==SQLITE_DONE ) return SQLITE_DONE;
  if( rc!=SQLITE_ROW ) return rc;

  /* NULL means that there were no inputs to max(). */
  if( SQLITE_NULL==sqlite3_column_type(s, 0) ){
    rc = sqlite3_step(s);
    if( rc==SQLITE_ROW ) return SQLITE_ERROR;
    return rc;
  }

  *pidx = sqlite3_column_int(s, 0);

  /* We expect only one row.  We must execute another sqlite3_step()
   * to complete the iteration; otherwise the table will remain locked. */
  rc = sqlite3_step(s);
  if( rc==SQLITE_ROW ) return SQLITE_ERROR;
  if( rc!=SQLITE_DONE ) return rc;
  return SQLITE_ROW;
}

/* insert into %_segdir values (
**   [iLevel], [idx],
**   [iStartBlockid], [iLeavesEndBlockid], [iEndBlockid],
**   [pRootData]
** )
*/
static int segdir_set(fulltext_vtab *v, int iLevel, int idx,
                      sqlite_int64 iStartBlockid,
                      sqlite_int64 iLeavesEndBlockid,
                      sqlite_int64 iEndBlockid,
                      const char *pRootData, int nRootData){
  sqlite3_stmt *s;
  int rc = sql_get_statement(v, SEGDIR_SET_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int(s, 1, iLevel);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int(s, 2, idx);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int64(s, 3, iStartBlockid);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int64(s, 4, iLeavesEndBlockid);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int64(s, 5, iEndBlockid);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_blob(s, 6, pRootData, nRootData, SQLITE_STATIC);
  if( rc!=SQLITE_OK ) return rc;

  return sql_single_step(s);
}

/* Queries %_segdir for the block span of the segments in level
** iLevel.  Returns SQLITE_DONE if there are no blocks for iLevel,
** SQLITE_ROW if there are blocks, else an error.
*/
static int segdir_span(fulltext_vtab *v, int iLevel,
                       sqlite_int64 *piStartBlockid,
                       sqlite_int64 *piEndBlockid){
  sqlite3_stmt *s;
  int rc = sql_get_statement(v, SEGDIR_SPAN_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int(s, 1, iLevel);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_step(s);
  if( rc==SQLITE_DONE ) return SQLITE_DONE;  /* Should never happen */
  if( rc!=SQLITE_ROW ) return rc;

  /* This happens if all segments at this level are entirely inline. */
  if( SQLITE_NULL==sqlite3_column_type(s, 0) ){
    /* We expect only one row.  We must execute another sqlite3_step()
     * to complete the iteration; otherwise the table will remain locked. */
    int rc2 = sqlite3_step(s);
    if( rc2==SQLITE_ROW ) return SQLITE_ERROR;
    return rc2;
  }

  *piStartBlockid = sqlite3_column_int64(s, 0);
  *piEndBlockid = sqlite3_column_int64(s, 1);

  /* We expect only one row.  We must execute another sqlite3_step()
   * to complete the iteration; otherwise the table will remain locked. */
  rc = sqlite3_step(s);
  if( rc==SQLITE_ROW ) return SQLITE_ERROR;
  if( rc!=SQLITE_DONE ) return rc;
  return SQLITE_ROW;
}

/* Delete the segment blocks and segment directory records for all
** segments at iLevel.
*/
static int segdir_delete(fulltext_vtab *v, int iLevel){
  sqlite3_stmt *s;
  sqlite_int64 iStartBlockid, iEndBlockid;
  int rc = segdir_span(v, iLevel, &iStartBlockid, &iEndBlockid);
  if( rc!=SQLITE_ROW && rc!=SQLITE_DONE ) return rc;

  if( rc==SQLITE_ROW ){
    rc = block_delete(v, iStartBlockid, iEndBlockid);
    if( rc!=SQLITE_OK ) return rc;
  }

  /* Delete the segment directory itself. */
  rc = sql_get_statement(v, SEGDIR_DELETE_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int64(s, 1, iLevel);
  if( rc!=SQLITE_OK ) return rc;

  return sql_single_step(s);
}

/* Delete entire fts index, SQLITE_OK on success, relevant error on
** failure.
*/
static int segdir_delete_all(fulltext_vtab *v){
  sqlite3_stmt *s;
  int rc = sql_get_statement(v, SEGDIR_DELETE_ALL_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sql_single_step(s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sql_get_statement(v, BLOCK_DELETE_ALL_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  return sql_single_step(s);
}

/* Returns SQLITE_OK with *pnSegments set to the number of entries in
** %_segdir and *piMaxLevel set to the highest level which has a
** segment.  Otherwise returns the SQLite error which caused failure.
*/
static int segdir_count(fulltext_vtab *v, int *pnSegments, int *piMaxLevel){
  sqlite3_stmt *s;
  int rc = sql_get_statement(v, SEGDIR_COUNT_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_step(s);
  /* TODO(shess): This case should not be possible?  Should stronger
  ** measures be taken if it happens?
  */
  if( rc==SQLITE_DONE ){
    *pnSegments = 0;
    *piMaxLevel = 0;
    return SQLITE_OK;
  }
  if( rc!=SQLITE_ROW ) return rc;

  *pnSegments = sqlite3_column_int(s, 0);
  *piMaxLevel = sqlite3_column_int(s, 1);

  /* We expect only one row.  We must execute another sqlite3_step()
   * to complete the iteration; otherwise the table will remain locked. */
  rc = sqlite3_step(s);
  if( rc==SQLITE_DONE ) return SQLITE_OK;
  if( rc==SQLITE_ROW ) return SQLITE_ERROR;
  return rc;
}

/* TODO(shess) clearPendingTerms() is far down the file because
** writeZeroSegment() is far down the file because LeafWriter is far
** down the file.  Consider refactoring the code to move the non-vtab
** code above the vtab code so that we don't need this forward
** reference.
*/
static int clearPendingTerms(fulltext_vtab *v);

/*
** Free the memory used to contain a fulltext_vtab structure.
*/
static void fulltext_vtab_destroy(fulltext_vtab *v){
  int iStmt, i;

  FTSTRACE(("FTS3 Destroy %p\n", v));
  for( iStmt=0; iStmt<MAX_STMT; iStmt++ ){
    if( v->pFulltextStatements[iStmt]!=NULL ){
      sqlite3_finalize(v->pFulltextStatements[iStmt]);
      v->pFulltextStatements[iStmt] = NULL;
    }
  }

  for( i=0; i<MERGE_COUNT; i++ ){
    if( v->pLeafSelectStmts[i]!=NULL ){
      sqlite3_finalize(v->pLeafSelectStmts[i]);
      v->pLeafSelectStmts[i] = NULL;
    }
  }

  if( v->pTokenizer!=NULL ){
    v->pTokenizer->pModule->xDestroy(v->pTokenizer);
    v->pTokenizer = NULL;
  }

  clearPendingTerms(v);

  sqlite3_free(v->azColumn);
  for(i = 0; i < v->nColumn; ++i) {
    sqlite3_free(v->azContentColumn[i]);
  }
  sqlite3_free(v->azContentColumn);
  sqlite3_free(v);
}

/*
** Token types for parsing the arguments to xConnect or xCreate.
*/
#define TOKEN_EOF         0    /* End of file */
#define TOKEN_SPACE       1    /* Any kind of whitespace */
#define TOKEN_ID          2    /* An identifier */
#define TOKEN_STRING      3    /* A string literal */
#define TOKEN_PUNCT       4    /* A single punctuation character */

/*
** If X is a character that can be used in an identifier then
** ftsIdChar(X) will be true.  Otherwise it is false.
**
** For ASCII, any character with the high-order bit set is
** allowed in an identifier.  For 7-bit characters, 
** isFtsIdChar[X] must be 1.
**
** Ticket #1066.  the SQL standard does not allow '$' in the
** middle of identfiers.  But many SQL implementations do. 
** SQLite will allow '$' in identifiers for compatibility.
** But the feature is undocumented.
*/
static const char isFtsIdChar[] = {
/* x0 x1 x2 x3 x4 x5 x6 x7 x8 x9 xA xB xC xD xE xF */
    0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  /* 2x */
    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,  /* 3x */
    0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,  /* 4x */
    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,  /* 5x */
    0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,  /* 6x */
    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0,  /* 7x */
};
#define ftsIdChar(C)  (((c=C)&0x80)!=0 || (c>0x1f && isFtsIdChar[c-0x20]))


/*
** Return the length of the token that begins at z[0]. 
** Store the token type in *tokenType before returning.
*/
static int ftsGetToken(const char *z, int *tokenType){
  int i, c;
  switch( *z ){
    case 0: {
      *tokenType = TOKEN_EOF;
      return 0;
    }
    case ' ': case '\t': case '\n': case '\f': case '\r': {
      for(i=1; safe_isspace(z[i]); i++){}
      *tokenType = TOKEN_SPACE;
      return i;
    }
    case '`':
    case '\'':
    case '"': {
      int delim = z[0];
      for(i=1; (c=z[i])!=0; i++){
        if( c==delim ){
          if( z[i+1]==delim ){

            i++;
          }else{
            break;
          }
        }
      }
      *tokenType = TOKEN_STRING;
      return i + (c!=0);
    }
    case '[': {
      for(i=1, c=z[0]; c!=']' && (c=z[i])!=0; i++){}
      *tokenType = TOKEN_ID;
      return i;
    }
    default: {
      if( !ftsIdChar(*z) ){
        break;
      }
      for(i=1; ftsIdChar(z[i]); i++){}
      *tokenType = TOKEN_ID;
      return i;
    }
  }
  *tokenType = TOKEN_PUNCT;
  return 1;
}

/*
** A token extracted from a string is an instance of the following
** structure.
*/
typedef struct FtsToken {
  const char *z;       /* Pointer to token text.  Not '\000' terminated */
  short int n;         /* Length of the token text in bytes. */
} FtsToken;

/*
** Given a input string (which is really one of the argv[] parameters
** passed into xConnect or xCreate) split the string up into tokens.
** Return an array of pointers to '\000' terminated strings, one string
** for each non-whitespace token.
**
** The returned array is terminated by a single NULL pointer.
**
** Space to hold the returned array is obtained from a single
** malloc and should be freed by passing the return value to free().
** The individual strings within the token list are all a part of
** the single memory allocation and will all be freed at once.
*/
static char **tokenizeString(const char *z, int *pnToken){
  int nToken = 0;
  FtsToken *aToken = sqlite3_malloc( strlen(z) * sizeof(aToken[0]) );
  int n = 1;
  int e, i;
  int totalSize = 0;
  char **azToken;
  char *zCopy;

  while( n>0 ){
    n = ftsGetToken(z, &e);
    if( e!=TOKEN_SPACE ){
      aToken[nToken].z = z;
      aToken[nToken].n = n;
      nToken++;
      totalSize += n+1;
    }
    z += n;
  }
  azToken = (char**)sqlite3_malloc( nToken*sizeof(char*) + totalSize );
  zCopy = (char*)&azToken[nToken];
  nToken--;
  for(i=0; i<nToken; i++){
    azToken[i] = zCopy;
    n = aToken[i].n;
    memcpy(zCopy, aToken[i].z, n);
    zCopy[n] = 0;
    zCopy += n+1;
  }
  azToken[nToken] = 0;
  sqlite3_free(aToken);
  *pnToken = nToken;
  return azToken;
}

/*
** Convert an SQL-style quoted string into a normal string by removing
** the quote characters.  The conversion is done in-place.  If the
** input does not begin with a quote character, then this routine
** is a no-op.
................................................................................
** Examples:
**
**     "abc"   becomes   abc
**     'xyz'   becomes   xyz
**     [pqr]   becomes   pqr
**     `mno`   becomes   mno
*/
static void dequoteString(char *z){
  int quote;
  int i, j;
  if( z==0 ) return;

  quote = z[0];
  switch( quote ){
    case '\'':  break;
    case '"':   break;
    case '`':   break;                /* For MySQL compatibility */
    case '[':   quote = ']';  break;  /* For MS SqlServer compatibility */
    default:    return;
  }
  for(i=1, j=0; z[i]; i++){
    if( z[i]==quote ){
      if( z[i+1]==quote ){
        z[j++] = quote;
        i++;
      }else{
        z[j++] = 0;
        break;
      }
    }else{
      z[j++] = z[i];
    }
  }
}

/*
** The input azIn is a NULL-terminated list of tokens.  Remove the first
** token and all punctuation tokens.  Remove the quotes from
** around string literal tokens.
**
** Example:
**
**     input:      tokenize chinese ( 'simplifed' , 'mixed' )
**     output:     chinese simplifed mixed
**
** Another example:
**
**     input:      delimiters ( '[' , ']' , '...' )
**     output:     [ ] ...
*/
static void tokenListToIdList(char **azIn){
  int i, j;
  if( azIn ){
    for(i=0, j=-1; azIn[i]; i++){
      if( safe_isalnum(azIn[i][0]) || azIn[i][1] ){
        dequoteString(azIn[i]);
        if( j>=0 ){
          azIn[j] = azIn[i];
        }
        j++;
      }
    }
    azIn[j] = 0;
  }
}


/*
** Find the first alphanumeric token in the string zIn.  Null-terminate
** this token.  Remove any quotation marks.  And return a pointer to
** the result.
*/
static char *firstToken(char *zIn, char **pzTail){
  int n, ttype;
  while(1){
    n = ftsGetToken(zIn, &ttype);
    if( ttype==TOKEN_SPACE ){
      zIn += n;
    }else if( ttype==TOKEN_EOF ){
      *pzTail = zIn;
      return 0;
    }else{
      zIn[n] = 0;
      *pzTail = &zIn[1];
      dequoteString(zIn);
      return zIn;
    }
  }
  /*NOTREACHED*/
}

/* Return true if...
**
**   *  s begins with the string t, ignoring case
**   *  s is longer than t
**   *  The first character of s beyond t is not a alphanumeric
** 
** Ignore leading space in *s.
**
** To put it another way, return true if the first token of
** s[] is t[].
*/
static int startsWith(const char *s, const char *t){
  while( safe_isspace(*s) ){ s++; }
  while( *t ){
    if( safe_tolower(*s++)!=safe_tolower(*t++) ) return 0;
  }
  return *s!='_' && !safe_isalnum(*s);
}

/*
** An instance of this structure defines the "spec" of a
** full text index.  This structure is populated by parseSpec
** and use by fulltextConnect and fulltextCreate.
*/
typedef struct TableSpec {
  const char *zDb;         /* Logical database name */
  const char *zName;       /* Name of the full-text index */
  int nColumn;             /* Number of columns to be indexed */
  char **azColumn;         /* Original names of columns to be indexed */
  char **azContentColumn;  /* Column names for %_content */
  char **azTokenizer;      /* Name of tokenizer and its arguments */
} TableSpec;

/*
** Reclaim all of the memory used by a TableSpec
*/
static void clearTableSpec(TableSpec *p) {
  sqlite3_free(p->azColumn);
  sqlite3_free(p->azContentColumn);
  sqlite3_free(p->azTokenizer);
}

/* Parse a CREATE VIRTUAL TABLE statement, which looks like this:
 *
 * CREATE VIRTUAL TABLE email
 *        USING fts3(subject, body, tokenize mytokenizer(myarg))
 *
 * We return parsed information in a TableSpec structure.
 * 
 */
static int parseSpec(TableSpec *pSpec, int argc, const char *const*argv,
                     char**pzErr){
  int i, n;
  char *z, *zDummy;
  char **azArg;
  const char *zTokenizer = 0;    /* argv[] entry describing the tokenizer */

  assert( argc>=3 );
  /* Current interface:
  ** argv[0] - module name
  ** argv[1] - database name
  ** argv[2] - table name
  ** argv[3..] - columns, optionally followed by tokenizer specification
  **             and snippet delimiters specification.
  */

  /* Make a copy of the complete argv[][] array in a single allocation.
  ** The argv[][] array is read-only and transient.  We can write to the
  ** copy in order to modify things and the copy is persistent.
  */
  CLEAR(pSpec);
  for(i=n=0; i<argc; i++){
    n += strlen(argv[i]) + 1;
  }
  azArg = sqlite3_malloc( sizeof(char*)*argc + n );
  if( azArg==0 ){
    return SQLITE_NOMEM;
  }
  z = (char*)&azArg[argc];
  for(i=0; i<argc; i++){
    azArg[i] = z;
    strcpy(z, argv[i]);
    z += strlen(z)+1;
  }

  /* Identify the column names and the tokenizer and delimiter arguments
  ** in the argv[][] array.
  */
  pSpec->zDb = azArg[1];
  pSpec->zName = azArg[2];
  pSpec->nColumn = 0;
  pSpec->azColumn = azArg;
  zTokenizer = "tokenize simple";
  for(i=3; i<argc; ++i){
    if( startsWith(azArg[i],"tokenize") ){
      zTokenizer = azArg[i];
    }else{
      z = azArg[pSpec->nColumn] = firstToken(azArg[i], &zDummy);
      pSpec->nColumn++;
    }
  }
  if( pSpec->nColumn==0 ){
    azArg[0] = "content";
    pSpec->nColumn = 1;
  }

  /*
  ** Construct the list of content column names.
  **
  ** Each content column name will be of the form cNNAAAA
  ** where NN is the column number and AAAA is the sanitized
  ** column name.  "sanitized" means that special characters are
  ** converted to "_".  The cNN prefix guarantees that all column
  ** names are unique.
  **
  ** The AAAA suffix is not strictly necessary.  It is included
  ** for the convenience of people who might examine the generated
  ** %_content table and wonder what the columns are used for.
  */
  pSpec->azContentColumn = sqlite3_malloc( pSpec->nColumn * sizeof(char *) );
  if( pSpec->azContentColumn==0 ){
    clearTableSpec(pSpec);
    return SQLITE_NOMEM;
  }
  for(i=0; i<pSpec->nColumn; i++){
    char *p;
    pSpec->azContentColumn[i] = sqlite3_mprintf("c%d%s", i, azArg[i]);
    for (p = pSpec->azContentColumn[i]; *p ; ++p) {
      if( !safe_isalnum(*p) ) *p = '_';
    }
  }

  /*
  ** Parse the tokenizer specification string.
  */
  pSpec->azTokenizer = tokenizeString(zTokenizer, &n);
  tokenListToIdList(pSpec->azTokenizer);

  return SQLITE_OK;
}

/*
** Generate a CREATE TABLE statement that describes the schema of
** the virtual table.  Return a pointer to this schema string.
**
** Space is obtained from sqlite3_mprintf() and should be freed
** using sqlite3_free().
*/
static char *fulltextSchema(
  int nColumn,                  /* Number of columns */
  const char *const* azColumn,  /* List of columns */
  const char *zTableName        /* Name of the table */
){
  int i;
  char *zSchema, *zNext;
  const char *zSep = "(";
  zSchema = sqlite3_mprintf("CREATE TABLE x");
  for(i=0; i<nColumn; i++){
    zNext = sqlite3_mprintf("%s%s%Q", zSchema, zSep, azColumn[i]);
    sqlite3_free(zSchema);
    zSchema = zNext;
    zSep = ",";
  }
  zNext = sqlite3_mprintf("%s,%Q HIDDEN", zSchema, zTableName);
  sqlite3_free(zSchema);
  zSchema = zNext;
  zNext = sqlite3_mprintf("%s,docid HIDDEN)", zSchema);
  sqlite3_free(zSchema);
  return zNext;
}

/*
** Build a new sqlite3_vtab structure that will describe the
** fulltext index defined by spec.
*/
static int constructVtab(
  sqlite3 *db,              /* The SQLite database connection */
  fts3Hash *pHash,          /* Hash table containing tokenizers */
  TableSpec *spec,          /* Parsed spec information from parseSpec() */
  sqlite3_vtab **ppVTab,    /* Write the resulting vtab structure here */
  char **pzErr              /* Write any error message here */
){
  int rc;
  int n;
  fulltext_vtab *v = 0;
  const sqlite3_tokenizer_module *m = NULL;
  char *schema;

  char const *zTok;         /* Name of tokenizer to use for this fts table */
  int nTok;                 /* Length of zTok, including nul terminator */

  v = (fulltext_vtab *) sqlite3_malloc(sizeof(fulltext_vtab));
  if( v==0 ) return SQLITE_NOMEM;
  CLEAR(v);
  /* sqlite will initialize v->base */
  v->db = db;
  v->zDb = spec->zDb;       /* Freed when azColumn is freed */
  v->zName = spec->zName;   /* Freed when azColumn is freed */
  v->nColumn = spec->nColumn;
  v->azContentColumn = spec->azContentColumn;
  spec->azContentColumn = 0;
  v->azColumn = spec->azColumn;
  spec->azColumn = 0;

  if( spec->azTokenizer==0 ){
    return SQLITE_NOMEM;
  }

  zTok = spec->azTokenizer[0]; 
  if( !zTok ){
    zTok = "simple";
  }
  nTok = strlen(zTok)+1;

  m = (sqlite3_tokenizer_module *)sqlite3Fts3HashFind(pHash, zTok, nTok);
  if( !m ){
    *pzErr = sqlite3_mprintf("unknown tokenizer: %s", spec->azTokenizer[0]);
    rc = SQLITE_ERROR;
    goto err;
  }

  for(n=0; spec->azTokenizer[n]; n++){}
  if( n ){
    rc = m->xCreate(n-1, (const char*const*)&spec->azTokenizer[1],
                    &v->pTokenizer);
  }else{
    rc = m->xCreate(0, 0, &v->pTokenizer);
  }
  if( rc!=SQLITE_OK ) goto err;
  v->pTokenizer->pModule = m;

  /* TODO: verify the existence of backing tables foo_content, foo_term */

  schema = fulltextSchema(v->nColumn, (const char*const*)v->azColumn,
                          spec->zName);
  rc = sqlite3_declare_vtab(db, schema);
  sqlite3_free(schema);
  if( rc!=SQLITE_OK ) goto err;

  memset(v->pFulltextStatements, 0, sizeof(v->pFulltextStatements));

  /* Indicate that the buffer is not live. */
  v->nPendingData = -1;

  *ppVTab = &v->base;
  FTSTRACE(("FTS3 Connect %p\n", v));

  return rc;

err:
  fulltext_vtab_destroy(v);
  return rc;
}

static int fulltextConnect(
  sqlite3 *db,
  void *pAux,
  int argc, const char *const*argv,
  sqlite3_vtab **ppVTab,
  char **pzErr
){
  TableSpec spec;
  int rc = parseSpec(&spec, argc, argv, pzErr);
  if( rc!=SQLITE_OK ) return rc;

  rc = constructVtab(db, (fts3Hash *)pAux, &spec, ppVTab, pzErr);
  clearTableSpec(&spec);
  return rc;
}

/* The %_content table holds the text of each document, with
** the docid column exposed as the SQLite rowid for the table.
*/
/* TODO(shess) This comment needs elaboration to match the updated
** code.  Work it into the top-of-file comment at that time.
*/
static int fulltextCreate(sqlite3 *db, void *pAux,
                          int argc, const char * const *argv,
                          sqlite3_vtab **ppVTab, char **pzErr){
  int rc;
  TableSpec spec;
  StringBuffer schema;
  FTSTRACE(("FTS3 Create\n"));

  rc = parseSpec(&spec, argc, argv, pzErr);
  if( rc!=SQLITE_OK ) return rc;

  initStringBuffer(&schema);
  append(&schema, "CREATE TABLE %_content(");
  append(&schema, "  docid INTEGER PRIMARY KEY,");
  appendList(&schema, spec.nColumn, spec.azContentColumn);
  append(&schema, ")");
  rc = sql_exec(db, spec.zDb, spec.zName, stringBufferData(&schema));
  stringBufferDestroy(&schema);
  if( rc!=SQLITE_OK ) goto out;

  rc = sql_exec(db, spec.zDb, spec.zName,
                "create table %_segments("
                "  blockid INTEGER PRIMARY KEY,"
                "  block blob"
                ");"
                );
  if( rc!=SQLITE_OK ) goto out;

  rc = sql_exec(db, spec.zDb, spec.zName,
                "create table %_segdir("
                "  level integer,"
                "  idx integer,"
                "  start_block integer,"
                "  leaves_end_block integer,"
                "  end_block integer,"
                "  root blob,"
                "  primary key(level, idx)"
                ");");
  if( rc!=SQLITE_OK ) goto out;

  rc = constructVtab(db, (fts3Hash *)pAux, &spec, ppVTab, pzErr);

out:
  clearTableSpec(&spec);
  return rc;
}

/* Decide how to handle an SQL query. */
static int fulltextBestIndex(sqlite3_vtab *pVTab, sqlite3_index_info *pInfo){
  fulltext_vtab *v = (fulltext_vtab *)pVTab;
  int i;
  FTSTRACE(("FTS3 BestIndex\n"));

  for(i=0; i<pInfo->nConstraint; ++i){
    const struct sqlite3_index_constraint *pConstraint;
    pConstraint = &pInfo->aConstraint[i];
    if( pConstraint->usable ) {
      if( (pConstraint->iColumn==-1 || pConstraint->iColumn==v->nColumn+1) &&
          pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ ){
        pInfo->idxNum = QUERY_DOCID;      /* lookup by docid */
        FTSTRACE(("FTS3 QUERY_DOCID\n"));
      } else if( pConstraint->iColumn>=0 && pConstraint->iColumn<=v->nColumn &&
                 pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH ){
        /* full-text search */
        pInfo->idxNum = QUERY_FULLTEXT + pConstraint->iColumn;
        FTSTRACE(("FTS3 QUERY_FULLTEXT %d\n", pConstraint->iColumn));
      } else continue;

      pInfo->aConstraintUsage[i].argvIndex = 1;
      pInfo->aConstraintUsage[i].omit = 1;

      /* An arbitrary value for now.
       * TODO: Perhaps docid matches should be considered cheaper than
       * full-text searches. */
      pInfo->estimatedCost = 1.0;   

      return SQLITE_OK;
    }
  }
  pInfo->idxNum = QUERY_GENERIC;
  return SQLITE_OK;
}

static int fulltextDisconnect(sqlite3_vtab *pVTab){
  FTSTRACE(("FTS3 Disconnect %p\n", pVTab));
  fulltext_vtab_destroy((fulltext_vtab *)pVTab);
  return SQLITE_OK;
}

static int fulltextDestroy(sqlite3_vtab *pVTab){
  fulltext_vtab *v = (fulltext_vtab *)pVTab;
  int rc;

  FTSTRACE(("FTS3 Destroy %p\n", pVTab));
  rc = sql_exec(v->db, v->zDb, v->zName,
                "drop table if exists %_content;"
                "drop table if exists %_segments;"
                "drop table if exists %_segdir;"
                );
  if( rc!=SQLITE_OK ) return rc;

  fulltext_vtab_destroy((fulltext_vtab *)pVTab);
  return SQLITE_OK;
}

static int fulltextOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){
  fulltext_cursor *c;

  c = (fulltext_cursor *) sqlite3_malloc(sizeof(fulltext_cursor));
  if( c ){
    memset(c, 0, sizeof(fulltext_cursor));
    /* sqlite will initialize c->base */
    *ppCursor = &c->base;
    FTSTRACE(("FTS3 Open %p: %p\n", pVTab, c));
    return SQLITE_OK;
  }else{
    return SQLITE_NOMEM;
  }
}

/* Free all of the dynamically allocated memory held by the
** Snippet
*/
static void snippetClear(Snippet *p){
  sqlite3_free(p->aMatch);
  sqlite3_free(p->zOffset);
  sqlite3_free(p->zSnippet);
  CLEAR(p);
}

/*
** Append a single entry to the p->aMatch[] log.
*/
static void snippetAppendMatch(
  Snippet *p,               /* Append the entry to this snippet */
  int iCol, int iTerm,      /* The column and query term */
  int iToken,               /* Matching token in document */
  int iStart, int nByte     /* Offset and size of the match */
){
  int i;
  struct snippetMatch *pMatch;
  if( p->nMatch+1>=p->nAlloc ){
    p->nAlloc = p->nAlloc*2 + 10;
    p->aMatch = sqlite3_realloc(p->aMatch, p->nAlloc*sizeof(p->aMatch[0]) );
    if( p->aMatch==0 ){
      p->nMatch = 0;
      p->nAlloc = 0;
      return;
    }
  }
  i = p->nMatch++;
  pMatch = &p->aMatch[i];
  pMatch->iCol = iCol;
  pMatch->iTerm = iTerm;
  pMatch->iToken = iToken;
  pMatch->iStart = iStart;
  pMatch->nByte = nByte;
}

/*
** Sizing information for the circular buffer used in snippetOffsetsOfColumn()
*/
#define FTS3_ROTOR_SZ   (32)
#define FTS3_ROTOR_MASK (FTS3_ROTOR_SZ-1)

/*
** Function to iterate through the tokens of a compiled expression.
**
** Except, skip all tokens on the right-hand side of a NOT operator.
** This function is used to find tokens as part of snippet and offset
** generation and we do nt want snippets and offsets to report matches
** for tokens on the RHS of a NOT.
*/
static int fts3NextExprToken(Fts3Expr **ppExpr, int *piToken){
  Fts3Expr *p = *ppExpr;
  int iToken = *piToken;
  if( iToken<0 ){
    /* In this case the expression p is the root of an expression tree.
    ** Move to the first token in the expression tree.
    */
    while( p->pLeft ){
      p = p->pLeft;
    }
    iToken = 0;
  }else{
    assert(p && p->eType==FTSQUERY_PHRASE );
    if( iToken<(p->pPhrase->nToken-1) ){
      iToken++;
    }else{
      iToken = 0;
      while( p->pParent && p->pParent->pLeft!=p ){
        assert( p->pParent->pRight==p );
        p = p->pParent;
      }
      p = p->pParent;
      if( p ){
        assert( p->pRight!=0 );
        p = p->pRight;
        while( p->pLeft ){
          p = p->pLeft;
        }
      }
    }
  }

  *ppExpr = p;
  *piToken = iToken;
  return p?1:0;
}

/*
** Return TRUE if the expression node pExpr is located beneath the
** RHS of a NOT operator.
*/
static int fts3ExprBeneathNot(Fts3Expr *p){
  Fts3Expr *pParent;
  while( p ){
    pParent = p->pParent;
    if( pParent && pParent->eType==FTSQUERY_NOT && pParent->pRight==p ){
      return 1;
    }
    p = pParent;
  }
  return 0;
}

/*
** Add entries to pSnippet->aMatch[] for every match that occurs against
** document zDoc[0..nDoc-1] which is stored in column iColumn.
*/
static void snippetOffsetsOfColumn(
  fulltext_cursor *pCur,         /* The fulltest search cursor */
  Snippet *pSnippet,             /* The Snippet object to be filled in */
  int iColumn,                   /* Index of fulltext table column */
  const char *zDoc,              /* Text of the fulltext table column */
  int nDoc                       /* Length of zDoc in bytes */
){
  const sqlite3_tokenizer_module *pTModule;  /* The tokenizer module */
  sqlite3_tokenizer *pTokenizer;             /* The specific tokenizer */
  sqlite3_tokenizer_cursor *pTCursor;        /* Tokenizer cursor */
  fulltext_vtab *pVtab;                /* The full text index */
  int nColumn;                         /* Number of columns in the index */
  int i, j;                            /* Loop counters */
  int rc;                              /* Return code */
  unsigned int match, prevMatch;       /* Phrase search bitmasks */
  const char *zToken;                  /* Next token from the tokenizer */
  int nToken;                          /* Size of zToken */
  int iBegin, iEnd, iPos;              /* Offsets of beginning and end */

  /* The following variables keep a circular buffer of the last
  ** few tokens */
  unsigned int iRotor = 0;             /* Index of current token */
  int iRotorBegin[FTS3_ROTOR_SZ];      /* Beginning offset of token */
  int iRotorLen[FTS3_ROTOR_SZ];        /* Length of token */

  pVtab = cursor_vtab(pCur);
  nColumn = pVtab->nColumn;
  pTokenizer = pVtab->pTokenizer;
  pTModule = pTokenizer->pModule;
  rc = pTModule->xOpen(pTokenizer, zDoc, nDoc, &pTCursor);
  if( rc ) return;
  pTCursor->pTokenizer = pTokenizer;

  prevMatch = 0;
  while( !pTModule->xNext(pTCursor, &zToken, &nToken, &iBegin, &iEnd, &iPos) ){
    Fts3Expr *pIter = pCur->pExpr;
    int iIter = -1;
    iRotorBegin[iRotor&FTS3_ROTOR_MASK] = iBegin;
    iRotorLen[iRotor&FTS3_ROTOR_MASK] = iEnd-iBegin;
    match = 0;
    for(i=0; i<(FTS3_ROTOR_SZ-1) && fts3NextExprToken(&pIter, &iIter); i++){
      int nPhrase;                    /* Number of tokens in current phrase */
      struct PhraseToken *pToken;     /* Current token */
      int iCol;                       /* Column index */

      if( fts3ExprBeneathNot(pIter) ) continue;
      nPhrase = pIter->pPhrase->nToken;
      pToken = &pIter->pPhrase->aToken[iIter];
      iCol = pIter->pPhrase->iColumn;
      if( iCol>=0 && iCol<nColumn && iCol!=iColumn ) continue;
      if( pToken->n>nToken ) continue;
      if( !pToken->isPrefix && pToken->n<nToken ) continue;
      assert( pToken->n<=nToken );
      if( memcmp(pToken->z, zToken, pToken->n) ) continue;
      if( iIter>0 && (prevMatch & (1<<i))==0 ) continue;
      match |= 1<<i;
      if( i==(FTS3_ROTOR_SZ-2) || nPhrase==iIter+1 ){
        for(j=nPhrase-1; j>=0; j--){
          int k = (iRotor-j) & FTS3_ROTOR_MASK;
          snippetAppendMatch(pSnippet, iColumn, i-j, iPos-j,
                iRotorBegin[k], iRotorLen[k]);
        }
      }
    }
    prevMatch = match<<1;
    iRotor++;
  }
  pTModule->xClose(pTCursor);  
}

/*
** Remove entries from the pSnippet structure to account for the NEAR
** operator. When this is called, pSnippet contains the list of token 
** offsets produced by treating all NEAR operators as AND operators.
** This function removes any entries that should not be present after
** accounting for the NEAR restriction. For example, if the queried
** document is:
**
**     "A B C D E A"
**
** and the query is:
** 
**     A NEAR/0 E
**
** then when this function is called the Snippet contains token offsets
** 0, 4 and 5. This function removes the "0" entry (because the first A
** is not near enough to an E).
**
** When this function is called, the value pointed to by parameter piLeft is
** the integer id of the left-most token in the expression tree headed by
** pExpr. This function increments *piLeft by the total number of tokens
** in the expression tree headed by pExpr.
**
** Return 1 if any trimming occurs.  Return 0 if no trimming is required.
*/
static int trimSnippetOffsets(
  Fts3Expr *pExpr,      /* The search expression */
  Snippet *pSnippet,    /* The set of snippet offsets to be trimmed */
  int *piLeft           /* Index of left-most token in pExpr */
){
  if( pExpr ){
    if( trimSnippetOffsets(pExpr->pLeft, pSnippet, piLeft) ){
      return 1;
    }

    switch( pExpr->eType ){
      case FTSQUERY_PHRASE:
        *piLeft += pExpr->pPhrase->nToken;
        break;
      case FTSQUERY_NEAR: {
        /* The right-hand-side of a NEAR operator is always a phrase. The
        ** left-hand-side is either a phrase or an expression tree that is 
        ** itself headed by a NEAR operator. The following initializations
        ** set local variable iLeft to the token number of the left-most
        ** token in the right-hand phrase, and iRight to the right most
        ** token in the same phrase. For example, if we had:
        **
        **     <col> MATCH '"abc def" NEAR/2 "ghi jkl"'
        **
        ** then iLeft will be set to 2 (token number of ghi) and nToken will
        ** be set to 4.
        */
        Fts3Expr *pLeft = pExpr->pLeft;
        Fts3Expr *pRight = pExpr->pRight;
        int iLeft = *piLeft;
        int nNear = pExpr->nNear;
        int nToken = pRight->pPhrase->nToken;
        int jj, ii;
        if( pLeft->eType==FTSQUERY_NEAR ){
          pLeft = pLeft->pRight;
        }
        assert( pRight->eType==FTSQUERY_PHRASE );
        assert( pLeft->eType==FTSQUERY_PHRASE );
        nToken += pLeft->pPhrase->nToken;

        for(ii=0; ii<pSnippet->nMatch; ii++){
          struct snippetMatch *p = &pSnippet->aMatch[ii];
          if( p->iTerm==iLeft ){
            int isOk = 0;
            /* Snippet ii is an occurence of query term iLeft in the document.
            ** It occurs at position (p->iToken) of the document. We now
            ** search for an instance of token (iLeft-1) somewhere in the 
            ** range (p->iToken - nNear)...(p->iToken + nNear + nToken) within 
            ** the set of snippetMatch structures. If one is found, proceed. 
            ** If one cannot be found, then remove snippets ii..(ii+N-1) 
            ** from the matching snippets, where N is the number of tokens 
            ** in phrase pRight->pPhrase.
            */
            for(jj=0; isOk==0 && jj<pSnippet->nMatch; jj++){
              struct snippetMatch *p2 = &pSnippet->aMatch[jj];
              if( p2->iTerm==(iLeft-1) ){
                if( p2->iToken>=(p->iToken-nNear-1) 
                 && p2->iToken<(p->iToken+nNear+nToken) 
                ){
                  isOk = 1;
                }
              }
            }
            if( !isOk ){
              int kk;
              for(kk=0; kk<pRight->pPhrase->nToken; kk++){
                pSnippet->aMatch[kk+ii].iTerm = -2;
              }
              return 1;
            }
          }
          if( p->iTerm==(iLeft-1) ){
            int isOk = 0;
            for(jj=0; isOk==0 && jj<pSnippet->nMatch; jj++){
              struct snippetMatch *p2 = &pSnippet->aMatch[jj];
              if( p2->iTerm==iLeft ){
                if( p2->iToken<=(p->iToken+nNear+1) 
                 && p2->iToken>(p->iToken-nNear-nToken) 
                ){
                  isOk = 1;
                }
              }
            }
            if( !isOk ){
              int kk;
              for(kk=0; kk<pLeft->pPhrase->nToken; kk++){
                pSnippet->aMatch[ii-kk].iTerm = -2;
              }
              return 1;
            }
          }
        }
        break;
      }
    }

    if( trimSnippetOffsets(pExpr->pRight, pSnippet, piLeft) ){
      return 1;
    }
  }
  return 0;
}

/*
** Compute all offsets for the current row of the query.  
** If the offsets have already been computed, this routine is a no-op.
*/
static void snippetAllOffsets(fulltext_cursor *p){
  int nColumn;
  int iColumn, i;
  int iFirst, iLast;
  int iTerm = 0;
  fulltext_vtab *pFts = cursor_vtab(p);

  if( p->snippet.nMatch || p->pExpr==0 ){
    return;
  }
  nColumn = pFts->nColumn;
  iColumn = (p->iCursorType - QUERY_FULLTEXT);
  if( iColumn<0 || iColumn>=nColumn ){
    /* Look for matches over all columns of the full-text index */
    iFirst = 0;
    iLast = nColumn-1;
  }else{
    /* Look for matches in the iColumn-th column of the index only */
    iFirst = iColumn;
    iLast = iColumn;
  }
  for(i=iFirst; i<=iLast; i++){
    const char *zDoc;
    int nDoc;
    zDoc = (const char*)sqlite3_column_text(p->pStmt, i+1);
    nDoc = sqlite3_column_bytes(p->pStmt, i+1);
    snippetOffsetsOfColumn(p, &p->snippet, i, zDoc, nDoc);
  }

  while( trimSnippetOffsets(p->pExpr, &p->snippet, &iTerm) ){
    iTerm = 0;
  }
}

/*
** Convert the information in the aMatch[] array of the snippet
** into the string zOffset[0..nOffset-1]. This string is used as
** the return of the SQL offsets() function.
*/
static void snippetOffsetText(Snippet *p){
  int i;
  int cnt = 0;
  StringBuffer sb;
  char zBuf[200];
  if( p->zOffset ) return;
  initStringBuffer(&sb);
  for(i=0; i<p->nMatch; i++){
    struct snippetMatch *pMatch = &p->aMatch[i];
    if( pMatch->iTerm>=0 ){
      /* If snippetMatch.iTerm is less than 0, then the match was 
      ** discarded as part of processing the NEAR operator (see the 
      ** trimSnippetOffsetsForNear() function for details). Ignore 
      ** it in this case
      */
      zBuf[0] = ' ';
      sqlite3_snprintf(sizeof(zBuf)-1, &zBuf[cnt>0], "%d %d %d %d",
          pMatch->iCol, pMatch->iTerm, pMatch->iStart, pMatch->nByte);
      append(&sb, zBuf);
      cnt++;
    }
  }
  p->zOffset = stringBufferData(&sb);
  p->nOffset = stringBufferLength(&sb);
}

/*
** zDoc[0..nDoc-1] is phrase of text.  aMatch[0..nMatch-1] are a set
** of matching words some of which might be in zDoc.  zDoc is column
** number iCol.
**
** iBreak is suggested spot in zDoc where we could begin or end an
** excerpt.  Return a value similar to iBreak but possibly adjusted
** to be a little left or right so that the break point is better.
*/
static int wordBoundary(
  int iBreak,                   /* The suggested break point */
  const char *zDoc,             /* Document text */
  int nDoc,                     /* Number of bytes in zDoc[] */
  struct snippetMatch *aMatch,  /* Matching words */
  int nMatch,                   /* Number of entries in aMatch[] */
  int iCol                      /* The column number for zDoc[] */
){
  int i;
  if( iBreak<=10 ){
    return 0;
  }
  if( iBreak>=nDoc-10 ){
    return nDoc;
  }
  for(i=0; i<nMatch && aMatch[i].iCol<iCol; i++){}
  while( i<nMatch && aMatch[i].iStart+aMatch[i].nByte<iBreak ){ i++; }
  if( i<nMatch ){
    if( aMatch[i].iStart<iBreak+10 ){
      return aMatch[i].iStart;
    }
    if( i>0 && aMatch[i-1].iStart+aMatch[i-1].nByte>=iBreak ){
      return aMatch[i-1].iStart;
    }
  }
  for(i=1; i<=10; i++){
    if( safe_isspace(zDoc[iBreak-i]) ){
      return iBreak - i + 1;
    }
    if( safe_isspace(zDoc[iBreak+i]) ){
      return iBreak + i + 1;
    }
  }
  return iBreak;
}



/*
** Allowed values for Snippet.aMatch[].snStatus
*/
#define SNIPPET_IGNORE  0   /* It is ok to omit this match from the snippet */
#define SNIPPET_DESIRED 1   /* We want to include this match in the snippet */

/*
** Generate the text of a snippet.
*/
static void snippetText(
  fulltext_cursor *pCursor,   /* The cursor we need the snippet for */
  const char *zStartMark,     /* Markup to appear before each match */
  const char *zEndMark,       /* Markup to appear after each match */
  const char *zEllipsis       /* Ellipsis mark */
){
  int i, j;
  struct snippetMatch *aMatch;
  int nMatch;
  int nDesired;
  StringBuffer sb;
  int tailCol;
  int tailOffset;
  int iCol;
  int nDoc;
  const char *zDoc;
  int iStart, iEnd;
  int tailEllipsis = 0;
  int iMatch;
  

  sqlite3_free(pCursor->snippet.zSnippet);
  pCursor->snippet.zSnippet = 0;
  aMatch = pCursor->snippet.aMatch;
  nMatch = pCursor->snippet.nMatch;
  initStringBuffer(&sb);

  for(i=0; i<nMatch; i++){
    aMatch[i].snStatus = SNIPPET_IGNORE;
  }
  nDesired = 0;
  for(i=0; i<FTS3_ROTOR_SZ; i++){
    for(j=0; j<nMatch; j++){
      if( aMatch[j].iTerm==i ){
        aMatch[j].snStatus = SNIPPET_DESIRED;
        nDesired++;
        break;
      }
    }
  }

  iMatch = 0;
  tailCol = -1;
  tailOffset = 0;
  for(i=0; i<nMatch && nDesired>0; i++){
    if( aMatch[i].snStatus!=SNIPPET_DESIRED ) continue;
    nDesired--;
    iCol = aMatch[i].iCol;
    zDoc = (const char*)sqlite3_column_text(pCursor->pStmt, iCol+1);
    nDoc = sqlite3_column_bytes(pCursor->pStmt, iCol+1);
    iStart = aMatch[i].iStart - 40;
    iStart = wordBoundary(iStart, zDoc, nDoc, aMatch, nMatch, iCol);
    if( iStart<=10 ){
      iStart = 0;
    }
    if( iCol==tailCol && iStart<=tailOffset+20 ){
      iStart = tailOffset;
    }
    if( (iCol!=tailCol && tailCol>=0) || iStart!=tailOffset ){
      trimWhiteSpace(&sb);
      appendWhiteSpace(&sb);
      append(&sb, zEllipsis);
      appendWhiteSpace(&sb);
    }
    iEnd = aMatch[i].iStart + aMatch[i].nByte + 40;
    iEnd = wordBoundary(iEnd, zDoc, nDoc, aMatch, nMatch, iCol);
    if( iEnd>=nDoc-10 ){
      iEnd = nDoc;
      tailEllipsis = 0;
    }else{
      tailEllipsis = 1;
    }
    while( iMatch<nMatch && aMatch[iMatch].iCol<iCol ){ iMatch++; }
    while( iStart<iEnd ){
      while( iMatch<nMatch && aMatch[iMatch].iStart<iStart
             && aMatch[iMatch].iCol<=iCol ){
        iMatch++;
      }
      if( iMatch<nMatch && aMatch[iMatch].iStart<iEnd
             && aMatch[iMatch].iCol==iCol ){
        nappend(&sb, &zDoc[iStart], aMatch[iMatch].iStart - iStart);
        iStart = aMatch[iMatch].iStart;
        append(&sb, zStartMark);
        nappend(&sb, &zDoc[iStart], aMatch[iMatch].nByte);
        append(&sb, zEndMark);
        iStart += aMatch[iMatch].nByte;
        for(j=iMatch+1; j<nMatch; j++){
          if( aMatch[j].iTerm==aMatch[iMatch].iTerm
              && aMatch[j].snStatus==SNIPPET_DESIRED ){
            nDesired--;
            aMatch[j].snStatus = SNIPPET_IGNORE;
          }
        }
      }else{
        nappend(&sb, &zDoc[iStart], iEnd - iStart);
        iStart = iEnd;
      }
    }
    tailCol = iCol;
    tailOffset = iEnd;
  }
  trimWhiteSpace(&sb);
  if( tailEllipsis ){
    appendWhiteSpace(&sb);
    append(&sb, zEllipsis);
  }
  pCursor->snippet.zSnippet = stringBufferData(&sb);
  pCursor->snippet.nSnippet = stringBufferLength(&sb);
}


/*
** Close the cursor.  For additional information see the documentation
** on the xClose method of the virtual table interface.
*/
static int fulltextClose(sqlite3_vtab_cursor *pCursor){
  fulltext_cursor *c = (fulltext_cursor *) pCursor;
  FTSTRACE(("FTS3 Close %p\n", c));
  sqlite3_finalize(c->pStmt);
  sqlite3Fts3ExprFree(c->pExpr);
  snippetClear(&c->snippet);
  if( c->result.nData!=0 ){
    dlrDestroy(&c->reader);
  }
  dataBufferDestroy(&c->result);
  sqlite3_free(c);
  return SQLITE_OK;
}

static int fulltextNext(sqlite3_vtab_cursor *pCursor){
  fulltext_cursor *c = (fulltext_cursor *) pCursor;
  int rc;

  FTSTRACE(("FTS3 Next %p\n", pCursor));
  snippetClear(&c->snippet);
  if( c->iCursorType < QUERY_FULLTEXT ){
    /* TODO(shess) Handle SQLITE_SCHEMA AND SQLITE_BUSY. */
    rc = sqlite3_step(c->pStmt);
    switch( rc ){
      case SQLITE_ROW:
        c->eof = 0;
        return SQLITE_OK;
      case SQLITE_DONE:
        c->eof = 1;
        return SQLITE_OK;
      default:
        c->eof = 1;
        return rc;
    }
  } else {  /* full-text query */
    rc = sqlite3_reset(c->pStmt);
    if( rc!=SQLITE_OK ) return rc;

    if( c->result.nData==0 || dlrAtEnd(&c->reader) ){
      c->eof = 1;
      return SQLITE_OK;
    }
    rc = sqlite3_bind_int64(c->pStmt, 1, dlrDocid(&c->reader));
    dlrStep(&c->reader);
    if( rc!=SQLITE_OK ) return rc;
    /* TODO(shess) Handle SQLITE_SCHEMA AND SQLITE_BUSY. */
    rc = sqlite3_step(c->pStmt);
    if( rc==SQLITE_ROW ){   /* the case we expect */
      c->eof = 0;
      return SQLITE_OK;
    }
    /* an error occurred; abort */
    return rc==SQLITE_DONE ? SQLITE_ERROR : rc;
  }
}


/* TODO(shess) If we pushed LeafReader to the top of the file, or to
** another file, term_select() could be pushed above
** docListOfTerm().
*/
static int termSelect(fulltext_vtab *v, int iColumn,
                      const char *pTerm, int nTerm, int isPrefix,
                      DocListType iType, DataBuffer *out);

/* 
** Return a DocList corresponding to the phrase *pPhrase.
**
** The resulting DL_DOCIDS doclist is stored in pResult, which is
** overwritten.
*/
static int docListOfPhrase(
  fulltext_vtab *pTab,   /* The full text index */
  Fts3Phrase *pPhrase,   /* Phrase to return a doclist corresponding to */
  DocListType eListType, /* Either DL_DOCIDS or DL_POSITIONS */
  DataBuffer *pResult    /* Write the result here */
){
  int ii;
  int rc = SQLITE_OK;
  int iCol = pPhrase->iColumn;
  DocListType eType = eListType;
  assert( eType==DL_POSITIONS || eType==DL_DOCIDS );
  if( pPhrase->nToken>1 ){
    eType = DL_POSITIONS;
  }

  /* This code should never be called with buffered updates. */
  assert( pTab->nPendingData<0 );

  for(ii=0; rc==SQLITE_OK && ii<pPhrase->nToken; ii++){
    DataBuffer tmp;
    struct PhraseToken *p = &pPhrase->aToken[ii];
    rc = termSelect(pTab, iCol, p->z, p->n, p->isPrefix, eType, &tmp);
    if( rc==SQLITE_OK ){
      if( ii==0 ){
        *pResult = tmp;
      }else{
        DataBuffer res = *pResult;
        dataBufferInit(pResult, 0);
        if( ii==(pPhrase->nToken-1) ){
          eType = eListType;
        }
        docListPhraseMerge(
          res.pData, res.nData, tmp.pData, tmp.nData, 0, 0, eType, pResult
        );
        dataBufferDestroy(&res);
        dataBufferDestroy(&tmp);
      }























































































































































































    }
  }







  return rc;
}

/*
** Evaluate the full-text expression pExpr against fts3 table pTab. Write
** the results into pRes.
*/
static int evalFts3Expr(
  fulltext_vtab *pTab,           /* Fts3 Virtual table object */
  Fts3Expr *pExpr,               /* Parsed fts3 expression */
  DataBuffer *pRes               /* OUT: Write results of the expression here */

){
  int rc = SQLITE_OK;

  /* Initialize the output buffer. If this is an empty query (pExpr==0), 
  ** this is all that needs to be done. Empty queries produce empty 
  ** result sets.
  */
  dataBufferInit(pRes, 0);

  if( pExpr ){
    if( pExpr->eType==FTSQUERY_PHRASE ){
      DocListType eType = DL_DOCIDS;
      if( pExpr->pParent && pExpr->pParent->eType==FTSQUERY_NEAR ){
        eType = DL_POSITIONS;
      }
      rc = docListOfPhrase(pTab, pExpr->pPhrase, eType, pRes);

    }else{



      DataBuffer lhs;
      DataBuffer rhs;

      dataBufferInit(&rhs, 0);
      if( SQLITE_OK==(rc = evalFts3Expr(pTab, pExpr->pLeft, &lhs)) 
       && SQLITE_OK==(rc = evalFts3Expr(pTab, pExpr->pRight, &rhs)) 
      ){



        switch( pExpr->eType ){
          case FTSQUERY_NEAR: {
            int nToken;
            Fts3Expr *pLeft;
            DocListType eType = DL_DOCIDS;





            if( pExpr->pParent && pExpr->pParent->eType==FTSQUERY_NEAR ){
              eType = DL_POSITIONS;

            }
            pLeft = pExpr->pLeft;
            while( pLeft->eType==FTSQUERY_NEAR ){ 
              pLeft=pLeft->pRight;
            }

            assert( pExpr->pRight->eType==FTSQUERY_PHRASE );
            assert( pLeft->eType==FTSQUERY_PHRASE );
            nToken = pLeft->pPhrase->nToken + pExpr->pRight->pPhrase->nToken;
            docListPhraseMerge(lhs.pData, lhs.nData, rhs.pData, rhs.nData, 

                pExpr->nNear+1, nToken, eType, pRes




            );
            break;




          }
          case FTSQUERY_NOT: {
            docListExceptMerge(lhs.pData, lhs.nData, rhs.pData, rhs.nData,pRes);

            break;
          }

          case FTSQUERY_AND: {
            docListAndMerge(lhs.pData, lhs.nData, rhs.pData, rhs.nData, pRes);











            break;
          }
          case FTSQUERY_OR: {
            docListOrMerge(lhs.pData, lhs.nData, rhs.pData, rhs.nData, pRes);







            break;
          }
        }
      }
      dataBufferDestroy(&lhs);
      dataBufferDestroy(&rhs);
    }
  }

  return rc;
}

/* TODO(shess) Refactor the code to remove this forward decl. */
static int flushPendingTerms(fulltext_vtab *v);

/* Perform a full-text query using the search expression in
** zInput[0..nInput-1].  Return a list of matching documents
** in pResult.
**
** Queries must match column iColumn.  Or if iColumn>=nColumn
** they are allowed to match against any column.
*/
static int fulltextQuery(
  fulltext_vtab *v,      /* The full text index */
  int iColumn,           /* Match against this column by default */
  const char *zInput,    /* The query string */
  int nInput,            /* Number of bytes in zInput[] */
  DataBuffer *pResult,   /* Write the result doclist here */
  Fts3Expr **ppExpr        /* Put parsed query string here */
){
  int rc;

  /* TODO(shess) Instead of flushing pendingTerms, we could query for
  ** the relevant term and merge the doclist into what we receive from
  ** the database.  Wait and see if this is a common issue, first.
  **
  ** A good reason not to flush is to not generate update-related
  ** error codes from here.
  */

  /* Flush any buffered updates before executing the query. */
  rc = flushPendingTerms(v);
  if( rc!=SQLITE_OK ){
    return rc;
  }

  /* Parse the query passed to the MATCH operator. */
  rc = sqlite3Fts3ExprParse(v->pTokenizer, 
      v->azColumn, v->nColumn, iColumn, zInput, nInput, ppExpr
  );
  if( rc!=SQLITE_OK ){
    assert( 0==(*ppExpr) );
    return rc;
  }

  return evalFts3Expr(v, *ppExpr, pResult);
}

/*
** This is the xFilter interface for the virtual table.  See
** the virtual table xFilter method documentation for additional
** information.
**
** If idxNum==QUERY_GENERIC then do a full table scan against
** the %_content table.
**
** If idxNum==QUERY_DOCID then do a docid lookup for a single entry
** in the %_content table.
**
** If idxNum>=QUERY_FULLTEXT then use the full text index.  The
** column on the left-hand side of the MATCH operator is column
** number idxNum-QUERY_FULLTEXT, 0 indexed.  argv[0] is the right-hand
** side of the MATCH operator.
*/
/* TODO(shess) Upgrade the cursor initialization and destruction to
** account for fulltextFilter() being called multiple times on the
** same cursor.  The current solution is very fragile.  Apply fix to
** fts3 as appropriate.
*/
static int fulltextFilter(
  sqlite3_vtab_cursor *pCursor,     /* The cursor used for this query */
  int idxNum, const char *idxStr,   /* Which indexing scheme to use */
  int argc, sqlite3_value **argv    /* Arguments for the indexing scheme */
){
  fulltext_cursor *c = (fulltext_cursor *) pCursor;
  fulltext_vtab *v = cursor_vtab(c);
  int rc;

  FTSTRACE(("FTS3 Filter %p\n",pCursor));

  /* If the cursor has a statement that was not prepared according to
  ** idxNum, clear it.  I believe all calls to fulltextFilter with a
  ** given cursor will have the same idxNum , but in this case it's
  ** easy to be safe.
  */
  if( c->pStmt && c->iCursorType!=idxNum ){
    sqlite3_finalize(c->pStmt);
    c->pStmt = NULL;
  }

  /* Get a fresh statement appropriate to idxNum. */
  /* TODO(shess): Add a prepared-statement cache in the vt structure.
  ** The cache must handle multiple open cursors.  Easier to cache the
  ** statement variants at the vt to reduce malloc/realloc/free here.
  ** Or we could have a StringBuffer variant which allowed stack
  ** construction for small values.
  */
  if( !c->pStmt ){
    StringBuffer sb;
    initStringBuffer(&sb);
    append(&sb, "SELECT docid, ");
    appendList(&sb, v->nColumn, v->azContentColumn);
    append(&sb, " FROM %_content");
    if( idxNum!=QUERY_GENERIC ) append(&sb, " WHERE docid = ?");
    rc = sql_prepare(v->db, v->zDb, v->zName, &c->pStmt,
                     stringBufferData(&sb));
    stringBufferDestroy(&sb);
    if( rc!=SQLITE_OK ) return rc;
    c->iCursorType = idxNum;
  }else{
    sqlite3_reset(c->pStmt);
    assert( c->iCursorType==idxNum );
  }

  switch( idxNum ){
    case QUERY_GENERIC:
      break;

    case QUERY_DOCID:
      rc = sqlite3_bind_int64(c->pStmt, 1, sqlite3_value_int64(argv[0]));
      if( rc!=SQLITE_OK ) return rc;
      break;

    default:   /* full-text search */
    {
      int iCol = idxNum-QUERY_FULLTEXT;
      const char *zQuery = (const char *)sqlite3_value_text(argv[0]);
      assert( idxNum<=QUERY_FULLTEXT+v->nColumn);
      assert( argc==1 );
      if( c->result.nData!=0 ){
        /* This case happens if the same cursor is used repeatedly. */
        dlrDestroy(&c->reader);
        dataBufferReset(&c->result);
      }else{
        dataBufferInit(&c->result, 0);
      }
      rc = fulltextQuery(v, iCol, zQuery, -1, &c->result, &c->pExpr);
      if( rc!=SQLITE_OK ) return rc;
      if( c->result.nData!=0 ){
        dlrInit(&c->reader, DL_DOCIDS, c->result.pData, c->result.nData);
      }
      break;
    }
  }

  return fulltextNext(pCursor);
}

/* This is the xEof method of the virtual table.  The SQLite core
** calls this routine to find out if it has reached the end of
** a query's results set.
*/
static int fulltextEof(sqlite3_vtab_cursor *pCursor){
  fulltext_cursor *c = (fulltext_cursor *) pCursor;
  return c->eof;
}

/* This is the xColumn method of the virtual table.  The SQLite
** core calls this method during a query when it needs the value
** of a column from the virtual table.  This method needs to use
** one of the sqlite3_result_*() routines to store the requested
** value back in the pContext.
*/
static int fulltextColumn(sqlite3_vtab_cursor *pCursor,
                          sqlite3_context *pContext, int idxCol){
  fulltext_cursor *c = (fulltext_cursor *) pCursor;
  fulltext_vtab *v = cursor_vtab(c);

  if( idxCol<v->nColumn ){
    sqlite3_value *pVal = sqlite3_column_value(c->pStmt, idxCol+1);
    sqlite3_result_value(pContext, pVal);
  }else if( idxCol==v->nColumn ){
    /* The extra column whose name is the same as the table.
    ** Return a blob which is a pointer to the cursor
    */
    sqlite3_result_blob(pContext, &c, sizeof(c), SQLITE_TRANSIENT);
  }else if( idxCol==v->nColumn+1 ){
    /* The docid column, which is an alias for rowid. */
    sqlite3_value *pVal = sqlite3_column_value(c->pStmt, 0);
    sqlite3_result_value(pContext, pVal);
  }
  return SQLITE_OK;
}

/* This is the xRowid method.  The SQLite core calls this routine to
** retrieve the rowid for the current row of the result set.  fts3
** exposes %_content.docid as the rowid for the virtual table.  The
** rowid should be written to *pRowid.
*/
static int fulltextRowid(sqlite3_vtab_cursor *pCursor, sqlite_int64 *pRowid){
  fulltext_cursor *c = (fulltext_cursor *) pCursor;

  *pRowid = sqlite3_column_int64(c->pStmt, 0);
  return SQLITE_OK;
}

/* Add all terms in [zText] to pendingTerms table.  If [iColumn] > 0,
** we also store positions and offsets in the hash table using that
** column number.
*/
static int buildTerms(fulltext_vtab *v, sqlite_int64 iDocid,
                      const char *zText, int iColumn){
  sqlite3_tokenizer *pTokenizer = v->pTokenizer;
  sqlite3_tokenizer_cursor *pCursor;
  const char *pToken;
  int nTokenBytes;
  int iStartOffset, iEndOffset, iPosition;
  int rc;

  rc = pTokenizer->pModule->xOpen(pTokenizer, zText, -1, &pCursor);
  if( rc!=SQLITE_OK ) return rc;

  pCursor->pTokenizer = pTokenizer;
  while( SQLITE_OK==(rc=pTokenizer->pModule->xNext(pCursor,
                                                   &pToken, &nTokenBytes,
                                                   &iStartOffset, &iEndOffset,
                                                   &iPosition)) ){
    DLCollector *p;
    int nData;                   /* Size of doclist before our update. */

    /* Positions can't be negative; we use -1 as a terminator
     * internally.  Token can't be NULL or empty. */
    if( iPosition<0 || pToken == NULL || nTokenBytes == 0 ){
      rc = SQLITE_ERROR;
      break;
    }

    p = fts3HashFind(&v->pendingTerms, pToken, nTokenBytes);
    if( p==NULL ){
      nData = 0;
      p = dlcNew(iDocid, DL_DEFAULT);
      fts3HashInsert(&v->pendingTerms, pToken, nTokenBytes, p);

      /* Overhead for our hash table entry, the key, and the value. */
      v->nPendingData += sizeof(struct fts3HashElem)+sizeof(*p)+nTokenBytes;
    }else{
      nData = p->b.nData;
      if( p->dlw.iPrevDocid!=iDocid ) dlcNext(p, iDocid);
    }
    if( iColumn>=0 ){
      dlcAddPos(p, iColumn, iPosition, iStartOffset, iEndOffset);
    }

    /* Accumulate data added by dlcNew or dlcNext, and dlcAddPos. */
    v->nPendingData += p->b.nData-nData;
  }

  /* TODO(shess) Check return?  Should this be able to cause errors at
  ** this point?  Actually, same question about sqlite3_finalize(),
  ** though one could argue that failure there means that the data is
  ** not durable.  *ponder*
  */
  pTokenizer->pModule->xClose(pCursor);
  if( SQLITE_DONE == rc ) return SQLITE_OK;
  return rc;
}

/* Add doclists for all terms in [pValues] to pendingTerms table. */
static int insertTerms(fulltext_vtab *v, sqlite_int64 iDocid,
                       sqlite3_value **pValues){
  int i;
  for(i = 0; i < v->nColumn ; ++i){
    char *zText = (char*)sqlite3_value_text(pValues[i]);
    int rc = buildTerms(v, iDocid, zText, i);
    if( rc!=SQLITE_OK ) return rc;
  }
  return SQLITE_OK;
}

/* Add empty doclists for all terms in the given row's content to
** pendingTerms.
*/
static int deleteTerms(fulltext_vtab *v, sqlite_int64 iDocid){
  const char **pValues;
  int i, rc;

  /* TODO(shess) Should we allow such tables at all? */
  if( DL_DEFAULT==DL_DOCIDS ) return SQLITE_ERROR;

  rc = content_select(v, iDocid, &pValues);
  if( rc!=SQLITE_OK ) return rc;

  for(i = 0 ; i < v->nColumn; ++i) {
    rc = buildTerms(v, iDocid, pValues[i], -1);
    if( rc!=SQLITE_OK ) break;
  }

  freeStringArray(v->nColumn, pValues);
  return SQLITE_OK;
}

/* TODO(shess) Refactor the code to remove this forward decl. */
static int initPendingTerms(fulltext_vtab *v, sqlite_int64 iDocid);

/* Insert a row into the %_content table; set *piDocid to be the ID of the
** new row.  Add doclists for terms to pendingTerms.
*/
static int index_insert(fulltext_vtab *v, sqlite3_value *pRequestDocid,
                        sqlite3_value **pValues, sqlite_int64 *piDocid){
  int rc;

  rc = content_insert(v, pRequestDocid, pValues);  /* execute an SQL INSERT */
  if( rc!=SQLITE_OK ) return rc;

  /* docid column is an alias for rowid. */
  *piDocid = sqlite3_last_insert_rowid(v->db);
  rc = initPendingTerms(v, *piDocid);
  if( rc!=SQLITE_OK ) return rc;

  return insertTerms(v, *piDocid, pValues);
}

/* Delete a row from the %_content table; add empty doclists for terms
** to pendingTerms.
*/
static int index_delete(fulltext_vtab *v, sqlite_int64 iRow){
  int rc = initPendingTerms(v, iRow);
  if( rc!=SQLITE_OK ) return rc;

  rc = deleteTerms(v, iRow);
  if( rc!=SQLITE_OK ) return rc;

  return content_delete(v, iRow);  /* execute an SQL DELETE */
}

/* Update a row in the %_content table; add delete doclists to
** pendingTerms for old terms not in the new data, add insert doclists
** to pendingTerms for terms in the new data.
*/
static int index_update(fulltext_vtab *v, sqlite_int64 iRow,
                        sqlite3_value **pValues){
  int rc = initPendingTerms(v, iRow);
  if( rc!=SQLITE_OK ) return rc;

  /* Generate an empty doclist for each term that previously appeared in this
   * row. */
  rc = deleteTerms(v, iRow);
  if( rc!=SQLITE_OK ) return rc;

  rc = content_update(v, pValues, iRow);  /* execute an SQL UPDATE */
  if( rc!=SQLITE_OK ) return rc;

  /* Now add positions for terms which appear in the updated row. */
  return insertTerms(v, iRow, pValues);
}

/*******************************************************************/
/* InteriorWriter is used to collect terms and block references into
** interior nodes in %_segments.  See commentary at top of file for
** format.
*/

/* How large interior nodes can grow. */
#define INTERIOR_MAX 2048

/* Minimum number of terms per interior node (except the root). This
** prevents large terms from making the tree too skinny - must be >0
** so that the tree always makes progress.  Note that the min tree
** fanout will be INTERIOR_MIN_TERMS+1.
*/
#define INTERIOR_MIN_TERMS 7
#if INTERIOR_MIN_TERMS<1
# error INTERIOR_MIN_TERMS must be greater than 0.
#endif

/* ROOT_MAX controls how much data is stored inline in the segment
** directory.
*/
/* TODO(shess) Push ROOT_MAX down to whoever is writing things.  It's
** only here so that interiorWriterRootInfo() and leafWriterRootInfo()
** can both see it, but if the caller passed it in, we wouldn't even
** need a define.
*/
#define ROOT_MAX 1024
#if ROOT_MAX<VARINT_MAX*2
# error ROOT_MAX must have enough space for a header.
#endif

/* InteriorBlock stores a linked-list of interior blocks while a lower
** layer is being constructed.
*/
typedef struct InteriorBlock {
  DataBuffer term;           /* Leftmost term in block's subtree. */
  DataBuffer data;           /* Accumulated data for the block. */
  struct InteriorBlock *next;
} InteriorBlock;

static InteriorBlock *interiorBlockNew(int iHeight, sqlite_int64 iChildBlock,
                                       const char *pTerm, int nTerm){
  InteriorBlock *block = sqlite3_malloc(sizeof(InteriorBlock));
  char c[VARINT_MAX+VARINT_MAX];
  int n;

  if( block ){
    memset(block, 0, sizeof(*block));
    dataBufferInit(&block->term, 0);
    dataBufferReplace(&block->term, pTerm, nTerm);

    n = fts3PutVarint(c, iHeight);
    n += fts3PutVarint(c+n, iChildBlock);
    dataBufferInit(&block->data, INTERIOR_MAX);
    dataBufferReplace(&block->data, c, n);
  }
  return block;
}

#ifndef NDEBUG
/* Verify that the data is readable as an interior node. */
static void interiorBlockValidate(InteriorBlock *pBlock){
  const char *pData = pBlock->data.pData;
  int nData = pBlock->data.nData;
  int n, iDummy;
  sqlite_int64 iBlockid;

  assert( nData>0 );
  assert( pData!=0 );
  assert( pData+nData>pData );

  /* Must lead with height of node as a varint(n), n>0 */
  n = fts3GetVarint32(pData, &iDummy);
  assert( n>0 );
  assert( iDummy>0 );
  assert( n<nData );
  pData += n;
  nData -= n;

  /* Must contain iBlockid. */
  n = fts3GetVarint(pData, &iBlockid);
  assert( n>0 );
  assert( n<=nData );
  pData += n;
  nData -= n;

  /* Zero or more terms of positive length */
  if( nData!=0 ){
    /* First term is not delta-encoded. */
    n = fts3GetVarint32(pData, &iDummy);
    assert( n>0 );
    assert( iDummy>0 );
    assert( n+iDummy>0);
    assert( n+iDummy<=nData );
    pData += n+iDummy;
    nData -= n+iDummy;

    /* Following terms delta-encoded. */
    while( nData!=0 ){
      /* Length of shared prefix. */
      n = fts3GetVarint32(pData, &iDummy);
      assert( n>0 );
      assert( iDummy>=0 );
      assert( n<nData );
      pData += n;
      nData -= n;

      /* Length and data of distinct suffix. */
      n = fts3GetVarint32(pData, &iDummy);
      assert( n>0 );
      assert( iDummy>0 );
      assert( n+iDummy>0);
      assert( n+iDummy<=nData );
      pData += n+iDummy;
      nData -= n+iDummy;
    }
  }
}
#define ASSERT_VALID_INTERIOR_BLOCK(x) interiorBlockValidate(x)
#else
#define ASSERT_VALID_INTERIOR_BLOCK(x) assert( 1 )
#endif

typedef struct InteriorWriter {
  int iHeight;                   /* from 0 at leaves. */
  InteriorBlock *first, *last;
  struct InteriorWriter *parentWriter;

  DataBuffer term;               /* Last term written to block "last". */
  sqlite_int64 iOpeningChildBlock; /* First child block in block "last". */
#ifndef NDEBUG
  sqlite_int64 iLastChildBlock;  /* for consistency checks. */
#endif
} InteriorWriter;

/* Initialize an interior node where pTerm[nTerm] marks the leftmost
** term in the tree.  iChildBlock is the leftmost child block at the
** next level down the tree.
*/
static void interiorWriterInit(int iHeight, const char *pTerm, int nTerm,
                               sqlite_int64 iChildBlock,
                               InteriorWriter *pWriter){
  InteriorBlock *block;
  assert( iHeight>0 );
  CLEAR(pWriter);

  pWriter->iHeight = iHeight;
  pWriter->iOpeningChildBlock = iChildBlock;
#ifndef NDEBUG
  pWriter->iLastChildBlock = iChildBlock;
#endif
  block = interiorBlockNew(iHeight, iChildBlock, pTerm, nTerm);
  pWriter->last = pWriter->first = block;
  ASSERT_VALID_INTERIOR_BLOCK(pWriter->last);
  dataBufferInit(&pWriter->term, 0);
}

/* Append the child node rooted at iChildBlock to the interior node,
** with pTerm[nTerm] as the leftmost term in iChildBlock's subtree.
*/
static void interiorWriterAppend(InteriorWriter *pWriter,
                                 const char *pTerm, int nTerm,
                                 sqlite_int64 iChildBlock){
  char c[VARINT_MAX+VARINT_MAX];
  int n, nPrefix = 0;

  ASSERT_VALID_INTERIOR_BLOCK(pWriter->last);

  /* The first term written into an interior node is actually
  ** associated with the second child added (the first child was added
  ** in interiorWriterInit, or in the if clause at the bottom of this
  ** function).  That term gets encoded straight up, with nPrefix left
  ** at 0.
  */
  if( pWriter->term.nData==0 ){
    n = fts3PutVarint(c, nTerm);
  }else{
    while( nPrefix<pWriter->term.nData &&
           pTerm[nPrefix]==pWriter->term.pData[nPrefix] ){
      nPrefix++;
    }

    n = fts3PutVarint(c, nPrefix);
    n += fts3PutVarint(c+n, nTerm-nPrefix);
  }

#ifndef NDEBUG
  pWriter->iLastChildBlock++;
#endif
  assert( pWriter->iLastChildBlock==iChildBlock );

  /* Overflow to a new block if the new term makes the current block
  ** too big, and the current block already has enough terms.
  */
  if( pWriter->last->data.nData+n+nTerm-nPrefix>INTERIOR_MAX &&
      iChildBlock-pWriter->iOpeningChildBlock>INTERIOR_MIN_TERMS ){
    pWriter->last->next = interiorBlockNew(pWriter->iHeight, iChildBlock,
                                           pTerm, nTerm);
    pWriter->last = pWriter->last->next;
    pWriter->iOpeningChildBlock = iChildBlock;
    dataBufferReset(&pWriter->term);
  }else{
    dataBufferAppend2(&pWriter->last->data, c, n,
                      pTerm+nPrefix, nTerm-nPrefix);
    dataBufferReplace(&pWriter->term, pTerm, nTerm);
  }
  ASSERT_VALID_INTERIOR_BLOCK(pWriter->last);
}

/* Free the space used by pWriter, including the linked-list of
** InteriorBlocks, and parentWriter, if present.
*/
static int interiorWriterDestroy(InteriorWriter *pWriter){
  InteriorBlock *block = pWriter->first;

  while( block!=NULL ){
    InteriorBlock *b = block;
    block = block->next;
    dataBufferDestroy(&b->term);
    dataBufferDestroy(&b->data);
    sqlite3_free(b);
  }
  if( pWriter->parentWriter!=NULL ){
    interiorWriterDestroy(pWriter->parentWriter);
    sqlite3_free(pWriter->parentWriter);
  }
  dataBufferDestroy(&pWriter->term);
  SCRAMBLE(pWriter);
  return SQLITE_OK;
}

/* If pWriter can fit entirely in ROOT_MAX, return it as the root info
** directly, leaving *piEndBlockid unchanged.  Otherwise, flush
** pWriter to %_segments, building a new layer of interior nodes, and
** recursively ask for their root into.
*/
static int interiorWriterRootInfo(fulltext_vtab *v, InteriorWriter *pWriter,
                                  char **ppRootInfo, int *pnRootInfo,
                                  sqlite_int64 *piEndBlockid){
  InteriorBlock *block = pWriter->first;
  sqlite_int64 iBlockid = 0;
  int rc;

  /* If we can fit the segment inline */
  if( block==pWriter->last && block->data.nData<ROOT_MAX ){
    *ppRootInfo = block->data.pData;
    *pnRootInfo = block->data.nData;
    return SQLITE_OK;
  }

  /* Flush the first block to %_segments, and create a new level of
  ** interior node.
  */
  ASSERT_VALID_INTERIOR_BLOCK(block);
  rc = block_insert(v, block->data.pData, block->data.nData, &iBlockid);
  if( rc!=SQLITE_OK ) return rc;
  *piEndBlockid = iBlockid;

  pWriter->parentWriter = sqlite3_malloc(sizeof(*pWriter->parentWriter));
  interiorWriterInit(pWriter->iHeight+1,
                     block->term.pData, block->term.nData,
                     iBlockid, pWriter->parentWriter);

  /* Flush additional blocks and append to the higher interior
  ** node.
  */
  for(block=block->next; block!=NULL; block=block->next){
    ASSERT_VALID_INTERIOR_BLOCK(block);
    rc = block_insert(v, block->data.pData, block->data.nData, &iBlockid);
    if( rc!=SQLITE_OK ) return rc;
    *piEndBlockid = iBlockid;

    interiorWriterAppend(pWriter->parentWriter,
                         block->term.pData, block->term.nData, iBlockid);
  }

  /* Parent node gets the chance to be the root. */
  return interiorWriterRootInfo(v, pWriter->parentWriter,
                                ppRootInfo, pnRootInfo, piEndBlockid);
}

/****************************************************************/
/* InteriorReader is used to read off the data from an interior node
** (see comment at top of file for the format).
*/
typedef struct InteriorReader {
  const char *pData;
  int nData;

  DataBuffer term;          /* previous term, for decoding term delta. */

  sqlite_int64 iBlockid;
} InteriorReader;

static void interiorReaderDestroy(InteriorReader *pReader){
  dataBufferDestroy(&pReader->term);
  SCRAMBLE(pReader);
}

/* TODO(shess) The assertions are great, but what if we're in NDEBUG
** and the blob is empty or otherwise contains suspect data?
*/
static void interiorReaderInit(const char *pData, int nData,
                               InteriorReader *pReader){
  int n, nTerm;

  /* Require at least the leading flag byte */
  assert( nData>0 );
  assert( pData[0]!='\0' );

  CLEAR(pReader);

  /* Decode the base blockid, and set the cursor to the first term. */
  n = fts3GetVarint(pData+1, &pReader->iBlockid);
  assert( 1+n<=nData );
  pReader->pData = pData+1+n;
  pReader->nData = nData-(1+n);

  /* A single-child interior node (such as when a leaf node was too
  ** large for the segment directory) won't have any terms.
  ** Otherwise, decode the first term.
  */
  if( pReader->nData==0 ){
    dataBufferInit(&pReader->term, 0);
  }else{
    n = fts3GetVarint32(pReader->pData, &nTerm);
    dataBufferInit(&pReader->term, nTerm);
    dataBufferReplace(&pReader->term, pReader->pData+n, nTerm);
    assert( n+nTerm<=pReader->nData );
    pReader->pData += n+nTerm;
    pReader->nData -= n+nTerm;
  }
}

static int interiorReaderAtEnd(InteriorReader *pReader){
  return pReader->term.nData==0;
}

static sqlite_int64 interiorReaderCurrentBlockid(InteriorReader *pReader){
  return pReader->iBlockid;
}

static int interiorReaderTermBytes(InteriorReader *pReader){
  assert( !interiorReaderAtEnd(pReader) );
  return pReader->term.nData;
}
static const char *interiorReaderTerm(InteriorReader *pReader){
  assert( !interiorReaderAtEnd(pReader) );
  return pReader->term.pData;
}

/* Step forward to the next term in the node. */
static void interiorReaderStep(InteriorReader *pReader){
  assert( !interiorReaderAtEnd(pReader) );

  /* If the last term has been read, signal eof, else construct the
  ** next term.
  */
  if( pReader->nData==0 ){
    dataBufferReset(&pReader->term);
  }else{
    int n, nPrefix, nSuffix;

    n = fts3GetVarint32(pReader->pData, &nPrefix);
    n += fts3GetVarint32(pReader->pData+n, &nSuffix);

    /* Truncate the current term and append suffix data. */
    pReader->term.nData = nPrefix;
    dataBufferAppend(&pReader->term, pReader->pData+n, nSuffix);

    assert( n+nSuffix<=pReader->nData );
    pReader->pData += n+nSuffix;
    pReader->nData -= n+nSuffix;
  }
  pReader->iBlockid++;
}

/* Compare the current term to pTerm[nTerm], returning strcmp-style
** results.  If isPrefix, equality means equal through nTerm bytes.
*/
static int interiorReaderTermCmp(InteriorReader *pReader,
                                 const char *pTerm, int nTerm, int isPrefix){
  const char *pReaderTerm = interiorReaderTerm(pReader);
  int nReaderTerm = interiorReaderTermBytes(pReader);
  int c, n = nReaderTerm<nTerm ? nReaderTerm : nTerm;

  if( n==0 ){
    if( nReaderTerm>0 ) return -1;
    if( nTerm>0 ) return 1;
    return 0;
  }

  c = memcmp(pReaderTerm, pTerm, n);
  if( c!=0 ) return c;
  if( isPrefix && n==nTerm ) return 0;
  return nReaderTerm - nTerm;
}

/****************************************************************/
/* LeafWriter is used to collect terms and associated doclist data
** into leaf blocks in %_segments (see top of file for format info).
** Expected usage is:
**
** LeafWriter writer;
** leafWriterInit(0, 0, &writer);
** while( sorted_terms_left_to_process ){
**   // data is doclist data for that term.
**   rc = leafWriterStep(v, &writer, pTerm, nTerm, pData, nData);
**   if( rc!=SQLITE_OK ) goto err;
** }
** rc = leafWriterFinalize(v, &writer);
**err:
** leafWriterDestroy(&writer);
** return rc;
**
** leafWriterStep() may write a collected leaf out to %_segments.
** leafWriterFinalize() finishes writing any buffered data and stores
** a root node in %_segdir.  leafWriterDestroy() frees all buffers and
** InteriorWriters allocated as part of writing this segment.
**
** TODO(shess) Document leafWriterStepMerge().
*/

/* Put terms with data this big in their own block. */
#define STANDALONE_MIN 1024

/* Keep leaf blocks below this size. */
#define LEAF_MAX 2048

typedef struct LeafWriter {
  int iLevel;
  int idx;
  sqlite_int64 iStartBlockid;     /* needed to create the root info */
  sqlite_int64 iEndBlockid;       /* when we're done writing. */

  DataBuffer term;                /* previous encoded term */
  DataBuffer data;                /* encoding buffer */

  /* bytes of first term in the current node which distinguishes that
  ** term from the last term of the previous node.
  */
  int nTermDistinct;

  InteriorWriter parentWriter;    /* if we overflow */
  int has_parent;
} LeafWriter;

static void leafWriterInit(int iLevel, int idx, LeafWriter *pWriter){
  CLEAR(pWriter);
  pWriter->iLevel = iLevel;
  pWriter->idx = idx;

  dataBufferInit(&pWriter->term, 32);

  /* Start out with a reasonably sized block, though it can grow. */
  dataBufferInit(&pWriter->data, LEAF_MAX);
}

#ifndef NDEBUG
/* Verify that the data is readable as a leaf node. */
static void leafNodeValidate(const char *pData, int nData){
  int n, iDummy;

  if( nData==0 ) return;
  assert( nData>0 );
  assert( pData!=0 );
  assert( pData+nData>pData );

  /* Must lead with a varint(0) */
  n = fts3GetVarint32(pData, &iDummy);
  assert( iDummy==0 );
  assert( n>0 );
  assert( n<nData );
  pData += n;
  nData -= n;

  /* Leading term length and data must fit in buffer. */
  n = fts3GetVarint32(pData, &iDummy);
  assert( n>0 );
  assert( iDummy>0 );
  assert( n+iDummy>0 );
  assert( n+iDummy<nData );
  pData += n+iDummy;
  nData -= n+iDummy;

  /* Leading term's doclist length and data must fit. */
  n = fts3GetVarint32(pData, &iDummy);
  assert( n>0 );
  assert( iDummy>0 );
  assert( n+iDummy>0 );
  assert( n+iDummy<=nData );
  ASSERT_VALID_DOCLIST(DL_DEFAULT, pData+n, iDummy, NULL);
  pData += n+iDummy;
  nData -= n+iDummy;

  /* Verify that trailing terms and doclists also are readable. */
  while( nData!=0 ){
    n = fts3GetVarint32(pData, &iDummy);
    assert( n>0 );
    assert( iDummy>=0 );
    assert( n<nData );
    pData += n;
    nData -= n;
    n = fts3GetVarint32(pData, &iDummy);
    assert( n>0 );
    assert( iDummy>0 );
    assert( n+iDummy>0 );
    assert( n+iDummy<nData );
    pData += n+iDummy;
    nData -= n+iDummy;

    n = fts3GetVarint32(pData, &iDummy);
    assert( n>0 );
    assert( iDummy>0 );
    assert( n+iDummy>0 );
    assert( n+iDummy<=nData );
    ASSERT_VALID_DOCLIST(DL_DEFAULT, pData+n, iDummy, NULL);
    pData += n+iDummy;
    nData -= n+iDummy;
  }
}
#define ASSERT_VALID_LEAF_NODE(p, n) leafNodeValidate(p, n)
#else
#define ASSERT_VALID_LEAF_NODE(p, n) assert( 1 )
#endif

/* Flush the current leaf node to %_segments, and adding the resulting
** blockid and the starting term to the interior node which will
** contain it.
*/
static int leafWriterInternalFlush(fulltext_vtab *v, LeafWriter *pWriter,
                                   int iData, int nData){
  sqlite_int64 iBlockid = 0;
  const char *pStartingTerm;
  int nStartingTerm, rc, n;

  /* Must have the leading varint(0) flag, plus at least some
  ** valid-looking data.
  */
  assert( nData>2 );
  assert( iData>=0 );
  assert( iData+nData<=pWriter->data.nData );
  ASSERT_VALID_LEAF_NODE(pWriter->data.pData+iData, nData);

  rc = block_insert(v, pWriter->data.pData+iData, nData, &iBlockid);
  if( rc!=SQLITE_OK ) return rc;
  assert( iBlockid!=0 );

  /* Reconstruct the first term in the leaf for purposes of building
  ** the interior node.
  */
  n = fts3GetVarint32(pWriter->data.pData+iData+1, &nStartingTerm);
  pStartingTerm = pWriter->data.pData+iData+1+n;
  assert( pWriter->data.nData>iData+1+n+nStartingTerm );
  assert( pWriter->nTermDistinct>0 );
  assert( pWriter->nTermDistinct<=nStartingTerm );
  nStartingTerm = pWriter->nTermDistinct;

  if( pWriter->has_parent ){
    interiorWriterAppend(&pWriter->parentWriter,
                         pStartingTerm, nStartingTerm, iBlockid);
  }else{
    interiorWriterInit(1, pStartingTerm, nStartingTerm, iBlockid,
                       &pWriter->parentWriter);
    pWriter->has_parent = 1;
  }

  /* Track the span of this segment's leaf nodes. */
  if( pWriter->iEndBlockid==0 ){
    pWriter->iEndBlockid = pWriter->iStartBlockid = iBlockid;
  }else{
    pWriter->iEndBlockid++;
    assert( iBlockid==pWriter->iEndBlockid );
  }

  return SQLITE_OK;
}
static int leafWriterFlush(fulltext_vtab *v, LeafWriter *pWriter){
  int rc = leafWriterInternalFlush(v, pWriter, 0, pWriter->data.nData);
  if( rc!=SQLITE_OK ) return rc;

  /* Re-initialize the output buffer. */
  dataBufferReset(&pWriter->data);

  return SQLITE_OK;
}

/* Fetch the root info for the segment.  If the entire leaf fits
** within ROOT_MAX, then it will be returned directly, otherwise it
** will be flushed and the root info will be returned from the
** interior node.  *piEndBlockid is set to the blockid of the last
** interior or leaf node written to disk (0 if none are written at
** all).
*/
static int leafWriterRootInfo(fulltext_vtab *v, LeafWriter *pWriter,
                              char **ppRootInfo, int *pnRootInfo,
                              sqlite_int64 *piEndBlockid){
  /* we can fit the segment entirely inline */
  if( !pWriter->has_parent && pWriter->data.nData<ROOT_MAX ){
    *ppRootInfo = pWriter->data.pData;
    *pnRootInfo = pWriter->data.nData;
    *piEndBlockid = 0;
    return SQLITE_OK;
  }

  /* Flush remaining leaf data. */
  if( pWriter->data.nData>0 ){
    int rc = leafWriterFlush(v, pWriter);
    if( rc!=SQLITE_OK ) return rc;
  }

  /* We must have flushed a leaf at some point. */
  assert( pWriter->has_parent );

  /* Tenatively set the end leaf blockid as the end blockid.  If the
  ** interior node can be returned inline, this will be the final
  ** blockid, otherwise it will be overwritten by
  ** interiorWriterRootInfo().
  */
  *piEndBlockid = pWriter->iEndBlockid;

  return interiorWriterRootInfo(v, &pWriter->parentWriter,
                                ppRootInfo, pnRootInfo, piEndBlockid);
}

/* Collect the rootInfo data and store it into the segment directory.
** This has the effect of flushing the segment's leaf data to
** %_segments, and also flushing any interior nodes to %_segments.
*/
static int leafWriterFinalize(fulltext_vtab *v, LeafWriter *pWriter){
  sqlite_int64 iEndBlockid;
  char *pRootInfo;
  int rc, nRootInfo;

  rc = leafWriterRootInfo(v, pWriter, &pRootInfo, &nRootInfo, &iEndBlockid);
  if( rc!=SQLITE_OK ) return rc;

  /* Don't bother storing an entirely empty segment. */
  if( iEndBlockid==0 && nRootInfo==0 ) return SQLITE_OK;

  return segdir_set(v, pWriter->iLevel, pWriter->idx,
                    pWriter->iStartBlockid, pWriter->iEndBlockid,
                    iEndBlockid, pRootInfo, nRootInfo);
}

static void leafWriterDestroy(LeafWriter *pWriter){
  if( pWriter->has_parent ) interiorWriterDestroy(&pWriter->parentWriter);
  dataBufferDestroy(&pWriter->term);
  dataBufferDestroy(&pWriter->data);
}

/* Encode a term into the leafWriter, delta-encoding as appropriate.
** Returns the length of the new term which distinguishes it from the
** previous term, which can be used to set nTermDistinct when a node
** boundary is crossed.
*/
static int leafWriterEncodeTerm(LeafWriter *pWriter,
                                const char *pTerm, int nTerm){
  char c[VARINT_MAX+VARINT_MAX];
  int n, nPrefix = 0;

  assert( nTerm>0 );
  while( nPrefix<pWriter->term.nData &&
         pTerm[nPrefix]==pWriter->term.pData[nPrefix] ){
    nPrefix++;
    /* Failing this implies that the terms weren't in order. */
    assert( nPrefix<nTerm );
  }

  if( pWriter->data.nData==0 ){
    /* Encode the node header and leading term as:
    **  varint(0)
    **  varint(nTerm)
    **  char pTerm[nTerm]
    */
    n = fts3PutVarint(c, '\0');
    n += fts3PutVarint(c+n, nTerm);
    dataBufferAppend2(&pWriter->data, c, n, pTerm, nTerm);
  }else{
    /* Delta-encode the term as:
    **  varint(nPrefix)
    **  varint(nSuffix)
    **  char pTermSuffix[nSuffix]
    */
    n = fts3PutVarint(c, nPrefix);
    n += fts3PutVarint(c+n, nTerm-nPrefix);
    dataBufferAppend2(&pWriter->data, c, n, pTerm+nPrefix, nTerm-nPrefix);
  }
  dataBufferReplace(&pWriter->term, pTerm, nTerm);

  return nPrefix+1;
}

/* Used to avoid a memmove when a large amount of doclist data is in
** the buffer.  This constructs a node and term header before
** iDoclistData and flushes the resulting complete node using
** leafWriterInternalFlush().
*/
static int leafWriterInlineFlush(fulltext_vtab *v, LeafWriter *pWriter,
                                 const char *pTerm, int nTerm,
                                 int iDoclistData){
  char c[VARINT_MAX+VARINT_MAX];
  int iData, n = fts3PutVarint(c, 0);
  n += fts3PutVarint(c+n, nTerm);

  /* There should always be room for the header.  Even if pTerm shared
  ** a substantial prefix with the previous term, the entire prefix
  ** could be constructed from earlier data in the doclist, so there
  ** should be room.
  */
  assert( iDoclistData>=n+nTerm );

  iData = iDoclistData-(n+nTerm);
  memcpy(pWriter->data.pData+iData, c, n);
  memcpy(pWriter->data.pData+iData+n, pTerm, nTerm);

  return leafWriterInternalFlush(v, pWriter, iData, pWriter->data.nData-iData);
}

/* Push pTerm[nTerm] along with the doclist data to the leaf layer of
** %_segments.
*/
static int leafWriterStepMerge(fulltext_vtab *v, LeafWriter *pWriter,
                               const char *pTerm, int nTerm,
                               DLReader *pReaders, int nReaders){
  char c[VARINT_MAX+VARINT_MAX];
  int iTermData = pWriter->data.nData, iDoclistData;
  int i, nData, n, nActualData, nActual, rc, nTermDistinct;

  ASSERT_VALID_LEAF_NODE(pWriter->data.pData, pWriter->data.nData);
  nTermDistinct = leafWriterEncodeTerm(pWriter, pTerm, nTerm);

  /* Remember nTermDistinct if opening a new node. */
  if( iTermData==0 ) pWriter->nTermDistinct = nTermDistinct;

  iDoclistData = pWriter->data.nData;

  /* Estimate the length of the merged doclist so we can leave space
  ** to encode it.
  */
  for(i=0, nData=0; i<nReaders; i++){
    nData += dlrAllDataBytes(&pReaders[i]);
  }
  n = fts3PutVarint(c, nData);
  dataBufferAppend(&pWriter->data, c, n);

  docListMerge(&pWriter->data, pReaders, nReaders);
  ASSERT_VALID_DOCLIST(DL_DEFAULT,
                       pWriter->data.pData+iDoclistData+n,
                       pWriter->data.nData-iDoclistData-n, NULL);

  /* The actual amount of doclist data at this point could be smaller
  ** than the length we encoded.  Additionally, the space required to
  ** encode this length could be smaller.  For small doclists, this is
  ** not a big deal, we can just use memmove() to adjust things.
  */
  nActualData = pWriter->data.nData-(iDoclistData+n);
  nActual = fts3PutVarint(c, nActualData);
  assert( nActualData<=nData );
  assert( nActual<=n );

  /* If the new doclist is big enough for force a standalone leaf
  ** node, we can immediately flush it inline without doing the
  ** memmove().
  */
  /* TODO(shess) This test matches leafWriterStep(), which does this
  ** test before it knows the cost to varint-encode the term and
  ** doclist lengths.  At some point, change to
  ** pWriter->data.nData-iTermData>STANDALONE_MIN.
  */
  if( nTerm+nActualData>STANDALONE_MIN ){
    /* Push leaf node from before this term. */
    if( iTermData>0 ){
      rc = leafWriterInternalFlush(v, pWriter, 0, iTermData);
      if( rc!=SQLITE_OK ) return rc;

      pWriter->nTermDistinct = nTermDistinct;
    }

    /* Fix the encoded doclist length. */
    iDoclistData += n - nActual;
    memcpy(pWriter->data.pData+iDoclistData, c, nActual);

    /* Push the standalone leaf node. */
    rc = leafWriterInlineFlush(v, pWriter, pTerm, nTerm, iDoclistData);
    if( rc!=SQLITE_OK ) return rc;

    /* Leave the node empty. */
    dataBufferReset(&pWriter->data);

    return rc;
  }

  /* At this point, we know that the doclist was small, so do the
  ** memmove if indicated.
  */
  if( nActual<n ){
    memmove(pWriter->data.pData+iDoclistData+nActual,
            pWriter->data.pData+iDoclistData+n,
            pWriter->data.nData-(iDoclistData+n));
    pWriter->data.nData -= n-nActual;
  }

  /* Replace written length with actual length. */
  memcpy(pWriter->data.pData+iDoclistData, c, nActual);

  /* If the node is too large, break things up. */
  /* TODO(shess) This test matches leafWriterStep(), which does this
  ** test before it knows the cost to varint-encode the term and
  ** doclist lengths.  At some point, change to
  ** pWriter->data.nData>LEAF_MAX.
  */
  if( iTermData+nTerm+nActualData>LEAF_MAX ){
    /* Flush out the leading data as a node */
    rc = leafWriterInternalFlush(v, pWriter, 0, iTermData);
    if( rc!=SQLITE_OK ) return rc;

    pWriter->nTermDistinct = nTermDistinct;

    /* Rebuild header using the current term */
    n = fts3PutVarint(pWriter->data.pData, 0);
    n += fts3PutVarint(pWriter->data.pData+n, nTerm);
    memcpy(pWriter->data.pData+n, pTerm, nTerm);
    n += nTerm;

    /* There should always be room, because the previous encoding
    ** included all data necessary to construct the term.
    */
    assert( n<iDoclistData );
    /* So long as STANDALONE_MIN is half or less of LEAF_MAX, the
    ** following memcpy() is safe (as opposed to needing a memmove).
    */
    assert( 2*STANDALONE_MIN<=LEAF_MAX );
    assert( n+pWriter->data.nData-iDoclistData<iDoclistData );
    memcpy(pWriter->data.pData+n,
           pWriter->data.pData+iDoclistData,
           pWriter->data.nData-iDoclistData);
    pWriter->data.nData -= iDoclistData-n;
  }
  ASSERT_VALID_LEAF_NODE(pWriter->data.pData, pWriter->data.nData);

  return SQLITE_OK;
}

/* Push pTerm[nTerm] along with the doclist data to the leaf layer of
** %_segments.
*/
/* TODO(shess) Revise writeZeroSegment() so that doclists are
** constructed directly in pWriter->data.
*/
static int leafWriterStep(fulltext_vtab *v, LeafWriter *pWriter,
                          const char *pTerm, int nTerm,
                          const char *pData, int nData){
  int rc;
  DLReader reader;

  dlrInit(&reader, DL_DEFAULT, pData, nData);
  rc = leafWriterStepMerge(v, pWriter, pTerm, nTerm, &reader, 1);
  dlrDestroy(&reader);

  return rc;
}


/****************************************************************/
/* LeafReader is used to iterate over an individual leaf node. */
typedef struct LeafReader {
  DataBuffer term;          /* copy of current term. */

  const char *pData;        /* data for current term. */
  int nData;
} LeafReader;

static void leafReaderDestroy(LeafReader *pReader){
  dataBufferDestroy(&pReader->term);
  SCRAMBLE(pReader);
}

static int leafReaderAtEnd(LeafReader *pReader){
  return pReader->nData<=0;
}

/* Access the current term. */
static int leafReaderTermBytes(LeafReader *pReader){
  return pReader->term.nData;
}
static const char *leafReaderTerm(LeafReader *pReader){
  assert( pReader->term.nData>0 );
  return pReader->term.pData;
}

/* Access the doclist data for the current term. */
static int leafReaderDataBytes(LeafReader *pReader){
  int nData;
  assert( pReader->term.nData>0 );
  fts3GetVarint32(pReader->pData, &nData);
  return nData;
}
static const char *leafReaderData(LeafReader *pReader){
  int n, nData;
  assert( pReader->term.nData>0 );
  n = fts3GetVarint32(pReader->pData, &nData);
  return pReader->pData+n;
}

static void leafReaderInit(const char *pData, int nData,
                           LeafReader *pReader){
  int nTerm, n;

  assert( nData>0 );
  assert( pData[0]=='\0' );

  CLEAR(pReader);

  /* Read the first term, skipping the header byte. */
  n = fts3GetVarint32(pData+1, &nTerm);
  dataBufferInit(&pReader->term, nTerm);
  dataBufferReplace(&pReader->term, pData+1+n, nTerm);

  /* Position after the first term. */
  assert( 1+n+nTerm<nData );
  pReader->pData = pData+1+n+nTerm;
  pReader->nData = nData-1-n-nTerm;
}

/* Step the reader forward to the next term. */
static void leafReaderStep(LeafReader *pReader){
  int n, nData, nPrefix, nSuffix;
  assert( !leafReaderAtEnd(pReader) );

  /* Skip previous entry's data block. */
  n = fts3GetVarint32(pReader->pData, &nData);
  assert( n+nData<=pReader->nData );
  pReader->pData += n+nData;
  pReader->nData -= n+nData;

  if( !leafReaderAtEnd(pReader) ){
    /* Construct the new term using a prefix from the old term plus a
    ** suffix from the leaf data.
    */
    n = fts3GetVarint32(pReader->pData, &nPrefix);
    n += fts3GetVarint32(pReader->pData+n, &nSuffix);
    assert( n+nSuffix<pReader->nData );
    pReader->term.nData = nPrefix;
    dataBufferAppend(&pReader->term, pReader->pData+n, nSuffix);

    pReader->pData += n+nSuffix;
    pReader->nData -= n+nSuffix;
  }
}

/* strcmp-style comparison of pReader's current term against pTerm.
** If isPrefix, equality means equal through nTerm bytes.
*/
static int leafReaderTermCmp(LeafReader *pReader,
                             const char *pTerm, int nTerm, int isPrefix){
  int c, n = pReader->term.nData<nTerm ? pReader->term.nData : nTerm;
  if( n==0 ){
    if( pReader->term.nData>0 ) return -1;
    if(nTerm>0 ) return 1;
    return 0;
  }

  c = memcmp(pReader->term.pData, pTerm, n);
  if( c!=0 ) return c;
  if( isPrefix && n==nTerm ) return 0;
  return pReader->term.nData - nTerm;
}


/****************************************************************/
/* LeavesReader wraps LeafReader to allow iterating over the entire
** leaf layer of the tree.
*/
typedef struct LeavesReader {
  int idx;                  /* Index within the segment. */

  sqlite3_stmt *pStmt;      /* Statement we're streaming leaves from. */
  int eof;                  /* we've seen SQLITE_DONE from pStmt. */

  LeafReader leafReader;    /* reader for the current leaf. */
  DataBuffer rootData;      /* root data for inline. */
} LeavesReader;

/* Access the current term. */
static int leavesReaderTermBytes(LeavesReader *pReader){
  assert( !pReader->eof );
  return leafReaderTermBytes(&pReader->leafReader);
}
static const char *leavesReaderTerm(LeavesReader *pReader){
  assert( !pReader->eof );
  return leafReaderTerm(&pReader->leafReader);
}

/* Access the doclist data for the current term. */
static int leavesReaderDataBytes(LeavesReader *pReader){
  assert( !pReader->eof );
  return leafReaderDataBytes(&pReader->leafReader);
}
static const char *leavesReaderData(LeavesReader *pReader){
  assert( !pReader->eof );
  return leafReaderData(&pReader->leafReader);
}

static int leavesReaderAtEnd(LeavesReader *pReader){
  return pReader->eof;
}

/* loadSegmentLeaves() may not read all the way to SQLITE_DONE, thus
** leaving the statement handle open, which locks the table.
*/
/* TODO(shess) This "solution" is not satisfactory.  Really, there
** should be check-in function for all statement handles which
** arranges to call sqlite3_reset().  This most likely will require
** modification to control flow all over the place, though, so for now
** just punt.
**
** Note the the current system assumes that segment merges will run to
** completion, which is why this particular probably hasn't arisen in
** this case.  Probably a brittle assumption.
*/
static int leavesReaderReset(LeavesReader *pReader){
  return sqlite3_reset(pReader->pStmt);
}

static void leavesReaderDestroy(LeavesReader *pReader){
  /* If idx is -1, that means we're using a non-cached statement
  ** handle in the optimize() case, so we need to release it.
  */
  if( pReader->pStmt!=NULL && pReader->idx==-1 ){
    sqlite3_finalize(pReader->pStmt);
  }
  leafReaderDestroy(&pReader->leafReader);
  dataBufferDestroy(&pReader->rootData);
  SCRAMBLE(pReader);
}

/* Initialize pReader with the given root data (if iStartBlockid==0
** the leaf data was entirely contained in the root), or from the
** stream of blocks between iStartBlockid and iEndBlockid, inclusive.
*/
static int leavesReaderInit(fulltext_vtab *v,
                            int idx,
                            sqlite_int64 iStartBlockid,
                            sqlite_int64 iEndBlockid,
                            const char *pRootData, int nRootData,
                            LeavesReader *pReader){
  CLEAR(pReader);
  pReader->idx = idx;

  dataBufferInit(&pReader->rootData, 0);
  if( iStartBlockid==0 ){
    /* Entire leaf level fit in root data. */
    dataBufferReplace(&pReader->rootData, pRootData, nRootData);
    leafReaderInit(pReader->rootData.pData, pReader->rootData.nData,
                   &pReader->leafReader);
  }else{
    sqlite3_stmt *s;
    int rc = sql_get_leaf_statement(v, idx, &s);
    if( rc!=SQLITE_OK ) return rc;

    rc = sqlite3_bind_int64(s, 1, iStartBlockid);
    if( rc!=SQLITE_OK ) return rc;

    rc = sqlite3_bind_int64(s, 2, iEndBlockid);
    if( rc!=SQLITE_OK ) return rc;

    rc = sqlite3_step(s);
    if( rc==SQLITE_DONE ){
      pReader->eof = 1;
      return SQLITE_OK;
    }
    if( rc!=SQLITE_ROW ) return rc;

    pReader->pStmt = s;
    leafReaderInit(sqlite3_column_blob(pReader->pStmt, 0),
                   sqlite3_column_bytes(pReader->pStmt, 0),
                   &pReader->leafReader);
  }
  return SQLITE_OK;
}

/* Step the current leaf forward to the next term.  If we reach the
** end of the current leaf, step forward to the next leaf block.
*/
static int leavesReaderStep(fulltext_vtab *v, LeavesReader *pReader){
  assert( !leavesReaderAtEnd(pReader) );
  leafReaderStep(&pReader->leafReader);

  if( leafReaderAtEnd(&pReader->leafReader) ){
    int rc;
    if( pReader->rootData.pData ){
      pReader->eof = 1;
      return SQLITE_OK;
    }
    rc = sqlite3_step(pReader->pStmt);
    if( rc!=SQLITE_ROW ){
      pReader->eof = 1;
      return rc==SQLITE_DONE ? SQLITE_OK : rc;
    }
    leafReaderDestroy(&pReader->leafReader);
    leafReaderInit(sqlite3_column_blob(pReader->pStmt, 0),
                   sqlite3_column_bytes(pReader->pStmt, 0),
                   &pReader->leafReader);
  }
  return SQLITE_OK;
}

/* Order LeavesReaders by their term, ignoring idx.  Readers at eof
** always sort to the end.
*/
static int leavesReaderTermCmp(LeavesReader *lr1, LeavesReader *lr2){
  if( leavesReaderAtEnd(lr1) ){
    if( leavesReaderAtEnd(lr2) ) return 0;
    return 1;
  }
  if( leavesReaderAtEnd(lr2) ) return -1;

  return leafReaderTermCmp(&lr1->leafReader,
                           leavesReaderTerm(lr2), leavesReaderTermBytes(lr2),
                           0);
}

/* Similar to leavesReaderTermCmp(), with additional ordering by idx
** so that older segments sort before newer segments.
*/
static int leavesReaderCmp(LeavesReader *lr1, LeavesReader *lr2){
  int c = leavesReaderTermCmp(lr1, lr2);
  if( c!=0 ) return c;
  return lr1->idx-lr2->idx;
}

/* Assume that pLr[1]..pLr[nLr] are sorted.  Bubble pLr[0] into its
** sorted position.
*/
static void leavesReaderReorder(LeavesReader *pLr, int nLr){
  while( nLr>1 && leavesReaderCmp(pLr, pLr+1)>0 ){
    LeavesReader tmp = pLr[0];
    pLr[0] = pLr[1];
    pLr[1] = tmp;
    nLr--;
    pLr++;
  }
}

/* Initializes pReaders with the segments from level iLevel, returning
** the number of segments in *piReaders.  Leaves pReaders in sorted
** order.
*/
static int leavesReadersInit(fulltext_vtab *v, int iLevel,
                             LeavesReader *pReaders, int *piReaders){
  sqlite3_stmt *s;
  int i, rc = sql_get_statement(v, SEGDIR_SELECT_LEVEL_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int(s, 1, iLevel);
  if( rc!=SQLITE_OK ) return rc;

  i = 0;
  while( (rc = sqlite3_step(s))==SQLITE_ROW ){
    sqlite_int64 iStart = sqlite3_column_int64(s, 0);
    sqlite_int64 iEnd = sqlite3_column_int64(s, 1);
    const char *pRootData = sqlite3_column_blob(s, 2);
    int nRootData = sqlite3_column_bytes(s, 2);

    assert( i<MERGE_COUNT );
    rc = leavesReaderInit(v, i, iStart, iEnd, pRootData, nRootData,
                          &pReaders[i]);
    if( rc!=SQLITE_OK ) break;

    i++;
  }
  if( rc!=SQLITE_DONE ){
    while( i-->0 ){
      leavesReaderDestroy(&pReaders[i]);
    }
    return rc;
  }

  *piReaders = i;

  /* Leave our results sorted by term, then age. */
  while( i-- ){
    leavesReaderReorder(pReaders+i, *piReaders-i);
  }
  return SQLITE_OK;
}

/* Merge doclists from pReaders[nReaders] into a single doclist, which
** is written to pWriter.  Assumes pReaders is ordered oldest to
** newest.
*/
/* TODO(shess) Consider putting this inline in segmentMerge(). */
static int leavesReadersMerge(fulltext_vtab *v,
                              LeavesReader *pReaders, int nReaders,
                              LeafWriter *pWriter){
  DLReader dlReaders[MERGE_COUNT];
  const char *pTerm = leavesReaderTerm(pReaders);
  int i, nTerm = leavesReaderTermBytes(pReaders);

  assert( nReaders<=MERGE_COUNT );

  for(i=0; i<nReaders; i++){
    dlrInit(&dlReaders[i], DL_DEFAULT,
            leavesReaderData(pReaders+i),
            leavesReaderDataBytes(pReaders+i));
  }

  return leafWriterStepMerge(v, pWriter, pTerm, nTerm, dlReaders, nReaders);
}

/* Forward ref due to mutual recursion with segdirNextIndex(). */
static int segmentMerge(fulltext_vtab *v, int iLevel);

/* Put the next available index at iLevel into *pidx.  If iLevel
** already has MERGE_COUNT segments, they are merged to a higher
** level to make room.
*/
static int segdirNextIndex(fulltext_vtab *v, int iLevel, int *pidx){
  int rc = segdir_max_index(v, iLevel, pidx);
  if( rc==SQLITE_DONE ){              /* No segments at iLevel. */
    *pidx = 0;
  }else if( rc==SQLITE_ROW ){
    if( *pidx==(MERGE_COUNT-1) ){
      rc = segmentMerge(v, iLevel);
      if( rc!=SQLITE_OK ) return rc;
      *pidx = 0;
    }else{
      (*pidx)++;
    }
  }else{
    return rc;
  }
  return SQLITE_OK;
}

/* Merge MERGE_COUNT segments at iLevel into a new segment at
** iLevel+1.  If iLevel+1 is already full of segments, those will be
** merged to make room.
*/
static int segmentMerge(fulltext_vtab *v, int iLevel){
  LeafWriter writer;
  LeavesReader lrs[MERGE_COUNT];
  int i, rc, idx = 0;

  /* Determine the next available segment index at the next level,
  ** merging as necessary.
  */
  rc = segdirNextIndex(v, iLevel+1, &idx);
  if( rc!=SQLITE_OK ) return rc;

  /* TODO(shess) This assumes that we'll always see exactly
  ** MERGE_COUNT segments to merge at a given level.  That will be
  ** broken if we allow the developer to request preemptive or
  ** deferred merging.
  */
  memset(&lrs, '\0', sizeof(lrs));
  rc = leavesReadersInit(v, iLevel, lrs, &i);
  if( rc!=SQLITE_OK ) return rc;
  assert( i==MERGE_COUNT );

  leafWriterInit(iLevel+1, idx, &writer);

  /* Since leavesReaderReorder() pushes readers at eof to the end,
  ** when the first reader is empty, all will be empty.
  */
  while( !leavesReaderAtEnd(lrs) ){
    /* Figure out how many readers share their next term. */
    for(i=1; i<MERGE_COUNT && !leavesReaderAtEnd(lrs+i); i++){
      if( 0!=leavesReaderTermCmp(lrs, lrs+i) ) break;
    }

    rc = leavesReadersMerge(v, lrs, i, &writer);
    if( rc!=SQLITE_OK ) goto err;

    /* Step forward those that were merged. */
    while( i-->0 ){
      rc = leavesReaderStep(v, lrs+i);
      if( rc!=SQLITE_OK ) goto err;

      /* Reorder by term, then by age. */
      leavesReaderReorder(lrs+i, MERGE_COUNT-i);
    }
  }

  for(i=0; i<MERGE_COUNT; i++){
    leavesReaderDestroy(&lrs[i]);
  }

  rc = leafWriterFinalize(v, &writer);
  leafWriterDestroy(&writer);
  if( rc!=SQLITE_OK ) return rc;

  /* Delete the merged segment data. */
  return segdir_delete(v, iLevel);

 err:
  for(i=0; i<MERGE_COUNT; i++){
    leavesReaderDestroy(&lrs[i]);
  }
  leafWriterDestroy(&writer);
  return rc;
}

/* Accumulate the union of *acc and *pData into *acc. */
static void docListAccumulateUnion(DataBuffer *acc,
                                   const char *pData, int nData) {
  DataBuffer tmp = *acc;
  dataBufferInit(acc, tmp.nData+nData);
  docListUnion(tmp.pData, tmp.nData, pData, nData, acc);
  dataBufferDestroy(&tmp);
}

/* TODO(shess) It might be interesting to explore different merge
** strategies, here.  For instance, since this is a sorted merge, we
** could easily merge many doclists in parallel.  With some
** comprehension of the storage format, we could merge all of the
** doclists within a leaf node directly from the leaf node's storage.
** It may be worthwhile to merge smaller doclists before larger
** doclists, since they can be traversed more quickly - but the
** results may have less overlap, making them more expensive in a
** different way.
*/

/* Scan pReader for pTerm/nTerm, and merge the term's doclist over
** *out (any doclists with duplicate docids overwrite those in *out).
** Internal function for loadSegmentLeaf().
*/
static int loadSegmentLeavesInt(fulltext_vtab *v, LeavesReader *pReader,
                                const char *pTerm, int nTerm, int isPrefix,
                                DataBuffer *out){
  /* doclist data is accumulated into pBuffers similar to how one does
  ** increment in binary arithmetic.  If index 0 is empty, the data is
  ** stored there.  If there is data there, it is merged and the
  ** results carried into position 1, with further merge-and-carry
  ** until an empty position is found.
  */
  DataBuffer *pBuffers = NULL;
  int nBuffers = 0, nMaxBuffers = 0, rc;

  assert( nTerm>0 );

  for(rc=SQLITE_OK; rc==SQLITE_OK && !leavesReaderAtEnd(pReader);
      rc=leavesReaderStep(v, pReader)){
    /* TODO(shess) Really want leavesReaderTermCmp(), but that name is
    ** already taken to compare the terms of two LeavesReaders.  Think
    ** on a better name.  [Meanwhile, break encapsulation rather than
    ** use a confusing name.]
    */
    int c = leafReaderTermCmp(&pReader->leafReader, pTerm, nTerm, isPrefix);
    if( c>0 ) break;      /* Past any possible matches. */
    if( c==0 ){
      const char *pData = leavesReaderData(pReader);
      int iBuffer, nData = leavesReaderDataBytes(pReader);

      /* Find the first empty buffer. */
      for(iBuffer=0; iBuffer<nBuffers; ++iBuffer){
        if( 0==pBuffers[iBuffer].nData ) break;
      }

      /* Out of buffers, add an empty one. */
      if( iBuffer==nBuffers ){
        if( nBuffers==nMaxBuffers ){
          DataBuffer *p;
          nMaxBuffers += 20;

          /* Manual realloc so we can handle NULL appropriately. */
          p = sqlite3_malloc(nMaxBuffers*sizeof(*pBuffers));
          if( p==NULL ){
            rc = SQLITE_NOMEM;
            break;
          }

          if( nBuffers>0 ){
            assert(pBuffers!=NULL);
            memcpy(p, pBuffers, nBuffers*sizeof(*pBuffers));
            sqlite3_free(pBuffers);
          }
          pBuffers = p;
        }
        dataBufferInit(&(pBuffers[nBuffers]), 0);
        nBuffers++;
      }

      /* At this point, must have an empty at iBuffer. */
      assert(iBuffer<nBuffers && pBuffers[iBuffer].nData==0);

      /* If empty was first buffer, no need for merge logic. */
      if( iBuffer==0 ){
        dataBufferReplace(&(pBuffers[0]), pData, nData);
      }else{
        /* pAcc is the empty buffer the merged data will end up in. */
        DataBuffer *pAcc = &(pBuffers[iBuffer]);
        DataBuffer *p = &(pBuffers[0]);

        /* Handle position 0 specially to avoid need to prime pAcc
        ** with pData/nData.
        */
        dataBufferSwap(p, pAcc);
        docListAccumulateUnion(pAcc, pData, nData);

        /* Accumulate remaining doclists into pAcc. */
        for(++p; p<pAcc; ++p){
          docListAccumulateUnion(pAcc, p->pData, p->nData);

          /* dataBufferReset() could allow a large doclist to blow up
          ** our memory requirements.
          */
          if( p->nCapacity<1024 ){
            dataBufferReset(p);
          }else{
            dataBufferDestroy(p);
            dataBufferInit(p, 0);
          }
        }
      }
    }
  }

  /* Union all the doclists together into *out. */
  /* TODO(shess) What if *out is big?  Sigh. */
  if( rc==SQLITE_OK && nBuffers>0 ){
    int iBuffer;
    for(iBuffer=0; iBuffer<nBuffers; ++iBuffer){
      if( pBuffers[iBuffer].nData>0 ){
        if( out->nData==0 ){
          dataBufferSwap(out, &(pBuffers[iBuffer]));
        }else{
          docListAccumulateUnion(out, pBuffers[iBuffer].pData,
                                 pBuffers[iBuffer].nData);
        }
      }
    }
  }

  while( nBuffers-- ){
    dataBufferDestroy(&(pBuffers[nBuffers]));
  }
  if( pBuffers!=NULL ) sqlite3_free(pBuffers);

  return rc;
}

/* Call loadSegmentLeavesInt() with pData/nData as input. */
static int loadSegmentLeaf(fulltext_vtab *v, const char *pData, int nData,
                           const char *pTerm, int nTerm, int isPrefix,
                           DataBuffer *out){
  LeavesReader reader;
  int rc;

  assert( nData>1 );
  assert( *pData=='\0' );
  rc = leavesReaderInit(v, 0, 0, 0, pData, nData, &reader);
  if( rc!=SQLITE_OK ) return rc;

  rc = loadSegmentLeavesInt(v, &reader, pTerm, nTerm, isPrefix, out);
  leavesReaderReset(&reader);
  leavesReaderDestroy(&reader);
  return rc;
}

/* Call loadSegmentLeavesInt() with the leaf nodes from iStartLeaf to
** iEndLeaf (inclusive) as input, and merge the resulting doclist into
** out.
*/
static int loadSegmentLeaves(fulltext_vtab *v,
                             sqlite_int64 iStartLeaf, sqlite_int64 iEndLeaf,
                             const char *pTerm, int nTerm, int isPrefix,
                             DataBuffer *out){
  int rc;
  LeavesReader reader;

  assert( iStartLeaf<=iEndLeaf );
  rc = leavesReaderInit(v, 0, iStartLeaf, iEndLeaf, NULL, 0, &reader);
  if( rc!=SQLITE_OK ) return rc;

  rc = loadSegmentLeavesInt(v, &reader, pTerm, nTerm, isPrefix, out);
  leavesReaderReset(&reader);
  leavesReaderDestroy(&reader);
  return rc;
}

/* Taking pData/nData as an interior node, find the sequence of child
** nodes which could include pTerm/nTerm/isPrefix.  Note that the
** interior node terms logically come between the blocks, so there is
** one more blockid than there are terms (that block contains terms >=
** the last interior-node term).
*/
/* TODO(shess) The calling code may already know that the end child is
** not worth calculating, because the end may be in a later sibling
** node.  Consider whether breaking symmetry is worthwhile.  I suspect
** it is not worthwhile.
*/
static void getChildrenContaining(const char *pData, int nData,
                                  const char *pTerm, int nTerm, int isPrefix,
                                  sqlite_int64 *piStartChild,
                                  sqlite_int64 *piEndChild){
  InteriorReader reader;

  assert( nData>1 );
  assert( *pData!='\0' );
  interiorReaderInit(pData, nData, &reader);

  /* Scan for the first child which could contain pTerm/nTerm. */
  while( !interiorReaderAtEnd(&reader) ){
    if( interiorReaderTermCmp(&reader, pTerm, nTerm, 0)>0 ) break;
    interiorReaderStep(&reader);
  }
  *piStartChild = interiorReaderCurrentBlockid(&reader);

  /* Keep scanning to find a term greater than our term, using prefix
  ** comparison if indicated.  If isPrefix is false, this will be the
  ** same blockid as the starting block.
  */
  while( !interiorReaderAtEnd(&reader) ){
    if( interiorReaderTermCmp(&reader, pTerm, nTerm, isPrefix)>0 ) break;
    interiorReaderStep(&reader);
  }
  *piEndChild = interiorReaderCurrentBlockid(&reader);

  interiorReaderDestroy(&reader);

  /* Children must ascend, and if !prefix, both must be the same. */
  assert( *piEndChild>=*piStartChild );
  assert( isPrefix || *piStartChild==*piEndChild );
}

/* Read block at iBlockid and pass it with other params to
** getChildrenContaining().
*/
static int loadAndGetChildrenContaining(
  fulltext_vtab *v,
  sqlite_int64 iBlockid,
  const char *pTerm, int nTerm, int isPrefix,
  sqlite_int64 *piStartChild, sqlite_int64 *piEndChild
){
  sqlite3_stmt *s = NULL;
  int rc;

  assert( iBlockid!=0 );
  assert( pTerm!=NULL );
  assert( nTerm!=0 );        /* TODO(shess) Why not allow this? */
  assert( piStartChild!=NULL );
  assert( piEndChild!=NULL );

  rc = sql_get_statement(v, BLOCK_SELECT_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_int64(s, 1, iBlockid);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_step(s);
  if( rc==SQLITE_DONE ) return SQLITE_ERROR;
  if( rc!=SQLITE_ROW ) return rc;

  getChildrenContaining(sqlite3_column_blob(s, 0), sqlite3_column_bytes(s, 0),
                        pTerm, nTerm, isPrefix, piStartChild, piEndChild);

  /* We expect only one row.  We must execute another sqlite3_step()
   * to complete the iteration; otherwise the table will remain
   * locked. */
  rc = sqlite3_step(s);
  if( rc==SQLITE_ROW ) return SQLITE_ERROR;
  if( rc!=SQLITE_DONE ) return rc;

  return SQLITE_OK;
}

/* Traverse the tree represented by pData[nData] looking for
** pTerm[nTerm], placing its doclist into *out.  This is internal to
** loadSegment() to make error-handling cleaner.
*/
static int loadSegmentInt(fulltext_vtab *v, const char *pData, int nData,
                          sqlite_int64 iLeavesEnd,
                          const char *pTerm, int nTerm, int isPrefix,
                          DataBuffer *out){
  /* Special case where root is a leaf. */
  if( *pData=='\0' ){
    return loadSegmentLeaf(v, pData, nData, pTerm, nTerm, isPrefix, out);
  }else{
    int rc;
    sqlite_int64 iStartChild, iEndChild;

    /* Process pData as an interior node, then loop down the tree
    ** until we find the set of leaf nodes to scan for the term.
    */
    getChildrenContaining(pData, nData, pTerm, nTerm, isPrefix,
                          &iStartChild, &iEndChild);
    while( iStartChild>iLeavesEnd ){
      sqlite_int64 iNextStart, iNextEnd;
      rc = loadAndGetChildrenContaining(v, iStartChild, pTerm, nTerm, isPrefix,
                                        &iNextStart, &iNextEnd);
      if( rc!=SQLITE_OK ) return rc;

      /* If we've branched, follow the end branch, too. */
      if( iStartChild!=iEndChild ){
        sqlite_int64 iDummy;
        rc = loadAndGetChildrenContaining(v, iEndChild, pTerm, nTerm, isPrefix,
                                          &iDummy, &iNextEnd);
        if( rc!=SQLITE_OK ) return rc;
      }

      assert( iNextStart<=iNextEnd );
      iStartChild = iNextStart;
      iEndChild = iNextEnd;
    }
    assert( iStartChild<=iLeavesEnd );
    assert( iEndChild<=iLeavesEnd );

    /* Scan through the leaf segments for doclists. */
    return loadSegmentLeaves(v, iStartChild, iEndChild,
                             pTerm, nTerm, isPrefix, out);
  }
}

/* Call loadSegmentInt() to collect the doclist for pTerm/nTerm, then
** merge its doclist over *out (any duplicate doclists read from the
** segment rooted at pData will overwrite those in *out).
*/
/* TODO(shess) Consider changing this to determine the depth of the
** leaves using either the first characters of interior nodes (when
** ==1, we're one level above the leaves), or the first character of
** the root (which will describe the height of the tree directly).
** Either feels somewhat tricky to me.
*/
/* TODO(shess) The current merge is likely to be slow for large
** doclists (though it should process from newest/smallest to
** oldest/largest, so it may not be that bad).  It might be useful to
** modify things to allow for N-way merging.  This could either be
** within a segment, with pairwise merges across segments, or across
** all segments at once.
*/
static int loadSegment(fulltext_vtab *v, const char *pData, int nData,
                       sqlite_int64 iLeavesEnd,
                       const char *pTerm, int nTerm, int isPrefix,
                       DataBuffer *out){
  DataBuffer result;
  int rc;

  assert( nData>1 );

  /* This code should never be called with buffered updates. */
  assert( v->nPendingData<0 );

  dataBufferInit(&result, 0);
  rc = loadSegmentInt(v, pData, nData, iLeavesEnd,
                      pTerm, nTerm, isPrefix, &result);
  if( rc==SQLITE_OK && result.nData>0 ){
    if( out->nData==0 ){
      DataBuffer tmp = *out;
      *out = result;
      result = tmp;
    }else{
      DataBuffer merged;
      DLReader readers[2];

      dlrInit(&readers[0], DL_DEFAULT, out->pData, out->nData);
      dlrInit(&readers[1], DL_DEFAULT, result.pData, result.nData);
      dataBufferInit(&merged, out->nData+result.nData);
      docListMerge(&merged, readers, 2);
      dataBufferDestroy(out);
      *out = merged;
      dlrDestroy(&readers[0]);
      dlrDestroy(&readers[1]);
    }
  }
  dataBufferDestroy(&result);
  return rc;
}

/* Scan the database and merge together the posting lists for the term
** into *out.
*/
static int termSelect(
  fulltext_vtab *v, 
  int iColumn,
  const char *pTerm, int nTerm,             /* Term to query for */
  int isPrefix,                             /* True for a prefix search */
  DocListType iType, 
  DataBuffer *out                           /* Write results here */
){
  DataBuffer doclist;
  sqlite3_stmt *s;
  int rc = sql_get_statement(v, SEGDIR_SELECT_ALL_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  /* This code should never be called with buffered updates. */
  assert( v->nPendingData<0 );

  dataBufferInit(&doclist, 0);
  dataBufferInit(out, 0);

  /* Traverse the segments from oldest to newest so that newer doclist
  ** elements for given docids overwrite older elements.
  */
  while( (rc = sqlite3_step(s))==SQLITE_ROW ){
    const char *pData = sqlite3_column_blob(s, 2);
    const int nData = sqlite3_column_bytes(s, 2);
    const sqlite_int64 iLeavesEnd = sqlite3_column_int64(s, 1);
    rc = loadSegment(v, pData, nData, iLeavesEnd, pTerm, nTerm, isPrefix,
                     &doclist);
    if( rc!=SQLITE_OK ) goto err;
  }
  if( rc==SQLITE_DONE ){
    if( doclist.nData!=0 ){
      /* TODO(shess) The old term_select_all() code applied the column
      ** restrict as we merged segments, leading to smaller buffers.
      ** This is probably worthwhile to bring back, once the new storage
      ** system is checked in.
      */
      if( iColumn==v->nColumn) iColumn = -1;
      docListTrim(DL_DEFAULT, doclist.pData, doclist.nData,
                  iColumn, iType, out);
    }
    rc = SQLITE_OK;
  }

 err:
  dataBufferDestroy(&doclist);
  return rc;
}

/****************************************************************/
/* Used to hold hashtable data for sorting. */
typedef struct TermData {
  const char *pTerm;
  int nTerm;
  DLCollector *pCollector;
} TermData;

/* Orders TermData elements in strcmp fashion ( <0 for less-than, 0
** for equal, >0 for greater-than).
*/
static int termDataCmp(const void *av, const void *bv){
  const TermData *a = (const TermData *)av;
  const TermData *b = (const TermData *)bv;
  int n = a->nTerm<b->nTerm ? a->nTerm : b->nTerm;
  int c = memcmp(a->pTerm, b->pTerm, n);
  if( c!=0 ) return c;
  return a->nTerm-b->nTerm;
}

/* Order pTerms data by term, then write a new level 0 segment using
** LeafWriter.
*/
static int writeZeroSegment(fulltext_vtab *v, fts3Hash *pTerms){
  fts3HashElem *e;
  int idx, rc, i, n;
  TermData *pData;
  LeafWriter writer;
  DataBuffer dl;

  /* Determine the next index at level 0, merging as necessary. */
  rc = segdirNextIndex(v, 0, &idx);
  if( rc!=SQLITE_OK ) return rc;

  n = fts3HashCount(pTerms);
  pData = sqlite3_malloc(n*sizeof(TermData));

  for(i = 0, e = fts3HashFirst(pTerms); e; i++, e = fts3HashNext(e)){
    assert( i<n );
    pData[i].pTerm = fts3HashKey(e);
    pData[i].nTerm = fts3HashKeysize(e);
    pData[i].pCollector = fts3HashData(e);
  }
  assert( i==n );

  /* TODO(shess) Should we allow user-defined collation sequences,
  ** here?  I think we only need that once we support prefix searches.
  */
  if( n>1 ) qsort(pData, n, sizeof(*pData), termDataCmp);

  /* TODO(shess) Refactor so that we can write directly to the segment
  ** DataBuffer, as happens for segment merges.
  */
  leafWriterInit(0, idx, &writer);
  dataBufferInit(&dl, 0);
  for(i=0; i<n; i++){
    dataBufferReset(&dl);
    dlcAddDoclist(pData[i].pCollector, &dl);
    rc = leafWriterStep(v, &writer,
                        pData[i].pTerm, pData[i].nTerm, dl.pData, dl.nData);
    if( rc!=SQLITE_OK ) goto err;
  }
  rc = leafWriterFinalize(v, &writer);

 err:
  dataBufferDestroy(&dl);
  sqlite3_free(pData);
  leafWriterDestroy(&writer);
  return rc;
}

/* If pendingTerms has data, free it. */
static int clearPendingTerms(fulltext_vtab *v){
  if( v->nPendingData>=0 ){
    fts3HashElem *e;
    for(e=fts3HashFirst(&v->pendingTerms); e; e=fts3HashNext(e)){
      dlcDelete(fts3HashData(e));
    }
    fts3HashClear(&v->pendingTerms);
    v->nPendingData = -1;
  }
  return SQLITE_OK;
}

/* If pendingTerms has data, flush it to a level-zero segment, and
** free it.
*/
static int flushPendingTerms(fulltext_vtab *v){
  if( v->nPendingData>=0 ){
    int rc = writeZeroSegment(v, &v->pendingTerms);
    if( rc==SQLITE_OK ) clearPendingTerms(v);
    return rc;
  }
  return SQLITE_OK;
}

/* If pendingTerms is "too big", or docid is out of order, flush it.
** Regardless, be certain that pendingTerms is initialized for use.
*/
static int initPendingTerms(fulltext_vtab *v, sqlite_int64 iDocid){
  /* TODO(shess) Explore whether partially flushing the buffer on
  ** forced-flush would provide better performance.  I suspect that if
  ** we ordered the doclists by size and flushed the largest until the
  ** buffer was half empty, that would let the less frequent terms
  ** generate longer doclists.
  */
  if( iDocid<=v->iPrevDocid || v->nPendingData>kPendingThreshold ){
    int rc = flushPendingTerms(v);
    if( rc!=SQLITE_OK ) return rc;
  }
  if( v->nPendingData<0 ){
    fts3HashInit(&v->pendingTerms, FTS3_HASH_STRING, 1);
    v->nPendingData = 0;
  }
  v->iPrevDocid = iDocid;
  return SQLITE_OK;
}

/* This function implements the xUpdate callback; it is the top-level entry
 * point for inserting, deleting or updating a row in a full-text table. */
static int fulltextUpdate(sqlite3_vtab *pVtab, int nArg, sqlite3_value **ppArg,
                          sqlite_int64 *pRowid){
  fulltext_vtab *v = (fulltext_vtab *) pVtab;
  int rc;

  FTSTRACE(("FTS3 Update %p\n", pVtab));

  if( nArg<2 ){
    rc = index_delete(v, sqlite3_value_int64(ppArg[0]));
    if( rc==SQLITE_OK ){
      /* If we just deleted the last row in the table, clear out the
      ** index data.
      */
      rc = content_exists(v);
      if( rc==SQLITE_ROW ){
        rc = SQLITE_OK;
      }else if( rc==SQLITE_DONE ){
        /* Clear the pending terms so we don't flush a useless level-0
        ** segment when the transaction closes.
        */
        rc = clearPendingTerms(v);
        if( rc==SQLITE_OK ){
          rc = segdir_delete_all(v);
        }
      }
    }
  } else if( sqlite3_value_type(ppArg[0]) != SQLITE_NULL ){
    /* An update:
     * ppArg[0] = old rowid
     * ppArg[1] = new rowid
     * ppArg[2..2+v->nColumn-1] = values
     * ppArg[2+v->nColumn] = value for magic column (we ignore this)
     * ppArg[2+v->nColumn+1] = value for docid
     */
    sqlite_int64 rowid = sqlite3_value_int64(ppArg[0]);
    if( sqlite3_value_type(ppArg[1]) != SQLITE_INTEGER ||
        sqlite3_value_int64(ppArg[1]) != rowid ){
      rc = SQLITE_ERROR;  /* we don't allow changing the rowid */
    }else if( sqlite3_value_type(ppArg[2+v->nColumn+1]) != SQLITE_INTEGER ||
              sqlite3_value_int64(ppArg[2+v->nColumn+1]) != rowid ){
      rc = SQLITE_ERROR;  /* we don't allow changing the docid */
    }else{
      assert( nArg==2+v->nColumn+2);
      rc = index_update(v, rowid, &ppArg[2]);
    }
  } else {
    /* An insert:
     * ppArg[1] = requested rowid
     * ppArg[2..2+v->nColumn-1] = values
     * ppArg[2+v->nColumn] = value for magic column (we ignore this)
     * ppArg[2+v->nColumn+1] = value for docid
     */
    sqlite3_value *pRequestDocid = ppArg[2+v->nColumn+1];
    assert( nArg==2+v->nColumn+2);
    if( SQLITE_NULL != sqlite3_value_type(pRequestDocid) &&
        SQLITE_NULL != sqlite3_value_type(ppArg[1]) ){
      /* TODO(shess) Consider allowing this to work if the values are
      ** identical.  I'm inclined to discourage that usage, though,
      ** given that both rowid and docid are special columns.  Better
      ** would be to define one or the other as the default winner,
      ** but should it be fts3-centric (docid) or SQLite-centric
      ** (rowid)?
      */
      rc = SQLITE_ERROR;
    }else{
      if( SQLITE_NULL == sqlite3_value_type(pRequestDocid) ){
        pRequestDocid = ppArg[1];
      }
      rc = index_insert(v, pRequestDocid, &ppArg[2], pRowid);
    }
  }

  return rc;
}

static int fulltextSync(sqlite3_vtab *pVtab){
  FTSTRACE(("FTS3 xSync()\n"));
  return flushPendingTerms((fulltext_vtab *)pVtab);
}

static int fulltextBegin(sqlite3_vtab *pVtab){
  fulltext_vtab *v = (fulltext_vtab *) pVtab;
  FTSTRACE(("FTS3 xBegin()\n"));

  /* Any buffered updates should have been cleared by the previous
  ** transaction.
  */
  assert( v->nPendingData<0 );
  return clearPendingTerms(v);
}

static int fulltextCommit(sqlite3_vtab *pVtab){
  fulltext_vtab *v = (fulltext_vtab *) pVtab;
  FTSTRACE(("FTS3 xCommit()\n"));

  /* Buffered updates should have been cleared by fulltextSync(). */
  assert( v->nPendingData<0 );
  return clearPendingTerms(v);
}

static int fulltextRollback(sqlite3_vtab *pVtab){
  FTSTRACE(("FTS3 xRollback()\n"));
  return clearPendingTerms((fulltext_vtab *)pVtab);
}

/*
** Implementation of the snippet() function for FTS3
*/
static void snippetFunc(
  sqlite3_context *pContext,
  int argc,
  sqlite3_value **argv
){
  fulltext_cursor *pCursor;
  if( argc<1 ) return;
  if( sqlite3_value_type(argv[0])!=SQLITE_BLOB ||
      sqlite3_value_bytes(argv[0])!=sizeof(pCursor) ){
    sqlite3_result_error(pContext, "illegal first argument to html_snippet",-1);
  }else{
    const char *zStart = "<b>";
    const char *zEnd = "</b>";
    const char *zEllipsis = "<b>...</b>";
    memcpy(&pCursor, sqlite3_value_blob(argv[0]), sizeof(pCursor));
    if( argc>=2 ){
      zStart = (const char*)sqlite3_value_text(argv[1]);
      if( argc>=3 ){
        zEnd = (const char*)sqlite3_value_text(argv[2]);
        if( argc>=4 ){
          zEllipsis = (const char*)sqlite3_value_text(argv[3]);
        }
      }
    }
    snippetAllOffsets(pCursor);
    snippetText(pCursor, zStart, zEnd, zEllipsis);
    sqlite3_result_text(pContext, pCursor->snippet.zSnippet,
                        pCursor->snippet.nSnippet, SQLITE_STATIC);
  }
}

/*
** Implementation of the offsets() function for FTS3
*/
static void snippetOffsetsFunc(
  sqlite3_context *pContext,
  int argc,
  sqlite3_value **argv
){
  fulltext_cursor *pCursor;
  if( argc<1 ) return;
  if( sqlite3_value_type(argv[0])!=SQLITE_BLOB ||
      sqlite3_value_bytes(argv[0])!=sizeof(pCursor) ){
    sqlite3_result_error(pContext, "illegal first argument to offsets",-1);
  }else{
    memcpy(&pCursor, sqlite3_value_blob(argv[0]), sizeof(pCursor));
    snippetAllOffsets(pCursor);
    snippetOffsetText(&pCursor->snippet);
    sqlite3_result_text(pContext,
                        pCursor->snippet.zOffset, pCursor->snippet.nOffset,
                        SQLITE_STATIC);
  }

}

/* OptLeavesReader is nearly identical to LeavesReader, except that
** where LeavesReader is geared towards the merging of complete
** segment levels (with exactly MERGE_COUNT segments), OptLeavesReader
** is geared towards implementation of the optimize() function, and
** can merge all segments simultaneously.  This version may be
** somewhat less efficient than LeavesReader because it merges into an
** accumulator rather than doing an N-way merge, but since segment
** size grows exponentially (so segment count logrithmically) this is
** probably not an immediate problem.
*/
/* TODO(shess): Prove that assertion, or extend the merge code to
** merge tree fashion (like the prefix-searching code does).
*/
/* TODO(shess): OptLeavesReader and LeavesReader could probably be
** merged with little or no loss of performance for LeavesReader.  The
** merged code would need to handle >MERGE_COUNT segments, and would
** also need to be able to optionally optimize away deletes.
*/
typedef struct OptLeavesReader {
  /* Segment number, to order readers by age. */
  int segment;
  LeavesReader reader;
} OptLeavesReader;

static int optLeavesReaderAtEnd(OptLeavesReader *pReader){
  return leavesReaderAtEnd(&pReader->reader);
}
static int optLeavesReaderTermBytes(OptLeavesReader *pReader){
  return leavesReaderTermBytes(&pReader->reader);
}
static const char *optLeavesReaderData(OptLeavesReader *pReader){
  return leavesReaderData(&pReader->reader);
}
static int optLeavesReaderDataBytes(OptLeavesReader *pReader){
  return leavesReaderDataBytes(&pReader->reader);
}
static const char *optLeavesReaderTerm(OptLeavesReader *pReader){
  return leavesReaderTerm(&pReader->reader);
}
static int optLeavesReaderStep(fulltext_vtab *v, OptLeavesReader *pReader){
  return leavesReaderStep(v, &pReader->reader);
}
static int optLeavesReaderTermCmp(OptLeavesReader *lr1, OptLeavesReader *lr2){
  return leavesReaderTermCmp(&lr1->reader, &lr2->reader);
}
/* Order by term ascending, segment ascending (oldest to newest), with
** exhausted readers to the end.
*/
static int optLeavesReaderCmp(OptLeavesReader *lr1, OptLeavesReader *lr2){
  int c = optLeavesReaderTermCmp(lr1, lr2);
  if( c!=0 ) return c;
  return lr1->segment-lr2->segment;
}
/* Bubble pLr[0] to appropriate place in pLr[1..nLr-1].  Assumes that
** pLr[1..nLr-1] is already sorted.
*/
static void optLeavesReaderReorder(OptLeavesReader *pLr, int nLr){
  while( nLr>1 && optLeavesReaderCmp(pLr, pLr+1)>0 ){
    OptLeavesReader tmp = pLr[0];
    pLr[0] = pLr[1];
    pLr[1] = tmp;
    nLr--;
    pLr++;
  }
}

/* optimize() helper function.  Put the readers in order and iterate
** through them, merging doclists for matching terms into pWriter.
** Returns SQLITE_OK on success, or the SQLite error code which
** prevented success.
*/
static int optimizeInternal(fulltext_vtab *v,
                            OptLeavesReader *readers, int nReaders,
                            LeafWriter *pWriter){
  int i, rc = SQLITE_OK;
  DataBuffer doclist, merged, tmp;

  /* Order the readers. */
  i = nReaders;
  while( i-- > 0 ){
    optLeavesReaderReorder(&readers[i], nReaders-i);
  }

  dataBufferInit(&doclist, LEAF_MAX);
  dataBufferInit(&merged, LEAF_MAX);

  /* Exhausted readers bubble to the end, so when the first reader is
  ** at eof, all are at eof.
  */
  while( !optLeavesReaderAtEnd(&readers[0]) ){

    /* Figure out how many readers share the next term. */
    for(i=1; i<nReaders && !optLeavesReaderAtEnd(&readers[i]); i++){
      if( 0!=optLeavesReaderTermCmp(&readers[0], &readers[i]) ) break;
    }

    /* Special-case for no merge. */
    if( i==1 ){
      /* Trim deletions from the doclist. */
      dataBufferReset(&merged);
      docListTrim(DL_DEFAULT,
                  optLeavesReaderData(&readers[0]),
                  optLeavesReaderDataBytes(&readers[0]),
                  -1, DL_DEFAULT, &merged);
    }else{
      DLReader dlReaders[MERGE_COUNT];
      int iReader, nReaders;

      /* Prime the pipeline with the first reader's doclist.  After
      ** one pass index 0 will reference the accumulated doclist.
      */
      dlrInit(&dlReaders[0], DL_DEFAULT,
              optLeavesReaderData(&readers[0]),
              optLeavesReaderDataBytes(&readers[0]));
      iReader = 1;

      assert( iReader<i );  /* Must execute the loop at least once. */
      while( iReader<i ){
        /* Merge 16 inputs per pass. */
        for( nReaders=1; iReader<i && nReaders<MERGE_COUNT;
             iReader++, nReaders++ ){
          dlrInit(&dlReaders[nReaders], DL_DEFAULT,
                  optLeavesReaderData(&readers[iReader]),
                  optLeavesReaderDataBytes(&readers[iReader]));
        }

        /* Merge doclists and swap result into accumulator. */
        dataBufferReset(&merged);
        docListMerge(&merged, dlReaders, nReaders);
        tmp = merged;
        merged = doclist;
        doclist = tmp;

        while( nReaders-- > 0 ){
          dlrDestroy(&dlReaders[nReaders]);
        }

        /* Accumulated doclist to reader 0 for next pass. */
        dlrInit(&dlReaders[0], DL_DEFAULT, doclist.pData, doclist.nData);
      }

      /* Destroy reader that was left in the pipeline. */
      dlrDestroy(&dlReaders[0]);

      /* Trim deletions from the doclist. */
      dataBufferReset(&merged);
      docListTrim(DL_DEFAULT, doclist.pData, doclist.nData,
                  -1, DL_DEFAULT, &merged);
    }

    /* Only pass doclists with hits (skip if all hits deleted). */
    if( merged.nData>0 ){
      rc = leafWriterStep(v, pWriter,
                          optLeavesReaderTerm(&readers[0]),
                          optLeavesReaderTermBytes(&readers[0]),
                          merged.pData, merged.nData);
      if( rc!=SQLITE_OK ) goto err;
    }

    /* Step merged readers to next term and reorder. */
    while( i-- > 0 ){
      rc = optLeavesReaderStep(v, &readers[i]);
      if( rc!=SQLITE_OK ) goto err;

      optLeavesReaderReorder(&readers[i], nReaders-i);
    }
  }

 err:
  dataBufferDestroy(&doclist);
  dataBufferDestroy(&merged);
  return rc;
}

/* Implement optimize() function for FTS3.  optimize(t) merges all
** segments in the fts index into a single segment.  't' is the magic
** table-named column.
*/
static void optimizeFunc(sqlite3_context *pContext,
                         int argc, sqlite3_value **argv){
  fulltext_cursor *pCursor;
  if( argc>1 ){
    sqlite3_result_error(pContext, "excess arguments to optimize()",-1);
  }else if( sqlite3_value_type(argv[0])!=SQLITE_BLOB ||
            sqlite3_value_bytes(argv[0])!=sizeof(pCursor) ){
    sqlite3_result_error(pContext, "illegal first argument to optimize",-1);
  }else{
    fulltext_vtab *v;
    int i, rc, iMaxLevel;
    OptLeavesReader *readers;
    int nReaders;
    LeafWriter writer;
    sqlite3_stmt *s;

    memcpy(&pCursor, sqlite3_value_blob(argv[0]), sizeof(pCursor));
    v = cursor_vtab(pCursor);

    /* Flush any buffered updates before optimizing. */
    rc = flushPendingTerms(v);
    if( rc!=SQLITE_OK ) goto err;

    rc = segdir_count(v, &nReaders, &iMaxLevel);
    if( rc!=SQLITE_OK ) goto err;
    if( nReaders==0 || nReaders==1 ){
      sqlite3_result_text(pContext, "Index already optimal", -1,
                          SQLITE_STATIC);
      return;
    }

    rc = sql_get_statement(v, SEGDIR_SELECT_ALL_STMT, &s);
    if( rc!=SQLITE_OK ) goto err;

    readers = sqlite3_malloc(nReaders*sizeof(readers[0]));
    if( readers==NULL ) goto err;

    /* Note that there will already be a segment at this position
    ** until we call segdir_delete() on iMaxLevel.
    */
    leafWriterInit(iMaxLevel, 0, &writer);

    i = 0;
    while( (rc = sqlite3_step(s))==SQLITE_ROW ){
      sqlite_int64 iStart = sqlite3_column_int64(s, 0);
      sqlite_int64 iEnd = sqlite3_column_int64(s, 1);
      const char *pRootData = sqlite3_column_blob(s, 2);
      int nRootData = sqlite3_column_bytes(s, 2);

      assert( i<nReaders );
      rc = leavesReaderInit(v, -1, iStart, iEnd, pRootData, nRootData,
                            &readers[i].reader);
      if( rc!=SQLITE_OK ) break;

      readers[i].segment = i;
      i++;
    }

    /* If we managed to successfully read them all, optimize them. */
    if( rc==SQLITE_DONE ){

      assert( i==nReaders );
      rc = optimizeInternal(v, readers, nReaders, &writer);

    }

    while( i-- > 0 ){
      leavesReaderDestroy(&readers[i].reader);
    }
    sqlite3_free(readers);

    /* If we've successfully gotten to here, delete the old segments
    ** and flush the interior structure of the new segment.
    */
    if( rc==SQLITE_OK ){
      for( i=0; i<=iMaxLevel; i++ ){
        rc = segdir_delete(v, i);
        if( rc!=SQLITE_OK ) break;
      }

      if( rc==SQLITE_OK ) rc = leafWriterFinalize(v, &writer);
    }

    leafWriterDestroy(&writer);

    if( rc!=SQLITE_OK ) goto err;

    sqlite3_result_text(pContext, "Index optimized", -1, SQLITE_STATIC);
    return;

    /* TODO(shess): Error-handling needs to be improved along the
    ** lines of the dump_ functions.



    */
 err:
    {
      char buf[512];
      sqlite3_snprintf(sizeof(buf), buf, "Error in optimize: %s",
                       sqlite3_errmsg(sqlite3_context_db_handle(pContext)));
      sqlite3_result_error(pContext, buf, -1);
    }
  }
}

#ifdef SQLITE_TEST
/* Generate an error of the form "<prefix>: <msg>".  If msg is NULL,
** pull the error from the context's db handle.

*/
static void generateError(sqlite3_context *pContext,
                          const char *prefix, const char *msg){
  char buf[512];
  if( msg==NULL ) msg = sqlite3_errmsg(sqlite3_context_db_handle(pContext));
  sqlite3_snprintf(sizeof(buf), buf, "%s: %s", prefix, msg);
  sqlite3_result_error(pContext, buf, -1);
}

/* Helper function to collect the set of terms in the segment into
** pTerms.  The segment is defined by the leaf nodes between
** iStartBlockid and iEndBlockid, inclusive, or by the contents of
** pRootData if iStartBlockid is 0 (in which case the entire segment
** fit in a leaf).

*/
static int collectSegmentTerms(fulltext_vtab *v, sqlite3_stmt *s,
                               fts3Hash *pTerms){
  const sqlite_int64 iStartBlockid = sqlite3_column_int64(s, 0);
  const sqlite_int64 iEndBlockid = sqlite3_column_int64(s, 1);
  const char *pRootData = sqlite3_column_blob(s, 2);
  const int nRootData = sqlite3_column_bytes(s, 2);
  LeavesReader reader;
  int rc = leavesReaderInit(v, 0, iStartBlockid, iEndBlockid,
                            pRootData, nRootData, &reader);
  if( rc!=SQLITE_OK ) return rc;

  while( rc==SQLITE_OK && !leavesReaderAtEnd(&reader) ){
    const char *pTerm = leavesReaderTerm(&reader);
    const int nTerm = leavesReaderTermBytes(&reader);
    void *oldValue = sqlite3Fts3HashFind(pTerms, pTerm, nTerm);
    void *newValue = (void *)((char *)oldValue+1);

    /* From the comment before sqlite3Fts3HashInsert in fts3_hash.c,
    ** the data value passed is returned in case of malloc failure.
    */
    if( newValue==sqlite3Fts3HashInsert(pTerms, pTerm, nTerm, newValue) ){
      rc = SQLITE_NOMEM;
    }else{
      rc = leavesReaderStep(v, &reader);
    }
  }

  leavesReaderDestroy(&reader);
  return rc;
}

/* Helper function to build the result string for dump_terms(). */
static int generateTermsResult(sqlite3_context *pContext, fts3Hash *pTerms){
  int iTerm, nTerms, nResultBytes, iByte;
  char *result;
  TermData *pData;
  fts3HashElem *e;

  /* Iterate pTerms to generate an array of terms in pData for
  ** sorting.
  */
  nTerms = fts3HashCount(pTerms);
  assert( nTerms>0 );
  pData = sqlite3_malloc(nTerms*sizeof(TermData));
  if( pData==NULL ) return SQLITE_NOMEM;

  nResultBytes = 0;
  for(iTerm = 0, e = fts3HashFirst(pTerms); e; iTerm++, e = fts3HashNext(e)){
    nResultBytes += fts3HashKeysize(e)+1;   /* Term plus trailing space */
    assert( iTerm<nTerms );
    pData[iTerm].pTerm = fts3HashKey(e);
    pData[iTerm].nTerm = fts3HashKeysize(e);
    pData[iTerm].pCollector = fts3HashData(e);  /* unused */
  }
  assert( iTerm==nTerms );

  assert( nResultBytes>0 );   /* nTerms>0, nResultsBytes must be, too. */
  result = sqlite3_malloc(nResultBytes);
  if( result==NULL ){
    sqlite3_free(pData);
    return SQLITE_NOMEM;
  }

  if( nTerms>1 ) qsort(pData, nTerms, sizeof(*pData), termDataCmp);

  /* Read the terms in order to build the result. */
  iByte = 0;
  for(iTerm=0; iTerm<nTerms; ++iTerm){
    memcpy(result+iByte, pData[iTerm].pTerm, pData[iTerm].nTerm);
    iByte += pData[iTerm].nTerm;
    result[iByte++] = ' ';
  }
  assert( iByte==nResultBytes );
  assert( result[nResultBytes-1]==' ' );
  result[nResultBytes-1] = '\0';

  /* Passes away ownership of result. */
  sqlite3_result_text(pContext, result, nResultBytes-1, sqlite3_free);
  sqlite3_free(pData);
  return SQLITE_OK;
}

/* Implements dump_terms() for use in inspecting the fts3 index from
** tests.  TEXT result containing the ordered list of terms joined by
** spaces.  dump_terms(t, level, idx) dumps the terms for the segment
** specified by level, idx (in %_segdir), while dump_terms(t) dumps
** all terms in the index.  In both cases t is the fts table's magic
** table-named column.
*/
static void dumpTermsFunc(
  sqlite3_context *pContext,

  int argc, sqlite3_value **argv
){
  fulltext_cursor *pCursor;
  if( argc!=3 && argc!=1 ){
    generateError(pContext, "dump_terms", "incorrect arguments");
  }else if( sqlite3_value_type(argv[0])!=SQLITE_BLOB ||
            sqlite3_value_bytes(argv[0])!=sizeof(pCursor) ){
    generateError(pContext, "dump_terms", "illegal first argument");
  }else{
    fulltext_vtab *v;
    fts3Hash terms;
    sqlite3_stmt *s = NULL;
    int rc;

    memcpy(&pCursor, sqlite3_value_blob(argv[0]), sizeof(pCursor));
    v = cursor_vtab(pCursor);

    /* If passed only the cursor column, get all segments.  Otherwise
    ** get the segment described by the following two arguments.
    */
    if( argc==1 ){
      rc = sql_get_statement(v, SEGDIR_SELECT_ALL_STMT, &s);
    }else{
      rc = sql_get_statement(v, SEGDIR_SELECT_SEGMENT_STMT, &s);
      if( rc==SQLITE_OK ){
        rc = sqlite3_bind_int(s, 1, sqlite3_value_int(argv[1]));
        if( rc==SQLITE_OK ){
          rc = sqlite3_bind_int(s, 2, sqlite3_value_int(argv[2]));
        }
      }
    }

    if( rc!=SQLITE_OK ){
      generateError(pContext, "dump_terms", NULL);
      return;
    }

    /* Collect the terms for each segment. */
    sqlite3Fts3HashInit(&terms, FTS3_HASH_STRING, 1);
    while( (rc = sqlite3_step(s))==SQLITE_ROW ){
      rc = collectSegmentTerms(v, s, &terms);
      if( rc!=SQLITE_OK ) break;
    }

    if( rc!=SQLITE_DONE ){
      sqlite3_reset(s);
      generateError(pContext, "dump_terms", NULL);
    }else{
      const int nTerms = fts3HashCount(&terms);
      if( nTerms>0 ){
        rc = generateTermsResult(pContext, &terms);
        if( rc==SQLITE_NOMEM ){
          generateError(pContext, "dump_terms", "out of memory");
        }else{
          assert( rc==SQLITE_OK );
        }
      }else if( argc==3 ){
        /* The specific segment asked for could not be found. */
        generateError(pContext, "dump_terms", "segment not found");
      }else{
        /* No segments found. */
        /* TODO(shess): It should be impossible to reach this.  This
        ** case can only happen for an empty table, in which case
        ** SQLite has no rows to call this function on.
        */
        sqlite3_result_null(pContext);
      }
    }
    sqlite3Fts3HashClear(&terms);
  }
}

/* Expand the DL_DEFAULT doclist in pData into a text result in
** pContext.
*/
static void createDoclistResult(sqlite3_context *pContext,
                                const char *pData, int nData){
  DataBuffer dump;
  DLReader dlReader;

  assert( pData!=NULL && nData>0 );

  dataBufferInit(&dump, 0);
  dlrInit(&dlReader, DL_DEFAULT, pData, nData);
  for( ; !dlrAtEnd(&dlReader); dlrStep(&dlReader) ){
    char buf[256];
    PLReader plReader;

    plrInit(&plReader, &dlReader);
    if( DL_DEFAULT==DL_DOCIDS || plrAtEnd(&plReader) ){
      sqlite3_snprintf(sizeof(buf), buf, "[%lld] ", dlrDocid(&dlReader));
      dataBufferAppend(&dump, buf, strlen(buf));
    }else{
      int iColumn = plrColumn(&plReader);

      sqlite3_snprintf(sizeof(buf), buf, "[%lld %d[",
                       dlrDocid(&dlReader), iColumn);
      dataBufferAppend(&dump, buf, strlen(buf));

      for( ; !plrAtEnd(&plReader); plrStep(&plReader) ){
        if( plrColumn(&plReader)!=iColumn ){
          iColumn = plrColumn(&plReader);
          sqlite3_snprintf(sizeof(buf), buf, "] %d[", iColumn);
          assert( dump.nData>0 );
          dump.nData--;                     /* Overwrite trailing space. */
          assert( dump.pData[dump.nData]==' ');
          dataBufferAppend(&dump, buf, strlen(buf));


        }
        if( DL_DEFAULT==DL_POSITIONS_OFFSETS ){
          sqlite3_snprintf(sizeof(buf), buf, "%d,%d,%d ",
                           plrPosition(&plReader),
                           plrStartOffset(&plReader), plrEndOffset(&plReader));
        }else if( DL_DEFAULT==DL_POSITIONS ){
          sqlite3_snprintf(sizeof(buf), buf, "%d ", plrPosition(&plReader));
        }else{
          assert( NULL=="Unhandled DL_DEFAULT value");

        }
        dataBufferAppend(&dump, buf, strlen(buf));
      }
      plrDestroy(&plReader);

      assert( dump.nData>0 );
      dump.nData--;                     /* Overwrite trailing space. */
      assert( dump.pData[dump.nData]==' ');
      dataBufferAppend(&dump, "]] ", 3);
    }
  }
  dlrDestroy(&dlReader);



  assert( dump.nData>0 );
  dump.nData--;                     /* Overwrite trailing space. */
  assert( dump.pData[dump.nData]==' ');
  dump.pData[dump.nData] = '\0';
  assert( dump.nData>0 );

  /* Passes ownership of dump's buffer to pContext. */
  sqlite3_result_text(pContext, dump.pData, dump.nData, sqlite3_free);
  dump.pData = NULL;
  dump.nData = dump.nCapacity = 0;
}

/* Implements dump_doclist() for use in inspecting the fts3 index from
** tests.  TEXT result containing a string representation of the
** doclist for the indicated term.  dump_doclist(t, term, level, idx)
** dumps the doclist for term from the segment specified by level, idx
** (in %_segdir), while dump_doclist(t, term) dumps the logical
** doclist for the term across all segments.  The per-segment doclist
** can contain deletions, while the full-index doclist will not
** (deletions are omitted).
**
** Result formats differ with the setting of DL_DEFAULTS.  Examples:
**
** DL_DOCIDS: [1] [3] [7]
** DL_POSITIONS: [1 0[0 4] 1[17]] [3 1[5]]
** DL_POSITIONS_OFFSETS: [1 0[0,0,3 4,23,26] 1[17,102,105]] [3 1[5,20,23]]
**
** In each case the number after the outer '[' is the docid.  In the
** latter two cases, the number before the inner '[' is the column
** associated with the values within.  For DL_POSITIONS the numbers
** within are the positions, for DL_POSITIONS_OFFSETS they are the
** position, the start offset, and the end offset.
*/
static void dumpDoclistFunc(
  sqlite3_context *pContext,
  int argc, sqlite3_value **argv
){
  fulltext_cursor *pCursor;
  if( argc!=2 && argc!=4 ){
    generateError(pContext, "dump_doclist", "incorrect arguments");
  }else if( sqlite3_value_type(argv[0])!=SQLITE_BLOB ||
            sqlite3_value_bytes(argv[0])!=sizeof(pCursor) ){
    generateError(pContext, "dump_doclist", "illegal first argument");
  }else if( sqlite3_value_text(argv[1])==NULL ||
            sqlite3_value_text(argv[1])[0]=='\0' ){
    generateError(pContext, "dump_doclist", "empty second argument");
  }else{
    const char *pTerm = (const char *)sqlite3_value_text(argv[1]);
    const int nTerm = strlen(pTerm);
    fulltext_vtab *v;
    int rc;
    DataBuffer doclist;

    memcpy(&pCursor, sqlite3_value_blob(argv[0]), sizeof(pCursor));
    v = cursor_vtab(pCursor);

    dataBufferInit(&doclist, 0);

    /* termSelect() yields the same logical doclist that queries are
    ** run against.
    */
    if( argc==2 ){
      rc = termSelect(v, v->nColumn, pTerm, nTerm, 0, DL_DEFAULT, &doclist);
    }else{
      sqlite3_stmt *s = NULL;

      /* Get our specific segment's information. */
      rc = sql_get_statement(v, SEGDIR_SELECT_SEGMENT_STMT, &s);
      if( rc==SQLITE_OK ){
        rc = sqlite3_bind_int(s, 1, sqlite3_value_int(argv[2]));
        if( rc==SQLITE_OK ){
          rc = sqlite3_bind_int(s, 2, sqlite3_value_int(argv[3]));
        }
      }

      if( rc==SQLITE_OK ){
        rc = sqlite3_step(s);

        if( rc==SQLITE_DONE ){
          dataBufferDestroy(&doclist);
          generateError(pContext, "dump_doclist", "segment not found");
          return;
        }

        /* Found a segment, load it into doclist. */
        if( rc==SQLITE_ROW ){
          const sqlite_int64 iLeavesEnd = sqlite3_column_int64(s, 1);
          const char *pData = sqlite3_column_blob(s, 2);
          const int nData = sqlite3_column_bytes(s, 2);

          /* loadSegment() is used by termSelect() to load each
          ** segment's data.
          */
          rc = loadSegment(v, pData, nData, iLeavesEnd, pTerm, nTerm, 0,
                           &doclist);

          if( rc==SQLITE_OK ){
            rc = sqlite3_step(s);

            /* Should not have more than one matching segment. */


            if( rc!=SQLITE_DONE ){
              sqlite3_reset(s);
              dataBufferDestroy(&doclist);
              generateError(pContext, "dump_doclist", "invalid segdir");
              return;
            }
            rc = SQLITE_OK;
          }
        }
      }

      sqlite3_reset(s);
    }

    if( rc==SQLITE_OK ){
      if( doclist.nData>0 ){
        createDoclistResult(pContext, doclist.pData, doclist.nData);
      }else{
        /* TODO(shess): This can happen if the term is not present, or
        ** if all instances of the term have been deleted and this is
        ** an all-index dump.  It may be interesting to distinguish
        ** these cases.
        */
        sqlite3_result_text(pContext, "", 0, SQLITE_STATIC);




      }
    }else if( rc==SQLITE_NOMEM ){
      /* Handle out-of-memory cases specially because if they are
      ** generated in fts3 code they may not be reflected in the db
      ** handle.
      */
      /* TODO(shess): Handle this more comprehensively.
      ** sqlite3ErrStr() has what I need, but is internal.
      */
      generateError(pContext, "dump_doclist", "out of memory");
    }else{
      generateError(pContext, "dump_doclist", NULL);
    }

    dataBufferDestroy(&doclist);
  }
}
#endif

/*
** This routine implements the xFindFunction method for the FTS3
** virtual table.
*/
static int fulltextFindFunction(
  sqlite3_vtab *pVtab,
  int nArg,

  const char *zName,
  void (**pxFunc)(sqlite3_context*,int,sqlite3_value**),
  void **ppArg
){
  if( strcmp(zName,"snippet")==0 ){
    *pxFunc = snippetFunc;
    return 1;
  }else if( strcmp(zName,"offsets")==0 ){





    *pxFunc = snippetOffsetsFunc;
    return 1;









  }else if( strcmp(zName,"optimize")==0 ){
    *pxFunc = optimizeFunc;
    return 1;
#ifdef SQLITE_TEST
    /* NOTE(shess): These functions are present only for testing
    ** purposes.  No particular effort is made to optimize their
    ** execution or how they build their results.
    */
  }else if( strcmp(zName,"dump_terms")==0 ){
    /* fprintf(stderr, "Found dump_terms\n"); */
    *pxFunc = dumpTermsFunc;
    return 1;
  }else if( strcmp(zName,"dump_doclist")==0 ){
    /* fprintf(stderr, "Found dump_doclist\n"); */
    *pxFunc = dumpDoclistFunc;
    return 1;
#endif
  }



  return 0;
}

/*
** Rename an fts3 table.
*/
static int fulltextRename(
  sqlite3_vtab *pVtab,
  const char *zName
){
  fulltext_vtab *p = (fulltext_vtab *)pVtab;
  int rc = SQLITE_NOMEM;


  char *zSql = sqlite3_mprintf(
    "ALTER TABLE %Q.'%q_content'  RENAME TO '%q_content';"
    "ALTER TABLE %Q.'%q_segments' RENAME TO '%q_segments';"
    "ALTER TABLE %Q.'%q_segdir'   RENAME TO '%q_segdir';"
    , p->zDb, p->zName, zName 
    , p->zDb, p->zName, zName 
    , p->zDb, p->zName, zName
  );
................................................................................
    sqlite3_free(zSql);
  }
  return rc;
}

static const sqlite3_module fts3Module = {
  /* iVersion      */ 0,
  /* xCreate       */ fulltextCreate,
  /* xConnect      */ fulltextConnect,
  /* xBestIndex    */ fulltextBestIndex,
  /* xDisconnect   */ fulltextDisconnect,
  /* xDestroy      */ fulltextDestroy,
  /* xOpen         */ fulltextOpen,
  /* xClose        */ fulltextClose,
  /* xFilter       */ fulltextFilter,
  /* xNext         */ fulltextNext,
  /* xEof          */ fulltextEof,
  /* xColumn       */ fulltextColumn,
  /* xRowid        */ fulltextRowid,
  /* xUpdate       */ fulltextUpdate,
  /* xBegin        */ fulltextBegin,
  /* xSync         */ fulltextSync,
  /* xCommit       */ fulltextCommit,
  /* xRollback     */ fulltextRollback,
  /* xFindFunction */ fulltextFindFunction,
  /* xRename */       fulltextRename,
};






static void hashDestroy(void *p){
  fts3Hash *pHash = (fts3Hash *)p;
  sqlite3Fts3HashClear(pHash);
  sqlite3_free(pHash);
}

/*
** The fts3 built-in tokenizers - "simple" and "porter" - are implemented
** in files fts3_tokenizer1.c and fts3_porter.c respectively. The following
................................................................................
** Function ...PorterTokenizerModule() sets *pModule to point to the
** porter tokenizer/stemmer implementation.
*/
void sqlite3Fts3SimpleTokenizerModule(sqlite3_tokenizer_module const**ppModule);
void sqlite3Fts3PorterTokenizerModule(sqlite3_tokenizer_module const**ppModule);
void sqlite3Fts3IcuTokenizerModule(sqlite3_tokenizer_module const**ppModule);

int sqlite3Fts3InitHashTable(sqlite3 *, fts3Hash *, const char *);

/*
** Initialise the fts3 extension. If this extension is built as part
** of the sqlite library, then this function is called directly by
** SQLite. If fts3 is built as a dynamically loadable extension, this
** function is called by the sqlite3_extension_init() entry point.
*/
int sqlite3Fts3Init(sqlite3 *db){
  int rc = SQLITE_OK;
  fts3Hash *pHash = 0;
  const sqlite3_tokenizer_module *pSimple = 0;
  const sqlite3_tokenizer_module *pPorter = 0;
  const sqlite3_tokenizer_module *pIcu = 0;

  sqlite3Fts3SimpleTokenizerModule(&pSimple);
  sqlite3Fts3PorterTokenizerModule(&pPorter);
#ifdef SQLITE_ENABLE_ICU
  sqlite3Fts3IcuTokenizerModule(&pIcu);
#endif

  /* Allocate and initialise the hash-table used to store tokenizers. */
  pHash = sqlite3_malloc(sizeof(fts3Hash));
  if( !pHash ){
    rc = SQLITE_NOMEM;
  }else{
    sqlite3Fts3HashInit(pHash, FTS3_HASH_STRING, 1);
  }

  /* Load the built-in tokenizers into the hash table */
................................................................................
  /* Create the virtual table wrapper around the hash-table and overload 
  ** the two scalar functions. If this is successful, register the
  ** module with sqlite.
  */
  if( SQLITE_OK==rc 
   && SQLITE_OK==(rc = sqlite3Fts3InitHashTable(db, pHash, "fts3_tokenizer"))
   && SQLITE_OK==(rc = sqlite3_overload_function(db, "snippet", -1))
   && SQLITE_OK==(rc = sqlite3_overload_function(db, "offsets", -1))
   && SQLITE_OK==(rc = sqlite3_overload_function(db, "optimize", -1))
#ifdef SQLITE_TEST
   && SQLITE_OK==(rc = sqlite3_overload_function(db, "dump_terms", -1))
   && SQLITE_OK==(rc = sqlite3_overload_function(db, "dump_doclist", -1))
#endif
  ){
    return sqlite3_create_module_v2(
        db, "fts3", &fts3Module, (void *)pHash, hashDestroy
    );
  }

  /* An error has occurred. Delete the hash table and return the error code. */
................................................................................
  const sqlite3_api_routines *pApi
){
  SQLITE_EXTENSION_INIT2(pApi)
  return sqlite3Fts3Init(db);
}
#endif

#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */







>
>


>


<


<
<
<




>







 







<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<







 







<
<
<
<
<







 







|
<
<
|
|
|
>
|







|



>
|
|
|
>
|





|









>
>
>
>
|

|





<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
>
>

<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
>
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
>
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
>
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|







 







|


<
>











|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>



>
>
>
>
>
>




|
|


|
|
|
>

|

<
|
|
|
<



<
|
<
<
<
>

>
>
>
|
<

<
|
|

>
>
>


<

|
>
>
>
>
>

<
>





>
|

<
<
>
|
>
>
>
>

<
>
>
>
>
|
<
<
>


>
|
<
>
>
>
>
>
>
>
>
>
>
>


<
<
>
>
>
>
>
>
>




|
<






<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<





|


|


|

|



|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<





|
|
|
|

|
<
<
<
<
<
<
<
<
<
<
<
|
>
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
>
|
<
>
|

<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
>
>
>
|
<
<
<
<
<
<
<
<
<
<
<
<
<
>
|
<
<
<
<
<
<
<
<
<
<
<
<
<
>

<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
|
>
|

<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
>
>
|
<
<
<
<
<
<
<
<
>
|
<
<
<
<
|
<
<
<
<
<
<
<
>
>
|
<
<
<
<

<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|

<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
>
|
<
<
<
>
>
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
>
>
>
>
|
<
<
<
<
<
<
<
<
<
<
<
|

<
<
<
<
<




|
|
<
>
|
|
|

<
<
<
<
>
>
>
>
>
|
<
>
>
>
>
>
>
>
>
>
|
|
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
>
>
>




|

|
|
|

|
|
>
>
|







 







|
|
|
|
|
|

|
|
|
|
|
|
|
|
|
|
|
|


>
>
>
>
>

|







 







<
<








|











|







 







|
|
<
<
<
<







 







|
274
275
276
277
278
279
280
281
282
283
284
285
286
287

288
289



290
291
292
293
294
295
296
297
298
299
300
301
...
309
310
311
312
313
314
315



















316
317
318
319
320
321
322
...
334
335
336
337
338
339
340





341
342
343
344
345
346
347
...
352
353
354
355
356
357
358
359


360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408




















































































































































































































































































































































































































































































































































































































































































































































































































































































































































































409



















410
411
412














































































































































































































































































































































































































































































































































































413
414

















































































































































































































































































































































































































































































































































































415
416
























































417
418






















419
420
421
422
423
424
425
426
...
428
429
430
431
432
433
434
435
436
437

438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776

1777
1778
1779

1780
1781
1782

1783



1784
1785
1786
1787
1788
1789

1790

1791
1792
1793
1794
1795
1796
1797
1798

1799
1800
1801
1802
1803
1804
1805
1806

1807
1808
1809
1810
1811
1812
1813
1814
1815


1816
1817
1818
1819
1820
1821
1822

1823
1824
1825
1826
1827


1828
1829
1830
1831
1832

1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845


1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857

1858
1859
1860
1861
1862
1863














































1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127






























































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138











2139
2140
2141




































































































































































































































2142










2143
2144

2145
2146
2147







2148


















2149
2150
2151
2152













2153
2154













2155
2156

























































































2157
2158
2159
2160
2161






































































































2162


2163
2164
2165








2166
2167




2168







2169
2170
2171




2172






































































2173
2174

















2175
2176



2177
2178
2179






















2180
2181
2182
2183
2184
2185











2186
2187





2188
2189
2190
2191
2192
2193

2194
2195
2196
2197
2198




2199
2200
2201
2202
2203
2204

2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216














2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
....
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
....
2292
2293
2294
2295
2296
2297
2298


2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
....
2340
2341
2342
2343
2344
2345
2346
2347
2348




2349
2350
2351
2352
2353
2354
2355
....
2368
2369
2370
2371
2372
2373
2374
2375

#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)

#if defined(SQLITE_ENABLE_FTS3) && !defined(SQLITE_CORE)
# define SQLITE_CORE 1
#endif

#include "fts3Int.h"

#include <assert.h>
#include <stdlib.h>
#include <stddef.h>
#include <stdio.h>
#include <string.h>


#include "fts3.h"



#ifndef SQLITE_CORE 
# include "sqlite3ext.h"
  SQLITE_EXTENSION_INIT1
#endif



/* TODO(shess) MAN, this thing needs some refactoring.  At minimum, it
** would be nice to order the file better, perhaps something along the
** lines of:
**
**  - utility functions
................................................................................

#if 0
# define FTSTRACE(A)  printf A; fflush(stdout)
#else
# define FTSTRACE(A)
#endif




















typedef enum DocListType {
  DL_DOCIDS,              /* docids only */
  DL_POSITIONS,           /* docids + positions */
  DL_POSITIONS_OFFSETS    /* docids + positions + offsets */
} DocListType;

/*
................................................................................

enum {
  POS_END = 0,        /* end of this position list */
  POS_COLUMN,         /* followed by new column number */
  POS_BASE
};






/* utility functions */

/* CLEAR() and SCRAMBLE() abstract memset() on a pointer to a single
** record to prevent errors of the form:
**
** my_function(SomeType *b){
**   memset(b, '\0', sizeof(b));  // sizeof(b)!=sizeof(*b)
................................................................................

#ifndef NDEBUG
#  define SCRAMBLE(b) memset(b, 0x55, sizeof(*(b)))
#else
#  define SCRAMBLE(b)
#endif

/* 


** Write a 64-bit variable-length integer to memory starting at p[0].
** The length of data written will be between 1 and FTS3_VARINT_MAX bytes.
** The number of bytes written is returned.
*/
int sqlite3Fts3PutVarint(char *p, sqlite_int64 v){
  unsigned char *q = (unsigned char *) p;
  sqlite_uint64 vu = v;
  do{
    *q++ = (unsigned char) ((vu & 0x7f) | 0x80);
    vu >>= 7;
  }while( vu!=0 );
  q[-1] &= 0x7f;  /* turn off high bit in final byte */
  assert( q - (unsigned char *)p <= FTS3_VARINT_MAX );
  return (int) (q - (unsigned char *)p);
}

/* 
** Read a 64-bit variable-length integer from memory starting at p[0].
** Return the number of bytes read, or 0 on error.
** The value is stored in *v.
*/
int sqlite3Fts3GetVarint(const char *p, sqlite_int64 *v){
  const unsigned char *q = (const unsigned char *) p;
  sqlite_uint64 x = 0, y = 1;
  while( (*q & 0x80) == 0x80 ){
    x += y * (*q++ & 0x7f);
    y <<= 7;
    if( q - (unsigned char *)p >= FTS3_VARINT_MAX ){  /* bad data */
      assert( 0 );
      return 0;
    }
  }
  x += y * (*q++);
  *v = (sqlite_int64) x;
  return (int) (q - (unsigned char *)p);
}

/*
** Similar to sqlite3Fts3GetVarint(), except that the output is truncated to a
** 32-bit integer before it is returned.
*/
int sqlite3Fts3GetVarint32(const char *p, int *pi){
 sqlite_int64 i;
 int ret = sqlite3Fts3GetVarint(p, &i);
 *pi = (int) i;
 assert( *pi==i );
 return ret;
}





















































































































































































































































































































































































































































































































































































































































































































































































































































































































































































/*



















** Return the number of bytes required to store the value passed as the
** first argument in varint form.
*/














































































































































































































































































































































































































































































































































































int sqlite3Fts3VarintLen(sqlite3_uint64 v){
  int i = 0;

















































































































































































































































































































































































































































































































































































  do{
    i++;
























































    v >>= 7;
  }while( v!=0 );






















  return i;
}

/*
** Convert an SQL-style quoted string into a normal string by removing
** the quote characters.  The conversion is done in-place.  If the
** input does not begin with a quote character, then this routine
** is a no-op.
................................................................................
** Examples:
**
**     "abc"   becomes   abc
**     'xyz'   becomes   xyz
**     [pqr]   becomes   pqr
**     `mno`   becomes   mno
*/
void sqlite3Fts3Dequote(char *z){
  int quote;
  int i, j;


  quote = z[0];
  switch( quote ){
    case '\'':  break;
    case '"':   break;
    case '`':   break;                /* For MySQL compatibility */
    case '[':   quote = ']';  break;  /* For MS SqlServer compatibility */
    default:    return;
  }
  for(i=1, j=0; z[i]; i++){
    if( z[i]==quote ){
      if( z[i+1]==quote ){
        z[j++] = (char)quote;
        i++;
      }else{
        z[j++] = 0;
        break;
      }
    }else{
      z[j++] = z[i];
    }
  }
}

static void fts3GetDeltaVarint(char **pp, sqlite3_int64 *pVal){
  sqlite3_int64 iVal;
  *pp += sqlite3Fts3GetVarint(*pp, &iVal);
  *pVal += iVal;
}

static void fts3GetDeltaVarint2(char **pp, char *pEnd, sqlite3_int64 *pVal){
  if( *pp>=pEnd ){
    *pp = 0;
  }else{
    fts3GetDeltaVarint(pp, pVal);
  }
}

/*
** The xDisconnect() virtual table method.
*/
static int fts3DisconnectMethod(sqlite3_vtab *pVtab){
  Fts3Table *p = (Fts3Table *)pVtab;
  int i;

  assert( p->nPendingData==0 );

  /* Free any prepared statements held */
  for(i=0; i<SizeofArray(p->aStmt); i++){
    sqlite3_finalize(p->aStmt[i]);
  }
  for(i=0; i<p->nLeavesStmt; i++){
    sqlite3_finalize(p->aLeavesStmt[i]);
  }
  sqlite3_free(p->zSelectLeaves);
  sqlite3_free(p->aLeavesStmt);

  /* Invoke the tokenizer destructor to free the tokenizer. */
  p->pTokenizer->pModule->xDestroy(p->pTokenizer);

  sqlite3_free(p);
  return SQLITE_OK;
}

/*
** The xDestroy() virtual table method.
*/
static int fts3DestroyMethod(sqlite3_vtab *pVtab){
  int rc;                         /* Return code */
  Fts3Table *p = (Fts3Table *)pVtab;

  /* Create a script to drop the underlying three storage tables. */
  char *zSql = sqlite3_mprintf(
      "DROP TABLE IF EXISTS %Q.'%q_content';"
      "DROP TABLE IF EXISTS %Q.'%q_segments';"
      "DROP TABLE IF EXISTS %Q.'%q_segdir';", 
      p->zDb, p->zName, p->zDb, p->zName, p->zDb, p->zName
  );

  /* If malloc has failed, set rc to SQLITE_NOMEM. Otherwise, try to
  ** execute the SQL script created above.
  */
  if( zSql ){
    rc = sqlite3_exec(p->db, zSql, 0, 0, 0);
    sqlite3_free(zSql);
  }else{
    rc = SQLITE_NOMEM;
  }

  /* If everything has worked, invoke fts3DisconnectMethod() to free the
  ** memory associated with the Fts3Table structure and return SQLITE_OK.
  ** Otherwise, return an SQLite error code.
  */
  return (rc==SQLITE_OK ? fts3DisconnectMethod(pVtab) : rc);
}


/*
** Invoke sqlite3_declare_vtab() to declare the schema for the FTS3 table
** passed as the first argument. This is done as part of the xConnect()
** and xCreate() methods.
*/
static int fts3DeclareVtab(Fts3Table *p){
  int i;                          /* Iterator variable */
  int rc;                         /* Return code */
  char *zSql;                     /* SQL statement passed to declare_vtab() */
  char *zCols;                    /* List of user defined columns */

  /* Create a list of user columns for the virtual table */
  zCols = sqlite3_mprintf("%Q, ", p->azColumn[0]);
  for(i=1; zCols && i<p->nColumn; i++){
    zCols = sqlite3_mprintf("%z%Q, ", zCols, p->azColumn[i]);
  }

  /* Create the whole "CREATE TABLE" statement to pass to SQLite */
  zSql = sqlite3_mprintf(
      "CREATE TABLE x(%s %Q HIDDEN, docid HIDDEN)", zCols, p->zName
  );

  if( !zCols || !zSql ){
    rc = SQLITE_NOMEM;
  }else{
    rc = sqlite3_declare_vtab(p->db, zSql);
  }

  sqlite3_free(zSql);
  sqlite3_free(zCols);
  return rc;
}

/*
** Create the backing store tables (%_content, %_segments and %_segdir)
** required by the FTS3 table passed as the only argument. This is done
** as part of the vtab xCreate() method.
*/
static int fts3CreateTables(Fts3Table *p){
  int rc;                         /* Return code */
  int i;                          /* Iterator variable */
  char *zContentCols;             /* Columns of %_content table */
  char *zSql;                     /* SQL script to create required tables */

  /* Create a list of user columns for the content table */
  zContentCols = sqlite3_mprintf("docid INTEGER PRIMARY KEY");
  for(i=0; zContentCols && i<p->nColumn; i++){
    char *z = p->azColumn[i];
    zContentCols = sqlite3_mprintf("%z, 'c%d%q'", zContentCols, i, z);
  }

  /* Create the whole SQL script */
  zSql = sqlite3_mprintf(
      "CREATE TABLE %Q.'%q_content'(%s);"
      "CREATE TABLE %Q.'%q_segments'(blockid INTEGER PRIMARY KEY, block BLOB);"
      "CREATE TABLE %Q.'%q_segdir'("
        "level INTEGER,"
        "idx INTEGER,"
        "start_block INTEGER,"
        "leaves_end_block INTEGER,"
        "end_block INTEGER,"
        "root BLOB,"
        "PRIMARY KEY(level, idx)"
      ");",
      p->zDb, p->zName, zContentCols, p->zDb, p->zName, p->zDb, p->zName
  );

  /* Unless a malloc() failure has occurred, execute the SQL script to 
  ** create the tables used to store data for this FTS3 virtual table.
  */
  if( zContentCols==0 || zSql==0 ){
    rc = SQLITE_NOMEM;
  }else{
    rc = sqlite3_exec(p->db, zSql, 0, 0, 0);
  }

  sqlite3_free(zSql);
  sqlite3_free(zContentCols);
  return rc;
}

/*
** This function is the implementation of both the xConnect and xCreate
** methods of the FTS3 virtual table.
**
** The argv[] array contains the following:
**
**   argv[0]   -> module name
**   argv[1]   -> database name
**   argv[2]   -> table name
**   argv[...] -> "column name" and other module argument fields.
*/
int fts3InitVtab(
  int isCreate,                   /* True for xCreate, false for xConnect */
  sqlite3 *db,                    /* The SQLite database connection */
  void *pAux,                     /* Hash table containing tokenizers */
  int argc,                       /* Number of elements in argv array */
  const char * const *argv,       /* xCreate/xConnect argument array */
  sqlite3_vtab **ppVTab,          /* Write the resulting vtab structure here */
  char **pzErr                    /* Write any error message here */
){
  Fts3Hash *pHash = (Fts3Hash *)pAux;
  Fts3Table *p;                   /* Pointer to allocated vtab */
  int rc;                         /* Return code */
  int i;                          /* Iterator variable */
  int nByte;                      /* Size of allocation used for *p */
  int iCol;
  int nString = 0;
  int nCol = 0;
  char *zCsr;
  int nDb;
  int nName;

  const char *zTokenizer = 0;               /* Name of tokenizer to use */
  sqlite3_tokenizer *pTokenizer = 0;        /* Tokenizer for this table */

#ifdef SQLITE_TEST
  const char *zTestParam = 0;
  if( strncmp(argv[argc-1], "test:", 5)==0 ){
    zTestParam = argv[argc-1];
    argc--;
  }
#endif

  nDb = (int)strlen(argv[1]) + 1;
  nName = (int)strlen(argv[2]) + 1;
  for(i=3; i<argc; i++){
    char const *z = argv[i];
    rc = sqlite3Fts3InitTokenizer(pHash, z, &pTokenizer, &zTokenizer, pzErr);
    if( rc!=SQLITE_OK ){
      return rc;
    }
    if( z!=zTokenizer ){
      nString += (int)(strlen(z) + 1);
    }
  }
  nCol = argc - 3 - (zTokenizer!=0);
  if( zTokenizer==0 ){
    rc = sqlite3Fts3InitTokenizer(pHash, 0, &pTokenizer, 0, pzErr);
    if( rc!=SQLITE_OK ){
      return rc;
    }
    assert( pTokenizer );
  }

  if( nCol==0 ){
    nCol = 1;
  }

  /* Allocate and populate the Fts3Table structure. */
  nByte = sizeof(Fts3Table) +              /* Fts3Table */
          nCol * sizeof(char *) +              /* azColumn */
          nName +                              /* zName */
          nDb +                                /* zDb */
          nString;                             /* Space for azColumn strings */
  p = (Fts3Table*)sqlite3_malloc(nByte);
  if( p==0 ){
    rc = SQLITE_NOMEM;
    goto fts3_init_out;
  }
  memset(p, 0, nByte);

  p->db = db;
  p->nColumn = nCol;
  p->nPendingData = 0;
  p->azColumn = (char **)&p[1];
  p->pTokenizer = pTokenizer;
  p->nNodeSize = 1000;
  zCsr = (char *)&p->azColumn[nCol];

  fts3HashInit(&p->pendingTerms, FTS3_HASH_STRING, 1);

  /* Fill in the zName and zDb fields of the vtab structure. */
  p->zName = zCsr;
  memcpy(zCsr, argv[2], nName);
  zCsr += nName;
  p->zDb = zCsr;
  memcpy(zCsr, argv[1], nDb);
  zCsr += nDb;

  /* Fill in the azColumn array */
  iCol = 0;
  for(i=3; i<argc; i++){
    if( argv[i]!=zTokenizer ){
      char *z; 
      int n;
      z = (char *)sqlite3Fts3NextToken(argv[i], &n);
      memcpy(zCsr, z, n);
      zCsr[n] = '\0';
      sqlite3Fts3Dequote(zCsr);
      p->azColumn[iCol++] = zCsr;
      zCsr += n+1;
      assert( zCsr <= &((char *)p)[nByte] );
    }
  }
  if( iCol==0 ){
    assert( nCol==1 );
    p->azColumn[0] = "content";
  }

  /* If this is an xCreate call, create the underlying tables in the 
  ** database. TODO: For xConnect(), it could verify that said tables exist.
  */
  if( isCreate ){
    rc = fts3CreateTables(p);
    if( rc!=SQLITE_OK ) goto fts3_init_out;
  }

  rc = fts3DeclareVtab(p);
  if( rc!=SQLITE_OK ) goto fts3_init_out;

#ifdef SQLITE_TEST
  if( zTestParam ){
    p->nNodeSize = atoi(&zTestParam[5]);
  }
#endif
  *ppVTab = &p->base;

fts3_init_out:
  assert( p || (pTokenizer && rc!=SQLITE_OK) );
  if( rc!=SQLITE_OK ){
    if( p ){
      fts3DisconnectMethod((sqlite3_vtab *)p);
    }else{
      pTokenizer->pModule->xDestroy(pTokenizer);
    }
  }
  return rc;
}

/*
** The xConnect() and xCreate() methods for the virtual table. All the
** work is done in function fts3InitVtab().
*/
static int fts3ConnectMethod(
  sqlite3 *db,                    /* Database connection */
  void *pAux,                     /* Pointer to tokenizer hash table */
  int argc,                       /* Number of elements in argv array */
  const char * const *argv,       /* xCreate/xConnect argument array */
  sqlite3_vtab **ppVtab,          /* OUT: New sqlite3_vtab object */
  char **pzErr                    /* OUT: sqlite3_malloc'd error message */
){
  return fts3InitVtab(0, db, pAux, argc, argv, ppVtab, pzErr);
}
static int fts3CreateMethod(
  sqlite3 *db,                    /* Database connection */
  void *pAux,                     /* Pointer to tokenizer hash table */
  int argc,                       /* Number of elements in argv array */
  const char * const *argv,       /* xCreate/xConnect argument array */
  sqlite3_vtab **ppVtab,          /* OUT: New sqlite3_vtab object */
  char **pzErr                    /* OUT: sqlite3_malloc'd error message */
){
  return fts3InitVtab(1, db, pAux, argc, argv, ppVtab, pzErr);
}

/* 
** Implementation of the xBestIndex method for FTS3 tables. There
** are three possible strategies, in order of preference:
**
**   1. Direct lookup by rowid or docid. 
**   2. Full-text search using a MATCH operator on a non-docid column.
**   3. Linear scan of %_content table.
*/
static int fts3BestIndexMethod(sqlite3_vtab *pVTab, sqlite3_index_info *pInfo){
  Fts3Table *p = (Fts3Table *)pVTab;
  int i;                          /* Iterator variable */
  int iCons = -1;                 /* Index of constraint to use */

  /* By default use a full table scan. This is an expensive option,
  ** so search through the constraints to see if a more efficient 
  ** strategy is possible.
  */
  pInfo->idxNum = FTS3_FULLSCAN_SEARCH;
  pInfo->estimatedCost = 500000;
  for(i=0; i<pInfo->nConstraint; i++){
    struct sqlite3_index_constraint *pCons = &pInfo->aConstraint[i];
    if( pCons->usable==0 ) continue;

    /* A direct lookup on the rowid or docid column. This is the best
    ** strategy in all cases. Assign a cost of 1.0 and return early.
    */
    if( pCons->op==SQLITE_INDEX_CONSTRAINT_EQ 
     && (pCons->iColumn<0 || pCons->iColumn==p->nColumn+1 )
    ){
      pInfo->idxNum = FTS3_DOCID_SEARCH;
      pInfo->estimatedCost = 1.0;
      iCons = i;
      break;
    }

    /* A MATCH constraint. Use a full-text search.
    **
    ** If there is more than one MATCH constraint available, use the first
    ** one encountered. If there is both a MATCH constraint and a direct
    ** rowid/docid lookup, prefer the rowid/docid strategy.
    */
    if( iCons<0 
     && pCons->op==SQLITE_INDEX_CONSTRAINT_MATCH 
     && pCons->iColumn>=0 && pCons->iColumn<=p->nColumn
    ){
      pInfo->idxNum = FTS3_FULLTEXT_SEARCH + pCons->iColumn;
      pInfo->estimatedCost = 2.0;
      iCons = i;
    }
  }

  if( iCons>=0 ){
    pInfo->aConstraintUsage[iCons].argvIndex = 1;
    pInfo->aConstraintUsage[iCons].omit = 1;
  } 
  return SQLITE_OK;
}

/*
** Implementation of xOpen method.
*/
static int fts3OpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){
  sqlite3_vtab_cursor *pCsr;               /* Allocated cursor */

  UNUSED_PARAMETER(pVTab);

  /* Allocate a buffer large enough for an Fts3Cursor structure. If the
  ** allocation succeeds, zero it and return SQLITE_OK. Otherwise, 
  ** if the allocation fails, return SQLITE_NOMEM.
  */
  *ppCsr = pCsr = (sqlite3_vtab_cursor *)sqlite3_malloc(sizeof(Fts3Cursor));
  if( !pCsr ){
    return SQLITE_NOMEM;
  }
  memset(pCsr, 0, sizeof(Fts3Cursor));
  return SQLITE_OK;
}

/****************************************************************/
/****************************************************************/
/****************************************************************/
/****************************************************************/


/*
** Close the cursor.  For additional information see the documentation
** on the xClose method of the virtual table interface.
*/
static int fulltextClose(sqlite3_vtab_cursor *pCursor){
  Fts3Cursor *pCsr = (Fts3Cursor *)pCursor;
  sqlite3_finalize(pCsr->pStmt);
  sqlite3Fts3ExprFree(pCsr->pExpr);
  sqlite3_free(pCsr->aDoclist);
  sqlite3_free(pCsr);
  return SQLITE_OK;
}

static int fts3CursorSeek(Fts3Cursor *pCsr){
  if( pCsr->isRequireSeek ){
    pCsr->isRequireSeek = 0;
    sqlite3_bind_int64(pCsr->pStmt, 1, pCsr->iPrevId);
    if( SQLITE_ROW==sqlite3_step(pCsr->pStmt) ){
      return SQLITE_OK;
    }else{
      int rc;
      pCsr->isEof = 1;
      if( SQLITE_OK==(rc = sqlite3_reset(pCsr->pStmt)) ){
        rc = SQLITE_ERROR;
      }
      return rc;
    }
  }else{
    return SQLITE_OK;
  }
}

static int fts3NextMethod(sqlite3_vtab_cursor *pCursor){
  int rc = SQLITE_OK;             /* Return code */
  Fts3Cursor *pCsr = (Fts3Cursor *)pCursor;

  if( pCsr->aDoclist==0 ){
    if( SQLITE_ROW!=sqlite3_step(pCsr->pStmt) ){
      pCsr->isEof = 1;
      rc = sqlite3_reset(pCsr->pStmt);
    }
  }else if( pCsr->pNextId>=&pCsr->aDoclist[pCsr->nDoclist] ){
    pCsr->isEof = 1;
  }else{
    sqlite3_reset(pCsr->pStmt);
    fts3GetDeltaVarint(&pCsr->pNextId, &pCsr->iPrevId);
    pCsr->isRequireSeek = 1;
  }
  return rc;
}


/*
** The buffer pointed to by argument zNode (size nNode bytes) contains the
** root node of a b-tree segment. The segment is guaranteed to be at least
** one level high (i.e. the root node is not also a leaf). If successful,
** this function locates the leaf node of the segment that may contain the 
** term specified by arguments zTerm and nTerm and writes its block number 
** to *piLeaf.
**
** It is possible that the returned leaf node does not contain the specified
** term. However, if the segment does contain said term, it is stored on
** the identified leaf node. Because this function only inspects interior
** segment nodes (and never loads leaf nodes into memory), it is not possible
** to be sure.
**
** If an error occurs, an error code other than SQLITE_OK is returned.
*/ 
static int fts3SelectLeaf(
  Fts3Table *p,                   /* Virtual table handle */
  const char *zTerm,              /* Term to select leaves for */
  int nTerm,                      /* Size of term zTerm in bytes */
  const char *zNode,              /* Buffer containing segment interior node */
  int nNode,                      /* Size of buffer at zNode */
  sqlite3_int64 *piLeaf           /* Selected leaf node */
){
  int rc = SQLITE_OK;             /* Return code */
  const char *zCsr = zNode;       /* Cursor to iterate through node */
  const char *zEnd = &zCsr[nNode];/* End of interior node buffer */
  char *zBuffer = 0;              /* Buffer to load terms into */
  int nAlloc = 0;                 /* Size of allocated buffer */

  while( 1 ){
    int isFirstTerm = 1;          /* True when processing first term on page */
    int iHeight;                  /* Height of this node in tree */
    sqlite3_int64 iChild;         /* Block id of child node to descend to */
    int nBlock;                   /* Size of child node in bytes */

    zCsr += sqlite3Fts3GetVarint32(zCsr, &iHeight);
    zCsr += sqlite3Fts3GetVarint(zCsr, &iChild);
  
    while( zCsr<zEnd ){
      int cmp;                    /* memcmp() result */
      int nSuffix;                /* Size of term suffix */
      int nPrefix = 0;            /* Size of term prefix */
      int nBuffer;                /* Total term size */
  
      /* Load the next term on the node into zBuffer */
      if( !isFirstTerm ){
        zCsr += sqlite3Fts3GetVarint32(zCsr, &nPrefix);
      }
      isFirstTerm = 0;
      zCsr += sqlite3Fts3GetVarint32(zCsr, &nSuffix);
      if( nPrefix+nSuffix>nAlloc ){
        char *zNew;
        nAlloc = (nPrefix+nSuffix) * 2;
        zNew = (char *)sqlite3_realloc(zBuffer, nAlloc);
        if( !zNew ){
          sqlite3_free(zBuffer);
          return SQLITE_NOMEM;
        }
        zBuffer = zNew;
      }
      memcpy(&zBuffer[nPrefix], zCsr, nSuffix);
      nBuffer = nPrefix + nSuffix;
      zCsr += nSuffix;
  
      /* Compare the term we are searching for with the term just loaded from
      ** the interior node. If the specified term is greater than or equal
      ** to the term from the interior node, then all terms on the sub-tree 
      ** headed by node iChild are smaller than zTerm. No need to search 
      ** iChild.
      **
      ** If the interior node term is larger than the specified term, then
      ** the tree headed by iChild may contain the specified term.
      */
      cmp = memcmp(zTerm, zBuffer, (nBuffer>nTerm ? nTerm : nBuffer));
      if( cmp<0 || (cmp==0 && nBuffer>nTerm) ) break;
      iChild++;
    };

    /* If (iHeight==1), the children of this interior node are leaves. The
    ** specified term may be present on leaf node iChild.
    */
    if( iHeight==1 ){
      *piLeaf = iChild;
      break;
    }

    /* Descend to interior node iChild. */
    rc = sqlite3Fts3ReadBlock(p, iChild, &zCsr, &nBlock);
    if( rc!=SQLITE_OK ) break;
    zEnd = &zCsr[nBlock];
  }
  sqlite3_free(zBuffer);
  return rc;
}

/*
** This function is used to create delta-encoded serialized lists of FTS3 
** varints. Each call to this function appends a single varint to a list.
*/
static void fts3PutDeltaVarint(
  char **pp,                      /* IN/OUT: Output pointer */
  sqlite3_int64 *piPrev,          /* IN/OUT: Previous value written to list */
  sqlite3_int64 iVal              /* Write this value to the list */
){
  assert( iVal-*piPrev > 0 || (*piPrev==0 && iVal==0) );
  *pp += sqlite3Fts3PutVarint(*pp, iVal-*piPrev);
  *piPrev = iVal;
}

static void fts3PoslistCopy(char **pp, char **ppPoslist){
  char *pEnd = *ppPoslist;
  char c = 0;
  while( *pEnd | c ) c = *pEnd++ & 0x80;
  pEnd++;
  if( pp ){
    int n = (int)(pEnd - *ppPoslist);
    char *p = *pp;
    memcpy(p, *ppPoslist, n);
    p += n;
    *pp = p;
  }
  *ppPoslist = pEnd;
}

static void fts3ColumnlistCopy(char **pp, char **ppPoslist){
  char *pEnd = *ppPoslist;
  char c = 0;

  /* A column-list is terminated by either a 0x01 or 0x00. */
  while( 0xFE & (*pEnd | c) ) c = *pEnd++ & 0x80;
  if( pp ){
    int n = (int)(pEnd - *ppPoslist);
    char *p = *pp;
    memcpy(p, *ppPoslist, n);
    p += n;
    *pp = p;
  }
  *ppPoslist = pEnd;
}

/*
** Value used to signify the end of an offset-list. This is safe because
** it is not possible to have a document with 2^31 terms.
*/
#define OFFSET_LIST_END 0x7fffffff

/*
** This function is used to help parse offset-lists. When this function is
** called, *pp may point to the start of the next varint in the offset-list
** being parsed, or it may point to 1 byte past the end of the offset-list
** (in which case **pp will be 0x00 or 0x01).
**
** If *pp points past the end of the current offset list, set *pi to 
** OFFSET_LIST_END and return. Otherwise, read the next varint from *pp,
** increment the current value of *pi by the value read, and set *pp to
** point to the next value before returning.
*/
static void fts3ReadNextPos(
  char **pp,                      /* IN/OUT: Pointer into offset-list buffer */
  sqlite3_int64 *pi               /* IN/OUT: Value read from offset-list */
){
  if( **pp&0xFE ){
    fts3GetDeltaVarint(pp, pi);
    *pi -= 2;
  }else{
    *pi = OFFSET_LIST_END;
  }
}

/*
** If parameter iCol is not 0, write an 0x01 byte followed by the value of
** iCol encoded as a varint to *pp. 
**
** Set *pp to point to the byte just after the last byte written before 
** returning (do not modify it if iCol==0). Return the total number of bytes
** written (0 if iCol==0).
*/
static int fts3PutColNumber(char **pp, int iCol){
  int n = 0;                      /* Number of bytes written */
  if( iCol ){
    char *p = *pp;                /* Output pointer */
    n = 1 + sqlite3Fts3PutVarint(&p[1], iCol);
    *p = 0x01;
    *pp = &p[n];
  }
  return n;
}

/*
**
*/
static void fts3PoslistMerge(
  char **pp,                      /* Output buffer */
  char **pp1,                     /* Left input list */
  char **pp2                      /* Right input list */
){
  char *p = *pp;
  char *p1 = *pp1;
  char *p2 = *pp2;

  while( *p1 || *p2 ){
    int iCol1;
    int iCol2;

    if( *p1==0x01 ) sqlite3Fts3GetVarint32(&p1[1], &iCol1);
    else if( *p1==0x00 ) iCol1 = OFFSET_LIST_END;
    else iCol1 = 0;

    if( *p2==0x01 ) sqlite3Fts3GetVarint32(&p2[1], &iCol2);
    else if( *p2==0x00 ) iCol2 = OFFSET_LIST_END;
    else iCol2 = 0;

    if( iCol1==iCol2 ){
      sqlite3_int64 i1 = 0;
      sqlite3_int64 i2 = 0;
      sqlite3_int64 iPrev = 0;
      int n = fts3PutColNumber(&p, iCol1);
      p1 += n;
      p2 += n;

      /* At this point, both p1 and p2 point to the start of offset-lists.
      ** An offset-list is a list of non-negative delta-encoded varints, each 
      ** incremented by 2 before being stored. Each list is terminated by a 0 
      ** or 1 value (0x00 or 0x01). The following block merges the two lists
      ** and writes the results to buffer p. p is left pointing to the byte
      ** after the list written. No terminator (0x00 or 0x01) is written to
      ** the output.
      */
      fts3GetDeltaVarint(&p1, &i1);
      fts3GetDeltaVarint(&p2, &i2);
      do {
        fts3PutDeltaVarint(&p, &iPrev, (i1<i2) ? i1 : i2); 
        iPrev -= 2;
        if( i1==i2 ){
          fts3ReadNextPos(&p1, &i1);
          fts3ReadNextPos(&p2, &i2);
        }else if( i1<i2 ){
          fts3ReadNextPos(&p1, &i1);
        }else{
          fts3ReadNextPos(&p2, &i2);
        }
      }while( i1!=OFFSET_LIST_END || i2!=OFFSET_LIST_END );
    }else if( iCol1<iCol2 ){
      p1 += fts3PutColNumber(&p, iCol1);
      fts3ColumnlistCopy(&p, &p1);
    }else{
      p2 += fts3PutColNumber(&p, iCol2);
      fts3ColumnlistCopy(&p, &p2);
    }
  }

  *p++ = '\0';
  *pp = p;
  *pp1 = p1 + 1;
  *pp2 = p2 + 1;
}

/*
** nToken==1 searches for adjacent positions.
*/
static int fts3PoslistPhraseMerge(
  char **pp,                      /* Output buffer */
  int nToken,                     /* Maximum difference in token positions */
  int isSaveLeft,                 /* Save the left position */
  char **pp1,                     /* Left input list */
  char **pp2                      /* Right input list */
){
  char *p = (pp ? *pp : 0);
  char *p1 = *pp1;
  char *p2 = *pp2;

  int iCol1 = 0;
  int iCol2 = 0;
  assert( *p1!=0 && *p2!=0 );
  if( *p1==0x01 ){ 
    p1++;
    p1 += sqlite3Fts3GetVarint32(p1, &iCol1);
  }
  if( *p2==0x01 ){ 
    p2++;
    p2 += sqlite3Fts3GetVarint32(p2, &iCol2);
  }

  while( 1 ){
    if( iCol1==iCol2 ){
      char *pSave = p;
      sqlite3_int64 iPrev = 0;
      sqlite3_int64 iPos1 = 0;
      sqlite3_int64 iPos2 = 0;

      if( pp && iCol1 ){
        *p++ = 0x01;
        p += sqlite3Fts3PutVarint(p, iCol1);
      }

      assert( *p1!=0x00 && *p2!=0x00 && *p1!=0x01 && *p2!=0x01 );
      fts3GetDeltaVarint(&p1, &iPos1); iPos1 -= 2;
      fts3GetDeltaVarint(&p2, &iPos2); iPos2 -= 2;

      while( 1 ){
        if( iPos2>iPos1 && iPos2<=iPos1+nToken ){
          sqlite3_int64 iSave;
          if( !pp ){
            fts3PoslistCopy(0, &p2);
            fts3PoslistCopy(0, &p1);
            *pp1 = p1;
            *pp2 = p2;
            return 1;
          }
          iSave = isSaveLeft ? iPos1 : iPos2;
          fts3PutDeltaVarint(&p, &iPrev, iSave+2); iPrev -= 2;
          pSave = 0;
        }
        if( (!isSaveLeft && iPos2<=(iPos1+nToken)) || iPos2<=iPos1 ){
          if( (*p2&0xFE)==0 ) break;
          fts3GetDeltaVarint(&p2, &iPos2); iPos2 -= 2;
        }else{
          if( (*p1&0xFE)==0 ) break;
          fts3GetDeltaVarint(&p1, &iPos1); iPos1 -= 2;
        }

      }
      if( pSave && pp ){
        p = pSave;
      }

      fts3ColumnlistCopy(0, &p1);
      fts3ColumnlistCopy(0, &p2);
      assert( (*p1&0xFE)==0 && (*p2&0xFE)==0 );
      if( 0==*p1 || 0==*p2 ) break;

      p1++;
      p1 += sqlite3Fts3GetVarint32(p1, &iCol1);
      p2++;
      p2 += sqlite3Fts3GetVarint32(p2, &iCol2);
    }

    /* Advance pointer p1 or p2 (whichever corresponds to the smaller of
    ** iCol1 and iCol2) so that it points to either the 0x00 that marks the
    ** end of the position list, or the 0x01 that precedes the next 
    ** column-number in the position list. 
    */
    else if( iCol1<iCol2 ){
      fts3ColumnlistCopy(0, &p1);
      if( 0==*p1 ) break;
      p1++;
      p1 += sqlite3Fts3GetVarint32(p1, &iCol1);
    }else{
      fts3ColumnlistCopy(0, &p2);
      if( 0==*p2 ) break;
      p2++;
      p2 += sqlite3Fts3GetVarint32(p2, &iCol2);
    }
  }

  fts3PoslistCopy(0, &p2);
  fts3PoslistCopy(0, &p1);
  *pp1 = p1;
  *pp2 = p2;
  if( !pp || *pp==p ){
    return 0;
  }
  *p++ = 0x00;
  *pp = p;
  return 1;
}

/*
** Merge two position-lists as required by the NEAR operator.
*/
static int fts3PoslistNearMerge(
  char **pp,                      /* Output buffer */
  char *aTmp,                     /* Temporary buffer space */
  int nRight,                     /* Maximum difference in token positions */
  int nLeft,                      /* Maximum difference in token positions */
  char **pp1,                     /* IN/OUT: Left input list */
  char **pp2                      /* IN/OUT: Right input list */
){
  char *p1 = *pp1;
  char *p2 = *pp2;

  if( !pp ){
    if( fts3PoslistPhraseMerge(0, nRight, 0, pp1, pp2) ) return 1;
    *pp1 = p1;
    *pp2 = p2;
    return fts3PoslistPhraseMerge(0, nLeft, 0, pp2, pp1);
  }else{
    char *pTmp1 = aTmp;
    char *pTmp2;
    char *aTmp2;
    int res = 1;

    fts3PoslistPhraseMerge(&pTmp1, nRight, 0, pp1, pp2);
    aTmp2 = pTmp2 = pTmp1;
    *pp1 = p1;
    *pp2 = p2;
    fts3PoslistPhraseMerge(&pTmp2, nLeft, 1, pp2, pp1);
    if( pTmp1!=aTmp && pTmp2!=aTmp2 ){
      fts3PoslistMerge(pp, &aTmp, &aTmp2);
    }else if( pTmp1!=aTmp ){
      fts3PoslistCopy(pp, &aTmp);
    }else if( pTmp2!=aTmp2 ){
      fts3PoslistCopy(pp, &aTmp2);
    }else{
      res = 0;
    }

    return res;
  }
}

/*
** Values that may be used as the first parameter to fts3DoclistMerge().
*/
#define MERGE_NOT        2        /* D + D -> D */
#define MERGE_AND        3        /* D + D -> D */
#define MERGE_OR         4        /* D + D -> D */
#define MERGE_POS_OR     5        /* P + P -> P */
#define MERGE_PHRASE     6        /* P + P -> D */
#define MERGE_POS_PHRASE 7        /* P + P -> P */
#define MERGE_NEAR       8        /* P + P -> D */
#define MERGE_POS_NEAR   9        /* P + P -> P */

/*
** Merge the two doclists passed in buffer a1 (size n1 bytes) and a2
** (size n2 bytes). The output is written to pre-allocated buffer aBuffer,
** which is guaranteed to be large enough to hold the results. The number
** of bytes written to aBuffer is stored in *pnBuffer before returning.
**
** If successful, SQLITE_OK is returned. Otherwise, if a malloc error
** occurs while allocating a temporary buffer as part of the merge operation,
** SQLITE_NOMEM is returned.
*/
static int fts3DoclistMerge(
  int mergetype,                  /* One of the MERGE_XXX constants */
  int nParam1,                    /* Used by MERGE_NEAR and MERGE_POS_NEAR */
  int nParam2,                    /* Used by MERGE_NEAR and MERGE_POS_NEAR */
  char *aBuffer,                  /* Pre-allocated output buffer */
  int *pnBuffer,                  /* OUT: Bytes written to aBuffer */
  char *a1,                       /* Buffer containing first doclist */
  int n1,                         /* Size of buffer a1 */
  char *a2,                       /* Buffer containing second doclist */
  int n2                          /* Size of buffer a2 */
){
  sqlite3_int64 i1 = 0;
  sqlite3_int64 i2 = 0;
  sqlite3_int64 iPrev = 0;

  char *p = aBuffer;
  char *p1 = a1;
  char *p2 = a2;
  char *pEnd1 = &a1[n1];
  char *pEnd2 = &a2[n2];

  assert( mergetype==MERGE_OR     || mergetype==MERGE_POS_OR 
       || mergetype==MERGE_AND    || mergetype==MERGE_NOT
       || mergetype==MERGE_PHRASE || mergetype==MERGE_POS_PHRASE
       || mergetype==MERGE_NEAR   || mergetype==MERGE_POS_NEAR
  );

  if( !aBuffer ){
    return SQLITE_NOMEM;
  }
  if( n1==0 && n2==0 ){
    *pnBuffer = 0;
    return SQLITE_OK;
  }

  /* Read the first docid from each doclist */
  fts3GetDeltaVarint2(&p1, pEnd1, &i1);
  fts3GetDeltaVarint2(&p2, pEnd2, &i2);

  switch( mergetype ){
    case MERGE_OR:
    case MERGE_POS_OR:
      while( p1 || p2 ){
        if( p2 && p1 && i1==i2 ){
          fts3PutDeltaVarint(&p, &iPrev, i1);
          if( mergetype==MERGE_POS_OR ) fts3PoslistMerge(&p, &p1, &p2);
          fts3GetDeltaVarint2(&p1, pEnd1, &i1);
          fts3GetDeltaVarint2(&p2, pEnd2, &i2);
        }else if( !p2 || (p1 && i1<i2) ){
          fts3PutDeltaVarint(&p, &iPrev, i1);
          if( mergetype==MERGE_POS_OR ) fts3PoslistCopy(&p, &p1);
          fts3GetDeltaVarint2(&p1, pEnd1, &i1);
        }else{
          fts3PutDeltaVarint(&p, &iPrev, i2);
          if( mergetype==MERGE_POS_OR ) fts3PoslistCopy(&p, &p2);
          fts3GetDeltaVarint2(&p2, pEnd2, &i2);
        }
      }
      break;

    case MERGE_AND:
      while( p1 && p2 ){
        if( i1==i2 ){
          fts3PutDeltaVarint(&p, &iPrev, i1);
          fts3GetDeltaVarint2(&p1, pEnd1, &i1);
          fts3GetDeltaVarint2(&p2, pEnd2, &i2);
        }else if( i1<i2 ){
          fts3GetDeltaVarint2(&p1, pEnd1, &i1);
        }else{
          fts3GetDeltaVarint2(&p2, pEnd2, &i2);
        }
      }
      break;

    case MERGE_NOT:
      while( p1 ){
        if( p2 && i1==i2 ){
          fts3GetDeltaVarint2(&p1, pEnd1, &i1);
          fts3GetDeltaVarint2(&p2, pEnd2, &i2);
        }else if( !p2 || i1<i2 ){
          fts3PutDeltaVarint(&p, &iPrev, i1);
          fts3GetDeltaVarint2(&p1, pEnd1, &i1);
        }else{
          fts3GetDeltaVarint2(&p2, pEnd2, &i2);
        }
      }
      break;

    case MERGE_POS_PHRASE:
    case MERGE_PHRASE: {
      char **ppPos = (mergetype==MERGE_PHRASE ? 0 : &p);
      while( p1 && p2 ){
        if( i1==i2 ){
          char *pSave = p;
          sqlite3_int64 iPrevSave = iPrev;
          fts3PutDeltaVarint(&p, &iPrev, i1);
          if( 0==fts3PoslistPhraseMerge(ppPos, 1, 0, &p1, &p2) ){
            p = pSave;
            iPrev = iPrevSave;
          }
          fts3GetDeltaVarint2(&p1, pEnd1, &i1);
          fts3GetDeltaVarint2(&p2, pEnd2, &i2);
        }else if( i1<i2 ){
          fts3PoslistCopy(0, &p1);
          fts3GetDeltaVarint2(&p1, pEnd1, &i1);
        }else{
          fts3PoslistCopy(0, &p2);
          fts3GetDeltaVarint2(&p2, pEnd2, &i2);
        }
      }
      break;
    }

    default: assert( mergetype==MERGE_POS_NEAR || mergetype==MERGE_NEAR ); {
      char *aTmp = 0;
      char **ppPos = 0;
      if( mergetype==MERGE_POS_NEAR ){
        ppPos = &p;
        aTmp = sqlite3_malloc(2*(n1+n2));
        if( !aTmp ){
          return SQLITE_NOMEM;
        }
      }

      while( p1 && p2 ){
        if( i1==i2 ){
          char *pSave = p;
          sqlite3_int64 iPrevSave = iPrev;
          fts3PutDeltaVarint(&p, &iPrev, i1);

          if( !fts3PoslistNearMerge(ppPos, aTmp, nParam1, nParam2, &p1, &p2) ){
            iPrev = iPrevSave;
            p = pSave;
          }

          fts3GetDeltaVarint2(&p1, pEnd1, &i1);
          fts3GetDeltaVarint2(&p2, pEnd2, &i2);
        }else if( i1<i2 ){
          fts3PoslistCopy(0, &p1);
          fts3GetDeltaVarint2(&p1, pEnd1, &i1);
        }else{
          fts3PoslistCopy(0, &p2);
          fts3GetDeltaVarint2(&p2, pEnd2, &i2);
        }
      }
      sqlite3_free(aTmp);
      break;
    }
  }

  *pnBuffer = (int)(p-aBuffer);
  return SQLITE_OK;
}

/* 
** A pointer to an instance of this structure is used as the context 
** argument to sqlite3Fts3SegReaderIterate()
*/
typedef struct TermSelect TermSelect;
struct TermSelect {
  int isReqPos;
  char *aOutput;                  /* Malloc'd output buffer */
  int nOutput;                    /* Size of output in bytes */
};

/*
** This function is used as the sqlite3Fts3SegReaderIterate() callback when
** querying the full-text index for a doclist associated with a term or
** term-prefix.
*/
static int fts3TermSelectCb(
  Fts3Table *p,                   /* Virtual table object */
  void *pContext,                 /* Pointer to TermSelect structure */
  char *zTerm,
  int nTerm,
  char *aDoclist,
  int nDoclist
){
  TermSelect *pTS = (TermSelect *)pContext;
  int nNew = pTS->nOutput + nDoclist;
  char *aNew = sqlite3_malloc(nNew);

  UNUSED_PARAMETER(p);
  UNUSED_PARAMETER(zTerm);
  UNUSED_PARAMETER(nTerm);

  if( !aNew ){
    return SQLITE_NOMEM;
  }

  if( pTS->nOutput==0 ){
    /* If this is the first term selected, copy the doclist to the output
    ** buffer using memcpy(). TODO: Add a way to transfer control of the
    ** aDoclist buffer from the caller so as to avoid the memcpy().
    */
    memcpy(aNew, aDoclist, nDoclist);
  }else{
    /* The output buffer is not empty. Merge doclist aDoclist with the
    ** existing output. This can only happen with prefix-searches (as
    ** searches for exact terms return exactly one doclist).
    */
    int mergetype = (pTS->isReqPos ? MERGE_POS_OR : MERGE_OR);
    fts3DoclistMerge(mergetype, 0, 0,
        aNew, &nNew, pTS->aOutput, pTS->nOutput, aDoclist, nDoclist
    );
  }

  sqlite3_free(pTS->aOutput);
  pTS->aOutput = aNew;
  pTS->nOutput = nNew;

  return SQLITE_OK;
}

/*
** This function retreives the doclist for the specified term (or term
** prefix) from the database. 
**
** The returned doclist may be in one of two formats, depending on the 
** value of parameter isReqPos. If isReqPos is zero, then the doclist is
** a sorted list of delta-compressed docids. If isReqPos is non-zero, 
** then the returned list is in the same format as is stored in the
** database without the found length specifier at the start of on-disk
** doclists.
*/
static int fts3TermSelect(
  Fts3Table *p,                   /* Virtual table handle */
  int iColumn,                    /* Column to query (or -ve for all columns) */
  const char *zTerm,              /* Term to query for */
  int nTerm,                      /* Size of zTerm in bytes */
  int isPrefix,                   /* True for a prefix search */
  int isReqPos,                   /* True to include position lists in output */
  int *pnOut,                     /* OUT: Size of buffer at *ppOut */
  char **ppOut                    /* OUT: Malloced result buffer */
){
  int i;
  TermSelect tsc;
  Fts3SegFilter filter;           /* Segment term filter configuration */
  Fts3SegReader **apSegment = 0;  /* Array of segments to read data from */
  int nSegment = 0;               /* Size of apSegment array */
  int nAlloc = 0;                 /* Allocated size of segment array */
  int rc;                         /* Return code */
  sqlite3_stmt *pStmt;            /* SQL statement to scan %_segdir table */
  int iAge = 0;                   /* Used to assign ages to segments */

  /* Loop through the entire %_segdir table. For each segment, create a
  ** Fts3SegReader to iterate through the subset of the segment leaves
  ** that may contain a term that matches zTerm/nTerm. For non-prefix
  ** searches, this is always a single leaf. For prefix searches, this
  ** may be a contiguous block of leaves.
  **
  ** The code in this loop does not actually load any leaves into memory
  ** (unless the root node happens to be a leaf). It simply examines the
  ** b-tree structure to determine which leaves need to be inspected.
  */
  rc = sqlite3Fts3AllSegdirs(p, &pStmt);
  while( rc==SQLITE_OK && SQLITE_ROW==(rc = sqlite3_step(pStmt)) ){
    Fts3SegReader *pNew = 0;
    int nRoot = sqlite3_column_bytes(pStmt, 4);
    char const *zRoot = sqlite3_column_blob(pStmt, 4);
    if( sqlite3_column_int64(pStmt, 1)==0 ){
      /* The entire segment is stored on the root node (which must be a
      ** leaf). Do not bother inspecting any data in this case, just
      ** create a Fts3SegReader to scan the single leaf. 
      */
      rc = sqlite3Fts3SegReaderNew(p, iAge, 0, 0, 0, zRoot, nRoot, &pNew);
    }else{
      int rc2;                    /* Return value of sqlite3Fts3ReadBlock() */
      sqlite3_int64 i1;           /* Blockid of leaf that may contain zTerm */
      rc = fts3SelectLeaf(p, zTerm, nTerm, zRoot, nRoot, &i1);
      if( rc==SQLITE_OK ){
        sqlite3_int64 i2 = sqlite3_column_int64(pStmt, 2);
        rc = sqlite3Fts3SegReaderNew(p, iAge, i1, i2, 0, 0, 0, &pNew);
      }

      /* The following call to ReadBlock() serves to reset the SQL statement
      ** used to retrieve blocks of data from the %_segments table. If it is
      ** not reset here, then it may remain classified as an active statement 
      ** by SQLite, which may lead to "DROP TABLE" or "DETACH" commands 
      ** failing.
      */ 
      rc2 = sqlite3Fts3ReadBlock(p, 0, 0, 0);
      if( rc==SQLITE_OK ){
        rc = rc2;
      }
    }
    iAge++;

    /* If a new Fts3SegReader was allocated, add it to the apSegment array. */
    assert( pNew!=0 || rc!=SQLITE_OK );
    if( pNew ){
      if( nSegment==nAlloc ){
        Fts3SegReader **pArray;
        nAlloc += 16;
        pArray = (Fts3SegReader **)sqlite3_realloc(
            apSegment, nAlloc*sizeof(Fts3SegReader *)
        );
        if( !pArray ){
          sqlite3Fts3SegReaderFree(p, pNew);
          rc = SQLITE_NOMEM;
          goto finished;
        }
        apSegment = pArray;
      }
      apSegment[nSegment++] = pNew;
    }
  }
  if( rc!=SQLITE_DONE ){
    assert( rc!=SQLITE_OK );
    goto finished;
  }

  memset(&tsc, 0, sizeof(TermSelect));
  tsc.isReqPos = isReqPos;

  filter.flags = FTS3_SEGMENT_IGNORE_EMPTY 
        | (isPrefix ? FTS3_SEGMENT_PREFIX : 0)
        | (isReqPos ? FTS3_SEGMENT_REQUIRE_POS : 0)
        | (iColumn<p->nColumn ? FTS3_SEGMENT_COLUMN_FILTER : 0);
  filter.iCol = iColumn;
  filter.zTerm = zTerm;
  filter.nTerm = nTerm;

  rc = sqlite3Fts3SegReaderIterate(p, apSegment, nSegment, &filter,
      fts3TermSelectCb, (void *)&tsc
  );

  if( rc==SQLITE_OK ){
    *ppOut = tsc.aOutput;
    *pnOut = tsc.nOutput;
  }else{
    sqlite3_free(tsc.aOutput);
  }

finished:
  sqlite3_reset(pStmt);
  for(i=0; i<nSegment; i++){
    sqlite3Fts3SegReaderFree(p, apSegment[i]);
  }
  sqlite3_free(apSegment);
  return rc;
}


/* 
** Return a DocList corresponding to the phrase *pPhrase.
*/
static int fts3PhraseSelect(
  Fts3Table *p,                   /* Virtual table handle */
  Fts3Phrase *pPhrase,            /* Phrase to return a doclist for */
  int isReqPos,                   /* True if output should contain positions */
  char **paOut,                   /* OUT: Pointer to malloc'd result buffer */
  int *pnOut                      /* OUT: Size of buffer at *paOut */
){
  char *pOut = 0;
  int nOut = 0;
  int rc = SQLITE_OK;
  int ii;
  int iCol = pPhrase->iColumn;
  int isTermPos = (pPhrase->nToken>1 || isReqPos);

  assert( p->nPendingData==0 );

  for(ii=0; ii<pPhrase->nToken; ii++){
    struct PhraseToken *pTok = &pPhrase->aToken[ii];
    char *z = pTok->z;            /* Next token of the phrase */
    int n = pTok->n;              /* Size of z in bytes */
    int isPrefix = pTok->isPrefix;/* True if token is a prefix */
    char *pList;                  /* Pointer to token doclist */
    int nList;                    /* Size of buffer at pList */

    rc = fts3TermSelect(p, iCol, z, n, isPrefix, isTermPos, &nList, &pList);
    if( rc!=SQLITE_OK ) break;

    if( ii==0 ){
      pOut = pList;
      nOut = nList;
    }else{
      /* Merge the new term list and the current output. If this is the
      ** last term in the phrase, and positions are not required in the
      ** output of this function, the positions can be dropped as part
      ** of this merge. Either way, the result of this merge will be
      ** smaller than nList bytes. The code in fts3DoclistMerge() is written
      ** so that it is safe to use pList as the output as well as an input
      ** in this case.
      */
      int mergetype = MERGE_POS_PHRASE;
      if( ii==pPhrase->nToken-1 && !isReqPos ){
        mergetype = MERGE_PHRASE;
      }
      fts3DoclistMerge(mergetype, 0, 0, pList, &nOut, pOut, nOut, pList, nList);
      sqlite3_free(pOut);
      pOut = pList;
    }
  }

  if( rc==SQLITE_OK ){
    *paOut = pOut;
    *pnOut = nOut;
  }else{
    sqlite3_free(pOut);
  }
  return rc;
}

/*
** Evaluate the full-text expression pExpr against fts3 table pTab. Store
** the resulting doclist in *paOut and *pnOut.
*/
static int evalFts3Expr(
  Fts3Table *p,                   /* Virtual table handle */
  Fts3Expr *pExpr,                /* Parsed fts3 expression */
  char **paOut,                   /* OUT: Pointer to malloc'd result buffer */
  int *pnOut                      /* OUT: Size of buffer at *paOut */
){
  int rc = SQLITE_OK;             /* Return code */


  /* Zero the output parameters. */
  *paOut = 0;
  *pnOut = 0;


  if( pExpr ){
    if( pExpr->eType==FTSQUERY_PHRASE ){

      int isReqPos = (pExpr->pParent && pExpr->pParent->eType==FTSQUERY_NEAR);



      rc = fts3PhraseSelect(p, pExpr->pPhrase, isReqPos, paOut, pnOut);
    }else{
      char *aLeft;
      char *aRight;
      int nLeft;
      int nRight;



      if( SQLITE_OK==(rc = evalFts3Expr(p, pExpr->pRight, &aRight, &nRight))
       && SQLITE_OK==(rc = evalFts3Expr(p, pExpr->pLeft, &aLeft, &nLeft))
      ){
        assert( pExpr->eType==FTSQUERY_NEAR || pExpr->eType==FTSQUERY_OR     
            || pExpr->eType==FTSQUERY_AND  || pExpr->eType==FTSQUERY_NOT
        );
        switch( pExpr->eType ){
          case FTSQUERY_NEAR: {

            Fts3Expr *pLeft;
            Fts3Expr *pRight;
            int mergetype = MERGE_NEAR;
            int nParam1;
            int nParam2;
            char *aBuffer;
           
            if( pExpr->pParent && pExpr->pParent->eType==FTSQUERY_NEAR ){

              mergetype = MERGE_POS_NEAR;
            }
            pLeft = pExpr->pLeft;
            while( pLeft->eType==FTSQUERY_NEAR ){ 
              pLeft=pLeft->pRight;
            }
            pRight = pExpr->pRight;
            assert( pRight->eType==FTSQUERY_PHRASE );
            assert( pLeft->eType==FTSQUERY_PHRASE );



            nParam1 = pExpr->nNear+1;
            nParam2 = nParam1+pLeft->pPhrase->nToken+pRight->pPhrase->nToken-2;
            aBuffer = sqlite3_malloc(nLeft+nRight+1);
            rc = fts3DoclistMerge(mergetype, nParam1, nParam2, aBuffer,
                pnOut, aLeft, nLeft, aRight, nRight
            );

            if( rc!=SQLITE_OK ){
              sqlite3_free(aBuffer);
            }else{
              *paOut = aBuffer;
            }


            sqlite3_free(aLeft);
            break;
          }

          case FTSQUERY_OR: {

            /* Allocate a buffer for the output. The maximum size is the
            ** sum of the sizes of the two input buffers. The +1 term is
            ** so that a buffer of zero bytes is never allocated - this can
            ** cause fts3DoclistMerge() to incorrectly return SQLITE_NOMEM.
            */
            char *aBuffer = sqlite3_malloc(nRight+nLeft+1);
            rc = fts3DoclistMerge(MERGE_OR, 0, 0, aBuffer, pnOut,
                aLeft, nLeft, aRight, nRight
            );
            *paOut = aBuffer;
            sqlite3_free(aLeft);
            break;
          }



          default: {
            assert( FTSQUERY_NOT==MERGE_NOT && FTSQUERY_AND==MERGE_AND );
            fts3DoclistMerge(pExpr->eType, 0, 0, aLeft, pnOut,
                aLeft, nLeft, aRight, nRight
            );
            *paOut = aLeft;
            break;
          }
        }
      }
      sqlite3_free(aRight);

    }
  }

  return rc;
}















































/*
** This is the xFilter interface for the virtual table.  See
** the virtual table xFilter method documentation for additional
** information.
**
** If idxNum==FTS3_FULLSCAN_SEARCH then do a full table scan against
** the %_content table.
**
** If idxNum==FTS3_DOCID_SEARCH then do a docid lookup for a single entry
** in the %_content table.
**
** If idxNum>=FTS3_FULLTEXT_SEARCH then use the full text index.  The
** column on the left-hand side of the MATCH operator is column
** number idxNum-FTS3_FULLTEXT_SEARCH, 0 indexed.  argv[0] is the right-hand
** side of the MATCH operator.
*/
/* TODO(shess) Upgrade the cursor initialization and destruction to
** account for fts3FilterMethod() being called multiple times on the
** same cursor. The current solution is very fragile. Apply fix to
** fts3 as appropriate.
*/
static int fts3FilterMethod(
  sqlite3_vtab_cursor *pCursor,   /* The cursor used for this query */
  int idxNum,                     /* Strategy index */
  const char *idxStr,             /* Unused */
  int nVal,                       /* Number of elements in apVal */
  sqlite3_value **apVal           /* Arguments for the indexing scheme */
){
  const char *azSql[] = {
    "SELECT * FROM %Q.'%q_content' WHERE docid = ?", /* non-full-table-scan */
    "SELECT * FROM %Q.'%q_content'",                 /* full-table-scan */
  };
  int rc;                         /* Return code */
  char *zSql;                     /* SQL statement used to access %_content */
  Fts3Table *p = (Fts3Table *)pCursor->pVtab;
  Fts3Cursor *pCsr = (Fts3Cursor *)pCursor;

  UNUSED_PARAMETER(idxStr);
  UNUSED_PARAMETER(nVal);

  assert( idxNum>=0 && idxNum<=(FTS3_FULLTEXT_SEARCH+p->nColumn) );
  assert( nVal==0 || nVal==1 );
  assert( (nVal==0)==(idxNum==FTS3_FULLSCAN_SEARCH) );

  /* In case the cursor has been used before, clear it now. */
  sqlite3_finalize(pCsr->pStmt);
  sqlite3_free(pCsr->aDoclist);
  memset(&pCursor[1], 0, sizeof(Fts3Cursor)-sizeof(sqlite3_vtab_cursor));

  /* Compile a SELECT statement for this cursor. For a full-table-scan, the
  ** statement loops through all rows of the %_content table. For a
  ** full-text query or docid lookup, the statement retrieves a single
  ** row by docid.
  */
  zSql = sqlite3_mprintf(azSql[idxNum==FTS3_FULLSCAN_SEARCH], p->zDb, p->zName);
  if( !zSql ){
    rc = SQLITE_NOMEM;
  }else{
    rc = sqlite3_prepare_v2(p->db, zSql, -1, &pCsr->pStmt, 0);
    sqlite3_free(zSql);
  }
  if( rc!=SQLITE_OK ) return rc;
  pCsr->eSearch = (i16)idxNum;

  if( idxNum==FTS3_DOCID_SEARCH ){
    rc = sqlite3_bind_value(pCsr->pStmt, 1, apVal[0]);
  }else if( idxNum!=FTS3_FULLSCAN_SEARCH ){
    int iCol = idxNum-FTS3_FULLTEXT_SEARCH;
    const char *zQuery = (const char *)sqlite3_value_text(apVal[0]);

    if( zQuery==0 && sqlite3_value_type(apVal[0])!=SQLITE_NULL ){
      return SQLITE_NOMEM;
    }
    rc = sqlite3Fts3PendingTermsFlush(p);
    if( rc!=SQLITE_OK ) return rc;

    rc = sqlite3Fts3ExprParse(p->pTokenizer, p->azColumn, p->nColumn, 
        iCol, zQuery, -1, &pCsr->pExpr
    );
    if( rc!=SQLITE_OK ) return rc;

    rc = evalFts3Expr(p, pCsr->pExpr, &pCsr->aDoclist, &pCsr->nDoclist);
    pCsr->pNextId = pCsr->aDoclist;
    pCsr->iPrevId = 0;
  }

  if( rc!=SQLITE_OK ) return rc;
  return fts3NextMethod(pCursor);
}

/* 
** This is the xEof method of the virtual table. SQLite calls this 
** routine to find out if it has reached the end of a result set.
*/
static int fts3EofMethod(sqlite3_vtab_cursor *pCursor){
  return ((Fts3Cursor *)pCursor)->isEof;
}

/* 
** This is the xRowid method. The SQLite core calls this routine to
** retrieve the rowid for the current row of the result set. fts3
** exposes %_content.docid as the rowid for the virtual table. The
** rowid should be written to *pRowid.
*/
static int fts3RowidMethod(sqlite3_vtab_cursor *pCursor, sqlite_int64 *pRowid){
  Fts3Cursor *pCsr = (Fts3Cursor *) pCursor;
  if( pCsr->aDoclist ){
    *pRowid = pCsr->iPrevId;
  }else{
    *pRowid = sqlite3_column_int64(pCsr->pStmt, 0);
  }
  return SQLITE_OK;
}

/* 
** This is the xColumn method, called by SQLite to request a value from
** the row that the supplied cursor currently points to.
*/
static int fts3ColumnMethod(
  sqlite3_vtab_cursor *pCursor,   /* Cursor to retrieve value from */
  sqlite3_context *pContext,      /* Context for sqlite3_result_xxx() calls */
  int iCol                        /* Index of column to read value from */
){
  int rc;                         /* Return Code */
  Fts3Cursor *pCsr = (Fts3Cursor *) pCursor;
  Fts3Table *p = (Fts3Table *)pCursor->pVtab;

  /* The column value supplied by SQLite must be in range. */
  assert( iCol>=0 && iCol<=p->nColumn+1 );

  rc = fts3CursorSeek(pCsr);
  if( rc==SQLITE_OK ){
    if( iCol==p->nColumn+1 ){
      /* This call is a request for the "docid" column. Since "docid" is an 
      ** alias for "rowid", use the xRowid() method to obtain the value.
      */
      sqlite3_int64 iRowid;
      rc = fts3RowidMethod(pCursor, &iRowid);
      sqlite3_result_int64(pContext, iRowid);
    }else if( iCol==p->nColumn ){
      /* The extra column whose name is the same as the table.
      ** Return a blob which is a pointer to the cursor.
      */
      sqlite3_result_blob(pContext, &pCsr, sizeof(pCsr), SQLITE_TRANSIENT);
    }else{
      sqlite3_result_value(pContext, sqlite3_column_value(pCsr->pStmt, iCol+1));
    }
  }
  return rc;
}

/* 
** This function is the implementation of the xUpdate callback used by 
** FTS3 virtual tables. It is invoked by SQLite each time a row is to be
** inserted, updated or deleted.
*/
static int fts3UpdateMethod(
  sqlite3_vtab *pVtab,            /* Virtual table handle */
  int nArg,                       /* Size of argument array */
  sqlite3_value **apVal,          /* Array of arguments */
  sqlite_int64 *pRowid            /* OUT: The affected (or effected) rowid */
){
  return sqlite3Fts3UpdateMethod(pVtab, nArg, apVal, pRowid);
}

/*
** Implementation of xSync() method. Flush the contents of the pending-terms
** hash-table to the database.
*/
static int fts3SyncMethod(sqlite3_vtab *pVtab){
  return sqlite3Fts3PendingTermsFlush((Fts3Table *)pVtab);
}

/*
** Implementation of xBegin() method. This is a no-op.
*/
static int fts3BeginMethod(sqlite3_vtab *pVtab){
  UNUSED_PARAMETER(pVtab);
  assert( ((Fts3Table *)pVtab)->nPendingData==0 );
  return SQLITE_OK;
}

/*
** Implementation of xCommit() method. This is a no-op. The contents of
** the pending-terms hash-table have already been flushed into the database
** by fts3SyncMethod().
*/
static int fts3CommitMethod(sqlite3_vtab *pVtab){
  UNUSED_PARAMETER(pVtab);
  assert( ((Fts3Table *)pVtab)->nPendingData==0 );
  return SQLITE_OK;
}

/*
** Implementation of xRollback(). Discard the contents of the pending-terms
** hash-table. Any changes made to the database are reverted by SQLite.
*/
static int fts3RollbackMethod(sqlite3_vtab *pVtab){
  sqlite3Fts3PendingTermsClear((Fts3Table *)pVtab);
  return SQLITE_OK;
}

/*
** Helper function used by the implementation of the overloaded snippet(),
** offsets() and optimize() SQL functions.
**
** If the value passed as the third argument is a blob of size
** sizeof(Fts3Cursor*), then the blob contents are copied to the 
** output variable *ppCsr and SQLITE_OK is returned. Otherwise, an error
** message is written to context pContext and SQLITE_ERROR returned. The
** string passed via zFunc is used as part of the error message.
*/
static int fts3FunctionArg(
  sqlite3_context *pContext,      /* SQL function call context */
  const char *zFunc,              /* Function name */
  sqlite3_value *pVal,            /* argv[0] passed to function */
  Fts3Cursor **ppCsr         /* OUT: Store cursor handle here */
){
  Fts3Cursor *pRet;
  if( sqlite3_value_type(pVal)!=SQLITE_BLOB 
   || sqlite3_value_bytes(pVal)!=sizeof(Fts3Cursor *)
  ){
    char *zErr = sqlite3_mprintf("illegal first argument to %s", zFunc);
    sqlite3_result_error(pContext, zErr, -1);
    sqlite3_free(zErr);
    return SQLITE_ERROR;
  }
  memcpy(&pRet, sqlite3_value_blob(pVal), sizeof(Fts3Cursor *));
  *ppCsr = pRet;
  return SQLITE_OK;
}

/*
** Implementation of the snippet() function for FTS3
*/
static void fts3SnippetFunc(
  sqlite3_context *pContext,      /* SQLite function call context */
  int nVal,                       /* Size of apVal[] array */
  sqlite3_value **apVal           /* Array of arguments */
){
  Fts3Cursor *pCsr;               /* Cursor handle passed through apVal[0] */
  const char *zStart = "<b>";
  const char *zEnd = "</b>";
  const char *zEllipsis = "<b>...</b>";

  /* There must be at least one argument passed to this function (otherwise
  ** the non-overloaded version would have been called instead of this one).
  */
  assert( nVal>=1 );

  if( nVal>4 ){
    sqlite3_result_error(pContext, 
        "wrong number of arguments to function snippet()", -1);
    return;
  }
  if( fts3FunctionArg(pContext, "snippet", apVal[0], &pCsr) ) return;

  switch( nVal ){
    case 4: zEllipsis = (const char*)sqlite3_value_text(apVal[3]);
    case 3: zEnd = (const char*)sqlite3_value_text(apVal[2]);
    case 2: zStart = (const char*)sqlite3_value_text(apVal[1]);
  }

  sqlite3Fts3Snippet(pContext, pCsr, zStart, zEnd, zEllipsis);






























































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































}

/*
** Implementation of the offsets() function for FTS3
*/
static void fts3OffsetsFunc(
  sqlite3_context *pContext,      /* SQLite function call context */
  int nVal,                       /* Size of argument array */
  sqlite3_value **apVal           /* Array of arguments */
){
  Fts3Cursor *pCsr;               /* Cursor handle passed through apVal[0] */












  UNUSED_PARAMETER(nVal);





































































































































































































































  assert( nVal==1 );










  if( fts3FunctionArg(pContext, "offsets", apVal[0], &pCsr) ) return;
  assert( pCsr );

  sqlite3Fts3Offsets(pContext, pCsr);
}








/* 


















** Implementation of the special optimize() function for FTS3. This 
** function merges all segments in the database to a single segment.
** Example usage is:
**













**   SELECT optimize(t) FROM t LIMIT 1;
**













** where 't' is the name of an FTS3 table.
*/

























































































static void fts3OptimizeFunc(
  sqlite3_context *pContext,      /* SQLite function call context */
  int nVal,                       /* Size of argument array */
  sqlite3_value **apVal           /* Array of arguments */
){






































































































  int rc;                         /* Return code */


  Fts3Table *p;                   /* Virtual table handle */
  Fts3Cursor *pCursor;            /* Cursor handle passed through apVal[0] */









  UNUSED_PARAMETER(nVal);





  assert( nVal==1 );







  if( fts3FunctionArg(pContext, "optimize", apVal[0], &pCursor) ) return;
  p = (Fts3Table *)pCursor->base.pVtab;
  assert( p );











































































  rc = sqlite3Fts3Optimize(p);


















  switch( rc ){
    case SQLITE_OK:



      sqlite3_result_text(pContext, "Index optimized", -1, SQLITE_STATIC);
      break;
    case SQLITE_DONE:






















      sqlite3_result_text(pContext, "Index already optimal", -1, SQLITE_STATIC);
      break;
    default:
      sqlite3_result_error_code(pContext, rc);
      break;
  }











}






/*
** This routine implements the xFindFunction method for the FTS3
** virtual table.
*/
static int fts3FindFunctionMethod(
  sqlite3_vtab *pVtab,            /* Virtual table handle */

  int nArg,                       /* Number of SQL function arguments */
  const char *zName,              /* Name of SQL function */
  void (**pxFunc)(sqlite3_context*,int,sqlite3_value**), /* OUT: Result */
  void **ppArg                    /* Unused */
){




  struct Overloaded {
    const char *zName;
    void (*xFunc)(sqlite3_context*,int,sqlite3_value**);
  } aOverload[] = {
    { "snippet", fts3SnippetFunc },
    { "offsets", fts3OffsetsFunc },

    { "optimize", fts3OptimizeFunc },
  };
  int i;                          /* Iterator variable */

  UNUSED_PARAMETER(pVtab);
  UNUSED_PARAMETER(nArg);
  UNUSED_PARAMETER(ppArg);

  for(i=0; i<SizeofArray(aOverload); i++){
    if( strcmp(zName, aOverload[i].zName)==0 ){
      *pxFunc = aOverload[i].xFunc;
      return 1;














    }
  }

  /* No function of the specified name was found. Return 0. */
  return 0;
}

/*
** Implementation of FTS3 xRename method. Rename an fts3 table.
*/
static int fts3RenameMethod(
  sqlite3_vtab *pVtab,            /* Virtual table handle */
  const char *zName               /* New name of table */
){
  Fts3Table *p = (Fts3Table *)pVtab;     
  int rc = SQLITE_NOMEM;          /* Return Code */
  char *zSql;                     /* SQL script to run to rename tables */
 
  zSql = sqlite3_mprintf(
    "ALTER TABLE %Q.'%q_content'  RENAME TO '%q_content';"
    "ALTER TABLE %Q.'%q_segments' RENAME TO '%q_segments';"
    "ALTER TABLE %Q.'%q_segdir'   RENAME TO '%q_segdir';"
    , p->zDb, p->zName, zName 
    , p->zDb, p->zName, zName 
    , p->zDb, p->zName, zName
  );
................................................................................
    sqlite3_free(zSql);
  }
  return rc;
}

static const sqlite3_module fts3Module = {
  /* iVersion      */ 0,
  /* xCreate       */ fts3CreateMethod,
  /* xConnect      */ fts3ConnectMethod,
  /* xBestIndex    */ fts3BestIndexMethod,
  /* xDisconnect   */ fts3DisconnectMethod,
  /* xDestroy      */ fts3DestroyMethod,
  /* xOpen         */ fts3OpenMethod,
  /* xClose        */ fulltextClose,
  /* xFilter       */ fts3FilterMethod,
  /* xNext         */ fts3NextMethod,
  /* xEof          */ fts3EofMethod,
  /* xColumn       */ fts3ColumnMethod,
  /* xRowid        */ fts3RowidMethod,
  /* xUpdate       */ fts3UpdateMethod,
  /* xBegin        */ fts3BeginMethod,
  /* xSync         */ fts3SyncMethod,
  /* xCommit       */ fts3CommitMethod,
  /* xRollback     */ fts3RollbackMethod,
  /* xFindFunction */ fts3FindFunctionMethod,
  /* xRename */       fts3RenameMethod,
};

/*
** This function is registered as the module destructor (called when an
** FTS3 enabled database connection is closed). It frees the memory
** allocated for the tokenizer hash table.
*/
static void hashDestroy(void *p){
  Fts3Hash *pHash = (Fts3Hash *)p;
  sqlite3Fts3HashClear(pHash);
  sqlite3_free(pHash);
}

/*
** The fts3 built-in tokenizers - "simple" and "porter" - are implemented
** in files fts3_tokenizer1.c and fts3_porter.c respectively. The following
................................................................................
** Function ...PorterTokenizerModule() sets *pModule to point to the
** porter tokenizer/stemmer implementation.
*/
void sqlite3Fts3SimpleTokenizerModule(sqlite3_tokenizer_module const**ppModule);
void sqlite3Fts3PorterTokenizerModule(sqlite3_tokenizer_module const**ppModule);
void sqlite3Fts3IcuTokenizerModule(sqlite3_tokenizer_module const**ppModule);



/*
** Initialise the fts3 extension. If this extension is built as part
** of the sqlite library, then this function is called directly by
** SQLite. If fts3 is built as a dynamically loadable extension, this
** function is called by the sqlite3_extension_init() entry point.
*/
int sqlite3Fts3Init(sqlite3 *db){
  int rc = SQLITE_OK;
  Fts3Hash *pHash = 0;
  const sqlite3_tokenizer_module *pSimple = 0;
  const sqlite3_tokenizer_module *pPorter = 0;
  const sqlite3_tokenizer_module *pIcu = 0;

  sqlite3Fts3SimpleTokenizerModule(&pSimple);
  sqlite3Fts3PorterTokenizerModule(&pPorter);
#ifdef SQLITE_ENABLE_ICU
  sqlite3Fts3IcuTokenizerModule(&pIcu);
#endif

  /* Allocate and initialise the hash-table used to store tokenizers. */
  pHash = sqlite3_malloc(sizeof(Fts3Hash));
  if( !pHash ){
    rc = SQLITE_NOMEM;
  }else{
    sqlite3Fts3HashInit(pHash, FTS3_HASH_STRING, 1);
  }

  /* Load the built-in tokenizers into the hash table */
................................................................................
  /* Create the virtual table wrapper around the hash-table and overload 
  ** the two scalar functions. If this is successful, register the
  ** module with sqlite.
  */
  if( SQLITE_OK==rc 
   && SQLITE_OK==(rc = sqlite3Fts3InitHashTable(db, pHash, "fts3_tokenizer"))
   && SQLITE_OK==(rc = sqlite3_overload_function(db, "snippet", -1))
   && SQLITE_OK==(rc = sqlite3_overload_function(db, "offsets", 1))
   && SQLITE_OK==(rc = sqlite3_overload_function(db, "optimize", 1))




  ){
    return sqlite3_create_module_v2(
        db, "fts3", &fts3Module, (void *)pHash, hashDestroy
    );
  }

  /* An error has occurred. Delete the hash table and return the error code. */
................................................................................
  const sqlite3_api_routines *pApi
){
  SQLITE_EXTENSION_INIT2(pApi)
  return sqlite3Fts3Init(db);
}
#endif

#endif

Added ext/fts3/fts3Int.h.















































































































































































































































































































































































































































































































































































>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
/*
** 2009 Nov 12
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
******************************************************************************
**
*/

#ifndef _FTSINT_H
#define _FTSINT_H

#if !defined(NDEBUG) && !defined(SQLITE_DEBUG) 
# define NDEBUG 1
#endif

#include "sqlite3.h"
#include "fts3_tokenizer.h"
#include "fts3_hash.h"

/*
** This constant controls how often segments are merged. Once there are
** FTS3_MERGE_COUNT segments of level N, they are merged into a single
** segment of level N+1.
*/
#define FTS3_MERGE_COUNT 16

/*
** This is the maximum amount of data (in bytes) to store in the 
** Fts3Table.pendingTerms hash table. Normally, the hash table is
** populated as documents are inserted/updated/deleted in a transaction
** and used to create a new segment when the transaction is committed.
** However if this limit is reached midway through a transaction, a new 
** segment is created and the hash table cleared immediately.
*/
#define FTS3_MAX_PENDING_DATA (1*1024*1024)

/*
** Macro to return the number of elements in an array. SQLite has a
** similar macro called ArraySize(). Use a different name to avoid
** a collision when building an amalgamation with built-in FTS3.
*/
#define SizeofArray(X) ((int)(sizeof(X)/sizeof(X[0])))

/*
** Maximum length of a varint encoded integer. The varint format is different
** from that used by SQLite, so the maximum length is 10, not 9.
*/
#define FTS3_VARINT_MAX 10

/*
** This section provides definitions to allow the
** FTS3 extension to be compiled outside of the 
** amalgamation.
*/
#ifndef SQLITE_AMALGAMATION
/*
** Macros indicating that conditional expressions are always true or
** false.
*/
# define ALWAYS(x) (x)
# define NEVER(X)  (x)
/*
** Internal types used by SQLite.
*/
typedef unsigned char u8;         /* 1-byte (or larger) unsigned integer */
typedef short int i16;            /* 2-byte (or larger) signed integer */
/*
** Macro used to suppress compiler warnings for unused parameters.
*/
#define UNUSED_PARAMETER(x) (void)(x)
#endif

typedef struct Fts3Table Fts3Table;
typedef struct Fts3Cursor Fts3Cursor;
typedef struct Fts3Expr Fts3Expr;
typedef struct Fts3Phrase Fts3Phrase;
typedef struct Fts3SegReader Fts3SegReader;
typedef struct Fts3SegFilter Fts3SegFilter;

/*
** A connection to a fulltext index is an instance of the following
** structure. The xCreate and xConnect methods create an instance
** of this structure and xDestroy and xDisconnect free that instance.
** All other methods receive a pointer to the structure as one of their
** arguments.
*/
struct Fts3Table {
  sqlite3_vtab base;              /* Base class used by SQLite core */
  sqlite3 *db;                    /* The database connection */
  const char *zDb;                /* logical database name */
  const char *zName;              /* virtual table name */
  int nColumn;                    /* number of named columns in virtual table */
  char **azColumn;                /* column names.  malloced */
  sqlite3_tokenizer *pTokenizer;  /* tokenizer for inserts and queries */

  /* Precompiled statements used by the implementation. Each of these 
  ** statements is run and reset within a single virtual table API call. 
  */
  sqlite3_stmt *aStmt[18];

  /* Pointer to string containing the SQL:
  **
  ** "SELECT block FROM %_segments WHERE blockid BETWEEN ? AND ? 
  **    ORDER BY blockid"
  */
  char *zSelectLeaves;
  int nLeavesStmt;                /* Valid statements in aLeavesStmt */
  int nLeavesTotal;               /* Total number of prepared leaves stmts */
  int nLeavesAlloc;               /* Allocated size of aLeavesStmt */
  sqlite3_stmt **aLeavesStmt;     /* Array of prepared zSelectLeaves stmts */

  int nNodeSize;                  /* Soft limit for node size */

  /* The following hash table is used to buffer pending index updates during
  ** transactions. Variable nPendingData estimates the memory size of the 
  ** pending data, including hash table overhead, but not malloc overhead. 
  ** When nPendingData exceeds FTS3_MAX_PENDING_DATA, the buffer is flushed 
  ** automatically. Variable iPrevDocid is the docid of the most recently
  ** inserted record.
  */
  int nPendingData;
  sqlite_int64 iPrevDocid;
  Fts3Hash pendingTerms;
};

/*
** When the core wants to read from the virtual table, it creates a
** virtual table cursor (an instance of the following structure) using
** the xOpen method. Cursors are destroyed using the xClose method.
*/
struct Fts3Cursor {
  sqlite3_vtab_cursor base;       /* Base class used by SQLite core */
  i16 eSearch;                    /* Search strategy (see below) */
  u8 isEof;                       /* True if at End Of Results */
  u8 isRequireSeek;               /* True if must seek pStmt to %_content row */
  sqlite3_stmt *pStmt;            /* Prepared statement in use by the cursor */
  Fts3Expr *pExpr;                /* Parsed MATCH query string */
  sqlite3_int64 iPrevId;          /* Previous id read from aDoclist */
  char *pNextId;                  /* Pointer into the body of aDoclist */
  char *aDoclist;                 /* List of docids for full-text queries */
  int nDoclist;                   /* Size of buffer at aDoclist */
};

/*
** The Fts3Cursor.eSearch member is always set to one of the following.
** Actualy, Fts3Cursor.eSearch can be greater than or equal to
** FTS3_FULLTEXT_SEARCH.  If so, then Fts3Cursor.eSearch - 2 is the index
** of the column to be searched.  For example, in
**
**     CREATE VIRTUAL TABLE ex1 USING fts3(a,b,c,d);
**     SELECT docid FROM ex1 WHERE b MATCH 'one two three';
** 
** Because the LHS of the MATCH operator is 2nd column "b",
** Fts3Cursor.eSearch will be set to FTS3_FULLTEXT_SEARCH+1.  (+0 for a,
** +1 for b, +2 for c, +3 for d.)  If the LHS of MATCH were "ex1" 
** indicating that all columns should be searched,
** then eSearch would be set to FTS3_FULLTEXT_SEARCH+4.
*/
#define FTS3_FULLSCAN_SEARCH 0    /* Linear scan of %_content table */
#define FTS3_DOCID_SEARCH    1    /* Lookup by rowid on %_content table */
#define FTS3_FULLTEXT_SEARCH 2    /* Full-text index search */

/*
** A "phrase" is a sequence of one or more tokens that must match in
** sequence.  A single token is the base case and the most common case.
** For a sequence of tokens contained in "...", nToken will be the number
** of tokens in the string.
*/
struct Fts3Phrase {
  int nToken;                /* Number of tokens in the phrase */
  int iColumn;               /* Index of column this phrase must match */
  int isNot;                 /* Phrase prefixed by unary not (-) operator */
  struct PhraseToken {
    char *z;                 /* Text of the token */
    int n;                   /* Number of bytes in buffer pointed to by z */
    int isPrefix;            /* True if token ends in with a "*" character */
  } aToken[1];               /* One entry for each token in the phrase */
};

/*
** A tree of these objects forms the RHS of a MATCH operator.
*/
struct Fts3Expr {
  int eType;                 /* One of the FTSQUERY_XXX values defined below */
  int nNear;                 /* Valid if eType==FTSQUERY_NEAR */
  Fts3Expr *pParent;         /* pParent->pLeft==this or pParent->pRight==this */
  Fts3Expr *pLeft;           /* Left operand */
  Fts3Expr *pRight;          /* Right operand */
  Fts3Phrase *pPhrase;       /* Valid if eType==FTSQUERY_PHRASE */
};

/*
** Candidate values for Fts3Query.eType. Note that the order of the first
** four values is in order of precedence when parsing expressions. For 
** example, the following:
**
**   "a OR b AND c NOT d NEAR e"
**
** is equivalent to:
**
**   "a OR (b AND (c NOT (d NEAR e)))"
*/
#define FTSQUERY_NEAR   1
#define FTSQUERY_NOT    2
#define FTSQUERY_AND    3
#define FTSQUERY_OR     4
#define FTSQUERY_PHRASE 5


/* fts3_init.c */
int sqlite3Fts3DeleteVtab(int, sqlite3_vtab *);
int sqlite3Fts3InitVtab(int, sqlite3*, void*, int, const char*const*, 
                        sqlite3_vtab **, char **);

/* fts3_write.c */
int sqlite3Fts3UpdateMethod(sqlite3_vtab*,int,sqlite3_value**,sqlite3_int64*);
int sqlite3Fts3PendingTermsFlush(Fts3Table *);
void sqlite3Fts3PendingTermsClear(Fts3Table *);
int sqlite3Fts3Optimize(Fts3Table *);
int sqlite3Fts3SegReaderNew(Fts3Table *,int, sqlite3_int64,
  sqlite3_int64, sqlite3_int64, const char *, int, Fts3SegReader**);
void sqlite3Fts3SegReaderFree(Fts3Table *, Fts3SegReader *);
int sqlite3Fts3SegReaderIterate(
  Fts3Table *, Fts3SegReader **, int, Fts3SegFilter *,
  int (*)(Fts3Table *, void *, char *, int, char *, int),  void *
);
int sqlite3Fts3ReadBlock(Fts3Table*, sqlite3_int64, char const**, int*);
int sqlite3Fts3AllSegdirs(Fts3Table*, sqlite3_stmt **);

/* Flags allowed as part of the 4th argument to SegmentReaderIterate() */
#define FTS3_SEGMENT_REQUIRE_POS   0x00000001
#define FTS3_SEGMENT_IGNORE_EMPTY  0x00000002
#define FTS3_SEGMENT_COLUMN_FILTER 0x00000004
#define FTS3_SEGMENT_PREFIX        0x00000008

/* Type passed as 4th argument to SegmentReaderIterate() */
struct Fts3SegFilter {
  const char *zTerm;
  int nTerm;
  int iCol;
  int flags;
};

/* fts3.c */
int sqlite3Fts3PutVarint(char *, sqlite3_int64);
int sqlite3Fts3GetVarint(const char *, sqlite_int64 *);
int sqlite3Fts3GetVarint32(const char *, int *);
int sqlite3Fts3VarintLen(sqlite3_uint64);
void sqlite3Fts3Dequote(char *);

/* fts3_tokenizer.c */
const char *sqlite3Fts3NextToken(const char *, int *);
int sqlite3Fts3InitHashTable(sqlite3 *, Fts3Hash *, const char *);
int sqlite3Fts3InitTokenizer(Fts3Hash *pHash, 
  const char *, sqlite3_tokenizer **, const char **, char **
);

/* fts3_snippet.c */
void sqlite3Fts3Offsets(sqlite3_context*, Fts3Cursor*);
void sqlite3Fts3Snippet(sqlite3_context*, Fts3Cursor*, 
  const char *, const char *, const char *
);

/* fts3_expr.c */
int sqlite3Fts3ExprParse(sqlite3_tokenizer *, 
  char **, int, int, const char *, int, Fts3Expr **
);
void sqlite3Fts3ExprFree(Fts3Expr *);
#ifdef SQLITE_TEST
void sqlite3Fts3ExprInitTestInterface(sqlite3 *db);
#endif

#endif /* _FTSINT_H */

Changes to ext/fts3/fts3_expr.c.

9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
..
36
37
38
39
40
41
42





















43

44
45
46
47
48
49
50
..
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
...
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
...
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
...
350
351
352
353
354
355
356



357
358
359
360
361
362
363
364
365
366
367
368
...
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
...
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
...
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
...
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
**    May you share freely, never taking more than you give.
**
******************************************************************************
**
** This module contains code that implements a parser for fts3 query strings
** (the right-hand argument to the MATCH operator). Because the supported 
** syntax is relatively simple, the whole tokenizer/parser system is
** hand-coded. The public interface to this module is declared in source
** code file "fts3_expr.h".
*/
#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)

/*
** By default, this module parses the legacy syntax that has been 
** traditionally used by fts3. Or, if SQLITE_ENABLE_FTS3_PARENTHESIS
** is defined, then it uses the new syntax. The differences between
................................................................................
**     AND operators have a higher precedence than OR.
**
** If compiled with SQLITE_TEST defined, then this module exports the
** symbol "int sqlite3_fts3_enable_parentheses". Setting this variable
** to zero causes the module to use the old syntax. If it is set to 
** non-zero the new syntax is activated. This is so both syntaxes can
** be tested using a single build of testfixture.





















*/

#ifdef SQLITE_TEST
int sqlite3_fts3_enable_parentheses = 0;
#else
# ifdef SQLITE_ENABLE_FTS3_PARENTHESIS 
#  define sqlite3_fts3_enable_parentheses 1
# else
#  define sqlite3_fts3_enable_parentheses 0
................................................................................
#endif

/*
** Default span for NEAR operators.
*/
#define SQLITE_FTS3_DEFAULT_NEAR_PARAM 10

#include "fts3_expr.h"
#include "sqlite3.h"
#include <ctype.h>
#include <string.h>
#include <assert.h>

typedef struct ParseContext ParseContext;
struct ParseContext {
  sqlite3_tokenizer *pTokenizer;      /* Tokenizer module */
................................................................................

    pModule->xClose(pCursor);
    pCursor = 0;
  }

  if( rc==SQLITE_DONE ){
    int jj;
    char *zNew;
    int nNew = 0;
    int nByte = sizeof(Fts3Expr) + sizeof(Fts3Phrase);
    nByte += (p?(p->pPhrase->nToken-1):0) * sizeof(struct PhraseToken);
    p = fts3ReallocOrFree(p, nByte + nTemp);
    if( !p ){
      goto no_mem;
    }
................................................................................
static int getNextNode(
  ParseContext *pParse,                   /* fts3 query parse context */
  const char *z, int n,                   /* Input string */
  Fts3Expr **ppExpr,                      /* OUT: expression */
  int *pnConsumed                         /* OUT: Number of bytes consumed */
){
  static const struct Fts3Keyword {
    char z[4];                            /* Keyword text */
    unsigned char n;                      /* Length of the keyword */
    unsigned char parenOnly;              /* Only valid in paren mode */
    unsigned char eType;                  /* Keyword code */
  } aKeyword[] = {
    { "OR" ,  2, 0, FTSQUERY_OR   },
    { "AND",  3, 1, FTSQUERY_AND  },
    { "NOT",  3, 1, FTSQUERY_NOT  },
................................................................................
      ** parenthesis, a quote character, or EOF. 
      */
      cNext = zInput[nKey];
      if( fts3isspace(cNext) 
       || cNext=='"' || cNext=='(' || cNext==')' || cNext==0
      ){
        pRet = (Fts3Expr *)sqlite3_malloc(sizeof(Fts3Expr));



        memset(pRet, 0, sizeof(Fts3Expr));
        pRet->eType = pKey->eType;
        pRet->nNear = nNear;
        *ppExpr = pRet;
        *pnConsumed = (zInput - z) + nKey;
        return SQLITE_OK;
      }

      /* Turns out that wasn't a keyword after all. This happens if the
      ** user has supplied a token such as "ORacle". Continue.
      */
    }
................................................................................
      int nConsumed;
      int rc;
      pParse->nNest++;
      rc = fts3ExprParse(pParse, &zInput[1], nInput-1, ppExpr, &nConsumed);
      if( rc==SQLITE_OK && !*ppExpr ){
        rc = SQLITE_DONE;
      }
      *pnConsumed = (zInput - z) + 1 + nConsumed;
      return rc;
    }
  
    /* Check for a close bracket. */
    if( *zInput==')' ){
      pParse->nNest--;
      *pnConsumed = (zInput - z) + 1;
      return SQLITE_DONE;
    }
  }

  /* See if we are dealing with a quoted phrase. If this is the case, then
  ** search for the closing quote and pass the whole string to getNextString()
  ** for processing. This is easy to do, as fts3 has no syntax for escaping
  ** a quote character embedded in a string.
  */
  if( *zInput=='"' ){
    for(ii=1; ii<nInput && zInput[ii]!='"'; ii++);
    *pnConsumed = (zInput - z) + ii + 1;
    if( ii==nInput ){
      return SQLITE_ERROR;
    }
    return getNextString(pParse, &zInput[1], ii-1, ppExpr);
  }


................................................................................
  ** first implemented. Whichever it was, this module duplicates the 
  ** limitation.
  */
  iCol = pParse->iDefaultCol;
  iColLen = 0;
  for(ii=0; ii<pParse->nCol; ii++){
    const char *zStr = pParse->azCol[ii];
    int nStr = strlen(zStr);
    if( nInput>nStr && zInput[nStr]==':' 
     && sqlite3_strnicmp(zStr, zInput, nStr)==0 
    ){
      iCol = ii;
      iColLen = ((zInput - z) + nStr + 1);
      break;
    }
  }
  rc = getNextToken(pParse, iCol, &z[iColLen], n-iColLen, ppExpr, pnConsumed);
  *pnConsumed += iColLen;
  return rc;
}
................................................................................
  sParse.iDefaultCol = iDefaultCol;
  sParse.nNest = 0;
  if( z==0 ){
    *ppExpr = 0;
    return SQLITE_OK;
  }
  if( n<0 ){
    n = strlen(z);
  }
  rc = fts3ExprParse(&sParse, z, n, ppExpr, &nParsed);

  /* Check for mismatched parenthesis */
  if( rc==SQLITE_OK && sParse.nNest ){
    rc = SQLITE_ERROR;
    sqlite3Fts3ExprFree(*ppExpr);
................................................................................
  if( rc!=SQLITE_OK ){
    return rc;
  }

  sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC);
  if( SQLITE_ROW==sqlite3_step(pStmt) ){
    if( sqlite3_column_type(pStmt, 0)==SQLITE_BLOB ){
      memcpy(pp, sqlite3_column_blob(pStmt, 0), sizeof(*pp));
    }
  }

  return sqlite3_finalize(pStmt);
}

/*







|
<







 







>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>

>







 







|
<







 







|







 







|







 







>
>
>




|







 







|






|











|







 







|




|







 







|







 







|







9
10
11
12
13
14
15
16

17
18
19
20
21
22
23
..
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
..
73
74
75
76
77
78
79
80

81
82
83
84
85
86
87
...
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
...
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
...
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
...
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
...
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
...
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
...
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
**    May you share freely, never taking more than you give.
**
******************************************************************************
**
** This module contains code that implements a parser for fts3 query strings
** (the right-hand argument to the MATCH operator). Because the supported 
** syntax is relatively simple, the whole tokenizer/parser system is
** hand-coded. 

*/
#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)

/*
** By default, this module parses the legacy syntax that has been 
** traditionally used by fts3. Or, if SQLITE_ENABLE_FTS3_PARENTHESIS
** is defined, then it uses the new syntax. The differences between
................................................................................
**     AND operators have a higher precedence than OR.
**
** If compiled with SQLITE_TEST defined, then this module exports the
** symbol "int sqlite3_fts3_enable_parentheses". Setting this variable
** to zero causes the module to use the old syntax. If it is set to 
** non-zero the new syntax is activated. This is so both syntaxes can
** be tested using a single build of testfixture.
**
** The following describes the syntax supported by the fts3 MATCH
** operator in a similar format to that used by the lemon parser
** generator. This module does not use actually lemon, it uses a
** custom parser.
**
**   query ::= andexpr (OR andexpr)*.
**
**   andexpr ::= notexpr (AND? notexpr)*.
**
**   notexpr ::= nearexpr (NOT nearexpr|-TOKEN)*.
**   notexpr ::= LP query RP.
**
**   nearexpr ::= phrase (NEAR distance_opt nearexpr)*.
**
**   distance_opt ::= .
**   distance_opt ::= / INTEGER.
**
**   phrase ::= TOKEN.
**   phrase ::= COLUMN:TOKEN.
**   phrase ::= "TOKEN TOKEN TOKEN...".
*/

#ifdef SQLITE_TEST
int sqlite3_fts3_enable_parentheses = 0;
#else
# ifdef SQLITE_ENABLE_FTS3_PARENTHESIS 
#  define sqlite3_fts3_enable_parentheses 1
# else
#  define sqlite3_fts3_enable_parentheses 0
................................................................................
#endif

/*
** Default span for NEAR operators.
*/
#define SQLITE_FTS3_DEFAULT_NEAR_PARAM 10

#include "fts3Int.h"

#include <ctype.h>
#include <string.h>
#include <assert.h>

typedef struct ParseContext ParseContext;
struct ParseContext {
  sqlite3_tokenizer *pTokenizer;      /* Tokenizer module */
................................................................................

    pModule->xClose(pCursor);
    pCursor = 0;
  }

  if( rc==SQLITE_DONE ){
    int jj;
    char *zNew = NULL;
    int nNew = 0;
    int nByte = sizeof(Fts3Expr) + sizeof(Fts3Phrase);
    nByte += (p?(p->pPhrase->nToken-1):0) * sizeof(struct PhraseToken);
    p = fts3ReallocOrFree(p, nByte + nTemp);
    if( !p ){
      goto no_mem;
    }
................................................................................
static int getNextNode(
  ParseContext *pParse,                   /* fts3 query parse context */
  const char *z, int n,                   /* Input string */
  Fts3Expr **ppExpr,                      /* OUT: expression */
  int *pnConsumed                         /* OUT: Number of bytes consumed */
){
  static const struct Fts3Keyword {
    char *z;                              /* Keyword text */
    unsigned char n;                      /* Length of the keyword */
    unsigned char parenOnly;              /* Only valid in paren mode */
    unsigned char eType;                  /* Keyword code */
  } aKeyword[] = {
    { "OR" ,  2, 0, FTSQUERY_OR   },
    { "AND",  3, 1, FTSQUERY_AND  },
    { "NOT",  3, 1, FTSQUERY_NOT  },
................................................................................
      ** parenthesis, a quote character, or EOF. 
      */
      cNext = zInput[nKey];
      if( fts3isspace(cNext) 
       || cNext=='"' || cNext=='(' || cNext==')' || cNext==0
      ){
        pRet = (Fts3Expr *)sqlite3_malloc(sizeof(Fts3Expr));
        if( !pRet ){
          return SQLITE_NOMEM;
        }
        memset(pRet, 0, sizeof(Fts3Expr));
        pRet->eType = pKey->eType;
        pRet->nNear = nNear;
        *ppExpr = pRet;
        *pnConsumed = (int)((zInput - z) + nKey);
        return SQLITE_OK;
      }

      /* Turns out that wasn't a keyword after all. This happens if the
      ** user has supplied a token such as "ORacle". Continue.
      */
    }
................................................................................
      int nConsumed;
      int rc;
      pParse->nNest++;
      rc = fts3ExprParse(pParse, &zInput[1], nInput-1, ppExpr, &nConsumed);
      if( rc==SQLITE_OK && !*ppExpr ){
        rc = SQLITE_DONE;
      }
      *pnConsumed = (int)((zInput - z) + 1 + nConsumed);
      return rc;
    }
  
    /* Check for a close bracket. */
    if( *zInput==')' ){
      pParse->nNest--;
      *pnConsumed = (int)((zInput - z) + 1);
      return SQLITE_DONE;
    }
  }

  /* See if we are dealing with a quoted phrase. If this is the case, then
  ** search for the closing quote and pass the whole string to getNextString()
  ** for processing. This is easy to do, as fts3 has no syntax for escaping
  ** a quote character embedded in a string.
  */
  if( *zInput=='"' ){
    for(ii=1; ii<nInput && zInput[ii]!='"'; ii++);
    *pnConsumed = (int)((zInput - z) + ii + 1);
    if( ii==nInput ){
      return SQLITE_ERROR;
    }
    return getNextString(pParse, &zInput[1], ii-1, ppExpr);
  }


................................................................................
  ** first implemented. Whichever it was, this module duplicates the 
  ** limitation.
  */
  iCol = pParse->iDefaultCol;
  iColLen = 0;
  for(ii=0; ii<pParse->nCol; ii++){
    const char *zStr = pParse->azCol[ii];
    int nStr = (int)strlen(zStr);
    if( nInput>nStr && zInput[nStr]==':' 
     && sqlite3_strnicmp(zStr, zInput, nStr)==0 
    ){
      iCol = ii;
      iColLen = (int)((zInput - z) + nStr + 1);
      break;
    }
  }
  rc = getNextToken(pParse, iCol, &z[iColLen], n-iColLen, ppExpr, pnConsumed);
  *pnConsumed += iColLen;
  return rc;
}
................................................................................
  sParse.iDefaultCol = iDefaultCol;
  sParse.nNest = 0;
  if( z==0 ){
    *ppExpr = 0;
    return SQLITE_OK;
  }
  if( n<0 ){
    n = (int)strlen(z);
  }
  rc = fts3ExprParse(&sParse, z, n, ppExpr, &nParsed);

  /* Check for mismatched parenthesis */
  if( rc==SQLITE_OK && sParse.nNest ){
    rc = SQLITE_ERROR;
    sqlite3Fts3ExprFree(*ppExpr);
................................................................................
  if( rc!=SQLITE_OK ){
    return rc;
  }

  sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC);
  if( SQLITE_ROW==sqlite3_step(pStmt) ){
    if( sqlite3_column_type(pStmt, 0)==SQLITE_BLOB ){
      memcpy((void *)pp, sqlite3_column_blob(pStmt, 0), sizeof(*pp));
    }
  }

  return sqlite3_finalize(pStmt);
}

/*

Deleted ext/fts3/fts3_expr.h.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
/*
** 2008 Nov 28
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
******************************************************************************
**
*/

#include "fts3_tokenizer.h"
#include "sqlite3.h"

/*
** The following describes the syntax supported by the fts3 MATCH
** operator in a similar format to that used by the lemon parser
** generator. This module does not use actually lemon, it uses a
** custom parser.
**
**   query ::= andexpr (OR andexpr)*.
**
**   andexpr ::= notexpr (AND? notexpr)*.
**
**   notexpr ::= nearexpr (NOT nearexpr|-TOKEN)*.
**   notexpr ::= LP query RP.
**
**   nearexpr ::= phrase (NEAR distance_opt nearexpr)*.
**
**   distance_opt ::= .
**   distance_opt ::= / INTEGER.
**
**   phrase ::= TOKEN.
**   phrase ::= COLUMN:TOKEN.
**   phrase ::= "TOKEN TOKEN TOKEN...".
*/

typedef struct Fts3Expr Fts3Expr;
typedef struct Fts3Phrase Fts3Phrase;

/*
** A "phrase" is a sequence of one or more tokens that must match in
** sequence.  A single token is the base case and the most common case.
** For a sequence of tokens contained in "...", nToken will be the number
** of tokens in the string.
*/
struct Fts3Phrase {
  int nToken;          /* Number of tokens in the phrase */
  int iColumn;         /* Index of column this phrase must match */
  int isNot;           /* Phrase prefixed by unary not (-) operator */
  struct PhraseToken {
    char *z;              /* Text of the token */
    int n;                /* Number of bytes in buffer pointed to by z */
    int isPrefix;         /* True if token ends in with a "*" character */
  } aToken[1];         /* One entry for each token in the phrase */
};

/*
** A tree of these objects forms the RHS of a MATCH operator.
*/
struct Fts3Expr {
  int eType;                 /* One of the FTSQUERY_XXX values defined below */
  int nNear;                 /* Valid if eType==FTSQUERY_NEAR */
  Fts3Expr *pParent;         /* pParent->pLeft==this or pParent->pRight==this */
  Fts3Expr *pLeft;           /* Left operand */
  Fts3Expr *pRight;          /* Right operand */
  Fts3Phrase *pPhrase;       /* Valid if eType==FTSQUERY_PHRASE */
};

int sqlite3Fts3ExprParse(sqlite3_tokenizer *, char **, int, int, 
                         const char *, int, Fts3Expr **);
void sqlite3Fts3ExprFree(Fts3Expr *);

/*
** Candidate values for Fts3Query.eType. Note that the order of the first
** four values is in order of precedence when parsing expressions. For 
** example, the following:
**
**   "a OR b AND c NOT d NEAR e"
**
** is equivalent to:
**
**   "a OR (b AND (c NOT (d NEAR e)))"
*/
#define FTSQUERY_NEAR   1
#define FTSQUERY_NOT    2
#define FTSQUERY_AND    3
#define FTSQUERY_OR     4
#define FTSQUERY_PHRASE 5

#ifdef SQLITE_TEST
void sqlite3Fts3ExprInitTestInterface(sqlite3 *db);
#endif
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
































































































































































































Changes to ext/fts3/fts3_hash.c.

52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
..
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
...
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
...
186
187
188
189
190
191
192


193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210

211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
...
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
...
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
...
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
...
334
335
336
337
338
339
340
341
342
343
344
345
346
347

348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
** "pNew" is a pointer to the hash table that is to be initialized.
** keyClass is one of the constants 
** FTS3_HASH_BINARY or FTS3_HASH_STRING.  The value of keyClass 
** determines what kind of key the hash table will use.  "copyKey" is
** true if the hash table should make its own private copy of keys and
** false if it should just use the supplied pointer.
*/
void sqlite3Fts3HashInit(fts3Hash *pNew, int keyClass, int copyKey){
  assert( pNew!=0 );
  assert( keyClass>=FTS3_HASH_STRING && keyClass<=FTS3_HASH_BINARY );
  pNew->keyClass = keyClass;
  pNew->copyKey = copyKey;
  pNew->first = 0;
  pNew->count = 0;
  pNew->htsize = 0;
................................................................................
  pNew->ht = 0;
}

/* Remove all entries from a hash table.  Reclaim all memory.
** Call this routine to delete a hash table or to reset a hash table
** to the empty state.
*/
void sqlite3Fts3HashClear(fts3Hash *pH){
  fts3HashElem *elem;         /* For looping over all elements of the table */

  assert( pH!=0 );
  elem = pH->first;
  pH->first = 0;
  fts3HashFree(pH->ht);
  pH->ht = 0;
  pH->htsize = 0;
  while( elem ){
    fts3HashElem *next_elem = elem->next;
    if( pH->copyKey && elem->pKey ){
      fts3HashFree(elem->pKey);
    }
    fts3HashFree(elem);
    elem = next_elem;
  }
  pH->count = 0;
................................................................................
    return &fts3BinCompare;
  }
}

/* Link an element into the hash table
*/
static void fts3HashInsertElement(
  fts3Hash *pH,            /* The complete hash table */
  struct _fts3ht *pEntry,  /* The entry into which pNew is inserted */
  fts3HashElem *pNew       /* The element to be inserted */
){
  fts3HashElem *pHead;     /* First element already in pEntry */
  pHead = pEntry->chain;
  if( pHead ){
    pNew->next = pHead;
    pNew->prev = pHead->prev;
    if( pHead->prev ){ pHead->prev->next = pNew; }
    else             { pH->first = pNew; }
    pHead->prev = pNew;
................................................................................
  pEntry->chain = pNew;
}


/* Resize the hash table so that it cantains "new_size" buckets.
** "new_size" must be a power of 2.  The hash table might fail 
** to resize if sqliteMalloc() fails.


*/
static void fts3Rehash(fts3Hash *pH, int new_size){
  struct _fts3ht *new_ht;          /* The new hash table */
  fts3HashElem *elem, *next_elem;  /* For looping over existing elements */
  int (*xHash)(const void*,int);   /* The hash function */

  assert( (new_size & (new_size-1))==0 );
  new_ht = (struct _fts3ht *)fts3HashMalloc( new_size*sizeof(struct _fts3ht) );
  if( new_ht==0 ) return;
  fts3HashFree(pH->ht);
  pH->ht = new_ht;
  pH->htsize = new_size;
  xHash = ftsHashFunction(pH->keyClass);
  for(elem=pH->first, pH->first=0; elem; elem = next_elem){
    int h = (*xHash)(elem->pKey, elem->nKey) & (new_size-1);
    next_elem = elem->next;
    fts3HashInsertElement(pH, &new_ht[h], elem);
  }

}

/* This function (for internal use only) locates an element in an
** hash table that matches the given key.  The hash for this key has
** already been computed and is passed as the 4th parameter.
*/
static fts3HashElem *fts3FindElementByHash(
  const fts3Hash *pH, /* The pH to be searched */
  const void *pKey,   /* The key we are searching for */
  int nKey,
  int h               /* The hash for this key. */
){
  fts3HashElem *elem;            /* Used to loop thru the element list */
  int count;                     /* Number of elements left to test */
  int (*xCompare)(const void*,int,const void*,int);  /* comparison function */

  if( pH->ht ){
    struct _fts3ht *pEntry = &pH->ht[h];
    elem = pEntry->chain;
    count = pEntry->count;
................................................................................
  return 0;
}

/* Remove a single entry from the hash table given a pointer to that
** element and a hash on the element's key.
*/
static void fts3RemoveElementByHash(
  fts3Hash *pH,         /* The pH containing "elem" */
  fts3HashElem* elem,   /* The element to be removed from the pH */
  int h                 /* Hash value for the element */
){
  struct _fts3ht *pEntry;
  if( elem->prev ){
    elem->prev->next = elem->next; 
  }else{
    pH->first = elem->next;
................................................................................
  }
}

/* Attempt to locate an element of the hash table pH with a key
** that matches pKey,nKey.  Return the data for this element if it is
** found, or NULL if there is no match.
*/
void *sqlite3Fts3HashFind(const fts3Hash *pH, const void *pKey, int nKey){
  int h;                 /* A hash on key */
  fts3HashElem *elem;    /* The element that matches key */
  int (*xHash)(const void*,int);  /* The hash function */

  if( pH==0 || pH->ht==0 ) return 0;
  xHash = ftsHashFunction(pH->keyClass);
  assert( xHash!=0 );
  h = (*xHash)(pKey,nKey);
  assert( (pH->htsize & (pH->htsize-1))==0 );
................................................................................
** The key is not copied in this instance.  If a malloc fails, then
** the new data is returned and the hash table is unchanged.
**
** If the "data" parameter to this function is NULL, then the
** element corresponding to "key" is removed from the hash table.
*/
void *sqlite3Fts3HashInsert(
  fts3Hash *pH,        /* The hash table to insert into */
  const void *pKey,    /* The key */
  int nKey,            /* Number of bytes in the key */
  void *data           /* The data */
){
  int hraw;                 /* Raw hash value of the key */
  int h;                    /* the hash of the key modulo hash table size */
  fts3HashElem *elem;       /* Used to loop thru the element list */
  fts3HashElem *new_elem;   /* New element added to the pH */
  int (*xHash)(const void*,int);  /* The hash function */

  assert( pH!=0 );
  xHash = ftsHashFunction(pH->keyClass);
  assert( xHash!=0 );
  hraw = (*xHash)(pKey, nKey);
  assert( (pH->htsize & (pH->htsize-1))==0 );
................................................................................
      fts3RemoveElementByHash(pH,elem,h);
    }else{
      elem->data = data;
    }
    return old_data;
  }
  if( data==0 ) return 0;
  if( pH->htsize==0 ){
    fts3Rehash(pH,8);
    if( pH->htsize==0 ){
      pH->count = 0;
      return data;
    }
  }

  new_elem = (fts3HashElem*)fts3HashMalloc( sizeof(fts3HashElem) );
  if( new_elem==0 ) return data;
  if( pH->copyKey && pKey!=0 ){
    new_elem->pKey = fts3HashMalloc( nKey );
    if( new_elem->pKey==0 ){
      fts3HashFree(new_elem);
      return data;
    }
    memcpy((void*)new_elem->pKey, pKey, nKey);
  }else{
    new_elem->pKey = (void*)pKey;
  }
  new_elem->nKey = nKey;
  pH->count++;
  if( pH->count > pH->htsize ){
    fts3Rehash(pH,pH->htsize*2);
  }
  assert( pH->htsize>0 );
  assert( (pH->htsize & (pH->htsize-1))==0 );
  h = hraw & (pH->htsize-1);
  fts3HashInsertElement(pH, &pH->ht[h], new_elem);
  new_elem->data = data;
  return 0;
}

#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */







|







 







|
|








|







 







|

|

|







 







>
>

|

|




|









>






|
|




|







 







|
|







 







|

|







 







|






|
|







 







|
|
|
|
|
|
<
>
|













<
<
<









52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
..
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
...
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
...
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
...
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
...
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
...
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
...
337
338
339
340
341
342
343
344
345
346
347
348
349

350
351
352
353
354
355
356
357
358
359
360
361
362
363
364



365
366
367
368
369
370
371
372
373
** "pNew" is a pointer to the hash table that is to be initialized.
** keyClass is one of the constants 
** FTS3_HASH_BINARY or FTS3_HASH_STRING.  The value of keyClass 
** determines what kind of key the hash table will use.  "copyKey" is
** true if the hash table should make its own private copy of keys and
** false if it should just use the supplied pointer.
*/
void sqlite3Fts3HashInit(Fts3Hash *pNew, char keyClass, char copyKey){
  assert( pNew!=0 );
  assert( keyClass>=FTS3_HASH_STRING && keyClass<=FTS3_HASH_BINARY );
  pNew->keyClass = keyClass;
  pNew->copyKey = copyKey;
  pNew->first = 0;
  pNew->count = 0;
  pNew->htsize = 0;
................................................................................
  pNew->ht = 0;
}

/* Remove all entries from a hash table.  Reclaim all memory.
** Call this routine to delete a hash table or to reset a hash table
** to the empty state.
*/
void sqlite3Fts3HashClear(Fts3Hash *pH){
  Fts3HashElem *elem;         /* For looping over all elements of the table */

  assert( pH!=0 );
  elem = pH->first;
  pH->first = 0;
  fts3HashFree(pH->ht);
  pH->ht = 0;
  pH->htsize = 0;
  while( elem ){
    Fts3HashElem *next_elem = elem->next;
    if( pH->copyKey && elem->pKey ){
      fts3HashFree(elem->pKey);
    }
    fts3HashFree(elem);
    elem = next_elem;
  }
  pH->count = 0;
................................................................................
    return &fts3BinCompare;
  }
}

/* Link an element into the hash table
*/
static void fts3HashInsertElement(
  Fts3Hash *pH,            /* The complete hash table */
  struct _fts3ht *pEntry,  /* The entry into which pNew is inserted */
  Fts3HashElem *pNew       /* The element to be inserted */
){
  Fts3HashElem *pHead;     /* First element already in pEntry */
  pHead = pEntry->chain;
  if( pHead ){
    pNew->next = pHead;
    pNew->prev = pHead->prev;
    if( pHead->prev ){ pHead->prev->next = pNew; }
    else             { pH->first = pNew; }
    pHead->prev = pNew;
................................................................................
  pEntry->chain = pNew;
}


/* Resize the hash table so that it cantains "new_size" buckets.
** "new_size" must be a power of 2.  The hash table might fail 
** to resize if sqliteMalloc() fails.
**
** Return non-zero if a memory allocation error occurs.
*/
static int fts3Rehash(Fts3Hash *pH, int new_size){
  struct _fts3ht *new_ht;          /* The new hash table */
  Fts3HashElem *elem, *next_elem;  /* For looping over existing elements */
  int (*xHash)(const void*,int);   /* The hash function */

  assert( (new_size & (new_size-1))==0 );
  new_ht = (struct _fts3ht *)fts3HashMalloc( new_size*sizeof(struct _fts3ht) );
  if( new_ht==0 ) return 1;
  fts3HashFree(pH->ht);
  pH->ht = new_ht;
  pH->htsize = new_size;
  xHash = ftsHashFunction(pH->keyClass);
  for(elem=pH->first, pH->first=0; elem; elem = next_elem){
    int h = (*xHash)(elem->pKey, elem->nKey) & (new_size-1);
    next_elem = elem->next;
    fts3HashInsertElement(pH, &new_ht[h], elem);
  }
  return 0;
}

/* This function (for internal use only) locates an element in an
** hash table that matches the given key.  The hash for this key has
** already been computed and is passed as the 4th parameter.
*/
static Fts3HashElem *fts3FindElementByHash(
  const Fts3Hash *pH, /* The pH to be searched */
  const void *pKey,   /* The key we are searching for */
  int nKey,
  int h               /* The hash for this key. */
){
  Fts3HashElem *elem;            /* Used to loop thru the element list */
  int count;                     /* Number of elements left to test */
  int (*xCompare)(const void*,int,const void*,int);  /* comparison function */

  if( pH->ht ){
    struct _fts3ht *pEntry = &pH->ht[h];
    elem = pEntry->chain;
    count = pEntry->count;
................................................................................
  return 0;
}

/* Remove a single entry from the hash table given a pointer to that
** element and a hash on the element's key.
*/
static void fts3RemoveElementByHash(
  Fts3Hash *pH,         /* The pH containing "elem" */
  Fts3HashElem* elem,   /* The element to be removed from the pH */
  int h                 /* Hash value for the element */
){
  struct _fts3ht *pEntry;
  if( elem->prev ){
    elem->prev->next = elem->next; 
  }else{
    pH->first = elem->next;
................................................................................
  }
}

/* Attempt to locate an element of the hash table pH with a key
** that matches pKey,nKey.  Return the data for this element if it is
** found, or NULL if there is no match.
*/
void *sqlite3Fts3HashFind(const Fts3Hash *pH, const void *pKey, int nKey){
  int h;                 /* A hash on key */
  Fts3HashElem *elem;    /* The element that matches key */
  int (*xHash)(const void*,int);  /* The hash function */

  if( pH==0 || pH->ht==0 ) return 0;
  xHash = ftsHashFunction(pH->keyClass);
  assert( xHash!=0 );
  h = (*xHash)(pKey,nKey);
  assert( (pH->htsize & (pH->htsize-1))==0 );
................................................................................
** The key is not copied in this instance.  If a malloc fails, then
** the new data is returned and the hash table is unchanged.
**
** If the "data" parameter to this function is NULL, then the
** element corresponding to "key" is removed from the hash table.
*/
void *sqlite3Fts3HashInsert(
  Fts3Hash *pH,        /* The hash table to insert into */
  const void *pKey,    /* The key */
  int nKey,            /* Number of bytes in the key */
  void *data           /* The data */
){
  int hraw;                 /* Raw hash value of the key */
  int h;                    /* the hash of the key modulo hash table size */
  Fts3HashElem *elem;       /* Used to loop thru the element list */
  Fts3HashElem *new_elem;   /* New element added to the pH */
  int (*xHash)(const void*,int);  /* The hash function */

  assert( pH!=0 );
  xHash = ftsHashFunction(pH->keyClass);
  assert( xHash!=0 );
  hraw = (*xHash)(pKey, nKey);
  assert( (pH->htsize & (pH->htsize-1))==0 );
................................................................................
      fts3RemoveElementByHash(pH,elem,h);
    }else{
      elem->data = data;
    }
    return old_data;
  }
  if( data==0 ) return 0;
  if( (pH->htsize==0 && fts3Rehash(pH,8))
   || (pH->count>=pH->htsize && fts3Rehash(pH, pH->htsize*2))
  ){
    pH->count = 0;
    return data;
  }

  assert( pH->htsize>0 );
  new_elem = (Fts3HashElem*)fts3HashMalloc( sizeof(Fts3HashElem) );
  if( new_elem==0 ) return data;
  if( pH->copyKey && pKey!=0 ){
    new_elem->pKey = fts3HashMalloc( nKey );
    if( new_elem->pKey==0 ){
      fts3HashFree(new_elem);
      return data;
    }
    memcpy((void*)new_elem->pKey, pKey, nKey);
  }else{
    new_elem->pKey = (void*)pKey;
  }
  new_elem->nKey = nKey;
  pH->count++;



  assert( pH->htsize>0 );
  assert( (pH->htsize & (pH->htsize-1))==0 );
  h = hraw & (pH->htsize-1);
  fts3HashInsertElement(pH, &pH->ht[h], new_elem);
  new_elem->data = data;
  return 0;
}

#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */

Changes to ext/fts3/fts3_hash.h.

14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
..
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
** hash table implementation for the full-text indexing module.
**
*/
#ifndef _FTS3_HASH_H_
#define _FTS3_HASH_H_

/* Forward declarations of structures. */
typedef struct fts3Hash fts3Hash;
typedef struct fts3HashElem fts3HashElem;

/* A complete hash table is an instance of the following structure.
** The internals of this structure are intended to be opaque -- client
** code should not attempt to access or modify the fields of this structure
** directly.  Change this structure only by using the routines below.
** However, many of the "procedures" and "functions" for modifying and
** accessing this structure are really macros, so we can't really make
** this structure opaque.
*/
struct fts3Hash {
  char keyClass;          /* HASH_INT, _POINTER, _STRING, _BINARY */
  char copyKey;           /* True if copy of key made on insert */
  int count;              /* Number of entries in this table */
  fts3HashElem *first;    /* The first element of the array */
  int htsize;             /* Number of buckets in the hash table */
  struct _fts3ht {        /* the hash table */
    int count;               /* Number of entries with this hash */
    fts3HashElem *chain;     /* Pointer to first entry with this hash */
  } *ht;
};

/* Each element in the hash table is an instance of the following 
** structure.  All elements are stored on a single doubly-linked list.
**
** Again, this structure is intended to be opaque, but it can't really
** be opaque because it is used by macros.
*/
struct fts3HashElem {
  fts3HashElem *next, *prev; /* Next and previous elements in the table */
  void *data;                /* Data associated with this element */
  void *pKey; int nKey;      /* Key associated with this element */
};

/*
** There are 2 different modes of operation for a hash table:
**
................................................................................
*/
#define FTS3_HASH_STRING    1
#define FTS3_HASH_BINARY    2

/*
** Access routines.  To delete, insert a NULL pointer.
*/
void sqlite3Fts3HashInit(fts3Hash*, int keytype, int copyKey);
void *sqlite3Fts3HashInsert(fts3Hash*, const void *pKey, int nKey, void *pData);
void *sqlite3Fts3HashFind(const fts3Hash*, const void *pKey, int nKey);
void sqlite3Fts3HashClear(fts3Hash*);

/*
** Shorthand for the functions above
*/
#define fts3HashInit   sqlite3Fts3HashInit
#define fts3HashInsert sqlite3Fts3HashInsert
#define fts3HashFind   sqlite3Fts3HashFind
#define fts3HashClear  sqlite3Fts3HashClear

/*
** Macros for looping over all elements of a hash table.  The idiom is
** like this:
**
**   fts3Hash h;
**   fts3HashElem *p;
**   ...
**   for(p=fts3HashFirst(&h); p; p=fts3HashNext(p)){
**     SomeStructure *pData = fts3HashData(p);
**     // do something with pData
**   }
*/
#define fts3HashFirst(H)  ((H)->first)







|
|









|



|



|









|
|







 







|
|
|
|













|
|







14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
..
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
** hash table implementation for the full-text indexing module.
**
*/
#ifndef _FTS3_HASH_H_
#define _FTS3_HASH_H_

/* Forward declarations of structures. */
typedef struct Fts3Hash Fts3Hash;
typedef struct Fts3HashElem Fts3HashElem;

/* A complete hash table is an instance of the following structure.
** The internals of this structure are intended to be opaque -- client
** code should not attempt to access or modify the fields of this structure
** directly.  Change this structure only by using the routines below.
** However, many of the "procedures" and "functions" for modifying and
** accessing this structure are really macros, so we can't really make
** this structure opaque.
*/
struct Fts3Hash {
  char keyClass;          /* HASH_INT, _POINTER, _STRING, _BINARY */
  char copyKey;           /* True if copy of key made on insert */
  int count;              /* Number of entries in this table */
  Fts3HashElem *first;    /* The first element of the array */
  int htsize;             /* Number of buckets in the hash table */
  struct _fts3ht {        /* the hash table */
    int count;               /* Number of entries with this hash */
    Fts3HashElem *chain;     /* Pointer to first entry with this hash */
  } *ht;
};

/* Each element in the hash table is an instance of the following 
** structure.  All elements are stored on a single doubly-linked list.
**
** Again, this structure is intended to be opaque, but it can't really
** be opaque because it is used by macros.
*/
struct Fts3HashElem {
  Fts3HashElem *next, *prev; /* Next and previous elements in the table */
  void *data;                /* Data associated with this element */
  void *pKey; int nKey;      /* Key associated with this element */
};

/*
** There are 2 different modes of operation for a hash table:
**
................................................................................
*/
#define FTS3_HASH_STRING    1
#define FTS3_HASH_BINARY    2

/*
** Access routines.  To delete, insert a NULL pointer.
*/
void sqlite3Fts3HashInit(Fts3Hash *pNew, char keyClass, char copyKey);
void *sqlite3Fts3HashInsert(Fts3Hash*, const void *pKey, int nKey, void *pData);
void *sqlite3Fts3HashFind(const Fts3Hash*, const void *pKey, int nKey);
void sqlite3Fts3HashClear(Fts3Hash*);

/*
** Shorthand for the functions above
*/
#define fts3HashInit   sqlite3Fts3HashInit
#define fts3HashInsert sqlite3Fts3HashInsert
#define fts3HashFind   sqlite3Fts3HashFind
#define fts3HashClear  sqlite3Fts3HashClear

/*
** Macros for looping over all elements of a hash table.  The idiom is
** like this:
**
**   Fts3Hash h;
**   Fts3HashElem *p;
**   ...
**   for(p=fts3HashFirst(&h); p; p=fts3HashNext(p)){
**     SomeStructure *pData = fts3HashData(p);
**     // do something with pData
**   }
*/
#define fts3HashFirst(H)  ((H)->first)

Changes to ext/fts3/fts3_porter.c.

20
21
22
23
24
25
26

27
28
29
30
31
32
33
..
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68




69
70
71
72
73
74
75
..
89
90
91
92
93
94
95


96
97
98
99
100
101
102
...
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
...
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
...
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
...
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
**       (in which case SQLITE_CORE is not defined), or
**
**     * The FTS3 module is being built into the core of
**       SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
*/
#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)



#include <assert.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <ctype.h>

................................................................................
  int nInput;                  /* size of the input */
  int iOffset;                 /* current position in zInput */
  int iToken;                  /* index of next token to be returned */
  char *zToken;                /* storage for current token */
  int nAllocated;              /* space allocated to zToken buffer */
} porter_tokenizer_cursor;


/* Forward declaration */
static const sqlite3_tokenizer_module porterTokenizerModule;


/*
** Create a new tokenizer instance.
*/
static int porterCreate(
  int argc, const char * const *argv,
  sqlite3_tokenizer **ppTokenizer
){
  porter_tokenizer *t;




  t = (porter_tokenizer *) sqlite3_malloc(sizeof(*t));
  if( t==NULL ) return SQLITE_NOMEM;
  memset(t, 0, sizeof(*t));
  *ppTokenizer = &t->base;
  return SQLITE_OK;
}

................................................................................
*/
static int porterOpen(
  sqlite3_tokenizer *pTokenizer,         /* The tokenizer */
  const char *zInput, int nInput,        /* String to be tokenized */
  sqlite3_tokenizer_cursor **ppCursor    /* OUT: Tokenization cursor */
){
  porter_tokenizer_cursor *c;



  c = (porter_tokenizer_cursor *) sqlite3_malloc(sizeof(*c));
  if( c==NULL ) return SQLITE_NOMEM;

  c->zInput = zInput;
  if( zInput==0 ){
    c->nInput = 0;
................................................................................
/*
** Return TRUE if the word ends in a double consonant.
**
** The text is reversed here. So we are really looking at
** the first two characters of z[].
*/
static int doubleConsonant(const char *z){
  return isConsonant(z) && z[0]==z[1] && isConsonant(z+1);
}

/*
** Return TRUE if the word ends with three letters which
** are consonant-vowel-consonent and where the final consonant
** is not 'w', 'x', or 'y'.
**
** The word is reversed here.  So we are really checking the
** first three letters and the first one cannot be in [wxy].
*/
static int star_oh(const char *z){
  return
    z[0]!=0 && isConsonant(z) &&
    z[0]!='w' && z[0]!='x' && z[0]!='y' &&
    z[1]!=0 && isVowel(z+1) &&
    z[2]!=0 && isConsonant(z+2);
}

/*
** If the word ends with zFrom and xCond() is true for the stem
** of the word that preceeds the zFrom ending, then change the 
** ending to zTo.
**
................................................................................
** it contains digits) then word is truncated to 20 or 6 bytes
** by taking 10 or 3 bytes from the beginning and end.
*/
static void copy_stemmer(const char *zIn, int nIn, char *zOut, int *pnOut){
  int i, mx, j;
  int hasDigit = 0;
  for(i=0; i<nIn; i++){
    int c = zIn[i];
    if( c>='A' && c<='Z' ){
      zOut[i] = c - 'A' + 'a';
    }else{
      if( c>='0' && c<='9' ) hasDigit = 1;
      zOut[i] = c;
    }
  }
................................................................................
** copies the input into the input into the output with US-ASCII
** case folding.
**
** Stemming never increases the length of the word.  So there is
** no chance of overflowing the zOut buffer.
*/
static void porter_stemmer(const char *zIn, int nIn, char *zOut, int *pnOut){
  int i, j, c;
  char zReverse[28];
  char *z, *z2;
  if( nIn<3 || nIn>=sizeof(zReverse)-7 ){
    /* The word is too big or too small for the porter stemmer.
    ** Fallback to the copy stemmer */
    copy_stemmer(zIn, nIn, zOut, pnOut);
    return;
  }
  for(i=0, j=sizeof(zReverse)-6; i<nIn; i++, j--){
    c = zIn[i];
    if( c>='A' && c<='Z' ){
      zReverse[j] = c + 'a' - 'A';
    }else if( c>='a' && c<='z' ){
      zReverse[j] = c;
    }else{
      /* The use of a character not in [a-zA-Z] means that we fallback
      ** to the copy stemmer */
................................................................................
  if( m_gt_1(z) && z[0]=='l' && z[1]=='l' ){
    z++;
  }

  /* z[] is now the stemmed word in reverse order.  Flip it back
  ** around into forward order and return.
  */
  *pnOut = i = strlen(z);
  zOut[i] = 0;
  while( *z ){
    zOut[--i] = *(z++);
  }
}

/*







>







 







<
<
<
<









>
>
>
>







 







>
>







 







|












|

|
|







 







|







 







|









|







 







|







20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
..
50
51
52
53
54
55
56




57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
..
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
...
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
...
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
...
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
...
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
**       (in which case SQLITE_CORE is not defined), or
**
**     * The FTS3 module is being built into the core of
**       SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
*/
#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)

#include "fts3Int.h"

#include <assert.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <ctype.h>

................................................................................
  int nInput;                  /* size of the input */
  int iOffset;                 /* current position in zInput */
  int iToken;                  /* index of next token to be returned */
  char *zToken;                /* storage for current token */
  int nAllocated;              /* space allocated to zToken buffer */
} porter_tokenizer_cursor;






/*
** Create a new tokenizer instance.
*/
static int porterCreate(
  int argc, const char * const *argv,
  sqlite3_tokenizer **ppTokenizer
){
  porter_tokenizer *t;

  UNUSED_PARAMETER(argc);
  UNUSED_PARAMETER(argv);

  t = (porter_tokenizer *) sqlite3_malloc(sizeof(*t));
  if( t==NULL ) return SQLITE_NOMEM;
  memset(t, 0, sizeof(*t));
  *ppTokenizer = &t->base;
  return SQLITE_OK;
}

................................................................................
*/
static int porterOpen(
  sqlite3_tokenizer *pTokenizer,         /* The tokenizer */
  const char *zInput, int nInput,        /* String to be tokenized */
  sqlite3_tokenizer_cursor **ppCursor    /* OUT: Tokenization cursor */
){
  porter_tokenizer_cursor *c;

  UNUSED_PARAMETER(pTokenizer);

  c = (porter_tokenizer_cursor *) sqlite3_malloc(sizeof(*c));
  if( c==NULL ) return SQLITE_NOMEM;

  c->zInput = zInput;
  if( zInput==0 ){
    c->nInput = 0;
................................................................................
/*
** Return TRUE if the word ends in a double consonant.
**
** The text is reversed here. So we are really looking at
** the first two characters of z[].
*/
static int doubleConsonant(const char *z){
  return isConsonant(z) && z[0]==z[1];
}

/*
** Return TRUE if the word ends with three letters which
** are consonant-vowel-consonent and where the final consonant
** is not 'w', 'x', or 'y'.
**
** The word is reversed here.  So we are really checking the
** first three letters and the first one cannot be in [wxy].
*/
static int star_oh(const char *z){
  return
    isConsonant(z) &&
    z[0]!='w' && z[0]!='x' && z[0]!='y' &&
    isVowel(z+1) &&
    isConsonant(z+2);
}

/*
** If the word ends with zFrom and xCond() is true for the stem
** of the word that preceeds the zFrom ending, then change the 
** ending to zTo.
**
................................................................................
** it contains digits) then word is truncated to 20 or 6 bytes
** by taking 10 or 3 bytes from the beginning and end.
*/
static void copy_stemmer(const char *zIn, int nIn, char *zOut, int *pnOut){
  int i, mx, j;
  int hasDigit = 0;
  for(i=0; i<nIn; i++){
    char c = zIn[i];
    if( c>='A' && c<='Z' ){
      zOut[i] = c - 'A' + 'a';
    }else{
      if( c>='0' && c<='9' ) hasDigit = 1;
      zOut[i] = c;
    }
  }
................................................................................
** copies the input into the input into the output with US-ASCII
** case folding.
**
** Stemming never increases the length of the word.  So there is
** no chance of overflowing the zOut buffer.
*/
static void porter_stemmer(const char *zIn, int nIn, char *zOut, int *pnOut){
  int i, j;
  char zReverse[28];
  char *z, *z2;
  if( nIn<3 || nIn>=sizeof(zReverse)-7 ){
    /* The word is too big or too small for the porter stemmer.
    ** Fallback to the copy stemmer */
    copy_stemmer(zIn, nIn, zOut, pnOut);
    return;
  }
  for(i=0, j=sizeof(zReverse)-6; i<nIn; i++, j--){
    char c = zIn[i];
    if( c>='A' && c<='Z' ){
      zReverse[j] = c + 'a' - 'A';
    }else if( c>='a' && c<='z' ){
      zReverse[j] = c;
    }else{
      /* The use of a character not in [a-zA-Z] means that we fallback
      ** to the copy stemmer */
................................................................................
  if( m_gt_1(z) && z[0]=='l' && z[1]=='l' ){
    z++;
  }

  /* z[] is now the stemmed word in reverse order.  Flip it back
  ** around into forward order and return.
  */
  *pnOut = i = (int)strlen(z);
  zOut[i] = 0;
  while( *z ){
    zOut[--i] = *(z++);
  }
}

/*

Added ext/fts3/fts3_snippet.c.





























































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
/*
** 2009 Oct 23
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
******************************************************************************
*/

#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)

#include "fts3Int.h"
#include <string.h>
#include <assert.h>
#include <ctype.h>

typedef struct Snippet Snippet;

/*
** An instance of the following structure keeps track of generated
** matching-word offset information and snippets.
*/
struct Snippet {
  int nMatch;                     /* Total number of matches */
  int nAlloc;                     /* Space allocated for aMatch[] */
  struct snippetMatch {  /* One entry for each matching term */
    char snStatus;       /* Status flag for use while constructing snippets */
    short int nByte;     /* Number of bytes in the term */
    short int iCol;      /* The column that contains the match */
    short int iTerm;     /* The index in Query.pTerms[] of the matching term */
    int iToken;          /* The index of the matching document token */
    int iStart;          /* The offset to the first character of the term */
  } *aMatch;                      /* Points to space obtained from malloc */
  char *zOffset;                  /* Text rendering of aMatch[] */
  int nOffset;                    /* strlen(zOffset) */
  char *zSnippet;                 /* Snippet text */
  int nSnippet;                   /* strlen(zSnippet) */
};


/* It is not safe to call isspace(), tolower(), or isalnum() on
** hi-bit-set characters.  This is the same solution used in the
** tokenizer.
*/
static int fts3snippetIsspace(char c){
  return (c&0x80)==0 ? isspace(c) : 0;
}


/*
** A StringBuffer object holds a zero-terminated string that grows
** arbitrarily by appending.  Space to hold the string is obtained
** from sqlite3_malloc().  After any memory allocation failure, 
** StringBuffer.z is set to NULL and no further allocation is attempted.
*/
typedef struct StringBuffer {
  char *z;         /* Text of the string.  Space from malloc. */
  int nUsed;       /* Number bytes of z[] used, not counting \000 terminator */
  int nAlloc;      /* Bytes allocated for z[] */
} StringBuffer;


/*
** Initialize a new StringBuffer.
*/
static void fts3SnippetSbInit(StringBuffer *p){
  p->nAlloc = 100;
  p->nUsed = 0;
  p->z = sqlite3_malloc( p->nAlloc );
}

/*
** Append text to the string buffer.
*/
static void fts3SnippetAppend(StringBuffer *p, const char *zNew, int nNew){
  if( p->z==0 ) return;
  if( nNew<0 ) nNew = (int)strlen(zNew);
  if( p->nUsed + nNew >= p->nAlloc ){
    int nAlloc;
    char *zNew;

    nAlloc = p->nUsed + nNew + p->nAlloc;
    zNew = sqlite3_realloc(p->z, nAlloc);
    if( zNew==0 ){
      sqlite3_free(p->z);
      p->z = 0;
      return;
    }
    p->z = zNew;
    p->nAlloc = nAlloc;
  }
  memcpy(&p->z[p->nUsed], zNew, nNew);
  p->nUsed += nNew;
  p->z[p->nUsed] = 0;
}

/* If the StringBuffer ends in something other than white space, add a
** single space character to the end.
*/
static void fts3SnippetAppendWhiteSpace(StringBuffer *p){
  if( p->z && p->nUsed && !fts3snippetIsspace(p->z[p->nUsed-1]) ){
    fts3SnippetAppend(p, " ", 1);
  }
}

/* Remove white space from the end of the StringBuffer */
static void fts3SnippetTrimWhiteSpace(StringBuffer *p){
  if( p->z ){
    while( p->nUsed && fts3snippetIsspace(p->z[p->nUsed-1]) ){
      p->nUsed--;
    }
    p->z[p->nUsed] = 0;
  }
}

/* 
** Release all memory associated with the Snippet structure passed as
** an argument.
*/
static void fts3SnippetFree(Snippet *p){
  if( p ){
    sqlite3_free(p->aMatch);
    sqlite3_free(p->zOffset);
    sqlite3_free(p->zSnippet);
    sqlite3_free(p);
  }
}

/*
** Append a single entry to the p->aMatch[] log.
*/
static int snippetAppendMatch(
  Snippet *p,               /* Append the entry to this snippet */
  int iCol, int iTerm,      /* The column and query term */
  int iToken,               /* Matching token in document */
  int iStart, int nByte     /* Offset and size of the match */
){
  int i;
  struct snippetMatch *pMatch;
  if( p->nMatch+1>=p->nAlloc ){
    struct snippetMatch *pNew;
    p->nAlloc = p->nAlloc*2 + 10;
    pNew = sqlite3_realloc(p->aMatch, p->nAlloc*sizeof(p->aMatch[0]) );
    if( pNew==0 ){
      p->aMatch = 0;
      p->nMatch = 0;
      p->nAlloc = 0;
      return SQLITE_NOMEM;
    }
    p->aMatch = pNew;
  }
  i = p->nMatch++;
  pMatch = &p->aMatch[i];
  pMatch->iCol = (short)iCol;
  pMatch->iTerm = (short)iTerm;
  pMatch->iToken = iToken;
  pMatch->iStart = iStart;
  pMatch->nByte = (short)nByte;
  return SQLITE_OK;
}

/*
** Sizing information for the circular buffer used in snippetOffsetsOfColumn()
*/
#define FTS3_ROTOR_SZ   (32)
#define FTS3_ROTOR_MASK (FTS3_ROTOR_SZ-1)

/*
** Function to iterate through the tokens of a compiled expression.
**
** Except, skip all tokens on the right-hand side of a NOT operator.
** This function is used to find tokens as part of snippet and offset
** generation and we do nt want snippets and offsets to report matches
** for tokens on the RHS of a NOT.
*/
static int fts3NextExprToken(Fts3Expr **ppExpr, int *piToken){
  Fts3Expr *p = *ppExpr;
  int iToken = *piToken;
  if( iToken<0 ){
    /* In this case the expression p is the root of an expression tree.
    ** Move to the first token in the expression tree.
    */
    while( p->pLeft ){
      p = p->pLeft;
    }
    iToken = 0;
  }else{
    assert(p && p->eType==FTSQUERY_PHRASE );
    if( iToken<(p->pPhrase->nToken-1) ){
      iToken++;
    }else{
      iToken = 0;
      while( p->pParent && p->pParent->pLeft!=p ){
        assert( p->pParent->pRight==p );
        p = p->pParent;
      }
      p = p->pParent;
      if( p ){
        assert( p->pRight!=0 );
        p = p->pRight;
        while( p->pLeft ){
          p = p->pLeft;
        }
      }
    }
  }

  *ppExpr = p;
  *piToken = iToken;
  return p?1:0;
}

/*
** Return TRUE if the expression node pExpr is located beneath the
** RHS of a NOT operator.
*/
static int fts3ExprBeneathNot(Fts3Expr *p){
  Fts3Expr *pParent;
  while( p ){
    pParent = p->pParent;
    if( pParent && pParent->eType==FTSQUERY_NOT && pParent->pRight==p ){
      return 1;
    }
    p = pParent;
  }
  return 0;
}

/*
** Add entries to pSnippet->aMatch[] for every match that occurs against
** document zDoc[0..nDoc-1] which is stored in column iColumn.
*/
static int snippetOffsetsOfColumn(
  Fts3Cursor *pCur,         /* The fulltest search cursor */
  Snippet *pSnippet,             /* The Snippet object to be filled in */
  int iColumn,                   /* Index of fulltext table column */
  const char *zDoc,              /* Text of the fulltext table column */
  int nDoc                       /* Length of zDoc in bytes */
){
  const sqlite3_tokenizer_module *pTModule;  /* The tokenizer module */
  sqlite3_tokenizer *pTokenizer;             /* The specific tokenizer */
  sqlite3_tokenizer_cursor *pTCursor;        /* Tokenizer cursor */
  Fts3Table *pVtab;                /* The full text index */
  int nColumn;                         /* Number of columns in the index */
  int i, j;                            /* Loop counters */
  int rc;                              /* Return code */
  unsigned int match, prevMatch;       /* Phrase search bitmasks */
  const char *zToken;                  /* Next token from the tokenizer */
  int nToken;                          /* Size of zToken */
  int iBegin, iEnd, iPos;              /* Offsets of beginning and end */

  /* The following variables keep a circular buffer of the last
  ** few tokens */
  unsigned int iRotor = 0;             /* Index of current token */
  int iRotorBegin[FTS3_ROTOR_SZ];      /* Beginning offset of token */
  int iRotorLen[FTS3_ROTOR_SZ];        /* Length of token */

  pVtab =  (Fts3Table *)pCur->base.pVtab;
  nColumn = pVtab->nColumn;
  pTokenizer = pVtab->pTokenizer;
  pTModule = pTokenizer->pModule;
  rc = pTModule->xOpen(pTokenizer, zDoc, nDoc, &pTCursor);
  if( rc ) return rc;
  pTCursor->pTokenizer = pTokenizer;

  prevMatch = 0;
  while( (rc = pTModule->xNext(pTCursor, &zToken, &nToken,
                               &iBegin, &iEnd, &iPos))==SQLITE_OK ){
    Fts3Expr *pIter = pCur->pExpr;
    int iIter = -1;
    iRotorBegin[iRotor&FTS3_ROTOR_MASK] = iBegin;
    iRotorLen[iRotor&FTS3_ROTOR_MASK] = iEnd-iBegin;
    match = 0;
    for(i=0; i<(FTS3_ROTOR_SZ-1) && fts3NextExprToken(&pIter, &iIter); i++){
      int nPhrase;                    /* Number of tokens in current phrase */
      struct PhraseToken *pToken;     /* Current token */
      int iCol;                       /* Column index */

      if( fts3ExprBeneathNot(pIter) ) continue;
      nPhrase = pIter->pPhrase->nToken;
      pToken = &pIter->pPhrase->aToken[iIter];
      iCol = pIter->pPhrase->iColumn;
      if( iCol>=0 && iCol<nColumn && iCol!=iColumn ) continue;
      if( pToken->n>nToken ) continue;
      if( !pToken->isPrefix && pToken->n<nToken ) continue;
      assert( pToken->n<=nToken );
      if( memcmp(pToken->z, zToken, pToken->n) ) continue;
      if( iIter>0 && (prevMatch & (1<<i))==0 ) continue;
      match |= 1<<i;
      if( i==(FTS3_ROTOR_SZ-2) || nPhrase==iIter+1 ){
        for(j=nPhrase-1; j>=0; j--){
          int k = (iRotor-j) & FTS3_ROTOR_MASK;
          rc = snippetAppendMatch(pSnippet, iColumn, i-j, iPos-j,
                                  iRotorBegin[k], iRotorLen[k]);
          if( rc ) goto end_offsets_of_column;
        }
      }
    }
    prevMatch = match<<1;
    iRotor++;
  }
end_offsets_of_column:
  pTModule->xClose(pTCursor);  
  return rc==SQLITE_DONE ? SQLITE_OK : rc;
}

/*
** Remove entries from the pSnippet structure to account for the NEAR
** operator. When this is called, pSnippet contains the list of token 
** offsets produced by treating all NEAR operators as AND operators.
** This function removes any entries that should not be present after
** accounting for the NEAR restriction. For example, if the queried
** document is:
**
**     "A B C D E A"
**
** and the query is:
** 
**     A NEAR/0 E
**
** then when this function is called the Snippet contains token offsets
** 0, 4 and 5. This function removes the "0" entry (because the first A
** is not near enough to an E).
**
** When this function is called, the value pointed to by parameter piLeft is
** the integer id of the left-most token in the expression tree headed by
** pExpr. This function increments *piLeft by the total number of tokens
** in the expression tree headed by pExpr.
**
** Return 1 if any trimming occurs.  Return 0 if no trimming is required.
*/
static int trimSnippetOffsets(
  Fts3Expr *pExpr,      /* The search expression */
  Snippet *pSnippet,    /* The set of snippet offsets to be trimmed */
  int *piLeft           /* Index of left-most token in pExpr */
){
  if( pExpr ){
    if( trimSnippetOffsets(pExpr->pLeft, pSnippet, piLeft) ){
      return 1;
    }

    switch( pExpr->eType ){
      case FTSQUERY_PHRASE:
        *piLeft += pExpr->pPhrase->nToken;
        break;
      case FTSQUERY_NEAR: {
        /* The right-hand-side of a NEAR operator is always a phrase. The
        ** left-hand-side is either a phrase or an expression tree that is 
        ** itself headed by a NEAR operator. The following initializations
        ** set local variable iLeft to the token number of the left-most
        ** token in the right-hand phrase, and iRight to the right most
        ** token in the same phrase. For example, if we had:
        **
        **     <col> MATCH '"abc def" NEAR/2 "ghi jkl"'
        **
        ** then iLeft will be set to 2 (token number of ghi) and nToken will
        ** be set to 4.
        */
        Fts3Expr *pLeft = pExpr->pLeft;
        Fts3Expr *pRight = pExpr->pRight;
        int iLeft = *piLeft;
        int nNear = pExpr->nNear;
        int nToken = pRight->pPhrase->nToken;
        int jj, ii;
        if( pLeft->eType==FTSQUERY_NEAR ){
          pLeft = pLeft->pRight;
        }
        assert( pRight->eType==FTSQUERY_PHRASE );
        assert( pLeft->eType==FTSQUERY_PHRASE );
        nToken += pLeft->pPhrase->nToken;

        for(ii=0; ii<pSnippet->nMatch; ii++){
          struct snippetMatch *p = &pSnippet->aMatch[ii];
          if( p->iTerm==iLeft ){
            int isOk = 0;
            /* Snippet ii is an occurence of query term iLeft in the document.
            ** It occurs at position (p->iToken) of the document. We now
            ** search for an instance of token (iLeft-1) somewhere in the 
            ** range (p->iToken - nNear)...(p->iToken + nNear + nToken) within 
            ** the set of snippetMatch structures. If one is found, proceed. 
            ** If one cannot be found, then remove snippets ii..(ii+N-1) 
            ** from the matching snippets, where N is the number of tokens 
            ** in phrase pRight->pPhrase.
            */
            for(jj=0; isOk==0 && jj<pSnippet->nMatch; jj++){
              struct snippetMatch *p2 = &pSnippet->aMatch[jj];
              if( p2->iTerm==(iLeft-1) ){
                if( p2->iToken>=(p->iToken-nNear-1) 
                 && p2->iToken<(p->iToken+nNear+nToken) 
                ){
                  isOk = 1;
                }
              }
            }
            if( !isOk ){
              int kk;
              for(kk=0; kk<pRight->pPhrase->nToken; kk++){
                pSnippet->aMatch[kk+ii].iTerm = -2;
              }
              return 1;
            }
          }
          if( p->iTerm==(iLeft-1) ){
            int isOk = 0;
            for(jj=0; isOk==0 && jj<pSnippet->nMatch; jj++){
              struct snippetMatch *p2 = &pSnippet->aMatch[jj];
              if( p2->iTerm==iLeft ){
                if( p2->iToken<=(p->iToken+nNear+1) 
                 && p2->iToken>(p->iToken-nNear-nToken) 
                ){
                  isOk = 1;
                }
              }
            }
            if( !isOk ){
              int kk;
              for(kk=0; kk<pLeft->pPhrase->nToken; kk++){
                pSnippet->aMatch[ii-kk].iTerm = -2;
              }
              return 1;
            }
          }
        }
        break;
      }
    }

    if( trimSnippetOffsets(pExpr->pRight, pSnippet, piLeft) ){
      return 1;
    }
  }
  return 0;
}

/*
** Compute all offsets for the current row of the query.  
** If the offsets have already been computed, this routine is a no-op.
*/
static int snippetAllOffsets(Fts3Cursor *pCsr, Snippet **ppSnippet){
  Fts3Table *p = (Fts3Table *)pCsr->base.pVtab;  /* The FTS3 virtual table */
  int nColumn;           /* Number of columns.  Docid does count */
  int iColumn;           /* Index of of a column */
  int i;                 /* Loop index */
  int iFirst;            /* First column to search */
  int iLast;             /* Last coumn to search */
  int iTerm = 0;
  Snippet *pSnippet;
  int rc = SQLITE_OK;

  if( pCsr->pExpr==0 ){
    return SQLITE_OK;
  }

  pSnippet = (Snippet *)sqlite3_malloc(sizeof(Snippet));
  *ppSnippet = pSnippet;
  if( !pSnippet ){
    return SQLITE_NOMEM;
  }
  memset(pSnippet, 0, sizeof(Snippet));

  nColumn = p->nColumn;
  iColumn = (pCsr->eSearch - 2);
  if( iColumn<0 || iColumn>=nColumn ){
    /* Look for matches over all columns of the full-text index */
    iFirst = 0;
    iLast = nColumn-1;
  }else{
    /* Look for matches in the iColumn-th column of the index only */
    iFirst = iColumn;
    iLast = iColumn;
  }
  for(i=iFirst; rc==SQLITE_OK && i<=iLast; i++){
    const char *zDoc;
    int nDoc;
    zDoc = (const char*)sqlite3_column_text(pCsr->pStmt, i+1);
    nDoc = sqlite3_column_bytes(pCsr->pStmt, i+1);
    if( zDoc==0 && sqlite3_column_type(pCsr->pStmt, i+1)!=SQLITE_NULL ){
      rc = SQLITE_NOMEM;
    }else{
      rc = snippetOffsetsOfColumn(pCsr, pSnippet, i, zDoc, nDoc);
    }
  }

  while( trimSnippetOffsets(pCsr->pExpr, pSnippet, &iTerm) ){
    iTerm = 0;
  }

  return rc;
}

/*
** Convert the information in the aMatch[] array of the snippet
** into the string zOffset[0..nOffset-1]. This string is used as
** the return of the SQL offsets() function.
*/
static void snippetOffsetText(Snippet *p){
  int i;
  int cnt = 0;
  StringBuffer sb;
  char zBuf[200];
  if( p->zOffset ) return;
  fts3SnippetSbInit(&sb);
  for(i=0; i<p->nMatch; i++){
    struct snippetMatch *pMatch = &p->aMatch[i];
    if( pMatch->iTerm>=0 ){
      /* If snippetMatch.iTerm is less than 0, then the match was 
      ** discarded as part of processing the NEAR operator (see the 
      ** trimSnippetOffsetsForNear() function for details). Ignore 
      ** it in this case
      */
      zBuf[0] = ' ';
      sqlite3_snprintf(sizeof(zBuf)-1, &zBuf[cnt>0], "%d %d %d %d",
          pMatch->iCol, pMatch->iTerm, pMatch->iStart, pMatch->nByte);
      fts3SnippetAppend(&sb, zBuf, -1);
      cnt++;
    }
  }
  p->zOffset = sb.z;
  p->nOffset = sb.z ? sb.nUsed : 0;
}

/*
** zDoc[0..nDoc-1] is phrase of text.  aMatch[0..nMatch-1] are a set
** of matching words some of which might be in zDoc.  zDoc is column
** number iCol.
**
** iBreak is suggested spot in zDoc where we could begin or end an
** excerpt.  Return a value similar to iBreak but possibly adjusted
** to be a little left or right so that the break point is better.
*/
static int wordBoundary(
  int iBreak,                   /* The suggested break point */
  const char *zDoc,             /* Document text */
  int nDoc,                     /* Number of bytes in zDoc[] */
  struct snippetMatch *aMatch,  /* Matching words */
  int nMatch,                   /* Number of entries in aMatch[] */
  int iCol                      /* The column number for zDoc[] */
){
  int i;
  if( iBreak<=10 ){
    return 0;
  }
  if( iBreak>=nDoc-10 ){
    return nDoc;
  }
  for(i=0; ALWAYS(i<nMatch) && aMatch[i].iCol<iCol; i++){}
  while( i<nMatch && aMatch[i].iStart+aMatch[i].nByte<iBreak ){ i++; }
  if( i<nMatch ){
    if( aMatch[i].iStart<iBreak+10 ){
      return aMatch[i].iStart;
    }
    if( i>0 && aMatch[i-1].iStart+aMatch[i-1].nByte>=iBreak ){
      return aMatch[i-1].iStart;
    }
  }
  for(i=1; i<=10; i++){
    if( fts3snippetIsspace(zDoc[iBreak-i]) ){
      return iBreak - i + 1;
    }
    if( fts3snippetIsspace(zDoc[iBreak+i]) ){
      return iBreak + i + 1;
    }
  }
  return iBreak;
}



/*
** Allowed values for Snippet.aMatch[].snStatus
*/
#define SNIPPET_IGNORE  0   /* It is ok to omit this match from the snippet */
#define SNIPPET_DESIRED 1   /* We want to include this match in the snippet */

/*
** Generate the text of a snippet.
*/
static void snippetText(
  Fts3Cursor *pCursor,   /* The cursor we need the snippet for */
  Snippet *pSnippet,
  const char *zStartMark,     /* Markup to appear before each match */
  const char *zEndMark,       /* Markup to appear after each match */
  const char *zEllipsis       /* Ellipsis mark */
){
  int i, j;
  struct snippetMatch *aMatch;
  int nMatch;
  int nDesired;
  StringBuffer sb;
  int tailCol;
  int tailOffset;
  int iCol;
  int nDoc;
  const char *zDoc;
  int iStart, iEnd;
  int tailEllipsis = 0;
  int iMatch;
  

  sqlite3_free(pSnippet->zSnippet);
  pSnippet->zSnippet = 0;
  aMatch = pSnippet->aMatch;
  nMatch = pSnippet->nMatch;
  fts3SnippetSbInit(&sb);

  for(i=0; i<nMatch; i++){
    aMatch[i].snStatus = SNIPPET_IGNORE;
  }
  nDesired = 0;
  for(i=0; i<FTS3_ROTOR_SZ; i++){
    for(j=0; j<nMatch; j++){
      if( aMatch[j].iTerm==i ){
        aMatch[j].snStatus = SNIPPET_DESIRED;
        nDesired++;
        break;
      }
    }
  }

  iMatch = 0;
  tailCol = -1;
  tailOffset = 0;
  for(i=0; i<nMatch && nDesired>0; i++){
    if( aMatch[i].snStatus!=SNIPPET_DESIRED ) continue;
    nDesired--;
    iCol = aMatch[i].iCol;
    zDoc = (const char*)sqlite3_column_text(pCursor->pStmt, iCol+1);
    nDoc = sqlite3_column_bytes(pCursor->pStmt, iCol+1);
    iStart = aMatch[i].iStart - 40;
    iStart = wordBoundary(iStart, zDoc, nDoc, aMatch, nMatch, iCol);
    if( iStart<=10 ){
      iStart = 0;
    }
    if( iCol==tailCol && iStart<=tailOffset+20 ){
      iStart = tailOffset;
    }
    if( (iCol!=tailCol && tailCol>=0) || iStart!=tailOffset ){
      fts3SnippetTrimWhiteSpace(&sb);
      fts3SnippetAppendWhiteSpace(&sb);
      fts3SnippetAppend(&sb, zEllipsis, -1);
      fts3SnippetAppendWhiteSpace(&sb);
    }
    iEnd = aMatch[i].iStart + aMatch[i].nByte + 40;
    iEnd = wordBoundary(iEnd, zDoc, nDoc, aMatch, nMatch, iCol);
    if( iEnd>=nDoc-10 ){
      iEnd = nDoc;
      tailEllipsis = 0;
    }else{
      tailEllipsis = 1;
    }
    while( iMatch<nMatch && aMatch[iMatch].iCol<iCol ){ iMatch++; }
    while( iStart<iEnd ){
      while( iMatch<nMatch && aMatch[iMatch].iStart<iStart
             && aMatch[iMatch].iCol<=iCol ){
        iMatch++;
      }
      if( iMatch<nMatch && aMatch[iMatch].iStart<iEnd
             && aMatch[iMatch].iCol==iCol ){
        fts3SnippetAppend(&sb, &zDoc[iStart], aMatch[iMatch].iStart - iStart);
        iStart = aMatch[iMatch].iStart;
        fts3SnippetAppend(&sb, zStartMark, -1);
        fts3SnippetAppend(&sb, &zDoc[iStart], aMatch[iMatch].nByte);
        fts3SnippetAppend(&sb, zEndMark, -1);
        iStart += aMatch[iMatch].nByte;
        for(j=iMatch+1; j<nMatch; j++){
          if( aMatch[j].iTerm==aMatch[iMatch].iTerm
              && aMatch[j].snStatus==SNIPPET_DESIRED ){
            nDesired--;
            aMatch[j].snStatus = SNIPPET_IGNORE;
          }
        }
      }else{
        fts3SnippetAppend(&sb, &zDoc[iStart], iEnd - iStart);
        iStart = iEnd;
      }
    }
    tailCol = iCol;
    tailOffset = iEnd;
  }
  fts3SnippetTrimWhiteSpace(&sb);
  if( tailEllipsis ){
    fts3SnippetAppendWhiteSpace(&sb);
    fts3SnippetAppend(&sb, zEllipsis, -1);
  }
  pSnippet->zSnippet = sb.z;
  pSnippet->nSnippet = sb.z ? sb.nUsed : 0;
}

void sqlite3Fts3Offsets(
  sqlite3_context *pCtx,          /* SQLite function call context */
  Fts3Cursor *pCsr                /* Cursor object */
){
  Snippet *p;                     /* Snippet structure */
  int rc = snippetAllOffsets(pCsr, &p);
  if( rc==SQLITE_OK ){
    snippetOffsetText(p);
    if( p->zOffset ){
      sqlite3_result_text(pCtx, p->zOffset, p->nOffset, SQLITE_TRANSIENT);
    }else{
      sqlite3_result_error_nomem(pCtx);
    }
  }else{
    sqlite3_result_error_nomem(pCtx);
  }
  fts3SnippetFree(p);
}

void sqlite3Fts3Snippet(
  sqlite3_context *pCtx,          /* SQLite function call context */
  Fts3Cursor *pCsr,               /* Cursor object */
  const char *zStart,             /* Snippet start text - "<b>" */
  const char *zEnd,               /* Snippet end text - "</b>" */
  const char *zEllipsis           /* Snippet ellipsis text - "<b>...</b>" */
){
  Snippet *p;                     /* Snippet structure */
  int rc = snippetAllOffsets(pCsr, &p);
  if( rc==SQLITE_OK ){
    snippetText(pCsr, p, zStart, zEnd, zEllipsis);
    if( p->zSnippet ){
      sqlite3_result_text(pCtx, p->zSnippet, p->nSnippet, SQLITE_TRANSIENT);
    }else{
      sqlite3_result_error_nomem(pCtx);
    }
  }else{
    sqlite3_result_error_nomem(pCtx);
  }
  fts3SnippetFree(p);
}

#endif

Changes to ext/fts3/fts3_tokenizer.c.

26
27
28
29
30
31
32
33
34
35


36
37
38
39
40
41
42
..
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
..
92
93
94
95
96
97
98

























































































































99
100
101
102
103
104
105
...
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
...
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
...
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
...
289
290
291
292
293
294
295



296
297
298
299
300
301
302
...
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
...
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)

#include "sqlite3ext.h"
#ifndef SQLITE_CORE
  SQLITE_EXTENSION_INIT1
#endif

#include "fts3_hash.h"
#include "fts3_tokenizer.h"
#include <assert.h>



/*
** Implementation of the SQL scalar function for accessing the underlying 
** hash table. This function may be called as follows:
**
**   SELECT <function-name>(<key-name>);
**   SELECT <function-name>(<key-name>, <pointer>);
................................................................................
** to string <key-name> (after the hash-table is updated, if applicable).
*/
static void scalarFunc(
  sqlite3_context *context,
  int argc,
  sqlite3_value **argv
){
  fts3Hash *pHash;
  void *pPtr = 0;
  const unsigned char *zName;
  int nName;

  assert( argc==1 || argc==2 );

  pHash = (fts3Hash *)sqlite3_user_data(context);

  zName = sqlite3_value_text(argv[0]);
  nName = sqlite3_value_bytes(argv[0])+1;

  if( argc==2 ){
    void *pOld;
    int n = sqlite3_value_bytes(argv[1]);
................................................................................
      sqlite3_free(zErr);
      return;
    }
  }

  sqlite3_result_blob(context, (void *)&pPtr, sizeof(pPtr), SQLITE_TRANSIENT);
}


























































































































#ifdef SQLITE_TEST

#include <tcl.h>
#include <string.h>

/*
................................................................................
**   
*/
static void testFunc(
  sqlite3_context *context,
  int argc,
  sqlite3_value **argv
){
  fts3Hash *pHash;
  sqlite3_tokenizer_module *p;
  sqlite3_tokenizer *pTokenizer = 0;
  sqlite3_tokenizer_cursor *pCsr = 0;

  const char *zErr = 0;

  const char *zName;
................................................................................
  nInput = sqlite3_value_bytes(argv[argc-1]);
  zInput = (const char *)sqlite3_value_text(argv[argc-1]);

  if( argc==3 ){
    zArg = (const char *)sqlite3_value_text(argv[1]);
  }

  pHash = (fts3Hash *)sqlite3_user_data(context);
  p = (sqlite3_tokenizer_module *)sqlite3Fts3HashFind(pHash, zName, nName+1);

  if( !p ){
    char *zErr = sqlite3_mprintf("unknown tokenizer: %s", zName);
    sqlite3_result_error(context, zErr, -1);
    sqlite3_free(zErr);
    return;
................................................................................
  if( rc!=SQLITE_OK ){
    return rc;
  }

  sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC);
  if( SQLITE_ROW==sqlite3_step(pStmt) ){
    if( sqlite3_column_type(pStmt, 0)==SQLITE_BLOB ){
      memcpy(pp, sqlite3_column_blob(pStmt, 0), sizeof(*pp));
    }
  }

  return sqlite3_finalize(pStmt);
}

void sqlite3Fts3SimpleTokenizerModule(sqlite3_tokenizer_module const**ppModule);
................................................................................
  int argc,
  sqlite3_value **argv
){
  int rc;
  const sqlite3_tokenizer_module *p1;
  const sqlite3_tokenizer_module *p2;
  sqlite3 *db = (sqlite3 *)sqlite3_user_data(context);




  /* Test the query function */
  sqlite3Fts3SimpleTokenizerModule(&p1);
  rc = queryTokenizer(db, "simple", &p2);
  assert( rc==SQLITE_OK );
  assert( p1==p2 );
  rc = queryTokenizer(db, "nosuchtokenizer", &p2);
................................................................................
** provide read/write access to the contents of *pHash.
**
** The third argument to this function, zName, is used as the name
** of both the scalar and, if created, the virtual table.
*/
int sqlite3Fts3InitHashTable(
  sqlite3 *db, 
  fts3Hash *pHash, 
  const char *zName
){
  int rc = SQLITE_OK;
  void *p = (void *)pHash;
  const int any = SQLITE_ANY;
  char *zTest = 0;
  char *zTest2 = 0;
................................................................................
  zTest = sqlite3_mprintf("%s_test", zName);
  zTest2 = sqlite3_mprintf("%s_internal_test", zName);
  if( !zTest || !zTest2 ){
    rc = SQLITE_NOMEM;
  }
#endif

  if( rc!=SQLITE_OK
   || (rc = sqlite3_create_function(db, zName, 1, any, p, scalarFunc, 0, 0))
   || (rc = sqlite3_create_function(db, zName, 2, any, p, scalarFunc, 0, 0))
#ifdef SQLITE_TEST
   || (rc = sqlite3_create_function(db, zTest, 2, any, p, testFunc, 0, 0))
   || (rc = sqlite3_create_function(db, zTest, 3, any, p, testFunc, 0, 0))
   || (rc = sqlite3_create_function(db, zTest2, 0, any, pdb, intTestFunc, 0, 0))
#endif
  );

  sqlite3_free(zTest);
  sqlite3_free(zTest2);
  return rc;
}

#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */







|
<

>
>







 







|






|







 







>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







 







|







 







|







 







|







 







>
>
>







 







|







 







|
|
|

|
|
|









26
27
28
29
30
31
32
33

34
35
36
37
38
39
40
41
42
43
..
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
..
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
...
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
...
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
...
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
...
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
...
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
...
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)

#include "sqlite3ext.h"
#ifndef SQLITE_CORE
  SQLITE_EXTENSION_INIT1
#endif

#include "fts3Int.h"

#include <assert.h>
#include <ctype.h>
#include <string.h>

/*
** Implementation of the SQL scalar function for accessing the underlying 
** hash table. This function may be called as follows:
**
**   SELECT <function-name>(<key-name>);
**   SELECT <function-name>(<key-name>, <pointer>);
................................................................................
** to string <key-name> (after the hash-table is updated, if applicable).
*/
static void scalarFunc(
  sqlite3_context *context,
  int argc,
  sqlite3_value **argv
){
  Fts3Hash *pHash;
  void *pPtr = 0;
  const unsigned char *zName;
  int nName;

  assert( argc==1 || argc==2 );

  pHash = (Fts3Hash *)sqlite3_user_data(context);

  zName = sqlite3_value_text(argv[0]);
  nName = sqlite3_value_bytes(argv[0])+1;

  if( argc==2 ){
    void *pOld;
    int n = sqlite3_value_bytes(argv[1]);
................................................................................
      sqlite3_free(zErr);
      return;
    }
  }

  sqlite3_result_blob(context, (void *)&pPtr, sizeof(pPtr), SQLITE_TRANSIENT);
}

static int fts3IsIdChar(char c){
  static const char isFtsIdChar[] = {
      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  /* 0x */
      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  /* 1x */
      0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,  /* 2x */
      1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,  /* 3x */
      0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,  /* 4x */
      1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,  /* 5x */
      0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,  /* 6x */
      1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0,  /* 7x */
  };
  return (c&0x80 || isFtsIdChar[(int)(c)]);
}

const char *sqlite3Fts3NextToken(const char *zStr, int *pn){
  const char *z1;
  const char *z2 = 0;

  /* Find the start of the next token. */
  z1 = zStr;
  while( z2==0 ){
    switch( *z1 ){
      case '\0': return 0;        /* No more tokens here */
      case '\'':
      case '"':
      case '`': {
        z2 = &z1[1];
        while( *z2 && (z2[0]!=*z1 || z2[1]==*z1) ) z2++;
        if( *z2 ) z2++;
        break;
      }
      case '[':
        z2 = &z1[1];
        while( *z2 && z2[0]!=']' ) z2++;
        if( *z2 ) z2++;
        break;

      default:
        if( fts3IsIdChar(*z1) ){
          z2 = &z1[1];
          while( fts3IsIdChar(*z2) ) z2++;
        }else{
          z1++;
        }
    }
  }

  *pn = (int)(z2-z1);
  return z1;
}

int sqlite3Fts3InitTokenizer(
  Fts3Hash *pHash,                /* Tokenizer hash table */
  const char *zArg,               /* Possible tokenizer specification */
  sqlite3_tokenizer **ppTok,      /* OUT: Tokenizer (if applicable) */
  const char **pzTokenizer,       /* OUT: Set to zArg if is tokenizer */
  char **pzErr                    /* OUT: Set to malloced error message */
){
  int rc;
  char *z = (char *)zArg;
  int n;
  char *zCopy;
  char *zEnd;                     /* Pointer to nul-term of zCopy */
  sqlite3_tokenizer_module *m;

  if( !z ){
    zCopy = sqlite3_mprintf("simple");
  }else{
    if( sqlite3_strnicmp(z, "tokenize", 8) || fts3IsIdChar(z[8])){
      return SQLITE_OK;
    }
    zCopy = sqlite3_mprintf("%s", &z[8]);
    *pzTokenizer = zArg;
  }
  if( !zCopy ){
    return SQLITE_NOMEM;
  }

  zEnd = &zCopy[strlen(zCopy)];

  z = (char *)sqlite3Fts3NextToken(zCopy, &n);
  z[n] = '\0';
  sqlite3Fts3Dequote(z);

  m = (sqlite3_tokenizer_module *)sqlite3Fts3HashFind(pHash, z, (int)strlen(z)+1);
  if( !m ){
    *pzErr = sqlite3_mprintf("unknown tokenizer: %s", z);
    rc = SQLITE_ERROR;
  }else{
    char const **aArg = 0;
    int iArg = 0;
    z = &z[n+1];
    while( z<zEnd && (NULL!=(z = (char *)sqlite3Fts3NextToken(z, &n))) ){
      int nNew = sizeof(char *)*(iArg+1);
      char const **aNew = (const char **)sqlite3_realloc((void *)aArg, nNew);
      if( !aNew ){
        sqlite3_free(zCopy);
        sqlite3_free((void *)aArg);
        return SQLITE_NOMEM;
      }
      aArg = aNew;
      aArg[iArg++] = z;
      z[n] = '\0';
      sqlite3Fts3Dequote(z);
      z = &z[n+1];
    }
    rc = m->xCreate(iArg, aArg, ppTok);
    assert( rc!=SQLITE_OK || *ppTok );
    if( rc!=SQLITE_OK ){
      *pzErr = sqlite3_mprintf("unknown tokenizer");
    }else{
      (*ppTok)->pModule = m; 
    }
    sqlite3_free((void *)aArg);
  }

  sqlite3_free(zCopy);
  return rc;
}


#ifdef SQLITE_TEST

#include <tcl.h>
#include <string.h>

/*
................................................................................
**   
*/
static void testFunc(
  sqlite3_context *context,
  int argc,
  sqlite3_value **argv
){
  Fts3Hash *pHash;
  sqlite3_tokenizer_module *p;
  sqlite3_tokenizer *pTokenizer = 0;
  sqlite3_tokenizer_cursor *pCsr = 0;

  const char *zErr = 0;

  const char *zName;
................................................................................
  nInput = sqlite3_value_bytes(argv[argc-1]);
  zInput = (const char *)sqlite3_value_text(argv[argc-1]);

  if( argc==3 ){
    zArg = (const char *)sqlite3_value_text(argv[1]);
  }

  pHash = (Fts3Hash *)sqlite3_user_data(context);
  p = (sqlite3_tokenizer_module *)sqlite3Fts3HashFind(pHash, zName, nName+1);

  if( !p ){
    char *zErr = sqlite3_mprintf("unknown tokenizer: %s", zName);
    sqlite3_result_error(context, zErr, -1);
    sqlite3_free(zErr);
    return;
................................................................................
  if( rc!=SQLITE_OK ){
    return rc;
  }

  sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC);
  if( SQLITE_ROW==sqlite3_step(pStmt) ){
    if( sqlite3_column_type(pStmt, 0)==SQLITE_BLOB ){
      memcpy((void *)pp, sqlite3_column_blob(pStmt, 0), sizeof(*pp));
    }
  }

  return sqlite3_finalize(pStmt);
}

void sqlite3Fts3SimpleTokenizerModule(sqlite3_tokenizer_module const**ppModule);
................................................................................
  int argc,
  sqlite3_value **argv
){
  int rc;
  const sqlite3_tokenizer_module *p1;
  const sqlite3_tokenizer_module *p2;
  sqlite3 *db = (sqlite3 *)sqlite3_user_data(context);

  UNUSED_PARAMETER(argc);
  UNUSED_PARAMETER(argv);

  /* Test the query function */
  sqlite3Fts3SimpleTokenizerModule(&p1);
  rc = queryTokenizer(db, "simple", &p2);
  assert( rc==SQLITE_OK );
  assert( p1==p2 );
  rc = queryTokenizer(db, "nosuchtokenizer", &p2);
................................................................................
** provide read/write access to the contents of *pHash.
**
** The third argument to this function, zName, is used as the name
** of both the scalar and, if created, the virtual table.
*/
int sqlite3Fts3InitHashTable(
  sqlite3 *db, 
  Fts3Hash *pHash, 
  const char *zName
){
  int rc = SQLITE_OK;
  void *p = (void *)pHash;
  const int any = SQLITE_ANY;
  char *zTest = 0;
  char *zTest2 = 0;
................................................................................
  zTest = sqlite3_mprintf("%s_test", zName);
  zTest2 = sqlite3_mprintf("%s_internal_test", zName);
  if( !zTest || !zTest2 ){
    rc = SQLITE_NOMEM;
  }
#endif

  if( SQLITE_OK!=rc
   || SQLITE_OK!=(rc = sqlite3_create_function(db, zName, 1, any, p, scalarFunc, 0, 0))
   || SQLITE_OK!=(rc = sqlite3_create_function(db, zName, 2, any, p, scalarFunc, 0, 0))
#ifdef SQLITE_TEST
   || SQLITE_OK!=(rc = sqlite3_create_function(db, zTest, 2, any, p, testFunc, 0, 0))
   || SQLITE_OK!=(rc = sqlite3_create_function(db, zTest, 3, any, p, testFunc, 0, 0))
   || SQLITE_OK!=(rc = sqlite3_create_function(db, zTest2, 0, any, pdb, intTestFunc, 0, 0))
#endif
  );

  sqlite3_free(zTest);
  sqlite3_free(zTest2);
  return rc;
}

#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */

Changes to ext/fts3/fts3_tokenizer1.c.

20
21
22
23
24
25
26

27
28
29
30
31
32
33
..
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
..
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
...
113
114
115
116
117
118
119


120
121
122
123
124
125
126
...
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
**       (in which case SQLITE_CORE is not defined), or
**
**     * The FTS3 module is being built into the core of
**       SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
*/
#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)



#include <assert.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <ctype.h>

................................................................................
  int iOffset;                 /* current position in pInput */
  int iToken;                  /* index of next token to be returned */
  char *pToken;                /* storage for current token */
  int nTokenAllocated;         /* space allocated to zToken buffer */
} simple_tokenizer_cursor;


/* Forward declaration */
static const sqlite3_tokenizer_module simpleTokenizerModule;

static int simpleDelim(simple_tokenizer *t, unsigned char c){
  return c<0x80 && t->delim[c];
}

/*
** Create a new tokenizer instance.
*/
................................................................................

  /* TODO(shess) Delimiters need to remain the same from run to run,
  ** else we need to reindex.  One solution would be a meta-table to
  ** track such information in the database, then we'd only want this
  ** information on the initial create.
  */
  if( argc>1 ){
    int i, n = strlen(argv[1]);
    for(i=0; i<n; i++){
      unsigned char ch = argv[1][i];
      /* We explicitly don't support UTF-8 delimiters for now. */
      if( ch>=0x80 ){
        sqlite3_free(t);
        return SQLITE_ERROR;
      }
      t->delim[ch] = 1;
    }
  } else {
    /* Mark non-alphanumeric ASCII characters as delimiters */
    int i;
    for(i=1; i<0x80; i++){
      t->delim[i] = !isalnum(i);
    }
  }

  *ppTokenizer = &t->base;
  return SQLITE_OK;
}

................................................................................
*/
static int simpleOpen(
  sqlite3_tokenizer *pTokenizer,         /* The tokenizer */
  const char *pInput, int nBytes,        /* String to be tokenized */
  sqlite3_tokenizer_cursor **ppCursor    /* OUT: Tokenization cursor */
){
  simple_tokenizer_cursor *c;



  c = (simple_tokenizer_cursor *) sqlite3_malloc(sizeof(*c));
  if( c==NULL ) return SQLITE_NOMEM;

  c->pInput = pInput;
  if( pInput==0 ){
    c->nBytes = 0;
................................................................................
        if( c->pToken==NULL ) return SQLITE_NOMEM;
      }
      for(i=0; i<n; i++){
        /* TODO(shess) This needs expansion to handle UTF-8
        ** case-insensitivity.
        */
        unsigned char ch = p[iStartOffset+i];
        c->pToken[i] = ch<0x80 ? tolower(ch) : ch;
      }
      *ppToken = c->pToken;
      *pnBytes = n;
      *piStartOffset = iStartOffset;
      *piEndOffset = c->iOffset;
      *piPosition = c->iToken++;








>







 







<
<
<







 







|













|







 







>
>







 







|







20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
..
46
47
48
49
50
51
52



53
54
55
56
57
58
59
..
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
...
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
...
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
**       (in which case SQLITE_CORE is not defined), or
**
**     * The FTS3 module is being built into the core of
**       SQLite (in which case SQLITE_ENABLE_FTS3 is defined).
*/
#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)

#include "fts3Int.h"

#include <assert.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <ctype.h>

................................................................................
  int iOffset;                 /* current position in pInput */
  int iToken;                  /* index of next token to be returned */
  char *pToken;                /* storage for current token */
  int nTokenAllocated;         /* space allocated to zToken buffer */
} simple_tokenizer_cursor;





static int simpleDelim(simple_tokenizer *t, unsigned char c){
  return c<0x80 && t->delim[c];
}

/*
** Create a new tokenizer instance.
*/
................................................................................

  /* TODO(shess) Delimiters need to remain the same from run to run,
  ** else we need to reindex.  One solution would be a meta-table to
  ** track such information in the database, then we'd only want this
  ** information on the initial create.
  */
  if( argc>1 ){
    int i, n = (int)strlen(argv[1]);
    for(i=0; i<n; i++){
      unsigned char ch = argv[1][i];
      /* We explicitly don't support UTF-8 delimiters for now. */
      if( ch>=0x80 ){
        sqlite3_free(t);
        return SQLITE_ERROR;
      }
      t->delim[ch] = 1;
    }
  } else {
    /* Mark non-alphanumeric ASCII characters as delimiters */
    int i;
    for(i=1; i<0x80; i++){
      t->delim[i] = !isalnum(i) ? -1 : 0;
    }
  }

  *ppTokenizer = &t->base;
  return SQLITE_OK;
}

................................................................................
*/
static int simpleOpen(
  sqlite3_tokenizer *pTokenizer,         /* The tokenizer */
  const char *pInput, int nBytes,        /* String to be tokenized */
  sqlite3_tokenizer_cursor **ppCursor    /* OUT: Tokenization cursor */
){
  simple_tokenizer_cursor *c;

  UNUSED_PARAMETER(pTokenizer);

  c = (simple_tokenizer_cursor *) sqlite3_malloc(sizeof(*c));
  if( c==NULL ) return SQLITE_NOMEM;

  c->pInput = pInput;
  if( pInput==0 ){
    c->nBytes = 0;
................................................................................
        if( c->pToken==NULL ) return SQLITE_NOMEM;
      }
      for(i=0; i<n; i++){
        /* TODO(shess) This needs expansion to handle UTF-8
        ** case-insensitivity.
        */
        unsigned char ch = p[iStartOffset+i];
        c->pToken[i] = (char)(ch<0x80 ? tolower(ch) : ch);
      }
      *ppToken = c->pToken;
      *pnBytes = n;
      *piStartOffset = iStartOffset;
      *piEndOffset = c->iOffset;
      *piPosition = c->iToken++;

Added ext/fts3/fts3_write.c.























































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
/*
** 2009 Oct 23
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
******************************************************************************
**
** This file is part of the SQLite FTS3 extension module. Specifically,
** this file contains code to insert, update and delete rows from FTS3
** tables. It also contains code to merge FTS3 b-tree segments. Some
** of the sub-routines used to merge segments are also used by the query 
** code in fts3.c.
*/

#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)

#include "fts3Int.h"
#include <string.h>
#include <assert.h>
#include <stdlib.h>

typedef struct PendingList PendingList;
typedef struct SegmentNode SegmentNode;
typedef struct SegmentWriter SegmentWriter;

/*
** Data structure used while accumulating terms in the pending-terms hash
** table. The hash table entry maps from term (a string) to a malloced
** instance of this structure.
*/
struct PendingList {
  int nData;
  char *aData;
  int nSpace;
  sqlite3_int64 iLastDocid;
  sqlite3_int64 iLastCol;
  sqlite3_int64 iLastPos;
};

/*
** An instance of this structure is used to iterate through the terms on
** a contiguous set of segment b-tree leaf nodes. Although the details of
** this structure are only manipulated by code in this file, opaque handles
** of type Fts3SegReader* are also used by code in fts3.c to iterate through
** terms when querying the full-text index. See functions:
**
**   sqlite3Fts3SegReaderNew()
**   sqlite3Fts3SegReaderFree()
**   sqlite3Fts3SegReaderIterate()
*/
struct Fts3SegReader {
  int iIdx;                       /* Index within level */
  sqlite3_int64 iStartBlock;
  sqlite3_int64 iEndBlock;
  sqlite3_stmt *pStmt;            /* SQL Statement to access leaf nodes */
  char *aNode;                    /* Pointer to node data (or NULL) */
  int nNode;                      /* Size of buffer at aNode (or 0) */
  int nTermAlloc;                 /* Allocated size of zTerm buffer */

  /* Variables set by fts3SegReaderNext(). These may be read directly
  ** by the caller. They are valid from the time SegmentReaderNew() returns
  ** until SegmentReaderNext() returns something other than SQLITE_OK
  ** (i.e. SQLITE_DONE).
  */
  int nTerm;                      /* Number of bytes in current term */
  char *zTerm;                    /* Pointer to current term */
  char *aDoclist;                 /* Pointer to doclist of current entry */
  int nDoclist;                   /* Size of doclist in current entry */

  /* The following variables are used to iterate through the current doclist */
  char *pOffsetList;
  sqlite3_int64 iDocid;
};

/*
** An instance of this structure is used to create a segment b-tree in the
** database. The internal details of this type are only accessed by the
** following functions:
**
**   fts3SegWriterAdd()
**   fts3SegWriterFlush()
**   fts3SegWriterFree()
*/
struct SegmentWriter {
  SegmentNode *pTree;             /* Pointer to interior tree structure */
  sqlite3_int64 iFirst;           /* First slot in %_segments written */
  sqlite3_int64 iFree;            /* Next free slot in %_segments */
  char *zTerm;                    /* Pointer to previous term buffer */
  int nTerm;                      /* Number of bytes in zTerm */
  int nMalloc;                    /* Size of malloc'd buffer at zMalloc */
  char *zMalloc;                  /* Malloc'd space (possibly) used for zTerm */
  int nSize;                      /* Size of allocation at aData */
  int nData;                      /* Bytes of data in aData */
  char *aData;                    /* Pointer to block from malloc() */
};

/*
** Type SegmentNode is used by the following three functions to create
** the interior part of the segment b+-tree structures (everything except
** the leaf nodes). These functions and type are only ever used by code
** within the fts3SegWriterXXX() family of functions described above.
**
**   fts3NodeAddTerm()
**   fts3NodeWrite()
**   fts3NodeFree()
*/
struct SegmentNode {
  SegmentNode *pParent;           /* Parent node (or NULL for root node) */
  SegmentNode *pRight;            /* Pointer to right-sibling */
  SegmentNode *pLeftmost;         /* Pointer to left-most node of this depth */
  int nEntry;                     /* Number of terms written to node so far */
  char *zTerm;                    /* Pointer to previous term buffer */
  int nTerm;                      /* Number of bytes in zTerm */
  int nMalloc;                    /* Size of malloc'd buffer at zMalloc */
  char *zMalloc;                  /* Malloc'd space (possibly) used for zTerm */
  int nData;                      /* Bytes of valid data so far */
  char *aData;                    /* Node data */
};

/*
** Valid values for the second argument to fts3SqlStmt().
*/
#define SQL_DELETE_CONTENT             0
#define SQL_IS_EMPTY                   1
#define SQL_DELETE_ALL_CONTENT         2 
#define SQL_DELETE_ALL_SEGMENTS        3
#define SQL_DELETE_ALL_SEGDIR          4
#define SQL_SELECT_CONTENT_BY_ROWID    5
#define SQL_NEXT_SEGMENT_INDEX         6
#define SQL_INSERT_SEGMENTS            7
#define SQL_NEXT_SEGMENTS_ID           8
#define SQL_INSERT_SEGDIR              9
#define SQL_SELECT_LEVEL              10
#define SQL_SELECT_ALL_LEVEL          11
#define SQL_SELECT_LEVEL_COUNT        12
#define SQL_SELECT_SEGDIR_COUNT_MAX   13
#define SQL_DELETE_SEGDIR_BY_LEVEL    14
#define SQL_DELETE_SEGMENTS_RANGE     15
#define SQL_CONTENT_INSERT            16
#define SQL_GET_BLOCK                 17

/*
** This function is used to obtain an SQLite prepared statement handle
** for the statement identified by the second argument. If successful,
** *pp is set to the requested statement handle and SQLITE_OK returned.
** Otherwise, an SQLite error code is returned and *pp is set to 0.
**
** If argument apVal is not NULL, then it must point to an array with
** at least as many entries as the requested statement has bound 
** parameters. The values are bound to the statements parameters before
** returning.
*/
static int fts3SqlStmt(
  Fts3Table *p,                   /* Virtual table handle */
  int eStmt,                      /* One of the SQL_XXX constants above */
  sqlite3_stmt **pp,              /* OUT: Statement handle */
  sqlite3_value **apVal           /* Values to bind to statement */
){
  const char *azSql[] = {
/* 0  */  "DELETE FROM %Q.'%q_content' WHERE rowid = ?",
/* 1  */  "SELECT NOT EXISTS(SELECT docid FROM %Q.'%q_content' WHERE rowid!=?)",
/* 2  */  "DELETE FROM %Q.'%q_content'",
/* 3  */  "DELETE FROM %Q.'%q_segments'",
/* 4  */  "DELETE FROM %Q.'%q_segdir'",
/* 5  */  "SELECT * FROM %Q.'%q_content' WHERE rowid=?",
/* 6  */  "SELECT coalesce(max(idx)+1, 0) FROM %Q.'%q_segdir' WHERE level=?",
/* 7  */  "INSERT INTO %Q.'%q_segments'(blockid, block) VALUES(?, ?)",
/* 8  */  "SELECT coalesce(max(blockid)+1, 1) FROM %Q.'%q_segments'",
/* 9  */  "INSERT INTO %Q.'%q_segdir' VALUES(?,?,?,?,?,?)",

          /* Return segments in order from oldest to newest.*/ 
/* 10 */  "SELECT idx, start_block, leaves_end_block, end_block, root "
            "FROM %Q.'%q_segdir' WHERE level = ? ORDER BY idx ASC",
/* 11 */  "SELECT idx, start_block, leaves_end_block, end_block, root "
            "FROM %Q.'%q_segdir' ORDER BY level DESC, idx ASC",

/* 12 */  "SELECT count(*) FROM %Q.'%q_segdir' WHERE level = ?",
/* 13 */  "SELECT count(*), max(level) FROM %Q.'%q_segdir'",

/* 14 */  "DELETE FROM %Q.'%q_segdir' WHERE level = ?",
/* 15 */  "DELETE FROM %Q.'%q_segments' WHERE blockid BETWEEN ? AND ?",
/* 16 */  "INSERT INTO %Q.'%q_content' VALUES(%z)",
/* 17 */  "SELECT block FROM %Q.'%q_segments' WHERE blockid = ?",
  };
  int rc = SQLITE_OK;
  sqlite3_stmt *pStmt;

  assert( SizeofArray(azSql)==SizeofArray(p->aStmt) );
  assert( eStmt<SizeofArray(azSql) && eStmt>=0 );
  
  pStmt = p->aStmt[eStmt];
  if( !pStmt ){
    char *zSql;
    if( eStmt==SQL_CONTENT_INSERT ){
      int i;                      /* Iterator variable */  
      char *zVarlist;             /* The "?, ?, ..." string */
      zVarlist = (char *)sqlite3_malloc(2*p->nColumn+2);
      if( !zVarlist ){
        *pp = 0;
        return SQLITE_NOMEM;
      }
      zVarlist[0] = '?';
      zVarlist[p->nColumn*2+1] = '\0';
      for(i=1; i<=p->nColumn; i++){
        zVarlist[i*2-1] = ',';
        zVarlist[i*2] = '?';
      }
      zSql = sqlite3_mprintf(azSql[eStmt], p->zDb, p->zName, zVarlist);
    }else{
      zSql = sqlite3_mprintf(azSql[eStmt], p->zDb, p->zName);
    }
    if( !zSql ){
      rc = SQLITE_NOMEM;
    }else{
      rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, NULL);
      sqlite3_free(zSql);
      assert( rc==SQLITE_OK || pStmt==0 );
      p->aStmt[eStmt] = pStmt;
    }
  }
  if( apVal ){
    int i;
    int nParam = sqlite3_bind_parameter_count(pStmt);
    for(i=0; rc==SQLITE_OK && i<nParam; i++){
      rc = sqlite3_bind_value(pStmt, i+1, apVal[i]);
    }
  }
  *pp = pStmt;
  return rc;
}

/*
** Similar to fts3SqlStmt(). Except, after binding the parameters in
** array apVal[] to the SQL statement identified by eStmt, the statement
** is executed.
**
** Returns SQLITE_OK if the statement is successfully executed, or an
** SQLite error code otherwise.
*/
static int fts3SqlExec(Fts3Table *p, int eStmt, sqlite3_value **apVal){
  sqlite3_stmt *pStmt;
  int rc = fts3SqlStmt(p, eStmt, &pStmt, apVal); 
  if( rc==SQLITE_OK ){
    sqlite3_step(pStmt);
    rc = sqlite3_reset(pStmt);
  }
  return rc;
}


/*
** Read a single block from the %_segments table. If the specified block
** does not exist, return SQLITE_CORRUPT. If some other error (malloc, IO 
** etc.) occurs, return the appropriate SQLite error code.
**
** Otherwise, if successful, set *pzBlock to point to a buffer containing
** the block read from the database, and *pnBlock to the size of the read
** block in bytes.
**
** WARNING: The returned buffer is only valid until the next call to 
** sqlite3Fts3ReadBlock().
*/
int sqlite3Fts3ReadBlock(
  Fts3Table *p,
  sqlite3_int64 iBlock,
  char const **pzBlock,
  int *pnBlock
){
  sqlite3_stmt *pStmt;
  int rc = fts3SqlStmt(p, SQL_GET_BLOCK, &pStmt, 0);
  if( rc!=SQLITE_OK ) return rc;
  sqlite3_reset(pStmt);

  if( pzBlock ){
    sqlite3_bind_int64(pStmt, 1, iBlock);
    rc = sqlite3_step(pStmt); 
    if( rc!=SQLITE_ROW ){
      return SQLITE_CORRUPT;
    }
  
    *pnBlock = sqlite3_column_bytes(pStmt, 0);
    *pzBlock = (char *)sqlite3_column_blob(pStmt, 0);
    if( !*pzBlock ){
      return SQLITE_NOMEM;
    }
  }
  return SQLITE_OK;
}

/*
** Set *ppStmt to a statement handle that may be used to iterate through
** all rows in the %_segdir table, from oldest to newest. If successful,
** return SQLITE_OK. If an error occurs while preparing the statement, 
** return an SQLite error code.
**
** There is only ever one instance of this SQL statement compiled for
** each FTS3 table.
**
** The statement returns the following columns from the %_segdir table:
**
**   0: idx
**   1: start_block
**   2: leaves_end_block
**   3: end_block
**   4: root
*/
int sqlite3Fts3AllSegdirs(Fts3Table *p, sqlite3_stmt **ppStmt){
  return fts3SqlStmt(p, SQL_SELECT_ALL_LEVEL, ppStmt, 0);
}


/*
** Append a single varint to a PendingList buffer. SQLITE_OK is returned
** if successful, or an SQLite error code otherwise.
**
** This function also serves to allocate the PendingList structure itself.
** For example, to create a new PendingList structure containing two
** varints:
**
**   PendingList *p = 0;
**   fts3PendingListAppendVarint(&p, 1);
**   fts3PendingListAppendVarint(&p, 2);
*/
static int fts3PendingListAppendVarint(
  PendingList **pp,               /* IN/OUT: Pointer to PendingList struct */
  sqlite3_int64 i                 /* Value to append to data */
){
  PendingList *p = *pp;

  /* Allocate or grow the PendingList as required. */
  if( !p ){
    p = sqlite3_malloc(sizeof(*p) + 100);
    if( !p ){
      return SQLITE_NOMEM;
    }
    p->nSpace = 100;
    p->aData = (char *)&p[1];
    p->nData = 0;
  }
  else if( p->nData+FTS3_VARINT_MAX+1>p->nSpace ){
    int nNew = p->nSpace * 2;
    p = sqlite3_realloc(p, sizeof(*p) + nNew);
    if( !p ){
      sqlite3_free(*pp);
      *pp = 0;
      return SQLITE_NOMEM;
    }
    p->nSpace = nNew;
    p->aData = (char *)&p[1];
  }

  /* Append the new serialized varint to the end of the list. */
  p->nData += sqlite3Fts3PutVarint(&p->aData[p->nData], i);
  p->aData[p->nData] = '\0';
  *pp = p;
  return SQLITE_OK;
}

/*
** Add a docid/column/position entry to a PendingList structure. Non-zero
** is returned if the structure is sqlite3_realloced as part of adding
** the entry. Otherwise, zero.
**
** If an OOM error occurs, *pRc is set to SQLITE_NOMEM before returning.
** Zero is always returned in this case. Otherwise, if no OOM error occurs,
** it is set to SQLITE_OK.
*/
static int fts3PendingListAppend(
  PendingList **pp,               /* IN/OUT: PendingList structure */
  sqlite3_int64 iDocid,           /* Docid for entry to add */
  sqlite3_int64 iCol,             /* Column for entry to add */
  sqlite3_int64 iPos,             /* Position of term for entry to add */
  int *pRc                        /* OUT: Return code */
){
  PendingList *p = *pp;
  int rc = SQLITE_OK;

  assert( !p || p->iLastDocid<=iDocid );

  if( !p || p->iLastDocid!=iDocid ){
    sqlite3_int64 iDelta = iDocid - (p ? p->iLastDocid : 0);
    if( p ){
      assert( p->nData<p->nSpace );
      assert( p->aData[p->nData]==0 );
      p->nData++;
    }
    if( SQLITE_OK!=(rc = fts3PendingListAppendVarint(&p, iDelta)) ){
      goto pendinglistappend_out;
    }
    p->iLastCol = -1;
    p->iLastPos = 0;
    p->iLastDocid = iDocid;
  }
  if( iCol>0 && p->iLastCol!=iCol ){
    if( SQLITE_OK!=(rc = fts3PendingListAppendVarint(&p, 1))
     || SQLITE_OK!=(rc = fts3PendingListAppendVarint(&p, iCol))
    ){
      goto pendinglistappend_out;
    }
    p->iLastCol = iCol;
    p->iLastPos = 0;
  }
  if( iCol>=0 ){
    assert( iPos>p->iLastPos || (iPos==0 && p->iLastPos==0) );
    rc = fts3PendingListAppendVarint(&p, 2+iPos-p->iLastPos);
    if( rc==SQLITE_OK ){
      p->iLastPos = iPos;
    }
  }

 pendinglistappend_out:
  *pRc = rc;
  if( p!=*pp ){
    *pp = p;
    return 1;
  }
  return 0;
}

/*
** Tokenize the nul-terminated string zText and add all tokens to the
** pending-terms hash-table. The docid used is that currently stored in
** p->iPrevDocid, and the column is specified by argument iCol.
**
** If successful, SQLITE_OK is returned. Otherwise, an SQLite error code.
*/
static int fts3PendingTermsAdd(Fts3Table *p, const char *zText, int iCol){
  int rc;
  int iStart;
  int iEnd;
  int iPos;

  char const *zToken;
  int nToken;

  sqlite3_tokenizer *pTokenizer = p->pTokenizer;
  sqlite3_tokenizer_module const *pModule = pTokenizer->pModule;
  sqlite3_tokenizer_cursor *pCsr;
  int (*xNext)(sqlite3_tokenizer_cursor *pCursor,
      const char**,int*,int*,int*,int*);

  assert( pTokenizer && pModule );

  rc = pModule->xOpen(pTokenizer, zText, -1, &pCsr);
  if( rc!=SQLITE_OK ){
    return rc;
  }
  pCsr->pTokenizer = pTokenizer;

  xNext = pModule->xNext;
  while( SQLITE_OK==rc
      && SQLITE_OK==(rc = xNext(pCsr, &zToken, &nToken, &iStart, &iEnd, &iPos))
  ){
    PendingList *pList;

    /* Positions cannot be negative; we use -1 as a terminator internally.
    ** Tokens must have a non-zero length.
    */
    if( iPos<0 || !zToken || nToken<=0 ){
      rc = SQLITE_ERROR;
      break;
    }

    pList = (PendingList *)fts3HashFind(&p->pendingTerms, zToken, nToken);
    if( pList ){
      p->nPendingData -= (pList->nData + nToken + sizeof(Fts3HashElem));
    }
    if( fts3PendingListAppend(&pList, p->iPrevDocid, iCol, iPos, &rc) ){
      if( pList==fts3HashInsert(&p->pendingTerms, zToken, nToken, pList) ){
        /* Malloc failed while inserting the new entry. This can only 
        ** happen if there was no previous entry for this token.
        */
        assert( 0==fts3HashFind(&p->pendingTerms, zToken, nToken) );
        sqlite3_free(pList);
        rc = SQLITE_NOMEM;
      }
    }
    if( rc==SQLITE_OK ){
      p->nPendingData += (pList->nData + nToken + sizeof(Fts3HashElem));
    }
  }

  pModule->xClose(pCsr);
  return (rc==SQLITE_DONE ? SQLITE_OK : rc);
}

/* 
** Calling this function indicates that subsequent calls to 
** fts3PendingTermsAdd() are to add term/position-list pairs for the
** contents of the document with docid iDocid.
*/
static int fts3PendingTermsDocid(Fts3Table *p, sqlite_int64 iDocid){
  /* TODO(shess) Explore whether partially flushing the buffer on
  ** forced-flush would provide better performance.  I suspect that if
  ** we ordered the doclists by size and flushed the largest until the
  ** buffer was half empty, that would let the less frequent terms
  ** generate longer doclists.
  */
  if( iDocid<=p->iPrevDocid || p->nPendingData>FTS3_MAX_PENDING_DATA ){
    int rc = sqlite3Fts3PendingTermsFlush(p);
    if( rc!=SQLITE_OK ) return rc;
  }
  p->iPrevDocid = iDocid;
  return SQLITE_OK;
}

void sqlite3Fts3PendingTermsClear(Fts3Table *p){
  Fts3HashElem *pElem;
  for(pElem=fts3HashFirst(&p->pendingTerms); pElem; pElem=fts3HashNext(pElem)){
    sqlite3_free(fts3HashData(pElem));
  }
  fts3HashClear(&p->pendingTerms);
  p->nPendingData = 0;
}

/*
** This function is called by the xUpdate() method as part of an INSERT
** operation. It adds entries for each term in the new record to the
** pendingTerms hash table.
**
** Argument apVal is the same as the similarly named argument passed to
** fts3InsertData(). Parameter iDocid is the docid of the new row.
*/
static int fts3InsertTerms(Fts3Table *p, sqlite3_value **apVal){
  int i;                          /* Iterator variable */
  for(i=2; i<p->nColumn+2; i++){
    const char *zText = (const char *)sqlite3_value_text(apVal[i]);
    if( zText ){
      int rc = fts3PendingTermsAdd(p, zText, i-2);
      if( rc!=SQLITE_OK ){
        return rc;
      }
    }
  }
  return SQLITE_OK;
}

/*
** This function is called by the xUpdate() method for an INSERT operation.
** The apVal parameter is passed a copy of the apVal argument passed by
** SQLite to the xUpdate() method. i.e:
**
**   apVal[0]                Not used for INSERT.
**   apVal[1]                rowid
**   apVal[2]                Left-most user-defined column
**   ...
**   apVal[p->nColumn+1]     Right-most user-defined column
**   apVal[p->nColumn+2]     Hidden column with same name as table
**   apVal[p->nColumn+3]     Hidden "docid" column (alias for rowid)
*/
static int fts3InsertData(
  Fts3Table *p,                   /* Full-text table */
  sqlite3_value **apVal,          /* Array of values to insert */
  sqlite3_int64 *piDocid          /* OUT: Docid for row just inserted */
){
  int rc;                         /* Return code */
  sqlite3_stmt *pContentInsert;   /* INSERT INTO %_content VALUES(...) */

  /* Locate the statement handle used to insert data into the %_content
  ** table. The SQL for this statement is:
  **
  **   INSERT INTO %_content VALUES(?, ?, ?, ...)
  **
  ** The statement features N '?' variables, where N is the number of user
  ** defined columns in the FTS3 table, plus one for the docid field.
  */
  rc = fts3SqlStmt(p, SQL_CONTENT_INSERT, &pContentInsert, &apVal[1]);
  if( rc!=SQLITE_OK ){
    return rc;
  }

  /* There is a quirk here. The users INSERT statement may have specified
  ** a value for the "rowid" field, for the "docid" field, or for both.
  ** Which is a problem, since "rowid" and "docid" are aliases for the
  ** same value. For example:
  **
  **   INSERT INTO fts3tbl(rowid, docid) VALUES(1, 2);
  **
  ** In FTS3, this is an error. It is an error to specify non-NULL values
  ** for both docid and some other rowid alias.
  */
  if( SQLITE_NULL!=sqlite3_value_type(apVal[3+p->nColumn]) ){
    if( SQLITE_NULL==sqlite3_value_type(apVal[0])
     && SQLITE_NULL!=sqlite3_value_type(apVal[1])
    ){
      /* A rowid/docid conflict. */
      return SQLITE_ERROR;
    }
    rc = sqlite3_bind_value(pContentInsert, 1, apVal[3+p->nColumn]);
    if( rc!=SQLITE_OK ) return rc;
  }

  /* Execute the statement to insert the record. Set *piDocid to the 
  ** new docid value. 
  */
  sqlite3_step(pContentInsert);
  rc = sqlite3_reset(pContentInsert);

  *piDocid = sqlite3_last_insert_rowid(p->db);
  return rc;
}



/*
** Remove all data from the FTS3 table. Clear the hash table containing
** pending terms.
*/
static int fts3DeleteAll(Fts3Table *p){
  int rc;                         /* Return code */

  /* Discard the contents of the pending-terms hash table. */
  sqlite3Fts3PendingTermsClear(p);

  /* Delete everything from the %_content, %_segments and %_segdir tables. */
  rc = fts3SqlExec(p, SQL_DELETE_ALL_CONTENT, 0);
  if( rc==SQLITE_OK ){
    rc = fts3SqlExec(p, SQL_DELETE_ALL_SEGMENTS, 0);
  }
  if( rc==SQLITE_OK ){
    rc = fts3SqlExec(p, SQL_DELETE_ALL_SEGDIR, 0);
  }
  return rc;
}

/*
** The first element in the apVal[] array is assumed to contain the docid
** (an integer) of a row about to be deleted. Remove all terms from the
** full-text index.
*/
static int fts3DeleteTerms(Fts3Table *p, sqlite3_value **apVal){
  int rc;
  sqlite3_stmt *pSelect;

  rc = fts3SqlStmt(p, SQL_SELECT_CONTENT_BY_ROWID, &pSelect, apVal);
  if( rc==SQLITE_OK ){
    if( SQLITE_ROW==sqlite3_step(pSelect) ){
      int i;
      for(i=1; i<=p->nColumn; i++){
        const char *zText = (const char *)sqlite3_column_text(pSelect, i);
        rc = fts3PendingTermsAdd(p, zText, -1);
        if( rc!=SQLITE_OK ){
          sqlite3_reset(pSelect);
          return rc;
        }
      }
    }
    rc = sqlite3_reset(pSelect);
  }else{
    sqlite3_reset(pSelect);
  }
  return rc;
}

/*
** Forward declaration to account for the circular dependency between
** functions fts3SegmentMerge() and fts3AllocateSegdirIdx().
*/
static int fts3SegmentMerge(Fts3Table *, int);

/* 
** This function allocates a new level iLevel index in the segdir table.
** Usually, indexes are allocated within a level sequentially starting
** with 0, so the allocated index is one greater than the value returned
** by:
**
**   SELECT max(idx) FROM %_segdir WHERE level = :iLevel
**
** However, if there are already FTS3_MERGE_COUNT indexes at the requested
** level, they are merged into a single level (iLevel+1) segment and the 
** allocated index is 0.
**
** If successful, *piIdx is set to the allocated index slot and SQLITE_OK
** returned. Otherwise, an SQLite error code is returned.
*/
static int fts3AllocateSegdirIdx(Fts3Table *p, int iLevel, int *piIdx){
  int rc;                         /* Return Code */
  sqlite3_stmt *pNextIdx;         /* Query for next idx at level iLevel */
  int iNext = 0;                  /* Result of query pNextIdx */

  /* Set variable iNext to the next available segdir index at level iLevel. */
  rc = fts3SqlStmt(p, SQL_NEXT_SEGMENT_INDEX, &pNextIdx, 0);
  if( rc==SQLITE_OK ){
    sqlite3_bind_int(pNextIdx, 1, iLevel);
    if( SQLITE_ROW==sqlite3_step(pNextIdx) ){
      iNext = sqlite3_column_int(pNextIdx, 0);
    }
    rc = sqlite3_reset(pNextIdx);
  }

  if( rc==SQLITE_OK ){
    /* If iNext is FTS3_MERGE_COUNT, indicating that level iLevel is already
    ** full, merge all segments in level iLevel into a single iLevel+1
    ** segment and allocate (newly freed) index 0 at level iLevel. Otherwise,
    ** if iNext is less than FTS3_MERGE_COUNT, allocate index iNext.
    */
    if( iNext>=FTS3_MERGE_COUNT ){
      rc = fts3SegmentMerge(p, iLevel);
      *piIdx = 0;
    }else{
      *piIdx = iNext;
    }
  }

  return rc;
}

/*
** Move the iterator passed as the first argument to the next term in the
** segment. If successful, SQLITE_OK is returned. If there is no next term,
** SQLITE_DONE. Otherwise, an SQLite error code.
*/
static int fts3SegReaderNext(Fts3SegReader *pReader){
  char *pNext;                    /* Cursor variable */
  int nPrefix;                    /* Number of bytes in term prefix */
  int nSuffix;                    /* Number of bytes in term suffix */

  if( !pReader->aDoclist ){
    pNext = pReader->aNode;
  }else{
    pNext = &pReader->aDoclist[pReader->nDoclist];
  }

  if( !pNext || pNext>=&pReader->aNode[pReader->nNode] ){
    int rc;
    if( !pReader->pStmt ){
      pReader->aNode = 0;
      return SQLITE_OK;
    }
    rc = sqlite3_step(pReader->pStmt);
    if( rc!=SQLITE_ROW ){
      pReader->aNode = 0;
      return (rc==SQLITE_DONE ? SQLITE_OK : rc);
    }
    pReader->nNode = sqlite3_column_bytes(pReader->pStmt, 0);
    pReader->aNode = (char *)sqlite3_column_blob(pReader->pStmt, 0);
    pNext = pReader->aNode;
  }
  
  pNext += sqlite3Fts3GetVarint32(pNext, &nPrefix);
  pNext += sqlite3Fts3GetVarint32(pNext, &nSuffix);

  if( nPrefix+nSuffix>pReader->nTermAlloc ){
    int nNew = (nPrefix+nSuffix)*2;
    char *zNew = sqlite3_realloc(pReader->zTerm, nNew);
    if( !zNew ){
      return SQLITE_NOMEM;
    }
    pReader->zTerm = zNew;
    pReader->nTermAlloc = nNew;
  }
  memcpy(&pReader->zTerm[nPrefix], pNext, nSuffix);
  pReader->nTerm = nPrefix+nSuffix;
  pNext += nSuffix;
  pNext += sqlite3Fts3GetVarint32(pNext, &pReader->nDoclist);
  assert( pNext<&pReader->aNode[pReader->nNode] );
  pReader->aDoclist = pNext;
  pReader->pOffsetList = 0;
  return SQLITE_OK;
}

/*
** Set the SegReader to point to the first docid in the doclist associated
** with the current term.
*/
static void fts3SegReaderFirstDocid(Fts3SegReader *pReader){
  int n;
  assert( pReader->aDoclist );
  assert( !pReader->pOffsetList );
  n = sqlite3Fts3GetVarint(pReader->aDoclist, &pReader->iDocid);
  pReader->pOffsetList = &pReader->aDoclist[n];
}

/*
** Advance the SegReader to point to the next docid in the doclist
** associated with the current term.
** 
** If arguments ppOffsetList and pnOffsetList are not NULL, then 
** *ppOffsetList is set to point to the first column-offset list
** in the doclist entry (i.e. immediately past the docid varint).
** *pnOffsetList is set to the length of the set of column-offset
** lists, not including the nul-terminator byte. For example:
*/
static void fts3SegReaderNextDocid(
  Fts3SegReader *pReader,
  char **ppOffsetList,
  int *pnOffsetList
){
  char *p = pReader->pOffsetList;
  char c = 0;

  /* Pointer p currently points at the first byte of an offset list. The
  ** following two lines advance it to point one byte past the end of
  ** the same offset list.
  */
  while( *p | c ) c = *p++ & 0x80;
  p++;

  /* If required, populate the output variables with a pointer to and the
  ** size of the previous offset-list.
  */
  if( ppOffsetList ){
    *ppOffsetList = pReader->pOffsetList;
    *pnOffsetList = (int)(p - pReader->pOffsetList - 1);
  }

  /* If there are no more entries in the doclist, set pOffsetList to
  ** NULL. Otherwise, set Fts3SegReader.iDocid to the next docid and
  ** Fts3SegReader.pOffsetList to point to the next offset list before
  ** returning.
  */
  if( p>=&pReader->aDoclist[pReader->nDoclist] ){
    pReader->pOffsetList = 0;
  }else{
    sqlite3_int64 iDelta;
    pReader->pOffsetList = p + sqlite3Fts3GetVarint(p, &iDelta);
    pReader->iDocid += iDelta;
  }
}

/*
** Free all allocations associated with the iterator passed as the 
** second argument.
*/
void sqlite3Fts3SegReaderFree(Fts3Table *p, Fts3SegReader *pReader){
  if( pReader ){
    if( pReader->pStmt ){
      /* Move the leaf-range SELECT statement to the aLeavesStmt[] array,
      ** so that it can be reused when required by another query.
      */
      assert( p->nLeavesStmt<p->nLeavesTotal );
      sqlite3_reset(pReader->pStmt);
      p->aLeavesStmt[p->nLeavesStmt++] = pReader->pStmt;
    }
    sqlite3_free(pReader->zTerm);
    sqlite3_free(pReader);
  }
}

/*
** Allocate a new SegReader object.
*/
int sqlite3Fts3SegReaderNew(
  Fts3Table *p,                   /* Virtual table handle */
  int iAge,                       /* Segment "age". */
  sqlite3_int64 iStartLeaf,       /* First leaf to traverse */
  sqlite3_int64 iEndLeaf,         /* Final leaf to traverse */
  sqlite3_int64 iEndBlock,        /* Final block of segment */
  const char *zRoot,              /* Buffer containing root node */
  int nRoot,                      /* Size of buffer containing root node */
  Fts3SegReader **ppReader        /* OUT: Allocated Fts3SegReader */
){
  int rc = SQLITE_OK;             /* Return code */
  Fts3SegReader *pReader;         /* Newly allocated SegReader object */
  int nExtra = 0;                 /* Bytes to allocate segment root node */

  if( iStartLeaf==0 ){
    nExtra = nRoot;
  }

  pReader = (Fts3SegReader *)sqlite3_malloc(sizeof(Fts3SegReader) + nExtra);
  if( !pReader ){
    return SQLITE_NOMEM;
  }
  memset(pReader, 0, sizeof(Fts3SegReader));
  pReader->iStartBlock = iStartLeaf;
  pReader->iIdx = iAge;
  pReader->iEndBlock = iEndBlock;

  if( nExtra ){
    /* The entire segment is stored in the root node. */
    pReader->aNode = (char *)&pReader[1];
    pReader->nNode = nRoot;
    memcpy(pReader->aNode, zRoot, nRoot);
  }else{
    /* If the text of the SQL statement to iterate through a contiguous
    ** set of entries in the %_segments table has not yet been composed,
    ** compose it now.
    */
    if( !p->zSelectLeaves ){
      p->zSelectLeaves = sqlite3_mprintf(
          "SELECT block FROM %Q.'%q_segments' WHERE blockid BETWEEN ? AND ? "
          "ORDER BY blockid", p->zDb, p->zName
      );
      if( !p->zSelectLeaves ){
        rc = SQLITE_NOMEM;
        goto finished;
      }
    }

    /* If there are no free statements in the aLeavesStmt[] array, prepare
    ** a new statement now. Otherwise, reuse a prepared statement from
    ** aLeavesStmt[].
    */
    if( p->nLeavesStmt==0 ){
      if( p->nLeavesTotal==p->nLeavesAlloc ){
        int nNew = p->nLeavesAlloc + 16;
        sqlite3_stmt **aNew = (sqlite3_stmt **)sqlite3_realloc(
            p->aLeavesStmt, nNew*sizeof(sqlite3_stmt *)
        );
        if( !aNew ){
          rc = SQLITE_NOMEM;
          goto finished;
        }
        p->nLeavesAlloc = nNew;
        p->aLeavesStmt = aNew;
      }
      rc = sqlite3_prepare_v2(p->db, p->zSelectLeaves, -1, &pReader->pStmt, 0);
      if( rc!=SQLITE_OK ){
        goto finished;
      }
      p->nLeavesTotal++;
    }else{
      pReader->pStmt = p->aLeavesStmt[--p->nLeavesStmt];
    }

    /* Bind the start and end leaf blockids to the prepared SQL statement. */
    sqlite3_bind_int64(pReader->pStmt, 1, iStartLeaf);
    sqlite3_bind_int64(pReader->pStmt, 2, iEndLeaf);
  }
  rc = fts3SegReaderNext(pReader);

 finished:
  if( rc==SQLITE_OK ){
    *ppReader = pReader;
  }else{
    sqlite3Fts3SegReaderFree(p, pReader);
  }
  return rc;
}


/*
** The second argument to this function is expected to be a statement of
** the form:
**
**   SELECT 
**     idx,                  -- col 0
**     start_block,          -- col 1
**     leaves_end_block,     -- col 2
**     end_block,            -- col 3
**     root                  -- col 4
**   FROM %_segdir ...
**
** This function allocates and initializes a Fts3SegReader structure to
** iterate through the terms stored in the segment identified by the
** current row that pStmt is pointing to. 
**
** If successful, the Fts3SegReader is left pointing to the first term
** in the segment and SQLITE_OK is returned. Otherwise, an SQLite error
** code is returned.
*/
static int fts3SegReaderNew(
  Fts3Table *p,                   /* Virtual table handle */
  sqlite3_stmt *pStmt,            /* See above */
  int iAge,                       /* Segment "age". */
  Fts3SegReader **ppReader        /* OUT: Allocated Fts3SegReader */
){
  return sqlite3Fts3SegReaderNew(p, iAge, 
      sqlite3_column_int64(pStmt, 1),
      sqlite3_column_int64(pStmt, 2),
      sqlite3_column_int64(pStmt, 3),
      sqlite3_column_blob(pStmt, 4),
      sqlite3_column_bytes(pStmt, 4),
      ppReader
  );
}

/*
** Compare the entries pointed to by two Fts3SegReader structures. 
** Comparison is as follows:
**
**   1) EOF is greater than not EOF.
**
**   2) The current terms (if any) are compared with memcmp(). If one
**      term is a prefix of another, the longer term is considered the
**      larger.
**
**   3) By segment age. An older segment is considered larger.
*/
static int fts3SegReaderCmp(Fts3SegReader *pLhs, Fts3SegReader *pRhs){
  int rc;
  if( pLhs->aNode && pRhs->aNode ){
    int rc2 = pLhs->nTerm - pRhs->nTerm;
    if( rc2<0 ){
      rc = memcmp(pLhs->zTerm, pRhs->zTerm, pLhs->nTerm);
    }else{
      rc = memcmp(pLhs->zTerm, pRhs->zTerm, pRhs->nTerm);
    }
    if( rc==0 ){
      rc = rc2;
    }
  }else{
    rc = (pLhs->aNode==0) - (pRhs->aNode==0);
  }
  if( rc==0 ){
    rc = pRhs->iIdx - pLhs->iIdx;
  }
  assert( rc!=0 );
  return rc;
}

/*
** A different comparison function for SegReader structures. In this
** version, it is assumed that each SegReader points to an entry in
** a doclist for identical terms. Comparison is made as follows:
**
**   1) EOF (end of doclist in this case) is greater than not EOF.
**
**   2) By current docid.
**
**   3) By segment age. An older segment is considered larger.
*/
static int fts3SegReaderDoclistCmp(Fts3SegReader *pLhs, Fts3SegReader *pRhs){
  int rc = (pLhs->pOffsetList==0)-(pRhs->pOffsetList==0);
  if( rc==0 ){
    if( pLhs->iDocid==pRhs->iDocid ){
      rc = pRhs->iIdx - pLhs->iIdx;
    }else{
      rc = (pLhs->iDocid > pRhs->iDocid) ? 1 : -1;
    }
  }
  assert( pLhs->aNode && pRhs->aNode );
  return rc;
}

/*
** Compare the term that the Fts3SegReader object passed as the first argument
** points to with the term specified by arguments zTerm and nTerm. 
**
** If the pSeg iterator is already at EOF, return 0. Otherwise, return
** -ve if the pSeg term is less than zTerm/nTerm, 0 if the two terms are
** equal, or +ve if the pSeg term is greater than zTerm/nTerm.
*/
static int fts3SegReaderTermCmp(
  Fts3SegReader *pSeg,            /* Segment reader object */
  const char *zTerm,              /* Term to compare to */
  int nTerm                       /* Size of term zTerm in bytes */
){
  int res = 0;
  if( pSeg->aNode ){
    if( pSeg->nTerm>nTerm ){
      res = memcmp(pSeg->zTerm, zTerm, nTerm);
    }else{
      res = memcmp(pSeg->zTerm, zTerm, pSeg->nTerm);
    }
    if( res==0 ){
      res = pSeg->nTerm-nTerm;
    }
  }
  return res;
}

/*
** Argument apSegment is an array of nSegment elements. It is known that
** the final (nSegment-nSuspect) members are already in sorted order
** (according to the comparison function provided). This function shuffles
** the array around until all entries are in sorted order.
*/
static void fts3SegReaderSort(
  Fts3SegReader **apSegment,                     /* Array to sort entries of */
  int nSegment,                                  /* Size of apSegment array */
  int nSuspect,                                  /* Unsorted entry count */
  int (*xCmp)(Fts3SegReader *, Fts3SegReader *)  /* Comparison function */
){
  int i;                          /* Iterator variable */

  assert( nSuspect<=nSegment );

  if( nSuspect==nSegment ) nSuspect--;
  for(i=nSuspect-1; i>=0; i--){
    int j;
    for(j=i; j<(nSegment-1); j++){
      Fts3SegReader *pTmp;
      if( xCmp(apSegment[j], apSegment[j+1])<0 ) break;
      pTmp = apSegment[j+1];
      apSegment[j+1] = apSegment[j];
      apSegment[j] = pTmp;
    }
  }

#ifndef NDEBUG
  /* Check that the list really is sorted now. */
  for(i=0; i<(nSuspect-1); i++){
    assert( xCmp(apSegment[i], apSegment[i+1])<0 );
  }
#endif
}

/* 
** Insert a record into the %_segments table.
*/
static int fts3WriteSegment(
  Fts3Table *p,                   /* Virtual table handle */
  sqlite3_int64 iBlock,           /* Block id for new block */
  char *z,                        /* Pointer to buffer containing block data */
  int n                           /* Size of buffer z in bytes */
){
  sqlite3_stmt *pStmt;
  int rc = fts3SqlStmt(p, SQL_INSERT_SEGMENTS, &pStmt, 0);
  if( rc==SQLITE_OK ){
    sqlite3_bind_int64(pStmt, 1, iBlock);
    rc = sqlite3_bind_blob(pStmt, 2, z, n, SQLITE_STATIC);
    if( rc==SQLITE_OK ){
      sqlite3_step(pStmt);
      rc = sqlite3_reset(pStmt);
    }
  }
  return rc;
}

/* 
** Insert a record into the %_segdir table.
*/
static int fts3WriteSegdir(
  Fts3Table *p,                   /* Virtual table handle */
  int iLevel,                     /* Value for "level" field */
  int iIdx,                       /* Value for "idx" field */
  sqlite3_int64 iStartBlock,      /* Value for "start_block" field */
  sqlite3_int64 iLeafEndBlock,    /* Value for "leaves_end_block" field */
  sqlite3_int64 iEndBlock,        /* Value for "end_block" field */
  char *zRoot,                    /* Blob value for "root" field */
  int nRoot                       /* Number of bytes in buffer zRoot */
){
  sqlite3_stmt *pStmt;
  int rc = fts3SqlStmt(p, SQL_INSERT_SEGDIR, &pStmt, 0);
  if( rc==SQLITE_OK ){
    sqlite3_bind_int(pStmt, 1, iLevel);
    sqlite3_bind_int(pStmt, 2, iIdx);
    sqlite3_bind_int64(pStmt, 3, iStartBlock);
    sqlite3_bind_int64(pStmt, 4, iLeafEndBlock);
    sqlite3_bind_int64(pStmt, 5, iEndBlock);
    rc = sqlite3_bind_blob(pStmt, 6, zRoot, nRoot, SQLITE_STATIC);
    if( rc==SQLITE_OK ){
      sqlite3_step(pStmt);
      rc = sqlite3_reset(pStmt);
    }
  }
  return rc;
}

/*
** Return the size of the common prefix (if any) shared by zPrev and
** zNext, in bytes. For example, 
**
**   fts3PrefixCompress("abc", 3, "abcdef", 6)   // returns 3
**   fts3PrefixCompress("abX", 3, "abcdef", 6)   // returns 2
**   fts3PrefixCompress("abX", 3, "Xbcdef", 6)   // returns 0
*/
static int fts3PrefixCompress(
  const char *zPrev,              /* Buffer containing previous term */
  int nPrev,                      /* Size of buffer zPrev in bytes */
  const char *zNext,              /* Buffer containing next term */
  int nNext                       /* Size of buffer zNext in bytes */
){
  int n;
  UNUSED_PARAMETER(nNext);
  for(n=0; n<nPrev && zPrev[n]==zNext[n]; n++);
  return n;
}

/*
** Add term zTerm to the SegmentNode. It is guaranteed that zTerm is larger
** (according to memcmp) than the previous term.
*/
static int fts3NodeAddTerm(
  Fts3Table *p,               /* Virtual table handle */
  SegmentNode **ppTree,           /* IN/OUT: SegmentNode handle */ 
  int isCopyTerm,                 /* True if zTerm/nTerm is transient */
  const char *zTerm,              /* Pointer to buffer containing term */
  int nTerm                       /* Size of term in bytes */
){
  SegmentNode *pTree = *ppTree;
  int rc;
  SegmentNode *pNew;

  /* First try to append the term to the current node. Return early if 
  ** this is possible.
  */
  if( pTree ){
    int nData = pTree->nData;     /* Current size of node in bytes */
    int nReq = nData;             /* Required space after adding zTerm */
    int nPrefix;                  /* Number of bytes of prefix compression */
    int nSuffix;                  /* Suffix length */

    nPrefix = fts3PrefixCompress(pTree->zTerm, pTree->nTerm, zTerm, nTerm);
    nSuffix = nTerm-nPrefix;

    nReq += sqlite3Fts3VarintLen(nPrefix)+sqlite3Fts3VarintLen(nSuffix)+nSuffix;
    if( nReq<=p->nNodeSize || !pTree->zTerm ){

      if( nReq>p->nNodeSize ){
        /* An unusual case: this is the first term to be added to the node
        ** and the static node buffer (p->nNodeSize bytes) is not large
        ** enough. Use a separately malloced buffer instead This wastes
        ** p->nNodeSize bytes, but since this scenario only comes about when
        ** the database contain two terms that share a prefix of almost 2KB, 
        ** this is not expected to be a serious problem. 
        */
        assert( pTree->aData==(char *)&pTree[1] );
        pTree->aData = (char *)sqlite3_malloc(nReq);
        if( !pTree->aData ){
          return SQLITE_NOMEM;
        }
      }

      if( pTree->zTerm ){
        /* There is no prefix-length field for first term in a node */
        nData += sqlite3Fts3PutVarint(&pTree->aData[nData], nPrefix);
      }

      nData += sqlite3Fts3PutVarint(&pTree->aData[nData], nSuffix);
      memcpy(&pTree->aData[nData], &zTerm[nPrefix], nSuffix);
      pTree->nData = nData + nSuffix;
      pTree->nEntry++;

      if( isCopyTerm ){
        if( pTree->nMalloc<nTerm ){
          char *zNew = sqlite3_realloc(pTree->zMalloc, nTerm*2);
          if( !zNew ){
            return SQLITE_NOMEM;
          }
          pTree->nMalloc = nTerm*2;
          pTree->zMalloc = zNew;
        }
        pTree->zTerm = pTree->zMalloc;
        memcpy(pTree->zTerm, zTerm, nTerm);
        pTree->nTerm = nTerm;
      }else{
        pTree->zTerm = (char *)zTerm;
        pTree->nTerm = nTerm;
      }
      return SQLITE_OK;
    }
  }

  /* If control flows to here, it was not possible to append zTerm to the
  ** current node. Create a new node (a right-sibling of the current node).
  ** If this is the first node in the tree, the term is added to it.
  **
  ** Otherwise, the term is not added to the new node, it is left empty for
  ** now. Instead, the term is inserted into the parent of pTree. If pTree 
  ** has no parent, one is created here.
  */
  pNew = (SegmentNode *)sqlite3_malloc(sizeof(SegmentNode) + p->nNodeSize);
  if( !pNew ){
    return SQLITE_NOMEM;
  }
  memset(pNew, 0, sizeof(SegmentNode));
  pNew->nData = 1 + FTS3_VARINT_MAX;
  pNew->aData = (char *)&pNew[1];

  if( pTree ){
    SegmentNode *pParent = pTree->pParent;
    rc = fts3NodeAddTerm(p, &pParent, isCopyTerm, zTerm, nTerm);
    if( pTree->pParent==0 ){
      pTree->pParent = pParent;
    }
    pTree->pRight = pNew;
    pNew->pLeftmost = pTree->pLeftmost;
    pNew->pParent = pParent;
    pNew->zMalloc = pTree->zMalloc;
    pNew->nMalloc = pTree->nMalloc;
    pTree->zMalloc = 0;
  }else{
    pNew->pLeftmost = pNew;
    rc = fts3NodeAddTerm(p, &pNew, isCopyTerm, zTerm, nTerm); 
  }

  *ppTree = pNew;
  return rc;
}

/*
** Helper function for fts3NodeWrite().
*/
static int fts3TreeFinishNode(
  SegmentNode *pTree, 
  int iHeight, 
  sqlite3_int64 iLeftChild
){
  int nStart;
  assert( iHeight>=1 && iHeight<128 );
  nStart = FTS3_VARINT_MAX - sqlite3Fts3VarintLen(iLeftChild);
  pTree->aData[nStart] = (char)iHeight;
  sqlite3Fts3PutVarint(&pTree->aData[nStart+1], iLeftChild);
  return nStart;
}

/*
** Write the buffer for the segment node pTree and all of its peers to the
** database. Then call this function recursively to write the parent of 
** pTree and its peers to the database. 
**
** Except, if pTree is a root node, do not write it to the database. Instead,
** set output variables *paRoot and *pnRoot to contain the root node.
**
** If successful, SQLITE_OK is returned and output variable *piLast is
** set to the largest blockid written to the database (or zero if no
** blocks were written to the db). Otherwise, an SQLite error code is 
** returned.
*/
static int fts3NodeWrite(
  Fts3Table *p,                   /* Virtual table handle */
  SegmentNode *pTree,             /* SegmentNode handle */
  int iHeight,                    /* Height of this node in tree */
  sqlite3_int64 iLeaf,            /* Block id of first leaf node */
  sqlite3_int64 iFree,            /* Block id of next free slot in %_segments */
  sqlite3_int64 *piLast,          /* OUT: Block id of last entry written */
  char **paRoot,                  /* OUT: Data for root node */
  int *pnRoot                     /* OUT: Size of root node in bytes */
){
  int rc = SQLITE_OK;

  if( !pTree->pParent ){
    /* Root node of the tree. */
    int nStart = fts3TreeFinishNode(pTree, iHeight, iLeaf);
    *piLast = iFree-1;
    *pnRoot = pTree->nData - nStart;
    *paRoot = &pTree->aData[nStart];
  }else{
    SegmentNode *pIter;
    sqlite3_int64 iNextFree = iFree;
    sqlite3_int64 iNextLeaf = iLeaf;
    for(pIter=pTree->pLeftmost; pIter && rc==SQLITE_OK; pIter=pIter->pRight){
      int nStart = fts3TreeFinishNode(pIter, iHeight, iNextLeaf);
      int nWrite = pIter->nData - nStart;
  
      rc = fts3WriteSegment(p, iNextFree, &pIter->aData[nStart], nWrite);
      iNextFree++;
      iNextLeaf += (pIter->nEntry+1);
    }
    if( rc==SQLITE_OK ){
      assert( iNextLeaf==iFree );
      rc = fts3NodeWrite(
          p, pTree->pParent, iHeight+1, iFree, iNextFree, piLast, paRoot, pnRoot
      );
    }
  }

  return rc;
}

/*
** Free all memory allocations associated with the tree pTree.
*/
static void fts3NodeFree(SegmentNode *pTree){
  if( pTree ){
    SegmentNode *p = pTree->pLeftmost;
    fts3NodeFree(p->pParent);
    while( p ){
      SegmentNode *pRight = p->pRight;
      if( p->aData!=(char *)&p[1] ){
        sqlite3_free(p->aData);
      }
      assert( pRight==0 || p->zMalloc==0 );
      sqlite3_free(p->zMalloc);
      sqlite3_free(p);
      p = pRight;
    }
  }
}

/*
** Add a term to the segment being constructed by the SegmentWriter object
** *ppWriter. When adding the first term to a segment, *ppWriter should
** be passed NULL. This function will allocate a new SegmentWriter object
** and return it via the input/output variable *ppWriter in this case.
**
** If successful, SQLITE_OK is returned. Otherwise, an SQLite error code.
*/
static int fts3SegWriterAdd(
  Fts3Table *p,                   /* Virtual table handle */
  SegmentWriter **ppWriter,       /* IN/OUT: SegmentWriter handle */ 
  int isCopyTerm,                 /* True if buffer zTerm must be copied */
  const char *zTerm,              /* Pointer to buffer containing term */
  int nTerm,                      /* Size of term in bytes */
  const char *aDoclist,           /* Pointer to buffer containing doclist */
  int nDoclist                    /* Size of doclist in bytes */
){
  int nPrefix;                    /* Size of term prefix in bytes */
  int nSuffix;                    /* Size of term suffix in bytes */
  int nReq;                       /* Number of bytes required on leaf page */
  int nData;
  SegmentWriter *pWriter = *ppWriter;

  if( !pWriter ){
    int rc;
    sqlite3_stmt *pStmt;

    /* Allocate the SegmentWriter structure */
    pWriter = (SegmentWriter *)sqlite3_malloc(sizeof(SegmentWriter));
    if( !pWriter ) return SQLITE_NOMEM;
    memset(pWriter, 0, sizeof(SegmentWriter));
    *ppWriter = pWriter;

    /* Allocate a buffer in which to accumulate data */
    pWriter->aData = (char *)sqlite3_malloc(p->nNodeSize);
    if( !pWriter->aData ) return SQLITE_NOMEM;
    pWriter->nSize = p->nNodeSize;

    /* Find the next free blockid in the %_segments table */
    rc = fts3SqlStmt(p, SQL_NEXT_SEGMENTS_ID, &pStmt, 0);
    if( rc!=SQLITE_OK ) return rc;
    if( SQLITE_ROW==sqlite3_step(pStmt) ){
      pWriter->iFree = sqlite3_column_int64(pStmt, 0);
      pWriter->iFirst = pWriter->iFree;
    }
    rc = sqlite3_reset(pStmt);
    if( rc!=SQLITE_OK ) return rc;
  }
  nData = pWriter->nData;

  nPrefix = fts3PrefixCompress(pWriter->zTerm, pWriter->nTerm, zTerm, nTerm);
  nSuffix = nTerm-nPrefix;

  /* Figure out how many bytes are required by this new entry */
  nReq = sqlite3Fts3VarintLen(nPrefix) +    /* varint containing prefix size */
    sqlite3Fts3VarintLen(nSuffix) +         /* varint containing suffix size */
    nSuffix +                               /* Term suffix */
    sqlite3Fts3VarintLen(nDoclist) +        /* Size of doclist */
    nDoclist;                               /* Doclist data */

  if( nData>0 && nData+nReq>p->nNodeSize ){
    int rc;

    /* The current leaf node is full. Write it out to the database. */
    rc = fts3WriteSegment(p, pWriter->iFree++, pWriter->aData, nData);
    if( rc!=SQLITE_OK ) return rc;

    /* Add the current term to the interior node tree. The term added to
    ** the interior tree must:
    **
    **   a) be greater than the largest term on the leaf node just written
    **      to the database (still available in pWriter->zTerm), and
    **
    **   b) be less than or equal to the term about to be added to the new
    **      leaf node (zTerm/nTerm).
    **
    ** In other words, it must be the prefix of zTerm 1 byte longer than
    ** the common prefix (if any) of zTerm and pWriter->zTerm.
    */
    assert( nPrefix<nTerm );
    rc = fts3NodeAddTerm(p, &pWriter->pTree, isCopyTerm, zTerm, nPrefix+1);
    if( rc!=SQLITE_OK ) return rc;

    nData = 0;
    pWriter->nTerm = 0;

    nPrefix = 0;
    nSuffix = nTerm;
    nReq = 1 +                              /* varint containing prefix size */
      sqlite3Fts3VarintLen(nTerm) +         /* varint containing suffix size */
      nTerm +                               /* Term suffix */
      sqlite3Fts3VarintLen(nDoclist) +      /* Size of doclist */
      nDoclist;                             /* Doclist data */
  }

  /* If the buffer currently allocated is too small for this entry, realloc
  ** the buffer to make it large enough.
  */
  if( nReq>pWriter->nSize ){
    char *aNew = sqlite3_realloc(pWriter->aData, nReq);
    if( !aNew ) return SQLITE_NOMEM;
    pWriter->aData = aNew;
    pWriter->nSize = nReq;
  }
  assert( nData+nReq<=pWriter->nSize );

  /* Append the prefix-compressed term and doclist to the buffer. */
  nData += sqlite3Fts3PutVarint(&pWriter->aData[nData], nPrefix);
  nData += sqlite3Fts3PutVarint(&pWriter->aData[nData], nSuffix);
  memcpy(&pWriter->aData[nData], &zTerm[nPrefix], nSuffix);
  nData += nSuffix;
  nData += sqlite3Fts3PutVarint(&pWriter->aData[nData], nDoclist);
  memcpy(&pWriter->aData[nData], aDoclist, nDoclist);
  pWriter->nData = nData + nDoclist;

  /* Save the current term so that it can be used to prefix-compress the next.
  ** If the isCopyTerm parameter is true, then the buffer pointed to by
  ** zTerm is transient, so take a copy of the term data. Otherwise, just
  ** store a copy of the pointer.
  */
  if( isCopyTerm ){
    if( nTerm>pWriter->nMalloc ){
      char *zNew = sqlite3_realloc(pWriter->zMalloc, nTerm*2);
      if( !zNew ){
        return SQLITE_NOMEM;
      }
      pWriter->nMalloc = nTerm*2;
      pWriter->zMalloc = zNew;
      pWriter->zTerm = zNew;
    }
    assert( pWriter->zTerm==pWriter->zMalloc );
    memcpy(pWriter->zTerm, zTerm, nTerm);
  }else{
    pWriter->zTerm = (char *)zTerm;
  }
  pWriter->nTerm = nTerm;

  return SQLITE_OK;
}

/*
** Flush all data associated with the SegmentWriter object pWriter to the
** database. This function must be called after all terms have been added
** to the segment using fts3SegWriterAdd(). If successful, SQLITE_OK is
** returned. Otherwise, an SQLite error code.
*/
static int fts3SegWriterFlush(
  Fts3Table *p,                   /* Virtual table handle */
  SegmentWriter *pWriter,         /* SegmentWriter to flush to the db */
  int iLevel,                     /* Value for 'level' column of %_segdir */
  int iIdx                        /* Value for 'idx' column of %_segdir */
){
  int rc;                         /* Return code */
  if( pWriter->pTree ){
    sqlite3_int64 iLast = 0;      /* Largest block id written to database */
    sqlite3_int64 iLastLeaf;      /* Largest leaf block id written to db */
    char *zRoot = NULL;           /* Pointer to buffer containing root node */
    int nRoot = 0;                /* Size of buffer zRoot */

    iLastLeaf = pWriter->iFree;
    rc = fts3WriteSegment(p, pWriter->iFree++, pWriter->aData, pWriter->nData);
    if( rc==SQLITE_OK ){
      rc = fts3NodeWrite(p, pWriter->pTree, 1,
          pWriter->iFirst, pWriter->iFree, &iLast, &zRoot, &nRoot);
    }
    if( rc==SQLITE_OK ){
      rc = fts3WriteSegdir(
          p, iLevel, iIdx, pWriter->iFirst, iLastLeaf, iLast, zRoot, nRoot);
    }
  }else{
    /* The entire tree fits on the root node. Write it to the segdir table. */
    rc = fts3WriteSegdir(
        p, iLevel, iIdx, 0, 0, 0, pWriter->aData, pWriter->nData);
  }
  return rc;
}

/*
** Release all memory held by the SegmentWriter object passed as the 
** first argument.
*/
static void fts3SegWriterFree(SegmentWriter *pWriter){
  if( pWriter ){
    sqlite3_free(pWriter->aData);
    sqlite3_free(pWriter->zMalloc);
    fts3NodeFree(pWriter->pTree);
    sqlite3_free(pWriter);
  }
}

/*
** The first value in the apVal[] array is assumed to contain an integer.
** This function tests if there exist any documents with docid values that
** are different from that integer. i.e. if deleting the document with docid
** apVal[0] would mean the FTS3 table were empty.
**
** If successful, *pisEmpty is set to true if the table is empty except for
** document apVal[0], or false otherwise, and SQLITE_OK is returned. If an
** error occurs, an SQLite error code is returned.
*/
static int fts3IsEmpty(Fts3Table *p, sqlite3_value **apVal, int *pisEmpty){
  sqlite3_stmt *pStmt;
  int rc;
  rc = fts3SqlStmt(p, SQL_IS_EMPTY, &pStmt, apVal);
  if( rc==SQLITE_OK ){
    if( SQLITE_ROW==sqlite3_step(pStmt) ){
      *pisEmpty = sqlite3_column_int(pStmt, 0);
    }
    rc = sqlite3_reset(pStmt);
  }
  return rc;
}

/*
** Set *pnSegment to the number of segments of level iLevel in the database.
**
** Return SQLITE_OK if successful, or an SQLite error code if not.
*/
static int fts3SegmentCount(Fts3Table *p, int iLevel, int *pnSegment){
  sqlite3_stmt *pStmt;
  int rc;

  assert( iLevel>=0 );
  rc = fts3SqlStmt(p, SQL_SELECT_LEVEL_COUNT, &pStmt, 0);
  if( rc!=SQLITE_OK ) return rc;
  sqlite3_bind_int(pStmt, 1, iLevel);
  if( SQLITE_ROW==sqlite3_step(pStmt) ){
    *pnSegment = sqlite3_column_int(pStmt, 0);
  }
  return sqlite3_reset(pStmt);
}

/*
** Set *pnSegment to the total number of segments in the database. Set
** *pnMax to the largest segment level in the database (segment levels
** are stored in the 'level' column of the %_segdir table).
**
** Return SQLITE_OK if successful, or an SQLite error code if not.
*/
static int fts3SegmentCountMax(Fts3Table *p, int *pnSegment, int *pnMax){
  sqlite3_stmt *pStmt;
  int rc;

  rc = fts3SqlStmt(p, SQL_SELECT_SEGDIR_COUNT_MAX, &pStmt, 0);
  if( rc!=SQLITE_OK ) return rc;
  if( SQLITE_ROW==sqlite3_step(pStmt) ){
    *pnSegment = sqlite3_column_int(pStmt, 0);
    *pnMax = sqlite3_column_int(pStmt, 1);
  }
  return sqlite3_reset(pStmt);
}

/*
** This function is used after merging multiple segments into a single large
** segment to delete the old, now redundant, segment b-trees. Specifically,
** it:
** 
**   1) Deletes all %_segments entries for the segments associated with 
**      each of the SegReader objects in the array passed as the third 
**      argument, and
**
**   2) deletes all %_segdir entries with level iLevel, or all %_segdir
**      entries regardless of level if (iLevel<0).
**
** SQLITE_OK is returned if successful, otherwise an SQLite error code.
*/
static int fts3DeleteSegdir(
  Fts3Table *p,                   /* Virtual table handle */
  int iLevel,                     /* Level of %_segdir entries to delete */
  Fts3SegReader **apSegment,      /* Array of SegReader objects */
  int nReader                     /* Size of array apSegment */
){
  int rc;                         /* Return Code */
  int i;                          /* Iterator variable */
  sqlite3_stmt *pDelete;          /* SQL statement to delete rows */

  rc = fts3SqlStmt(p, SQL_DELETE_SEGMENTS_RANGE, &pDelete, 0);
  for(i=0; rc==SQLITE_OK && i<nReader; i++){
    Fts3SegReader *pSegment = apSegment[i];
    if( pSegment->iStartBlock ){
      sqlite3_bind_int64(pDelete, 1, pSegment->iStartBlock);
      sqlite3_bind_int64(pDelete, 2, pSegment->iEndBlock);
      sqlite3_step(pDelete);
      rc = sqlite3_reset(pDelete);
    }
  }
  if( rc!=SQLITE_OK ){
    return rc;
  }

  if( iLevel>=0 ){
    rc = fts3SqlStmt(p, SQL_DELETE_SEGDIR_BY_LEVEL, &pDelete, 0);
    if( rc==SQLITE_OK ){
      sqlite3_bind_int(pDelete, 1, iLevel);
      sqlite3_step(pDelete);
      rc = sqlite3_reset(pDelete);
    }
  }else{
    rc = fts3SqlExec(p, SQL_DELETE_ALL_SEGDIR, 0);
  }

  return rc;
}

/*
** When this function is called, buffer *ppList (size *pnList bytes) contains 
** a position list that may (or may not) feature multiple columns. This
** function adjusts the pointer *ppList and the length *pnList so that they
** identify the subset of the position list that corresponds to column iCol.
**
** If there are no entries in the input position list for column iCol, then
** *pnList is set to zero before returning.
*/
static void fts3ColumnFilter(
  int iCol,                       /* Column to filter on */
  char **ppList,                  /* IN/OUT: Pointer to position list */
  int *pnList                     /* IN/OUT: Size of buffer *ppList in bytes */
){
  char *pList = *ppList;
  int nList = *pnList;
  char *pEnd = &pList[nList];
  int iCurrent = 0;
  char *p = pList;

  assert( iCol>=0 );
  while( 1 ){
    char c = 0;
    while( p<pEnd && (c | *p)&0xFE ) c = *p++ & 0x80;
  
    if( iCol==iCurrent ){
      nList = (int)(p - pList);
      break;
    }

    nList -= (int)(p - pList);
    pList = p;
    if( nList==0 ){
      break;
    }
    p = &pList[1];
    p += sqlite3Fts3GetVarint32(p, &iCurrent);
  }

  *ppList = pList;
  *pnList = nList;
}

/*
** sqlite3Fts3SegReaderIterate() callback used when merging multiple 
** segments to create a single, larger segment.
*/
static int fts3MergeCallback(
  Fts3Table *p,
  void *pContext,
  char *zTerm,
  int nTerm,
  char *aDoclist,
  int nDoclist
){
  SegmentWriter **ppW = (SegmentWriter **)pContext;
  return fts3SegWriterAdd(p, ppW, 1, zTerm, nTerm, aDoclist, nDoclist);
}

/*
** This function is used to iterate through a contiguous set of terms 
** stored in the full-text index. It merges data contained in one or 
** more segments to support this.
**
** The second argument is passed an array of pointers to SegReader objects
** allocated with sqlite3Fts3SegReaderNew(). This function merges the range 
** of terms selected by each SegReader. If a single term is present in
** more than one segment, the associated doclists are merged. For each
** term and (possibly merged) doclist in the merged range, the callback
** function xFunc is invoked with its arguments set as follows.
**
**   arg 0: Copy of 'p' parameter passed to this function
**   arg 1: Copy of 'pContext' parameter passed to this function
**   arg 2: Pointer to buffer containing term
**   arg 3: Size of arg 2 buffer in bytes
**   arg 4: Pointer to buffer containing doclist
**   arg 5: Size of arg 2 buffer in bytes
**
** The 4th argument to this function is a pointer to a structure of type
** Fts3SegFilter, defined in fts3Int.h. The contents of this structure
** further restrict the range of terms that callbacks are made for and
** modify the behaviour of this function. See comments above structure
** definition for details.
*/
int sqlite3Fts3SegReaderIterate(
  Fts3Table *p,                   /* Virtual table handle */
  Fts3SegReader **apSegment,      /* Array of Fts3SegReader objects */
  int nSegment,                   /* Size of apSegment array */
  Fts3SegFilter *pFilter,         /* Restrictions on range of iteration */
  int (*xFunc)(Fts3Table *, void *, char *, int, char *, int),  /* Callback */
  void *pContext                  /* Callback context (2nd argument) */
){
  int i;                          /* Iterator variable */
  char *aBuffer = 0;              /* Buffer to merge doclists in */
  int nAlloc = 0;                 /* Allocated size of aBuffer buffer */
  int rc = SQLITE_OK;             /* Return code */

  int isIgnoreEmpty =  (pFilter->flags & FTS3_SEGMENT_IGNORE_EMPTY);
  int isRequirePos =   (pFilter->flags & FTS3_SEGMENT_REQUIRE_POS);
  int isColFilter =    (pFilter->flags & FTS3_SEGMENT_COLUMN_FILTER);
  int isPrefix =       (pFilter->flags & FTS3_SEGMENT_PREFIX);

  /* If there are zero segments, this function is a no-op. This scenario
  ** comes about only when reading from an empty database.
  */
  if( nSegment==0 ) goto finished;

  /* If the Fts3SegFilter defines a specific term (or term prefix) to search 
  ** for, then advance each segment iterator until it points to a term of
  ** equal or greater value than the specified term. This prevents many
  ** unnecessary merge/sort operations for the case where single segment
  ** b-tree leaf nodes contain more than one term.
  */
  if( pFilter->zTerm ){
    int nTerm = pFilter->nTerm;
    const char *zTerm = pFilter->zTerm;
    for(i=0; i<nSegment; i++){
      Fts3SegReader *pSeg = apSegment[i];
      while( fts3SegReaderTermCmp(pSeg, zTerm, nTerm)<0 ){
        rc = fts3SegReaderNext(pSeg);
        if( rc!=SQLITE_OK ) goto finished;
      }
    }
  }

  fts3SegReaderSort(apSegment, nSegment, nSegment, fts3SegReaderCmp);
  while( apSegment[0]->aNode ){
    int nTerm = apSegment[0]->nTerm;
    char *zTerm = apSegment[0]->zTerm;
    int nMerge = 1;

    /* If this is a prefix-search, and if the term that apSegment[0] points
    ** to does not share a suffix with pFilter->zTerm/nTerm, then all 
    ** required callbacks have been made. In this case exit early.
    **
    ** Similarly, if this is a search for an exact match, and the first term
    ** of segment apSegment[0] is not a match, exit early.
    */
    if( pFilter->zTerm ){
      if( nTerm<pFilter->nTerm 
       || (!isPrefix && nTerm>pFilter->nTerm)
       || memcmp(zTerm, pFilter->zTerm, pFilter->nTerm) 
    ){
        goto finished;
      }
    }

    while( nMerge<nSegment 
        && apSegment[nMerge]->aNode
        && apSegment[nMerge]->nTerm==nTerm 
        && 0==memcmp(zTerm, apSegment[nMerge]->zTerm, nTerm)
    ){
      nMerge++;
    }

    if( nMerge==1 && !isIgnoreEmpty && !isColFilter && isRequirePos ){
      Fts3SegReader *p0 = apSegment[0];
      rc = xFunc(p, pContext, zTerm, nTerm, p0->aDoclist, p0->nDoclist);
      if( rc!=SQLITE_OK ) goto finished;
    }else{
      int nDoclist = 0;           /* Size of doclist */
      sqlite3_int64 iPrev = 0;    /* Previous docid stored in doclist */

      /* The current term of the first nMerge entries in the array
      ** of Fts3SegReader objects is the same. The doclists must be merged
      ** and a single term added to the new segment.
      */
      for(i=0; i<nMerge; i++){
        fts3SegReaderFirstDocid(apSegment[i]);
      }
      fts3SegReaderSort(apSegment, nMerge, nMerge, fts3SegReaderDoclistCmp);
      while( apSegment[0]->pOffsetList ){
        int j;                    /* Number of segments that share a docid */
        char *pList;
        int nList;
        int nByte;
        sqlite3_int64 iDocid = apSegment[0]->iDocid;
        fts3SegReaderNextDocid(apSegment[0], &pList, &nList);
        j = 1;
        while( j<nMerge 
            && apSegment[j]->pOffsetList 
            && apSegment[j]->iDocid==iDocid 
        ){
          fts3SegReaderNextDocid(apSegment[j], 0, 0);
          j++;
        }

        if( isColFilter ){
          fts3ColumnFilter(pFilter->iCol, &pList, &nList);
        }

        if( !isIgnoreEmpty || nList>0 ){
          nByte = sqlite3Fts3VarintLen(iDocid-iPrev) + (isRequirePos?nList+1:0);
          if( nDoclist+nByte>nAlloc ){
            char *aNew;
            nAlloc = nDoclist+nByte*2;
            aNew = sqlite3_realloc(aBuffer, nAlloc);
            if( !aNew ){
              rc = SQLITE_NOMEM;
              goto finished;
            }
            aBuffer = aNew;
          }
          nDoclist += sqlite3Fts3PutVarint(&aBuffer[nDoclist], iDocid-iPrev);
          iPrev = iDocid;
          if( isRequirePos ){
            memcpy(&aBuffer[nDoclist], pList, nList);
            nDoclist += nList;
            aBuffer[nDoclist++] = '\0';
          }
        }

        fts3SegReaderSort(apSegment, nMerge, j, fts3SegReaderDoclistCmp);
      }

      if( nDoclist>0 ){
        rc = xFunc(p, pContext, zTerm, nTerm, aBuffer, nDoclist);
        if( rc!=SQLITE_OK ) goto finished;
      }
    }

    /* If there is a term specified to filter on, and this is not a prefix
    ** search, return now. The callback that corresponds to the required
    ** term (if such a term exists in the index) has already been made.
    */
    if( pFilter->zTerm && !isPrefix ){
      goto finished;
    }

    for(i=0; i<nMerge; i++){
      rc = fts3SegReaderNext(apSegment[i]);
      if( rc!=SQLITE_OK ) goto finished;
    }
    fts3SegReaderSort(apSegment, nSegment, nMerge, fts3SegReaderCmp);
  }

 finished:
  sqlite3_free(aBuffer);
  return rc;
}

/*
** Merge all level iLevel segments in the database into a single 
** iLevel+1 segment. Or, if iLevel<0, merge all segments into a
** single segment with a level equal to the numerically largest level 
** currently present in the database.
**
** If this function is called with iLevel<0, but there is only one
** segment in the database, SQLITE_DONE is returned immediately. 
** Otherwise, if successful, SQLITE_OK is returned. If an error occurs, 
** an SQLite error code is returned.
*/
static int fts3SegmentMerge(Fts3Table *p, int iLevel){
  int i;                          /* Iterator variable */
  int rc;                         /* Return code */
  int iIdx;                       /* Index of new segment */
  int iNewLevel;                  /* Level to create new segment at */
  sqlite3_stmt *pStmt;
  SegmentWriter *pWriter = 0;
  int nSegment = 0;               /* Number of segments being merged */
  Fts3SegReader **apSegment = 0;  /* Array of Segment iterators */
  Fts3SegFilter filter;           /* Segment term filter condition */

  if( iLevel<0 ){
    /* This call is to merge all segments in the database to a single
    ** segment. The level of the new segment is equal to the the numerically 
    ** greatest segment level currently present in the database. The index
    ** of the new segment is always 0.
    */
    iIdx = 0;
    rc = fts3SegmentCountMax(p, &nSegment, &iNewLevel);
    if( nSegment==1 ){
      return SQLITE_DONE;
    }
  }else{
    /* This call is to merge all segments at level iLevel. Find the next
    ** available segment index at level iLevel+1. The call to
    ** fts3AllocateSegdirIdx() will merge the segments at level iLevel+1 to 
    ** a single iLevel+2 segment if necessary.
    */
    iNewLevel = iLevel+1;
    rc = fts3AllocateSegdirIdx(p, iNewLevel, &iIdx);
    if( rc!=SQLITE_OK ) return rc;
    rc = fts3SegmentCount(p, iLevel, &nSegment);
  }
  if( rc!=SQLITE_OK ) return rc;
  assert( nSegment>0 );
  assert( iNewLevel>=0 );

  /* Allocate space for an array of pointers to segment iterators. */
  apSegment = (Fts3SegReader**)sqlite3_malloc(sizeof(Fts3SegReader *)*nSegment);
  if( !apSegment ){
    return SQLITE_NOMEM;
  }
  memset(apSegment, 0, sizeof(Fts3SegReader *)*nSegment);

  /* Allocate a Fts3SegReader structure for each segment being merged. A 
  ** Fts3SegReader stores the state data required to iterate through all 
  ** entries on all leaves of a single segment. 
  */
  assert( SQL_SELECT_LEVEL+1==SQL_SELECT_ALL_LEVEL);
  rc = fts3SqlStmt(p, SQL_SELECT_LEVEL+(iLevel<0), &pStmt, 0);
  if( rc!=SQLITE_OK ) goto finished;
  sqlite3_bind_int(pStmt, 1, iLevel);
  for(i=0; SQLITE_ROW==(sqlite3_step(pStmt)); i++){
    rc = fts3SegReaderNew(p, pStmt, i, &apSegment[i]);
    if( rc!=SQLITE_OK ){
      goto finished;
    }
  }
  rc = sqlite3_reset(pStmt);
  pStmt = 0;
  if( rc!=SQLITE_OK ) goto finished;

  memset(&filter, 0, sizeof(Fts3SegFilter));
  filter.flags = FTS3_SEGMENT_REQUIRE_POS;
  filter.flags |= (iLevel<0 ? FTS3_SEGMENT_IGNORE_EMPTY : 0);
  rc = sqlite3Fts3SegReaderIterate(p, apSegment, nSegment,
      &filter, fts3MergeCallback, (void *)&pWriter
  );
  if( rc!=SQLITE_OK ) goto finished;

  rc = fts3DeleteSegdir(p, iLevel, apSegment, nSegment);
  if( rc==SQLITE_OK ){
    rc = fts3SegWriterFlush(p, pWriter, iNewLevel, iIdx);
  }

 finished:
  fts3SegWriterFree(pWriter);
  if( apSegment ){
    for(i=0; i<nSegment; i++){
      sqlite3Fts3SegReaderFree(p, apSegment[i]);
    }
    sqlite3_free(apSegment);
  }
  sqlite3_reset(pStmt);
  return rc;
}

/*
** This is a comparison function used as a qsort() callback when sorting
** an array of pending terms by term. This occurs as part of flushing
** the contents of the pending-terms hash table to the database.
*/
static int qsortCompare(const void *lhs, const void *rhs){
  char *z1 = fts3HashKey(*(Fts3HashElem **)lhs);
  char *z2 = fts3HashKey(*(Fts3HashElem **)rhs);
  int n1 = fts3HashKeysize(*(Fts3HashElem **)lhs);
  int n2 = fts3HashKeysize(*(Fts3HashElem **)rhs);

  int n = (n1<n2 ? n1 : n2);
  int c = memcmp(z1, z2, n);
  if( c==0 ){
    c = n1 - n2;
  }
  return c;
}


/* 
** Flush the contents of pendingTerms to a level 0 segment.
*/
int sqlite3Fts3PendingTermsFlush(Fts3Table *p){
  Fts3HashElem *pElem;
  int idx, rc, i;
  Fts3HashElem **apElem;          /* Array of pointers to hash elements */
  int nElem;                      /* Number of terms in new segment */
  SegmentWriter *pWriter = 0;     /* Used to write the segment */

  /* Find the number of terms that will make up the new segment. If there
  ** are no terms, return early (do not bother to write an empty segment).
  */
  nElem = fts3HashCount(&p->pendingTerms);
  if( nElem==0 ){
    assert( p->nPendingData==0 );
    return SQLITE_OK;
  }

  /* Determine the next index at level 0, merging as necessary. */
  rc = fts3AllocateSegdirIdx(p, 0, &idx);
  if( rc!=SQLITE_OK ){
    return rc;
  } 

  apElem = sqlite3_malloc(nElem*sizeof(Fts3HashElem *));
  if( !apElem ){
    return SQLITE_NOMEM;
  }

  i = 0;
  for(pElem=fts3HashFirst(&p->pendingTerms); pElem; pElem=fts3HashNext(pElem)){
    apElem[i++] = pElem;
  }
  assert( i==nElem );

  /* TODO(shess) Should we allow user-defined collation sequences,
  ** here?  I think we only need that once we support prefix searches.
  ** Also, should we be using qsort()?
  */
  if( nElem>1 ){
    qsort(apElem, nElem, sizeof(Fts3HashElem *), qsortCompare);
  }


  /* Write the segment tree into the database. */
  for(i=0; rc==SQLITE_OK && i<nElem; i++){
    const char *z = fts3HashKey(apElem[i]);
    int n = fts3HashKeysize(apElem[i]);
    PendingList *pList = fts3HashData(apElem[i]);
    rc = fts3SegWriterAdd(p, &pWriter, 0, z, n, pList->aData, pList->nData+1);
  }
  if( rc==SQLITE_OK ){
    rc = fts3SegWriterFlush(p, pWriter, 0, idx);
  }

  /* Free all allocated resources before returning */
  fts3SegWriterFree(pWriter);
  sqlite3_free(apElem);
  sqlite3Fts3PendingTermsClear(p);
  return rc;
}

/*
** This function does the work for the xUpdate method of FTS3 virtual
** tables.
*/
int sqlite3Fts3UpdateMethod(
  sqlite3_vtab *pVtab,            /* FTS3 vtab object */
  int nArg,                       /* Size of argument array */
  sqlite3_value **apVal,          /* Array of arguments */
  sqlite_int64 *pRowid            /* OUT: The affected (or effected) rowid */
){
  Fts3Table *p = (Fts3Table *)pVtab;
  int rc = SQLITE_OK;             /* Return Code */
  int isRemove = 0;               /* True for an UPDATE or DELETE */
  sqlite3_int64 iRemove = 0;      /* Rowid removed by UPDATE or DELETE */

  /* If this is a DELETE or UPDATE operation, remove the old record. */
  if( sqlite3_value_type(apVal[0])!=SQLITE_NULL ){
    int isEmpty;
    rc = fts3IsEmpty(p, apVal, &isEmpty);
    if( rc==SQLITE_OK ){
      if( isEmpty ){
        /* Deleting this row means the whole table is empty. In this case
        ** delete the contents of all three tables and throw away any
        ** data in the pendingTerms hash table.
        */
        rc = fts3DeleteAll(p);
      }else{
        isRemove = 1;
        iRemove = sqlite3_value_int64(apVal[0]);
        rc = fts3PendingTermsDocid(p, iRemove);
        if( rc==SQLITE_OK ){
          rc = fts3DeleteTerms(p, apVal);
          if( rc==SQLITE_OK ){
            rc = fts3SqlExec(p, SQL_DELETE_CONTENT, apVal);
          }
        }
      }
    }
  }
  
  /* If this is an INSERT or UPDATE operation, insert the new record. */
  if( nArg>1 && rc==SQLITE_OK ){
    rc = fts3InsertData(p, apVal, pRowid);
    if( rc==SQLITE_OK && (!isRemove || *pRowid!=iRemove) ){
      rc = fts3PendingTermsDocid(p, *pRowid);
    }
    if( rc==SQLITE_OK ){
      rc = fts3InsertTerms(p, apVal);
    }
  }

  return rc;
}

/* 
** Flush any data in the pending-terms hash table to disk. If successful,
** merge all segments in the database (including the new segment, if 
** there was any data to flush) into a single segment. 
*/
int sqlite3Fts3Optimize(Fts3Table *p){
  int rc;
  rc = sqlite3_exec(p->db, "SAVEPOINT fts3", 0, 0, 0);
  if( rc==SQLITE_OK ){
    rc = sqlite3Fts3PendingTermsFlush(p);
    if( rc==SQLITE_OK ){
      rc = fts3SegmentMerge(p, -1);
    }
    if( rc==SQLITE_OK ){
      rc = sqlite3_exec(p->db, "RELEASE fts3", 0, 0, 0);
    }else{
      sqlite3_exec(p->db, "ROLLBACK TO fts3 ; RELEASE fts3", 0, 0, 0);
    }
  }
  return rc;
}

#endif

Changes to ext/rtree/rtree.c.

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains code for implementations of the r-tree and r*-tree
** algorithms packaged as an SQLite virtual table module.
**
** $Id: rtree.c,v 1.14 2009/08/06 18:36:47 danielk1977 Exp $
*/

#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_RTREE)

/*
** This file contains an implementation of a couple of different variants
** of the r-tree algorithm. See the README file for further details. The 







<
<







7
8
9
10
11
12
13


14
15
16
17
18
19
20
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains code for implementations of the r-tree and r*-tree
** algorithms packaged as an SQLite virtual table module.


*/

#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_RTREE)

/*
** This file contains an implementation of a couple of different variants
** of the r-tree algorithm. See the README file for further details. The 

Changes to ext/rtree/rtree1.test.

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# The focus of this file is testing the r-tree extension.
#
# $Id: rtree1.test,v 1.7 2009/07/17 16:54:48 danielk1977 Exp $
#

if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
}
source [file join [file dirname [info script]] rtree_util.tcl]
source $testdir/tester.tcl








<
<







7
8
9
10
11
12
13


14
15
16
17
18
19
20
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# The focus of this file is testing the r-tree extension.
#



if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
}
source [file join [file dirname [info script]] rtree_util.tcl]
source $testdir/tester.tcl

Changes to ext/rtree/rtree2.test.

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# The focus of this file is testing the r-tree extension.
#
# $Id: rtree2.test,v 1.4 2008/07/14 15:37:01 danielk1977 Exp $
#

if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
} 
source [file join [file dirname [info script]] rtree_util.tcl]
source $testdir/tester.tcl








<
<







7
8
9
10
11
12
13


14
15
16
17
18
19
20
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# The focus of this file is testing the r-tree extension.
#



if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
} 
source [file join [file dirname [info script]] rtree_util.tcl]
source $testdir/tester.tcl

Changes to ext/rtree/rtree3.test.

8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
..
67
68
69
70
71
72
73
74
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# The focus of this file is testing that the r-tree correctly handles
# out-of-memory conditions.
#
# $Id: rtree3.test,v 1.2 2008/06/23 15:55:52 danielk1977 Exp $
#

if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
} 
source $testdir/tester.tcl

ifcapable !rtree {
................................................................................
    set f [expr rand()]
    db eval { DELETE FROM rt WHERE x1<($f*10.0) AND x1>($f*10.5) }
  }
  db eval COMMIT
} 

finish_test








<
<







 







<
8
9
10
11
12
13
14


15
16
17
18
19
20
21
..
65
66
67
68
69
70
71

#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# The focus of this file is testing that the r-tree correctly handles
# out-of-memory conditions.
#



if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
} 
source $testdir/tester.tcl

ifcapable !rtree {
................................................................................
    set f [expr rand()]
    db eval { DELETE FROM rt WHERE x1<($f*10.0) AND x1>($f*10.5) }
  }
  db eval COMMIT
} 

finish_test

Changes to ext/rtree/rtree4.test.

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# Randomized test cases for the rtree extension.
#
# $Id: rtree4.test,v 1.3 2008/06/23 15:55:52 danielk1977 Exp $
#

if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
} 
source $testdir/tester.tcl

ifcapable !rtree {







<
<







7
8
9
10
11
12
13


14
15
16
17
18
19
20
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# Randomized test cases for the rtree extension.
#



if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
} 
source $testdir/tester.tcl

ifcapable !rtree {

Changes to ext/rtree/rtree5.test.

8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# The focus of this file is testing the r-tree extension when it is
# configured to store values as 32 bit integers.
#
# $Id: rtree5.test,v 1.1 2008/07/14 15:37:01 danielk1977 Exp $
#

if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
} 
source $testdir/tester.tcl

ifcapable !rtree {







<
<







8
9
10
11
12
13
14


15
16
17
18
19
20
21
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# The focus of this file is testing the r-tree extension when it is
# configured to store values as 32 bit integers.
#



if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
} 
source $testdir/tester.tcl

ifcapable !rtree {

Changes to ext/rtree/rtree6.test.

4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
...
104
105
106
107
108
109
110
111
# a legal notice, here is a blessing:
#
#    May you do good and not evil.
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# $Id: rtree6.test,v 1.1 2008/09/01 12:47:00 danielk1977 Exp $
#

if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
} 
source $testdir/tester.tcl

................................................................................
  query_plan {SELECT * FROM t1,t2 WHERE k=ii AND x1<v}
} [list \
  {TABLE t2}                              \
  {TABLE t1 VIRTUAL TABLE INDEX 1:}   \
]

finish_test








|
<







 







<
4
5
6
7
8
9
10
11

12
13
14
15
16
17
18
...
103
104
105
106
107
108
109

# a legal notice, here is a blessing:
#
#    May you do good and not evil.
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
# 

#

if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
} 
source $testdir/tester.tcl

................................................................................
  query_plan {SELECT * FROM t1,t2 WHERE k=ii AND x1<v}
} [list \
  {TABLE t2}                              \
  {TABLE t1 VIRTUAL TABLE INDEX 1:}   \
]

finish_test

Changes to ext/rtree/rtree_perf.tcl.

68
69
70
71
72
73
74
75
76
flush stdout
set rtree_select_time [time {
  foreach {x1 x2 y1 y2} [lrange $data 0 [expr $NQUERY*4-1]] {
    db eval {SELECT * FROM rtree WHERE x1<$x1 AND x2>$x2 AND y1<$y1 AND y2>$y2}
  }
}]
puts "$rtree_select_time"









<
<
68
69
70
71
72
73
74


flush stdout
set rtree_select_time [time {
  foreach {x1 x2 y1 y2} [lrange $data 0 [expr $NQUERY*4-1]] {
    db eval {SELECT * FROM rtree WHERE x1<$x1 AND x2>$x2 AND y1<$y1 AND y2>$y2}
  }
}]
puts "$rtree_select_time"


Changes to ext/rtree/rtree_util.tcl.

9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
...
188
189
190
191
192
193
194
195
#
#***********************************************************************
#
# This file contains Tcl code that may be useful for testing or
# analyzing r-tree structures created with this module. It is
# used by both test procedures and the r-tree viewer application.
#
# $Id: rtree_util.tcl,v 1.1 2008/05/26 18:41:54 danielk1977 Exp $
#


#--------------------------------------------------------------------------
# PUBLIC API:
#
#   rtree_depth
#   rtree_ndim
................................................................................
  set ret
}

proc rtree_treedump {db zTab} {
  set d [rtree_depth $db $zTab]
  rtree_nodetreedump $db $zTab "" $d 1
}








<
<







 







<
9
10
11
12
13
14
15


16
17
18
19
20
21
22
...
186
187
188
189
190
191
192

#
#***********************************************************************
#
# This file contains Tcl code that may be useful for testing or
# analyzing r-tree structures created with this module. It is
# used by both test procedures and the r-tree viewer application.
#




#--------------------------------------------------------------------------
# PUBLIC API:
#
#   rtree_depth
#   rtree_ndim
................................................................................
  set ret
}

proc rtree_treedump {db zTab} {
  set d [rtree_depth $db $zTab]
  rtree_nodetreedump $db $zTab "" $d 1
}

Changes to ext/rtree/tkt3363.test.

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
..
46
47
48
49
50
51
52
53
54
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# The focus of this file is testing that ticket #3363 is fixed.
#
# $Id: tkt3363.test,v 1.1 2008/09/08 11:07:03 danielk1977 Exp $
#

if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
}
source [file join [file dirname [info script]] rtree_util.tcl]
source $testdir/tester.tcl

................................................................................
do_test tkt3363.1.4 {
  execsql { 
    SELECT count(*) FROM t1 WHERE y2>4000425.0;
  }
} {7}

finish_test









<
<







 







<
<
7
8
9
10
11
12
13


14
15
16
17
18
19
20
..
44
45
46
47
48
49
50


#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# The focus of this file is testing that ticket #3363 is fixed.
#



if {![info exists testdir]} {
  set testdir [file join [file dirname $argv0] .. .. test]
}
source [file join [file dirname [info script]] rtree_util.tcl]
source $testdir/tester.tcl

................................................................................
do_test tkt3363.1.4 {
  execsql { 
    SELECT count(*) FROM t1 WHERE y2>4000425.0;
  }
} {7}

finish_test


Changes to ext/rtree/viewrtree.tcl.

182
183
184
185
186
187
188
189
  }

  return $zReport
}

view_node
bind .c <Configure> view_node








<
182
183
184
185
186
187
188

  }

  return $zReport
}

view_node
bind .c <Configure> view_node

Changes to main.mk.

50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
...
150
151
152
153
154
155
156

157
158
159
160
161
162
163
...
178
179
180
181
182
183
184
185
186
187
188
189
190

191
192
193

194
195
196
197
198
199
200
...
227
228
229
230
231
232
233

234
235
236
237
238
239
240
...
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
...
429
430
431
432
433
434
435



436
437
438
439
440
441
442
443
444



445
446
447
448
449
450
451

# Object files for the SQLite library.
#
LIBOBJ+= alter.o analyze.o attach.o auth.o \
         backup.o bitvec.o btmutex.o btree.o build.o \
         callback.o complete.o date.o delete.o expr.o fault.o fkey.o \
         fts3.o fts3_expr.o fts3_hash.o fts3_icu.o fts3_porter.o \
         fts3_tokenizer.o fts3_tokenizer1.o \
         func.o global.o hash.o \
         icu.o insert.o journal.o legacy.o loadext.o \
         main.o malloc.o mem0.o mem1.o mem2.o mem3.o mem5.o \
         memjournal.o \
         mutex.o mutex_noop.o mutex_os2.o mutex_unix.o mutex_w32.o \
         notify.o opcodes.o os.o os_os2.o os_unix.o os_win.o \
         pager.o parse.o pcache.o pcache1.o pragma.o prepare.o printf.o \
         random.o resolve.o rowset.o rtree.o select.o status.o \
         table.o tokenize.o trigger.o \
         update.o util.o vacuum.o \
         vdbe.o vdbeapi.o vdbeaux.o vdbeblob.o vdbemem.o \
         walker.o where.o utf.o vtab.o



# All of the source code files.
#
SRC = \
................................................................................
  $(TOP)/src/vacuum.c \
  $(TOP)/src/vdbe.c \
  $(TOP)/src/vdbe.h \
  $(TOP)/src/vdbeapi.c \
  $(TOP)/src/vdbeaux.c \
  $(TOP)/src/vdbeblob.c \
  $(TOP)/src/vdbemem.c \

  $(TOP)/src/vdbeInt.h \
  $(TOP)/src/vtab.c \
  $(TOP)/src/walker.c \
  $(TOP)/src/where.c

# Source code for extensions
#
................................................................................
  $(TOP)/ext/fts2/fts2_porter.c \
  $(TOP)/ext/fts2/fts2_tokenizer.h \
  $(TOP)/ext/fts2/fts2_tokenizer.c \
  $(TOP)/ext/fts2/fts2_tokenizer1.c
SRC += \
  $(TOP)/ext/fts3/fts3.c \
  $(TOP)/ext/fts3/fts3.h \
  $(TOP)/ext/fts3/fts3_expr.c \
  $(TOP)/ext/fts3/fts3_expr.h \
  $(TOP)/ext/fts3/fts3_hash.c \
  $(TOP)/ext/fts3/fts3_hash.h \
  $(TOP)/ext/fts3/fts3_icu.c \
  $(TOP)/ext/fts3/fts3_porter.c \

  $(TOP)/ext/fts3/fts3_tokenizer.h \
  $(TOP)/ext/fts3/fts3_tokenizer.c \
  $(TOP)/ext/fts3/fts3_tokenizer1.c

SRC += \
  $(TOP)/ext/icu/sqliteicu.h \
  $(TOP)/ext/icu/icu.c
SRC += \
  $(TOP)/ext/rtree/rtree.h \
  $(TOP)/ext/rtree/rtree.c

................................................................................
  $(TOP)/src/test_backup.c \
  $(TOP)/src/test_btree.c \
  $(TOP)/src/test_config.c \
  $(TOP)/src/test_devsym.c \
  $(TOP)/src/test_func.c \
  $(TOP)/src/test_hexio.c \
  $(TOP)/src/test_init.c \

  $(TOP)/src/test_journal.c \
  $(TOP)/src/test_malloc.c \
  $(TOP)/src/test_mutex.c \
  $(TOP)/src/test_onefile.c \
  $(TOP)/src/test_osinst.c \
  $(TOP)/src/test_pcache.c \
  $(TOP)/src/test_schema.c \
................................................................................
  $(TOP)/ext/fts1/fts1_tokenizer.h
EXTHDR += \
  $(TOP)/ext/fts2/fts2.h \
  $(TOP)/ext/fts2/fts2_hash.h \
  $(TOP)/ext/fts2/fts2_tokenizer.h
EXTHDR += \
  $(TOP)/ext/fts3/fts3.h \
  $(TOP)/ext/fts3/fts3_expr.h \
  $(TOP)/ext/fts3/fts3_hash.h \
  $(TOP)/ext/fts3/fts3_tokenizer.h
EXTHDR += \
  $(TOP)/ext/rtree/rtree.h
EXTHDR += \
  $(TOP)/ext/icu/sqliteicu.h

................................................................................
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_expr.c

fts3_hash.o:	$(TOP)/ext/fts3/fts3_hash.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_hash.c

fts3_icu.o:	$(TOP)/ext/fts3/fts3_icu.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_icu.c




fts3_porter.o:	$(TOP)/ext/fts3/fts3_porter.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_porter.c

fts3_tokenizer.o:	$(TOP)/ext/fts3/fts3_tokenizer.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_tokenizer.c

fts3_tokenizer1.o:	$(TOP)/ext/fts3/fts3_tokenizer1.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_tokenizer1.c




rtree.o:	$(TOP)/ext/rtree/rtree.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/rtree/rtree.c


# Rules for building test programs and for running tests
#







|










|







 







>







 







|
|




>


|
>







 







>







 







|







 







>
>
>









>
>
>







50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
...
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
...
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
...
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
...
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
...
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461

# Object files for the SQLite library.
#
LIBOBJ+= alter.o analyze.o attach.o auth.o \
         backup.o bitvec.o btmutex.o btree.o build.o \
         callback.o complete.o date.o delete.o expr.o fault.o fkey.o \
         fts3.o fts3_expr.o fts3_hash.o fts3_icu.o fts3_porter.o \
         fts3_snippet.o fts3_tokenizer.o fts3_tokenizer1.o fts3_write.o \
         func.o global.o hash.o \
         icu.o insert.o journal.o legacy.o loadext.o \
         main.o malloc.o mem0.o mem1.o mem2.o mem3.o mem5.o \
         memjournal.o \
         mutex.o mutex_noop.o mutex_os2.o mutex_unix.o mutex_w32.o \
         notify.o opcodes.o os.o os_os2.o os_unix.o os_win.o \
         pager.o parse.o pcache.o pcache1.o pragma.o prepare.o printf.o \
         random.o resolve.o rowset.o rtree.o select.o status.o \
         table.o tokenize.o trigger.o \
         update.o util.o vacuum.o \
         vdbe.o vdbeapi.o vdbeaux.o vdbeblob.o vdbemem.o vdbetrace.o \
         walker.o where.o utf.o vtab.o



# All of the source code files.
#
SRC = \
................................................................................
  $(TOP)/src/vacuum.c \
  $(TOP)/src/vdbe.c \
  $(TOP)/src/vdbe.h \
  $(TOP)/src/vdbeapi.c \
  $(TOP)/src/vdbeaux.c \
  $(TOP)/src/vdbeblob.c \
  $(TOP)/src/vdbemem.c \
  $(TOP)/src/vdbetrace.c \
  $(TOP)/src/vdbeInt.h \
  $(TOP)/src/vtab.c \
  $(TOP)/src/walker.c \
  $(TOP)/src/where.c

# Source code for extensions
#
................................................................................
  $(TOP)/ext/fts2/fts2_porter.c \
  $(TOP)/ext/fts2/fts2_tokenizer.h \
  $(TOP)/ext/fts2/fts2_tokenizer.c \
  $(TOP)/ext/fts2/fts2_tokenizer1.c
SRC += \
  $(TOP)/ext/fts3/fts3.c \
  $(TOP)/ext/fts3/fts3.h \
  $(TOP)/ext/fts3/fts3Int.h \
  $(TOP)/ext/fts3/fts3_expr.c \
  $(TOP)/ext/fts3/fts3_hash.c \
  $(TOP)/ext/fts3/fts3_hash.h \
  $(TOP)/ext/fts3/fts3_icu.c \
  $(TOP)/ext/fts3/fts3_porter.c \
  $(TOP)/ext/fts3/fts3_snippet.c \
  $(TOP)/ext/fts3/fts3_tokenizer.h \
  $(TOP)/ext/fts3/fts3_tokenizer.c \
  $(TOP)/ext/fts3/fts3_tokenizer1.c \
  $(TOP)/ext/fts3/fts3_write.c
SRC += \
  $(TOP)/ext/icu/sqliteicu.h \
  $(TOP)/ext/icu/icu.c
SRC += \
  $(TOP)/ext/rtree/rtree.h \
  $(TOP)/ext/rtree/rtree.c

................................................................................
  $(TOP)/src/test_backup.c \
  $(TOP)/src/test_btree.c \
  $(TOP)/src/test_config.c \
  $(TOP)/src/test_devsym.c \
  $(TOP)/src/test_func.c \
  $(TOP)/src/test_hexio.c \
  $(TOP)/src/test_init.c \
  $(TOP)/src/test_intarray.c \
  $(TOP)/src/test_journal.c \
  $(TOP)/src/test_malloc.c \
  $(TOP)/src/test_mutex.c \
  $(TOP)/src/test_onefile.c \
  $(TOP)/src/test_osinst.c \
  $(TOP)/src/test_pcache.c \
  $(TOP)/src/test_schema.c \
................................................................................
  $(TOP)/ext/fts1/fts1_tokenizer.h
EXTHDR += \
  $(TOP)/ext/fts2/fts2.h \
  $(TOP)/ext/fts2/fts2_hash.h \
  $(TOP)/ext/fts2/fts2_tokenizer.h
EXTHDR += \
  $(TOP)/ext/fts3/fts3.h \
  $(TOP)/ext/fts3/fts3Int.h \
  $(TOP)/ext/fts3/fts3_hash.h \
  $(TOP)/ext/fts3/fts3_tokenizer.h
EXTHDR += \
  $(TOP)/ext/rtree/rtree.h
EXTHDR += \
  $(TOP)/ext/icu/sqliteicu.h

................................................................................
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_expr.c

fts3_hash.o:	$(TOP)/ext/fts3/fts3_hash.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_hash.c

fts3_icu.o:	$(TOP)/ext/fts3/fts3_icu.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_icu.c

fts3_snippet.o:	$(TOP)/ext/fts3/fts3_snippet.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_snippet.c

fts3_porter.o:	$(TOP)/ext/fts3/fts3_porter.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_porter.c

fts3_tokenizer.o:	$(TOP)/ext/fts3/fts3_tokenizer.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_tokenizer.c

fts3_tokenizer1.o:	$(TOP)/ext/fts3/fts3_tokenizer1.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_tokenizer1.c

fts3_write.o:	$(TOP)/ext/fts3/fts3_write.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_write.c

rtree.o:	$(TOP)/ext/rtree/rtree.c $(HDR) $(EXTHDR)
	$(TCCX) -DSQLITE_CORE -c $(TOP)/ext/rtree/rtree.c


# Rules for building test programs and for running tests
#

Changes to mkopcodeh.awk.

45
46
47
48
49
50
51

52
53
54
55
56
57
58
..
66
67
68
69
70
71
72


73
74
75
76
77
78
79
...
121
122
123
124
125
126
127
128
129
130
131
132
133

134
135
136
137
138
139
140
141
142
143
144
145
146
147

148
149
150
151
152
153
154
155
156
  sub("\r","",name)
  op[name] = -1
  jump[name] = 0
  out2_prerelease[name] = 0
  in1[name] = 0
  in2[name] = 0
  in3[name] = 0

  out3[name] = 0
  for(i=3; i<NF; i++){
    if($i=="same" && $(i+1)=="as"){
      sym = $(i+2)
      sub(/,/,"",sym)
      op[name] = tk[sym]
      used[op[name]] = 1
................................................................................
      out2_prerelease[name] = 1
    }else if(x=="in1"){
      in1[name] = 1
    }else if(x=="in2"){
      in2[name] = 1
    }else if(x=="in3"){
      in3[name] = 1


    }else if(x=="out3"){
      out3[name] = 1
    }
  }
  order[n_op++] = name;
}

................................................................................
  #  bit 2:     output to p1.  release p1 before opcode runs
  #
  for(i=0; i<=max; i++) bv[i] = 0;
  for(i=0; i<n_op; i++){
    name = order[i];
    x = op[name]
    a0 = a1 = a2 = a3 = a4 = a5 = a6 = a7 = 0
    # a8 = a9 = a10 = a11 = a12 = a13 = a14 = a15 = 0
    if( jump[name] ) a0 = 1;
    if( out2_prerelease[name] ) a1 = 2;
    if( in1[name] ) a2 = 4;
    if( in2[name] ) a3 = 8;
    if( in3[name] ) a4 = 16;

    if( out3[name] ) a5 = 32;
    # bv[x] = a0+a1+a2+a3+a4+a5+a6+a7+a8+a9+a10+a11+a12+a13+a14+a15;
    bv[x] = a0+a1+a2+a3+a4+a5+a6+a7;
  }
  print "\n"
  print "/* Properties such as \"out2\" or \"jump\" that are specified in"
  print "** comments following the \"case\" for each opcode in the vdbe.c"
  print "** are encoded into bitvectors as follows:"
  print "*/"
  print "#define OPFLG_JUMP            0x0001  /* jump:  P2 holds jmp target */"
  print "#define OPFLG_OUT2_PRERELEASE 0x0002  /* out2-prerelease: */"
  print "#define OPFLG_IN1             0x0004  /* in1:   P1 is an input */"
  print "#define OPFLG_IN2             0x0008  /* in2:   P2 is an input */"
  print "#define OPFLG_IN3             0x0010  /* in3:   P3 is an input */"

  print "#define OPFLG_OUT3            0x0020  /* out3:  P3 is an output */"
  print "#define OPFLG_INITIALIZER {\\"
  for(i=0; i<=max; i++){
    if( i%8==0 ) printf("/* %3d */",i)
    printf " 0x%02x,", bv[i]
    if( i%8==7 ) printf("\\\n");
  }
  print "}"
}







>







 







>
>







 







|





>
|













>
|








45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
..
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
...
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
  sub("\r","",name)
  op[name] = -1
  jump[name] = 0
  out2_prerelease[name] = 0
  in1[name] = 0
  in2[name] = 0
  in3[name] = 0
  out2[name] = 0
  out3[name] = 0
  for(i=3; i<NF; i++){
    if($i=="same" && $(i+1)=="as"){
      sym = $(i+2)
      sub(/,/,"",sym)
      op[name] = tk[sym]
      used[op[name]] = 1
................................................................................
      out2_prerelease[name] = 1
    }else if(x=="in1"){
      in1[name] = 1
    }else if(x=="in2"){
      in2[name] = 1
    }else if(x=="in3"){
      in3[name] = 1
    }else if(x=="out2"){
      out2[name] = 1
    }else if(x=="out3"){
      out3[name] = 1
    }
  }
  order[n_op++] = name;
}

................................................................................
  #  bit 2:     output to p1.  release p1 before opcode runs
  #
  for(i=0; i<=max; i++) bv[i] = 0;
  for(i=0; i<n_op; i++){
    name = order[i];
    x = op[name]
    a0 = a1 = a2 = a3 = a4 = a5 = a6 = a7 = 0
    # a7 = a9 = a10 = a11 = a12 = a13 = a14 = a15 = 0
    if( jump[name] ) a0 = 1;
    if( out2_prerelease[name] ) a1 = 2;
    if( in1[name] ) a2 = 4;
    if( in2[name] ) a3 = 8;
    if( in3[name] ) a4 = 16;
    if( out2[name] ) a5 = 32;
    if( out3[name] ) a6 = 64;
    # bv[x] = a0+a1+a2+a3+a4+a5+a6+a7+a8+a9+a10+a11+a12+a13+a14+a15;
    bv[x] = a0+a1+a2+a3+a4+a5+a6+a7;
  }
  print "\n"
  print "/* Properties such as \"out2\" or \"jump\" that are specified in"
  print "** comments following the \"case\" for each opcode in the vdbe.c"
  print "** are encoded into bitvectors as follows:"
  print "*/"
  print "#define OPFLG_JUMP            0x0001  /* jump:  P2 holds jmp target */"
  print "#define OPFLG_OUT2_PRERELEASE 0x0002  /* out2-prerelease: */"
  print "#define OPFLG_IN1             0x0004  /* in1:   P1 is an input */"
  print "#define OPFLG_IN2             0x0008  /* in2:   P2 is an input */"
  print "#define OPFLG_IN3             0x0010  /* in3:   P3 is an input */"
  print "#define OPFLG_OUT2            0x0020  /* out2:  P2 is an output */"
  print "#define OPFLG_OUT3            0x0040  /* out3:  P3 is an output */"
  print "#define OPFLG_INITIALIZER {\\"
  for(i=0; i<=max; i++){
    if( i%8==0 ) printf("/* %3d */",i)
    printf " 0x%02x,", bv[i]
    if( i%8==7 ) printf("\\\n");
  }
  print "}"
}

Changes to src/alter.c.

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains C code routines that used to generate VDBE code
** that implements the ALTER TABLE command.
**
** $Id: alter.c,v 1.62 2009/07/24 17:58:53 danielk1977 Exp $
*/
#include "sqliteInt.h"

/*
** The code in this file only exists if we are not omitting the
** ALTER TABLE logic from the build.
*/







<
<







7
8
9
10
11
12
13


14
15
16
17
18
19
20
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains C code routines that used to generate VDBE code
** that implements the ALTER TABLE command.


*/
#include "sqliteInt.h"

/*
** The code in this file only exists if we are not omitting the
** ALTER TABLE logic from the build.
*/

Changes to src/analyze.c.

6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains code associated with the ANALYZE command.
**
** @(#) $Id: analyze.c,v 1.52 2009/04/16 17:45:48 drh Exp $
*/
#ifndef SQLITE_OMIT_ANALYZE
#include "sqliteInt.h"

/*
** This routine generates code that opens the sqlite_stat1 table for
** writing with cursor iStatCur. If the library was built with the







<
<







6
7
8
9
10
11
12


13
14
15
16
17
18
19
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains code associated with the ANALYZE command.


*/
#ifndef SQLITE_OMIT_ANALYZE
#include "sqliteInt.h"

/*
** This routine generates code that opens the sqlite_stat1 table for
** writing with cursor iStatCur. If the library was built with the

Changes to src/attach.c.

6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
...
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
...
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains code used to implement the ATTACH and DETACH commands.
**
** $Id: attach.c,v 1.93 2009/05/31 21:21:41 drh Exp $
*/
#include "sqliteInt.h"

#ifndef SQLITE_OMIT_ATTACH
/*
** Resolve an expression that was part of an ATTACH or DETACH statement. This
** is slightly different from resolving a normal SQL expression, because simple
................................................................................
    sqlite3PagerLockingMode(pPager, db->dfltLockMode);
    sqlite3PagerJournalMode(pPager, db->dfltJournalMode);
  }
  aNew->zName = sqlite3DbStrDup(db, zName);
  aNew->safety_level = 3;

#if SQLITE_HAS_CODEC
  {
    extern int sqlite3CodecAttach(sqlite3*, int, const void*, int);
    extern void sqlite3CodecGetKey(sqlite3*, int, void**, int*);
    int nKey;
    char *zKey;
    int t = sqlite3_value_type(argv[2]);
    switch( t ){
      case SQLITE_INTEGER:
................................................................................
        rc = SQLITE_ERROR;
        break;
        
      case SQLITE_TEXT:
      case SQLITE_BLOB:
        nKey = sqlite3_value_bytes(argv[2]);
        zKey = (char *)sqlite3_value_blob(argv[2]);
        sqlite3CodecAttach(db, db->nDb-1, zKey, nKey);
        break;

      case SQLITE_NULL:
        /* No key specified.  Use the key from the main database */
        sqlite3CodecGetKey(db, 0, (void**)&zKey, &nKey);
        sqlite3CodecAttach(db, db->nDb-1, zKey, nKey);
        break;
    }
  }
#endif

  /* If the file was opened successfully, read the schema for the new database.
  ** If this fails, or if opening the file failed, then close the file and 







<
<







 







|







 







|





|







6
7
8
9
10
11
12


13
14
15
16
17
18
19
...
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
...
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains code used to implement the ATTACH and DETACH commands.


*/
#include "sqliteInt.h"

#ifndef SQLITE_OMIT_ATTACH
/*
** Resolve an expression that was part of an ATTACH or DETACH statement. This
** is slightly different from resolving a normal SQL expression, because simple
................................................................................
    sqlite3PagerLockingMode(pPager, db->dfltLockMode);
    sqlite3PagerJournalMode(pPager, db->dfltJournalMode);
  }
  aNew->zName = sqlite3DbStrDup(db, zName);
  aNew->safety_level = 3;

#if SQLITE_HAS_CODEC
  if( rc==SQLITE_OK ){
    extern int sqlite3CodecAttach(sqlite3*, int, const void*, int);
    extern void sqlite3CodecGetKey(sqlite3*, int, void**, int*);
    int nKey;
    char *zKey;
    int t = sqlite3_value_type(argv[2]);
    switch( t ){
      case SQLITE_INTEGER:
................................................................................
        rc = SQLITE_ERROR;
        break;
        
      case SQLITE_TEXT:
      case SQLITE_BLOB:
        nKey = sqlite3_value_bytes(argv[2]);
        zKey = (char *)sqlite3_value_blob(argv[2]);
        rc = sqlite3CodecAttach(db, db->nDb-1, zKey, nKey);
        break;

      case SQLITE_NULL:
        /* No key specified.  Use the key from the main database */
        sqlite3CodecGetKey(db, 0, (void**)&zKey, &nKey);
        rc = sqlite3CodecAttach(db, db->nDb-1, zKey, nKey);
        break;
    }
  }
#endif

  /* If the file was opened successfully, read the schema for the new database.
  ** If this fails, or if opening the file failed, then close the file and 

Changes to src/auth.c.

9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains code used to implement the sqlite3_set_authorizer()
** API.  This facility is an optional feature of the library.  Embedded
** systems that do not need this facility may omit it by recompiling
** the library with -DSQLITE_OMIT_AUTHORIZATION=1
**
** $Id: auth.c,v 1.32 2009/07/02 18:40:35 danielk1977 Exp $
*/
#include "sqliteInt.h"

/*
** All of the code in this file may be omitted by defining a single
** macro.
*/







<
<







9
10
11
12
13
14
15


16
17
18
19
20
21
22
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains code used to implement the sqlite3_set_authorizer()
** API.  This facility is an optional feature of the library.  Embedded
** systems that do not need this facility may omit it by recompiling
** the library with -DSQLITE_OMIT_AUTHORIZATION=1


*/
#include "sqliteInt.h"

/*
** All of the code in this file may be omitted by defining a single
** macro.
*/

Changes to src/backup.c.

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains the implementation of the sqlite3_backup_XXX() 
** API functions and the related features.
**
** $Id: backup.c,v 1.19 2009/07/06 19:03:13 drh Exp $
*/
#include "sqliteInt.h"
#include "btreeInt.h"

/* Macro to find the minimum of two numeric values.
*/
#ifndef MIN







<
<







7
8
9
10
11
12
13


14
15
16
17
18
19
20
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains the implementation of the sqlite3_backup_XXX() 
** API functions and the related features.


*/
#include "sqliteInt.h"
#include "btreeInt.h"

/* Macro to find the minimum of two numeric values.
*/
#ifndef MIN

Changes to src/bitvec.c.

29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
** Test operations are about 100 times more common that set operations.
** Clear operations are exceedingly rare.  There are usually between
** 5 and 500 set operations per Bitvec object, though the number of sets can
** sometimes grow into tens of thousands or larger.  The size of the
** Bitvec object is the number of pages in the database file at the
** start of a transaction, and is thus usually less than a few thousand,
** but can be as large as 2 billion for a really big database.
**
** @(#) $Id: bitvec.c,v 1.17 2009/07/25 17:33:26 drh Exp $
*/
#include "sqliteInt.h"

/* Size of the Bitvec structure in bytes. */
#define BITVEC_SZ        (sizeof(void*)*128)  /* 512 on 32bit.  1024 on 64bit */

/* Round the union size down to the nearest pointer boundary, since that's how 







<
<







29
30
31
32
33
34
35


36
37
38
39
40
41
42
** Test operations are about 100 times more common that set operations.
** Clear operations are exceedingly rare.  There are usually between
** 5 and 500 set operations per Bitvec object, though the number of sets can
** sometimes grow into tens of thousands or larger.  The size of the
** Bitvec object is the number of pages in the database file at the
** start of a transaction, and is thus usually less than a few thousand,
** but can be as large as 2 billion for a really big database.


*/
#include "sqliteInt.h"

/* Size of the Bitvec structure in bytes. */
#define BITVEC_SZ        (sizeof(void*)*128)  /* 512 on 32bit.  1024 on 64bit */

/* Round the union size down to the nearest pointer boundary, since that's how 

Changes to src/btmutex.c.

6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
**
** $Id: btmutex.c,v 1.17 2009/07/20 12:33:33 drh Exp $
**
** This file contains code used to implement mutexes on Btree objects.
** This code really belongs in btree.c.  But btree.c is getting too
** big and we want to break it down some.  This packaged seemed like
** a good breakout.
*/
#include "btreeInt.h"
#ifndef SQLITE_OMIT_SHARED_CACHE







<
<







6
7
8
9
10
11
12


13
14
15
16
17
18
19
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
**


** This file contains code used to implement mutexes on Btree objects.
** This code really belongs in btree.c.  But btree.c is getting too
** big and we want to break it down some.  This packaged seemed like
** a good breakout.
*/
#include "btreeInt.h"
#ifndef SQLITE_OMIT_SHARED_CACHE

Changes to src/btree.c.

5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
....
1141
1142
1143
1144
1145
1146
1147

1148
1149
1150
1151
1152
1153
1154
1155

1156
1157
1158
1159
1160
1161
1162
....
1171
1172
1173
1174
1175
1176
1177




1178
1179
1180
1181
1182
1183
1184
1185
1186
1187


1188
1189
1190
1191
1192
1193
1194
....
1473
1474
1475
1476
1477
1478
1479

1480

1481
1482
1483
1484
1485
1486
1487
....
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
....
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
....
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
....
3350
3351
3352
3353
3354
3355
3356
3357












3358
3359
3360
3361
3362
3363
3364
....
5282
5283
5284
5285
5286
5287
5288
5289
5290






5291
5292
5293
5294
5295
5296
5297
....
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584


5585
5586
5587



5588
5589
5590
5591
5592
5593
5594
....
6841
6842
6843
6844
6845
6846
6847
6848
6849
6850
6851
6852
6853
6854
6855
6856
6857
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** $Id: btree.c,v 1.705 2009/08/10 03:57:58 shane Exp $
**
** This file implements a external (disk-based) database using BTrees.
** See the header comment on "btreeInt.h" for additional information.
** Including a description of file format and an overview of operation.
*/
#include "btreeInt.h"

/*
................................................................................
static int allocateSpace(MemPage *pPage, int nByte, int *pIdx){
  const int hdr = pPage->hdrOffset;    /* Local cache of pPage->hdrOffset */
  u8 * const data = pPage->aData;      /* Local cache of pPage->aData */
  int nFrag;                           /* Number of fragmented bytes on pPage */
  int top;                             /* First byte of cell content area */
  int gap;        /* First byte of gap between cell pointers and cell content */
  int rc;         /* Integer return code */

  
  assert( sqlite3PagerIswriteable(pPage->pDbPage) );
  assert( pPage->pBt );
  assert( sqlite3_mutex_held(pPage->pBt->mutex) );
  assert( nByte>=0 );  /* Minimum cell size is 4 */
  assert( pPage->nFree>=nByte );
  assert( pPage->nOverflow==0 );
  assert( nByte<pPage->pBt->usableSize-8 );


  nFrag = data[hdr+7];
  assert( pPage->cellOffset == hdr + 12 - 4*pPage->leaf );
  gap = pPage->cellOffset + 2*pPage->nCell;
  top = get2byte(&data[hdr+5]);
  if( gap>top ) return SQLITE_CORRUPT_BKPT;
  testcase( gap+2==top );
................................................................................
  }else if( gap+2<=top ){
    /* Search the freelist looking for a free slot big enough to satisfy 
    ** the request. The allocation is made from the first free slot in 
    ** the list that is large enough to accomadate it.
    */
    int pc, addr;
    for(addr=hdr+1; (pc = get2byte(&data[addr]))>0; addr=pc){




      int size = get2byte(&data[pc+2]);     /* Size of free slot */
      if( size>=nByte ){
        int x = size - nByte;
        testcase( x==4 );
        testcase( x==3 );
        if( x<4 ){
          /* Remove the slot from the free-list. Update the number of
          ** fragmented bytes within the page. */
          memcpy(&data[addr], &data[pc], 2);
          data[hdr+7] = (u8)(nFrag + x);


        }else{
          /* The slot remains on the free-list. Reduce its size to account
          ** for the portion used by the new allocation. */
          put2byte(&data[pc+2], x);
        }
        *pIdx = pc + x;
        return SQLITE_OK;
................................................................................
  u16 first;

  assert( sqlite3PagerPagenumber(pPage->pDbPage)==pPage->pgno );
  assert( sqlite3PagerGetExtra(pPage->pDbPage) == (void*)pPage );
  assert( sqlite3PagerGetData(pPage->pDbPage) == data );
  assert( sqlite3PagerIswriteable(pPage->pDbPage) );
  assert( sqlite3_mutex_held(pBt->mutex) );

  /*memset(&data[hdr], 0, pBt->usableSize - hdr);*/

  data[hdr] = (char)flags;
  first = hdr + 8 + 4*((flags&PTF_LEAF)==0 ?1:0);
  memset(&data[hdr+1], 0, 4);
  data[hdr+7] = 0;
  put2byte(&data[hdr+5], pBt->usableSize);
  pPage->nFree = pBt->usableSize - first;
  decodeFlags(pPage, flags);
................................................................................

/*
** Release a MemPage.  This should be called once for each prior
** call to btreeGetPage.
*/
static void releasePage(MemPage *pPage){
  if( pPage ){
    assert( pPage->nOverflow==0 || sqlite3PagerPageRefcount(pPage->pDbPage)>1 );
    assert( pPage->aData );
    assert( pPage->pBt );
    assert( sqlite3PagerGetExtra(pPage->pDbPage) == (void*)pPage );
    assert( sqlite3PagerGetData(pPage->pDbPage)==pPage->aData );
    assert( sqlite3_mutex_held(pPage->pBt->mutex) );
    sqlite3PagerUnref(pPage->pDbPage);
  }
................................................................................
static int newDatabase(BtShared *pBt){
  MemPage *pP1;
  unsigned char *data;
  int rc;
  int nPage;

  assert( sqlite3_mutex_held(pBt->mutex) );
  /* The database size has already been measured and cached, so failure
  ** is impossible here.  If the original size measurement failed, then
  ** processing aborts before entering this routine. */
  rc = sqlite3PagerPagecount(pBt->pPager, &nPage);
  if( NEVER(rc!=SQLITE_OK) || nPage>0 ){
    return rc;
  }
  pP1 = pBt->pPage1;
  assert( pP1!=0 );
  data = pP1->aData;
  rc = sqlite3PagerWrite(pP1->pDbPage);
  if( rc ) return rc;
................................................................................
**
** 4:  There must be an active transaction.
**
** No checking is done to make sure that page iTable really is the
** root page of a b-tree.  If it is not, then the cursor acquired
** will not work correctly.
**
** It is assumed that the sqlite3BtreeCursorSize() bytes of memory 
** pointed to by pCur have been zeroed by the caller.
*/
static int btreeCursor(
  Btree *p,                              /* The btree */
  int iTable,                            /* Root page of table to open */
  int wrFlag,                            /* 1 to write. 0 read-only */
  struct KeyInfo *pKeyInfo,              /* First arg to comparison function */
  BtCursor *pCur                         /* Space for new cursor */
................................................................................
**
** This interfaces is needed so that users of cursors can preallocate
** sufficient storage to hold a cursor.  The BtCursor object is opaque
** to users so they cannot do the sizeof() themselves - they must call
** this routine.
*/
int sqlite3BtreeCursorSize(void){
  return sizeof(BtCursor);












}

/*
** Set the cached rowid value of every cursor in the same database file
** as pCur and having the same root page number as pCur.  The value is
** set to iRowid.
**
................................................................................
  int nSkip = (iChild ? 4 : 0);

  if( *pRC ) return;

  assert( i>=0 && i<=pPage->nCell+pPage->nOverflow );
  assert( pPage->nCell<=MX_CELL(pPage->pBt) && MX_CELL(pPage->pBt)<=5460 );
  assert( pPage->nOverflow<=ArraySize(pPage->aOvfl) );
  assert( sz==cellSizePtr(pPage, pCell) );
  assert( sqlite3_mutex_held(pPage->pBt->mutex) );






  if( pPage->nOverflow || sz+2>pPage->nFree ){
    if( pTemp ){
      memcpy(pTemp+nSkip, pCell+nSkip, sz-nSkip);
      pCell = pTemp;
    }
    if( iChild ){
      put4byte(pCell, iChild);
................................................................................
static void copyNodeContent(MemPage *pFrom, MemPage *pTo, int *pRC){
  if( (*pRC)==SQLITE_OK ){
    BtShared * const pBt = pFrom->pBt;
    u8 * const aFrom = pFrom->aData;
    u8 * const aTo = pTo->aData;
    int const iFromHdr = pFrom->hdrOffset;
    int const iToHdr = ((pTo->pgno==1) ? 100 : 0);
    TESTONLY(int rc;)
    int iData;
  
  
    assert( pFrom->isInit );
    assert( pFrom->nFree>=iToHdr );
    assert( get2byte(&aFrom[iFromHdr+5])<=pBt->usableSize );
  
    /* Copy the b-tree node content from page pFrom to page pTo. */
    iData = get2byte(&aFrom[iFromHdr+5]);
    memcpy(&aTo[iData], &aFrom[iData], pBt->usableSize-iData);
    memcpy(&aTo[iToHdr], &aFrom[iFromHdr], pFrom->cellOffset + 2*pFrom->nCell);
  
    /* Reinitialize page pTo so that the contents of the MemPage structure
    ** match the new data. The initialization of pTo "cannot" fail, as the
    ** data copied from pFrom is known to be valid.  */


    pTo->isInit = 0;
    TESTONLY(rc = ) btreeInitPage(pTo);
    assert( rc==SQLITE_OK );



  
    /* If this is an auto-vacuum database, update the pointer-map entries
    ** for any b-tree or overflow pages that pTo now contains the pointers to.
    */
    if( ISAUTOVACUUM ){
      *pRC = setChildPtrmaps(pTo);
    }
................................................................................

/*
** Erase the given database page and all its children.  Return
** the page to the freelist.
*/
static int clearDatabasePage(
  BtShared *pBt,           /* The BTree that contains the table */
  Pgno pgno,            /* Page number to clear */
  int freePageFlag,     /* Deallocate page if true */
  int *pnChange
){
  MemPage *pPage;
  int rc;
  unsigned char *pCell;
  int i;

  assert( sqlite3_mutex_held(pBt->mutex) );







<
<







 







>







|
>







 







>
>
>
>
|









>
>







 







>
|
>







 







<







 







<
<
<

|







 







|
|







 







|
>
>
>
>
>
>
>
>
>
>
>
>







 







<

>
>
>
>
>
>







 







|













|
|
>
>

|
|
>
>
>







 







|
|
|







5
6
7
8
9
10
11


12
13
14
15
16
17
18
....
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
....
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
....
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
....
1610
1611
1612
1613
1614
1615
1616

1617
1618
1619
1620
1621
1622
1623
....
2349
2350
2351
2352
2353
2354
2355



2356
2357
2358
2359
2360
2361
2362
2363
2364
....
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
....
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
....
5298
5299
5300
5301
5302
5303
5304

5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
....
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
....
6867
6868
6869
6870
6871
6872
6873
6874
6875
6876
6877
6878
6879
6880
6881
6882
6883
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************


** This file implements a external (disk-based) database using BTrees.
** See the header comment on "btreeInt.h" for additional information.
** Including a description of file format and an overview of operation.
*/
#include "btreeInt.h"

/*
................................................................................
static int allocateSpace(MemPage *pPage, int nByte, int *pIdx){
  const int hdr = pPage->hdrOffset;    /* Local cache of pPage->hdrOffset */
  u8 * const data = pPage->aData;      /* Local cache of pPage->aData */
  int nFrag;                           /* Number of fragmented bytes on pPage */
  int top;                             /* First byte of cell content area */
  int gap;        /* First byte of gap between cell pointers and cell content */
  int rc;         /* Integer return code */
  int usableSize; /* Usable size of the page */
  
  assert( sqlite3PagerIswriteable(pPage->pDbPage) );
  assert( pPage->pBt );
  assert( sqlite3_mutex_held(pPage->pBt->mutex) );
  assert( nByte>=0 );  /* Minimum cell size is 4 */
  assert( pPage->nFree>=nByte );
  assert( pPage->nOverflow==0 );
  usableSize = pPage->pBt->usableSize;
  assert( nByte < usableSize-8 );

  nFrag = data[hdr+7];
  assert( pPage->cellOffset == hdr + 12 - 4*pPage->leaf );
  gap = pPage->cellOffset + 2*pPage->nCell;
  top = get2byte(&data[hdr+5]);
  if( gap>top ) return SQLITE_CORRUPT_BKPT;
  testcase( gap+2==top );
................................................................................
  }else if( gap+2<=top ){
    /* Search the freelist looking for a free slot big enough to satisfy 
    ** the request. The allocation is made from the first free slot in 
    ** the list that is large enough to accomadate it.
    */
    int pc, addr;
    for(addr=hdr+1; (pc = get2byte(&data[addr]))>0; addr=pc){
      int size;            /* Size of the free slot */
      if( pc>usableSize-4 || pc<addr+4 ){
        return SQLITE_CORRUPT_BKPT;
      }
      size = get2byte(&data[pc+2]);
      if( size>=nByte ){
        int x = size - nByte;
        testcase( x==4 );
        testcase( x==3 );
        if( x<4 ){
          /* Remove the slot from the free-list. Update the number of
          ** fragmented bytes within the page. */
          memcpy(&data[addr], &data[pc], 2);
          data[hdr+7] = (u8)(nFrag + x);
        }else if( size+pc > usableSize ){
          return SQLITE_CORRUPT_BKPT;
        }else{
          /* The slot remains on the free-list. Reduce its size to account
          ** for the portion used by the new allocation. */
          put2byte(&data[pc+2], x);
        }
        *pIdx = pc + x;
        return SQLITE_OK;
................................................................................
  u16 first;

  assert( sqlite3PagerPagenumber(pPage->pDbPage)==pPage->pgno );
  assert( sqlite3PagerGetExtra(pPage->pDbPage) == (void*)pPage );
  assert( sqlite3PagerGetData(pPage->pDbPage) == data );
  assert( sqlite3PagerIswriteable(pPage->pDbPage) );
  assert( sqlite3_mutex_held(pBt->mutex) );
#ifdef SQLITE_SECURE_DELETE
  memset(&data[hdr], 0, pBt->usableSize - hdr);
#endif
  data[hdr] = (char)flags;
  first = hdr + 8 + 4*((flags&PTF_LEAF)==0 ?1:0);
  memset(&data[hdr+1], 0, 4);
  data[hdr+7] = 0;
  put2byte(&data[hdr+5], pBt->usableSize);
  pPage->nFree = pBt->usableSize - first;
  decodeFlags(pPage, flags);
................................................................................

/*
** Release a MemPage.  This should be called once for each prior
** call to btreeGetPage.
*/
static void releasePage(MemPage *pPage){
  if( pPage ){

    assert( pPage->aData );
    assert( pPage->pBt );
    assert( sqlite3PagerGetExtra(pPage->pDbPage) == (void*)pPage );
    assert( sqlite3PagerGetData(pPage->pDbPage)==pPage->aData );
    assert( sqlite3_mutex_held(pPage->pBt->mutex) );
    sqlite3PagerUnref(pPage->pDbPage);
  }
................................................................................
static int newDatabase(BtShared *pBt){
  MemPage *pP1;
  unsigned char *data;
  int rc;
  int nPage;

  assert( sqlite3_mutex_held(pBt->mutex) );



  rc = sqlite3PagerPagecount(pBt->pPager, &nPage);
  if( rc!=SQLITE_OK || nPage>0 ){
    return rc;
  }
  pP1 = pBt->pPage1;
  assert( pP1!=0 );
  data = pP1->aData;
  rc = sqlite3PagerWrite(pP1->pDbPage);
  if( rc ) return rc;
................................................................................
**
** 4:  There must be an active transaction.
**
** No checking is done to make sure that page iTable really is the
** root page of a b-tree.  If it is not, then the cursor acquired
** will not work correctly.
**
** It is assumed that the sqlite3BtreeCursorZero() has been called
** on pCur to initialize the memory space prior to invoking this routine.
*/
static int btreeCursor(
  Btree *p,                              /* The btree */
  int iTable,                            /* Root page of table to open */
  int wrFlag,                            /* 1 to write. 0 read-only */
  struct KeyInfo *pKeyInfo,              /* First arg to comparison function */
  BtCursor *pCur                         /* Space for new cursor */
................................................................................
**
** This interfaces is needed so that users of cursors can preallocate
** sufficient storage to hold a cursor.  The BtCursor object is opaque
** to users so they cannot do the sizeof() themselves - they must call
** this routine.
*/
int sqlite3BtreeCursorSize(void){
  return ROUND8(sizeof(BtCursor));
}

/*
** Initialize memory that will be converted into a BtCursor object.
**
** The simple approach here would be to memset() the entire object
** to zero.  But it turns out that the apPage[] and aiIdx[] arrays
** do not need to be zeroed and they are large, so we can save a lot
** of run-time by skipping the initialization of those elements.
*/
void sqlite3BtreeCursorZero(BtCursor *p){
  memset(p, 0, offsetof(BtCursor, iPage));
}

/*
** Set the cached rowid value of every cursor in the same database file
** as pCur and having the same root page number as pCur.  The value is
** set to iRowid.
**
................................................................................
  int nSkip = (iChild ? 4 : 0);

  if( *pRC ) return;

  assert( i>=0 && i<=pPage->nCell+pPage->nOverflow );
  assert( pPage->nCell<=MX_CELL(pPage->pBt) && MX_CELL(pPage->pBt)<=5460 );
  assert( pPage->nOverflow<=ArraySize(pPage->aOvfl) );

  assert( sqlite3_mutex_held(pPage->pBt->mutex) );
  /* The cell should normally be sized correctly.  However, when moving a
  ** malformed cell from a leaf page to an interior page, if the cell size
  ** wanted to be less than 4 but got rounded up to 4 on the leaf, then size
  ** might be less than 8 (leaf-size + pointer) on the interior node.  Hence
  ** the term after the || in the following assert(). */
  assert( sz==cellSizePtr(pPage, pCell) || (sz==8 && iChild>0) );
  if( pPage->nOverflow || sz+2>pPage->nFree ){
    if( pTemp ){
      memcpy(pTemp+nSkip, pCell+nSkip, sz-nSkip);
      pCell = pTemp;
    }
    if( iChild ){
      put4byte(pCell, iChild);
................................................................................
static void copyNodeContent(MemPage *pFrom, MemPage *pTo, int *pRC){
  if( (*pRC)==SQLITE_OK ){
    BtShared * const pBt = pFrom->pBt;
    u8 * const aFrom = pFrom->aData;
    u8 * const aTo = pTo->aData;
    int const iFromHdr = pFrom->hdrOffset;
    int const iToHdr = ((pTo->pgno==1) ? 100 : 0);
    int rc;
    int iData;
  
  
    assert( pFrom->isInit );
    assert( pFrom->nFree>=iToHdr );
    assert( get2byte(&aFrom[iFromHdr+5])<=pBt->usableSize );
  
    /* Copy the b-tree node content from page pFrom to page pTo. */
    iData = get2byte(&aFrom[iFromHdr+5]);
    memcpy(&aTo[iData], &aFrom[iData], pBt->usableSize-iData);
    memcpy(&aTo[iToHdr], &aFrom[iFromHdr], pFrom->cellOffset + 2*pFrom->nCell);
  
    /* Reinitialize page pTo so that the contents of the MemPage structure
    ** match the new data. The initialization of pTo can actually fail under
    ** fairly obscure circumstances, even though it is a copy of initialized 
    ** page pFrom.
    */
    pTo->isInit = 0;
    rc = btreeInitPage(pTo);
    if( rc!=SQLITE_OK ){
      *pRC = rc;
      return;
    }
  
    /* If this is an auto-vacuum database, update the pointer-map entries
    ** for any b-tree or overflow pages that pTo now contains the pointers to.
    */
    if( ISAUTOVACUUM ){
      *pRC = setChildPtrmaps(pTo);
    }
................................................................................

/*
** Erase the given database page and all its children.  Return
** the page to the freelist.
*/
static int clearDatabasePage(
  BtShared *pBt,           /* The BTree that contains the table */
  Pgno pgno,               /* Page number to clear */
  int freePageFlag,        /* Deallocate page if true */
  int *pnChange            /* Add number of Cells freed to this counter */
){
  MemPage *pPage;
  int rc;
  unsigned char *pCell;
  int i;

  assert( sqlite3_mutex_held(pBt->mutex) );

Changes to src/btree.h.

8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
...
146
147
148
149
150
151
152

153
154
155
156
157
158
159
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This header file defines the interface that the sqlite B-Tree file
** subsystem.  See comments in the source code for a detailed description
** of what each interface routine does.
**
** @(#) $Id: btree.h,v 1.120 2009/07/22 00:35:24 drh Exp $
*/
#ifndef _BTREE_H_
#define _BTREE_H_

/* TODO: This definition is just included so other modules compile. It
** needs to be revisited.
*/
................................................................................
  Btree*,                              /* BTree containing table to open */
  int iTable,                          /* Index of root page */
  int wrFlag,                          /* 1 for writing.  0 for read-only */
  struct KeyInfo*,                     /* First argument to compare function */
  BtCursor *pCursor                    /* Space to write cursor structure */
);
int sqlite3BtreeCursorSize(void);


int sqlite3BtreeCloseCursor(BtCursor*);
int sqlite3BtreeMovetoUnpacked(
  BtCursor*,
  UnpackedRecord *pUnKey,
  i64 intKey,
  int bias,







<
<







 







>







8
9
10
11
12
13
14


15
16
17
18
19
20
21
...
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This header file defines the interface that the sqlite B-Tree file
** subsystem.  See comments in the source code for a detailed description
** of what each interface routine does.


*/
#ifndef _BTREE_H_
#define _BTREE_H_

/* TODO: This definition is just included so other modules compile. It
** needs to be revisited.
*/
................................................................................
  Btree*,                              /* BTree containing table to open */
  int iTable,                          /* Index of root page */
  int wrFlag,                          /* 1 for writing.  0 for read-only */
  struct KeyInfo*,                     /* First argument to compare function */
  BtCursor *pCursor                    /* Space to write cursor structure */
);
int sqlite3BtreeCursorSize(void);
void sqlite3BtreeCursorZero(BtCursor*);

int sqlite3BtreeCloseCursor(BtCursor*);
int sqlite3BtreeMovetoUnpacked(
  BtCursor*,
  UnpackedRecord *pUnKey,
  i64 intKey,
  int bias,

Changes to src/btreeInt.h.

5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** $Id: btreeInt.h,v 1.52 2009/07/15 17:25:46 drh Exp $
**
** This file implements a external (disk-based) database using BTrees.
** For a detailed discussion of BTrees, refer to
**
**     Donald E. Knuth, THE ART OF COMPUTER PROGRAMMING, Volume 3:
**     "Sorting And Searching", pages 473-480. Addison-Wesley
**     Publishing Company, Reading, Massachusetts.
**







<
<







5
6
7
8
9
10
11


12
13
14
15
16
17
18
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************


** This file implements a external (disk-based) database using BTrees.
** For a detailed discussion of BTrees, refer to
**
**     Donald E. Knuth, THE ART OF COMPUTER PROGRAMMING, Volume 3:
**     "Sorting And Searching", pages 473-480. Addison-Wesley
**     Publishing Company, Reading, Massachusetts.
**

Changes to src/build.c.

17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
**     DROP TABLE
**     CREATE INDEX
**     DROP INDEX
**     creating ID lists
**     BEGIN TRANSACTION
**     COMMIT
**     ROLLBACK
**
** $Id: build.c,v 1.557 2009/07/24 17:58:53 danielk1977 Exp $
*/
#include "sqliteInt.h"

/*
** This routine is called when a new SQL statement is beginning to
** be parsed.  Initialize the pParse structure as needed.
*/







<
<







17
18
19
20
21
22
23


24
25
26
27
28
29
30
**     DROP TABLE
**     CREATE INDEX
**     DROP INDEX
**     creating ID lists
**     BEGIN TRANSACTION
**     COMMIT
**     ROLLBACK


*/
#include "sqliteInt.h"

/*
** This routine is called when a new SQL statement is beginning to
** be parsed.  Initialize the pParse structure as needed.
*/

Changes to src/callback.c.

8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
**
** This file contains functions used to access the internal hash tables
** of user defined functions and collation sequences.
**
** $Id: callback.c,v 1.42 2009/06/17 00:35:31 drh Exp $
*/

#include "sqliteInt.h"

/*
** Invoke the 'collation needed' callback to request a collation sequence
** in the encoding enc of name zName, length nName.







<
<







8
9
10
11
12
13
14


15
16
17
18
19
20
21
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
**
** This file contains functions used to access the internal hash tables
** of user defined functions and collation sequences.


*/

#include "sqliteInt.h"

/*
** Invoke the 'collation needed' callback to request a collation sequence
** in the encoding enc of name zName, length nName.

Changes to src/complete.c.

11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
...
180
181
182
183
184
185
186
187


188
189
190
191
192
193
194
*************************************************************************
** An tokenizer for SQL
**
** This file contains C code that implements the sqlite3_complete() API.
** This code used to be part of the tokenizer.c source file.  But by
** separating it out, the code will be automatically omitted from
** static links that do not use it.
**
** $Id: complete.c,v 1.8 2009/04/28 04:46:42 drh Exp $
*/
#include "sqliteInt.h"
#ifndef SQLITE_OMIT_COMPLETE

/*
** This is defined in tokenize.c.  We just have to import the definition.
*/
#ifndef SQLITE_AMALGAMATION
#ifdef SQLITE_ASCII
extern const char sqlite3IsAsciiIdChar[];
#define IdChar(C)  (((c=C)&0x80)!=0 || (c>0x1f && sqlite3IsAsciiIdChar[c-0x20]))
#endif
#ifdef SQLITE_EBCDIC
extern const char sqlite3IsEbcdicIdChar[];
#define IdChar(C)  (((c=C)>=0x42 && sqlite3IsEbcdicIdChar[c-0x40]))
#endif
#endif /* SQLITE_AMALGAMATION */

................................................................................
        zSql++;
        while( *zSql && *zSql!=c ){ zSql++; }
        if( *zSql==0 ) return 0;
        token = tkOTHER;
        break;
      }
      default: {
        int c;


        if( IdChar((u8)*zSql) ){
          /* Keywords and unquoted identifiers */
          int nId;
          for(nId=1; IdChar(zSql[nId]); nId++){}
#ifdef SQLITE_OMIT_TRIGGER
          token = tkOTHER;
#else







<
<









|
<







 







|
>
>







11
12
13
14
15
16
17


18
19
20
21
22
23
24
25
26
27

28
29
30
31
32
33
34
...
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
*************************************************************************
** An tokenizer for SQL
**
** This file contains C code that implements the sqlite3_complete() API.
** This code used to be part of the tokenizer.c source file.  But by
** separating it out, the code will be automatically omitted from
** static links that do not use it.


*/
#include "sqliteInt.h"
#ifndef SQLITE_OMIT_COMPLETE

/*
** This is defined in tokenize.c.  We just have to import the definition.
*/
#ifndef SQLITE_AMALGAMATION
#ifdef SQLITE_ASCII
#define IdChar(C)  ((sqlite3CtypeMap[(unsigned char)C]&0x46)!=0)

#endif
#ifdef SQLITE_EBCDIC
extern const char sqlite3IsEbcdicIdChar[];
#define IdChar(C)  (((c=C)>=0x42 && sqlite3IsEbcdicIdChar[c-0x40]))
#endif
#endif /* SQLITE_AMALGAMATION */

................................................................................
        zSql++;
        while( *zSql && *zSql!=c ){ zSql++; }
        if( *zSql==0 ) return 0;
        token = tkOTHER;
        break;
      }
      default: {
#ifdef SQLITE_EBCDIC
        unsigned char c;
#endif
        if( IdChar((u8)*zSql) ){
          /* Keywords and unquoted identifiers */
          int nId;
          for(nId=1; IdChar(zSql[nId]); nId++){}
#ifdef SQLITE_OMIT_TRIGGER
          token = tkOTHER;
#else

Changes to src/date.c.

12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
** This file contains the C functions that implement date and time
** functions for SQLite.  
**
** There is only one exported symbol in this file - the function
** sqlite3RegisterDateTimeFunctions() found at the bottom of the file.
** All other code has file scope.
**
** $Id: date.c,v 1.107 2009/05/03 20:23:53 drh Exp $
**
** SQLite processes all times and dates as Julian Day numbers.  The
** dates and times are stored as the number of days since noon
** in Greenwich on November 24, 4714 B.C. according to the Gregorian
** calendar system. 
**
** 1970-01-01 00:00:00 is JD 2440587.5
** 2000-01-01 00:00:00 is JD 2451544.5







<
<







12
13
14
15
16
17
18


19
20
21
22
23
24
25
** This file contains the C functions that implement date and time
** functions for SQLite.  
**
** There is only one exported symbol in this file - the function
** sqlite3RegisterDateTimeFunctions() found at the bottom of the file.
** All other code has file scope.
**


** SQLite processes all times and dates as Julian Day numbers.  The
** dates and times are stored as the number of days since noon
** in Greenwich on November 24, 4714 B.C. according to the Gregorian
** calendar system. 
**
** 1970-01-01 00:00:00 is JD 2440587.5
** 2000-01-01 00:00:00 is JD 2451544.5

Changes to src/delete.c.

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
...
494
495
496
497
498
499
500
501


502
503
504
505
506
507
508
...
631
632
633
634
635
636
637
638
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains C code routines that are called by the parser
** in order to generate code for DELETE FROM statements.
**
** $Id: delete.c,v 1.207 2009/08/08 18:01:08 drh Exp $
*/
#include "sqliteInt.h"

/*
** Look up every table that is named in pSrc.  If any table is not found,
** add an error message to pParse->zErrMsg and return NULL.  If all tables
** are found, return a pointer to the last table.
................................................................................
  ** use for the old.* references in the triggers.  */
  if( sqlite3FkRequired(pParse, pTab, 0, 0) || pTrigger ){
    u32 mask;                     /* Mask of OLD.* columns in use */
    int iCol;                     /* Iterator used while populating OLD.* */

    /* TODO: Could use temporary registers here. Also could attempt to
    ** avoid copying the contents of the rowid register.  */
    mask = sqlite3TriggerOldmask(pParse, pTrigger, 0, pTab, onconf);


    mask |= sqlite3FkOldmask(pParse, pTab);
    iOld = pParse->nMem+1;
    pParse->nMem += (1 + pTab->nCol);

    /* Populate the OLD.* pseudo-table register array. These values will be 
    ** used by any BEFORE and AFTER triggers that exist.  */
    sqlite3VdbeAddOp2(v, OP_Copy, iRowid, iOld);
................................................................................
    sqlite3VdbeAddOp3(v, OP_MakeRecord, regBase, nCol+1, regOut);
    sqlite3VdbeChangeP4(v, -1, sqlite3IndexAffinityStr(v, pIdx), 0);
    sqlite3ExprCacheAffinityChange(pParse, regBase, nCol+1);
  }
  sqlite3ReleaseTempRange(pParse, regBase, nCol+1);
  return regBase;
}








<
<







 







|
>
>







 







<
7
8
9
10
11
12
13


14
15
16
17
18
19
20
...
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
...
631
632
633
634
635
636
637

**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** This file contains C code routines that are called by the parser
** in order to generate code for DELETE FROM statements.


*/
#include "sqliteInt.h"

/*
** Look up every table that is named in pSrc.  If any table is not found,
** add an error message to pParse->zErrMsg and return NULL.  If all tables
** are found, return a pointer to the last table.
................................................................................
  ** use for the old.* references in the triggers.  */
  if( sqlite3FkRequired(pParse, pTab, 0, 0) || pTrigger ){
    u32 mask;                     /* Mask of OLD.* columns in use */
    int iCol;                     /* Iterator used while populating OLD.* */

    /* TODO: Could use temporary registers here. Also could attempt to
    ** avoid copying the contents of the rowid register.  */
    mask = sqlite3TriggerColmask(
        pParse, pTrigger, 0, 0, TRIGGER_BEFORE|TRIGGER_AFTER, pTab, onconf
    );
    mask |= sqlite3FkOldmask(pParse, pTab);
    iOld = pParse->nMem+1;
    pParse->nMem += (1 + pTab->nCol);

    /* Populate the OLD.* pseudo-table register array. These values will be 
    ** used by any BEFORE and AFTER triggers that exist.  */
    sqlite3VdbeAddOp2(v, OP_Copy, iRowid, iOld);
................................................................................
    sqlite3VdbeAddOp3(v, OP_MakeRecord, regBase, nCol+1, regOut);
    sqlite3VdbeChangeP4(v, -1, sqlite3IndexAffinityStr(v, pIdx), 0);
    sqlite3ExprCacheAffinityChange(pParse, regBase, nCol+1);
  }
  sqlite3ReleaseTempRange(pParse, regBase, nCol+1);
  return regBase;
}

Changes to src/expr.c.

567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
...
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
...
635
636
637
638
639
640
641
642
643
644
645
646

647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
....
1265
1266
1267
1268
1269
1270
1271
























































































1272
1273
1274
1275
1276
1277
1278
....
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
....
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
....
1702
1703
1704
1705
1706
1707
1708


























































































































1709
1710
1711
1712
1713
1714
1715
....
2402
2403
2404
2405
2406
2407
2408





















2409
2410
2411
2412
2413
2414
2415
....
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506

2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521

2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539


2540
2541
2542
2543
2544
2545

2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
....
2977
2978
2979
2980
2981
2982
2983
























































2984
2985
2986
2987
2988
2989
2990
....
3077
3078
3079
3080
3081
3082
3083
3084
3085

3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113






3114
3115
3116
3117
3118
3119
3120
....
3190
3191
3192
3193
3194
3195
3196

3197
3198
3199
3200
3201
3202
3203
....
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
  assert( !ExprHasAnyProperty(pExpr, EP_IntValue|EP_Reduced|EP_TokenOnly) );
  z = pExpr->u.zToken;
  assert( z!=0 );
  assert( z[0]!=0 );
  if( z[1]==0 ){
    /* Wildcard of the form "?".  Assign the next variable number */
    assert( z[0]=='?' );
#if SQLITE_MAX_VARIABLE_NUMBER<=32767
    pExpr->iColumn = (i16)(++pParse->nVar);
#else
    pExpr->iColumn = ++pParse->nVar;
#endif
  }else if( z[0]=='?' ){
    /* Wildcard of the form "?nnn".  Convert "nnn" to an integer and
    ** use it as the variable number */
    int i = atoi((char*)&z[1]);
#if SQLITE_MAX_VARIABLE_NUMBER<=32767
    pExpr->iColumn = (i16)i;
#else
    pExpr->iColumn = i;
#endif
    testcase( i==0 );
    testcase( i==1 );
    testcase( i==db->aLimit[SQLITE_LIMIT_VARIABLE_NUMBER]-1 );
    testcase( i==db->aLimit[SQLITE_LIMIT_VARIABLE_NUMBER] );
    if( i<1 || i>db->aLimit[SQLITE_LIMIT_VARIABLE_NUMBER] ){
      sqlite3ErrorMsg(pParse, "variable number must be between ?1 and ?%d",
          db->aLimit[SQLITE_LIMIT_VARIABLE_NUMBER]);
................................................................................
      assert( pE!=0 );
      if( memcmp(pE->u.zToken, z, n)==0 && pE->u.zToken[n]==0 ){
        pExpr->iColumn = pE->iColumn;
        break;
      }
    }
    if( i>=pParse->nVarExpr ){
#if SQLITE_MAX_VARIABLE_NUMBER<=32767
      pExpr->iColumn = (i16)(++pParse->nVar);
#else
      pExpr->iColumn = ++pParse->nVar;
#endif
      if( pParse->nVarExpr>=pParse->nVarExprAlloc-1 ){
        pParse->nVarExprAlloc += pParse->nVarExprAlloc + 10;
        pParse->apVarExpr =
            sqlite3DbReallocOrFree(
              db,
              pParse->apVarExpr,
              pParse->nVarExprAlloc*sizeof(pParse->apVarExpr[0])
................................................................................
  } 
  if( !pParse->nErr && pParse->nVar>db->aLimit[SQLITE_LIMIT_VARIABLE_NUMBER] ){
    sqlite3ErrorMsg(pParse, "too many SQL variables");
  }
}

/*
** Clear an expression structure without deleting the structure itself.
** Substructure is deleted.
*/
void sqlite3ExprClear(sqlite3 *db, Expr *p){
  assert( p!=0 );

  if( !ExprHasAnyProperty(p, EP_TokenOnly) ){
    sqlite3ExprDelete(db, p->pLeft);
    sqlite3ExprDelete(db, p->pRight);
    if( !ExprHasProperty(p, EP_Reduced) && (p->flags2 & EP2_MallocedToken)!=0 ){
      sqlite3DbFree(db, p->u.zToken);
    }
    if( ExprHasProperty(p, EP_xIsSelect) ){
      sqlite3SelectDelete(db, p->x.pSelect);
    }else{
      sqlite3ExprListDelete(db, p->x.pList);
    }
  }
}

/*
** Recursively delete an expression tree.
*/
void sqlite3ExprDelete(sqlite3 *db, Expr *p){
  if( p==0 ) return;
  sqlite3ExprClear(db, p);
  if( !ExprHasProperty(p, EP_Static) ){
    sqlite3DbFree(db, p);
  }
}

/*
** Return the number of bytes allocated for the expression structure 
................................................................................
               || (p->flags2 & EP2_MallocedToken)==0 );
    p->op = TK_INTEGER;
    p->flags |= EP_IntValue;
    p->u.iValue = *pValue;
  }
  return rc;
}

























































































/*
** Return TRUE if the given string is a row-id column name.
*/
int sqlite3IsRowid(const char *z){
  if( sqlite3StrICmp(z, "_ROWID_")==0 ) return 1;
  if( sqlite3StrICmp(z, "ROWID")==0 ) return 1;
................................................................................
** for fast set membership tests. In this case an epheremal table must 
** be used unless <column> is an INTEGER PRIMARY KEY or an index can 
** be found with <column> as its left-most column.
**
** When the b-tree is being used for membership tests, the calling function
** needs to know whether or not the structure contains an SQL NULL 
** value in order to correctly evaluate expressions like "X IN (Y, Z)".
** If there is a chance that the b-tree might contain a NULL value at
** runtime, then a register is allocated and the register number written
** to *prNotFound. If there is no chance that the b-tree contains a
** NULL value, then *prNotFound is left unchanged.
**
** If a register is allocated and its location stored in *prNotFound, then
** its initial value is NULL. If the b-tree does not remain constant
** for the duration of the query (i.e. the SELECT that generates the b-tree
** is a correlated subquery) then the value of the allocated register is
** reset to NULL each time the b-tree is repopulated. This allows the
** caller to use vdbe code equivalent to the following:
**
**   if( register==NULL ){
**     has_null = <test if data structure contains null>
**     register = 1
**   }
**
................................................................................
      if( rMayHaveNull ){
        sqlite3VdbeAddOp2(v, OP_Null, 0, rMayHaveNull);
      }

      affinity = sqlite3ExprAffinity(pLeft);

      /* Whether this is an 'x IN(SELECT...)' or an 'x IN(<exprlist>)'
      ** expression it is handled the same way. A virtual table is 
      ** filled with single-field index keys representing the results
      ** from the SELECT or the <exprlist>.
      **
      ** If the 'x' expression is a column value, or the SELECT...
      ** statement returns a column value, then the affinity of that
      ** column is used to build the index keys. If both 'x' and the
      ** SELECT... statement are columns, then numeric affinity is used
................................................................................
  }
  sqlite3ExprCachePop(pParse, 1);

  return rReg;
}
#endif /* SQLITE_OMIT_SUBQUERY */



























































































































/*
** Duplicate an 8-byte value
*/
static char *dup8bytes(Vdbe *v, const char *in){
  char *out = sqlite3DbMallocRaw(sqlite3VdbeDb(v), 8);
  if( out ){
    memcpy(out, in, 8);
................................................................................
      zId = pExpr->u.zToken;
      nId = sqlite3Strlen30(zId);
      pDef = sqlite3FindFunction(db, zId, nId, nFarg, enc, 0);
      if( pDef==0 ){
        sqlite3ErrorMsg(pParse, "unknown function: %.*s()", nId, zId);
        break;
      }





















      if( pFarg ){
        r1 = sqlite3GetTempRange(pParse, nFarg);
        sqlite3ExprCachePush(pParse);     /* Ticket 2ea2425d34be */
        sqlite3ExprCodeExprList(pParse, pFarg, r1, 1);
        sqlite3ExprCachePop(pParse, 1);   /* Ticket 2ea2425d34be */
      }else{
        r1 = 0;
................................................................................
    case TK_SELECT: {
      testcase( op==TK_EXISTS );
      testcase( op==TK_SELECT );
      inReg = sqlite3CodeSubselect(pParse, pExpr, 0, 0);
      break;
    }
    case TK_IN: {
      int rNotFound = 0;
      int rMayHaveNull = 0;
      int j2, j3, j4, j5;
      char affinity;
      int eType;

      VdbeNoopComment((v, "begin IN expr r%d", target));
      eType = sqlite3FindInIndex(pParse, pExpr, &rMayHaveNull);
      if( rMayHaveNull ){
        rNotFound = ++pParse->nMem;
      }

      /* Figure out the affinity to use to create a key from the results
      ** of the expression. affinityStr stores a static string suitable for
      ** P4 of OP_MakeRecord.
      */