#   <phrase number> . <column number> . <token offset>
#
# Options:
#
#   -near N        (NEAR distance. Default 10)
#   -col  C        (List of column indexes to match against)
#   -pc   VARNAME  (variable in caller frame to use for phrase numbering)

#
proc nearset {aCol args} {



  set O(-near) 10
  set O(-col)  {}
  set O(-pc)   ""


  set nOpt [lsearch -exact $args --]
  if {$nOpt<0} { error "no -- option" }





  foreach {k v} [lrange $args 0 [expr $nOpt-1]] {
    if {[info exists O($k)]==0} { error "unrecognized option $k" }
    set O($k) $v
  }

  if {$O(-pc) == ""} {
    set counter 0
  } else {
    upvar $O(-pc) counter
  }

  # Set $phraselist to be a list of phrases. $nPhrase its length.
  set phraselist [lrange $args [expr $nOpt+1] end]
  set nPhrase [llength $phraselist]

  for {set j 0} {$j < [llength $aCol]} {incr j} {
    for {set i 0} {$i < $nPhrase} {incr i} { 
      set A($j,$i) [list]
    }
  }

  set iCol -1
  foreach col $aCol {
    incr iCol




    if {$O(-col)!="" && [lsearch $O(-col) $iCol]<0} continue

    set nToken [llength $col]



















    set iFL [expr $O(-near) >= $nToken ? $nToken - 1 : $O(-near)]
    for { } {$iFL < $nToken} {incr iFL} {
      for {set iPhrase 0} {$iPhrase<$nPhrase} {incr iPhrase} {
        set B($iPhrase) [list]
      }

      
      for {set iPhrase 0} {$iPhrase<$nPhrase} {incr iPhrase} {
        set p [lindex $phraselist $iPhrase]
        set nPm1 [expr {[llength $p] - 1}]
        set iFirst [expr $iFL - $O(-near) - [llength $p]]

        for {set i $iFirst} {$i <= $iFL} {incr i} {
          set lCand [lrange $col $i [expr $i+$nPm1]]

          set bMatch 1
          foreach tok $p term $lCand {
            if {[string match $tok $term]==0} {
              #puts "$tok $term failed"
              set bMatch 0

            }
          }
          if {$bMatch} { 
            #puts "match at $i"
            lappend B($iPhrase) $i 
          }

          #if {$lCand == $p} { lappend B($iPhrase) $i }
        }
        if {[llength $B($iPhrase)] == 0} break

      }

      if {$iPhrase==$nPhrase} {
        for {set iPhrase 0} {$iPhrase<$nPhrase} {incr iPhrase} {
          set A($iCol,$iPhrase) [concat $A($iCol,$iPhrase) $B($iPhrase)]
          set A($iCol,$iPhrase) [lsort -integer -uniq $A($iCol,$iPhrase)]
        }
................................................................................
    }
    incr counter
  }

  #puts "$aCol -> $res"
  sort_poslist $res
}













#-------------------------------------------------------------------------
# Usage:
#
#   sort_poslist LIST
#
# Sort a position list of the type returned by command [nearset]
................................................................................
    lappend res $cand
  }
  set res [lsort -command fts5_collist_elem_compare -unique $res]
  return $res
}

# Comparison function used by fts5_poslist2collist to sort collist entries.
#
proc fts5_collist_elem_compare {a b} {
  foreach {a1 a2} [split $a .] {}
  foreach {b1 b2} [split $b .] {}

  if {$a1==$b1} { return [expr $a2 - $b2] }
  return [expr $a1 - $b1]
}
................................................................................
#   SELECT 
#     rowid, 
#     fts5_test_poslist($tbl),
#     fts5_test_collist($tbl) 
#   FROM $tbl('$expr')
#   ORDER BY rowid $order;
#
proc fts5_query_data {expr tbl {order ASC}} {

  # Figure out the set of columns in the FTS5 table. This routine does
  # not handle tables with UNINDEXED columns, but if it did, it would
  # have to be here.
  db eval "PRAGMA table_info = $tbl" x { lappend lCols $x(name) }







  set cols ""
  foreach e $lCols { append cols ", '$e'" }
  set tclexpr [db one [subst -novar {
    SELECT fts5_expr_tcl( $expr, 'nearset $cols -pc ::pc' [set cols] )
  }]]

  set res [list]
  db eval "SELECT rowid, * FROM $tbl ORDER BY rowid $order" x {
    set cols [list]
    foreach col $lCols { lappend cols $x($col) }
    
................................................................................

  set res
}

#-------------------------------------------------------------------------
# Similar to [fts5_query_data], but omit the collist field.
#
proc fts5_poslist_data {expr tbl {order ASC}} {
  set res [list]








  foreach {rowid poslist collist} [fts5_query_data $expr $tbl $order] {
    lappend res $rowid $poslist
  }
  set res
}

#-------------------------------------------------------------------------
#
proc nearset_rf {aCol args} {
  set idx [lsearch -exact $args --]
  if {$idx != [llength $args]-2 || [llength [lindex $args end]]!=1} {
    set ::expr_not_ok 1
  }
  list
}




proc nearset_rc {aCol args} {
  nearset_rf $aCol {*}$args
  if {[lsearch $args -col]>=0} { 
    set ::expr_not_ok 1
  }
  list
}




proc fts5_expr_ok {expr tbl} {

  if {![detail_is_full]} {
    set nearset "nearset_rc"
    if {[detail_is_col]} { set nearset "nearset_rf" }

    set ::expr_not_ok 0
................................................................................
    }]]
    eval $tclexpr
    if {$::expr_not_ok} { return 0 }
  }

  return 1
}

#   <phrase number> . <column number> . <token offset>
#
# Options:
#
#   -near N        (NEAR distance. Default 10)
#   -col  C        (List of column indexes to match against)
#   -pc   VARNAME  (variable in caller frame to use for phrase numbering)
#   -dict VARNAME  (array in caller frame to use for synonyms)
#
proc nearset {aCol args} {

  # Process the command line options.
  #
  set O(-near) 10
  set O(-col)  {}
  set O(-pc)   ""
  set O(-dict) ""

  set nOpt [lsearch -exact $args --]
  if {$nOpt<0} { error "no -- option" }

  # Set $lPhrase to be a list of phrases. $nPhrase its length.
  set lPhrase [lrange $args [expr $nOpt+1] end]
  set nPhrase [llength $lPhrase]

  foreach {k v} [lrange $args 0 [expr $nOpt-1]] {
    if {[info exists O($k)]==0} { error "unrecognized option $k" }
    set O($k) $v
  }

  if {$O(-pc) == ""} {
    set counter 0
  } else {
    upvar $O(-pc) counter
  }

  if {$O(-dict)!=""} { upvar $O(-dict) aDict }



  for {set j 0} {$j < [llength $aCol]} {incr j} {
    for {set i 0} {$i < $nPhrase} {incr i} { 
      set A($j,$i) [list]
    }
  }

  # Loop through each column of the current row.
  for {set iCol 0} {$iCol < [llength $aCol]} {incr iCol} {


    # If there is a column filter, test whether this column is excluded. If
    # so, skip to the next iteration of this loop. Otherwise, set zCol to the
    # column value and nToken to the number of tokens that comprise it.
    if {$O(-col)!="" && [lsearch $O(-col) $iCol]<0} continue
    set zCol [lindex $aCol $iCol]
    set nToken [llength $zCol]

    # Each iteration of the following loop searches a substring of the 
    # column value for phrase matches. The last token of the substring
    # is token $iLast of the column value. The first token is:
    #
    #   iFirst = ($iLast - $O(-near) - 1)
    #
    # where $sz is the length of the phrase being searched for. A phrase 
    # counts as matching the substring if its first token lies on or before
    # $iLast and its last token on or after $iFirst.
    #
    # For example, if the query is "NEAR(a+b c, 2)" and the column value:
    #
    #   "x x x x A B x x C x"
    #    0 1 2 3 4 5 6 7 8 9"
    #
    # when (iLast==8 && iFirst=5) the range will contain both phrases and
    # so both instances can be added to the output poslists.
    #
    set iLast [expr $O(-near) >= $nToken ? $nToken - 1 : $O(-near)]
    for { } {$iLast < $nToken} {incr iLast} {



      catch { array unset B }
      
      for {set iPhrase 0} {$iPhrase<$nPhrase} {incr iPhrase} {
        set p [lindex $lPhrase $iPhrase]
        set nPm1 [expr {[llength $p] - 1}]
        set iFirst [expr $iLast - $O(-near) - [llength $p]]

        for {set i $iFirst} {$i <= $iLast} {incr i} {
          set lCand [lrange $zCol $i [expr $i+$nPm1]]

          set bMatch 1
          foreach tok $p term $lCand {



            if {[nearset_match aDict $tok $term]==0} { set bMatch 0 ; break }
          }



          if {$bMatch} { lappend B($iPhrase) $i }
        }




        if {![info exists B($iPhrase)]} break
      }

      if {$iPhrase==$nPhrase} {
        for {set iPhrase 0} {$iPhrase<$nPhrase} {incr iPhrase} {
          set A($iCol,$iPhrase) [concat $A($iCol,$iPhrase) $B($iPhrase)]
          set A($iCol,$iPhrase) [lsort -integer -uniq $A($iCol,$iPhrase)]
        }
................................................................................
    }
    incr counter
  }

  #puts "$aCol -> $res"
  sort_poslist $res
}

proc nearset_match {aDictVar tok term} {
  if {[string match $tok $term]} { return 1 }

  upvar $aDictVar aDict
  if {[info exists aDict($tok)]} {
    foreach s $aDict($tok) {
      if {[string match $s $term]} { return 1 }
    }
  }
  return 0;
}

#-------------------------------------------------------------------------
# Usage:
#
#   sort_poslist LIST
#
# Sort a position list of the type returned by command [nearset]
................................................................................
    lappend res $cand
  }
  set res [lsort -command fts5_collist_elem_compare -unique $res]
  return $res
}

# Comparison function used by fts5_poslist2collist to sort collist entries.

proc fts5_collist_elem_compare {a b} {
  foreach {a1 a2} [split $a .] {}
  foreach {b1 b2} [split $b .] {}

  if {$a1==$b1} { return [expr $a2 - $b2] }
  return [expr $a1 - $b1]
}
................................................................................
#   SELECT 
#     rowid, 
#     fts5_test_poslist($tbl),
#     fts5_test_collist($tbl) 
#   FROM $tbl('$expr')
#   ORDER BY rowid $order;
#
proc fts5_query_data {expr tbl {order ASC} {aDictVar ""}} {

  # Figure out the set of columns in the FTS5 table. This routine does
  # not handle tables with UNINDEXED columns, but if it did, it would
  # have to be here.
  db eval "PRAGMA table_info = $tbl" x { lappend lCols $x(name) }

  set d ""
  if {$aDictVar != ""} {
    upvar $aDictVar aDict
    set d aDict
  }

  set cols ""
  foreach e $lCols { append cols ", '$e'" }
  set tclexpr [db one [subst -novar {
    SELECT fts5_expr_tcl( $expr, 'nearset $cols -dict $d -pc ::pc' [set cols] )
  }]]

  set res [list]
  db eval "SELECT rowid, * FROM $tbl ORDER BY rowid $order" x {
    set cols [list]
    foreach col $lCols { lappend cols $x($col) }
    
................................................................................

  set res
}

#-------------------------------------------------------------------------
# Similar to [fts5_query_data], but omit the collist field.
#
proc fts5_poslist_data {expr tbl {order ASC} {aDictVar ""}} {
  set res [list]

  if {$aDictVar!=""} {
    upvar $aDictVar aDict
    set dict aDict
  } else {
    set dict ""
  }

  foreach {rowid poslist collist} [fts5_query_data $expr $tbl $order $dict] {
    lappend res $rowid $poslist
  }
  set res
}

#-------------------------------------------------------------------------
#







# This command will only work inside a [foreach_detail_mode] block. It tests
# whether or not expression $expr run on FTS5 table $tbl is supported by
# the current mode. If so, 1 is returned. If not, 0.
#







#   detail=full    (all queries supported)
#   detail=col     (all but phrase queries and NEAR queries)
#   detail=none    (all but phrase queries, NEAR queries, and column filters)
#
proc fts5_expr_ok {expr tbl} {

  if {![detail_is_full]} {
    set nearset "nearset_rc"
    if {[detail_is_col]} { set nearset "nearset_rf" }

    set ::expr_not_ok 0
................................................................................
    }]]
    eval $tclexpr
    if {$::expr_not_ok} { return 0 }
  }

  return 1
}

# Helper for [fts5_expr_ok]
proc nearset_rf {aCol args} {
  set idx [lsearch -exact $args --]
  if {$idx != [llength $args]-2 || [llength [lindex $args end]]!=1} {
    set ::expr_not_ok 1
  }
  list
}

# Helper for [fts5_expr_ok]
proc nearset_rc {aCol args} {
  nearset_rf $aCol {*}$args
  if {[lsearch $args -col]>=0} { 
    set ::expr_not_ok 1
  }
  list
}

# 2014 Dec 20
#
# 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.
#
#***********************************************************************
#
# Tests focusing on custom tokenizers that support synonyms.
#

source [file join [file dirname [info script]] fts5_common.tcl]
set testprefix fts5synonym

# If SQLITE_ENABLE_FTS5 is defined, omit this file.
ifcapable !fts5 {
  finish_test
  return
}

#-------------------------------------------------------------------------
# Code for a simple Tcl tokenizer that supports synonyms at query time.
#
foreach SYNDICT {
  {zero  0}
  {one   1 i}
  {two   2 ii}
  {three 3 iii}
  {four  4 iv}
  {five  5 v}
  {six   6 vi}
  {seven 7 vii}
  {eight 8 viii}
  {nine  9 ix}
} {
  foreach s $SYNDICT {
    set o [list]
    foreach x $SYNDICT {if {$x!=$s} {lappend o $x}}
    set ::syn($s) $o
  }
}

proc tcl_tokenize {tflags text} {
  foreach {w iStart iEnd} [fts5_tokenize_split $text] {
    sqlite3_fts5_token $w $iStart $iEnd
    if {$tflags == "query"} {
      foreach s $::SYNDICT($w)  { sqlite3_fts5_token -colo $s $iStart $iEnd }
    }
  }
}

proc tcl_create {args} {
  return "tcl_tokenize"
}

#
# End of tokenizer code.
#-------------------------------------------------------------------------

foreach_detail_mode $testprefix {

sqlite3_fts5_create_tokenizer db tcl tcl_create
fts5_aux_test_functions db

do_execsql_test 1.0 {
  CREATE VIRTUAL TABLE ss USING fts5(a, b, tokenize=tcl, detail=%DETAIL%);

  INSERT INTO ss VALUES('5 5 five seven 3 seven i', '2 1 5 0 two 1 i');
  INSERT INTO ss VALUES('six ix iii 7 i vii iii', 'one seven nine 4 9 1 vi');
  INSERT INTO ss VALUES('6 viii i five six zero seven', '5 v iii iv iv 3');
  INSERT INTO ss VALUES('9 ii six 8 1 6', 'six 4 iv iv 7');
  INSERT INTO ss VALUES('1 5 4 eight ii iv iii', 'nine 2 eight ix v vii');
  INSERT INTO ss VALUES('one 7 seven six 2 two', '1 2 four 7 4 3 4');
  INSERT INTO ss VALUES('eight iv 4 nine vii six 1', '5 6 v one zero 4');
  INSERT INTO ss VALUES('v 9 8 iii 4', '9 4 seven two vi vii');
  INSERT INTO ss VALUES('3 ix two 9 0 nine i', 'five ii nine two viii i five');
  INSERT INTO ss VALUES('six iii 9 two eight 2', 'nine i nine vii nine');
  INSERT INTO ss VALUES('6 three zero seven vii five', '8 vii ix 0 7 seven');
  INSERT INTO ss VALUES('8 vii 8 7 3 4', 'eight iii four viii nine iv three');
  INSERT INTO ss VALUES('4 v 7 two 0 one 8', 'vii 1 two five i zero 9');
  INSERT INTO ss VALUES('3 ii vii vi eight', '8 4 ix one three eight');
  INSERT INTO ss VALUES('iv eight seven 6 9 seven', 'one vi two five seven');
  INSERT INTO ss VALUES('i i 5 i v vii eight', '2 seven i 2 2 four');
  INSERT INTO ss VALUES('0 i iii nine 3 ix five', '0 eight iv 0 six 2');
  INSERT INTO ss VALUES('iv vii three 3 9 one 8', '2 ii 6 eight ii six six');
  INSERT INTO ss VALUES('eight one two nine six', '8 9 3 viii vi');
  INSERT INTO ss VALUES('one 0 four ii eight one 3', 'iii eight vi vi vi');
  INSERT INTO ss VALUES('4 0 eight 0 0', '1 four one vii seven ii');
  INSERT INTO ss VALUES('1 zero nine 2 2', 'viii iv two vi nine v iii');
  INSERT INTO ss VALUES('5 five viii four four vi', '8 five 7 vii 6 4');
  INSERT INTO ss VALUES('7 ix four 8 vii', 'nine three nine ii ix vii');
  INSERT INTO ss VALUES('nine iv v i 0 v', 'two iv vii six i ix 4');
  INSERT INTO ss VALUES('one v v one viii 3 8', '2 1 3 five iii');
  INSERT INTO ss VALUES('six ii 5 nine 4 viii seven', 'eight i ix ix 7 four');
  INSERT INTO ss VALUES('9 ii two seven three 7 0', 'six viii seven 7 five');
  INSERT INTO ss VALUES('five two 4 viii nine', '9 7 nine zero 1 two one');
  INSERT INTO ss VALUES('viii 8 iii i ii 8 3', '4 2 7 v 8 8');
  INSERT INTO ss VALUES('four vii 4 iii zero 0 vii', '3 viii iii zero 9 i');
  INSERT INTO ss VALUES('0 seven v five i five v', 'one 4 2 ix 9');
  INSERT INTO ss VALUES('two 5 two two ix 4 1', '3 nine ii v nine 3 five');
  INSERT INTO ss VALUES('five 5 7 4 6 vii', 'three 2 ix 2 8 6');
  INSERT INTO ss VALUES('six iii vi iv seven eight', '8 six 7 0 4');
  INSERT INTO ss VALUES('vi vi iv 3 0 one one', '9 6 eight ix iv');
  INSERT INTO ss VALUES('7 2 2 iii 0', '0 0 seven 1 nine');
  INSERT INTO ss VALUES('8 6 iv six ii', 'iv 6 3 4 ii five');
  INSERT INTO ss VALUES('0 two two seven ii', 'vii ix four 4 zero vi vi');
  INSERT INTO ss VALUES('2 one eight 8 9 7', 'vi 3 0 3 vii');
  INSERT INTO ss VALUES('iii ii ix iv three', 'vi i 6 1 two');
  INSERT INTO ss VALUES('eight four nine 8 seven', 'one three i nine iii one');
  INSERT INTO ss VALUES('iii seven five ix 8', 'ii 7 seven 0 four ii');
  INSERT INTO ss VALUES('four 0 1 5 two', 'iii 9 5 ii ii 2 4');
  INSERT INTO ss VALUES('iii nine four vi 8 five six', 'i i ii seven vi vii');
  INSERT INTO ss VALUES('eight vii eight six 3', 'i vii 1 six 9 vii');
  INSERT INTO ss VALUES('9 0 viii viii five', 'i 1 viii ix 3 4');
  INSERT INTO ss VALUES('three nine 5 nine viii four zero', 'ii i 1 5 2 viii');
  INSERT INTO ss VALUES('5 vii three 9 four', 'three five one 7 2 eight one');
}

foreach {tn expr} {
  1 "eight"
} {
  if {[fts5_expr_ok $expr ss]==0} {
    do_test 1.$tn.OMITTED { list } [list]
    continue
  }

  set res [fts5_query_data $expr ss ASC ::SYNDICT]
  do_execsql_test 1.$tn.[llength $res].asc {
    SELECT rowid, fts5_test_poslist(ss), fts5_test_collist(ss) FROM ss($expr)
  } $res
}

}

finish_test