SQLite

3.5.0

** on how SQLite interfaces are suppose to operate.
**
** The name of this file under configuration management is "sqlite.h.in".
** The makefile makes some minor changes to this file (such as inserting
** the version number) and changes its name to "sqlite3.h" as
** part of the build process.
**
** @(#) $Id: sqlite.h.in,v 1.256 2007/09/03 20:32:45 drh Exp $
*/
#ifndef _SQLITE3_H_
#define _SQLITE3_H_
#include <stdarg.h>     /* Needed for the definition of va_list */

/*
** Make sure we can call this stuff from C++.

** the database for reading and writing if possible, or reading only if
** if the file is write protected.  In either case the database must already
** exist or an error is returned.  The third option opens the database
** for reading and writing and creates it if it does not already exist.
** The third options is behavior that is always used for [sqlite3_open()]
** and [sqlite3_open16()].
**
** If the filename is ":memory:", then an private
** in-memory database is created for the connection.  This in-memory
** database will vanish when the database connection is closed.  Future
** version of SQLite might make use of additional special filenames
** that begin with the ":" character.  It is recommended that 
** when a database filename really does begin with
** ":" that you prefix the filename with a pathname like "./" to
** avoid ambiguity.
**
** If the filename is an empty string, then a private temporary
** on-disk database will be created.  This private database will be
** automatically deleted as soon as the database connection is closed.
**
** The fourth parameter to sqlite3_open_v2() is the name of the
** [sqlite3_vfs] object that defines the operating system 
** interface that the new database connection should use.  If the
** fourth parameter is a NULL pointer then the default [sqlite3_vfs]
** object is used.
**
** <b>Note to windows users:</b>  The encoding used for the filename argument

#
# Run this TCL script to generate HTML for the goals.html file.
#
set rcsid {$Id: 34to35.tcl,v 1.3 2007/09/03 20:32:45 drh Exp $}
source common.tcl
header {SQLite Changes From Version 3.4.2 To 3.5.0}

proc CODE {text} {
  puts "<blockquote><pre>"
  puts $text
  puts "</pre></blockquote>"

    puts "<br>$caption"
  }
  puts "</center>"
}
proc PARAGRAPH {text} {
  # regsub -all "/(\[a-zA-Z0-9\]+)/" $text {<i>\1</i>} t2
  #regsub -all "\\*(\[^\n*\]+)\\*" $text {<tt><b><big>\1</big></b></tt>} t3
  regsub -all {\[([^]\n]+)\]} $text {[resolve_link \1]} t3
  puts "<p>[subst -novar -noback $t3]</p>\n"
}
proc resolve_link {args} {
  set a2 [split $args |]
  set id [string trim [lindex $a2 0]]
  if {[lindex $a2 1]==""} {
    set display [string trim [lindex $a2 0]]
  } else {
    set display [string trim [lrange $a2 1 end]]
  }
  regsub -all {[^a-zA-Z0-9_]} $id {} id
  return "<a href=\"capi3ref.html#$id\">$display</a>"
}
set level(0) 0
set level(1) 0
proc HEADING {n name {tag {}}} {
  if {$tag!=""} {
    puts "<a name=\"$tag\">"
  }

  is provide here.  Subsequent sections will describe these
  changes in more detail.
}
PARAGRAPH {
  <ol>
  <li>The OS interface layer has been completely reworked:
  <ol type="a">
  <li>The undocumented <b>sqlite3_os_switch()</b> interface has
      been removed.</li>
  <li>The <b>SQLITE_ENABLE_REDEF_IO</b> compile-time flag no longer functions.
      I/O procedures are now always redefinable.</li>
  <li>Three new objects are defined for specifying I/O procedures:
      [sqlite3_vfs], [sqlite3_file], and [sqlite3_io_methods].</li>
  <li>Three new interfaces are used to create alternative OS interfaces:
      [sqlite3_vfs_register()], [sqlite3_vfs_unregister()], and
      [sqlite3_vfs_find()].</li>
  <li>A new interface has been added to provided additional control over

  systems. 
}

HEADING 1 {The OS Interface Layer}

PARAGRAPH {
  If your system defines a custom OS interface for SQLite or if you
  were using the undocumented <b>sqlite3_os_switch()</b>
  interface, then you will need to make modifications in order to
  upgrade to SQLite version 3.5.0.  This may seem painful at first
  glance.  But as you look more closely, you will probably discover
  that your changes are made smaller and easier to understand and manage
  by the new SQLite interface.  It is likely that your changes will
  now also work seamlessly with the SQLite amalgamation.  You will
  no longer need to make any changes to the code SQLite source code.

  SQLite will guarantee that the zFilename string passed to
  xOpen() is a full pathname as generated by xFullPathname() and
  that the string will be valid and unchanged until xClose() is
  called.  So the [sqlite3_file] can store a pointer to the
   filename if it needs to remember the filename for some reason.
   The flags argument to xOpen() is a copy of the flags argument
   to sqlite3_open_v2().  If sqlite3_open() or sqlite3_open16()
   is used, then flags is [SQLITE_OPEN_READWRITE] | [SQLITE_OPEN_CREATE].
   If xOpen() opens a file read-only then it sets *pOutFlags to
   include [SQLITE_OPEN_READONLY].  Other bits in *pOutFlags may be
   set.
   SQLite will also add one of the following flags to the xOpen()
   call, depending on the object being opened:
   <ul>
   <li>  [SQLITE_OPEN_MAIN_DB]
   <li>  [SQLITE_OPEN_MAIN_JOURNAL]
   <li>  [SQLITE_OPEN_TEMP_DB]
   <li>  [SQLITE_OPEN_TEMP_JOURNAL]
   <li>  [SQLITE_OPEN_TRANSIENT_DB]
   <li>  [SQLITE_OPEN_SUBJOURNAL]
   <li>  [SQLITE_OPEN_MASTER_JOURNAL]
   </ul>
   The file I/O implementation can use the object type flags to
   changes the way it deals with files.  For example, an application
   that does not care about crash recovery or rollback, might make
   the open of a journal file a no-op.  Writes to this journal are
   also a no-op.  Any attempt to read the journal returns [SQLITE_IOERR].
   Or the implementation might recognize the a database file will
   be doing page-aligned sector reads and writes in a random order
   and set up its I/O subsystem accordingly.
   SQLite might also add one of the following flags to the xOpen
   method:
   <ul>
   <li> [SQLITE_OPEN_DELETEONCLOSE]
   <li> [SQLITE_OPEN_EXCLUSIVE]
   </ul>
   The [SQLITE_OPEN_DELETEONCLOSE] flag means the file should be
   deleted when it is closed.  This will always be set for TEMP 
   databases and journals and for subjournals.  The 
   [SQLITE_OPEN_EXCLUSIVE] flag means the file should be opened
   for exclusive access.  This flag is set for all files except
   for the main database file.
   The [sqlite3_file] structure passed as the third argument to
   xOpen is allocated by the caller.  xOpen just fills it in.  The
   caller allocates a minimum of szOsFile bytes for the [sqlite3_file]
   structure.
}

PARAGRAPH {
  The differences between an [SQLITE_OPEN_TEMP_DB] database and an
  [SQLITE_OPEN_TRANSIENT_DB] database is this:  The [SQLITE_OPEN_TEMP_DB]
  is used for explicitly declared and named TEMP tables (using the
  CREATE TEMP TABLE syntax) or for named tables in a temporary database
  that is created by opening a database with a filename that is an empty
  string.  An [SQLITE_OPEN_TRANSIENT_DB] holds an database table that
  SQLite creates automatically in order to evaluate a subquery or
  ORDER BY or GROUP BY clause.  Both TEMP_DB and TRANSIENT_DB databases
  are private and are deleted automatically.  TEMP_DB databases last
  for the duration of the database connection.  TRANSIENT_DB databases
  last only for the duration of a single SQL statement.
}

PARAGRAPH {
  The xDelete method is used delete a file.  The name of the file is
  given in the second parameter.  The filename will be in UTF-8.
  The VFS must convert the filename into whatever character representation
  the underlying operating system expects.  If the syncDir parameter is
  true, then the xDelete method should not return until the change
  to the directory contents for the directory containing the
  deleted file have been synced to disk in order to insure that the
  file does not "reappear" if a power failure occurs soon after.
}

PARAGRAPH {
  The xAccess method is used to check for access permissions on a file.
  The filename will be UTF-8 encoded.  The flags argument will be
  [SQLITE_ACCESS_EXISTS] to check for the existence of the file,
  [SQLITE_ACCESS_READWRITE] to check to see if the file is both readable
  and writable, or [SQLITE_ACCESS_READ] to check to see if the file is
  at least readable.  The "file" named by the second parameter might
  be a directory or folder name.
}

PARAGRAPH {
  The xGetTempName method computes the name of a temporary file that
  SQLite can use.  The name should be written into the buffer given

  int (*xWrite)(sqlite3_file*, const void*, int iAmt, sqlite_int64 iOfst);
  int (*xTruncate)(sqlite3_file*, sqlite_int64 size);
  int (*xSync)(sqlite3_file*, int flags);
  int (*xFileSize)(sqlite3_file*, sqlite_int64 *pSize);
  int (*xLock)(sqlite3_file*, int);
  int (*xUnlock)(sqlite3_file*, int);
  int (*xCheckReservedLock)(sqlite3_file*);
  int (*xFileControl)(sqlite3_file*, int op, void *pArg);

  int (*xSectorSize)(sqlite3_file*);
  int (*xDeviceCharacteristics)(sqlite3_file*);
  /* Additional methods may be added in future releases */
};
}

PARAGRAPH {

  contains pointers to other allocated memory or resources, those
  allocations should be released by the xClose method.
}

PARAGRAPH {
  The xRead method reads iAmt bytes from the file beginning at a byte
  offset to iOfst.  The data read is stored in the pointer of the
  second parameter.  xRead returns the [SQLITE_OK] on success,
  [SQLITE_IOERR_SHORT_READ] if it was not able to read the full number
  of bytes because it reached end-of-file, or [SQLITE_IOERR_READ] for
  any other error.
}

PARAGRAPH {
  The xWrite method writes iAmt bytes of data from the second parameter
  into the file beginning at an offset of iOfst bytes.  If the size of
  the file is less than iOfst bytes prior to the write, then xWrite should
  ensure that the file is extended with zeros up to iOfst bytes prior
  to beginning its write.  xWrite continues to extends the file as
  necessary so that the size of the file is at least iAmt+iOfst bytes 
  at the conclusion of the xWrite call.  The xWrite method returns
  [SQLITE_OK] on success.  If the write cannot complete because the
  underlying storage medium is full, then [SQLITE_FULL] is returned.
  [SQLITE_IOERR_WRITE] should be returned for any other error.
}

PARAGRAPH {
  The xTruncate method truncates a file to be nByte bytes in length.
  If the file is already nByte bytes or less in length then this
  method is a no-op.  The xTruncate method returns [SQLITE_OK] on
  success and [SQLITE_IOERR_TRUNCATE] if anything goes wrong.
}

PARAGRAPH {
  The xSync method is used to force previously written data out of
  operating system cache and into non-volatile memory.  The second
  parameter is usually [SQLITE_SYNC_NORMAL].  If the second parameter
  is [SQLITE_SYNC_FULL] then the xSync method should make sure that
  data has also been flushed through the disk controllers cache.
  The [SQLITE_SYNC_FULL] parameter is the equivalent of the F_FULLSYNC





  ioctl() on Mac OS X. The xSync method returns
  [SQLITE_OK] on success and [SQLITE_IOERR_FSYNC] if anything goes wrong.
}

PARAGRAPH {
  The xFileSize() method determines the current size of the file
  in bytes and writes that value into *pSize.  It returns [SQLITE_OK]
  on success and [SQLITE_IOERR_FSTAT] if something goes wrong.
}

PARAGRAPH {
  The xLock and xUnlock methods are used to set and clear file locks.
  SQLite supports five levels of file locks, in order:
  <ul>
  <li> [SQLITE_LOCK_NONE]
  <li> [SQLITE_LOCK_SHARED]
  <li> [SQLITE_LOCK_RESERVED]
  <li> [SQLITE_LOCK_PENDING]
  <li> [SQLITE_LOCK_EXCLUSIVE]
  </ul>
  The underlying implementation can support some subset of these locking
  levels as long as it meets the other requirements of this paragraph.
  The locking level is specified as the second argument to both xLock
  and xUnlock.  The xLock method increases the locking level to the
  specified locking level or higher.  The xUnlock method decreases the
  locking level to no lower than the level specified.  
  [SQLITE_LOCK_NONE] means that the file is unlocked.  [SQLITE_LOCK_SHARED]
  gives permission to read the file.  Multiple database connections can
  hold [SQLITE_LOCK_SHARED] at the same time.
  [SQLITE_LOCK_RESERVED] is like [SQLITE_LOCK_SHARED] in that its is permission
  to read the file.  But only a single connection can hold a reserved lock
  at any point in time.  The [SQLITE_LOCK_PENDING] is also permission to
  read the file.  Other connections can continue to read the file as well,
  but no other connection is allowed to escalate a lock from none to shared.
  [SQLITE_LOCK_EXCLUSIVE] is permission to write on the file.  Only a single
  connection can hold an exclusive lock and no other connection can hold
  any lock (other than "none") while one connection is hold an exclusive
  lock.  The xLock returns [SQLITE_OK] on success, [SQLITE_BUSY] if it
  is unable to obtain the lock, or [SQLITE_IOERR_RDLOCK] if something else
  goes wrong.  The xUnlock method returns [SQLITE_OK] on success and
  [SQLITE_IOERR_UNLOCK] for problems.
}

PARAGRAPH {
  The xCheckReservedLock method checks to see if another connection or
  another process is currently holding a reserved, pending, or exclusive
  lock on the file.  It returns true or false.
}

PARAGRAPH {
  The xFileControl() method is a generic interface that allows custom
  VFS implementations to directly control an open file using the
  (new and experimental)
  [sqlite3_file_control()] interface.  The second "op" argument
  is an integer opcode.   The third
  argument is a generic pointer which is intended to be a pointer
  to a structure that may contain arguments or space in which to
  write return values.  Potential uses for xFileControl() might be
  functions to enable blocking locks with timeouts, to change the
  locking strategy (for example to use dot-file locks), to inquire
  about the status of a lock, or to break stale locks.  The SQLite
  core reserves opcodes less than 100 for its own use. 
  A [SQLITE_FCNTL_LOCKSTATE | list of opcodes] less than 100 is available.
  Applications that define a custom xFileControl method should use opcodes 
  greater than 100 to avoid conflicts.
}

PARAGRAPH {
  The xSectorSize returns the "sector size" of the underlying
  non-volatile media.  A "sector" is defined as the smallest unit of
  storage that can be written without disturbing adjacent storage.
  On a disk drive the "sector size" has until recently been 512 bytes,
  though there is a push to increase this value to 4KiB.  SQLite needs
  to know the sector size so that it can write a full sector at a
  time, and thus avoid corrupting adjacent storage space if a power
  lose occurs in the middle of a write.
}

PARAGRAPH {
  The xDeviceCharacteristics method returns an integer bit vector that
  defines any special properties that the underlying storage medium might
  have that SQLite can use to increase performance.  The allowed return
  is the bit-wise OR of the following values:
  <ul>
  <li> [SQLITE_IOCAP_ATOMIC]
  <li> [SQLITE_IOCAP_ATOMIC512]
  <li> [SQLITE_IOCAP_ATOMIC1K]
  <li> [SQLITE_IOCAP_ATOMIC2K]
  <li> [SQLITE_IOCAP_ATOMIC4K]
  <li> [SQLITE_IOCAP_ATOMIC8K]
  <li> [SQLITE_IOCAP_ATOMIC16K]
  <li> [SQLITE_IOCAP_ATOMIC32K]
  <li> [SQLITE_IOCAP_ATOMIC64K]
  <li> [SQLITE_IOCAP_SAFE_APPEND]
  <li> [SQLITE_IOCAP_SEQUENTIAL]
  </ul>
  The [SQLITE_IOCAP_ATOMIC] bit means that all writes to this device are
  atomic in the sense that either the entire write occurs or none of it
  occurs.  The other 
  [SQLITE_IOCAP_ATOMIC | SQLITE_IOCAP_ATOMIC<i>nnn</i>] values indicate that
  writes of aligned blocks of the indicated size are atomic.
  [SQLITE_IOCAP_SAFE_APPEND] means that when extending a file with new
  data, the new data is written first and then the file size is updated.
  So if a power failure occurs, there is no chance that the file might have
  been extended with randomness.  The [SQLITE_IOCAP_SEQUENTIAL] bit means
  that all writes occur in the order that they are issued and are not
  reordered by the underlying file system.
}

HEADING 3 {Checklist For Constructing A New VFS}

PARAGRAPH {
  The preceding paragraphs contain a lot of information.
  To ease the task of constructing
  a new VFS for SQLite we offer the following implementation checklist:
}

PARAGRAPH {
  <ol>
  <li> Define an appropriate subclass of the [sqlite3_file] object.
  <li> Implement the methods required by the [sqlite3_io_methods] object.
  <li> Create a static and 
       constant [sqlite3_io_methods] object containing pointers
       to the methods from the previous step.
  <li> Implement the xOpen method that opens a file and populates an
       [sqlite3_file] object, including setting pMethods to
       point to the [sqlite3_io_methods] object from the previous step.
  <li> Implement the other methods required by [sqlite3_vfs].

  The argument to [sqlite3_mutex_alloc()] should be 
  [SQLITE_MUTEX_FAST] or [SQLITE_MUTEX_RECURSIVE] for non-recursive
  and recursive mutexes, respectively.  If the underlying system does
  not provide non-recursive mutexes, then a recursive mutex can be
  substituted in that case.  The argument to [sqlite3_mutex_alloc()]
  can also be a constant designating one of several static mutexes:
  <ul>
  <li>  [SQLITE_MUTEX_STATIC_MASTER]
  <li>  [SQLITE_MUTEX_STATIC_MEM]
  <li>  [SQLITE_MUTEX_STATIC_MEM2]
  <li>  [SQLITE_MUTEX_STATIC_PRNG]
  <li>  [SQLITE_MUTEX_STATIC_LRU]
  </ul>
  These static mutexes are reserved for use internally by SQLite
  and should not be used by the application.  The static mutexes
  are all non-recursive.
}

PARAGRAPH {
  The [sqlite3_mutex_free()] routine should be used to deallocate
  a non-static mutex.  If a static mutex is passed to this routine
  then the behavior is undefined.
}

PARAGRAPH {
  The [sqlite3_mutex_enter()] attempts to enter the mutex and blocks
  if another threads is already there.  [sqlite3_mutex_try()] attempts
  to enter and returns [SQLITE_OK] on success or [SQLITE_BUSY] if another
  thread is already there.  [sqlite3_mutex_leave()] exits a mutex.
  The mutex is held until the number of exits matches the number of
  entrances.  If [sqlite3_mutex_leave()] is called on a mutex that 
  the thread is not currently holding, then the behavior is undefined.
  If any routine is called for a deallocated mutex, then the behavior
  is undefined.
}

  }
  puts "<DT><B>$date</B></DT>"
  regsub -all {[Tt]icket #(\d+)} $desc \
      {<a href="http://www.sqlite.org/cvstrac/tktview?tn=\1">\0</a>} desc
  puts "<DD><P><UL>$desc</UL></P>"
  puts "</DD>"
}


chng {2007 Sep 3 (3.5.0 beta)} {
<li>Redesign the OS interface layer.  See
    <a href="34to35.html">34to35.html</a> for details.
    <font color="red">*** Potentially incompatible change ***</font>
<li>The <a href="capi3ref.html#sqlite3_release_memory">
    sqlite3_release_memory()</a>,
    <a href="capi3ref.html#sqlite3_soft_heap_limit">
    sqlite3_soft_heap_limit()</a>,
    and <a href="capi3ref.html#sqlite3_enable_shared_cache">
    sqlite3_enable_shared_cache()</a> interfaces now work cross all
    threads in the process, not just the single thread in which they
    are invoked.
    <font color="red">*** Potentially incompatible change ***</font>
<li>Added the 
    <a href="capi3ref.html#sqlite3_open_v2">sqlite3_open_v2()</a>
    interface.  
<li>Reimplemented the memory allocation subsystem and made it 
    replacable at compile-time.
<li>Created a new mutex subsystem and made it replacable at
    compile-time.
<li>The same database connection may now be used simultaneously by
    separate threads.
}


chng {2007 August 13 (3.4.2)} {
<li>Fix a database corruption bug that might occur if a ROLLBACK command
is executed in <a href="pragma.html#pragma_auto_vacuum">auto-vacuum mode</a>
and a very small <a href="capi3ref.html#sqlite3_soft_heap_limit">
soft_heap_limit</a> is set. 
<a href="http://www.sqlite.org/cvstrac/tktview?tn=2565">Ticket #2565</a>.

#
# Run this Tcl script to generate the formatchng.html file.
#
set rcsid {$Id: formatchng.tcl,v 1.20 2007/09/03 20:32:45 drh Exp $ }
source common.tcl
header {File Format Changes in SQLite}
puts {
<h2>File Format Changes in SQLite</h2>

<p>
Every effort is made to keep SQLite fully backwards compatible from

  is used by default instead of the new.  This might change
  again in some future release - we may go back to generating
  the new file format by default - but probably not until
  all users have upgraded to a version of SQLite that will
  understand the new file format.  That might take several
  years.</p></td>
</tr>
<tr>
  <td valign="top">3.4.2 to 3.5.0</td>
  <td valign="top">2007-Sep-3</td>
  <td><p>The design of the OS interface layer was changed for
  release 3.5.0.  Applications that implemented a custom OS
  interface will need to be modified in order to upgrade.
  There are also some subtly different semantics a few obscure
  APIs.  An <a href="34to35.html">article</a> is avilable which
  describing the changes in detail.</p>

  <p>The on-disk file format is unchanged.</p>
  </td>
</tr>
</table>
</blockquote>

<p>
To perform a database reload, have ready versions of the
<b>sqlite</b> command-line utility for both the old and new
version of SQLite.  Call these two executables "<b>sqlite-old</b>"

proc newsitem {date title text} {
  puts "<h3>$date - $title</h3>"
  regsub -all "\n( *\n)+" $text "</p>\n\n<p>" txt
  puts "<p>$txt</p>"
  puts "<hr width=\"50%\">"
}

newsitem {2007-Sep-3} {Version 3.5.0 beta} {
  The OS interface layer and the memory allocation subsystems in
  SQLite have been reimplemented.  The published API is largely unchanged
  but the (unpublished) OS interface has been modified extensively.  
  Application that implement their own OS interface will require
  modification.  See
  <a href="34to35.html">34to35.html</a> for details.<p>

  This is a large change approximately 1 line of count of out 10 was
  modified.  We are calling this first release "beta" in order to give
  the user community time to test and evaluate the changes before we
  freeze the new design.
}

newsitem {2007-Aug-13} {Version 3.4.2} {
  While stress-testing the 
  <a href="capi3ref.html#sqlite3_soft_heap_limit">soft_heap_limit</a>
  feature, a bug that could lead to
  <a href="http://www.sqlite.org/cvstrac/wiki?p=DatabaseCorruption">database
  corruption</a> was <a href="http://www.sqlite.org/cvstrac/tktview?tn=2565">

  for additional information.
}

puts {
<p align="right"><a href="oldnews.html">Old news...</a></p>
</td></tr></table>
}
footer {$Id: index.tcl,v 1.160 2007/09/03 20:32:46 drh Exp $}