Documentation Source Text

<p>The CREATE VIRTUAL TABLE statement creates a new table
called <yyterm>table-name</yyterm> derived from the class
class <yyterm>module-name</yyterm>.  The <yyterm>module-name</yyterm>
is the name that is registered for the virtual table by
the [sqlite3_create_module()] interface.

<blockquote><pre>
   CREATE VIRTUAL TABLE tablename USING modulename;
</pre></blockquote>

<p>One can also provide comma-separated arguments to the module following 
the module name:

<blockquote><pre>
   CREATE VIRTUAL TABLE tablename USING modulename(arg1, arg2, ...);
</pre></blockquote>

<p>The format of the arguments to the module is very general. Each 
<yyterm>module-argument</yyterm>
may contain keywords, string literals, identifiers, numbers, and 
punctuation. Each <yyterm>module-argument</yyterm> is passed as 
written (as text) into the
[xCreate | constructor method] of the virtual table implementation 

<h3>Temporary virtual tables</h3>

<p>There is no "CREATE TEMP VIRTUAL TABLE" statement.  To create a
temporary virtual table, add the "temp" schema
before the virtual table name.

<blockquote><pre>
   CREATE VIRTUAL TABLE <b>temp.</b>tablename USING module(arg1, ...);
</pre></blockquote>

<tcl>hd_fragment epovtab {eponymous virtual tables}  \
        {eponymous virtual table}</tcl>
<h3>Eponymous virtual tables</h3>

<p>Some virtual tables exist automatically in the "main" schema of
every database connection in which their

<p>An example of an eponymous virtual table is the [dbstat virtual table].
To use the dbstat virtual table as an eponymous virtual table, 
simply query against the "dbstat"
module name, as if it were an ordinary table.  (Note that SQLite
must be compiled with the [SQLITE_ENABLE_DBSTAT_VTAB] option to include
the dbstat virtual table in the build.)

<blockquote><pre>
   SELECT * FROM dbstat;
</pre></blockquote>

<p>A virtual table is eponymous if its [xCreate] method is the exact same
function as the [xConnect] method, or if the [xCreate] method is NULL.
The [xCreate] method is called when a virtual table is first created
using the [CREATE VIRTUAL TABLE] statement.  The [xConnect] method whenever
a database connection attaches to or reparses a schema. When these two methods
are the same, that indicates that the virtual table has no persistent

command is attempted against that virtual table module, a jump to a NULL
pointer will occur, resulting in a crash.

<h2>Implementation</h2>

<p>Several new C-level objects are used by the virtual table implementation:

<blockquote><pre>
  typedef struct sqlite3_vtab sqlite3_vtab;
  typedef struct sqlite3_index_info sqlite3_index_info;
  typedef struct sqlite3_vtab_cursor sqlite3_vtab_cursor;
  typedef struct sqlite3_module sqlite3_module;
</pre></blockquote>

<p>The [sqlite3_module] structure defines a module object used to implement
a virtual table. Think of a module as a class from which one can 
construct multiple virtual tables having similar properties. For example,
one might have a module that provides read-only access to 
comma-separated-value (CSV) files on disk. That one module can then be
used to create several virtual tables where each virtual table refers
to a different CSV file.

<p>The module structure contains methods that are invoked by SQLite to
perform various actions on the virtual table such as creating new
instances of a virtual table or destroying old ones, reading and
writing data, searching for and deleting, updating, or inserting rows. 
The module structure is explained in more detail below.

<p>Each virtual table instance is represented by an [sqlite3_vtab] structure. 
The sqlite3_vtab structure looks like this:

<blockquote><pre>
  struct sqlite3_vtab {
    const sqlite3_module *pModule;
    int nRef;
    char *zErrMsg;
  };
</pre></blockquote>

<p>Virtual table implementations will normally subclass this structure 
to add additional private and implementation-specific fields. 
The nRef field is used internally by the SQLite core and should not 
be altered by the virtual table implementation. The virtual table 
implementation may pass error message text to the core by putting 
an error message string in zErrMsg.

it destroys the virtual table. The virtual table implementation only 
needs to worry about freeing the zErrMsg content when it overwrites 
the content with a new, different error message.

<p>The [sqlite3_vtab_cursor] structure represents a pointer to a specific
row of a virtual table. This is what an sqlite3_vtab_cursor looks like:

<blockquote><pre>
  struct sqlite3_vtab_cursor {
    sqlite3_vtab *pVtab;
  };
</pre></blockquote>

<p>Once again, practical implementations will likely subclass this 
structure to add additional private fields.

<p>The [sqlite3_index_info] structure is used to pass information into
and out of the xBestIndex method of the module that implements a 
virtual table.

<p>Before a [CREATE VIRTUAL TABLE] statement can be run, the module 
specified in that statement must be registered with the database 
connection. This is accomplished using either of the [sqlite3_create_module()]
or [sqlite3_create_module_v2()] interfaces:

<blockquote><pre>
  int sqlite3_create_module(
    sqlite3 *db,               /* SQLite connection to register module with */
    const char *zName,         /* Name of the module */
    const sqlite3_module *,    /* Methods for the module */
    void *                     /* Client data for xCreate/xConnect */
  );
  int sqlite3_create_module_v2(
    sqlite3 *db,               /* SQLite connection to register module with */
    const char *zName,         /* Name of the module */
    const sqlite3_module *,    /* Methods for the module */
    void *,                    /* Client data for xCreate/xConnect */
    void(*xDestroy)(void*)     /* Client data destructor function */
  );
</pre></blockquote>

<p>The [sqlite3_create_module()] and [sqlite3_create_module_v2()]
routines associates a module name with 
an [sqlite3_module] structure and a separate client data that is specific 
to each module.  The only difference between the two create_module methods
is that the _v2 method includes an extra parameter that specifies a
destructor for client data pointer.  The module structure is what defines
the behavior of a virtual table.  The module structure looks like this:

<blockquote><pre>  
  struct sqlite3_module {
    int iVersion;
    int (*xCreate)(sqlite3*, void *pAux,
                 int argc, char **argv,
                 sqlite3_vtab **ppVTab,
                 char **pzErr);
    int (*xConnect)(sqlite3*, void *pAux,

    int (*Rename)(sqlite3_vtab *pVtab, const char *zNew);
    /* The methods above are in version 1 of the sqlite_module object. Those 
    ** below are for version 2 and greater. */
    int (*xSavepoint)(sqlite3_vtab *pVTab, int);
    int (*xRelease)(sqlite3_vtab *pVTab, int);
    int (*xRollbackTo)(sqlite3_vtab *pVTab, int);
  };
</pre></blockquote>

<p>The module structure defines all of the methods for each virtual 
table object. The module structure also contains the iVersion field which
defines the particular edition of the module table structure. Currently, 
iVersion is always 1, but in future releases of SQLite the module structure 
definition might be extended with additional methods and in that case 
the iVersion value will be increased.

[sqlite3_load_extension | loadable extension].

<h1>Virtual Table Methods</h1>

<tcl>hd_fragment xcreate {sqlite3_module.xCreate} {xCreate}</tcl>
<h2>The xCreate Method</h2>

<blockquote><pre>
  int (*xCreate)(sqlite3 *db, void *pAux,
               int argc, char **argv,
               sqlite3_vtab **ppVTab,
               char **pzErr);
</pre></blockquote>

<p>The xCreate method is called to create a new instance of a virtual table 
in response to a [CREATE VIRTUAL TABLE] statement.
If the xCreate method is the same pointer as the [xConnect] method, then the
virtual table is an [eponymous virtual table].
If the xCreate method is omitted (if it is a NULL pointer) then the virtual 
table is an [eponymous-only virtual table].

(an [sqlite3_vtab] object) and return a pointer to it in *ppVTab.

<p>As part of the task of creating a new [sqlite3_vtab] structure, this 
method <u>must</u> invoke [sqlite3_declare_vtab()] to tell the SQLite 
core about the columns and datatypes in the virtual table. 
The [sqlite3_declare_vtab()] API has the following prototype:

<blockquote><pre>
    int sqlite3_declare_vtab(sqlite3 *db, const char *zCreateTable)
</pre></blockquote>

<p>The first argument to [sqlite3_declare_vtab()] must be the same 
[database connection] pointer as the first parameter to this method.
The second argument to [sqlite3_declare_vtab()] must a zero-terminated 
UTF-8 string that contains a well-formed [CREATE TABLE] statement that 
defines the columns in the virtual table and their data types. 
The name of the table in this CREATE TABLE statement is ignored, 

     expression in the result set of a [SELECT], and
<li> Hidden columns are not included in the implicit column-list 
     used by an [INSERT] statement that lacks an explicit column-list. 
</ul>

<p>For example, if the following SQL is passed to [sqlite3_declare_vtab()]:

<blockquote><pre>
   CREATE TABLE x(a HIDDEN VARCHAR(12), b INTEGER, c INTEGER Hidden);
</pre></blockquote>

<p>Then the virtual table would be created with two hidden columns,
and with datatypes of "VARCHAR(12)" and "INTEGER".

<p>An example use of hidden columns can be seen in the [FTS3] virtual 
table implementation, where every FTS virtual table
contains an [FTS hidden column] that is used to pass information from the

the HIDDEN columns of the virtual table.

<p>For example, the "generate_series" extension (located in the
[http://www.sqlite.org/src/artifact?ci=trunk&filename=ext/misc/series.c|ext/misc/series.c]
file in the [http://www.sqlite.org/src/tree?ci=trunk|source tree])
implements an [eponymous virtual table] with the following schema:

<blockquote><pre>
CREATE TABLE generate_series(
  value,
  start HIDDEN,
  stop HIDDEN,
  step HIDDEN
);
</pre></blockquote>

<p>The [sqlite3_module.xBestIndex] method in the implementation of this
table checks for equality constraints against the HIDDEN columns, and uses
those as input parameters to determine the range of integer "value" outputs
to generate.  Reasonable defaults are used for any unconstrained columns.
For example, to list all integers between 5 and 50:

<blockquote><pre>
SELECT value FROM generate_series(5,50);
</pre></blockquote>

<p>The previous query is equivalent to the following:

<blockquote><pre>
SELECT value FROM generate_series WHERE start=5 AND stop=50;
</pre></blockquote>

<p>Arguments on the virtual table name are matched to [hidden columns]
in order.  The number of arguments can be less than the
number of hidden columns, in which case the latter hidden columns are
unconstrained.  However, an error results if there are more arguments
than there are hidden columns in the virtual table.

WITHOUT ROWID virtual tables must be read-only.


<tcl>############################################################# xConnect
hd_fragment xconnect {sqlite3_module.xConnect} {xConnect}</tcl>
<h2>The xConnect Method</h2>

<blockquote><pre>
  int (*xConnect)(sqlite3*, void *pAux,
               int argc, char **argv,
               sqlite3_vtab **ppVTab,
               char **pzErr);
</pre></blockquote>

<p>The xConnect method is very similar to [xCreate]. 
It has the same parameters and constructs a new [sqlite3_vtab] structure 
just like xCreate. 
And it must also call [sqlite3_declare_vtab()] like xCreate.

<p>The difference is that xConnect is called to establish a new 

hd_fragment xbestindex {sqlite3_module.xBestIndex} {xBestIndex}</tcl>
<h2>The xBestIndex Method</h2>

<p>SQLite uses the xBestIndex method of a virtual table module to determine
the best way to access the virtual table. 
The xBestIndex method has a prototype like this:

<blockquote><pre>
  int (*xBestIndex)(sqlite3_vtab *pVTab, sqlite3_index_info*);
</pre></blockquote>

<p>The SQLite core communicates with the xBestIndex method by filling 
in certain fields of the [sqlite3_index_info] structure and passing a 
pointer to that structure into xBestIndex as the second parameter. 
The xBestIndex method fills out other fields of this structure which
forms the reply. The [sqlite3_index_info] structure looks like this:

<blockquote><pre>
  struct sqlite3_index_info {
    /* Inputs */
    const int nConstraint;     /* Number of entries in aConstraint */
    const struct sqlite3_index_constraint {
       int iColumn;              /* Column constrained.  -1 for ROWID */
       unsigned char op;         /* Constraint operator */
       unsigned char usable;     /* True if this constraint is usable */

    <b>/* Fields below are only available in SQLite 3.8.2 and later */</b>
    sqlite3_int64 estimatedRows;    /* Estimated number of rows returned */
    <b>/* Fields below are only available in SQLite 3.9.0 and later */</b>
    int idxFlags;              /* Mask of SQLITE_INDEX_SCAN_* flags */
    <b>/* Fields below are only available in SQLite 3.10.0 and later */</b>
    sqlite3_uint64 colUsed;    /* Input: Mask of columns used by statement */
  };
</pre></blockquote>

<p>Note the warnings on the "estimatedRows", "idxFlags", and colUsed fields.
These fields were added with SQLite versions 3.8.2, 3.9.0, and 3.10.0, respectively. 
Any extension that reads or writes these fields must first check that the 
version of the SQLite library in use is greater than or equal to appropriate
version - perhaps comparing the value returned from [sqlite3_libversion_number()]
against constants 3008002, 3009000, and/or 3010000. The result of attempting 
to access these fields in an sqlite3_index_info structure created by an 
older version of SQLite are undefined.

<p>In addition, there are some defined constants:

<blockquote><pre>
  #define SQLITE_INDEX_CONSTRAINT_EQ      2
  #define SQLITE_INDEX_CONSTRAINT_GT      4
  #define SQLITE_INDEX_CONSTRAINT_LE      8
  #define SQLITE_INDEX_CONSTRAINT_LT     16
  #define SQLITE_INDEX_CONSTRAINT_GE     32
  #define SQLITE_INDEX_CONSTRAINT_MATCH  64
  #define SQLITE_INDEX_CONSTRAINT_LIKE   65     /* 3.10.0 and later only */
  #define SQLITE_INDEX_CONSTRAINT_GLOB   66     /* 3.10.0 and later only */
  #define SQLITE_INDEX_CONSTRAINT_REGEXP 67     /* 3.10.0 and later only */
  #define SQLITE_INDEX_SCAN_UNIQUE        1     /* Scan visits at most 1 row */
</pre></blockquote>

<p>The SQLite core calls the xBestIndex method when it is compiling a query
that involves a virtual table. In other words, SQLite calls this method 
when it is running [sqlite3_prepare()] or the equivalent. 
By calling this method, the 
SQLite core is saying to the virtual table that it needs to access 
some subset of the rows in the virtual table and it wants to know the

     column  OP  EXPR
</blockquote>

<p>Where "column" is a column in the virtual table, OP is an operator 
like "=" or "&lt;", and EXPR is an arbitrary expression. So, for example,
if the WHERE clause contained a term like this:

<blockquote><pre>
     a = 5
</pre></blockquote>

<p>Then one of the constraints would be on the "a" column with 
operator "=" and an expression of "5". Constraints need not have a
literal representation of the WHERE clause. The query optimizer might
make transformations to the 
WHERE clause in order to extract as many constraints 
as it can. So, for example, if the WHERE clause contained something 
like this:

<blockquote><pre>
     x BETWEEN 10 AND 100 AND 999&gt;y
</pre></blockquote>

<p>The query optimizer might translate this into three separate constraints:

<blockquote><pre>
     x &gt;= 10
     x &lt;= 100
     y &lt; 999
</pre></blockquote>

<p>For each constraint, the aConstraint[].iColumn field indicates which 
column appears on the left-hand side of the constraint.
The first column of the virtual table is column 0. 
The rowid of the virtual table is column -1. 
The aConstraint[].op field indicates which operator is used. 
The SQLITE_INDEX_CONSTRAINT_* constants map integer constants 

each row of the virtual table that it receives. If such a check 
is redundant, the xBestFilter method can suppress that double-check by 
setting aConstraintUsage[].omit.

<tcl>hd_fragment xdisconnect {sqlite3_module.xDisconnect} {xDisconnect}</tcl>
<h2>The xDisconnect Method</h2>

<blockquote><pre>
  int (*xDisconnect)(sqlite3_vtab *pVTab);
</pre></blockquote>

<p>This method releases a connection to a virtual table. 
Only the [sqlite3_vtab] object is destroyed.
The virtual table is not destroyed and any backing store 
associated with the virtual table persists. 

This method undoes the work of [xConnect].

<p>This method is a destructor for a connection to the virtual table.
Contrast this method with [xDestroy].  The xDestroy is a destructor
for the entire virtual table.

<p>The xDisconnect method is required for every virtual table implementation,
though it is acceptable for the xDisconnect and [xDestroy] methods to be
the same function if that makes sense for the particular virtual table.

<tcl>########################################################## xDestroy
hd_fragment {sqlite3_module.xDestroy} {xDestroy}</tcl>
<h2>The xDestroy Method</h2>

<blockquote><pre>
  int (*xDestroy)(sqlite3_vtab *pVTab);
</pre></blockquote>

<p>This method releases a connection to a virtual table, just like 
the [xDisconnect] method, and it also destroys the underlying 
table implementation. This method undoes the work of [xCreate].

<p>The [xDisconnect] method is called whenever a database connection
that uses a virtual table is closed. The xDestroy method is only 
called when a [DROP TABLE] statement is executed against the virtual table.

<p>The xDestroy method is required for every virtual table implementation,
though it is acceptable for the [xDisconnect] and xDestroy methods to be
the same function if that makes sense for the particular virtual table.

<tcl>########################################################## xOpen
hd_fragment xopen {sqlite3_module.xOpen}</tcl>
<h2>The xOpen Method</h2>

<blockquote><pre>
  int (*xOpen)(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor);
</pre></blockquote>

<p>The xOpen method creates a new cursor used for accessing (read and/or
writing) a virtual table.  A successful invocation of this method 
will allocate the memory for the [sqlite3_vtab_cursor] (or a subclass),
initialize the new object, and make *ppCursor point to the new object.
The successful call then returns [SQLITE_OK].

<p>The xOpen method is required for every virtual table implementation.

<tcl>############################################################### xClose
hd_fragment xclose {sqlite3_module.xClose}</tcl>
<h2>The xClose Method</h2>

<blockquote><pre>
  int (*xClose)(sqlite3_vtab_cursor*);
</pre></blockquote>

<p>The xClose method closes a cursor previously opened by 
[sqlite3_module.xOpen | xOpen]. 
The SQLite core will always call xClose once for each cursor opened 
using xOpen.

<p>This method must release all resources allocated by the
corresponding xOpen call. The routine will not be called again even if it
returns an error.  The SQLite core will not use the
[sqlite3_vtab_cursor] again after it has been closed.

<p>The xClose method is required for every virtual table implementation.

<tcl>############################################################## xEof
hd_fragment xeof {sqlite3_module.xEof} {xEof}</tcl>
<h2>The xEof Method</h2>

<blockquote><pre>
  int (*xEof)(sqlite3_vtab_cursor*);
</pre></blockquote>

<p>The xEof method must return false (zero) if the specified cursor 
currently points to a valid row of data, or true (non-zero) otherwise. 
This method is called by the SQL engine immediately after each 
[xFilter] and [xNext] invocation.

<p>The xEof method is required for every virtual table implementation.

<tcl>############################################################## xFilter
hd_fragment xfilter {sqlite3_module.xFilter} {xFilter}</tcl>
<h2>The xFilter Method</h2>

<blockquote><pre>
  int (*xFilter)(sqlite3_vtab_cursor*, int idxNum, const char *idxStr,
                int argc, sqlite3_value **argv);
</pre></blockquote>

<p>This method begins a search of a virtual table. 
The first argument is a cursor opened by [sqlite3_module.xOpen | xOpen]. 
The next two arguments define a particular search index previously 
chosen by [xBestIndex]. The specific meanings of idxNum and idxStr 
are unimportant as long as xFilter and xBestIndex agree on what 
that meaning is.

<p>The xFilter method is required for every virtual table implementation.

<tcl>############################################################### xNext
hd_fragment xnext {sqlite3_module.xNext} {xNext}</tcl>
<h2>The xNext Method</h2>

<blockquote><pre>
  int (*xNext)(sqlite3_vtab_cursor*);
</pre></blockquote>

<p>The xNext method advances a [sqlite3_vtab_cursor | virtual table cursor]
to the next row of a result set initiated by [xFilter]. 
If the cursor is already pointing at the last row when this 
routine is called, then the cursor no longer points to valid 
data and a subsequent call to the [xEof] method must return true (non-zero). 
If the cursor is successfully advanced to another row of content, then
subsequent calls to [xEof] must return false (zero).

<p>This method must return [SQLITE_OK] if successful, or an sqlite 
[error code] if an error occurs.

<p>The xNext method is required for every virtual table implementation.

<tcl>############################################################## xColumn
hd_fragment xcolumn {sqlite3_module.xColumn} {xColumn}</tcl>
<h2>The xColumn Method</h2>

<blockquote><pre>
  int (*xColumn)(sqlite3_vtab_cursor*, sqlite3_context*, int N);
</pre></blockquote>

<p>The SQLite core invokes this method in order to find the value for 
the N-th column of the current row. N is zero-based so the first column 
is numbered 0. 
The xColumn method may return its result back to SQLite using one of the
following interface:

<p>The xColumn method is required for every virtual table implementation.

<tcl>############################################################# xRowid
hd_fragment xrowid {sqlite3_module.xRowid} {xRowid}</tcl>
<h2>The xRowid Method</h2>

<blockquote><pre>
  int (*xRowid)(sqlite3_vtab_cursor *pCur, sqlite_int64 *pRowid);
</pre></blockquote>

<p>A successful invocation of this method will cause *pRowid to be
filled with the [rowid] of row that the
[sqlite3_vtab_cursor | virtual table cursor] pCur is currently pointing at.
This method returns [SQLITE_OK] on success.
It returns an appropriate [error code] on failure.</p>

<p>The xRowid method is required for every virtual table implementation.

<tcl>############################################################# xUpdate
hd_fragment xupdate {sqlite3_module.xUpdate} {xUpdate}</tcl>
<h2>The xUpdate Method</h2>

<blockquote><pre>
  int (*xUpdate)(
    sqlite3_vtab *pVTab,
    int argc,
    sqlite3_value **argv,
    sqlite_int64 *pRowid
  );
</pre></blockquote>

<p>All changes to a virtual table are made using the xUpdate method.
This one method can be used to insert, delete, or update.

<p>The argc parameter specifies the number of entries in the argv array. 
The value of argc will be 1 for a pure delete operation or N+2 for an insert
or replace or update where N is the number of columns in the table.  

is a NULL pointer, then the virtual table is read-only.


<tcl>########################################################## xFindFunction
hd_fragment xfindfunction {sqlite3_module.xFindFunction} {xFindFunction}</tcl>
<h2>The xFindFunction Method</h2>

<blockquote><pre>
  int (*xFindFunction)(
    sqlite3_vtab *pVtab,
    int nArg,
    const char *zName,
    void (**pxFunc)(sqlite3_context*,int,sqlite3_value**),
    void **ppArg
  );
</pre></blockquote>

<p>This method is called during [sqlite3_prepare()] to give the virtual
table implementation an opportunity to overload functions. 
This method may be set to NULL in which case no overloading occurs.

<p>When a function uses a column from a virtual table as its first 
argument, this method is called to see if the virtual table would 

<p>The function pointer returned by this routine must be valid for
the lifetime of the [sqlite3_vtab] object given in the first parameter.

<tcl>############################################################ xBegin
hd_fragment xBegin {sqlite3_module.xBegin} {xBegin}</tcl>
<h2>The xBegin Method</h2>

<blockquote><pre>
  int (*xBegin)(sqlite3_vtab *pVTab);
</pre></blockquote>

<p>This method begins a transaction on a virtual table.
This is method is optional.  The xBegin pointer of [sqlite3_module]
may be NULL.

<p>This method is always followed by one call to either the
[xCommit] or [xRollback] method.  Virtual table transactions do
not nest, so the xBegin method will not be invoked more than once
on a single virtual table
without an intervening call to either [xCommit] or [xRollback].
Multiple calls to other methods can and likely will occur in between
the xBegin and the corresponding [xCommit] or [xRollback].

<tcl>############################################################ xSync
hd_fragment xsync {sqlite3_module.xSync}</tcl>
<h2>The xSync Method</h2>

<blockquote><pre>
  int (*xSync)(sqlite3_vtab *pVTab);
</pre></blockquote>


<p>This method signals the start of a two-phase commit on a virtual
table.
This is method is optional.  The xSync pointer of [sqlite3_module]
may be NULL.

<p>This method is only invoked after call to the [xBegin] method and
prior to an [xCommit] or [xRollback].  In order to implement two-phase
commit, the xSync method on all virtual tables is invoked prior to
invoking the [xCommit] method on any virtual table.  If any of the 
xSync methods fail, the entire transaction is rolled back.

<tcl>########################################################### xCommit
hd_fragment xcommit {sqlite3_module.xCommit} {xCommit}</tcl>
<h2>The xCommit Method</h2>

<blockquote><pre>
  int (*xCommit)(sqlite3_vtab *pVTab);
</pre></blockquote>

<p>This method causes a virtual table transaction to commit.
This is method is optional.  The xCommit pointer of [sqlite3_module]
may be NULL.

<p>A call to this method always follows a prior call to [xBegin] and
[sqlite3_module.xSync|xSync].


<tcl>############################################################## xRollback
hd_fragment xrollback {sqlite3_module.xRollback} {xRollback}</tcl>
<h2>The xRollback Method</h2>

<blockquote><pre>
  int (*xRollback)(sqlite3_vtab *pVTab);
</pre></blockquote>

<p>This method causes a virtual table transaction to rollback.
This is method is optional.  The xRollback pointer of [sqlite3_module]
may be NULL.

<p>A call to this method always follows a prior call to [xBegin].


<tcl>############################################################# xRename
hd_fragment xrename {sqlite3_module.xRename} {xRename}</tcl>
<h2>The xRename Method</h2>

<blockquote><pre>
  int (*xRename)(sqlite3_vtab *pVtab, const char *zNew);
</pre></blockquote>

<p>This method provides notification that the virtual table implementation
that the virtual table will be given a new name. 
If this method returns [SQLITE_OK] then SQLite renames the table.
If this method returns an [error code] then the renaming is prevented.

<p>The xRename method is required for every virtual table implementation.

<tcl>############################################################# xSavepoint
hd_fragment xsavepoint {sqlite3_module.xSavepoint} {xSavepoint}\
  xRelease xRollbackTo</tcl>
<h2>The xSavepoint, xRelease, and xRollbackTo Methods</h2>

<blockquote><pre>
  int (*xSavepoint)(sqlite3_vtab *pVtab, int);
  int (*xRelease)(sqlite3_vtab *pVtab, int);
  int (*xRollbackTo)(sqlite3_vtab *pVtab, int);
</pre></blockquote>

<p>
These methods provide the virtual table implementation an opportunity to
implement nested transactions.  They are always optional and will only be
called in SQLite [version 3.7.7] and later.
</p>

  
  # All inline tags.
  variable aInline
  foreach x {
    tt i b big small u
    em strong dfn code samp kbd var cite abbr acronym
    a img object br script map q sub sup span bdo
    input select textarea label button tcl
  } { set aInline($x) 1 }

  variable aContentChecker
  set aContentChecker(p)        HtmlInlineContent
  set aContentChecker(th)       HtmlTableCellContent
  set aContentChecker(td)       HtmlTableCellContent
  set aContentChecker(tr)       HtmlTableRowContent