summaryrefslogtreecommitdiff
path: root/ext/sqlite/libsqlite/src/main.c
diff options
context:
space:
mode:
Diffstat (limited to 'ext/sqlite/libsqlite/src/main.c')
-rw-r--r--ext/sqlite/libsqlite/src/main.c989
1 files changed, 989 insertions, 0 deletions
diff --git a/ext/sqlite/libsqlite/src/main.c b/ext/sqlite/libsqlite/src/main.c
new file mode 100644
index 0000000000..b708b00c2a
--- /dev/null
+++ b/ext/sqlite/libsqlite/src/main.c
@@ -0,0 +1,989 @@
+/*
+** 2001 September 15
+**
+** 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.
+**
+*************************************************************************
+** Main file for the SQLite library. The routines in this file
+** implement the programmer interface to the library. Routines in
+** other files are for internal use by SQLite and should not be
+** accessed by users of the library.
+**
+** $Id$
+*/
+#include "sqliteInt.h"
+#include "os.h"
+#include <ctype.h>
+
+/*
+** A pointer to this structure is used to communicate information
+** from sqliteInit into the sqliteInitCallback.
+*/
+typedef struct {
+ sqlite *db; /* The database being initialized */
+ char **pzErrMsg; /* Error message stored here */
+} InitData;
+
+
+/*
+** This is the callback routine for the code that initializes the
+** database. See sqliteInit() below for additional information.
+**
+** Each callback contains the following information:
+**
+** argv[0] = "file-format" or "schema-cookie" or "table" or "index"
+** argv[1] = table or index name or meta statement type.
+** argv[2] = root page number for table or index. NULL for meta.
+** argv[3] = SQL text for a CREATE TABLE or CREATE INDEX statement.
+** argv[4] = "1" for temporary files, "0" for main database
+**
+*/
+static
+int sqliteInitCallback(void *pInit, int argc, char **argv, char **azColName){
+ InitData *pData = (InitData*)pInit;
+ Parse sParse;
+ int nErr = 0;
+
+ /* TODO: Do some validity checks on all fields. In particular,
+ ** make sure fields do not contain NULLs. Otherwise we might core
+ ** when attempting to initialize from a corrupt database file. */
+
+ assert( argc==5 );
+ switch( argv[0][0] ){
+ case 'v':
+ case 'i':
+ case 't': { /* CREATE TABLE, CREATE INDEX, or CREATE VIEW statements */
+ if( argv[3] && argv[3][0] ){
+ /* Call the parser to process a CREATE TABLE, INDEX or VIEW.
+ ** But because sParse.initFlag is set to 1, no VDBE code is generated
+ ** or executed. All the parser does is build the internal data
+ ** structures that describe the table, index, or view.
+ */
+ memset(&sParse, 0, sizeof(sParse));
+ sParse.db = pData->db;
+ sParse.initFlag = 1;
+ sParse.isTemp = argv[4][0] - '0';
+ sParse.newTnum = atoi(argv[2]);
+ sParse.useCallback = 1;
+ sqliteRunParser(&sParse, argv[3], pData->pzErrMsg);
+ }else{
+ /* If the SQL column is blank it means this is an index that
+ ** was created to be the PRIMARY KEY or to fulfill a UNIQUE
+ ** constraint for a CREATE TABLE. The index should have already
+ ** been created when we processed the CREATE TABLE. All we have
+ ** to do here is record the root page number for that index.
+ */
+ Index *pIndex = sqliteFindIndex(pData->db, argv[1]);
+ if( pIndex==0 || pIndex->tnum!=0 ){
+ /* This can occur if there exists an index on a TEMP table which
+ ** has the same name as another index on a permanent index. Since
+ ** the permanent table is hidden by the TEMP table, we can also
+ ** safely ignore the index on the permanent table.
+ */
+ /* Do Nothing */;
+ }else{
+ pIndex->tnum = atoi(argv[2]);
+ }
+ }
+ break;
+ }
+ default: {
+ /* This can not happen! */
+ nErr = 1;
+ assert( nErr==0 );
+ }
+ }
+ return nErr;
+}
+
+/*
+** This is a callback procedure used to reconstruct a table. The
+** name of the table to be reconstructed is passed in as argv[0].
+**
+** This routine is used to automatically upgrade a database from
+** format version 1 or 2 to version 3. The correct operation of
+** this routine relys on the fact that no indices are used when
+** copying a table out to a temporary file.
+*/
+static
+int upgrade_3_callback(void *pInit, int argc, char **argv, char **NotUsed){
+ InitData *pData = (InitData*)pInit;
+ int rc;
+ Table *pTab;
+ Trigger *pTrig;
+ char *zErr = 0;
+
+ pTab = sqliteFindTable(pData->db, argv[0]);
+ assert( pTab!=0 );
+ assert( sqliteStrICmp(pTab->zName, argv[0])==0 );
+ if( pTab ){
+ pTrig = pTab->pTrigger;
+ pTab->pTrigger = 0; /* Disable all triggers before rebuilding the table */
+ }
+ rc = sqlite_exec_printf(pData->db,
+ "CREATE TEMP TABLE sqlite_x AS SELECT * FROM '%q'; "
+ "DELETE FROM '%q'; "
+ "INSERT INTO '%q' SELECT * FROM sqlite_x; "
+ "DROP TABLE sqlite_x;",
+ 0, 0, &zErr, argv[0], argv[0], argv[0]);
+ if( zErr ){
+ sqliteSetString(pData->pzErrMsg, zErr, 0);
+ sqlite_freemem(zErr);
+ }
+
+ /* If an error occurred in the SQL above, then the transaction will
+ ** rollback which will delete the internal symbol tables. This will
+ ** cause the structure that pTab points to be deleted. In case that
+ ** happened, we need to refetch pTab.
+ */
+ pTab = sqliteFindTable(pData->db, argv[0]);
+ if( pTab ){
+ assert( sqliteStrICmp(pTab->zName, argv[0])==0 );
+ pTab->pTrigger = pTrig; /* Re-enable triggers */
+ }
+ return rc!=SQLITE_OK;
+}
+
+
+
+/*
+** Attempt to read the database schema and initialize internal
+** data structures. Return one of the SQLITE_ error codes to
+** indicate success or failure.
+**
+** After the database is initialized, the SQLITE_Initialized
+** bit is set in the flags field of the sqlite structure. An
+** attempt is made to initialize the database as soon as it
+** is opened. If that fails (perhaps because another process
+** has the sqlite_master table locked) than another attempt
+** is made the first time the database is accessed.
+*/
+int sqliteInit(sqlite *db, char **pzErrMsg){
+ int rc;
+ BtCursor *curMain;
+ int size;
+ Table *pTab;
+ char *azArg[6];
+ int meta[SQLITE_N_BTREE_META];
+ Parse sParse;
+ InitData initData;
+
+ /*
+ ** The master database table has a structure like this
+ */
+ static char master_schema[] =
+ "CREATE TABLE sqlite_master(\n"
+ " type text,\n"
+ " name text,\n"
+ " tbl_name text,\n"
+ " rootpage integer,\n"
+ " sql text\n"
+ ")"
+ ;
+ static char temp_master_schema[] =
+ "CREATE TEMP TABLE sqlite_temp_master(\n"
+ " type text,\n"
+ " name text,\n"
+ " tbl_name text,\n"
+ " rootpage integer,\n"
+ " sql text\n"
+ ")"
+ ;
+
+ /* The following SQL will read the schema from the master tables.
+ ** The first version works with SQLite file formats 2 or greater.
+ ** The second version is for format 1 files.
+ **
+ ** Beginning with file format 2, the rowid for new table entries
+ ** (including entries in sqlite_master) is an increasing integer.
+ ** So for file format 2 and later, we can play back sqlite_master
+ ** and all the CREATE statements will appear in the right order.
+ ** But with file format 1, table entries were random and so we
+ ** have to make sure the CREATE TABLEs occur before their corresponding
+ ** CREATE INDEXs. (We don't have to deal with CREATE VIEW or
+ ** CREATE TRIGGER in file format 1 because those constructs did
+ ** not exist then.)
+ */
+ static char init_script[] =
+ "SELECT type, name, rootpage, sql, 1 FROM sqlite_temp_master "
+ "UNION ALL "
+ "SELECT type, name, rootpage, sql, 0 FROM sqlite_master";
+ static char older_init_script[] =
+ "SELECT type, name, rootpage, sql, 1 FROM sqlite_temp_master "
+ "UNION ALL "
+ "SELECT type, name, rootpage, sql, 0 FROM sqlite_master "
+ "WHERE type='table' "
+ "UNION ALL "
+ "SELECT type, name, rootpage, sql, 0 FROM sqlite_master "
+ "WHERE type='index'";
+
+
+ /* Construct the schema tables: sqlite_master and sqlite_temp_master
+ */
+ azArg[0] = "table";
+ azArg[1] = MASTER_NAME;
+ azArg[2] = "2";
+ azArg[3] = master_schema;
+ azArg[4] = "0";
+ azArg[5] = 0;
+ initData.db = db;
+ initData.pzErrMsg = pzErrMsg;
+ sqliteInitCallback(&initData, 5, azArg, 0);
+ pTab = sqliteFindTable(db, MASTER_NAME);
+ if( pTab ){
+ pTab->readOnly = 1;
+ }
+ azArg[1] = TEMP_MASTER_NAME;
+ azArg[3] = temp_master_schema;
+ azArg[4] = "1";
+ sqliteInitCallback(&initData, 5, azArg, 0);
+ pTab = sqliteFindTable(db, TEMP_MASTER_NAME);
+ if( pTab ){
+ pTab->readOnly = 1;
+ }
+
+ /* Create a cursor to hold the database open
+ */
+ if( db->pBe==0 ) return SQLITE_OK;
+ rc = sqliteBtreeCursor(db->pBe, 2, 0, &curMain);
+ if( rc ){
+ sqliteSetString(pzErrMsg, sqlite_error_string(rc), 0);
+ sqliteResetInternalSchema(db);
+ return rc;
+ }
+
+ /* Get the database meta information
+ */
+ rc = sqliteBtreeGetMeta(db->pBe, meta);
+ if( rc ){
+ sqliteSetString(pzErrMsg, sqlite_error_string(rc), 0);
+ sqliteResetInternalSchema(db);
+ sqliteBtreeCloseCursor(curMain);
+ return rc;
+ }
+ db->schema_cookie = meta[1];
+ db->next_cookie = db->schema_cookie;
+ db->file_format = meta[2];
+ size = meta[3];
+ if( size==0 ){ size = MAX_PAGES; }
+ db->cache_size = size;
+ sqliteBtreeSetCacheSize(db->pBe, size);
+ db->safety_level = meta[4];
+ if( db->safety_level==0 ) db->safety_level = 2;
+ sqliteBtreeSetSafetyLevel(db->pBe, db->safety_level);
+
+ /*
+ ** file_format==1 Version 2.1.0.
+ ** file_format==2 Version 2.2.0. Add support for INTEGER PRIMARY KEY.
+ ** file_format==3 Version 2.6.0. Fix empty-string index bug.
+ ** file_format==4 Version 2.7.0. Add support for separate numeric and
+ ** text datatypes.
+ */
+ if( db->file_format==0 ){
+ /* This happens if the database was initially empty */
+ db->file_format = 4;
+ }else if( db->file_format>4 ){
+ sqliteBtreeCloseCursor(curMain);
+ sqliteSetString(pzErrMsg, "unsupported file format", 0);
+ return SQLITE_ERROR;
+ }
+
+ /* Read the schema information out of the schema tables
+ */
+ memset(&sParse, 0, sizeof(sParse));
+ sParse.db = db;
+ sParse.pBe = db->pBe;
+ sParse.xCallback = sqliteInitCallback;
+ sParse.pArg = (void*)&initData;
+ sParse.initFlag = 1;
+ sParse.useCallback = 1;
+ sqliteRunParser(&sParse,
+ db->file_format>=2 ? init_script : older_init_script,
+ pzErrMsg);
+ if( sqlite_malloc_failed ){
+ sqliteSetString(pzErrMsg, "out of memory", 0);
+ sParse.rc = SQLITE_NOMEM;
+ sqliteBtreeRollback(db->pBe);
+ sqliteResetInternalSchema(db);
+ }
+ if( sParse.rc==SQLITE_OK ){
+ db->flags |= SQLITE_Initialized;
+ sqliteCommitInternalChanges(db);
+ }else{
+ db->flags &= ~SQLITE_Initialized;
+ sqliteResetInternalSchema(db);
+ }
+ sqliteBtreeCloseCursor(curMain);
+ return sParse.rc;
+}
+
+/*
+** The version of the library
+*/
+const char rcsid[] = "@(#) \044Id: SQLite version " SQLITE_VERSION " $";
+const char sqlite_version[] = SQLITE_VERSION;
+
+/*
+** Does the library expect data to be encoded as UTF-8 or iso8859? The
+** following global constant always lets us know.
+*/
+#ifdef SQLITE_UTF8
+const char sqlite_encoding[] = "UTF-8";
+#else
+const char sqlite_encoding[] = "iso8859";
+#endif
+
+/*
+** Open a new SQLite database. Construct an "sqlite" structure to define
+** the state of this database and return a pointer to that structure.
+**
+** An attempt is made to initialize the in-memory data structures that
+** hold the database schema. But if this fails (because the schema file
+** is locked) then that step is deferred until the first call to
+** sqlite_exec().
+*/
+sqlite *sqlite_open(const char *zFilename, int mode, char **pzErrMsg){
+ sqlite *db;
+ int rc;
+
+ /* Allocate the sqlite data structure */
+ db = sqliteMalloc( sizeof(sqlite) );
+ if( pzErrMsg ) *pzErrMsg = 0;
+ if( db==0 ) goto no_mem_on_open;
+ sqliteHashInit(&db->tblHash, SQLITE_HASH_STRING, 0);
+ sqliteHashInit(&db->idxHash, SQLITE_HASH_STRING, 0);
+ sqliteHashInit(&db->trigHash, SQLITE_HASH_STRING, 0);
+ sqliteHashInit(&db->aFunc, SQLITE_HASH_STRING, 1);
+ sqliteHashInit(&db->aFKey, SQLITE_HASH_STRING, 1);
+ db->onError = OE_Default;
+ db->priorNewRowid = 0;
+ db->magic = SQLITE_MAGIC_BUSY;
+
+ /* Open the backend database driver */
+ rc = sqliteBtreeOpen(zFilename, 0, MAX_PAGES, &db->pBe);
+ if( rc!=SQLITE_OK ){
+ switch( rc ){
+ default: {
+ sqliteSetString(pzErrMsg, "unable to open database: ", zFilename, 0);
+ }
+ }
+ sqliteFree(db);
+ sqliteStrRealloc(pzErrMsg);
+ return 0;
+ }
+
+ /* Attempt to read the schema */
+ sqliteRegisterBuiltinFunctions(db);
+ rc = sqliteInit(db, pzErrMsg);
+ db->magic = SQLITE_MAGIC_OPEN;
+ if( sqlite_malloc_failed ){
+ sqlite_close(db);
+ goto no_mem_on_open;
+ }else if( rc!=SQLITE_OK && rc!=SQLITE_BUSY ){
+ sqlite_close(db);
+ sqliteStrRealloc(pzErrMsg);
+ return 0;
+ }else if( pzErrMsg ){
+ sqliteFree(*pzErrMsg);
+ *pzErrMsg = 0;
+ }
+
+ /* If the database is in formats 1 or 2, then upgrade it to
+ ** version 3. This will reconstruct all indices. If the
+ ** upgrade fails for any reason (ex: out of disk space, database
+ ** is read only, interrupt received, etc.) then refuse to open.
+ */
+ if( rc==SQLITE_OK && db->file_format<3 ){
+ char *zErr = 0;
+ InitData initData;
+ int meta[SQLITE_N_BTREE_META];
+
+ initData.db = db;
+ initData.pzErrMsg = &zErr;
+ db->file_format = 3;
+ rc = sqlite_exec(db,
+ "BEGIN; SELECT name FROM sqlite_master WHERE type='table';",
+ upgrade_3_callback,
+ &initData,
+ &zErr);
+ if( rc==SQLITE_OK ){
+ sqliteBtreeGetMeta(db->pBe, meta);
+ meta[2] = 4;
+ sqliteBtreeUpdateMeta(db->pBe, meta);
+ sqlite_exec(db, "COMMIT", 0, 0, 0);
+ }
+ if( rc!=SQLITE_OK ){
+ sqliteSetString(pzErrMsg,
+ "unable to upgrade database to the version 2.6 format",
+ zErr ? ": " : 0, zErr, 0);
+ sqlite_freemem(zErr);
+ sqliteStrRealloc(pzErrMsg);
+ sqlite_close(db);
+ return 0;
+ }
+ sqlite_freemem(zErr);
+ }
+
+ /* Return a pointer to the newly opened database structure */
+ return db;
+
+no_mem_on_open:
+ sqliteSetString(pzErrMsg, "out of memory", 0);
+ sqliteStrRealloc(pzErrMsg);
+ return 0;
+}
+
+/*
+** Return the ROWID of the most recent insert
+*/
+int sqlite_last_insert_rowid(sqlite *db){
+ return db->lastRowid;
+}
+
+/*
+** Return the number of changes in the most recent call to sqlite_exec().
+*/
+int sqlite_changes(sqlite *db){
+ return db->nChange;
+}
+
+/*
+** Close an existing SQLite database
+*/
+void sqlite_close(sqlite *db){
+ HashElem *i;
+ db->want_to_close = 1;
+ if( sqliteSafetyCheck(db) || sqliteSafetyOn(db) ){
+ /* printf("DID NOT CLOSE\n"); fflush(stdout); */
+ return;
+ }
+ db->magic = SQLITE_MAGIC_CLOSED;
+ sqliteBtreeClose(db->pBe);
+ sqliteResetInternalSchema(db);
+ if( db->pBeTemp ){
+ sqliteBtreeClose(db->pBeTemp);
+ }
+ for(i=sqliteHashFirst(&db->aFunc); i; i=sqliteHashNext(i)){
+ FuncDef *pFunc, *pNext;
+ for(pFunc = (FuncDef*)sqliteHashData(i); pFunc; pFunc=pNext){
+ pNext = pFunc->pNext;
+ sqliteFree(pFunc);
+ }
+ }
+ sqliteHashClear(&db->aFunc);
+ sqliteHashClear(&db->aFKey);
+ sqliteFree(db);
+}
+
+/*
+** Return TRUE if the given SQL string ends in a semicolon.
+**
+** Special handling is require for CREATE TRIGGER statements.
+** Whenever the CREATE TRIGGER keywords are seen, the statement
+** must end with ";END;".
+*/
+int sqlite_complete(const char *zSql){
+ int isComplete = 1;
+ int requireEnd = 0;
+ int seenText = 0;
+ int seenCreate = 0;
+ while( *zSql ){
+ switch( *zSql ){
+ case ';': {
+ isComplete = 1;
+ seenText = 1;
+ seenCreate = 0;
+ break;
+ }
+ case ' ':
+ case '\t':
+ case '\n':
+ case '\f': {
+ break;
+ }
+ case '[': {
+ isComplete = 0;
+ seenText = 1;
+ seenCreate = 0;
+ zSql++;
+ while( *zSql && *zSql!=']' ){ zSql++; }
+ if( *zSql==0 ) return 0;
+ break;
+ }
+ case '"':
+ case '\'': {
+ int c = *zSql;
+ isComplete = 0;
+ seenText = 1;
+ seenCreate = 0;
+ zSql++;
+ while( *zSql && *zSql!=c ){ zSql++; }
+ if( *zSql==0 ) return 0;
+ break;
+ }
+ case '-': {
+ if( zSql[1]!='-' ){
+ isComplete = 0;
+ seenCreate = 0;
+ break;
+ }
+ while( *zSql && *zSql!='\n' ){ zSql++; }
+ if( *zSql==0 ) return seenText && isComplete && requireEnd==0;
+ break;
+ }
+ case 'c':
+ case 'C': {
+ seenText = 1;
+ if( !isComplete ) break;
+ isComplete = 0;
+ if( sqliteStrNICmp(zSql, "create", 6)!=0 ) break;
+ if( !isspace(zSql[6]) ) break;
+ zSql += 5;
+ seenCreate = 1;
+ while( isspace(zSql[1]) ) zSql++;
+ if( sqliteStrNICmp(&zSql[1],"trigger", 7)!=0 ) break;
+ zSql += 7;
+ requireEnd++;
+ break;
+ }
+ case 't':
+ case 'T': {
+ seenText = 1;
+ if( !seenCreate ) break;
+ seenCreate = 0;
+ isComplete = 0;
+ if( sqliteStrNICmp(zSql, "trigger", 7)!=0 ) break;
+ if( !isspace(zSql[7]) ) break;
+ zSql += 6;
+ requireEnd++;
+ break;
+ }
+ case 'e':
+ case 'E': {
+ seenCreate = 0;
+ seenText = 1;
+ if( !isComplete ) break;
+ isComplete = 0;
+ if( requireEnd==0 ) break;
+ if( sqliteStrNICmp(zSql, "end", 3)!=0 ) break;
+ zSql += 2;
+ while( isspace(zSql[1]) ) zSql++;
+ if( zSql[1]==';' ){
+ zSql++;
+ isComplete = 1;
+ requireEnd--;
+ }
+ break;
+ }
+ default: {
+ seenCreate = 0;
+ seenText = 1;
+ isComplete = 0;
+ break;
+ }
+ }
+ zSql++;
+ }
+ return seenText && isComplete && requireEnd==0;
+}
+
+/*
+** This routine does the work of either sqlite_exec() or sqlite_compile().
+** It works like sqlite_exec() if pVm==NULL and it works like sqlite_compile()
+** otherwise.
+*/
+static int sqliteMain(
+ sqlite *db, /* The database on which the SQL executes */
+ const char *zSql, /* The SQL to be executed */
+ sqlite_callback xCallback, /* Invoke this callback routine */
+ void *pArg, /* First argument to xCallback() */
+ const char **pzTail, /* OUT: Next statement after the first */
+ sqlite_vm **ppVm, /* OUT: The virtual machine */
+ char **pzErrMsg /* OUT: Write error messages here */
+){
+ Parse sParse;
+
+ if( pzErrMsg ) *pzErrMsg = 0;
+ if( sqliteSafetyOn(db) ) goto exec_misuse;
+ if( (db->flags & SQLITE_Initialized)==0 ){
+ int rc, cnt = 1;
+ while( (rc = sqliteInit(db, pzErrMsg))==SQLITE_BUSY
+ && db->xBusyCallback && db->xBusyCallback(db->pBusyArg, "", cnt++)!=0 ){}
+ if( rc!=SQLITE_OK ){
+ sqliteStrRealloc(pzErrMsg);
+ sqliteSafetyOff(db);
+ return rc;
+ }
+ if( pzErrMsg ){
+ sqliteFree(*pzErrMsg);
+ *pzErrMsg = 0;
+ }
+ }
+ if( db->file_format<3 ){
+ sqliteSafetyOff(db);
+ sqliteSetString(pzErrMsg, "obsolete database file format", 0);
+ return SQLITE_ERROR;
+ }
+ if( db->pVdbe==0 ){ db->nChange = 0; }
+ memset(&sParse, 0, sizeof(sParse));
+ sParse.db = db;
+ sParse.pBe = db->pBe;
+ sParse.xCallback = xCallback;
+ sParse.pArg = pArg;
+ sParse.useCallback = ppVm==0;
+#ifndef SQLITE_OMIT_TRACE
+ if( db->xTrace ) db->xTrace(db->pTraceArg, zSql);
+#endif
+ sqliteRunParser(&sParse, zSql, pzErrMsg);
+ if( sqlite_malloc_failed ){
+ sqliteSetString(pzErrMsg, "out of memory", 0);
+ sParse.rc = SQLITE_NOMEM;
+ sqliteBtreeRollback(db->pBe);
+ if( db->pBeTemp ) sqliteBtreeRollback(db->pBeTemp);
+ db->flags &= ~SQLITE_InTrans;
+ sqliteResetInternalSchema(db);
+ }
+ if( sParse.rc==SQLITE_DONE ) sParse.rc = SQLITE_OK;
+ if( sParse.rc!=SQLITE_OK && pzErrMsg && *pzErrMsg==0 ){
+ sqliteSetString(pzErrMsg, sqlite_error_string(sParse.rc), 0);
+ }
+ sqliteStrRealloc(pzErrMsg);
+ if( sParse.rc==SQLITE_SCHEMA ){
+ sqliteResetInternalSchema(db);
+ }
+ if( sParse.useCallback==0 ){
+ assert( ppVm );
+ *ppVm = (sqlite_vm*)sParse.pVdbe;
+ *pzTail = sParse.zTail;
+ }
+ if( sqliteSafetyOff(db) ) goto exec_misuse;
+ return sParse.rc;
+
+exec_misuse:
+ if( pzErrMsg ){
+ *pzErrMsg = 0;
+ sqliteSetString(pzErrMsg, sqlite_error_string(SQLITE_MISUSE), 0);
+ sqliteStrRealloc(pzErrMsg);
+ }
+ return SQLITE_MISUSE;
+}
+
+/*
+** Execute SQL code. Return one of the SQLITE_ success/failure
+** codes. Also write an error message into memory obtained from
+** malloc() and make *pzErrMsg point to that message.
+**
+** If the SQL is a query, then for each row in the query result
+** the xCallback() function is called. pArg becomes the first
+** argument to xCallback(). If xCallback=NULL then no callback
+** is invoked, even for queries.
+*/
+int sqlite_exec(
+ sqlite *db, /* The database on which the SQL executes */
+ const char *zSql, /* The SQL to be executed */
+ sqlite_callback xCallback, /* Invoke this callback routine */
+ void *pArg, /* First argument to xCallback() */
+ char **pzErrMsg /* Write error messages here */
+){
+ return sqliteMain(db, zSql, xCallback, pArg, 0, 0, pzErrMsg);
+}
+
+/*
+** Compile a single statement of SQL into a virtual machine. Return one
+** of the SQLITE_ success/failure codes. Also write an error message into
+** memory obtained from malloc() and make *pzErrMsg point to that message.
+*/
+int sqlite_compile(
+ sqlite *db, /* The database on which the SQL executes */
+ const char *zSql, /* The SQL to be executed */
+ const char **pzTail, /* OUT: Next statement after the first */
+ sqlite_vm **ppVm, /* OUT: The virtual machine */
+ char **pzErrMsg /* OUT: Write error messages here */
+){
+ return sqliteMain(db, zSql, 0, 0, pzTail, ppVm, pzErrMsg);
+}
+
+/*
+** The following routine destroys a virtual machine that is created by
+** the sqlite_compile() routine.
+**
+** The integer returned is an SQLITE_ success/failure code that describes
+** the result of executing the virtual machine. An error message is
+** written into memory obtained from malloc and *pzErrMsg is made to
+** point to that error if pzErrMsg is not NULL. The calling routine
+** should use sqlite_freemem() to delete the message when it has finished
+** with it.
+*/
+int sqlite_finalize(
+ sqlite_vm *pVm, /* The virtual machine to be destroyed */
+ char **pzErrMsg /* OUT: Write error messages here */
+){
+ int rc = sqliteVdbeFinalize((Vdbe*)pVm, pzErrMsg);
+ sqliteStrRealloc(pzErrMsg);
+ return rc;
+}
+
+/*
+** Return a static string that describes the kind of error specified in the
+** argument.
+*/
+const char *sqlite_error_string(int rc){
+ const char *z;
+ switch( rc ){
+ case SQLITE_OK: z = "not an error"; break;
+ case SQLITE_ERROR: z = "SQL logic error or missing database"; break;
+ case SQLITE_INTERNAL: z = "internal SQLite implementation flaw"; break;
+ case SQLITE_PERM: z = "access permission denied"; break;
+ case SQLITE_ABORT: z = "callback requested query abort"; break;
+ case SQLITE_BUSY: z = "database is locked"; break;
+ case SQLITE_LOCKED: z = "database table is locked"; break;
+ case SQLITE_NOMEM: z = "out of memory"; break;
+ case SQLITE_READONLY: z = "attempt to write a readonly database"; break;
+ case SQLITE_INTERRUPT: z = "interrupted"; break;
+ case SQLITE_IOERR: z = "disk I/O error"; break;
+ case SQLITE_CORRUPT: z = "database disk image is malformed"; break;
+ case SQLITE_NOTFOUND: z = "table or record not found"; break;
+ case SQLITE_FULL: z = "database is full"; break;
+ case SQLITE_CANTOPEN: z = "unable to open database file"; break;
+ case SQLITE_PROTOCOL: z = "database locking protocol failure"; break;
+ case SQLITE_EMPTY: z = "table contains no data"; break;
+ case SQLITE_SCHEMA: z = "database schema has changed"; break;
+ case SQLITE_TOOBIG: z = "too much data for one table row"; break;
+ case SQLITE_CONSTRAINT: z = "constraint failed"; break;
+ case SQLITE_MISMATCH: z = "datatype mismatch"; break;
+ case SQLITE_MISUSE: z = "library routine called out of sequence";break;
+ case SQLITE_NOLFS: z = "kernel lacks large file support"; break;
+ case SQLITE_AUTH: z = "authorization denied"; break;
+ default: z = "unknown error"; break;
+ }
+ return z;
+}
+
+/*
+** This routine implements a busy callback that sleeps and tries
+** again until a timeout value is reached. The timeout value is
+** an integer number of milliseconds passed in as the first
+** argument.
+*/
+static int sqliteDefaultBusyCallback(
+ void *Timeout, /* Maximum amount of time to wait */
+ const char *NotUsed, /* The name of the table that is busy */
+ int count /* Number of times table has been busy */
+){
+#if SQLITE_MIN_SLEEP_MS==1
+ int delay = 10;
+ int prior_delay = 0;
+ int timeout = (int)Timeout;
+ int i;
+
+ for(i=1; i<count; i++){
+ prior_delay += delay;
+ delay = delay*2;
+ if( delay>=1000 ){
+ delay = 1000;
+ prior_delay += 1000*(count - i - 1);
+ break;
+ }
+ }
+ if( prior_delay + delay > timeout ){
+ delay = timeout - prior_delay;
+ if( delay<=0 ) return 0;
+ }
+ sqliteOsSleep(delay);
+ return 1;
+#else
+ int timeout = (int)Timeout;
+ if( (count+1)*1000 > timeout ){
+ return 0;
+ }
+ sqliteOsSleep(1000);
+ return 1;
+#endif
+}
+
+/*
+** This routine sets the busy callback for an Sqlite database to the
+** given callback function with the given argument.
+*/
+void sqlite_busy_handler(
+ sqlite *db,
+ int (*xBusy)(void*,const char*,int),
+ void *pArg
+){
+ db->xBusyCallback = xBusy;
+ db->pBusyArg = pArg;
+}
+
+/*
+** This routine installs a default busy handler that waits for the
+** specified number of milliseconds before returning 0.
+*/
+void sqlite_busy_timeout(sqlite *db, int ms){
+ if( ms>0 ){
+ sqlite_busy_handler(db, sqliteDefaultBusyCallback, (void*)ms);
+ }else{
+ sqlite_busy_handler(db, 0, 0);
+ }
+}
+
+/*
+** Cause any pending operation to stop at its earliest opportunity.
+*/
+void sqlite_interrupt(sqlite *db){
+ db->flags |= SQLITE_Interrupt;
+}
+
+/*
+** Windows systems should call this routine to free memory that
+** is returned in the in the errmsg parameter of sqlite_open() when
+** SQLite is a DLL. For some reason, it does not work to call free()
+** directly.
+**
+** Note that we need to call free() not sqliteFree() here, since every
+** string that is exported from SQLite should have already passed through
+** sqliteStrRealloc().
+*/
+void sqlite_freemem(void *p){ free(p); }
+
+/*
+** Windows systems need functions to call to return the sqlite_version
+** and sqlite_encoding strings since they are unable to access constants
+** within DLLs.
+*/
+const char *sqlite_libversion(void){ return sqlite_version; }
+const char *sqlite_libencoding(void){ return sqlite_encoding; }
+
+/*
+** Create new user-defined functions. The sqlite_create_function()
+** routine creates a regular function and sqlite_create_aggregate()
+** creates an aggregate function.
+**
+** Passing a NULL xFunc argument or NULL xStep and xFinalize arguments
+** disables the function. Calling sqlite_create_function() with the
+** same name and number of arguments as a prior call to
+** sqlite_create_aggregate() disables the prior call to
+** sqlite_create_aggregate(), and vice versa.
+**
+** If nArg is -1 it means that this function will accept any number
+** of arguments, including 0.
+*/
+int sqlite_create_function(
+ sqlite *db, /* Add the function to this database connection */
+ const char *zName, /* Name of the function to add */
+ int nArg, /* Number of arguments */
+ void (*xFunc)(sqlite_func*,int,const char**), /* The implementation */
+ void *pUserData /* User data */
+){
+ FuncDef *p;
+ int nName;
+ if( db==0 || zName==0 || sqliteSafetyCheck(db) ) return 1;
+ nName = strlen(zName);
+ if( nName>255 ) return 1;
+ p = sqliteFindFunction(db, zName, nName, nArg, 1);
+ if( p==0 ) return 1;
+ p->xFunc = xFunc;
+ p->xStep = 0;
+ p->xFinalize = 0;
+ p->pUserData = pUserData;
+ return 0;
+}
+int sqlite_create_aggregate(
+ sqlite *db, /* Add the function to this database connection */
+ const char *zName, /* Name of the function to add */
+ int nArg, /* Number of arguments */
+ void (*xStep)(sqlite_func*,int,const char**), /* The step function */
+ void (*xFinalize)(sqlite_func*), /* The finalizer */
+ void *pUserData /* User data */
+){
+ FuncDef *p;
+ int nName;
+ if( db==0 || zName==0 || sqliteSafetyCheck(db) ) return 1;
+ nName = strlen(zName);
+ if( nName>255 ) return 1;
+ p = sqliteFindFunction(db, zName, nName, nArg, 1);
+ if( p==0 ) return 1;
+ p->xFunc = 0;
+ p->xStep = xStep;
+ p->xFinalize = xFinalize;
+ p->pUserData = pUserData;
+ return 0;
+}
+
+/*
+** Change the datatype for all functions with a given name. See the
+** header comment for the prototype of this function in sqlite.h for
+** additional information.
+*/
+int sqlite_function_type(sqlite *db, const char *zName, int dataType){
+ FuncDef *p = (FuncDef*)sqliteHashFind(&db->aFunc, zName, strlen(zName));
+ while( p ){
+ p->dataType = dataType;
+ p = p->pNext;
+ }
+ return SQLITE_OK;
+}
+
+/*
+** Register a trace function. The pArg from the previously registered trace
+** is returned.
+**
+** A NULL trace function means that no tracing is executes. A non-NULL
+** trace is a pointer to a function that is invoked at the start of each
+** sqlite_exec().
+*/
+void *sqlite_trace(sqlite *db, void (*xTrace)(void*,const char*), void *pArg){
+#ifndef SQLITE_OMIT_TRACE
+ void *pOld = db->pTraceArg;
+ db->xTrace = xTrace;
+ db->pTraceArg = pArg;
+ return pOld;
+#else
+ return 0;
+#endif
+}
+
+
+/*
+** Attempt to open the file named in the argument as the auxiliary database
+** file. The auxiliary database file is used to store TEMP tables. But
+** by using this API, it is possible to trick SQLite into opening two
+** separate databases and acting on them as if they were one.
+**
+** This routine closes the existing auxiliary database file, which will
+** cause any previously created TEMP tables to be dropped.
+**
+** The zName parameter can be a NULL pointer or an empty string to cause
+** a temporary file to be opened and automatically deleted when closed.
+*/
+int sqlite_open_aux_file(sqlite *db, const char *zName, char **pzErrMsg){
+ int rc;
+ if( zName && zName[0]==0 ) zName = 0;
+ if( sqliteSafetyOn(db) ) goto openaux_misuse;
+ sqliteResetInternalSchema(db);
+ if( db->pBeTemp!=0 ){
+ sqliteBtreeClose(db->pBeTemp);
+ }
+ rc = sqliteBtreeOpen(zName, 0, MAX_PAGES, &db->pBeTemp);
+ if( rc ){
+ if( zName==0 ) zName = "a temporary file";
+ sqliteSetString(pzErrMsg, "unable to open ", zName,
+ ": ", sqlite_error_string(rc), 0);
+ sqliteStrRealloc(pzErrMsg);
+ sqliteSafetyOff(db);
+ return rc;
+ }
+ rc = sqliteInit(db, pzErrMsg);
+ if( sqliteSafetyOff(db) ) goto openaux_misuse;
+ sqliteStrRealloc(pzErrMsg);
+ return rc;
+
+openaux_misuse:
+ sqliteSetString(pzErrMsg, sqlite_error_string(SQLITE_MISUSE), 0);
+ sqliteStrRealloc(pzErrMsg);
+ return SQLITE_MISUSE;
+}