DBD-SQLcipher
view release on metacpan or search on metacpan
#ifndef O_LARGEFILE
# define O_LARGEFILE 0
#endif
#ifdef SQLITE_DISABLE_LFS
# undef O_LARGEFILE
# define O_LARGEFILE 0
#endif
#ifndef O_NOFOLLOW
# define O_NOFOLLOW 0
#endif
#ifndef O_BINARY
# define O_BINARY 0
#endif
/*
** The threadid macro resolves to the thread-id or to 0. Used for
** testing and debugging only.
*/
#if SQLITE_THREADSAFE
#define threadid pthread_self()
#else
#define threadid 0
#endif
/*
** HAVE_MREMAP defaults to true on Linux and false everywhere else.
*/
#if !defined(HAVE_MREMAP)
# if defined(__linux__) && defined(_GNU_SOURCE)
# define HAVE_MREMAP 1
# else
# define HAVE_MREMAP 0
# endif
#endif
/*
** Explicitly call the 64-bit version of lseek() on Android. Otherwise, lseek()
** is the 32-bit version, even if _FILE_OFFSET_BITS=64 is defined.
*/
#ifdef __ANDROID__
# define lseek lseek64
#endif
/*
** Different Unix systems declare open() in different ways. Same use
** open(const char*,int,mode_t). Others use open(const char*,int,...).
** The difference is important when using a pointer to the function.
**
** The safest way to deal with the problem is to always use this wrapper
** which always has the same well-defined interface.
*/
static int posixOpen(const char *zFile, int flags, int mode){
return open(zFile, flags, mode);
}
/*
** On some systems, calls to fchown() will trigger a message in a security
** log if they come from non-root processes. So avoid calling fchown() if
** we are not running as root.
*/
static int posixFchown(int fd, uid_t uid, gid_t gid){
#if OS_VXWORKS
return 0;
#else
return geteuid() ? 0 : fchown(fd,uid,gid);
#endif
}
/* Forward reference */
static int openDirectory(const char*, int*);
static int unixGetpagesize(void);
/*
** Many system calls are accessed through pointer-to-functions so that
** they may be overridden at runtime to facilitate fault injection during
** testing and sandboxing. The following array holds the names and pointers
** to all overrideable system calls.
*/
static struct unix_syscall {
const char *zName; /* Name of the system call */
sqlite3_syscall_ptr pCurrent; /* Current value of the system call */
sqlite3_syscall_ptr pDefault; /* Default value */
} aSyscall[] = {
{ "open", (sqlite3_syscall_ptr)posixOpen, 0 },
#define osOpen ((int(*)(const char*,int,int))aSyscall[0].pCurrent)
{ "close", (sqlite3_syscall_ptr)close, 0 },
#define osClose ((int(*)(int))aSyscall[1].pCurrent)
{ "access", (sqlite3_syscall_ptr)access, 0 },
#define osAccess ((int(*)(const char*,int))aSyscall[2].pCurrent)
{ "getcwd", (sqlite3_syscall_ptr)getcwd, 0 },
#define osGetcwd ((char*(*)(char*,size_t))aSyscall[3].pCurrent)
{ "stat", (sqlite3_syscall_ptr)stat, 0 },
#define osStat ((int(*)(const char*,struct stat*))aSyscall[4].pCurrent)
/*
** The DJGPP compiler environment looks mostly like Unix, but it
** lacks the fcntl() system call. So redefine fcntl() to be something
** that always succeeds. This means that locking does not occur under
** DJGPP. But it is DOS - what did you expect?
*/
#ifdef __DJGPP__
{ "fstat", 0, 0 },
#define osFstat(a,b,c) 0
#else
{ "fstat", (sqlite3_syscall_ptr)fstat, 0 },
#define osFstat ((int(*)(int,struct stat*))aSyscall[5].pCurrent)
#endif
{ "ftruncate", (sqlite3_syscall_ptr)ftruncate, 0 },
#define osFtruncate ((int(*)(int,off_t))aSyscall[6].pCurrent)
{ "fcntl", (sqlite3_syscall_ptr)fcntl, 0 },
#define osFcntl ((int(*)(int,int,...))aSyscall[7].pCurrent)
{ "read", (sqlite3_syscall_ptr)read, 0 },
#define osRead ((ssize_t(*)(int,void*,size_t))aSyscall[8].pCurrent)
#if defined(USE_PREAD) || (SQLITE_ENABLE_LOCKING_STYLE && !OS_VXWORKS)
{ "pread", (sqlite3_syscall_ptr)pread, 0 },
#else
{ "pread", (sqlite3_syscall_ptr)0, 0 },
#endif
#define osPread ((ssize_t(*)(int,void*,size_t,off_t))aSyscall[9].pCurrent)
#if defined(USE_PREAD64)
{ "pread64", (sqlite3_syscall_ptr)pread64, 0 },
#else
{ "pread64", (sqlite3_syscall_ptr)0, 0 },
#endif
#define osPread64 ((ssize_t(*)(int,void*,size_t,off_t))aSyscall[10].pCurrent)
{ "write", (sqlite3_syscall_ptr)write, 0 },
#define osWrite ((ssize_t(*)(int,const void*,size_t))aSyscall[11].pCurrent)
#if defined(USE_PREAD) || (SQLITE_ENABLE_LOCKING_STYLE && !OS_VXWORKS)
{ "pwrite", (sqlite3_syscall_ptr)pwrite, 0 },
#else
{ "pwrite", (sqlite3_syscall_ptr)0, 0 },
#endif
#define osPwrite ((ssize_t(*)(int,const void*,size_t,off_t))\
aSyscall[12].pCurrent)
#if defined(USE_PREAD64)
{ "pwrite64", (sqlite3_syscall_ptr)pwrite64, 0 },
#else
{ "pwrite64", (sqlite3_syscall_ptr)0, 0 },
#endif
#define osPwrite64 ((ssize_t(*)(int,const void*,size_t,off_t))\
aSyscall[13].pCurrent)
{ "fchmod", (sqlite3_syscall_ptr)fchmod, 0 },
#define osFchmod ((int(*)(int,mode_t))aSyscall[14].pCurrent)
#if defined(HAVE_POSIX_FALLOCATE) && HAVE_POSIX_FALLOCATE
{ "fallocate", (sqlite3_syscall_ptr)posix_fallocate, 0 },
#else
{ "fallocate", (sqlite3_syscall_ptr)0, 0 },
#endif
#define osFallocate ((int(*)(int,off_t,off_t))aSyscall[15].pCurrent)
{ "unlink", (sqlite3_syscall_ptr)unlink, 0 },
#define osUnlink ((int(*)(const char*))aSyscall[16].pCurrent)
{ "openDirectory", (sqlite3_syscall_ptr)openDirectory, 0 },
#define osOpenDirectory ((int(*)(const char*,int*))aSyscall[17].pCurrent)
{ "mkdir", (sqlite3_syscall_ptr)mkdir, 0 },
#define osMkdir ((int(*)(const char*,mode_t))aSyscall[18].pCurrent)
{ "rmdir", (sqlite3_syscall_ptr)rmdir, 0 },
#define osRmdir ((int(*)(const char*))aSyscall[19].pCurrent)
{ "fchown", (sqlite3_syscall_ptr)posixFchown, 0 },
#define osFchown ((int(*)(int,uid_t,gid_t))aSyscall[20].pCurrent)
#if !defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0
{ "mmap", (sqlite3_syscall_ptr)mmap, 0 },
#define osMmap ((void*(*)(void*,size_t,int,int,int,off_t))aSyscall[21].pCurrent)
{ "munmap", (sqlite3_syscall_ptr)munmap, 0 },
#define osMunmap ((void*(*)(void*,size_t))aSyscall[22].pCurrent)
#if HAVE_MREMAP
{ "mremap", (sqlite3_syscall_ptr)mremap, 0 },
#else
{ "mremap", (sqlite3_syscall_ptr)0, 0 },
#endif
#define osMremap ((void*(*)(void*,size_t,size_t,int,...))aSyscall[23].pCurrent)
{ "getpagesize", (sqlite3_syscall_ptr)unixGetpagesize, 0 },
#define osGetpagesize ((int(*)(void))aSyscall[24].pCurrent)
#endif
}; /* End of the overrideable system calls */
/*
** This is the xSetSystemCall() method of sqlite3_vfs for all of the
** "unix" VFSes. Return SQLITE_OK opon successfully updating the
** system call pointer, or SQLITE_NOTFOUND if there is no configurable
** system call named zName.
*/
static int unixSetSystemCall(
sqlite3_vfs *pNotUsed, /* The VFS pointer. Not used */
const char *zName, /* Name of system call to override */
sqlite3_syscall_ptr pNewFunc /* Pointer to new system call value */
){
unsigned int i;
int rc = SQLITE_NOTFOUND;
UNUSED_PARAMETER(pNotUsed);
if( zName==0 ){
/* If no zName is given, restore all system calls to their default
** settings and return NULL
*/
rc = SQLITE_OK;
for(i=0; i<sizeof(aSyscall)/sizeof(aSyscall[0]); i++){
if( aSyscall[i].pDefault ){
aSyscall[i].pCurrent = aSyscall[i].pDefault;
}
}
}else{
/* If zName is specified, operate on only the one system call
** specified.
*/
for(i=0; i<sizeof(aSyscall)/sizeof(aSyscall[0]); i++){
if( strcmp(zName, aSyscall[i].zName)==0 ){
if( aSyscall[i].pDefault==0 ){
aSyscall[i].pDefault = aSyscall[i].pCurrent;
}
rc = SQLITE_OK;
if( pNewFunc==0 ) pNewFunc = aSyscall[i].pDefault;
aSyscall[i].pCurrent = pNewFunc;
break;
}
unixEnterMutex();
pInode = pDbFd->pInode;
pShmNode = pInode->pShmNode;
if( pShmNode==0 ){
struct stat sStat; /* fstat() info for database file */
/* Call fstat() to figure out the permissions on the database file. If
** a new *-shm file is created, an attempt will be made to create it
** with the same permissions.
*/
if( osFstat(pDbFd->h, &sStat) && pInode->bProcessLock==0 ){
rc = SQLITE_IOERR_FSTAT;
goto shm_open_err;
}
#ifdef SQLITE_SHM_DIRECTORY
nShmFilename = sizeof(SQLITE_SHM_DIRECTORY) + 31;
#else
nShmFilename = 6 + (int)strlen(pDbFd->zPath);
#endif
pShmNode = sqlite3_malloc( sizeof(*pShmNode) + nShmFilename );
if( pShmNode==0 ){
rc = SQLITE_NOMEM;
goto shm_open_err;
}
memset(pShmNode, 0, sizeof(*pShmNode)+nShmFilename);
zShmFilename = pShmNode->zFilename = (char*)&pShmNode[1];
#ifdef SQLITE_SHM_DIRECTORY
sqlite3_snprintf(nShmFilename, zShmFilename,
SQLITE_SHM_DIRECTORY "/sqlite-shm-%x-%x",
(u32)sStat.st_ino, (u32)sStat.st_dev);
#else
sqlite3_snprintf(nShmFilename, zShmFilename, "%s-shm", pDbFd->zPath);
sqlite3FileSuffix3(pDbFd->zPath, zShmFilename);
#endif
pShmNode->h = -1;
pDbFd->pInode->pShmNode = pShmNode;
pShmNode->pInode = pDbFd->pInode;
pShmNode->mutex = sqlite3_mutex_alloc(SQLITE_MUTEX_FAST);
if( pShmNode->mutex==0 ){
rc = SQLITE_NOMEM;
goto shm_open_err;
}
if( pInode->bProcessLock==0 ){
int openFlags = O_RDWR | O_CREAT;
if( sqlite3_uri_boolean(pDbFd->zPath, "readonly_shm", 0) ){
openFlags = O_RDONLY;
pShmNode->isReadonly = 1;
}
pShmNode->h = robust_open(zShmFilename, openFlags, (sStat.st_mode&0777));
if( pShmNode->h<0 ){
rc = unixLogError(SQLITE_CANTOPEN_BKPT, "open", zShmFilename);
goto shm_open_err;
}
/* If this process is running as root, make sure that the SHM file
** is owned by the same user that owns the original database. Otherwise,
** the original owner will not be able to connect.
*/
osFchown(pShmNode->h, sStat.st_uid, sStat.st_gid);
/* Check to see if another process is holding the dead-man switch.
** If not, truncate the file to zero length.
*/
rc = SQLITE_OK;
if( unixShmSystemLock(pShmNode, F_WRLCK, UNIX_SHM_DMS, 1)==SQLITE_OK ){
if( robust_ftruncate(pShmNode->h, 0) ){
rc = unixLogError(SQLITE_IOERR_SHMOPEN, "ftruncate", zShmFilename);
}
}
if( rc==SQLITE_OK ){
rc = unixShmSystemLock(pShmNode, F_RDLCK, UNIX_SHM_DMS, 1);
}
if( rc ) goto shm_open_err;
}
}
/* Make the new connection a child of the unixShmNode */
p->pShmNode = pShmNode;
#ifdef SQLITE_DEBUG
p->id = pShmNode->nextShmId++;
#endif
pShmNode->nRef++;
pDbFd->pShm = p;
unixLeaveMutex();
/* The reference count on pShmNode has already been incremented under
** the cover of the unixEnterMutex() mutex and the pointer from the
** new (struct unixShm) object to the pShmNode has been set. All that is
** left to do is to link the new object into the linked list starting
** at pShmNode->pFirst. This must be done while holding the pShmNode->mutex
** mutex.
*/
sqlite3_mutex_enter(pShmNode->mutex);
p->pNext = pShmNode->pFirst;
pShmNode->pFirst = p;
sqlite3_mutex_leave(pShmNode->mutex);
return SQLITE_OK;
/* Jump here on any error */
shm_open_err:
unixShmPurge(pDbFd); /* This call frees pShmNode if required */
sqlite3_free(p);
unixLeaveMutex();
return rc;
}
/*
** This function is called to obtain a pointer to region iRegion of the
** shared-memory associated with the database file fd. Shared-memory regions
** are numbered starting from zero. Each shared-memory region is szRegion
** bytes in size.
**
** If an error occurs, an error code is returned and *pp is set to NULL.
**
** Otherwise, if the bExtend parameter is 0 and the requested shared-memory
** region has not been allocated (by any client, including one running in a
** separate process), then *pp is set to NULL and SQLITE_OK returned. If
** bExtend is non-zero and the requested shared-memory region has not yet
** been allocated, it is allocated by this function.
#if !OS_VXWORKS
struct stat sStat; /* Results of stat() call */
/* A stat() call may fail for various reasons. If this happens, it is
** almost certain that an open() call on the same path will also fail.
** For this reason, if an error occurs in the stat() call here, it is
** ignored and -1 is returned. The caller will try to open a new file
** descriptor on the same path, fail, and return an error to SQLite.
**
** Even if a subsequent open() call does succeed, the consequences of
** not searching for a reusable file descriptor are not dire. */
if( 0==osStat(zPath, &sStat) ){
unixInodeInfo *pInode;
unixEnterMutex();
pInode = inodeList;
while( pInode && (pInode->fileId.dev!=sStat.st_dev
|| pInode->fileId.ino!=sStat.st_ino) ){
pInode = pInode->pNext;
}
if( pInode ){
UnixUnusedFd **pp;
for(pp=&pInode->pUnused; *pp && (*pp)->flags!=flags; pp=&((*pp)->pNext));
pUnused = *pp;
if( pUnused ){
*pp = pUnused->pNext;
}
}
unixLeaveMutex();
}
#endif /* if !OS_VXWORKS */
return pUnused;
}
/*
** This function is called by unixOpen() to determine the unix permissions
** to create new files with. If no error occurs, then SQLITE_OK is returned
** and a value suitable for passing as the third argument to open(2) is
** written to *pMode. If an IO error occurs, an SQLite error code is
** returned and the value of *pMode is not modified.
**
** In most cases, this routine sets *pMode to 0, which will become
** an indication to robust_open() to create the file using
** SQLITE_DEFAULT_FILE_PERMISSIONS adjusted by the umask.
** But if the file being opened is a WAL or regular journal file, then
** this function queries the file-system for the permissions on the
** corresponding database file and sets *pMode to this value. Whenever
** possible, WAL and journal files are created using the same permissions
** as the associated database file.
**
** If the SQLITE_ENABLE_8_3_NAMES option is enabled, then the
** original filename is unavailable. But 8_3_NAMES is only used for
** FAT filesystems and permissions do not matter there, so just use
** the default permissions.
*/
static int findCreateFileMode(
const char *zPath, /* Path of file (possibly) being created */
int flags, /* Flags passed as 4th argument to xOpen() */
mode_t *pMode, /* OUT: Permissions to open file with */
uid_t *pUid, /* OUT: uid to set on the file */
gid_t *pGid /* OUT: gid to set on the file */
){
int rc = SQLITE_OK; /* Return Code */
*pMode = 0;
*pUid = 0;
*pGid = 0;
if( flags & (SQLITE_OPEN_WAL|SQLITE_OPEN_MAIN_JOURNAL) ){
char zDb[MAX_PATHNAME+1]; /* Database file path */
int nDb; /* Number of valid bytes in zDb */
struct stat sStat; /* Output of stat() on database file */
/* zPath is a path to a WAL or journal file. The following block derives
** the path to the associated database file from zPath. This block handles
** the following naming conventions:
**
** "<path to db>-journal"
** "<path to db>-wal"
** "<path to db>-journalNN"
** "<path to db>-walNN"
**
** where NN is a decimal number. The NN naming schemes are
** used by the test_multiplex.c module.
*/
nDb = sqlite3Strlen30(zPath) - 1;
#ifdef SQLITE_ENABLE_8_3_NAMES
while( nDb>0 && sqlite3Isalnum(zPath[nDb]) ) nDb--;
if( nDb==0 || zPath[nDb]!='-' ) return SQLITE_OK;
#else
while( zPath[nDb]!='-' ){
assert( nDb>0 );
assert( zPath[nDb]!='\n' );
nDb--;
}
#endif
memcpy(zDb, zPath, nDb);
zDb[nDb] = '\0';
if( 0==osStat(zDb, &sStat) ){
*pMode = sStat.st_mode & 0777;
*pUid = sStat.st_uid;
*pGid = sStat.st_gid;
}else{
rc = SQLITE_IOERR_FSTAT;
}
}else if( flags & SQLITE_OPEN_DELETEONCLOSE ){
*pMode = 0600;
}
return rc;
}
/*
** Open the file zPath.
**
** Previously, the SQLite OS layer used three functions in place of this
** one:
**
** sqlite3OsOpenReadWrite();
** sqlite3OsOpenReadOnly();
** sqlite3OsOpenExclusive();
**
** These calls correspond to the following combinations of flags:
**
** ReadWrite() -> (READWRITE | CREATE)
** ReadOnly() -> (READONLY)
** OpenExclusive() -> (READWRITE | CREATE | EXCLUSIVE)
**
** The old OpenExclusive() accepted a boolean argument - "delFlag". If
** true, the file was configured to be automatically deleted when the
** file handle closed. To achieve the same effect using this new
** interface, add the DELETEONCLOSE flag to those specified above for
** OpenExclusive().
*/
static int unixOpen(
sqlite3_vfs *pVfs, /* The VFS for which this is the xOpen method */
const char *zPath, /* Pathname of file to be opened */
sqlite3_file *pFile, /* The file descriptor to be filled in */
int flags, /* Input flags to control the opening */
int *pOutFlags /* Output flags returned to SQLite core */
){
unixFile *p = (unixFile *)pFile;
int fd = -1; /* File descriptor returned by open() */
int openFlags = 0; /* Flags to pass to open() */
int eType = flags&0xFFFFFF00; /* Type of file to open */
int noLock; /* True to omit locking primitives */
int rc = SQLITE_OK; /* Function Return Code */
int ctrlFlags = 0; /* UNIXFILE_* flags */
int isExclusive = (flags & SQLITE_OPEN_EXCLUSIVE);
int isDelete = (flags & SQLITE_OPEN_DELETEONCLOSE);
int isCreate = (flags & SQLITE_OPEN_CREATE);
int isReadonly = (flags & SQLITE_OPEN_READONLY);
int isReadWrite = (flags & SQLITE_OPEN_READWRITE);
#if SQLITE_ENABLE_LOCKING_STYLE
int isAutoProxy = (flags & SQLITE_OPEN_AUTOPROXY);
#endif
#if defined(__APPLE__) || SQLITE_ENABLE_LOCKING_STYLE
struct statfs fsInfo;
#endif
/* If creating a master or main-file journal, this function will open
** a file-descriptor on the directory too. The first time unixSync()
|| eType==SQLITE_OPEN_TRANSIENT_DB || eType==SQLITE_OPEN_WAL
);
/* Detect a pid change and reset the PRNG. There is a race condition
** here such that two or more threads all trying to open databases at
** the same instant might all reset the PRNG. But multiple resets
** are harmless.
*/
if( randomnessPid!=getpid() ){
randomnessPid = getpid();
sqlite3_randomness(0,0);
}
memset(p, 0, sizeof(unixFile));
if( eType==SQLITE_OPEN_MAIN_DB ){
UnixUnusedFd *pUnused;
pUnused = findReusableFd(zName, flags);
if( pUnused ){
fd = pUnused->fd;
}else{
pUnused = sqlite3_malloc(sizeof(*pUnused));
if( !pUnused ){
return SQLITE_NOMEM;
}
}
p->pUnused = pUnused;
/* Database filenames are double-zero terminated if they are not
** URIs with parameters. Hence, they can always be passed into
** sqlite3_uri_parameter(). */
assert( (flags & SQLITE_OPEN_URI) || zName[strlen(zName)+1]==0 );
}else if( !zName ){
/* If zName is NULL, the upper layer is requesting a temp file. */
assert(isDelete && !syncDir);
rc = unixGetTempname(MAX_PATHNAME+2, zTmpname);
if( rc!=SQLITE_OK ){
return rc;
}
zName = zTmpname;
/* Generated temporary filenames are always double-zero terminated
** for use by sqlite3_uri_parameter(). */
assert( zName[strlen(zName)+1]==0 );
}
/* Determine the value of the flags parameter passed to POSIX function
** open(). These must be calculated even if open() is not called, as
** they may be stored as part of the file handle and used by the
** 'conch file' locking functions later on. */
if( isReadonly ) openFlags |= O_RDONLY;
if( isReadWrite ) openFlags |= O_RDWR;
if( isCreate ) openFlags |= O_CREAT;
if( isExclusive ) openFlags |= (O_EXCL|O_NOFOLLOW);
openFlags |= (O_LARGEFILE|O_BINARY);
if( fd<0 ){
mode_t openMode; /* Permissions to create file with */
uid_t uid; /* Userid for the file */
gid_t gid; /* Groupid for the file */
rc = findCreateFileMode(zName, flags, &openMode, &uid, &gid);
if( rc!=SQLITE_OK ){
assert( !p->pUnused );
assert( eType==SQLITE_OPEN_WAL || eType==SQLITE_OPEN_MAIN_JOURNAL );
return rc;
}
fd = robust_open(zName, openFlags, openMode);
OSTRACE(("OPENX %-3d %s 0%o\n", fd, zName, openFlags));
if( fd<0 && errno!=EISDIR && isReadWrite && !isExclusive ){
/* Failed to open the file for read/write access. Try read-only. */
flags &= ~(SQLITE_OPEN_READWRITE|SQLITE_OPEN_CREATE);
openFlags &= ~(O_RDWR|O_CREAT);
flags |= SQLITE_OPEN_READONLY;
openFlags |= O_RDONLY;
isReadonly = 1;
fd = robust_open(zName, openFlags, openMode);
}
if( fd<0 ){
rc = unixLogError(SQLITE_CANTOPEN_BKPT, "open", zName);
goto open_finished;
}
/* If this process is running as root and if creating a new rollback
** journal or WAL file, set the ownership of the journal or WAL to be
** the same as the original database.
*/
if( flags & (SQLITE_OPEN_WAL|SQLITE_OPEN_MAIN_JOURNAL) ){
osFchown(fd, uid, gid);
}
}
assert( fd>=0 );
if( pOutFlags ){
*pOutFlags = flags;
}
if( p->pUnused ){
p->pUnused->fd = fd;
p->pUnused->flags = flags;
}
if( isDelete ){
#if OS_VXWORKS
zPath = zName;
#elif defined(SQLITE_UNLINK_AFTER_CLOSE)
zPath = sqlite3_mprintf("%s", zName);
if( zPath==0 ){
robust_close(p, fd, __LINE__);
return SQLITE_NOMEM;
}
#else
osUnlink(zName);
#endif
}
#if SQLITE_ENABLE_LOCKING_STYLE
else{
p->openFlags = openFlags;
}
#endif
noLock = eType!=SQLITE_OPEN_MAIN_DB;
#if defined(__APPLE__) || SQLITE_ENABLE_LOCKING_STYLE
if( fstatfs(fd, &fsInfo) == -1 ){
((unixFile*)pFile)->lastErrno = errno;
robust_close(p, fd, __LINE__);
return SQLITE_IOERR_ACCESS;
}
if (0 == strncmp("msdos", fsInfo.f_fstypename, 5)) {
((unixFile*)pFile)->fsFlags |= SQLITE_FSFLAGS_IS_MSDOS;
}
#endif
/* Set up appropriate ctrlFlags */
if( isDelete ) ctrlFlags |= UNIXFILE_DELETE;
if( isReadonly ) ctrlFlags |= UNIXFILE_RDONLY;
if( noLock ) ctrlFlags |= UNIXFILE_NOLOCK;
if( syncDir ) ctrlFlags |= UNIXFILE_DIRSYNC;
if( flags & SQLITE_OPEN_URI ) ctrlFlags |= UNIXFILE_URI;
#if SQLITE_ENABLE_LOCKING_STYLE
#if SQLITE_PREFER_PROXY_LOCKING
isAutoProxy = 1;
#endif
if( isAutoProxy && (zPath!=NULL) && (!noLock) && pVfs->xOpen ){
char *envforce = getenv("SQLITE_FORCE_PROXY_LOCKING");
int useProxy = 0;
** exactly once for each successful call to xCreate().
*/
int (*xDestroy)(sqlite3_tokenizer *pTokenizer);
/*
** Create a tokenizer cursor to tokenize an input buffer. The caller
** is responsible for ensuring that the input buffer remains valid
** until the cursor is closed (using the xClose() method).
*/
int (*xOpen)(
sqlite3_tokenizer *pTokenizer, /* Tokenizer object */
const char *pInput, int nBytes, /* Input buffer */
sqlite3_tokenizer_cursor **ppCursor /* OUT: Created tokenizer cursor */
);
/*
** Destroy an existing tokenizer cursor. The fts3 module calls this
** method exactly once for each successful call to xOpen().
*/
int (*xClose)(sqlite3_tokenizer_cursor *pCursor);
/*
** Retrieve the next token from the tokenizer cursor pCursor. This
** method should either return SQLITE_OK and set the values of the
** "OUT" variables identified below, or SQLITE_DONE to indicate that
** the end of the buffer has been reached, or an SQLite error code.
**
** *ppToken should be set to point at a buffer containing the
** normalized version of the token (i.e. after any case-folding and/or
** stemming has been performed). *pnBytes should be set to the length
** of this buffer in bytes. The input text that generated the token is
** identified by the byte offsets returned in *piStartOffset and
** *piEndOffset. *piStartOffset should be set to the index of the first
** byte of the token in the input buffer. *piEndOffset should be set
** to the index of the first byte just past the end of the token in
** the input buffer.
**
** The buffer *ppToken is set to point at is managed by the tokenizer
** implementation. It is only required to be valid until the next call
** to xNext() or xClose().
*/
/* TODO(shess) current implementation requires pInput to be
** nul-terminated. This should either be fixed, or pInput/nBytes
** should be converted to zInput.
*/
int (*xNext)(
sqlite3_tokenizer_cursor *pCursor, /* Tokenizer cursor */
const char **ppToken, int *pnBytes, /* OUT: Normalized text for token */
int *piStartOffset, /* OUT: Byte offset of token in input buffer */
int *piEndOffset, /* OUT: Byte offset of end of token in input buffer */
int *piPosition /* OUT: Number of tokens returned before this one */
);
/***********************************************************************
** Methods below this point are only available if iVersion>=1.
*/
/*
** Configure the language id of a tokenizer cursor.
*/
int (*xLanguageid)(sqlite3_tokenizer_cursor *pCsr, int iLangid);
};
struct sqlite3_tokenizer {
const sqlite3_tokenizer_module *pModule; /* The module for this tokenizer */
/* Tokenizer implementations will typically add additional fields */
};
struct sqlite3_tokenizer_cursor {
sqlite3_tokenizer *pTokenizer; /* Tokenizer for this cursor. */
/* Tokenizer implementations will typically add additional fields */
};
int fts3_global_term_cnt(int iTerm, int iCol);
int fts3_term_cnt(int iTerm, int iCol);
#endif /* _FTS3_TOKENIZER_H_ */
/************** End of fts3_tokenizer.h **************************************/
/************** Continuing where we left off in fts3Int.h ********************/
/************** Include fts3_hash.h in the middle of fts3Int.h ***************/
/************** Begin file fts3_hash.h ***************************************/
/*
** 2001 September 22
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
*************************************************************************
** This is the header file for the generic hash-table implementation
** used in SQLite. We've modified it slightly to serve as a standalone
** hash table implementation for the full-text indexing module.
**
*/
#ifndef _FTS3_HASH_H_
#define _FTS3_HASH_H_
/* Forward declarations of structures. */
typedef struct Fts3Hash Fts3Hash;
typedef struct Fts3HashElem Fts3HashElem;
/* A complete hash table is an instance of the following structure.
** The internals of this structure are intended to be opaque -- client
** code should not attempt to access or modify the fields of this structure
** directly. Change this structure only by using the routines below.
** However, many of the "procedures" and "functions" for modifying and
** accessing this structure are really macros, so we can't really make
** this structure opaque.
*/
struct Fts3Hash {
char keyClass; /* HASH_INT, _POINTER, _STRING, _BINARY */
char copyKey; /* True if copy of key made on insert */
int count; /* Number of entries in this table */
Fts3HashElem *first; /* The first element of the array */
int htsize; /* Number of buckets in the hash table */
struct _fts3ht { /* the hash table */
SQLITE_PRIVATE int sqlite3Fts3Corrupt(void);
# define FTS_CORRUPT_VTAB sqlite3Fts3Corrupt()
#else
# define FTS_CORRUPT_VTAB SQLITE_CORRUPT_VTAB
#endif
typedef struct Fts3Table Fts3Table;
typedef struct Fts3Cursor Fts3Cursor;
typedef struct Fts3Expr Fts3Expr;
typedef struct Fts3Phrase Fts3Phrase;
typedef struct Fts3PhraseToken Fts3PhraseToken;
typedef struct Fts3Doclist Fts3Doclist;
typedef struct Fts3SegFilter Fts3SegFilter;
typedef struct Fts3DeferredToken Fts3DeferredToken;
typedef struct Fts3SegReader Fts3SegReader;
typedef struct Fts3MultiSegReader Fts3MultiSegReader;
/*
** A connection to a fulltext index is an instance of the following
** structure. The xCreate and xConnect methods create an instance
** of this structure and xDestroy and xDisconnect free that instance.
** All other methods receive a pointer to the structure as one of their
** arguments.
*/
struct Fts3Table {
sqlite3_vtab base; /* Base class used by SQLite core */
sqlite3 *db; /* The database connection */
const char *zDb; /* logical database name */
const char *zName; /* virtual table name */
int nColumn; /* number of named columns in virtual table */
char **azColumn; /* column names. malloced */
u8 *abNotindexed; /* True for 'notindexed' columns */
sqlite3_tokenizer *pTokenizer; /* tokenizer for inserts and queries */
char *zContentTbl; /* content=xxx option, or NULL */
char *zLanguageid; /* languageid=xxx option, or NULL */
int nAutoincrmerge; /* Value configured by 'automerge' */
u32 nLeafAdd; /* Number of leaf blocks added this trans */
/* Precompiled statements used by the implementation. Each of these
** statements is run and reset within a single virtual table API call.
*/
sqlite3_stmt *aStmt[40];
char *zReadExprlist;
char *zWriteExprlist;
int nNodeSize; /* Soft limit for node size */
u8 bFts4; /* True for FTS4, false for FTS3 */
u8 bHasStat; /* True if %_stat table exists (2==unknown) */
u8 bHasDocsize; /* True if %_docsize table exists */
u8 bDescIdx; /* True if doclists are in reverse order */
u8 bIgnoreSavepoint; /* True to ignore xSavepoint invocations */
int nPgsz; /* Page size for host database */
char *zSegmentsTbl; /* Name of %_segments table */
sqlite3_blob *pSegments; /* Blob handle open on %_segments table */
/*
** The following array of hash tables is used to buffer pending index
** updates during transactions. All pending updates buffered at any one
** time must share a common language-id (see the FTS4 langid= feature).
** The current language id is stored in variable iPrevLangid.
**
** A single FTS4 table may have multiple full-text indexes. For each index
** there is an entry in the aIndex[] array. Index 0 is an index of all the
** terms that appear in the document set. Each subsequent index in aIndex[]
** is an index of prefixes of a specific length.
**
** Variable nPendingData contains an estimate the memory consumed by the
** pending data structures, including hash table overhead, but not including
** malloc overhead. When nPendingData exceeds nMaxPendingData, all hash
** tables are flushed to disk. Variable iPrevDocid is the docid of the most
** recently inserted record.
*/
int nIndex; /* Size of aIndex[] */
struct Fts3Index {
int nPrefix; /* Prefix length (0 for main terms index) */
Fts3Hash hPending; /* Pending terms table for this index */
} *aIndex;
int nMaxPendingData; /* Max pending data before flush to disk */
int nPendingData; /* Current bytes of pending data */
sqlite_int64 iPrevDocid; /* Docid of most recently inserted document */
int iPrevLangid; /* Langid of recently inserted document */
#if defined(SQLITE_DEBUG) || defined(SQLITE_COVERAGE_TEST)
/* State variables used for validating that the transaction control
** methods of the virtual table are called at appropriate times. These
** values do not contribute to FTS functionality; they are used for
** verifying the operation of the SQLite core.
*/
int inTransaction; /* True after xBegin but before xCommit/xRollback */
int mxSavepoint; /* Largest valid xSavepoint integer */
#endif
#ifdef SQLITE_TEST
/* True to disable the incremental doclist optimization. This is controled
** by special insert command 'test-no-incr-doclist'. */
int bNoIncrDoclist;
#endif
};
/*
** When the core wants to read from the virtual table, it creates a
** virtual table cursor (an instance of the following structure) using
** the xOpen method. Cursors are destroyed using the xClose method.
*/
struct Fts3Cursor {
sqlite3_vtab_cursor base; /* Base class used by SQLite core */
i16 eSearch; /* Search strategy (see below) */
u8 isEof; /* True if at End Of Results */
u8 isRequireSeek; /* True if must seek pStmt to %_content row */
sqlite3_stmt *pStmt; /* Prepared statement in use by the cursor */
Fts3Expr *pExpr; /* Parsed MATCH query string */
int iLangid; /* Language being queried for */
int nPhrase; /* Number of matchable phrases in query */
Fts3DeferredToken *pDeferred; /* Deferred search tokens, if any */
sqlite3_int64 iPrevId; /* Previous id read from aDoclist */
char *pNextId; /* Pointer into the body of aDoclist */
char *aDoclist; /* List of docids for full-text queries */
int nDoclist; /* Size of buffer at aDoclist */
u8 bDesc; /* True to sort in descending order */
int eEvalmode; /* An FTS3_EVAL_XX constant */
int nRowAvg; /* Average size of database rows, in pages */
sqlite3_int64 nDoc; /* Documents in table */
i64 iMinDocid; /* Minimum docid to return */
i64 iMaxDocid; /* Maximum docid to return */
int isMatchinfoNeeded; /* True when aMatchinfo[] needs filling in */
u32 *aMatchinfo; /* Information about most recent match */
int nMatchinfo; /* Number of elements in aMatchinfo[] */
char *zMatchinfo; /* Matchinfo specification */
};
#define FTS3_EVAL_FILTER 0
#define FTS3_EVAL_NEXT 1
#define FTS3_EVAL_MATCHINFO 2
/*
** The Fts3Cursor.eSearch member is always set to one of the following.
** Actualy, Fts3Cursor.eSearch can be greater than or equal to
** FTS3_FULLTEXT_SEARCH. If so, then Fts3Cursor.eSearch - 2 is the index
** of the column to be searched. For example, in
**
** CREATE VIRTUAL TABLE ex1 USING fts3(a,b,c,d);
** SELECT docid FROM ex1 WHERE b MATCH 'one two three';
**
** Because the LHS of the MATCH operator is 2nd column "b",
** Fts3Cursor.eSearch will be set to FTS3_FULLTEXT_SEARCH+1. (+0 for a,
** +1 for b, +2 for c, +3 for d.) If the LHS of MATCH were "ex1"
** indicating that all columns should be searched,
** then eSearch would be set to FTS3_FULLTEXT_SEARCH+4.
*/
#define FTS3_FULLSCAN_SEARCH 0 /* Linear scan of %_content table */
#define FTS3_DOCID_SEARCH 1 /* Lookup by rowid on %_content table */
#define FTS3_FULLTEXT_SEARCH 2 /* Full-text index search */
/*
** The lower 16-bits of the sqlite3_index_info.idxNum value set by
** the xBestIndex() method contains the Fts3Cursor.eSearch value described
** above. The upper 16-bits contain a combination of the following
** bits, used to describe extra constraints on full-text searches.
*/
#define FTS3_HAVE_LANGID 0x00010000 /* languageid=? */
#define FTS3_HAVE_DOCID_GE 0x00020000 /* docid>=? */
#define FTS3_HAVE_DOCID_LE 0x00040000 /* docid<=? */
struct Fts3Doclist {
char *aAll; /* Array containing doclist (or NULL) */
int nAll; /* Size of a[] in bytes */
char *pNextDocid; /* Pointer to next docid */
sqlite3_int64 iDocid; /* Current docid (if pList!=0) */
int bFreeList; /* True if pList should be sqlite3_free()d */
char *pList; /* Pointer to position list following iDocid */
int nList; /* Length of position list */
const char *zFormat, /* Format string for SQL */
... /* Arguments to the format string */
){
va_list ap;
char *zSql;
if( *pRc ) return;
va_start(ap, zFormat);
zSql = sqlite3_vmprintf(zFormat, ap);
va_end(ap);
if( zSql==0 ){
*pRc = SQLITE_NOMEM;
}else{
*pRc = sqlite3_exec(db, zSql, 0, 0, 0);
sqlite3_free(zSql);
}
}
/*
** The xDestroy() virtual table method.
*/
static int fts3DestroyMethod(sqlite3_vtab *pVtab){
Fts3Table *p = (Fts3Table *)pVtab;
int rc = SQLITE_OK; /* Return code */
const char *zDb = p->zDb; /* Name of database (e.g. "main", "temp") */
sqlite3 *db = p->db; /* Database handle */
/* Drop the shadow tables */
if( p->zContentTbl==0 ){
fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_content'", zDb, p->zName);
}
fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_segments'", zDb,p->zName);
fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_segdir'", zDb, p->zName);
fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_docsize'", zDb, p->zName);
fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_stat'", zDb, p->zName);
/* If everything has worked, invoke fts3DisconnectMethod() to free the
** memory associated with the Fts3Table structure and return SQLITE_OK.
** Otherwise, return an SQLite error code.
*/
return (rc==SQLITE_OK ? fts3DisconnectMethod(pVtab) : rc);
}
/*
** Invoke sqlite3_declare_vtab() to declare the schema for the FTS3 table
** passed as the first argument. This is done as part of the xConnect()
** and xCreate() methods.
**
** If *pRc is non-zero when this function is called, it is a no-op.
** Otherwise, if an error occurs, an SQLite error code is stored in *pRc
** before returning.
*/
static void fts3DeclareVtab(int *pRc, Fts3Table *p){
if( *pRc==SQLITE_OK ){
int i; /* Iterator variable */
int rc; /* Return code */
char *zSql; /* SQL statement passed to declare_vtab() */
char *zCols; /* List of user defined columns */
const char *zLanguageid;
zLanguageid = (p->zLanguageid ? p->zLanguageid : "__langid");
sqlite3_vtab_config(p->db, SQLITE_VTAB_CONSTRAINT_SUPPORT, 1);
/* Create a list of user columns for the virtual table */
zCols = sqlite3_mprintf("%Q, ", p->azColumn[0]);
for(i=1; zCols && i<p->nColumn; i++){
zCols = sqlite3_mprintf("%z%Q, ", zCols, p->azColumn[i]);
}
/* Create the whole "CREATE TABLE" statement to pass to SQLite */
zSql = sqlite3_mprintf(
"CREATE TABLE x(%s %Q HIDDEN, docid HIDDEN, %Q HIDDEN)",
zCols, p->zName, zLanguageid
);
if( !zCols || !zSql ){
rc = SQLITE_NOMEM;
}else{
rc = sqlite3_declare_vtab(p->db, zSql);
}
sqlite3_free(zSql);
sqlite3_free(zCols);
*pRc = rc;
}
}
/*
** Create the %_stat table if it does not already exist.
*/
SQLITE_PRIVATE void sqlite3Fts3CreateStatTable(int *pRc, Fts3Table *p){
fts3DbExec(pRc, p->db,
"CREATE TABLE IF NOT EXISTS %Q.'%q_stat'"
"(id INTEGER PRIMARY KEY, value BLOB);",
p->zDb, p->zName
);
if( (*pRc)==SQLITE_OK ) p->bHasStat = 1;
}
/*
** Create the backing store tables (%_content, %_segments and %_segdir)
** required by the FTS3 table passed as the only argument. This is done
** as part of the vtab xCreate() method.
**
** If the p->bHasDocsize boolean is true (indicating that this is an
** FTS4 table, not an FTS3 table) then also create the %_docsize and
** %_stat tables required by FTS4.
*/
static int fts3CreateTables(Fts3Table *p){
int rc = SQLITE_OK; /* Return code */
int i; /* Iterator variable */
sqlite3 *db = p->db; /* The database connection */
if( p->zContentTbl==0 ){
const char *zLanguageid = p->zLanguageid;
char *zContentCols; /* Columns of %_content table */
/* Create a list of user columns for the content table */
zContentCols = sqlite3_mprintf("docid INTEGER PRIMARY KEY");
for(i=0; zContentCols && i<p->nColumn; i++){
char *z = p->azColumn[i];
zContentCols = sqlite3_mprintf("%z, 'c%d%q'", zContentCols, i, z);
}
if( zLanguageid && zContentCols ){
zContentCols = sqlite3_mprintf("%z, langid", zContentCols, zLanguageid);
}
if( zContentCols==0 ) rc = SQLITE_NOMEM;
/* Create the content table */
fts3DbExec(&rc, db,
"CREATE TABLE %Q.'%q_content'(%s)",
p->zDb, p->zName, zContentCols
);
sqlite3_free(zContentCols);
}
/* Create other tables */
fts3DbExec(&rc, db,
"CREATE TABLE %Q.'%q_segments'(blockid INTEGER PRIMARY KEY, block BLOB);",
p->zDb, p->zName
);
fts3DbExec(&rc, db,
"CREATE TABLE %Q.'%q_segdir'("
"level INTEGER,"
"idx INTEGER,"
"start_block INTEGER,"
"leaves_end_block INTEGER,"
"end_block INTEGER,"
"root BLOB,"
"PRIMARY KEY(level, idx)"
");",
p->zDb, p->zName
);
if( p->bHasDocsize ){
fts3DbExec(&rc, db,
"CREATE TABLE %Q.'%q_docsize'(docid INTEGER PRIMARY KEY, size BLOB);",
p->zDb, p->zName
);
}
assert( p->bHasStat==p->bFts4 );
if( p->bHasStat ){
sqlite3Fts3CreateStatTable(&rc, p);
}
return rc;
}
/*
** Store the current database page-size in bytes in p->nPgsz.
**
** If *pRc is non-zero when this function is called, it is a no-op.
** Otherwise, if an error occurs, an SQLite error code is stored in *pRc
** before returning.
*/
static void fts3DatabasePageSize(int *pRc, Fts3Table *p){
if( *pRc==SQLITE_OK ){
int rc; /* Return code */
char *zSql; /* SQL text "PRAGMA %Q.page_size" */
sqlite3_stmt *pStmt; /* Compiled "PRAGMA %Q.page_size" statement */
zSql = sqlite3_mprintf("PRAGMA %Q.page_size", p->zDb);
if( !zSql ){
rc = SQLITE_NOMEM;
}else{
rc = sqlite3_prepare(p->db, zSql, -1, &pStmt, 0);
if( rc==SQLITE_OK ){
*/
static char *fts3QuoteId(char const *zInput){
int nRet;
char *zRet;
nRet = 2 + (int)strlen(zInput)*2 + 1;
zRet = sqlite3_malloc(nRet);
if( zRet ){
int i;
char *z = zRet;
*(z++) = '"';
for(i=0; zInput[i]; i++){
if( zInput[i]=='"' ) *(z++) = '"';
*(z++) = zInput[i];
}
*(z++) = '"';
*(z++) = '\0';
}
return zRet;
}
/*
** Return a list of comma separated SQL expressions and a FROM clause that
** could be used in a SELECT statement such as the following:
**
** SELECT <list of expressions> FROM %_content AS x ...
**
** to return the docid, followed by each column of text data in order
** from left to write. If parameter zFunc is not NULL, then instead of
** being returned directly each column of text data is passed to an SQL
** function named zFunc first. For example, if zFunc is "unzip" and the
** table has the three user-defined columns "a", "b", and "c", the following
** string is returned:
**
** "docid, unzip(x.'a'), unzip(x.'b'), unzip(x.'c') FROM %_content AS x"
**
** The pointer returned points to a buffer allocated by sqlite3_malloc(). It
** is the responsibility of the caller to eventually free it.
**
** If *pRc is not SQLITE_OK when this function is called, it is a no-op (and
** a NULL pointer is returned). Otherwise, if an OOM error is encountered
** by this function, NULL is returned and *pRc is set to SQLITE_NOMEM. If
** no error occurs, *pRc is left unmodified.
*/
static char *fts3ReadExprList(Fts3Table *p, const char *zFunc, int *pRc){
char *zRet = 0;
char *zFree = 0;
char *zFunction;
int i;
if( p->zContentTbl==0 ){
if( !zFunc ){
zFunction = "";
}else{
zFree = zFunction = fts3QuoteId(zFunc);
}
fts3Appendf(pRc, &zRet, "docid");
for(i=0; i<p->nColumn; i++){
fts3Appendf(pRc, &zRet, ",%s(x.'c%d%q')", zFunction, i, p->azColumn[i]);
}
if( p->zLanguageid ){
fts3Appendf(pRc, &zRet, ", x.%Q", "langid");
}
sqlite3_free(zFree);
}else{
fts3Appendf(pRc, &zRet, "rowid");
for(i=0; i<p->nColumn; i++){
fts3Appendf(pRc, &zRet, ", x.'%q'", p->azColumn[i]);
}
if( p->zLanguageid ){
fts3Appendf(pRc, &zRet, ", x.%Q", p->zLanguageid);
}
}
fts3Appendf(pRc, &zRet, " FROM '%q'.'%q%s' AS x",
p->zDb,
(p->zContentTbl ? p->zContentTbl : p->zName),
(p->zContentTbl ? "" : "_content")
);
return zRet;
}
/*
** Return a list of N comma separated question marks, where N is the number
** of columns in the %_content table (one for the docid plus one for each
** user-defined text column).
**
** If argument zFunc is not NULL, then all but the first question mark
** is preceded by zFunc and an open bracket, and followed by a closed
** bracket. For example, if zFunc is "zip" and the FTS3 table has three
** user-defined text columns, the following string is returned:
**
** "?, zip(?), zip(?), zip(?)"
**
** The pointer returned points to a buffer allocated by sqlite3_malloc(). It
** is the responsibility of the caller to eventually free it.
**
** If *pRc is not SQLITE_OK when this function is called, it is a no-op (and
** a NULL pointer is returned). Otherwise, if an OOM error is encountered
** by this function, NULL is returned and *pRc is set to SQLITE_NOMEM. If
** no error occurs, *pRc is left unmodified.
*/
static char *fts3WriteExprList(Fts3Table *p, const char *zFunc, int *pRc){
char *zRet = 0;
char *zFree = 0;
char *zFunction;
int i;
if( !zFunc ){
zFunction = "";
}else{
zFree = zFunction = fts3QuoteId(zFunc);
}
fts3Appendf(pRc, &zRet, "?");
for(i=0; i<p->nColumn; i++){
fts3Appendf(pRc, &zRet, ",%s(?)", zFunction);
}
if( p->zLanguageid ){
fts3Appendf(pRc, &zRet, ", ?");
}
sqlite3_free(zFree);
return zRet;
}
pTokenizer->pModule->xDestroy(pTokenizer);
}
}else{
assert( p->pSegments==0 );
*ppVTab = &p->base;
}
return rc;
}
/*
** The xConnect() and xCreate() methods for the virtual table. All the
** work is done in function fts3InitVtab().
*/
static int fts3ConnectMethod(
sqlite3 *db, /* Database connection */
void *pAux, /* Pointer to tokenizer hash table */
int argc, /* Number of elements in argv array */
const char * const *argv, /* xCreate/xConnect argument array */
sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */
char **pzErr /* OUT: sqlite3_malloc'd error message */
){
return fts3InitVtab(0, db, pAux, argc, argv, ppVtab, pzErr);
}
static int fts3CreateMethod(
sqlite3 *db, /* Database connection */
void *pAux, /* Pointer to tokenizer hash table */
int argc, /* Number of elements in argv array */
const char * const *argv, /* xCreate/xConnect argument array */
sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */
char **pzErr /* OUT: sqlite3_malloc'd error message */
){
return fts3InitVtab(1, db, pAux, argc, argv, ppVtab, pzErr);
}
/*
** Set the pIdxInfo->estimatedRows variable to nRow. Unless this
** extension is currently being used by a version of SQLite too old to
** support estimatedRows. In that case this function is a no-op.
*/
static void fts3SetEstimatedRows(sqlite3_index_info *pIdxInfo, i64 nRow){
#if SQLITE_VERSION_NUMBER>=3008002
if( sqlite3_libversion_number()>=3008002 ){
pIdxInfo->estimatedRows = nRow;
}
#endif
}
/*
** Implementation of the xBestIndex method for FTS3 tables. There
** are three possible strategies, in order of preference:
**
** 1. Direct lookup by rowid or docid.
** 2. Full-text search using a MATCH operator on a non-docid column.
** 3. Linear scan of %_content table.
*/
static int fts3BestIndexMethod(sqlite3_vtab *pVTab, sqlite3_index_info *pInfo){
Fts3Table *p = (Fts3Table *)pVTab;
int i; /* Iterator variable */
int iCons = -1; /* Index of constraint to use */
int iLangidCons = -1; /* Index of langid=x constraint, if present */
int iDocidGe = -1; /* Index of docid>=x constraint, if present */
int iDocidLe = -1; /* Index of docid<=x constraint, if present */
int iIdx;
/* By default use a full table scan. This is an expensive option,
** so search through the constraints to see if a more efficient
** strategy is possible.
*/
pInfo->idxNum = FTS3_FULLSCAN_SEARCH;
pInfo->estimatedCost = 5000000;
for(i=0; i<pInfo->nConstraint; i++){
int bDocid; /* True if this constraint is on docid */
struct sqlite3_index_constraint *pCons = &pInfo->aConstraint[i];
if( pCons->usable==0 ){
if( pCons->op==SQLITE_INDEX_CONSTRAINT_MATCH ){
/* There exists an unusable MATCH constraint. This means that if
** the planner does elect to use the results of this call as part
** of the overall query plan the user will see an "unable to use
** function MATCH in the requested context" error. To discourage
** this, return a very high cost here. */
pInfo->idxNum = FTS3_FULLSCAN_SEARCH;
pInfo->estimatedCost = 1e50;
fts3SetEstimatedRows(pInfo, ((sqlite3_int64)1) << 50);
return SQLITE_OK;
}
continue;
}
bDocid = (pCons->iColumn<0 || pCons->iColumn==p->nColumn+1);
/* A direct lookup on the rowid or docid column. Assign a cost of 1.0. */
if( iCons<0 && pCons->op==SQLITE_INDEX_CONSTRAINT_EQ && bDocid ){
pInfo->idxNum = FTS3_DOCID_SEARCH;
pInfo->estimatedCost = 1.0;
iCons = i;
}
/* A MATCH constraint. Use a full-text search.
**
** If there is more than one MATCH constraint available, use the first
** one encountered. If there is both a MATCH constraint and a direct
** rowid/docid lookup, prefer the MATCH strategy. This is done even
** though the rowid/docid lookup is faster than a MATCH query, selecting
** it would lead to an "unable to use function MATCH in the requested
** context" error.
*/
if( pCons->op==SQLITE_INDEX_CONSTRAINT_MATCH
&& pCons->iColumn>=0 && pCons->iColumn<=p->nColumn
){
pInfo->idxNum = FTS3_FULLTEXT_SEARCH + pCons->iColumn;
pInfo->estimatedCost = 2.0;
iCons = i;
}
/* Equality constraint on the langid column */
if( pCons->op==SQLITE_INDEX_CONSTRAINT_EQ
&& pCons->iColumn==p->nColumn + 2
){
iLangidCons = i;
}
if( bDocid ){
switch( pCons->op ){
case SQLITE_INDEX_CONSTRAINT_GE:
case SQLITE_INDEX_CONSTRAINT_GT:
iDocidGe = i;
break;
case SQLITE_INDEX_CONSTRAINT_LE:
case SQLITE_INDEX_CONSTRAINT_LT:
iDocidLe = i;
break;
}
}
}
iIdx = 1;
if( iCons>=0 ){
pInfo->aConstraintUsage[iCons].argvIndex = iIdx++;
pInfo->aConstraintUsage[iCons].omit = 1;
}
if( iLangidCons>=0 ){
pInfo->idxNum |= FTS3_HAVE_LANGID;
pInfo->aConstraintUsage[iLangidCons].argvIndex = iIdx++;
}
if( iDocidGe>=0 ){
pInfo->idxNum |= FTS3_HAVE_DOCID_GE;
pInfo->aConstraintUsage[iDocidGe].argvIndex = iIdx++;
}
if( iDocidLe>=0 ){
pInfo->idxNum |= FTS3_HAVE_DOCID_LE;
pInfo->aConstraintUsage[iDocidLe].argvIndex = iIdx++;
}
/* Regardless of the strategy selected, FTS can deliver rows in rowid (or
** docid) order. Both ascending and descending are possible.
*/
if( pInfo->nOrderBy==1 ){
struct sqlite3_index_orderby *pOrder = &pInfo->aOrderBy[0];
if( pOrder->iColumn<0 || pOrder->iColumn==p->nColumn+1 ){
if( pOrder->desc ){
pInfo->idxStr = "DESC";
}else{
pInfo->idxStr = "ASC";
}
pInfo->orderByConsumed = 1;
}
}
assert( p->pSegments==0 );
return SQLITE_OK;
}
/*
** Implementation of xOpen method.
*/
static int fts3OpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){
sqlite3_vtab_cursor *pCsr; /* Allocated cursor */
UNUSED_PARAMETER(pVTab);
/* Allocate a buffer large enough for an Fts3Cursor structure. If the
** allocation succeeds, zero it and return SQLITE_OK. Otherwise,
** if the allocation fails, return SQLITE_NOMEM.
*/
*ppCsr = pCsr = (sqlite3_vtab_cursor *)sqlite3_malloc(sizeof(Fts3Cursor));
if( !pCsr ){
return SQLITE_NOMEM;
}
memset(pCsr, 0, sizeof(Fts3Cursor));
return SQLITE_OK;
}
/*
** Close the cursor. For additional information see the documentation
** on the xClose method of the virtual table interface.
*/
static int fts3CloseMethod(sqlite3_vtab_cursor *pCursor){
Fts3Cursor *pCsr = (Fts3Cursor *)pCursor;
assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
sqlite3_finalize(pCsr->pStmt);
sqlite3Fts3ExprFree(pCsr->pExpr);
sqlite3Fts3FreeDeferredTokens(pCsr);
sqlite3_free(pCsr->aDoclist);
pTS->anOutput[iOut] = nMerge;
break;
}else{
char *aNew;
int nNew;
int rc = fts3DoclistOrMerge(p->bDescIdx, aMerge, nMerge,
pTS->aaOutput[iOut], pTS->anOutput[iOut], &aNew, &nNew
);
if( rc!=SQLITE_OK ){
if( aMerge!=aDoclist ) sqlite3_free(aMerge);
return rc;
}
if( aMerge!=aDoclist ) sqlite3_free(aMerge);
sqlite3_free(pTS->aaOutput[iOut]);
pTS->aaOutput[iOut] = 0;
aMerge = aNew;
nMerge = nNew;
if( (iOut+1)==SizeofArray(pTS->aaOutput) ){
pTS->aaOutput[iOut] = aMerge;
pTS->anOutput[iOut] = nMerge;
}
}
}
}
return SQLITE_OK;
}
/*
** Append SegReader object pNew to the end of the pCsr->apSegment[] array.
*/
static int fts3SegReaderCursorAppend(
Fts3MultiSegReader *pCsr,
Fts3SegReader *pNew
){
if( (pCsr->nSegment%16)==0 ){
Fts3SegReader **apNew;
int nByte = (pCsr->nSegment + 16)*sizeof(Fts3SegReader*);
apNew = (Fts3SegReader **)sqlite3_realloc(pCsr->apSegment, nByte);
if( !apNew ){
sqlite3Fts3SegReaderFree(pNew);
return SQLITE_NOMEM;
}
pCsr->apSegment = apNew;
}
pCsr->apSegment[pCsr->nSegment++] = pNew;
return SQLITE_OK;
}
/*
** Add seg-reader objects to the Fts3MultiSegReader object passed as the
** 8th argument.
**
** This function returns SQLITE_OK if successful, or an SQLite error code
** otherwise.
*/
static int fts3SegReaderCursor(
Fts3Table *p, /* FTS3 table handle */
int iLangid, /* Language id */
int iIndex, /* Index to search (from 0 to p->nIndex-1) */
int iLevel, /* Level of segments to scan */
const char *zTerm, /* Term to query for */
int nTerm, /* Size of zTerm in bytes */
int isPrefix, /* True for a prefix search */
int isScan, /* True to scan from zTerm to EOF */
Fts3MultiSegReader *pCsr /* Cursor object to populate */
){
int rc = SQLITE_OK; /* Error code */
sqlite3_stmt *pStmt = 0; /* Statement to iterate through segments */
int rc2; /* Result of sqlite3_reset() */
/* If iLevel is less than 0 and this is not a scan, include a seg-reader
** for the pending-terms. If this is a scan, then this call must be being
** made by an fts4aux module, not an FTS table. In this case calling
** Fts3SegReaderPending might segfault, as the data structures used by
** fts4aux are not completely populated. So it's easiest to filter these
** calls out here. */
if( iLevel<0 && p->aIndex ){
Fts3SegReader *pSeg = 0;
rc = sqlite3Fts3SegReaderPending(p, iIndex, zTerm, nTerm, isPrefix, &pSeg);
if( rc==SQLITE_OK && pSeg ){
rc = fts3SegReaderCursorAppend(pCsr, pSeg);
}
}
if( iLevel!=FTS3_SEGCURSOR_PENDING ){
if( rc==SQLITE_OK ){
rc = sqlite3Fts3AllSegdirs(p, iLangid, iIndex, iLevel, &pStmt);
}
while( rc==SQLITE_OK && SQLITE_ROW==(rc = sqlite3_step(pStmt)) ){
Fts3SegReader *pSeg = 0;
/* Read the values returned by the SELECT into local variables. */
sqlite3_int64 iStartBlock = sqlite3_column_int64(pStmt, 1);
sqlite3_int64 iLeavesEndBlock = sqlite3_column_int64(pStmt, 2);
sqlite3_int64 iEndBlock = sqlite3_column_int64(pStmt, 3);
int nRoot = sqlite3_column_bytes(pStmt, 4);
char const *zRoot = sqlite3_column_blob(pStmt, 4);
/* If zTerm is not NULL, and this segment is not stored entirely on its
** root node, the range of leaves scanned can be reduced. Do this. */
if( iStartBlock && zTerm ){
sqlite3_int64 *pi = (isPrefix ? &iLeavesEndBlock : 0);
rc = fts3SelectLeaf(p, zTerm, nTerm, zRoot, nRoot, &iStartBlock, pi);
if( rc!=SQLITE_OK ) goto finished;
if( isPrefix==0 && isScan==0 ) iLeavesEndBlock = iStartBlock;
}
rc = sqlite3Fts3SegReaderNew(pCsr->nSegment+1,
(isPrefix==0 && isScan==0),
iStartBlock, iLeavesEndBlock,
iEndBlock, zRoot, nRoot, &pSeg
);
if( rc!=SQLITE_OK ) goto finished;
rc = fts3SegReaderCursorAppend(pCsr, pSeg);
}
}
finished:
rc2 = sqlite3_reset(pStmt);
if( rc==SQLITE_DONE ) rc = rc2;
return rc;
}
/*
** Set up a cursor object for iterating through a full-text index or a
** single level therein.
*/
SQLITE_PRIVATE int sqlite3Fts3SegReaderCursor(
Fts3Table *p, /* FTS3 table handle */
int iLangid, /* Language-id to search */
int iIndex, /* Index to search (from 0 to p->nIndex-1) */
int iLevel, /* Level of segments to scan */
const char *zTerm, /* Term to query for */
int nTerm, /* Size of zTerm in bytes */
int isPrefix, /* True for a prefix search */
int isScan, /* True to scan from zTerm to EOF */
Fts3MultiSegReader *pCsr /* Cursor object to populate */
){
assert( iIndex>=0 && iIndex<p->nIndex );
assert( iLevel==FTS3_SEGCURSOR_ALL
|| iLevel==FTS3_SEGCURSOR_PENDING
|| iLevel>=0
);
assert( iLevel<FTS3_SEGDIR_MAXLEVEL );
assert( FTS3_SEGCURSOR_ALL<0 && FTS3_SEGCURSOR_PENDING<0 );
assert( isPrefix==0 || isScan==0 );
memset(pCsr, 0, sizeof(Fts3MultiSegReader));
return fts3SegReaderCursor(
p, iLangid, iIndex, iLevel, zTerm, nTerm, isPrefix, isScan, pCsr
);
}
/*
** In addition to its current configuration, have the Fts3MultiSegReader
** passed as the 4th argument also scan the doclist for term zTerm/nTerm.
**
** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
*/
static int fts3SegReaderCursorAddZero(
Fts3Table *p, /* FTS virtual table handle */
int iLangid,
const char *zTerm, /* Term to scan doclist of */
int nTerm, /* Number of bytes in zTerm */
Fts3MultiSegReader *pCsr /* Fts3MultiSegReader to modify */
){
return fts3SegReaderCursor(p,
iLangid, 0, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 0, 0,pCsr
);
}
/*
** Open an Fts3MultiSegReader to scan the doclist for term zTerm/nTerm. Or,
** if isPrefix is true, to scan the doclist for all terms for which
** zTerm/nTerm is a prefix. If successful, return SQLITE_OK and write
** a pointer to the new Fts3MultiSegReader to *ppSegcsr. Otherwise, return
** an SQLite error code.
**
** It is the responsibility of the caller to free this object by eventually
** passing it to fts3SegReaderCursorFree()
**
** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
** Output parameter *ppSegcsr is set to 0 if an error occurs.
*/
static int fts3TermSegReaderCursor(
Fts3Cursor *pCsr, /* Virtual table cursor handle */
const char *zTerm, /* Term to query for */
int nTerm, /* Size of zTerm in bytes */
int isPrefix, /* True for a prefix search */
Fts3MultiSegReader **ppSegcsr /* OUT: Allocated seg-reader cursor */
){
Fts3MultiSegReader *pSegcsr; /* Object to allocate and return */
int rc = SQLITE_NOMEM; /* Return code */
pSegcsr = sqlite3_malloc(sizeof(Fts3MultiSegReader));
if( pSegcsr ){
int i;
int bFound = 0; /* True once an index has been found */
Fts3Table *p = (Fts3Table *)pCsr->base.pVtab;
if( isPrefix ){
for(i=1; bFound==0 && i<p->nIndex; i++){
if( p->aIndex[i].nPrefix==nTerm ){
bFound = 1;
rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid,
i, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 0, 0, pSegcsr
);
pSegcsr->bLookup = 1;
}
}
for(i=1; bFound==0 && i<p->nIndex; i++){
if( p->aIndex[i].nPrefix==nTerm+1 ){
bFound = 1;
rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid,
i, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 1, 0, pSegcsr
);
if( rc==SQLITE_OK ){
rc = fts3SegReaderCursorAddZero(
p, pCsr->iLangid, zTerm, nTerm, pSegcsr
);
}
}
}
}
if( bFound==0 ){
rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid,
0, FTS3_SEGCURSOR_ALL, zTerm, nTerm, isPrefix, 0, pSegcsr
);
pSegcsr->bLookup = !isPrefix;
}
}
*ppSegcsr = pSegcsr;
return rc;
}
/*
** Free an Fts3MultiSegReader allocated by fts3TermSegReaderCursor().
*/
static void fts3SegReaderCursorFree(Fts3MultiSegReader *pSegcsr){
sqlite3Fts3SegReaderFinish(pSegcsr);
sqlite3_free(pSegcsr);
}
/*
** This function retrieves the doclist for the specified term (or term
** prefix) from the database.
*/
static int fts3TermSelect(
Fts3Table *p, /* Virtual table handle */
Fts3PhraseToken *pTok, /* Token to query for */
int iColumn, /* Column to query (or -ve for all columns) */
int *pnOut, /* OUT: Size of buffer at *ppOut */
char **ppOut /* OUT: Malloced result buffer */
){
int rc; /* Return code */
Fts3MultiSegReader *pSegcsr; /* Seg-reader cursor for this term */
TermSelect tsc; /* Object for pair-wise doclist merging */
Fts3SegFilter filter; /* Segment term filter configuration */
pSegcsr = pTok->pSegcsr;
memset(&tsc, 0, sizeof(TermSelect));
filter.flags = FTS3_SEGMENT_IGNORE_EMPTY | FTS3_SEGMENT_REQUIRE_POS
| (pTok->isPrefix ? FTS3_SEGMENT_PREFIX : 0)
| (pTok->bFirst ? FTS3_SEGMENT_FIRST : 0)
| (iColumn<p->nColumn ? FTS3_SEGMENT_COLUMN_FILTER : 0);
filter.iCol = iColumn;
filter.zTerm = pTok->z;
filter.nTerm = pTok->n;
rc = sqlite3Fts3SegReaderStart(p, pSegcsr, &filter);
while( SQLITE_OK==rc
&& SQLITE_ROW==(rc = sqlite3Fts3SegReaderStep(p, pSegcsr))
){
rc = fts3TermSelectMerge(p, &tsc, pSegcsr->aDoclist, pSegcsr->nDoclist);
}
if( rc==SQLITE_OK ){
rc = fts3TermSelectFinishMerge(p, &tsc);
}
if( rc==SQLITE_OK ){
*ppOut = tsc.aaOutput[0];
*pnOut = tsc.anOutput[0];
}else{
int i;
return rc;
}
/*
** The following are copied from sqliteInt.h.
**
** Constants for the largest and smallest possible 64-bit signed integers.
** These macros are designed to work correctly on both 32-bit and 64-bit
** compilers.
*/
#ifndef SQLITE_AMALGAMATION
# define LARGEST_INT64 (0xffffffff|(((sqlite3_int64)0x7fffffff)<<32))
# define SMALLEST_INT64 (((sqlite3_int64)-1) - LARGEST_INT64)
#endif
/*
** If the numeric type of argument pVal is "integer", then return it
** converted to a 64-bit signed integer. Otherwise, return a copy of
** the second parameter, iDefault.
*/
static sqlite3_int64 fts3DocidRange(sqlite3_value *pVal, i64 iDefault){
if( pVal ){
int eType = sqlite3_value_numeric_type(pVal);
if( eType==SQLITE_INTEGER ){
return sqlite3_value_int64(pVal);
}
}
return iDefault;
}
/*
** This is the xFilter interface for the virtual table. See
** the virtual table xFilter method documentation for additional
** information.
**
** If idxNum==FTS3_FULLSCAN_SEARCH then do a full table scan against
** the %_content table.
**
** If idxNum==FTS3_DOCID_SEARCH then do a docid lookup for a single entry
** in the %_content table.
**
** If idxNum>=FTS3_FULLTEXT_SEARCH then use the full text index. The
** column on the left-hand side of the MATCH operator is column
** number idxNum-FTS3_FULLTEXT_SEARCH, 0 indexed. argv[0] is the right-hand
** side of the MATCH operator.
*/
static int fts3FilterMethod(
sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */
int idxNum, /* Strategy index */
const char *idxStr, /* Unused */
int nVal, /* Number of elements in apVal */
sqlite3_value **apVal /* Arguments for the indexing scheme */
){
int rc = SQLITE_OK;
char *zSql; /* SQL statement used to access %_content */
int eSearch;
Fts3Table *p = (Fts3Table *)pCursor->pVtab;
Fts3Cursor *pCsr = (Fts3Cursor *)pCursor;
sqlite3_value *pCons = 0; /* The MATCH or rowid constraint, if any */
sqlite3_value *pLangid = 0; /* The "langid = ?" constraint, if any */
sqlite3_value *pDocidGe = 0; /* The "docid >= ?" constraint, if any */
sqlite3_value *pDocidLe = 0; /* The "docid <= ?" constraint, if any */
int iIdx;
UNUSED_PARAMETER(idxStr);
UNUSED_PARAMETER(nVal);
eSearch = (idxNum & 0x0000FFFF);
assert( eSearch>=0 && eSearch<=(FTS3_FULLTEXT_SEARCH+p->nColumn) );
assert( p->pSegments==0 );
/* Collect arguments into local variables */
iIdx = 0;
if( eSearch!=FTS3_FULLSCAN_SEARCH ) pCons = apVal[iIdx++];
if( idxNum & FTS3_HAVE_LANGID ) pLangid = apVal[iIdx++];
if( idxNum & FTS3_HAVE_DOCID_GE ) pDocidGe = apVal[iIdx++];
if( idxNum & FTS3_HAVE_DOCID_LE ) pDocidLe = apVal[iIdx++];
assert( iIdx==nVal );
/* In case the cursor has been used before, clear it now. */
sqlite3_finalize(pCsr->pStmt);
sqlite3_free(pCsr->aDoclist);
sqlite3_free(pCsr->aMatchinfo);
sqlite3Fts3ExprFree(pCsr->pExpr);
memset(&pCursor[1], 0, sizeof(Fts3Cursor)-sizeof(sqlite3_vtab_cursor));
/* Set the lower and upper bounds on docids to return */
pCsr->iMinDocid = fts3DocidRange(pDocidGe, SMALLEST_INT64);
pCsr->iMaxDocid = fts3DocidRange(pDocidLe, LARGEST_INT64);
if( idxStr ){
pCsr->bDesc = (idxStr[0]=='D');
}else{
pCsr->bDesc = p->bDescIdx;
}
pCsr->eSearch = (i16)eSearch;
if( eSearch!=FTS3_DOCID_SEARCH && eSearch!=FTS3_FULLSCAN_SEARCH ){
int iCol = eSearch-FTS3_FULLTEXT_SEARCH;
const char *zQuery = (const char *)sqlite3_value_text(pCons);
if( zQuery==0 && sqlite3_value_type(pCons)!=SQLITE_NULL ){
return SQLITE_NOMEM;
}
pCsr->iLangid = 0;
if( pLangid ) pCsr->iLangid = sqlite3_value_int(pLangid);
assert( p->base.zErrMsg==0 );
rc = sqlite3Fts3ExprParse(p->pTokenizer, pCsr->iLangid,
p->azColumn, p->bFts4, p->nColumn, iCol, zQuery, -1, &pCsr->pExpr,
&p->base.zErrMsg
);
if( rc!=SQLITE_OK ){
return rc;
}
rc = fts3EvalStart(pCsr);
sqlite3Fts3SegmentsClose(p);
if( rc!=SQLITE_OK ) return rc;
pCsr->pNextId = pCsr->aDoclist;
pCsr->iPrevId = 0;
}
/* Compile a SELECT statement for this cursor. For a full-table-scan, the
** statement loops through all rows of the %_content table. For a
** full-text query or docid lookup, the statement retrieves a single
** row by docid.
*/
if( eSearch==FTS3_FULLSCAN_SEARCH ){
zSql = sqlite3_mprintf(
"SELECT %s ORDER BY rowid %s",
p->zReadExprlist, (pCsr->bDesc ? "DESC" : "ASC")
);
if( zSql ){
rc = sqlite3_prepare_v2(p->db, zSql, -1, &pCsr->pStmt, 0);
sqlite3_free(zSql);
}else{
rc = SQLITE_NOMEM;
}
}else if( eSearch==FTS3_DOCID_SEARCH ){
rc = fts3CursorSeekStmt(pCsr, &pCsr->pStmt);
if( rc==SQLITE_OK ){
rc = sqlite3_bind_value(pCsr->pStmt, 1, pCons);
}
}
if( rc!=SQLITE_OK ) return rc;
return fts3NextMethod(pCursor);
}
/*
** This is the xEof method of the virtual table. SQLite calls this
** routine to find out if it has reached the end of a result set.
*/
static int fts3EofMethod(sqlite3_vtab_cursor *pCursor){
return ((Fts3Cursor *)pCursor)->isEof;
}
/*
** This is the xRowid method. The SQLite core calls this routine to
** retrieve the rowid for the current row of the result set. fts3
** exposes %_content.docid as the rowid for the virtual table. The
** rowid should be written to *pRowid.
*/
static int fts3RowidMethod(sqlite3_vtab_cursor *pCursor, sqlite_int64 *pRowid){
Fts3Cursor *pCsr = (Fts3Cursor *) pCursor;
*pRowid = pCsr->iPrevId;
return SQLITE_OK;
}
/*
** This is the xColumn method, called by SQLite to request a value from
** the row that the supplied cursor currently points to.
**
** If:
**
** (iCol < p->nColumn) -> The value of the iCol'th user column.
** (iCol == p->nColumn) -> Magic column with the same name as the table.
** (iCol == p->nColumn+1) -> Docid column
** (iCol == p->nColumn+2) -> Langid column
*/
static int fts3ColumnMethod(
sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */
sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */
int iCol /* Index of column to read value from */
){
int rc = SQLITE_OK; /* Return Code */
Fts3Cursor *pCsr = (Fts3Cursor *) pCursor;
Fts3Table *p = (Fts3Table *)pCursor->pVtab;
/* The column value supplied by SQLite must be in range. */
assert( iCol>=0 && iCol<=p->nColumn+2 );
if( iCol==p->nColumn+1 ){
/* This call is a request for the "docid" column. Since "docid" is an
** alias for "rowid", use the xRowid() method to obtain the value.
*/
sqlite3_result_int64(pCtx, pCsr->iPrevId);
}else if( iCol==p->nColumn ){
/* The extra column whose name is the same as the table.
** Return a blob which is a pointer to the cursor. */
sqlite3_result_blob(pCtx, &pCsr, sizeof(pCsr), SQLITE_TRANSIENT);
}else if( iCol==p->nColumn+2 && pCsr->pExpr ){
sqlite3_result_int64(pCtx, pCsr->iLangid);
}else{
/* The requested column is either a user column (one that contains
** indexed data), or the language-id column. */
rc = fts3CursorSeek(0, pCsr);
if( rc==SQLITE_OK ){
if( iCol==p->nColumn+2 ){
int iLangid = 0;
if( p->zLanguageid ){
iLangid = sqlite3_column_int(pCsr->pStmt, p->nColumn+1);
}
sqlite3_result_int(pCtx, iLangid);
}else if( sqlite3_data_count(pCsr->pStmt)>(iCol+1) ){
sqlite3_result_value(pCtx, sqlite3_column_value(pCsr->pStmt, iCol+1));
}
}
}
assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
return rc;
}
/*
** This function is the implementation of the xUpdate callback used by
** FTS3 virtual tables. It is invoked by SQLite each time a row is to be
** inserted, updated or deleted.
*/
static int fts3UpdateMethod(
sqlite3_vtab *pVtab, /* Virtual table handle */
int nArg, /* Size of argument array */
sqlite3_value **apVal, /* Array of arguments */
sqlite_int64 *pRowid /* OUT: The affected (or effected) rowid */
){
return sqlite3Fts3UpdateMethod(pVtab, nArg, apVal, pRowid);
}
/*
** Implementation of xSync() method. Flush the contents of the pending-terms
** hash-table to the database.
*/
static int fts3SyncMethod(sqlite3_vtab *pVtab){
/* Following an incremental-merge operation, assuming that the input
** segments are not completely consumed (the usual case), they are updated
** in place to remove the entries that have already been merged. This
** involves updating the leaf block that contains the smallest unmerged
** entry and each block (if any) between the leaf and the root node. So
** if the height of the input segment b-trees is N, and input segments
** are merged eight at a time, updating the input segments at the end
** of an incremental-merge requires writing (8*(1+N)) blocks. N is usually
** small - often between 0 and 2. So the overhead of the incremental
** merge is somewhere between 8 and 24 blocks. To avoid this overhead
** dwarfing the actual productive work accomplished, the incremental merge
** is only attempted if it will write at least 64 leaf blocks. Hence
** nMinMerge.
**
** Of course, updating the input segments also involves deleting a bunch
** of blocks from the segments table. But this is not considered overhead
** as it would also be required by a crisis-merge that used the same input
** segments.
*/
const u32 nMinMerge = 64; /* Minimum amount of incr-merge work to do */
Fts3Table *p = (Fts3Table*)pVtab;
int rc = sqlite3Fts3PendingTermsFlush(p);
if( rc==SQLITE_OK
&& p->nLeafAdd>(nMinMerge/16)
&& p->nAutoincrmerge && p->nAutoincrmerge!=0xff
){
int mxLevel = 0; /* Maximum relative level value in db */
int A; /* Incr-merge parameter A */
*/
#ifdef SQLITE_DEBUG
SQLITE_PRIVATE int sqlite3Fts3Corrupt(){
return SQLITE_CORRUPT_VTAB;
}
#endif
#if !SQLITE_CORE
/*
** Initialize API pointer table, if required.
*/
#ifdef _WIN32
__declspec(dllexport)
#endif
SQLITE_API int sqlite3_fts3_init(
sqlite3 *db,
char **pzErrMsg,
const sqlite3_api_routines *pApi
){
SQLITE_EXTENSION_INIT2(pApi)
return sqlite3Fts3Init(db);
}
#endif
#endif
/************** End of fts3.c ************************************************/
/************** Begin file fts3_aux.c ****************************************/
/*
** 2011 Jan 27
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
******************************************************************************
**
*/
#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
/* #include <string.h> */
/* #include <assert.h> */
typedef struct Fts3auxTable Fts3auxTable;
typedef struct Fts3auxCursor Fts3auxCursor;
struct Fts3auxTable {
sqlite3_vtab base; /* Base class used by SQLite core */
Fts3Table *pFts3Tab;
};
struct Fts3auxCursor {
sqlite3_vtab_cursor base; /* Base class used by SQLite core */
Fts3MultiSegReader csr; /* Must be right after "base" */
Fts3SegFilter filter;
char *zStop;
int nStop; /* Byte-length of string zStop */
int iLangid; /* Language id to query */
int isEof; /* True if cursor is at EOF */
sqlite3_int64 iRowid; /* Current rowid */
int iCol; /* Current value of 'col' column */
int nStat; /* Size of aStat[] array */
struct Fts3auxColstats {
sqlite3_int64 nDoc; /* 'documents' values for current csr row */
sqlite3_int64 nOcc; /* 'occurrences' values for current csr row */
} *aStat;
};
/*
** Schema of the terms table.
*/
#define FTS3_AUX_SCHEMA \
"CREATE TABLE x(term, col, documents, occurrences, languageid HIDDEN)"
/*
** This function does all the work for both the xConnect and xCreate methods.
** These tables have no persistent representation of their own, so xConnect
** and xCreate are identical operations.
*/
static int fts3auxConnectMethod(
sqlite3 *db, /* Database connection */
void *pUnused, /* Unused */
int argc, /* Number of elements in argv array */
const char * const *argv, /* xCreate/xConnect argument array */
sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */
char **pzErr /* OUT: sqlite3_malloc'd error message */
){
char const *zDb; /* Name of database (e.g. "main") */
char const *zFts3; /* Name of fts3 table */
int nDb; /* Result of strlen(zDb) */
int nFts3; /* Result of strlen(zFts3) */
int nByte; /* Bytes of space to allocate here */
int rc; /* value returned by declare_vtab() */
Fts3auxTable *p; /* Virtual table object to return */
UNUSED_PARAMETER(pUnused);
/* The user should invoke this in one of two forms:
**
** CREATE VIRTUAL TABLE xxx USING fts4aux(fts4-table);
** CREATE VIRTUAL TABLE xxx USING fts4aux(fts4-table-db, fts4-table);
*/
if( argc!=4 && argc!=5 ) goto bad_args;
zDb = argv[1];
nDb = (int)strlen(zDb);
if( argc==5 ){
if( nDb==4 && 0==sqlite3_strnicmp("temp", zDb, 4) ){
zDb = argv[3];
nDb = (int)strlen(zDb);
zFts3 = argv[4];
}else{
goto bad_args;
}
}else{
zFts3 = argv[3];
}
rc = sqlite3_declare_vtab(db, FTS3_AUX_SCHEMA);
if( rc!=SQLITE_OK ) return rc;
nByte = sizeof(Fts3auxTable) + sizeof(Fts3Table) + nDb + nFts3 + 2;
p = (Fts3auxTable *)sqlite3_malloc(nByte);
if( !p ) return SQLITE_NOMEM;
memset(p, 0, nByte);
p->pFts3Tab = (Fts3Table *)&p[1];
p->pFts3Tab->zDb = (char *)&p->pFts3Tab[1];
p->pFts3Tab->zName = &p->pFts3Tab->zDb[nDb+1];
p->pFts3Tab->db = db;
p->pFts3Tab->nIndex = 1;
memcpy((char *)p->pFts3Tab->zDb, zDb, nDb);
memcpy((char *)p->pFts3Tab->zName, zFts3, nFts3);
sqlite3Fts3Dequote((char *)p->pFts3Tab->zName);
*ppVtab = (sqlite3_vtab *)p;
return SQLITE_OK;
bad_args:
*pzErr = sqlite3_mprintf("invalid arguments to fts4aux constructor");
return SQLITE_ERROR;
}
/*
** This function does the work for both the xDisconnect and xDestroy methods.
** These tables have no persistent representation of their own, so xDisconnect
** and xDestroy are identical operations.
*/
static int fts3auxDisconnectMethod(sqlite3_vtab *pVtab){
Fts3auxTable *p = (Fts3auxTable *)pVtab;
Fts3Table *pFts3 = p->pFts3Tab;
int i;
/* Free any prepared statements held */
for(i=0; i<SizeofArray(pFts3->aStmt); i++){
sqlite3_finalize(pFts3->aStmt[i]);
}
sqlite3_free(pFts3->zSegmentsTbl);
sqlite3_free(p);
return SQLITE_OK;
}
#define FTS4AUX_EQ_CONSTRAINT 1
#define FTS4AUX_GE_CONSTRAINT 2
#define FTS4AUX_LE_CONSTRAINT 4
/*
** xBestIndex - Analyze a WHERE and ORDER BY clause.
*/
static int fts3auxBestIndexMethod(
sqlite3_vtab *pVTab,
sqlite3_index_info *pInfo
){
int i;
int iEq = -1;
int iGe = -1;
int iLe = -1;
int iLangid = -1;
int iNext = 1; /* Next free argvIndex value */
UNUSED_PARAMETER(pVTab);
/* This vtab delivers always results in "ORDER BY term ASC" order. */
if( pInfo->nOrderBy==1
&& pInfo->aOrderBy[0].iColumn==0
&& pInfo->aOrderBy[0].desc==0
){
pInfo->orderByConsumed = 1;
}
/* Search for equality and range constraints on the "term" column.
** And equality constraints on the hidden "languageid" column. */
for(i=0; i<pInfo->nConstraint; i++){
if( pInfo->aConstraint[i].usable ){
int op = pInfo->aConstraint[i].op;
int iCol = pInfo->aConstraint[i].iColumn;
if( iCol==0 ){
if( op==SQLITE_INDEX_CONSTRAINT_EQ ) iEq = i;
if( op==SQLITE_INDEX_CONSTRAINT_LT ) iLe = i;
if( op==SQLITE_INDEX_CONSTRAINT_LE ) iLe = i;
if( op==SQLITE_INDEX_CONSTRAINT_GT ) iGe = i;
if( op==SQLITE_INDEX_CONSTRAINT_GE ) iGe = i;
}
if( iCol==4 ){
if( op==SQLITE_INDEX_CONSTRAINT_EQ ) iLangid = i;
}
}
}
if( iEq>=0 ){
pInfo->idxNum = FTS4AUX_EQ_CONSTRAINT;
pInfo->aConstraintUsage[iEq].argvIndex = iNext++;
pInfo->estimatedCost = 5;
}else{
pInfo->idxNum = 0;
pInfo->estimatedCost = 20000;
if( iGe>=0 ){
pInfo->idxNum += FTS4AUX_GE_CONSTRAINT;
pInfo->aConstraintUsage[iGe].argvIndex = iNext++;
pInfo->estimatedCost /= 2;
}
if( iLe>=0 ){
pInfo->idxNum += FTS4AUX_LE_CONSTRAINT;
pInfo->aConstraintUsage[iLe].argvIndex = iNext++;
pInfo->estimatedCost /= 2;
}
}
if( iLangid>=0 ){
pInfo->aConstraintUsage[iLangid].argvIndex = iNext++;
pInfo->estimatedCost--;
}
return SQLITE_OK;
}
/*
** xOpen - Open a cursor.
*/
static int fts3auxOpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){
Fts3auxCursor *pCsr; /* Pointer to cursor object to return */
UNUSED_PARAMETER(pVTab);
pCsr = (Fts3auxCursor *)sqlite3_malloc(sizeof(Fts3auxCursor));
if( !pCsr ) return SQLITE_NOMEM;
memset(pCsr, 0, sizeof(Fts3auxCursor));
*ppCsr = (sqlite3_vtab_cursor *)pCsr;
return SQLITE_OK;
}
/*
** xClose - Close a cursor.
*/
static int fts3auxCloseMethod(sqlite3_vtab_cursor *pCursor){
Fts3Table *pFts3 = ((Fts3auxTable *)pCursor->pVtab)->pFts3Tab;
Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor;
sqlite3Fts3SegmentsClose(pFts3);
sqlite3Fts3SegReaderFinish(&pCsr->csr);
sqlite3_free((void *)pCsr->filter.zTerm);
sqlite3_free(pCsr->zStop);
sqlite3_free(pCsr->aStat);
sqlite3_free(pCsr);
return SQLITE_OK;
}
static int fts3auxGrowStatArray(Fts3auxCursor *pCsr, int nSize){
if( nSize>pCsr->nStat ){
struct Fts3auxColstats *aNew;
aNew = (struct Fts3auxColstats *)sqlite3_realloc(pCsr->aStat,
sizeof(struct Fts3auxColstats) * nSize
);
if( aNew==0 ) return SQLITE_NOMEM;
memset(&aNew[pCsr->nStat], 0,
sizeof(struct Fts3auxColstats) * (nSize - pCsr->nStat)
);
pCsr->aStat = aNew;
pCsr->nStat = nSize;
}
return SQLITE_OK;
}
/*
** xNext - Advance the cursor to the next row, if any.
*/
static int fts3auxNextMethod(sqlite3_vtab_cursor *pCursor){
Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor;
Fts3Table *pFts3 = ((Fts3auxTable *)pCursor->pVtab)->pFts3Tab;
** The only difference between state 1 and state 2 is that if the
** integer encountered in state 1 is not 0 or 1, then we need to
** increment the column 0 "nDoc" count for this term.
*/
case 1:
assert( iCol==0 );
if( v>1 ){
pCsr->aStat[1].nDoc++;
}
eState = 2;
/* fall through */
case 2:
if( v==0 ){ /* 0x00. Next integer will be a docid. */
eState = 0;
}else if( v==1 ){ /* 0x01. Next integer will be a column number. */
eState = 3;
}else{ /* 2 or greater. A position. */
pCsr->aStat[iCol+1].nOcc++;
pCsr->aStat[0].nOcc++;
}
break;
/* State 3. The integer just read is a column number. */
default: assert( eState==3 );
iCol = (int)v;
if( fts3auxGrowStatArray(pCsr, iCol+2) ) return SQLITE_NOMEM;
pCsr->aStat[iCol+1].nDoc++;
eState = 2;
break;
}
}
pCsr->iCol = 0;
rc = SQLITE_OK;
}else{
pCsr->isEof = 1;
}
return rc;
}
/*
** xFilter - Initialize a cursor to point at the start of its data.
*/
static int fts3auxFilterMethod(
sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */
int idxNum, /* Strategy index */
const char *idxStr, /* Unused */
int nVal, /* Number of elements in apVal */
sqlite3_value **apVal /* Arguments for the indexing scheme */
){
Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor;
Fts3Table *pFts3 = ((Fts3auxTable *)pCursor->pVtab)->pFts3Tab;
int rc;
int isScan = 0;
int iLangVal = 0; /* Language id to query */
int iEq = -1; /* Index of term=? value in apVal */
int iGe = -1; /* Index of term>=? value in apVal */
int iLe = -1; /* Index of term<=? value in apVal */
int iLangid = -1; /* Index of languageid=? value in apVal */
int iNext = 0;
UNUSED_PARAMETER(nVal);
UNUSED_PARAMETER(idxStr);
assert( idxStr==0 );
assert( idxNum==FTS4AUX_EQ_CONSTRAINT || idxNum==0
|| idxNum==FTS4AUX_LE_CONSTRAINT || idxNum==FTS4AUX_GE_CONSTRAINT
|| idxNum==(FTS4AUX_LE_CONSTRAINT|FTS4AUX_GE_CONSTRAINT)
);
if( idxNum==FTS4AUX_EQ_CONSTRAINT ){
iEq = iNext++;
}else{
isScan = 1;
if( idxNum & FTS4AUX_GE_CONSTRAINT ){
iGe = iNext++;
}
if( idxNum & FTS4AUX_LE_CONSTRAINT ){
iLe = iNext++;
}
}
if( iNext<nVal ){
iLangid = iNext++;
}
/* In case this cursor is being reused, close and zero it. */
testcase(pCsr->filter.zTerm);
sqlite3Fts3SegReaderFinish(&pCsr->csr);
sqlite3_free((void *)pCsr->filter.zTerm);
sqlite3_free(pCsr->aStat);
memset(&pCsr->csr, 0, ((u8*)&pCsr[1]) - (u8*)&pCsr->csr);
pCsr->filter.flags = FTS3_SEGMENT_REQUIRE_POS|FTS3_SEGMENT_IGNORE_EMPTY;
if( isScan ) pCsr->filter.flags |= FTS3_SEGMENT_SCAN;
if( iEq>=0 || iGe>=0 ){
const unsigned char *zStr = sqlite3_value_text(apVal[0]);
assert( (iEq==0 && iGe==-1) || (iEq==-1 && iGe==0) );
if( zStr ){
pCsr->filter.zTerm = sqlite3_mprintf("%s", zStr);
pCsr->filter.nTerm = sqlite3_value_bytes(apVal[0]);
if( pCsr->filter.zTerm==0 ) return SQLITE_NOMEM;
}
}
if( iLe>=0 ){
pCsr->zStop = sqlite3_mprintf("%s", sqlite3_value_text(apVal[iLe]));
pCsr->nStop = sqlite3_value_bytes(apVal[iLe]);
if( pCsr->zStop==0 ) return SQLITE_NOMEM;
}
if( iLangid>=0 ){
iLangVal = sqlite3_value_int(apVal[iLangid]);
/* If the user specified a negative value for the languageid, use zero
** instead. This works, as the "languageid=?" constraint will also
** be tested by the VDBE layer. The test will always be false (since
** this module will not return a row with a negative languageid), and
** so the overall query will return zero rows. */
if( iLangVal<0 ) iLangVal = 0;
}
pCsr->iLangid = iLangVal;
rc = sqlite3Fts3SegReaderCursor(pFts3, iLangVal, 0, FTS3_SEGCURSOR_ALL,
pCsr->filter.zTerm, pCsr->filter.nTerm, 0, isScan, &pCsr->csr
);
if( rc==SQLITE_OK ){
rc = sqlite3Fts3SegReaderStart(pFts3, &pCsr->csr, &pCsr->filter);
}
if( rc==SQLITE_OK ) rc = fts3auxNextMethod(pCursor);
return rc;
}
/*
** xEof - Return true if the cursor is at EOF, or false otherwise.
*/
static int fts3auxEofMethod(sqlite3_vtab_cursor *pCursor){
Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor;
return pCsr->isEof;
}
/*
** xColumn - Return a column value.
*/
static int fts3auxColumnMethod(
sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */
sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */
int iCol /* Index of column to read value from */
){
Fts3auxCursor *p = (Fts3auxCursor *)pCursor;
assert( p->isEof==0 );
switch( iCol ){
case 0: /* term */
sqlite3_result_text(pCtx, p->csr.zTerm, p->csr.nTerm, SQLITE_TRANSIENT);
break;
case 1: /* col */
if( p->iCol ){
sqlite3_result_int(pCtx, p->iCol-1);
}else{
sqlite3_result_text(pCtx, "*", -1, SQLITE_STATIC);
}
break;
case 2: /* documents */
sqlite3_result_int64(pCtx, p->aStat[p->iCol].nDoc);
break;
case 3: /* occurrences */
sqlite3_result_int64(pCtx, p->aStat[p->iCol].nOcc);
break;
default: /* languageid */
assert( iCol==4 );
sqlite3_result_int(pCtx, p->iLangid);
break;
}
return SQLITE_OK;
}
/*
** xRowid - Return the current rowid for the cursor.
*/
static int fts3auxRowidMethod(
sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */
sqlite_int64 *pRowid /* OUT: Rowid value */
){
Fts3auxCursor *pCsr = (Fts3auxCursor *)pCursor;
*pRowid = pCsr->iRowid;
return SQLITE_OK;
}
/*
** Register the fts3aux module with database connection db. Return SQLITE_OK
** if successful or an error code if sqlite3_create_module() fails.
*/
SQLITE_PRIVATE int sqlite3Fts3InitAux(sqlite3 *db){
static const sqlite3_module fts3aux_module = {
0, /* iVersion */
fts3auxConnectMethod, /* xCreate */
fts3auxConnectMethod, /* xConnect */
fts3auxBestIndexMethod, /* xBestIndex */
fts3auxDisconnectMethod, /* xDisconnect */
fts3auxDisconnectMethod, /* xDestroy */
fts3auxOpenMethod, /* xOpen */
fts3auxCloseMethod, /* xClose */
fts3auxFilterMethod, /* xFilter */
fts3auxNextMethod, /* xNext */
fts3auxEofMethod, /* xEof */
fts3auxColumnMethod, /* xColumn */
fts3auxRowidMethod, /* xRowid */
0, /* xUpdate */
0, /* xBegin */
0, /* xSync */
0, /* xCommit */
0, /* xRollback */
0, /* xFindFunction */
0, /* xRename */
0, /* xSavepoint */
0, /* xRelease */
0 /* xRollbackTo */
};
int rc; /* Return code */
rc = sqlite3_create_module(db, "fts4aux", &fts3aux_module, 0);
return rc;
}
#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) */
/************** End of fts3_aux.c ********************************************/
/************** Begin file fts3_expr.c ***************************************/
/*
** 2008 Nov 28
** than an implicit AND. When using the new, both implicity and explicit
** AND operators have a higher precedence than OR.
**
** If compiled with SQLITE_TEST defined, then this module exports the
** symbol "int sqlite3_fts3_enable_parentheses". Setting this variable
** to zero causes the module to use the old syntax. If it is set to
** non-zero the new syntax is activated. This is so both syntaxes can
** be tested using a single build of testfixture.
**
** The following describes the syntax supported by the fts3 MATCH
** operator in a similar format to that used by the lemon parser
** generator. This module does not use actually lemon, it uses a
** custom parser.
**
** query ::= andexpr (OR andexpr)*.
**
** andexpr ::= notexpr (AND? notexpr)*.
**
** notexpr ::= nearexpr (NOT nearexpr|-TOKEN)*.
** notexpr ::= LP query RP.
**
** nearexpr ::= phrase (NEAR distance_opt nearexpr)*.
**
** distance_opt ::= .
** distance_opt ::= / INTEGER.
**
** phrase ::= TOKEN.
** phrase ::= COLUMN:TOKEN.
** phrase ::= "TOKEN TOKEN TOKEN...".
*/
#ifdef SQLITE_TEST
SQLITE_API int sqlite3_fts3_enable_parentheses = 0;
#else
# ifdef SQLITE_ENABLE_FTS3_PARENTHESIS
# define sqlite3_fts3_enable_parentheses 1
# else
# define sqlite3_fts3_enable_parentheses 0
# endif
#endif
/*
** Default span for NEAR operators.
*/
#define SQLITE_FTS3_DEFAULT_NEAR_PARAM 10
/* #include <string.h> */
/* #include <assert.h> */
/*
** isNot:
** This variable is used by function getNextNode(). When getNextNode() is
** called, it sets ParseContext.isNot to true if the 'next node' is a
** FTSQUERY_PHRASE with a unary "-" attached to it. i.e. "mysql" in the
** FTS3 query "sqlite -mysql". Otherwise, ParseContext.isNot is set to
** zero.
*/
typedef struct ParseContext ParseContext;
struct ParseContext {
sqlite3_tokenizer *pTokenizer; /* Tokenizer module */
int iLangid; /* Language id used with tokenizer */
const char **azCol; /* Array of column names for fts3 table */
int bFts4; /* True to allow FTS4-only syntax */
int nCol; /* Number of entries in azCol[] */
int iDefaultCol; /* Default column to query */
int isNot; /* True if getNextNode() sees a unary - */
sqlite3_context *pCtx; /* Write error message here */
int nNest; /* Number of nested brackets */
};
/*
** This function is equivalent to the standard isspace() function.
**
** The standard isspace() can be awkward to use safely, because although it
** is defined to accept an argument of type int, its behavior when passed
** an integer that falls outside of the range of the unsigned char type
** is undefined (and sometimes, "undefined" means segfault). This wrapper
** is defined to accept an argument of type char, and always returns 0 for
** any values that fall outside of the range of the unsigned char type (i.e.
** negative values).
*/
static int fts3isspace(char c){
return c==' ' || c=='\t' || c=='\n' || c=='\r' || c=='\v' || c=='\f';
}
/*
** Allocate nByte bytes of memory using sqlite3_malloc(). If successful,
** zero the memory before returning a pointer to it. If unsuccessful,
** return NULL.
*/
static void *fts3MallocZero(int nByte){
void *pRet = sqlite3_malloc(nByte);
if( pRet ) memset(pRet, 0, nByte);
return pRet;
}
SQLITE_PRIVATE int sqlite3Fts3OpenTokenizer(
sqlite3_tokenizer *pTokenizer,
int iLangid,
const char *z,
int n,
sqlite3_tokenizer_cursor **ppCsr
){
sqlite3_tokenizer_module const *pModule = pTokenizer->pModule;
sqlite3_tokenizer_cursor *pCsr = 0;
int rc;
rc = pModule->xOpen(pTokenizer, z, n, &pCsr);
assert( rc==SQLITE_OK || pCsr==0 );
if( rc==SQLITE_OK ){
pCsr->pTokenizer = pTokenizer;
if( pModule->iVersion>=1 ){
rc = pModule->xLanguageid(pCsr, iLangid);
if( rc!=SQLITE_OK ){
pModule->xClose(pCsr);
pCsr = 0;
}
}
}
*ppCsr = pCsr;
return rc;
}
/*
** Function getNextNode(), which is called by fts3ExprParse(), may itself
** call fts3ExprParse(). So this forward declaration is required.
*/
static int fts3ExprParse(ParseContext *, const char *, int, Fts3Expr **, int *);
/*
** Extract the next token from buffer z (length n) using the tokenizer
** and other information (column names etc.) in pParse. Create an Fts3Expr
** structure of type FTSQUERY_PHRASE containing a phrase consisting of this
** single token and set *ppExpr to point to it. If the end of the buffer is
** reached before a token is found, set *ppExpr to zero. It is the
** responsibility of the caller to eventually deallocate the allocated
** Fts3Expr structure (if any) by passing it to sqlite3_free().
**
** Return SQLITE_OK if successful, or SQLITE_NOMEM if a memory allocation
** fails.
*/
static int getNextToken(
ParseContext *pParse, /* fts3 query parse context */
int iCol, /* Value for Fts3Phrase.iColumn */
const char *z, int n, /* Input string */
Fts3Expr **ppExpr, /* OUT: expression */
int *pnConsumed /* OUT: Number of bytes consumed */
){
sqlite3_tokenizer *pTokenizer = pParse->pTokenizer;
sqlite3_tokenizer_module const *pModule = pTokenizer->pModule;
int rc;
sqlite3_tokenizer_cursor *pCursor;
Fts3Expr *pRet = 0;
int i = 0;
/* Set variable i to the maximum number of bytes of input to tokenize. */
for(i=0; i<n; i++){
if( sqlite3_fts3_enable_parentheses && (z[i]=='(' || z[i]==')') ) break;
if( z[i]=='"' ) break;
}
*pnConsumed = i;
rc = sqlite3Fts3OpenTokenizer(pTokenizer, pParse->iLangid, z, i, &pCursor);
if( rc==SQLITE_OK ){
const char *zToken;
int nToken = 0, iStart = 0, iEnd = 0, iPosition = 0;
int nByte; /* total space to allocate */
rc = pModule->xNext(pCursor, &zToken, &nToken, &iStart, &iEnd, &iPosition);
if( rc==SQLITE_OK ){
nByte = sizeof(Fts3Expr) + sizeof(Fts3Phrase) + nToken;
pRet = (Fts3Expr *)fts3MallocZero(nByte);
if( !pRet ){
rc = SQLITE_NOMEM;
}else{
pRet->eType = FTSQUERY_PHRASE;
pRet->pPhrase = (Fts3Phrase *)&pRet[1];
pRet->pPhrase->nToken = 1;
pRet->pPhrase->iColumn = iCol;
pRet->pPhrase->aToken[0].n = nToken;
pRet->pPhrase->aToken[0].z = (char *)&pRet->pPhrase[1];
memcpy(pRet->pPhrase->aToken[0].z, zToken, nToken);
if( iEnd<n && z[iEnd]=='*' ){
pRet->pPhrase->aToken[0].isPrefix = 1;
iEnd++;
}
while( 1 ){
if( !sqlite3_fts3_enable_parentheses
&& iStart>0 && z[iStart-1]=='-'
){
pParse->isNot = 1;
iStart--;
}else if( pParse->bFts4 && iStart>0 && z[iStart-1]=='^' ){
pRet->pPhrase->aToken[0].bFirst = 1;
iStart--;
}else{
break;
}
}
}
*pnConsumed = iEnd;
}else if( i && rc==SQLITE_DONE ){
rc = SQLITE_OK;
}
pModule->xClose(pCursor);
}
*ppExpr = pRet;
return rc;
}
/*
** Enlarge a memory allocation. If an out-of-memory allocation occurs,
** then free the old allocation.
*/
static void *fts3ReallocOrFree(void *pOrig, int nNew){
void *pRet = sqlite3_realloc(pOrig, nNew);
if( !pRet ){
sqlite3_free(pOrig);
}
return pRet;
}
/*
** Buffer zInput, length nInput, contains the contents of a quoted string
** that appeared as part of an fts3 query expression. Neither quote character
** is included in the buffer. This function attempts to tokenize the entire
** input buffer and create an Fts3Expr structure of type FTSQUERY_PHRASE
** containing the results.
**
** If successful, SQLITE_OK is returned and *ppExpr set to point at the
** allocated Fts3Expr structure. Otherwise, either SQLITE_NOMEM (out of memory
** error) or SQLITE_ERROR (tokenization error) is returned and *ppExpr set
** to 0.
*/
static int getNextString(
ParseContext *pParse, /* fts3 query parse context */
const char *zInput, int nInput, /* Input string */
Fts3Expr **ppExpr /* OUT: expression */
){
sqlite3_tokenizer *pTokenizer = pParse->pTokenizer;
sqlite3_tokenizer_module const *pModule = pTokenizer->pModule;
int rc;
Fts3Expr *p = 0;
sqlite3_tokenizer_cursor *pCursor = 0;
char *zTemp = 0;
int nTemp = 0;
const int nSpace = sizeof(Fts3Expr) + sizeof(Fts3Phrase);
int nToken = 0;
/* The final Fts3Expr data structure, including the Fts3Phrase,
** Fts3PhraseToken structures token buffers are all stored as a single
** allocation so that the expression can be freed with a single call to
** sqlite3_free(). Setting this up requires a two pass approach.
**
** The first pass, in the block below, uses a tokenizer cursor to iterate
** through the tokens in the expression. This pass uses fts3ReallocOrFree()
** to assemble data in two dynamic buffers:
**
** Buffer p: Points to the Fts3Expr structure, followed by the Fts3Phrase
** structure, followed by the array of Fts3PhraseToken
** structures. This pass only populates the Fts3PhraseToken array.
**
** Buffer zTemp: Contains copies of all tokens.
**
** The second pass, in the block that begins "if( rc==SQLITE_DONE )" below,
** appends buffer zTemp to buffer p, and fills in the Fts3Expr and Fts3Phrase
** structures.
*/
rc = sqlite3Fts3OpenTokenizer(
pTokenizer, pParse->iLangid, zInput, nInput, &pCursor);
if( rc==SQLITE_OK ){
int ii;
for(ii=0; rc==SQLITE_OK; ii++){
const char *zByte;
int nByte = 0, iBegin = 0, iEnd = 0, iPos = 0;
rc = pModule->xNext(pCursor, &zByte, &nByte, &iBegin, &iEnd, &iPos);
if( rc==SQLITE_OK ){
Fts3PhraseToken *pToken;
p = fts3ReallocOrFree(p, nSpace + ii*sizeof(Fts3PhraseToken));
if( !p ) goto no_mem;
zTemp = fts3ReallocOrFree(zTemp, nTemp + nByte);
if( !zTemp ) goto no_mem;
assert( nToken==ii );
pToken = &((Fts3Phrase *)(&p[1]))->aToken[ii];
memset(pToken, 0, sizeof(Fts3PhraseToken));
memcpy(&zTemp[nTemp], zByte, nByte);
nTemp += nByte;
pToken->n = nByte;
pToken->isPrefix = (iEnd<nInput && zInput[iEnd]=='*');
pToken->bFirst = (iBegin>0 && zInput[iBegin-1]=='^');
nToken = ii+1;
}
}
pModule->xClose(pCursor);
pCursor = 0;
}
if( rc==SQLITE_DONE ){
int jj;
char *zBuf = 0;
p = fts3ReallocOrFree(p, nSpace + nToken*sizeof(Fts3PhraseToken) + nTemp);
if( !p ) goto no_mem;
memset(p, 0, (char *)&(((Fts3Phrase *)&p[1])->aToken[0])-(char *)p);
p->eType = FTSQUERY_PHRASE;
p->pPhrase = (Fts3Phrase *)&p[1];
p->pPhrase->iColumn = pParse->iDefaultCol;
p->pPhrase->nToken = nToken;
zBuf = (char *)&p->pPhrase->aToken[nToken];
if( zTemp ){
memcpy(zBuf, zTemp, nTemp);
sqlite3_free(zTemp);
}else{
assert( nTemp==0 );
}
for(jj=0; jj<p->pPhrase->nToken; jj++){
p->pPhrase->aToken[jj].z = zBuf;
zBuf += p->pPhrase->aToken[jj].n;
}
rc = SQLITE_OK;
}
p = 0;
for(i=0; i<nMaxDepth; i++){
if( apLeaf[i] ){
if( p==0 ){
p = apLeaf[i];
p->pParent = 0;
}else{
assert( pFree!=0 );
pFree->pRight = p;
pFree->pLeft = apLeaf[i];
pFree->pLeft->pParent = pFree;
pFree->pRight->pParent = pFree;
p = pFree;
pFree = pFree->pParent;
p->pParent = 0;
}
}
}
pRoot = p;
}else{
/* An error occurred. Delete the contents of the apLeaf[] array
** and pFree list. Everything else is cleaned up by the call to
** sqlite3Fts3ExprFree(pRoot) below. */
Fts3Expr *pDel;
for(i=0; i<nMaxDepth; i++){
sqlite3Fts3ExprFree(apLeaf[i]);
}
while( (pDel=pFree)!=0 ){
pFree = pDel->pParent;
sqlite3_free(pDel);
}
}
assert( pFree==0 );
sqlite3_free( apLeaf );
}
}
if( rc!=SQLITE_OK ){
sqlite3Fts3ExprFree(pRoot);
pRoot = 0;
}
*pp = pRoot;
return rc;
}
/*
** This function is similar to sqlite3Fts3ExprParse(), with the following
** differences:
**
** 1. It does not do expression rebalancing.
** 2. It does not check that the expression does not exceed the
** maximum allowable depth.
** 3. Even if it fails, *ppExpr may still be set to point to an
** expression tree. It should be deleted using sqlite3Fts3ExprFree()
** in this case.
*/
static int fts3ExprParseUnbalanced(
sqlite3_tokenizer *pTokenizer, /* Tokenizer module */
int iLangid, /* Language id for tokenizer */
char **azCol, /* Array of column names for fts3 table */
int bFts4, /* True to allow FTS4-only syntax */
int nCol, /* Number of entries in azCol[] */
int iDefaultCol, /* Default column to query */
const char *z, int n, /* Text of MATCH query */
Fts3Expr **ppExpr /* OUT: Parsed query structure */
){
int nParsed;
int rc;
ParseContext sParse;
memset(&sParse, 0, sizeof(ParseContext));
sParse.pTokenizer = pTokenizer;
sParse.iLangid = iLangid;
sParse.azCol = (const char **)azCol;
sParse.nCol = nCol;
sParse.iDefaultCol = iDefaultCol;
sParse.bFts4 = bFts4;
if( z==0 ){
*ppExpr = 0;
return SQLITE_OK;
}
if( n<0 ){
n = (int)strlen(z);
}
rc = fts3ExprParse(&sParse, z, n, ppExpr, &nParsed);
assert( rc==SQLITE_OK || *ppExpr==0 );
/* Check for mismatched parenthesis */
if( rc==SQLITE_OK && sParse.nNest ){
rc = SQLITE_ERROR;
}
return rc;
}
/*
** Parameters z and n contain a pointer to and length of a buffer containing
** an fts3 query expression, respectively. This function attempts to parse the
** query expression and create a tree of Fts3Expr structures representing the
** parsed expression. If successful, *ppExpr is set to point to the head
** of the parsed expression tree and SQLITE_OK is returned. If an error
** occurs, either SQLITE_NOMEM (out-of-memory error) or SQLITE_ERROR (parse
** error) is returned and *ppExpr is set to 0.
**
** If parameter n is a negative number, then z is assumed to point to a
** nul-terminated string and the length is determined using strlen().
**
** The first parameter, pTokenizer, is passed the fts3 tokenizer module to
** use to normalize query tokens while parsing the expression. The azCol[]
** array, which is assumed to contain nCol entries, should contain the names
** of each column in the target fts3 table, in order from left to right.
** Column names must be nul-terminated strings.
**
** The iDefaultCol parameter should be passed the index of the table column
** that appears on the left-hand-side of the MATCH operator (the default
** column to match against for tokens for which a column name is not explicitly
** specified as part of the query string), or -1 if tokens may by default
** match any table column.
*/
SQLITE_PRIVATE int sqlite3Fts3ExprParse(
sqlite3_tokenizer *pTokenizer, /* Tokenizer module */
int iLangid, /* Language id for tokenizer */
char **azCol, /* Array of column names for fts3 table */
int bFts4, /* True to allow FTS4-only syntax */
int nCol, /* Number of entries in azCol[] */
int iDefaultCol, /* Default column to query */
const char *z, int n, /* Text of MATCH query */
Fts3Expr **ppExpr, /* OUT: Parsed query structure */
char **pzErr /* OUT: Error message (sqlite3_malloc) */
){
int rc = fts3ExprParseUnbalanced(
pTokenizer, iLangid, azCol, bFts4, nCol, iDefaultCol, z, n, ppExpr
);
/* Rebalance the expression. And check that its depth does not exceed
** SQLITE_FTS3_MAX_EXPR_DEPTH. */
if( rc==SQLITE_OK && *ppExpr ){
rc = fts3ExprBalance(ppExpr, SQLITE_FTS3_MAX_EXPR_DEPTH);
if( rc==SQLITE_OK ){
rc = fts3ExprCheckDepth(*ppExpr, SQLITE_FTS3_MAX_EXPR_DEPTH);
}
}
if( rc!=SQLITE_OK ){
sqlite3Fts3ExprFree(*ppExpr);
*ppExpr = 0;
if( rc==SQLITE_TOOBIG ){
*pzErr = sqlite3_mprintf(
"FTS expression tree is too large (maximum depth %d)",
SQLITE_FTS3_MAX_EXPR_DEPTH
);
rc = SQLITE_ERROR;
}else if( rc==SQLITE_ERROR ){
*pzErr = sqlite3_mprintf("malformed MATCH expression: [%s]", z);
}
}
return rc;
}
/*
** Free a single node of an expression tree.
*/
static void fts3FreeExprNode(Fts3Expr *p){
assert( p->eType==FTSQUERY_PHRASE || p->pPhrase==0 );
sqlite3Fts3EvalPhraseCleanup(p->pPhrase);
sqlite3_free(p->aMI);
sqlite3_free(p);
}
/*
** Free a parsed fts3 query expression allocated by sqlite3Fts3ExprParse().
**
** This function would be simpler if it recursively called itself. But
** that would mean passing a sufficiently large expression to ExprParse()
** could cause a stack overflow.
*/
SQLITE_PRIVATE void sqlite3Fts3ExprFree(Fts3Expr *pDel){
Fts3Expr *p;
assert( pDel==0 || pDel->pParent==0 );
for(p=pDel; p && (p->pLeft||p->pRight); p=(p->pLeft ? p->pLeft : p->pRight)){
assert( p->pParent==0 || p==p->pParent->pRight || p==p->pParent->pLeft );
}
while( p ){
Fts3Expr *pParent = p->pParent;
fts3FreeExprNode(p);
if( pParent && p==pParent->pLeft && pParent->pRight ){
p = pParent->pRight;
while( p && (p->pLeft || p->pRight) ){
assert( p==p->pParent->pRight || p==p->pParent->pLeft );
p = (p->pLeft ? p->pLeft : p->pRight);
}
if( *pRC ) return;
rc = fts3SqlStmt(p, eStmt, &pStmt, apVal);
if( rc==SQLITE_OK ){
sqlite3_step(pStmt);
rc = sqlite3_reset(pStmt);
}
*pRC = rc;
}
/*
** This function ensures that the caller has obtained an exclusive
** shared-cache table-lock on the %_segdir table. This is required before
** writing data to the fts3 table. If this lock is not acquired first, then
** the caller may end up attempting to take this lock as part of committing
** a transaction, causing SQLite to return SQLITE_LOCKED or
** LOCKED_SHAREDCACHEto a COMMIT command.
**
** It is best to avoid this because if FTS3 returns any error when
** committing a transaction, the whole transaction will be rolled back.
** And this is not what users expect when they get SQLITE_LOCKED_SHAREDCACHE.
** It can still happen if the user locks the underlying tables directly
** instead of accessing them via FTS.
*/
static int fts3Writelock(Fts3Table *p){
int rc = SQLITE_OK;
if( p->nPendingData==0 ){
sqlite3_stmt *pStmt;
rc = fts3SqlStmt(p, SQL_DELETE_SEGDIR_LEVEL, &pStmt, 0);
if( rc==SQLITE_OK ){
sqlite3_bind_null(pStmt, 1);
sqlite3_step(pStmt);
rc = sqlite3_reset(pStmt);
}
}
return rc;
}
/*
** FTS maintains a separate indexes for each language-id (a 32-bit integer).
** Within each language id, a separate index is maintained to store the
** document terms, and each configured prefix size (configured the FTS
** "prefix=" option). And each index consists of multiple levels ("relative
** levels").
**
** All three of these values (the language id, the specific index and the
** level within the index) are encoded in 64-bit integer values stored
** in the %_segdir table on disk. This function is used to convert three
** separate component values into the single 64-bit integer value that
** can be used to query the %_segdir table.
**
** Specifically, each language-id/index combination is allocated 1024
** 64-bit integer level values ("absolute levels"). The main terms index
** for language-id 0 is allocate values 0-1023. The first prefix index
** (if any) for language-id 0 is allocated values 1024-2047. And so on.
** Language 1 indexes are allocated immediately following language 0.
**
** So, for a system with nPrefix prefix indexes configured, the block of
** absolute levels that corresponds to language-id iLangid and index
** iIndex starts at absolute level ((iLangid * (nPrefix+1) + iIndex) * 1024).
*/
static sqlite3_int64 getAbsoluteLevel(
Fts3Table *p, /* FTS3 table handle */
int iLangid, /* Language id */
int iIndex, /* Index in p->aIndex[] */
int iLevel /* Level of segments */
){
sqlite3_int64 iBase; /* First absolute level for iLangid/iIndex */
assert( iLangid>=0 );
assert( p->nIndex>0 );
assert( iIndex>=0 && iIndex<p->nIndex );
iBase = ((sqlite3_int64)iLangid * p->nIndex + iIndex) * FTS3_SEGDIR_MAXLEVEL;
return iBase + iLevel;
}
/*
** Set *ppStmt to a statement handle that may be used to iterate through
** all rows in the %_segdir table, from oldest to newest. If successful,
** return SQLITE_OK. If an error occurs while preparing the statement,
** return an SQLite error code.
**
** There is only ever one instance of this SQL statement compiled for
** each FTS3 table.
**
** The statement returns the following columns from the %_segdir table:
**
** 0: idx
** 1: start_block
** 2: leaves_end_block
** 3: end_block
** 4: root
*/
SQLITE_PRIVATE int sqlite3Fts3AllSegdirs(
Fts3Table *p, /* FTS3 table */
int iLangid, /* Language being queried */
int iIndex, /* Index for p->aIndex[] */
int iLevel, /* Level to select (relative level) */
sqlite3_stmt **ppStmt /* OUT: Compiled statement */
){
int rc;
sqlite3_stmt *pStmt = 0;
assert( iLevel==FTS3_SEGCURSOR_ALL || iLevel>=0 );
assert( iLevel<FTS3_SEGDIR_MAXLEVEL );
assert( iIndex>=0 && iIndex<p->nIndex );
if( iLevel<0 ){
/* "SELECT * FROM %_segdir WHERE level BETWEEN ? AND ? ORDER BY ..." */
rc = fts3SqlStmt(p, SQL_SELECT_LEVEL_RANGE, &pStmt, 0);
if( rc==SQLITE_OK ){
sqlite3_bind_int64(pStmt, 1, getAbsoluteLevel(p, iLangid, iIndex, 0));
sqlite3_bind_int64(pStmt, 2,
getAbsoluteLevel(p, iLangid, iIndex, FTS3_SEGDIR_MAXLEVEL-1)
);
}
}else{
/* "SELECT * FROM %_segdir WHERE level = ? ORDER BY ..." */
rc = fts3SqlStmt(p, SQL_SELECT_LEVEL, &pStmt, 0);
if( rc==SQLITE_OK ){
sqlite3_bind_int64(pStmt, 1, getAbsoluteLevel(p, iLangid, iIndex,iLevel));
}
}
*ppStmt = pStmt;
return rc;
}
/*
** Append a single varint to a PendingList buffer. SQLITE_OK is returned
** if successful, or an SQLite error code otherwise.
**
** This function also serves to allocate the PendingList structure itself.
** For example, to create a new PendingList structure containing two
** varints:
**
** PendingList *p = 0;
** fts3PendingListAppendVarint(&p, 1);
** fts3PendingListAppendVarint(&p, 2);
*/
static int fts3PendingListAppendVarint(
PendingList **pp, /* IN/OUT: Pointer to PendingList struct */
sqlite3_int64 i /* Value to append to data */
){
PendingList *p = *pp;
/* Allocate or grow the PendingList as required. */
if( !p ){
p = sqlite3_malloc(sizeof(*p) + 100);
if( !p ){
return SQLITE_NOMEM;
}
p->nSpace = 100;
p->aData = (char *)&p[1];
p->nData = 0;
}
else if( p->nData+FTS3_VARINT_MAX+1>p->nSpace ){
int nNew = p->nSpace * 2;
p = sqlite3_realloc(p, sizeof(*p) + nNew);
if( !p ){
sqlite3_free(*pp);
*pp = 0;
return SQLITE_NOMEM;
}
p->nSpace = nNew;
p->aData = (char *)&p[1];
}
/* Append the new serialized varint to the end of the list. */
p->nData += sqlite3Fts3PutVarint(&p->aData[p->nData], i);
p->aData[p->nData] = '\0';
*pp = p;
return SQLITE_OK;
}
/*
** Add a docid/column/position entry to a PendingList structure. Non-zero
** is returned if the structure is sqlite3_realloced as part of adding
** the entry. Otherwise, zero.
**
** If an OOM error occurs, *pRc is set to SQLITE_NOMEM before returning.
pendinglistappend_out:
*pRc = rc;
if( p!=*pp ){
*pp = p;
return 1;
}
return 0;
}
/*
** Free a PendingList object allocated by fts3PendingListAppend().
*/
static void fts3PendingListDelete(PendingList *pList){
sqlite3_free(pList);
}
/*
** Add an entry to one of the pending-terms hash tables.
*/
static int fts3PendingTermsAddOne(
Fts3Table *p,
int iCol,
int iPos,
Fts3Hash *pHash, /* Pending terms hash table to add entry to */
const char *zToken,
int nToken
){
PendingList *pList;
int rc = SQLITE_OK;
pList = (PendingList *)fts3HashFind(pHash, zToken, nToken);
if( pList ){
p->nPendingData -= (pList->nData + nToken + sizeof(Fts3HashElem));
}
if( fts3PendingListAppend(&pList, p->iPrevDocid, iCol, iPos, &rc) ){
if( pList==fts3HashInsert(pHash, zToken, nToken, pList) ){
/* Malloc failed while inserting the new entry. This can only
** happen if there was no previous entry for this token.
*/
assert( 0==fts3HashFind(pHash, zToken, nToken) );
sqlite3_free(pList);
rc = SQLITE_NOMEM;
}
}
if( rc==SQLITE_OK ){
p->nPendingData += (pList->nData + nToken + sizeof(Fts3HashElem));
}
return rc;
}
/*
** Tokenize the nul-terminated string zText and add all tokens to the
** pending-terms hash-table. The docid used is that currently stored in
** p->iPrevDocid, and the column is specified by argument iCol.
**
** If successful, SQLITE_OK is returned. Otherwise, an SQLite error code.
*/
static int fts3PendingTermsAdd(
Fts3Table *p, /* Table into which text will be inserted */
int iLangid, /* Language id to use */
const char *zText, /* Text of document to be inserted */
int iCol, /* Column into which text is being inserted */
u32 *pnWord /* IN/OUT: Incr. by number tokens inserted */
){
int rc;
int iStart = 0;
int iEnd = 0;
int iPos = 0;
int nWord = 0;
char const *zToken;
int nToken = 0;
sqlite3_tokenizer *pTokenizer = p->pTokenizer;
sqlite3_tokenizer_module const *pModule = pTokenizer->pModule;
sqlite3_tokenizer_cursor *pCsr;
int (*xNext)(sqlite3_tokenizer_cursor *pCursor,
const char**,int*,int*,int*,int*);
assert( pTokenizer && pModule );
/* If the user has inserted a NULL value, this function may be called with
** zText==0. In this case, add zero token entries to the hash table and
** return early. */
if( zText==0 ){
*pnWord = 0;
return SQLITE_OK;
}
rc = sqlite3Fts3OpenTokenizer(pTokenizer, iLangid, zText, -1, &pCsr);
if( rc!=SQLITE_OK ){
return rc;
}
xNext = pModule->xNext;
while( SQLITE_OK==rc
&& SQLITE_OK==(rc = xNext(pCsr, &zToken, &nToken, &iStart, &iEnd, &iPos))
){
int i;
if( iPos>=nWord ) nWord = iPos+1;
/* Positions cannot be negative; we use -1 as a terminator internally.
** Tokens must have a non-zero length.
*/
if( iPos<0 || !zToken || nToken<=0 ){
rc = SQLITE_ERROR;
break;
}
/* Add the term to the terms index */
rc = fts3PendingTermsAddOne(
p, iCol, iPos, &p->aIndex[0].hPending, zToken, nToken
);
/* Add the term to each of the prefix indexes that it is not too
** short for. */
for(i=1; rc==SQLITE_OK && i<p->nIndex; i++){
struct Fts3Index *pIndex = &p->aIndex[i];
if( nToken<pIndex->nPrefix ) continue;
rc = fts3PendingTermsAddOne(
p, iCol, iPos, &pIndex->hPending, zToken, pIndex->nPrefix
);
}
}
pModule->xClose(pCsr);
*pnWord += nWord;
return (rc==SQLITE_DONE ? SQLITE_OK : rc);
}
/*
** Calling this function indicates that subsequent calls to
** fts3PendingTermsAdd() are to add term/position-list pairs for the
** contents of the document with docid iDocid.
*/
static int fts3PendingTermsDocid(
Fts3Table *p, /* Full-text table handle */
int iLangid, /* Language id of row being written */
sqlite_int64 iDocid /* Docid of row being written */
){
assert( iLangid>=0 );
/* TODO(shess) Explore whether partially flushing the buffer on
** forced-flush would provide better performance. I suspect that if
** we ordered the doclists by size and flushed the largest until the
** buffer was half empty, that would let the less frequent terms
** generate longer doclists.
*/
if( iDocid<=p->iPrevDocid
|| p->iPrevLangid!=iLangid
|| p->nPendingData>p->nMaxPendingData
){
int rc = sqlite3Fts3PendingTermsFlush(p);
if( rc!=SQLITE_OK ) return rc;
}
p->iPrevDocid = iDocid;
p->iPrevLangid = iLangid;
return SQLITE_OK;
}
/*
** Discard the contents of the pending-terms hash tables.
*/
SQLITE_PRIVATE void sqlite3Fts3PendingTermsClear(Fts3Table *p){
int i;
for(i=0; i<p->nIndex; i++){
Fts3HashElem *pElem;
Fts3Hash *pHash = &p->aIndex[i].hPending;
for(pElem=fts3HashFirst(pHash); pElem; pElem=fts3HashNext(pElem)){
PendingList *pList = (PendingList *)fts3HashData(pElem);
fts3PendingListDelete(pList);
}
fts3HashClear(pHash);
}
p->nPendingData = 0;
}
/*
** This function is called by the xUpdate() method as part of an INSERT
** operation. It adds entries for each term in the new record to the
** pendingTerms hash table.
**
** Argument apVal is the same as the similarly named argument passed to
** fts3InsertData(). Parameter iDocid is the docid of the new row.
*/
static int fts3InsertTerms(
Fts3Table *p,
int iLangid,
sqlite3_value **apVal,
u32 *aSz
){
int i; /* Iterator variable */
for(i=2; i<p->nColumn+2; i++){
int iCol = i-2;
if( p->abNotindexed[iCol]==0 ){
const char *zText = (const char *)sqlite3_value_text(apVal[i]);
int rc = fts3PendingTermsAdd(p, iLangid, zText, iCol, &aSz[iCol]);
if( rc!=SQLITE_OK ){
return rc;
}
aSz[p->nColumn] += sqlite3_value_bytes(apVal[i]);
}
}
return SQLITE_OK;
}
/*
** This function is called by the xUpdate() method for an INSERT operation.
** The apVal parameter is passed a copy of the apVal argument passed by
** SQLite to the xUpdate() method. i.e:
**
** apVal[0] Not used for INSERT.
** apVal[1] rowid
** apVal[2] Left-most user-defined column
** ...
** apVal[p->nColumn+1] Right-most user-defined column
** apVal[p->nColumn+2] Hidden column with same name as table
** apVal[p->nColumn+3] Hidden "docid" column (alias for rowid)
** apVal[p->nColumn+4] Hidden languageid column
*/
static int fts3InsertData(
Fts3Table *p, /* Full-text table */
sqlite3_value **apVal, /* Array of values to insert */
sqlite3_int64 *piDocid /* OUT: Docid for row just inserted */
){
int rc; /* Return code */
sqlite3_stmt *pContentInsert; /* INSERT INTO %_content VALUES(...) */
if( p->zContentTbl ){
sqlite3_value *pRowid = apVal[p->nColumn+3];
if( sqlite3_value_type(pRowid)==SQLITE_NULL ){
pRowid = apVal[1];
}
if( sqlite3_value_type(pRowid)!=SQLITE_INTEGER ){
return SQLITE_CONSTRAINT;
}
*piDocid = sqlite3_value_int64(pRowid);
return SQLITE_OK;
}
/* Locate the statement handle used to insert data into the %_content
** table. The SQL for this statement is:
**
** INSERT INTO %_content VALUES(?, ?, ?, ...)
**
** The statement features N '?' variables, where N is the number of user
** defined columns in the FTS3 table, plus one for the docid field.
*/
rc = fts3SqlStmt(p, SQL_CONTENT_INSERT, &pContentInsert, &apVal[1]);
if( rc==SQLITE_OK && p->zLanguageid ){
rc = sqlite3_bind_int(
pContentInsert, p->nColumn+2,
sqlite3_value_int(apVal[p->nColumn+4])
);
}
if( rc!=SQLITE_OK ) return rc;
** a value for the "rowid" field, for the "docid" field, or for both.
** Which is a problem, since "rowid" and "docid" are aliases for the
** same value. For example:
**
** INSERT INTO fts3tbl(rowid, docid) VALUES(1, 2);
**
** In FTS3, this is an error. It is an error to specify non-NULL values
** for both docid and some other rowid alias.
*/
if( SQLITE_NULL!=sqlite3_value_type(apVal[3+p->nColumn]) ){
if( SQLITE_NULL==sqlite3_value_type(apVal[0])
&& SQLITE_NULL!=sqlite3_value_type(apVal[1])
){
/* A rowid/docid conflict. */
return SQLITE_ERROR;
}
rc = sqlite3_bind_value(pContentInsert, 1, apVal[3+p->nColumn]);
if( rc!=SQLITE_OK ) return rc;
}
/* Execute the statement to insert the record. Set *piDocid to the
** new docid value.
*/
sqlite3_step(pContentInsert);
rc = sqlite3_reset(pContentInsert);
*piDocid = sqlite3_last_insert_rowid(p->db);
return rc;
}
/*
** Remove all data from the FTS3 table. Clear the hash table containing
** pending terms.
*/
static int fts3DeleteAll(Fts3Table *p, int bContent){
int rc = SQLITE_OK; /* Return code */
/* Discard the contents of the pending-terms hash table. */
sqlite3Fts3PendingTermsClear(p);
/* Delete everything from the shadow tables. Except, leave %_content as
** is if bContent is false. */
assert( p->zContentTbl==0 || bContent==0 );
if( bContent ) fts3SqlExec(&rc, p, SQL_DELETE_ALL_CONTENT, 0);
fts3SqlExec(&rc, p, SQL_DELETE_ALL_SEGMENTS, 0);
fts3SqlExec(&rc, p, SQL_DELETE_ALL_SEGDIR, 0);
if( p->bHasDocsize ){
fts3SqlExec(&rc, p, SQL_DELETE_ALL_DOCSIZE, 0);
}
if( p->bHasStat ){
fts3SqlExec(&rc, p, SQL_DELETE_ALL_STAT, 0);
}
return rc;
}
/*
**
*/
static int langidFromSelect(Fts3Table *p, sqlite3_stmt *pSelect){
int iLangid = 0;
if( p->zLanguageid ) iLangid = sqlite3_column_int(pSelect, p->nColumn+1);
return iLangid;
}
/*
** The first element in the apVal[] array is assumed to contain the docid
** (an integer) of a row about to be deleted. Remove all terms from the
** full-text index.
*/
static void fts3DeleteTerms(
int *pRC, /* Result code */
Fts3Table *p, /* The FTS table to delete from */
sqlite3_value *pRowid, /* The docid to be deleted */
u32 *aSz, /* Sizes of deleted document written here */
int *pbFound /* OUT: Set to true if row really does exist */
){
int rc;
sqlite3_stmt *pSelect;
assert( *pbFound==0 );
if( *pRC ) return;
rc = fts3SqlStmt(p, SQL_SELECT_CONTENT_BY_ROWID, &pSelect, &pRowid);
if( rc==SQLITE_OK ){
if( SQLITE_ROW==sqlite3_step(pSelect) ){
int i;
int iLangid = langidFromSelect(p, pSelect);
rc = fts3PendingTermsDocid(p, iLangid, sqlite3_column_int64(pSelect, 0));
for(i=1; rc==SQLITE_OK && i<=p->nColumn; i++){
int iCol = i-1;
if( p->abNotindexed[iCol]==0 ){
const char *zText = (const char *)sqlite3_column_text(pSelect, i);
rc = fts3PendingTermsAdd(p, iLangid, zText, -1, &aSz[iCol]);
aSz[p->nColumn] += sqlite3_column_bytes(pSelect, i);
}
}
if( rc!=SQLITE_OK ){
sqlite3_reset(pSelect);
*pRC = rc;
return;
}
*pbFound = 1;
}
rc = sqlite3_reset(pSelect);
}else{
sqlite3_reset(pSelect);
}
*pRC = rc;
}
/*
** Forward declaration to account for the circular dependency between
** functions fts3SegmentMerge() and fts3AllocateSegdirIdx().
*/
static int fts3SegmentMerge(Fts3Table *, int, int, int);
/*
** This function allocates a new level iLevel index in the segdir table.
** Usually, indexes are allocated within a level sequentially starting
** with 0, so the allocated index is one greater than the value returned
** by:
**
** SELECT max(idx) FROM %_segdir WHERE level = :iLevel
**
** However, if there are already FTS3_MERGE_COUNT indexes at the requested
** level, they are merged into a single level (iLevel+1) segment and the
** allocated index is 0.
**
** If successful, *piIdx is set to the allocated index slot and SQLITE_OK
** returned. Otherwise, an SQLite error code is returned.
*/
static int fts3AllocateSegdirIdx(
Fts3Table *p,
int iLangid, /* Language id */
int iIndex, /* Index for p->aIndex */
int iLevel,
int *piIdx
){
int rc; /* Return Code */
sqlite3_stmt *pNextIdx; /* Query for next idx at level iLevel */
int iNext = 0; /* Result of query pNextIdx */
assert( iLangid>=0 );
assert( p->nIndex>=1 );
/* Set variable iNext to the next available segdir index at level iLevel. */
rc = fts3SqlStmt(p, SQL_NEXT_SEGMENT_INDEX, &pNextIdx, 0);
if( rc==SQLITE_OK ){
sqlite3_bind_int64(
pNextIdx, 1, getAbsoluteLevel(p, iLangid, iIndex, iLevel)
);
if( SQLITE_ROW==sqlite3_step(pNextIdx) ){
iNext = sqlite3_column_int(pNextIdx, 0);
}
rc = sqlite3_reset(pNextIdx);
}
if( rc==SQLITE_OK ){
/* If iNext is FTS3_MERGE_COUNT, indicating that level iLevel is already
** full, merge all segments in level iLevel into a single iLevel+1
** segment and allocate (newly freed) index 0 at level iLevel. Otherwise,
** if iNext is less than FTS3_MERGE_COUNT, allocate index iNext.
*/
if( iNext>=FTS3_MERGE_COUNT ){
fts3LogMerge(16, getAbsoluteLevel(p, iLangid, iIndex, iLevel));
rc = fts3SegmentMerge(p, iLangid, iIndex, iLevel);
*piIdx = 0;
}else{
*piIdx = iNext;
}
}
return rc;
}
/*
** The %_segments table is declared as follows:
**
** CREATE TABLE %_segments(blockid INTEGER PRIMARY KEY, block BLOB)
**
** This function reads data from a single row of the %_segments table. The
** specific row is identified by the iBlockid parameter. If paBlob is not
** NULL, then a buffer is allocated using sqlite3_malloc() and populated
** with the contents of the blob stored in the "block" column of the
** identified table row is. Whether or not paBlob is NULL, *pnBlob is set
** to the size of the blob in bytes before returning.
**
** If an error occurs, or the table does not contain the specified row,
** an SQLite error code is returned. Otherwise, SQLITE_OK is returned. If
** paBlob is non-NULL, then it is the responsibility of the caller to
** eventually free the returned buffer.
**
** This function may leave an open sqlite3_blob* handle in the
** Fts3Table.pSegments variable. This handle is reused by subsequent calls
** to this function. The handle may be closed by calling the
** sqlite3Fts3SegmentsClose() function. Reusing a blob handle is a handy
** performance improvement, but the blob handle should always be closed
** before control is returned to the user (to prevent a lock being held
** on the database file for longer than necessary). Thus, any virtual table
** method (xFilter etc.) that may directly or indirectly call this function
** must call sqlite3Fts3SegmentsClose() before returning.
*/
SQLITE_PRIVATE int sqlite3Fts3ReadBlock(
Fts3Table *p, /* FTS3 table handle */
sqlite3_int64 iBlockid, /* Access the row with blockid=$iBlockid */
char **paBlob, /* OUT: Blob data in malloc'd buffer */
int *pnBlob, /* OUT: Size of blob data */
int *pnLoad /* OUT: Bytes actually loaded */
){
int rc; /* Return code */
/* pnBlob must be non-NULL. paBlob may be NULL or non-NULL. */
assert( pnBlob );
if( p->pSegments ){
rc = sqlite3_blob_reopen(p->pSegments, iBlockid);
}else{
if( 0==p->zSegmentsTbl ){
p->zSegmentsTbl = sqlite3_mprintf("%s_segments", p->zName);
if( 0==p->zSegmentsTbl ) return SQLITE_NOMEM;
}
rc = sqlite3_blob_open(
p->db, p->zDb, p->zSegmentsTbl, "block", iBlockid, 0, &p->pSegments
);
}
/* The entire tree fits on the root node. Write it to the segdir table. */
rc = fts3WriteSegdir(p, iLevel, iIdx,
0, 0, 0, pWriter->nLeafData, pWriter->aData, pWriter->nData);
}
p->nLeafAdd++;
return rc;
}
/*
** Release all memory held by the SegmentWriter object passed as the
** first argument.
*/
static void fts3SegWriterFree(SegmentWriter *pWriter){
if( pWriter ){
sqlite3_free(pWriter->aData);
sqlite3_free(pWriter->zMalloc);
fts3NodeFree(pWriter->pTree);
sqlite3_free(pWriter);
}
}
/*
** The first value in the apVal[] array is assumed to contain an integer.
** This function tests if there exist any documents with docid values that
** are different from that integer. i.e. if deleting the document with docid
** pRowid would mean the FTS3 table were empty.
**
** If successful, *pisEmpty is set to true if the table is empty except for
** document pRowid, or false otherwise, and SQLITE_OK is returned. If an
** error occurs, an SQLite error code is returned.
*/
static int fts3IsEmpty(Fts3Table *p, sqlite3_value *pRowid, int *pisEmpty){
sqlite3_stmt *pStmt;
int rc;
if( p->zContentTbl ){
/* If using the content=xxx option, assume the table is never empty */
*pisEmpty = 0;
rc = SQLITE_OK;
}else{
rc = fts3SqlStmt(p, SQL_IS_EMPTY, &pStmt, &pRowid);
if( rc==SQLITE_OK ){
if( SQLITE_ROW==sqlite3_step(pStmt) ){
*pisEmpty = sqlite3_column_int(pStmt, 0);
}
rc = sqlite3_reset(pStmt);
}
}
return rc;
}
/*
** Set *pnMax to the largest segment level in the database for the index
** iIndex.
**
** Segment levels are stored in the 'level' column of the %_segdir table.
**
** Return SQLITE_OK if successful, or an SQLite error code if not.
*/
static int fts3SegmentMaxLevel(
Fts3Table *p,
int iLangid,
int iIndex,
sqlite3_int64 *pnMax
){
sqlite3_stmt *pStmt;
int rc;
assert( iIndex>=0 && iIndex<p->nIndex );
/* Set pStmt to the compiled version of:
**
** SELECT max(level) FROM %Q.'%q_segdir' WHERE level BETWEEN ? AND ?
**
** (1024 is actually the value of macro FTS3_SEGDIR_PREFIXLEVEL_STR).
*/
rc = fts3SqlStmt(p, SQL_SELECT_SEGDIR_MAX_LEVEL, &pStmt, 0);
if( rc!=SQLITE_OK ) return rc;
sqlite3_bind_int64(pStmt, 1, getAbsoluteLevel(p, iLangid, iIndex, 0));
sqlite3_bind_int64(pStmt, 2,
getAbsoluteLevel(p, iLangid, iIndex, FTS3_SEGDIR_MAXLEVEL-1)
);
if( SQLITE_ROW==sqlite3_step(pStmt) ){
*pnMax = sqlite3_column_int64(pStmt, 0);
}
return sqlite3_reset(pStmt);
}
/*
** iAbsLevel is an absolute level that may be assumed to exist within
** the database. This function checks if it is the largest level number
** within its index. Assuming no error occurs, *pbMax is set to 1 if
** iAbsLevel is indeed the largest level, or 0 otherwise, and SQLITE_OK
** is returned. If an error occurs, an error code is returned and the
** final value of *pbMax is undefined.
*/
static int fts3SegmentIsMaxLevel(Fts3Table *p, i64 iAbsLevel, int *pbMax){
/* Set pStmt to the compiled version of:
**
** SELECT max(level) FROM %Q.'%q_segdir' WHERE level BETWEEN ? AND ?
**
** (1024 is actually the value of macro FTS3_SEGDIR_PREFIXLEVEL_STR).
*/
sqlite3_stmt *pStmt;
int rc = fts3SqlStmt(p, SQL_SELECT_SEGDIR_MAX_LEVEL, &pStmt, 0);
if( rc!=SQLITE_OK ) return rc;
sqlite3_bind_int64(pStmt, 1, iAbsLevel+1);
sqlite3_bind_int64(pStmt, 2,
((iAbsLevel/FTS3_SEGDIR_MAXLEVEL)+1) * FTS3_SEGDIR_MAXLEVEL
);
*pbMax = 0;
if( SQLITE_ROW==sqlite3_step(pStmt) ){
*pbMax = sqlite3_column_type(pStmt, 0)==SQLITE_NULL;
}
return sqlite3_reset(pStmt);
}
/*
** Delete all entries in the %_segments table associated with the segment
** opened with seg-reader pSeg. This function does not affect the contents
** of the %_segdir table.
*/
static int fts3DeleteSegment(
Fts3Table *p, /* FTS table handle */
Fts3SegReader *pSeg /* Segment to delete */
){
int rc = SQLITE_OK; /* Return code */
if( pSeg->iStartBlock ){
sqlite3_stmt *pDelete; /* SQL statement to delete rows */
rc = fts3SqlStmt(p, SQL_DELETE_SEGMENTS_RANGE, &pDelete, 0);
if( rc==SQLITE_OK ){
sqlite3_bind_int64(pDelete, 1, pSeg->iStartBlock);
sqlite3_bind_int64(pDelete, 2, pSeg->iEndBlock);
sqlite3_step(pDelete);
rc = sqlite3_reset(pDelete);
}
}
return rc;
}
/*
** This function is used after merging multiple segments into a single large
** segment to delete the old, now redundant, segment b-trees. Specifically,
** it:
**
** 1) Deletes all %_segments entries for the segments associated with
** each of the SegReader objects in the array passed as the third
** argument, and
**
** 2) deletes all %_segdir entries with level iLevel, or all %_segdir
** entries regardless of level if (iLevel<0).
**
** SQLITE_OK is returned if successful, otherwise an SQLite error code.
*/
static int fts3DeleteSegdir(
Fts3Table *p, /* Virtual table handle */
int iLangid, /* Language id */
int iIndex, /* Index for p->aIndex */
int iLevel, /* Level of %_segdir entries to delete */
Fts3SegReader **apSegment, /* Array of SegReader objects */
int nReader /* Size of array apSegment */
){
int rc = SQLITE_OK; /* Return Code */
int i; /* Iterator variable */
sqlite3_stmt *pDelete = 0; /* SQL statement to delete rows */
for(i=0; rc==SQLITE_OK && i<nReader; i++){
rc = fts3DeleteSegment(p, apSegment[i]);
}
if( rc!=SQLITE_OK ){
return rc;
}
assert( iLevel>=0 || iLevel==FTS3_SEGCURSOR_ALL );
if( iLevel==FTS3_SEGCURSOR_ALL ){
rc = fts3SqlStmt(p, SQL_DELETE_SEGDIR_RANGE, &pDelete, 0);
if( rc==SQLITE_OK ){
sqlite3_bind_int64(pDelete, 1, getAbsoluteLevel(p, iLangid, iIndex, 0));
sqlite3_bind_int64(pDelete, 2,
getAbsoluteLevel(p, iLangid, iIndex, FTS3_SEGDIR_MAXLEVEL-1)
);
}
}else{
rc = fts3SqlStmt(p, SQL_DELETE_SEGDIR_LEVEL, &pDelete, 0);
if( rc==SQLITE_OK ){
sqlite3_bind_int64(
pDelete, 1, getAbsoluteLevel(p, iLangid, iIndex, iLevel)
);
}
}
if( rc==SQLITE_OK ){
sqlite3_step(pDelete);
rc = sqlite3_reset(pDelete);
}
return rc;
}
/*
** When this function is called, buffer *ppList (size *pnList bytes) contains
** a position list that may (or may not) feature multiple columns. This
** function adjusts the pointer *ppList and the length *pnList so that they
** identify the subset of the position list that corresponds to column iCol.
**
** If there are no entries in the input position list for column iCol, then
** *pnList is set to zero before returning.
**
** If parameter bZero is non-zero, then any part of the input list following
** the end of the output list is zeroed before returning.
*/
static void fts3ColumnFilter(
int iCol, /* Column to filter on */
int bZero, /* Zero out anything following *ppList */
char **ppList, /* IN/OUT: Pointer to position list */
int *pnList /* IN/OUT: Size of buffer *ppList in bytes */
){
char *pList = *ppList;
int nList = *pnList;
char *pEnd = &pList[nList];
int iCurrent = 0;
char *p = pList;
assert( iCol>=0 );
while( 1 ){
char c = 0;
while( p<pEnd && (c | *p)&0xFE ) c = *p++ & 0x80;
if( iCol==iCurrent ){
nList = (int)(p - pList);
break;
}
nList -= (int)(p - pList);
pList = p;
if( nList==0 ){
break;
}
p = &pList[1];
p += fts3GetVarint32(p, &iCurrent);
}
if( bZero && &pList[nList]!=pEnd ){
memset(&pList[nList], 0, pEnd - &pList[nList]);
}
*ppList = pList;
*pnList = nList;
rc = fts3SqlStmt(p, SQL_UPDATE_LEVEL_IDX, &pUpdate1, 0);
}
if( rc==SQLITE_OK ){
rc = fts3SqlStmt(p, SQL_UPDATE_LEVEL, &pUpdate2, 0);
}
if( rc==SQLITE_OK ){
/* Loop through all %_segdir entries for segments in this index with
** levels equal to or greater than iAbsLevel. As each entry is visited,
** updated it to set (level = -1) and (idx = N), where N is 0 for the
** oldest segment in the range, 1 for the next oldest, and so on.
**
** In other words, move all segments being promoted to level -1,
** setting the "idx" fields as appropriate to keep them in the same
** order. The contents of level -1 (which is never used, except
** transiently here), will be moved back to level iAbsLevel below. */
sqlite3_bind_int64(pRange, 1, iAbsLevel);
while( SQLITE_ROW==sqlite3_step(pRange) ){
sqlite3_bind_int(pUpdate1, 1, iIdx++);
sqlite3_bind_int(pUpdate1, 2, sqlite3_column_int(pRange, 0));
sqlite3_bind_int(pUpdate1, 3, sqlite3_column_int(pRange, 1));
sqlite3_step(pUpdate1);
rc = sqlite3_reset(pUpdate1);
if( rc!=SQLITE_OK ){
sqlite3_reset(pRange);
break;
}
}
}
if( rc==SQLITE_OK ){
rc = sqlite3_reset(pRange);
}
/* Move level -1 to level iAbsLevel */
if( rc==SQLITE_OK ){
sqlite3_bind_int64(pUpdate2, 1, iAbsLevel);
sqlite3_step(pUpdate2);
rc = sqlite3_reset(pUpdate2);
}
}
}
return rc;
}
/*
** Merge all level iLevel segments in the database into a single
** iLevel+1 segment. Or, if iLevel<0, merge all segments into a
** single segment with a level equal to the numerically largest level
** currently present in the database.
**
** If this function is called with iLevel<0, but there is only one
** segment in the database, SQLITE_DONE is returned immediately.
** Otherwise, if successful, SQLITE_OK is returned. If an error occurs,
** an SQLite error code is returned.
*/
static int fts3SegmentMerge(
Fts3Table *p,
int iLangid, /* Language id to merge */
int iIndex, /* Index in p->aIndex[] to merge */
int iLevel /* Level to merge */
){
int rc; /* Return code */
int iIdx = 0; /* Index of new segment */
sqlite3_int64 iNewLevel = 0; /* Level/index to create new segment at */
SegmentWriter *pWriter = 0; /* Used to write the new, merged, segment */
Fts3SegFilter filter; /* Segment term filter condition */
Fts3MultiSegReader csr; /* Cursor to iterate through level(s) */
int bIgnoreEmpty = 0; /* True to ignore empty segments */
i64 iMaxLevel = 0; /* Max level number for this index/langid */
assert( iLevel==FTS3_SEGCURSOR_ALL
|| iLevel==FTS3_SEGCURSOR_PENDING
|| iLevel>=0
);
assert( iLevel<FTS3_SEGDIR_MAXLEVEL );
assert( iIndex>=0 && iIndex<p->nIndex );
rc = sqlite3Fts3SegReaderCursor(p, iLangid, iIndex, iLevel, 0, 0, 1, 0, &csr);
if( rc!=SQLITE_OK || csr.nSegment==0 ) goto finished;
if( iLevel!=FTS3_SEGCURSOR_PENDING ){
rc = fts3SegmentMaxLevel(p, iLangid, iIndex, &iMaxLevel);
if( rc!=SQLITE_OK ) goto finished;
}
if( iLevel==FTS3_SEGCURSOR_ALL ){
/* This call is to merge all segments in the database to a single
** segment. The level of the new segment is equal to the numerically
** greatest segment level currently present in the database for this
** index. The idx of the new segment is always 0. */
if( csr.nSegment==1 ){
rc = SQLITE_DONE;
goto finished;
}
iNewLevel = iMaxLevel;
bIgnoreEmpty = 1;
}else{
/* This call is to merge all segments at level iLevel. find the next
** available segment index at level iLevel+1. The call to
** fts3AllocateSegdirIdx() will merge the segments at level iLevel+1 to
** a single iLevel+2 segment if necessary. */
assert( FTS3_SEGCURSOR_PENDING==-1 );
iNewLevel = getAbsoluteLevel(p, iLangid, iIndex, iLevel+1);
rc = fts3AllocateSegdirIdx(p, iLangid, iIndex, iLevel+1, &iIdx);
bIgnoreEmpty = (iLevel!=FTS3_SEGCURSOR_PENDING) && (iNewLevel>iMaxLevel);
}
if( rc!=SQLITE_OK ) goto finished;
assert( csr.nSegment>0 );
assert( iNewLevel>=getAbsoluteLevel(p, iLangid, iIndex, 0) );
assert( iNewLevel<getAbsoluteLevel(p, iLangid, iIndex,FTS3_SEGDIR_MAXLEVEL) );
memset(&filter, 0, sizeof(Fts3SegFilter));
filter.flags = FTS3_SEGMENT_REQUIRE_POS;
filter.flags |= (bIgnoreEmpty ? FTS3_SEGMENT_IGNORE_EMPTY : 0);
rc = sqlite3Fts3SegReaderStart(p, &csr, &filter);
while( SQLITE_OK==rc ){
rc = sqlite3Fts3SegReaderStep(p, &csr);
if( rc!=SQLITE_ROW ) break;
rc = fts3SegWriterAdd(p, &pWriter, 1,
csr.zTerm, csr.nTerm, csr.aDoclist, csr.nDoclist);
}
if( rc!=SQLITE_OK ) goto finished;
assert( pWriter || bIgnoreEmpty );
if( iLevel!=FTS3_SEGCURSOR_PENDING ){
rc = fts3DeleteSegdir(
p, iLangid, iIndex, iLevel, csr.apSegment, csr.nSegment
);
if( rc!=SQLITE_OK ) goto finished;
}
if( pWriter ){
rc = fts3SegWriterFlush(p, pWriter, iNewLevel, iIdx);
if( rc==SQLITE_OK ){
if( iLevel==FTS3_SEGCURSOR_PENDING || iNewLevel<iMaxLevel ){
rc = fts3PromoteSegments(p, iNewLevel, pWriter->nLeafData);
}
}
}
finished:
fts3SegWriterFree(pWriter);
sqlite3Fts3SegReaderFinish(&csr);
return rc;
}
/*
** Flush the contents of pendingTerms to level 0 segments.
*/
SQLITE_PRIVATE int sqlite3Fts3PendingTermsFlush(Fts3Table *p){
int rc = SQLITE_OK;
int i;
for(i=0; rc==SQLITE_OK && i<p->nIndex; i++){
rc = fts3SegmentMerge(p, p->iPrevLangid, i, FTS3_SEGCURSOR_PENDING);
if( rc==SQLITE_DONE ) rc = SQLITE_OK;
}
sqlite3Fts3PendingTermsClear(p);
/* Determine the auto-incr-merge setting if unknown. If enabled,
** estimate the number of leaf blocks of content to be written
*/
if( rc==SQLITE_OK && p->bHasStat
&& p->nAutoincrmerge==0xff && p->nLeafAdd>0
){
sqlite3_stmt *pStmt = 0;
rc = fts3SqlStmt(p, SQL_SELECT_STAT, &pStmt, 0);
if( rc==SQLITE_OK ){
sqlite3_bind_int(pStmt, 1, FTS_STAT_AUTOINCRMERGE);
rc = sqlite3_step(pStmt);
if( rc==SQLITE_ROW ){
p->nAutoincrmerge = sqlite3_column_int(pStmt, 0);
if( p->nAutoincrmerge==1 ) p->nAutoincrmerge = 8;
}else if( rc==SQLITE_DONE ){
p->nAutoincrmerge = 0;
}
rc = sqlite3_reset(pStmt);
}
}
return rc;
}
/*
** Encode N integers as varints into a blob.
*/
static void fts3EncodeIntArray(
int N, /* The number of integers to encode */
u32 *a, /* The integer values */
char *zBuf, /* Write the BLOB here */
int *pNBuf /* Write number of bytes if zBuf[] used here */
){
int i, j;
for(i=j=0; i<N; i++){
j += sqlite3Fts3PutVarint(&zBuf[j], (sqlite3_int64)a[i]);
}
*pNBuf = j;
}
/*
** Decode a blob of varints into N integers
*/
static void fts3DecodeIntArray(
int N, /* The number of integers to decode */
u32 *a, /* Write the integer values */
const char *zBuf, /* The BLOB containing the varints */
int nBuf /* size of the BLOB */
){
int i, j;
UNUSED_PARAMETER(nBuf);
for(i=j=0; i<N; i++){
sqlite3_int64 x;
j += sqlite3Fts3GetVarint(&zBuf[j], &x);
assert(j<=nBuf);
a[i] = (u32)(x & 0xffffffff);
}
const int nStat = p->nColumn+2;
if( *pRC ) return;
a = sqlite3_malloc( (sizeof(u32)+10)*nStat );
if( a==0 ){
*pRC = SQLITE_NOMEM;
return;
}
pBlob = (char*)&a[nStat];
rc = fts3SqlStmt(p, SQL_SELECT_STAT, &pStmt, 0);
if( rc ){
sqlite3_free(a);
*pRC = rc;
return;
}
sqlite3_bind_int(pStmt, 1, FTS_STAT_DOCTOTAL);
if( sqlite3_step(pStmt)==SQLITE_ROW ){
fts3DecodeIntArray(nStat, a,
sqlite3_column_blob(pStmt, 0),
sqlite3_column_bytes(pStmt, 0));
}else{
memset(a, 0, sizeof(u32)*(nStat) );
}
rc = sqlite3_reset(pStmt);
if( rc!=SQLITE_OK ){
sqlite3_free(a);
*pRC = rc;
return;
}
if( nChng<0 && a[0]<(u32)(-nChng) ){
a[0] = 0;
}else{
a[0] += nChng;
}
for(i=0; i<p->nColumn+1; i++){
u32 x = a[i+1];
if( x+aSzIns[i] < aSzDel[i] ){
x = 0;
}else{
x = x + aSzIns[i] - aSzDel[i];
}
a[i+1] = x;
}
fts3EncodeIntArray(nStat, a, pBlob, &nBlob);
rc = fts3SqlStmt(p, SQL_REPLACE_STAT, &pStmt, 0);
if( rc ){
sqlite3_free(a);
*pRC = rc;
return;
}
sqlite3_bind_int(pStmt, 1, FTS_STAT_DOCTOTAL);
sqlite3_bind_blob(pStmt, 2, pBlob, nBlob, SQLITE_STATIC);
sqlite3_step(pStmt);
*pRC = sqlite3_reset(pStmt);
sqlite3_free(a);
}
/*
** Merge the entire database so that there is one segment for each
** iIndex/iLangid combination.
*/
static int fts3DoOptimize(Fts3Table *p, int bReturnDone){
int bSeenDone = 0;
int rc;
sqlite3_stmt *pAllLangid = 0;
rc = fts3SqlStmt(p, SQL_SELECT_ALL_LANGID, &pAllLangid, 0);
if( rc==SQLITE_OK ){
int rc2;
sqlite3_bind_int(pAllLangid, 1, p->nIndex);
while( sqlite3_step(pAllLangid)==SQLITE_ROW ){
int i;
int iLangid = sqlite3_column_int(pAllLangid, 0);
for(i=0; rc==SQLITE_OK && i<p->nIndex; i++){
rc = fts3SegmentMerge(p, iLangid, i, FTS3_SEGCURSOR_ALL);
if( rc==SQLITE_DONE ){
bSeenDone = 1;
rc = SQLITE_OK;
}
}
}
rc2 = sqlite3_reset(pAllLangid);
if( rc==SQLITE_OK ) rc = rc2;
}
sqlite3Fts3SegmentsClose(p);
sqlite3Fts3PendingTermsClear(p);
return (rc==SQLITE_OK && bReturnDone && bSeenDone) ? SQLITE_DONE : rc;
}
/*
** This function is called when the user executes the following statement:
**
** INSERT INTO <tbl>(<tbl>) VALUES('rebuild');
**
** The entire FTS index is discarded and rebuilt. If the table is one
** created using the content=xxx option, then the new index is based on
** the current contents of the xxx table. Otherwise, it is rebuilt based
** on the contents of the %_content table.
*/
static int fts3DoRebuild(Fts3Table *p){
int rc; /* Return Code */
rc = fts3DeleteAll(p, 0);
if( rc==SQLITE_OK ){
u32 *aSz = 0;
u32 *aSzIns = 0;
u32 *aSzDel = 0;
sqlite3_stmt *pStmt = 0;
int nEntry = 0;
/* Compose and prepare an SQL statement to loop through the content table */
char *zSql = sqlite3_mprintf("SELECT %s" , p->zReadExprlist);
if( !zSql ){
rc = SQLITE_NOMEM;
}else{
rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, 0);
sqlite3_free(zSql);
}
if( rc==SQLITE_OK ){
int nByte = sizeof(u32) * (p->nColumn+1)*3;
aSz = (u32 *)sqlite3_malloc(nByte);
if( aSz==0 ){
rc = SQLITE_NOMEM;
}else{
memset(aSz, 0, nByte);
aSzIns = &aSz[p->nColumn+1];
aSzDel = &aSzIns[p->nColumn+1];
}
}
while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){
int iCol;
int iLangid = langidFromSelect(p, pStmt);
rc = fts3PendingTermsDocid(p, iLangid, sqlite3_column_int64(pStmt, 0));
memset(aSz, 0, sizeof(aSz[0]) * (p->nColumn+1));
for(iCol=0; rc==SQLITE_OK && iCol<p->nColumn; iCol++){
if( p->abNotindexed[iCol]==0 ){
const char *z = (const char *) sqlite3_column_text(pStmt, iCol+1);
rc = fts3PendingTermsAdd(p, iLangid, z, iCol, &aSz[iCol]);
aSz[p->nColumn] += sqlite3_column_bytes(pStmt, iCol+1);
}
}
if( p->bHasDocsize ){
fts3InsertDocsize(&rc, p, aSz);
}
if( rc!=SQLITE_OK ){
sqlite3_finalize(pStmt);
pStmt = 0;
}else{
nEntry++;
for(iCol=0; iCol<=p->nColumn; iCol++){
aSzIns[iCol] += aSz[iCol];
}
}
}
if( p->bFts4 ){
fts3UpdateDocTotals(&rc, p, aSzIns, aSzDel, nEntry);
}
sqlite3_free(aSz);
if( pStmt ){
int rc2 = sqlite3_finalize(pStmt);
if( rc==SQLITE_OK ){
rc = rc2;
}
}
}
return rc;
}
/*
** This function opens a cursor used to read the input data for an
** incremental merge operation. Specifically, it opens a cursor to scan
** the oldest nSeg segments (idx=0 through idx=(nSeg-1)) in absolute
** level iAbsLevel.
*/
static int fts3IncrmergeCsr(
Fts3Table *p, /* FTS3 table handle */
sqlite3_int64 iAbsLevel, /* Absolute level to open */
int nSeg, /* Number of segments to merge */
Fts3MultiSegReader *pCsr /* Cursor object to populate */
){
int rc; /* Return Code */
sqlite3_stmt *pStmt = 0; /* Statement used to read %_segdir entry */
int nByte; /* Bytes allocated at pCsr->apSegment[] */
/* Allocate space for the Fts3MultiSegReader.aCsr[] array */
memset(pCsr, 0, sizeof(*pCsr));
nByte = sizeof(Fts3SegReader *) * nSeg;
pCsr->apSegment = (Fts3SegReader **)sqlite3_malloc(nByte);
if( pCsr->apSegment==0 ){
rc = SQLITE_NOMEM;
}else{
memset(pCsr->apSegment, 0, nByte);
rc = fts3SqlStmt(p, SQL_SELECT_LEVEL, &pStmt, 0);
}
if( z[0]==',' && z[1]!='\0' ){
z++;
nMin = fts3Getint(&z);
}
if( z[0]!='\0' || nMin<2 ){
rc = SQLITE_ERROR;
}else{
rc = SQLITE_OK;
if( !p->bHasStat ){
assert( p->bFts4==0 );
sqlite3Fts3CreateStatTable(&rc, p);
}
if( rc==SQLITE_OK ){
rc = sqlite3Fts3Incrmerge(p, nMerge, nMin);
}
sqlite3Fts3SegmentsClose(p);
}
return rc;
}
/*
** Process statements of the form:
**
** INSERT INTO table(table) VALUES('automerge=X');
**
** where X is an integer. X==0 means to turn automerge off. X!=0 means
** turn it on. The setting is persistent.
*/
static int fts3DoAutoincrmerge(
Fts3Table *p, /* FTS3 table handle */
const char *zParam /* Nul-terminated string containing boolean */
){
int rc = SQLITE_OK;
sqlite3_stmt *pStmt = 0;
p->nAutoincrmerge = fts3Getint(&zParam);
if( p->nAutoincrmerge==1 || p->nAutoincrmerge>FTS3_MERGE_COUNT ){
p->nAutoincrmerge = 8;
}
if( !p->bHasStat ){
assert( p->bFts4==0 );
sqlite3Fts3CreateStatTable(&rc, p);
if( rc ) return rc;
}
rc = fts3SqlStmt(p, SQL_REPLACE_STAT, &pStmt, 0);
if( rc ) return rc;
sqlite3_bind_int(pStmt, 1, FTS_STAT_AUTOINCRMERGE);
sqlite3_bind_int(pStmt, 2, p->nAutoincrmerge);
sqlite3_step(pStmt);
rc = sqlite3_reset(pStmt);
return rc;
}
/*
** Return a 64-bit checksum for the FTS index entry specified by the
** arguments to this function.
*/
static u64 fts3ChecksumEntry(
const char *zTerm, /* Pointer to buffer containing term */
int nTerm, /* Size of zTerm in bytes */
int iLangid, /* Language id for current row */
int iIndex, /* Index (0..Fts3Table.nIndex-1) */
i64 iDocid, /* Docid for current row. */
int iCol, /* Column number */
int iPos /* Position */
){
int i;
u64 ret = (u64)iDocid;
ret += (ret<<3) + iLangid;
ret += (ret<<3) + iIndex;
ret += (ret<<3) + iCol;
ret += (ret<<3) + iPos;
for(i=0; i<nTerm; i++) ret += (ret<<3) + zTerm[i];
return ret;
}
/*
** Return a checksum of all entries in the FTS index that correspond to
** language id iLangid. The checksum is calculated by XORing the checksums
** of each individual entry (see fts3ChecksumEntry()) together.
**
** If successful, the checksum value is returned and *pRc set to SQLITE_OK.
** Otherwise, if an error occurs, *pRc is set to an SQLite error code. The
** return value is undefined in this case.
*/
static u64 fts3ChecksumIndex(
Fts3Table *p, /* FTS3 table handle */
int iLangid, /* Language id to return cksum for */
int iIndex, /* Index to cksum (0..p->nIndex-1) */
int *pRc /* OUT: Return code */
){
Fts3SegFilter filter;
Fts3MultiSegReader csr;
int rc;
u64 cksum = 0;
assert( *pRc==SQLITE_OK );
memset(&filter, 0, sizeof(filter));
memset(&csr, 0, sizeof(csr));
filter.flags = FTS3_SEGMENT_REQUIRE_POS|FTS3_SEGMENT_IGNORE_EMPTY;
filter.flags |= FTS3_SEGMENT_SCAN;
rc = sqlite3Fts3SegReaderCursor(
p, iLangid, iIndex, FTS3_SEGCURSOR_ALL, 0, 0, 0, 1,&csr
);
if( rc==SQLITE_OK ){
rc = sqlite3Fts3SegReaderStart(p, &csr, &filter);
}
if( rc==SQLITE_OK ){
while( SQLITE_ROW==(rc = sqlite3Fts3SegReaderStep(p, &csr)) ){
char *pCsr = csr.aDoclist;
char *pEnd = &pCsr[csr.nDoclist];
i64 iDocid = 0;
i64 iCol = 0;
i64 iPos = 0;
pCsr += sqlite3Fts3GetVarint(pCsr, &iDocid);
while( pCsr<pEnd ){
i64 iVal = 0;
pCsr += sqlite3Fts3GetVarint(pCsr, &iVal);
if( pCsr<pEnd ){
if( iVal==0 || iVal==1 ){
iCol = 0;
iPos = 0;
if( iVal ){
pCsr += sqlite3Fts3GetVarint(pCsr, &iCol);
}else{
pCsr += sqlite3Fts3GetVarint(pCsr, &iVal);
iDocid += iVal;
}
}else{
iPos += (iVal - 2);
cksum = cksum ^ fts3ChecksumEntry(
csr.zTerm, csr.nTerm, iLangid, iIndex, iDocid,
(int)iCol, (int)iPos
);
}
}
}
}
}
sqlite3Fts3SegReaderFinish(&csr);
*pRc = rc;
return cksum;
}
/*
** Check if the contents of the FTS index match the current contents of the
** content table. If no error occurs and the contents do match, set *pbOk
** to true and return SQLITE_OK. Or if the contents do not match, set *pbOk
** to false before returning.
**
** If an error occurs (e.g. an OOM or IO error), return an SQLite error
** code. The final value of *pbOk is undefined in this case.
*/
static int fts3IntegrityCheck(Fts3Table *p, int *pbOk){
int rc = SQLITE_OK; /* Return code */
u64 cksum1 = 0; /* Checksum based on FTS index contents */
u64 cksum2 = 0; /* Checksum based on %_content contents */
sqlite3_stmt *pAllLangid = 0; /* Statement to return all language-ids */
/* This block calculates the checksum according to the FTS index. */
rc = fts3SqlStmt(p, SQL_SELECT_ALL_LANGID, &pAllLangid, 0);
if( rc==SQLITE_OK ){
int rc2;
sqlite3_bind_int(pAllLangid, 1, p->nIndex);
while( rc==SQLITE_OK && sqlite3_step(pAllLangid)==SQLITE_ROW ){
int iLangid = sqlite3_column_int(pAllLangid, 0);
int i;
for(i=0; i<p->nIndex; i++){
cksum1 = cksum1 ^ fts3ChecksumIndex(p, iLangid, i, &rc);
}
}
rc2 = sqlite3_reset(pAllLangid);
if( rc==SQLITE_OK ) rc = rc2;
}
/* This block calculates the checksum according to the %_content table */
rc = fts3SqlStmt(p, SQL_SELECT_ALL_LANGID, &pAllLangid, 0);
if( rc==SQLITE_OK ){
sqlite3_tokenizer_module const *pModule = p->pTokenizer->pModule;
sqlite3_stmt *pStmt = 0;
char *zSql;
zSql = sqlite3_mprintf("SELECT %s" , p->zReadExprlist);
if( !zSql ){
rc = SQLITE_NOMEM;
}else{
rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, 0);
sqlite3_free(zSql);
}
while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pStmt) ){
i64 iDocid = sqlite3_column_int64(pStmt, 0);
int iLang = langidFromSelect(p, pStmt);
int iCol;
for(iCol=0; rc==SQLITE_OK && iCol<p->nColumn; iCol++){
if( p->abNotindexed[iCol]==0 ){
const char *zText = (const char *)sqlite3_column_text(pStmt, iCol+1);
int nText = sqlite3_column_bytes(pStmt, iCol+1);
sqlite3_tokenizer_cursor *pT = 0;
rc = sqlite3Fts3OpenTokenizer(p->pTokenizer, iLang, zText, nText,&pT);
while( rc==SQLITE_OK ){
char const *zToken; /* Buffer containing token */
int nToken = 0; /* Number of bytes in token */
int iDum1 = 0, iDum2 = 0; /* Dummy variables */
int iPos = 0; /* Position of token in zText */
rc = pModule->xNext(pT, &zToken, &nToken, &iDum1, &iDum2, &iPos);
if( rc==SQLITE_OK ){
int i;
cksum2 = cksum2 ^ fts3ChecksumEntry(
zToken, nToken, iLang, 0, iDocid, iCol, iPos
);
for(i=1; i<p->nIndex; i++){
if( p->aIndex[i].nPrefix<=nToken ){
cksum2 = cksum2 ^ fts3ChecksumEntry(
zToken, p->aIndex[i].nPrefix, iLang, i, iDocid, iCol, iPos
);
}
}
}
}
if( pT ) pModule->xClose(pT);
if( rc==SQLITE_DONE ) rc = SQLITE_OK;
}
}
}
sqlite3_finalize(pStmt);
}
*pbOk = (cksum1==cksum2);
return rc;
}
/*
** Run the integrity-check. If no error occurs and the current contents of
** the FTS index are correct, return SQLITE_OK. Or, if the contents of the
** FTS index are incorrect, return SQLITE_CORRUPT_VTAB.
**
** Or, if an error (e.g. an OOM or IO error) occurs, return an SQLite
** error code.
**
** The integrity-check works as follows. For each token and indexed token
** prefix in the document set, a 64-bit checksum is calculated (by code
** in fts3ChecksumEntry()) based on the following:
**
** + The index number (0 for the main index, 1 for the first prefix
** index etc.),
** + The token (or token prefix) text itself,
** + The language-id of the row it appears in,
** + The docid of the row it appears in,
}
return rc;
}
#ifndef SQLITE_DISABLE_FTS4_DEFERRED
/*
** Delete all cached deferred doclists. Deferred doclists are cached
** (allocated) by the sqlite3Fts3CacheDeferredDoclists() function.
*/
SQLITE_PRIVATE void sqlite3Fts3FreeDeferredDoclists(Fts3Cursor *pCsr){
Fts3DeferredToken *pDef;
for(pDef=pCsr->pDeferred; pDef; pDef=pDef->pNext){
fts3PendingListDelete(pDef->pList);
pDef->pList = 0;
}
}
/*
** Free all entries in the pCsr->pDeffered list. Entries are added to
** this list using sqlite3Fts3DeferToken().
*/
SQLITE_PRIVATE void sqlite3Fts3FreeDeferredTokens(Fts3Cursor *pCsr){
Fts3DeferredToken *pDef;
Fts3DeferredToken *pNext;
for(pDef=pCsr->pDeferred; pDef; pDef=pNext){
pNext = pDef->pNext;
fts3PendingListDelete(pDef->pList);
sqlite3_free(pDef);
}
pCsr->pDeferred = 0;
}
/*
** Generate deferred-doclists for all tokens in the pCsr->pDeferred list
** based on the row that pCsr currently points to.
**
** A deferred-doclist is like any other doclist with position information
** included, except that it only contains entries for a single row of the
** table, not for all rows.
*/
SQLITE_PRIVATE int sqlite3Fts3CacheDeferredDoclists(Fts3Cursor *pCsr){
int rc = SQLITE_OK; /* Return code */
if( pCsr->pDeferred ){
int i; /* Used to iterate through table columns */
sqlite3_int64 iDocid; /* Docid of the row pCsr points to */
Fts3DeferredToken *pDef; /* Used to iterate through deferred tokens */
Fts3Table *p = (Fts3Table *)pCsr->base.pVtab;
sqlite3_tokenizer *pT = p->pTokenizer;
sqlite3_tokenizer_module const *pModule = pT->pModule;
assert( pCsr->isRequireSeek==0 );
iDocid = sqlite3_column_int64(pCsr->pStmt, 0);
for(i=0; i<p->nColumn && rc==SQLITE_OK; i++){
if( p->abNotindexed[i]==0 ){
const char *zText = (const char *)sqlite3_column_text(pCsr->pStmt, i+1);
sqlite3_tokenizer_cursor *pTC = 0;
rc = sqlite3Fts3OpenTokenizer(pT, pCsr->iLangid, zText, -1, &pTC);
while( rc==SQLITE_OK ){
char const *zToken; /* Buffer containing token */
int nToken = 0; /* Number of bytes in token */
int iDum1 = 0, iDum2 = 0; /* Dummy variables */
int iPos = 0; /* Position of token in zText */
rc = pModule->xNext(pTC, &zToken, &nToken, &iDum1, &iDum2, &iPos);
for(pDef=pCsr->pDeferred; pDef && rc==SQLITE_OK; pDef=pDef->pNext){
Fts3PhraseToken *pPT = pDef->pToken;
if( (pDef->iCol>=p->nColumn || pDef->iCol==i)
&& (pPT->bFirst==0 || iPos==0)
&& (pPT->n==nToken || (pPT->isPrefix && pPT->n<nToken))
&& (0==memcmp(zToken, pPT->z, pPT->n))
){
fts3PendingListAppend(&pDef->pList, iDocid, i, iPos, &rc);
}
}
}
if( pTC ) pModule->xClose(pTC);
if( rc==SQLITE_DONE ) rc = SQLITE_OK;
}
}
for(pDef=pCsr->pDeferred; pDef && rc==SQLITE_OK; pDef=pDef->pNext){
if( pDef->pList ){
rc = fts3PendingListAppendVarint(&pDef->pList, 0);
}
}
}
return rc;
}
SQLITE_PRIVATE int sqlite3Fts3DeferredTokenList(
Fts3DeferredToken *p,
char **ppData,
int *pnData
){
char *pRet;
int nSkip;
sqlite3_int64 dummy;
*ppData = 0;
*pnData = 0;
if( p->pList==0 ){
return SQLITE_OK;
}
pRet = (char *)sqlite3_malloc(p->pList->nData);
if( !pRet ) return SQLITE_NOMEM;
nSkip = sqlite3Fts3GetVarint(p->pList->aData, &dummy);
*pnData = p->pList->nData - nSkip;
*ppData = pRet;
memcpy(pRet, &p->pList->aData[nSkip], *pnData);
return SQLITE_OK;
}
pDeferred->pNext = pCsr->pDeferred;
pDeferred->iCol = iCol;
pCsr->pDeferred = pDeferred;
assert( pToken->pDeferred==0 );
pToken->pDeferred = pDeferred;
return SQLITE_OK;
}
#endif
/*
** SQLite value pRowid contains the rowid of a row that may or may not be
** present in the FTS3 table. If it is, delete it and adjust the contents
** of subsiduary data structures accordingly.
*/
static int fts3DeleteByRowid(
Fts3Table *p,
sqlite3_value *pRowid,
int *pnChng, /* IN/OUT: Decrement if row is deleted */
u32 *aSzDel
){
int rc = SQLITE_OK; /* Return code */
int bFound = 0; /* True if *pRowid really is in the table */
fts3DeleteTerms(&rc, p, pRowid, aSzDel, &bFound);
if( bFound && rc==SQLITE_OK ){
int isEmpty = 0; /* Deleting *pRowid leaves the table empty */
rc = fts3IsEmpty(p, pRowid, &isEmpty);
if( rc==SQLITE_OK ){
if( isEmpty ){
/* Deleting this row means the whole table is empty. In this case
** delete the contents of all three tables and throw away any
** data in the pendingTerms hash table. */
rc = fts3DeleteAll(p, 1);
*pnChng = 0;
memset(aSzDel, 0, sizeof(u32) * (p->nColumn+1) * 2);
}else{
*pnChng = *pnChng - 1;
if( p->zContentTbl==0 ){
fts3SqlExec(&rc, p, SQL_DELETE_CONTENT, &pRowid);
}
if( p->bHasDocsize ){
fts3SqlExec(&rc, p, SQL_DELETE_DOCSIZE, &pRowid);
}
}
}
}
return rc;
}
/*
** This function does the work for the xUpdate method of FTS3 virtual
** tables. The schema of the virtual table being:
**
** CREATE TABLE <table name>(
** <user columns>,
** <table name> HIDDEN,
** docid HIDDEN,
** <langid> HIDDEN
** );
**
**
*/
SQLITE_PRIVATE int sqlite3Fts3UpdateMethod(
sqlite3_vtab *pVtab, /* FTS3 vtab object */
int nArg, /* Size of argument array */
sqlite3_value **apVal, /* Array of arguments */
sqlite_int64 *pRowid /* OUT: The affected (or effected) rowid */
){
Fts3Table *p = (Fts3Table *)pVtab;
int rc = SQLITE_OK; /* Return Code */
int isRemove = 0; /* True for an UPDATE or DELETE */
u32 *aSzIns = 0; /* Sizes of inserted documents */
u32 *aSzDel = 0; /* Sizes of deleted documents */
int nChng = 0; /* Net change in number of documents */
int bInsertDone = 0;
/* At this point it must be known if the %_stat table exists or not.
** So bHasStat may not be 2. */
assert( p->bHasStat==0 || p->bHasStat==1 );
assert( p->pSegments==0 );
assert(
nArg==1 /* DELETE operations */
|| nArg==(2 + p->nColumn + 3) /* INSERT or UPDATE operations */
);
/* Check for a "special" INSERT operation. One of the form:
**
** INSERT INTO xyz(xyz) VALUES('command');
*/
if( nArg>1
&& sqlite3_value_type(apVal[0])==SQLITE_NULL
&& sqlite3_value_type(apVal[p->nColumn+2])!=SQLITE_NULL
){
rc = fts3SpecialInsert(p, apVal[p->nColumn+2]);
goto update_out;
}
if( nArg>1 && sqlite3_value_int(apVal[2 + p->nColumn + 2])<0 ){
rc = SQLITE_CONSTRAINT;
goto update_out;
}
/* Allocate space to hold the change in document sizes */
aSzDel = sqlite3_malloc( sizeof(aSzDel[0])*(p->nColumn+1)*2 );
if( aSzDel==0 ){
rc = SQLITE_NOMEM;
goto update_out;
}
aSzIns = &aSzDel[p->nColumn+1];
memset(aSzDel, 0, sizeof(aSzDel[0])*(p->nColumn+1)*2);
rc = fts3Writelock(p);
if( rc!=SQLITE_OK ) goto update_out;
/* If this is an INSERT operation, or an UPDATE that modifies the rowid
** value, then this operation requires constraint handling.
**
** If the on-conflict mode is REPLACE, this means that the existing row
** should be deleted from the database before inserting the new row. Or,
** if the on-conflict mode is other than REPLACE, then this method must
** detect the conflict and return SQLITE_CONSTRAINT before beginning to
** modify the database file.
*/
if( nArg>1 && p->zContentTbl==0 ){
/* Find the value object that holds the new rowid value. */
sqlite3_value *pNewRowid = apVal[3+p->nColumn];
if( sqlite3_value_type(pNewRowid)==SQLITE_NULL ){
pNewRowid = apVal[1];
}
if( sqlite3_value_type(pNewRowid)!=SQLITE_NULL && (
sqlite3_value_type(apVal[0])==SQLITE_NULL
|| sqlite3_value_int64(apVal[0])!=sqlite3_value_int64(pNewRowid)
)){
/* The new rowid is not NULL (in this case the rowid will be
** automatically assigned and there is no chance of a conflict), and
** the statement is either an INSERT or an UPDATE that modifies the
** rowid column. So if the conflict mode is REPLACE, then delete any
** existing row with rowid=pNewRowid.
**
** Or, if the conflict mode is not REPLACE, insert the new record into
** the %_content table. If we hit the duplicate rowid constraint (or any
** other error) while doing so, return immediately.
**
** This branch may also run if pNewRowid contains a value that cannot
** be losslessly converted to an integer. In this case, the eventual
** call to fts3InsertData() (either just below or further on in this
** function) will return SQLITE_MISMATCH. If fts3DeleteByRowid is
** invoked, it will delete zero rows (since no row will have
** docid=$pNewRowid if $pNewRowid is not an integer value).
*/
if( sqlite3_vtab_on_conflict(p->db)==SQLITE_REPLACE ){
rc = fts3DeleteByRowid(p, pNewRowid, &nChng, aSzDel);
}else{
rc = fts3InsertData(p, apVal, pRowid);
bInsertDone = 1;
}
}
}
if( rc!=SQLITE_OK ){
goto update_out;
}
/* If this is a DELETE or UPDATE operation, remove the old record. */
if( sqlite3_value_type(apVal[0])!=SQLITE_NULL ){
assert( sqlite3_value_type(apVal[0])==SQLITE_INTEGER );
rc = fts3DeleteByRowid(p, apVal[0], &nChng, aSzDel);
isRemove = 1;
}
/* If this is an INSERT or UPDATE operation, insert the new record. */
if( nArg>1 && rc==SQLITE_OK ){
int iLangid = sqlite3_value_int(apVal[2 + p->nColumn + 2]);
if( bInsertDone==0 ){
rc = fts3InsertData(p, apVal, pRowid);
if( rc==SQLITE_CONSTRAINT && p->zContentTbl==0 ){
rc = FTS_CORRUPT_VTAB;
}
}
if( rc==SQLITE_OK && (!isRemove || *pRowid!=p->iPrevDocid ) ){
rc = fts3PendingTermsDocid(p, iLangid, *pRowid);
}
if( rc==SQLITE_OK ){
assert( p->iPrevDocid==*pRowid );
rc = fts3InsertTerms(p, iLangid, apVal, aSzIns);
}
if( p->bHasDocsize ){
fts3InsertDocsize(&rc, p, aSzIns);
}
nChng++;
}
if( p->bFts4 ){
fts3UpdateDocTotals(&rc, p, aSzIns, aSzDel, nChng);
}
update_out:
sqlite3_free(aSzDel);
sqlite3Fts3SegmentsClose(p);
return rc;
}
/*
** Flush any data in the pending-terms hash table to disk. If successful,
** merge all segments in the database (including the new segment, if
** there was any data to flush) into a single segment.
*/
SQLITE_PRIVATE int sqlite3Fts3Optimize(Fts3Table *p){
int rc;
rc = sqlite3_exec(p->db, "SAVEPOINT fts3", 0, 0, 0);
if( rc==SQLITE_OK ){
rc = fts3DoOptimize(p, 1);
if( rc==SQLITE_OK || rc==SQLITE_DONE ){
int rc2 = sqlite3_exec(p->db, "RELEASE fts3", 0, 0, 0);
if( rc2!=SQLITE_OK ) rc = rc2;
}else{
sqlite3_exec(p->db, "ROLLBACK TO fts3", 0, 0, 0);
sqlite3_exec(p->db, "RELEASE fts3", 0, 0, 0);
}
}
sqlite3Fts3SegmentsClose(p);
return rc;
}
#endif
/************** End of fts3_write.c ******************************************/
/************** Begin file fts3_snippet.c ************************************/
/*
** 2009 Oct 23
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
******************************************************************************
*/
#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
/* #include <string.h> */
/* #include <assert.h> */
/*
** Append a string to the string-buffer passed as the first argument.
**
** If nAppend is negative, then the length of the string zAppend is
** determined using strlen().
*/
static int fts3StringAppend(
StrBuffer *pStr, /* Buffer to append to */
const char *zAppend, /* Pointer to data to append to buffer */
int nAppend /* Size of zAppend in bytes (or -1) */
){
if( nAppend<0 ){
nAppend = (int)strlen(zAppend);
}
/* If there is insufficient space allocated at StrBuffer.z, use realloc()
** to grow the buffer until so that it is big enough to accomadate the
** appended data.
*/
if( pStr->n+nAppend+1>=pStr->nAlloc ){
int nAlloc = pStr->nAlloc+nAppend+100;
char *zNew = sqlite3_realloc(pStr->z, nAlloc);
if( !zNew ){
return SQLITE_NOMEM;
}
pStr->z = zNew;
pStr->nAlloc = nAlloc;
}
assert( pStr->z!=0 && (pStr->nAlloc >= pStr->n+nAppend+1) );
/* Append the data to the string buffer. */
memcpy(&pStr->z[pStr->n], zAppend, nAppend);
pStr->n += nAppend;
pStr->z[pStr->n] = '\0';
return SQLITE_OK;
}
/*
** The fts3BestSnippet() function often selects snippets that end with a
** query term. That is, the final term of the snippet is always a term
** that requires highlighting. For example, if 'X' is a highlighted term
** and '.' is a non-highlighted term, BestSnippet() may select:
**
** ........X.....X
**
** This function "shifts" the beginning of the snippet forward in the
** document so that there are approximately the same number of
** non-highlighted terms to the right of the final highlighted term as there
** are to the left of the first highlighted term. For example, to this:
**
** ....X.....X....
**
** This is done as part of extracting the snippet text, not when selecting
** the snippet. Snippet selection is done based on doclists only, so there
** is no way for fts3BestSnippet() to know whether or not the document
** actually contains terms that follow the final highlighted term.
*/
static int fts3SnippetShift(
Fts3Table *pTab, /* FTS3 table snippet comes from */
int iLangid, /* Language id to use in tokenizing */
int nSnippet, /* Number of tokens desired for snippet */
const char *zDoc, /* Document text to extract snippet from */
int nDoc, /* Size of buffer zDoc in bytes */
int *piPos, /* IN/OUT: First token of snippet */
u64 *pHlmask /* IN/OUT: Mask of tokens to highlight */
){
u64 hlmask = *pHlmask; /* Local copy of initial highlight-mask */
if( hlmask ){
int nLeft; /* Tokens to the left of first highlight */
int nRight; /* Tokens to the right of last highlight */
int nDesired; /* Ideal number of tokens to shift forward */
for(nLeft=0; !(hlmask & ((u64)1 << nLeft)); nLeft++);
for(nRight=0; !(hlmask & ((u64)1 << (nSnippet-1-nRight))); nRight++);
nDesired = (nLeft-nRight)/2;
/* Ideally, the start of the snippet should be pushed forward in the
** document nDesired tokens. This block checks if there are actually
** nDesired tokens to the right of the snippet. If so, *piPos and
** *pHlMask are updated to shift the snippet nDesired tokens to the
** right. Otherwise, the snippet is shifted by the number of tokens
** available.
*/
if( nDesired>0 ){
int nShift; /* Number of tokens to shift snippet by */
int iCurrent = 0; /* Token counter */
int rc; /* Return Code */
sqlite3_tokenizer_module *pMod;
sqlite3_tokenizer_cursor *pC;
pMod = (sqlite3_tokenizer_module *)pTab->pTokenizer->pModule;
/* Open a cursor on zDoc/nDoc. Check if there are (nSnippet+nDesired)
** or more tokens in zDoc/nDoc.
*/
rc = sqlite3Fts3OpenTokenizer(pTab->pTokenizer, iLangid, zDoc, nDoc, &pC);
if( rc!=SQLITE_OK ){
return rc;
}
while( rc==SQLITE_OK && iCurrent<(nSnippet+nDesired) ){
const char *ZDUMMY; int DUMMY1 = 0, DUMMY2 = 0, DUMMY3 = 0;
rc = pMod->xNext(pC, &ZDUMMY, &DUMMY1, &DUMMY2, &DUMMY3, &iCurrent);
}
pMod->xClose(pC);
if( rc!=SQLITE_OK && rc!=SQLITE_DONE ){ return rc; }
nShift = (rc==SQLITE_DONE)+iCurrent-nSnippet;
assert( nShift<=nDesired );
if( nShift>0 ){
*piPos += nShift;
*pHlmask = hlmask >> nShift;
}
}
}
return SQLITE_OK;
}
/*
** Extract the snippet text for fragment pFragment from cursor pCsr and
** append it to string buffer pOut.
*/
static int fts3SnippetText(
Fts3Cursor *pCsr, /* FTS3 Cursor */
SnippetFragment *pFragment, /* Snippet to extract */
int iFragment, /* Fragment number */
int isLast, /* True for final fragment in snippet */
int nSnippet, /* Number of tokens in extracted snippet */
const char *zOpen, /* String inserted before highlighted term */
const char *zClose, /* String inserted after highlighted term */
const char *zEllipsis, /* String inserted between snippets */
StrBuffer *pOut /* Write output here */
){
Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
int rc; /* Return code */
const char *zDoc; /* Document text to extract snippet from */
int nDoc; /* Size of zDoc in bytes */
int iCurrent = 0; /* Current token number of document */
int iEnd = 0; /* Byte offset of end of current token */
int isShiftDone = 0; /* True after snippet is shifted */
int iPos = pFragment->iPos; /* First token of snippet */
u64 hlmask = pFragment->hlmask; /* Highlight-mask for snippet */
int iCol = pFragment->iCol+1; /* Query column to extract text from */
sqlite3_tokenizer_module *pMod; /* Tokenizer module methods object */
sqlite3_tokenizer_cursor *pC; /* Tokenizer cursor open on zDoc/nDoc */
zDoc = (const char *)sqlite3_column_text(pCsr->pStmt, iCol);
if( zDoc==0 ){
if( sqlite3_column_type(pCsr->pStmt, iCol)!=SQLITE_NULL ){
return SQLITE_NOMEM;
}
return SQLITE_OK;
}
nDoc = sqlite3_column_bytes(pCsr->pStmt, iCol);
/* Open a token cursor on the document. */
pMod = (sqlite3_tokenizer_module *)pTab->pTokenizer->pModule;
rc = sqlite3Fts3OpenTokenizer(pTab->pTokenizer, pCsr->iLangid, zDoc,nDoc,&pC);
if( rc!=SQLITE_OK ){
return rc;
}
while( rc==SQLITE_OK ){
const char *ZDUMMY; /* Dummy argument used with tokenizer */
int DUMMY1 = -1; /* Dummy argument used with tokenizer */
int iBegin = 0; /* Offset in zDoc of start of token */
int iFin = 0; /* Offset in zDoc of end of token */
int isHighlight = 0; /* True for highlighted terms */
/* Variable DUMMY1 is initialized to a negative value above. Elsewhere
** in the FTS code the variable that the third argument to xNext points to
** is initialized to zero before the first (*but not necessarily
** subsequent*) call to xNext(). This is done for a particular application
** that needs to know whether or not the tokenizer is being used for
** snippet generation or for some other purpose.
**
** Extreme care is required when writing code to depend on this
** initialization. It is not a documented part of the tokenizer interface.
** If a tokenizer is used directly by any code outside of FTS, this
** convention might not be respected. */
rc = pMod->xNext(pC, &ZDUMMY, &DUMMY1, &iBegin, &iFin, &iCurrent);
if( rc!=SQLITE_OK ){
if( rc==SQLITE_DONE ){
/* Special case - the last token of the snippet is also the last token
** of the column. Append any punctuation that occurred between the end
** of the previous token and the end of the document to the output.
** Then break out of the loop. */
rc = fts3StringAppend(pOut, &zDoc[iEnd], -1);
}
break;
}
if( iCurrent<iPos ){ continue; }
if( !isShiftDone ){
int n = nDoc - iBegin;
rc = fts3SnippetShift(
pTab, pCsr->iLangid, nSnippet, &zDoc[iBegin], n, &iPos, &hlmask
);
isShiftDone = 1;
/* Now that the shift has been done, check if the initial "..." are
** required. They are required if (a) this is not the first fragment,
** or (b) this fragment does not begin at position 0 of its column.
*/
if( rc==SQLITE_OK && (iPos>0 || iFragment>0) ){
rc = fts3StringAppend(pOut, zEllipsis, -1);
}
if( rc!=SQLITE_OK || iCurrent<iPos ) continue;
}
if( iCurrent>=(iPos+nSnippet) ){
if( isLast ){
rc = fts3StringAppend(pOut, zEllipsis, -1);
}
break;
}
/* Set isHighlight to true if this term should be highlighted. */
isHighlight = (hlmask & ((u64)1 << (iCurrent-iPos)))!=0;
if( iCurrent>iPos ) rc = fts3StringAppend(pOut, &zDoc[iEnd], iBegin-iEnd);
if( rc==SQLITE_OK && isHighlight ) rc = fts3StringAppend(pOut, zOpen, -1);
if( rc==SQLITE_OK ) rc = fts3StringAppend(pOut, &zDoc[iBegin], iFin-iBegin);
if( rc==SQLITE_OK && isHighlight ) rc = fts3StringAppend(pOut, zClose, -1);
iEnd = iFin;
}
pMod->xClose(pC);
return rc;
}
/*
** This function is used to count the entries in a column-list (a
** delta-encoded list of term offsets within a single column of a single
** row). When this function is called, *ppCollist should point to the
** beginning of the first varint in the column-list (the varint that
** contains the position of the first matching term in the column data).
** Before returning, *ppCollist is set to point to the first byte after
** the last varint in the column-list (either the 0x00 signifying the end
** of the position-list, or the 0x01 that precedes the column number of
** the next column in the position-list).
**
** The number of elements in the column-list is returned.
*/
static int fts3ColumnlistCount(char **ppCollist){
char *pEnd = *ppCollist;
char c = 0;
int nEntry = 0;
/* A column-list is terminated by either a 0x01 or 0x00. */
while( 0xFE & (*pEnd | c) ){
c = *pEnd++ & 0x80;
if( !c ) nEntry++;
}
if( !pCsr->pExpr ){
sqlite3_result_text(pCtx, "", 0, SQLITE_STATIC);
return;
}
memset(&sCtx, 0, sizeof(sCtx));
assert( pCsr->isRequireSeek==0 );
/* Count the number of terms in the query */
rc = fts3ExprLoadDoclists(pCsr, 0, &nToken);
if( rc!=SQLITE_OK ) goto offsets_out;
/* Allocate the array of TermOffset iterators. */
sCtx.aTerm = (TermOffset *)sqlite3_malloc(sizeof(TermOffset)*nToken);
if( 0==sCtx.aTerm ){
rc = SQLITE_NOMEM;
goto offsets_out;
}
sCtx.iDocid = pCsr->iPrevId;
sCtx.pCsr = pCsr;
/* Loop through the table columns, appending offset information to
** string-buffer res for each column.
*/
for(iCol=0; iCol<pTab->nColumn; iCol++){
sqlite3_tokenizer_cursor *pC; /* Tokenizer cursor */
const char *ZDUMMY; /* Dummy argument used with xNext() */
int NDUMMY = 0; /* Dummy argument used with xNext() */
int iStart = 0;
int iEnd = 0;
int iCurrent = 0;
const char *zDoc;
int nDoc;
/* Initialize the contents of sCtx.aTerm[] for column iCol. There is
** no way that this operation can fail, so the return code from
** fts3ExprIterate() can be discarded.
*/
sCtx.iCol = iCol;
sCtx.iTerm = 0;
(void)fts3ExprIterate(pCsr->pExpr, fts3ExprTermOffsetInit, (void *)&sCtx);
/* Retreive the text stored in column iCol. If an SQL NULL is stored
** in column iCol, jump immediately to the next iteration of the loop.
** If an OOM occurs while retrieving the data (this can happen if SQLite
** needs to transform the data from utf-16 to utf-8), return SQLITE_NOMEM
** to the caller.
*/
zDoc = (const char *)sqlite3_column_text(pCsr->pStmt, iCol+1);
nDoc = sqlite3_column_bytes(pCsr->pStmt, iCol+1);
if( zDoc==0 ){
if( sqlite3_column_type(pCsr->pStmt, iCol+1)==SQLITE_NULL ){
continue;
}
rc = SQLITE_NOMEM;
goto offsets_out;
}
/* Initialize a tokenizer iterator to iterate through column iCol. */
rc = sqlite3Fts3OpenTokenizer(pTab->pTokenizer, pCsr->iLangid,
zDoc, nDoc, &pC
);
if( rc!=SQLITE_OK ) goto offsets_out;
rc = pMod->xNext(pC, &ZDUMMY, &NDUMMY, &iStart, &iEnd, &iCurrent);
while( rc==SQLITE_OK ){
int i; /* Used to loop through terms */
int iMinPos = 0x7FFFFFFF; /* Position of next token */
TermOffset *pTerm = 0; /* TermOffset associated with next token */
for(i=0; i<nToken; i++){
TermOffset *pT = &sCtx.aTerm[i];
if( pT->pList && (pT->iPos-pT->iOff)<iMinPos ){
iMinPos = pT->iPos-pT->iOff;
pTerm = pT;
}
}
if( !pTerm ){
/* All offsets for this column have been gathered. */
rc = SQLITE_DONE;
}else{
assert( iCurrent<=iMinPos );
if( 0==(0xFE&*pTerm->pList) ){
pTerm->pList = 0;
}else{
fts3GetDeltaPosition(&pTerm->pList, &pTerm->iPos);
}
while( rc==SQLITE_OK && iCurrent<iMinPos ){
rc = pMod->xNext(pC, &ZDUMMY, &NDUMMY, &iStart, &iEnd, &iCurrent);
}
if( rc==SQLITE_OK ){
char aBuffer[64];
sqlite3_snprintf(sizeof(aBuffer), aBuffer,
"%d %d %d %d ", iCol, pTerm-sCtx.aTerm, iStart, iEnd-iStart
);
rc = fts3StringAppend(&res, aBuffer, -1);
}else if( rc==SQLITE_DONE && pTab->zContentTbl==0 ){
rc = FTS_CORRUPT_VTAB;
}
}
}
if( rc==SQLITE_DONE ){
rc = SQLITE_OK;
}
pMod->xClose(pC);
if( rc!=SQLITE_OK ) goto offsets_out;
}
offsets_out:
sqlite3_free(sCtx.aTerm);
assert( rc!=SQLITE_DONE );
sqlite3Fts3SegmentsClose(pTab);
if( rc!=SQLITE_OK ){
sqlite3_result_error_code(pCtx, rc);
sqlite3_free(res.z);
}else{
sqlite3_result_text(pCtx, res.z, res.n-1, sqlite3_free);
}
( run in 0.703 second using v1.01-cache-2.11-cpan-97f6503c9c8 )