fts1.c

sqlite 3.3.8 支持加密的版本

**     output:     [ ] ...
*/
void tokenListToIdList(char **azIn){
  int i, j;
  if( azIn ){
    for(i=0, j=-1; azIn[i]; i++){
      if( isalnum(azIn[i][0]) || azIn[i][1] ){
        dequoteString(azIn[i]);
        if( j>=0 ){
          azIn[j] = azIn[i];
        }
        j++;
      }
    }
    azIn[j] = 0;
  }
}


/*
** Find the first alphanumeric token in the string zIn.  Null-terminate
** this token.  Remove any quotation marks.  And return a pointer to
** the result.
*/
static char *firstToken(char *zIn, char **pzTail){
  int i, n, ttype;
  i = 0;
  while(1){
    n = getToken(zIn, &ttype);
    if( ttype==TOKEN_SPACE ){
      zIn += n;
    }else if( ttype==TOKEN_EOF ){
      *pzTail = zIn;
      return 0;
    }else{
      zIn[n] = 0;
      *pzTail = &zIn[1];
      dequoteString(zIn);
      return zIn;
    }
  }
  /*NOTREACHED*/
}

/* Return true if...
**
**   *  s begins with the string t, ignoring case
**   *  s is longer than t
**   *  The first character of s beyond t is not a alphanumeric
** 
** Ignore leading space in *s.
**
** To put it another way, return true if the first token of
** s[] is t[].
*/
static int startsWith(const char *s, const char *t){
  while( isspace(*s) ){ s++; }
  while( *t ){
    if( tolower(*s++)!=tolower(*t++) ) return 0;
  }
  return *s!='_' && !isalnum(*s);
}

/*
** An instance of this structure defines the "spec" of a
** full text index.  This structure is populated by parseSpec
** and use by fulltextConnect and fulltextCreate.
*/
typedef struct TableSpec {
  const char *zName;       /* Name of the full-text index */
  int nColumn;             /* Number of columns to be indexed */
  char **azColumn;         /* Original names of columns to be indexed */
  char **azContentColumn;  /* Column names for %_content */
  char **azTokenizer;      /* Name of tokenizer and its arguments */
} TableSpec;

/*
** Reclaim all of the memory used by a TableSpec
*/
void clearTableSpec(TableSpec *p) {
  free(p->azColumn);
  free(p->azContentColumn);
  free(p->azTokenizer);
}

/* Parse a CREATE VIRTUAL TABLE statement, which looks like this:
 *
 * CREATE VIRTUAL TABLE email
 *        USING fts1(subject, body, tokenize mytokenizer(myarg))
 *
 * We return parsed information in a TableSpec structure.
 * 
 */
int parseSpec(TableSpec *pSpec, int argc, const char *const*argv, char**pzErr){
  int i, j, n;
  char *z, *zDummy;
  char **azArg;
  const char *zTokenizer = 0;    /* argv[] entry describing the tokenizer */

  assert( argc>=3 );
  /* Current interface:
  ** argv[0] - module name
  ** argv[1] - database name
  ** argv[2] - table name
  ** argv[3..] - columns, optionally followed by tokenizer specification
  **             and snippet delimiters specification.
  */

  /* Make a copy of the complete argv[][] array in a single allocation.
  ** The argv[][] array is read-only and transient.  We can write to the
  ** copy in order to modify things and the copy is persistent.
  */
  memset(pSpec, 0, sizeof(*pSpec));
  for(i=n=0; i<argc; i++){
    n += strlen(argv[i]) + 1;
  }
  azArg = malloc( sizeof(char*)*argc + n );
  if( azArg==0 ){
    return SQLITE_NOMEM;
  }
  z = (char*)&azArg[argc];
  for(i=0; i<argc; i++){
    azArg[i] = z;
    strcpy(z, argv[i]);
    z += strlen(z)+1;
  }

  /* Identify the column names and the tokenizer and delimiter arguments
  ** in the argv[][] array.
  */
  pSpec->zName = azArg[2];
  pSpec->nColumn = 0;
  pSpec->azColumn = azArg;
  zTokenizer = "tokenize simple";
  for(i=3, j=0; i<argc; ++i){
    if( startsWith(azArg[i],"tokenize") ){
      zTokenizer = azArg[i];
    }else{
      z = azArg[pSpec->nColumn] = firstToken(azArg[i], &zDummy);
      pSpec->nColumn++;
    }
  }
  if( pSpec->nColumn==0 ){
    azArg[0] = "content";
    pSpec->nColumn = 1;
  }

  /*
  ** Construct the list of content column names.
  **
  ** Each content column name will be of the form cNNAAAA
  ** where NN is the column number and AAAA is the sanitized
  ** column name.  "sanitized" means that special characters are
  ** converted to "_".  The cNN prefix guarantees that all column
  ** names are unique.
  **
  ** The AAAA suffix is not strictly necessary.  It is included
  ** for the convenience of people who might examine the generated
  ** %_content table and wonder what the columns are used for.
  */
  pSpec->azContentColumn = malloc( pSpec->nColumn * sizeof(char *) );
  if( pSpec->azContentColumn==0 ){
    clearTableSpec(pSpec);
    return SQLITE_NOMEM;
  }
  for(i=0; i<pSpec->nColumn; i++){
    char *p;
    pSpec->azContentColumn[i] = sqlite3_mprintf("c%d%s", i, azArg[i]);
    for (p = pSpec->azContentColumn[i]; *p ; ++p) {
      if( !isalnum(*p) ) *p = '_';
    }
  }

  /*
  ** Parse the tokenizer specification string.
  */
  pSpec->azTokenizer = tokenizeString(zTokenizer, &n);
  tokenListToIdList(pSpec->azTokenizer);

  return SQLITE_OK;
}

/*
** Generate a CREATE TABLE statement that describes the schema of
** the virtual table.  Return a pointer to this schema string.
**
** Space is obtained from sqlite3_mprintf() and should be freed
** using sqlite3_free().
*/
static char *fulltextSchema(
  int nColumn,                  /* Number of columns */
  const char *const* azColumn,  /* List of columns */
  const char *zTableName        /* Name of the table */
){
  int i;
  char *zSchema, *zNext;
  const char *zSep = "(";
  zSchema = sqlite3_mprintf("CREATE TABLE x");
  for(i=0; i<nColumn; i++){
    zNext = sqlite3_mprintf("%s%s%Q", zSchema, zSep, azColumn[i]);
    sqlite3_free(zSchema);
    zSchema = zNext;
    zSep = ",";
  }
  zNext = sqlite3_mprintf("%s,%Q)", zSchema, zTableName);
  sqlite3_free(zSchema);
  return zNext;
}

/*
** Build a new sqlite3_vtab structure that will describe the
** fulltext index defined by spec.
*/
static int constructVtab(
  sqlite3 *db,              /* The SQLite database connection */
  TableSpec *spec,          /* Parsed spec information from parseSpec() */
  sqlite3_vtab **ppVTab,    /* Write the resulting vtab structure here */
  char **pzErr              /* Write any error message here */
){
  int rc;
  int n;
  fulltext_vtab *v = 0;
  const sqlite3_tokenizer_module *m = NULL;
  char *schema;

  v = (fulltext_vtab *) malloc(sizeof(fulltext_vtab));
  if( v==0 ) return SQLITE_NOMEM;
  memset(v, 0, sizeof(*v));
  /* sqlite will initialize v->base */
  v->db = db;
  v->zName = spec->zName;   /* Freed when azColumn is freed */
  v->nColumn = spec->nColumn;
  v->azContentColumn = spec->azContentColumn;
  spec->azContentColumn = 0;
  v->azColumn = spec->azColumn;
  spec->azColumn = 0;

  if( spec->azTokenizer==0 ){
    return SQLITE_NOMEM;
  }
  /* TODO(shess) For now, add new tokenizers as else if clauses. */
  if( spec->azTokenizer[0]==0 || startsWith(spec->azTokenizer[0], "simple") ){
    sqlite3Fts1SimpleTokenizerModule(&m);
  }else if( startsWith(spec->azTokenizer[0], "porter") ){
    sqlite3Fts1PorterTokenizerModule(&m);
  }else{
    *pzErr = sqlite3_mprintf("unknown tokenizer: %s", spec->azTokenizer[0]);
    rc = SQLITE_ERROR;
    goto err;
  }
  for(n=0; spec->azTokenizer[n]; n++){}
  if( n ){
    rc = m->xCreate(n-1, (const char*const*)&spec->azTokenizer[1],
                    &v->pTokenizer);
  }else{
    rc = m->xCreate(0, 0, &v->pTokenizer);
  }
  if( rc!=SQLITE_OK ) goto err;
  v->pTokenizer->pModule = m;

  /* TODO: verify the existence of backing tables foo_content, foo_term */

  schema = fulltextSchema(v->nColumn, (const char*const*)v->azColumn,
                          spec->zName);
  rc = sqlite3_declare_vtab(db, schema);
  sqlite3_free(schema);
  if( rc!=SQLITE_OK ) goto err;

  memset(v->pFulltextStatements, 0, sizeof(v->pFulltextStatements));

  *ppVTab = &v->base;
  TRACE(("FTS1 Connect %p\n", v));

  return rc;

err:
  fulltext_vtab_destroy(v);
  return rc;
}

static int fulltextConnect(
  sqlite3 *db,
  void *pAux,
  int argc, const char *const*argv,
  sqlite3_vtab **ppVTab,
  char **pzErr
){
  TableSpec spec;
  int rc = parseSpec(&spec, argc, argv, pzErr);
  if( rc!=SQLITE_OK ) return rc;

  rc = constructVtab(db, &spec, ppVTab, pzErr);
  clearTableSpec(&spec);
  return rc;
}

  /* The %_content table holds the text of each document, with
  ** the rowid used as the docid.
  **
  ** The %_term table maps each term to a document list blob
  ** containing elements sorted by ascending docid, each element
  ** encoded as:
  **
  **   docid varint-encoded
  **   token elements:
  **     position+1 varint-encoded as delta from previous position
  **     start offset varint-encoded as delta from previous start offset
  **     end offset varint-encoded as delta from start offset
  **
  ** The sentinel position of 0 indicates the end of the token list.
  **
  ** Additionally, doclist blobs are chunked into multiple segments,
  ** using segment to order the segments.  New elements are added to
  ** the segment at segment 0, until it exceeds CHUNK_MAX.  Then
  ** segment 0 is deleted, and the doclist is inserted at segment 1.
  ** If there is already a doclist at segment 1, the segment 0 doclist
  ** is merged with it, the segment 1 doclist is deleted, and the
  ** merged doclist is inserted at segment 2, repeating those
  ** operations until an insert succeeds.
  **
  ** Since this structure doesn't allow us to update elements in place
  ** in case of deletion or update, these are simply written to
  ** segment 0 (with an empty token list in case of deletion), with
  ** docListAccumulate() taking care to retain lower-segment
  ** information in preference to higher-segment information.
  */
  /* TODO(shess) Provide a VACUUM type operation which both removes
  ** deleted elements which are no longer necessary, and duplicated
  ** elements.  I suspect this will probably not be necessary in
  ** practice, though.
  */
static int fulltextCreate(sqlite3 *db, void *pAux,
                          int argc, const char * const *argv,
                          sqlite3_vtab **ppVTab, char **pzErr){
  int rc;
  TableSpec spec;
  StringBuffer schema;
  TRACE(("FTS1 Create\n"));

  rc = parseSpec(&spec, argc, argv, pzErr);
  if( rc!=SQLITE_OK ) return rc;

  initStringBuffer(&schema);
  append(&schema, "CREATE TABLE %_content(");
  appendList(&schema, spec.nColumn, spec.azContentColumn);
  append(&schema, ")");
  rc = sql_exec(db, spec.zName, schema.s);
  free(schema.s);
  if( rc!=SQLITE_OK ) goto out;

  rc = sql_exec(db, spec.zName,
    "create table %_term(term text, segment integer, doclist blob, "
                        "primary key(term, segment));");
  if( rc!=SQLITE_OK ) goto out;

  rc = constructVtab(db, &spec, ppVTab, pzErr);

out:
  clearTableSpec(&spec);
  return rc;
}

/* Decide how to handle an SQL query. */
static int fulltextBestIndex(sqlite3_vtab *pVTab, sqlite3_index_info *pInfo){
  int i;

  for(i=0; i<pInfo->nConstraint; ++i){
    const struct sqlite3_index_constraint *pConstraint;
    pConstraint = &pInfo->aConstraint[i];
    if( pConstraint->usable ) {
      if( pConstraint->iColumn==-1 &&
          pConstraint->op==SQLITE_INDEX_CONSTRAINT_EQ ){
        pInfo->idxNum = QUERY_ROWID;      /* lookup by rowid */
      } else if( pConstraint->iColumn>=0 &&
                 pConstraint->op==SQLITE_INDEX_CONSTRAINT_MATCH ){
        /* full-text search */
        pInfo->idxNum = QUERY_FULLTEXT + pConstraint->iColumn;
      } else continue;

      pInfo->aConstraintUsage[i].argvIndex = 1;
      pInfo->aConstraintUsage[i].omit = 1;

      /* An arbitrary value for now.
       * TODO: Perhaps rowid matches should be considered cheaper than
       * full-text searches. */
      pInfo->estimatedCost = 1.0;   

      return SQLITE_OK;
    }
  }
  pInfo->idxNum = QUERY_GENERIC;
  TRACE(("FTS1 BestIndex\n"));
  return SQLITE_OK;
}

static int fulltextDisconnect(sqlite3_vtab *pVTab){
  TRACE(("FTS1 Disconnect %p\n", pVTab));
  fulltext_vtab_destroy((fulltext_vtab *)pVTab);
  return SQLITE_OK;
}

static int fulltextDestroy(sqlite3_vtab *pVTab){
  fulltext_vtab *v = (fulltext_vtab *)pVTab;
  int rc;

  TRACE(("FTS1 Destroy %p\n", pVTab));
  rc = sql_exec(v->db, v->zName,
                    "drop table %_content; drop table %_term");
  if( rc!=SQLITE_OK ) return rc;

  fulltext_vtab_destroy((fulltext_vtab *)pVTab);
  return SQLITE_OK;
}

static int fulltextOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){
  fulltext_cursor *c;

  c = (fulltext_cursor *) calloc(sizeof(fulltext_cursor