SQLite

** in the decision may argue for a larger value.
*/
#define CHUNK_MAX 256

typedef enum fulltext_statement {
  CONTENT_INSERT_STMT,
  CONTENT_SELECT_STMT,
  CONTENT_UPDATE_STMT,
  CONTENT_DELETE_STMT,

  TERM_SELECT_STMT,
  TERM_SELECT_ALL_STMT,
  TERM_INSERT_STMT,
  TERM_UPDATE_STMT,
  TERM_DELETE_STMT,

  MAX_STMT                     /* Always at end! */
} fulltext_statement;

/* These must exactly match the enum above. */
/* TODO(adam): Is there some risk that a statement (in particular,
** pTermSelectStmt) will be used in two cursors at once, e.g.  if a
** query joins a virtual table to itself?  If so perhaps we should
** move some of these to the cursor object.
*/
static const char *const fulltext_zStatement[MAX_STMT] = {
  /* CONTENT_INSERT */ NULL,  /* generated in contentInsertStatement() */
  /* CONTENT_SELECT */ "select * from %_content where rowid = ?",
  /* CONTENT_UPDATE */ NULL,  /* generated in contentUpdateStatement() */
  /* CONTENT_DELETE */ "delete from %_content where rowid = ?",

  /* TERM_SELECT */
  "select rowid, doclist from %_term where term = ? and segment = ?",
  /* TERM_SELECT_ALL */
  "select doclist from %_term where term = ? order by segment",
  /* TERM_INSERT */

*/
struct fulltext_vtab {
  sqlite3_vtab base;               /* Base class used by SQLite core */
  sqlite3 *db;                     /* The database connection */
  const char *zName;               /* virtual table name */
  int nColumn;                     /* number of columns in virtual table */
  char **azColumn;                 /* column names.  malloced */
  char **azContentColumn;          /* column names in content table; malloced */
  sqlite3_tokenizer *pTokenizer;   /* tokenizer for inserts and queries */

  /* Precompiled statements which we keep as long as the table is
  ** open.
  */
  sqlite3_stmt *pFulltextStatements[MAX_STMT];
};

} fulltext_cursor;

static struct fulltext_vtab *cursor_vtab(fulltext_cursor *c){
  return (fulltext_vtab *) c->base.pVtab;
}

static const sqlite3_module fulltextModule;   /* forward declaration */

/* Append a list of strings separated by commas to a StringBuffer. */
static void appendList(StringBuffer *sb, int nString, char **azString){
  int i;
  for(i=0; i<nString; ++i){
    if( i>0 ) append(sb, ", ");
    append(sb, azString[i]);
  }
}

/* Return a dynamically generated statement of the form
 *   insert into %_content (rowid, ...) values (?, ...)
 */
static const char *contentInsertStatement(fulltext_vtab *v){
  StringBuffer sb;
  int i;

  initStringBuffer(&sb);
  append(&sb, "insert into %_content (rowid, ");
  appendList(&sb, v->nColumn, v->azContentColumn);
  append(&sb, ") values (?");
  for(i=0; i<v->nColumn; ++i)
    append(&sb, ", ?");
  append(&sb, ")");
  return sb.s;
}

/* Return a dynamically generated statement of the form
 *   update %_content set [col_0] = ?, [col_1] = ?, ...
 *                    where rowid = ?
 */
static const char *contentUpdateStatement(fulltext_vtab *v){
  StringBuffer sb;
  int i;

  initStringBuffer(&sb);
  append(&sb, "update %_content set ");
  for(i=0; i<v->nColumn; ++i) {
    if( i>0 ){
      append(&sb, ", ");
    }
    append(&sb, v->azContentColumn[i]);
    append(&sb, " = ?");
  }
  append(&sb, " where rowid = ?");
  return sb.s;
}

/* Puts a freshly-prepared statement determined by iStmt in *ppStmt.
** If the indicated statement has never been prepared, it is prepared
** and cached, otherwise the cached version is reset.
*/
static int sql_get_statement(fulltext_vtab *v, fulltext_statement iStmt,
                             sqlite3_stmt **ppStmt){
  assert( iStmt<MAX_STMT );
  if( v->pFulltextStatements[iStmt]==NULL ){
    const char *zStmt;
    int rc;
    switch( iStmt ){
      case CONTENT_INSERT_STMT:
        zStmt = contentInsertStatement(v); break;
      case CONTENT_UPDATE_STMT:
        zStmt = contentUpdateStatement(v); break;
      default:
        zStmt = fulltext_zStatement[iStmt];
    }
    rc = sql_prepare(v->db, v->zName, &v->pFulltextStatements[iStmt],
                         zStmt);
    if( zStmt != fulltext_zStatement[iStmt]) free((void *) zStmt);
    if( rc!=SQLITE_OK ) return rc;
  } else {
    int rc = sqlite3_reset(v->pFulltextStatements[iStmt]);
    if( rc!=SQLITE_OK ) return rc;
  }

  *ppStmt = v->pFulltextStatements[iStmt];

  for(i=0; i<v->nColumn; ++i){
    rc = sqlite3_bind_value(s, 2+i, pValues[i]);
    if( rc!=SQLITE_OK ) return rc;
  }

  return sql_single_step_statement(v, CONTENT_INSERT_STMT, &s);
}

/* update %_content set col0 = pValues[0], col1 = pValues[1], ...
 *                  where rowid = [iRowid] */
static int content_update(fulltext_vtab *v, sqlite3_value **pValues,
                          sqlite_int64 iRowid){
  sqlite3_stmt *s;
  int i;
  int rc = sql_get_statement(v, CONTENT_UPDATE_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  for(i=0; i<v->nColumn; ++i){
    rc = sqlite3_bind_value(s, 1+i, pValues[i]);
    if( rc!=SQLITE_OK ) return rc;
  }

  rc = sqlite3_bind_int64(s, 1+v->nColumn, iRowid);
  if( rc!=SQLITE_OK ) return rc;

  return sql_single_step_statement(v, CONTENT_UPDATE_STMT, &s);
}

void freeStringArray(int nString, const char **pString){
  int i;

  for (i=0 ; i < nString ; ++i) {
    free((void *) pString[i]);
  }

  return sql_single_step_statement(v, TERM_DELETE_STMT, &s);
}

/*
** Free the memory used to contain a fulltext_vtab structure.
*/
static void fulltext_vtab_destroy(fulltext_vtab *v){
  int iStmt, i;

  TRACE(("FTS1 Destroy %p\n", v));
  for( iStmt=0; iStmt<MAX_STMT; iStmt++ ){
    if( v->pFulltextStatements[iStmt]!=NULL ){
      sqlite3_finalize(v->pFulltextStatements[iStmt]);
      v->pFulltextStatements[iStmt] = NULL;
    }
  }

  if( v->pTokenizer!=NULL ){
    v->pTokenizer->pModule->xDestroy(v->pTokenizer);
    v->pTokenizer = NULL;
  }
  
  free(v->azColumn);
  for(i = 0; i < v->nColumn; ++i) {
    sqlite3_free(v->azContentColumn[i]);
  }
  free(v->azContentColumn);
  free(v);
}

/*
** Token types for parsing the arguments to xConnect or xCreate.
*/
#define TOKEN_EOF         0    /* End of file */

  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))

  /* 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);

  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. */

  ** 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;

static int fulltextRowid(sqlite3_vtab_cursor *pCursor, sqlite_int64 *pRowid){
  fulltext_cursor *c = (fulltext_cursor *) pCursor;

  *pRowid = sqlite3_column_int64(c->pStmt, 0);
  return SQLITE_OK;
}

/* Add all terms in [zText] to the given hash table.  If [iColumn] > 0,
 * we also store positions and offsets in the hash table using the given
 * column number. */
static int buildTerms(fulltext_vtab *v, fts1Hash *terms, sqlite_int64 iDocid,
                      const char *zText, int iColumn){
  sqlite3_tokenizer *pTokenizer = v->pTokenizer;
  sqlite3_tokenizer_cursor *pCursor;
  const char *pToken;
  int nTokenBytes;
  int iStartOffset, iEndOffset, iPosition;
  int rc;

  rc = pTokenizer->pModule->xOpen(pTokenizer, zText, -1, &pCursor);
  if( rc!=SQLITE_OK ) return rc;

  pCursor->pTokenizer = pTokenizer;
  while( SQLITE_OK==pTokenizer->pModule->xNext(pCursor,
                                               &pToken, &nTokenBytes,
                                               &iStartOffset, &iEndOffset,
                                               &iPosition) ){

    p = fts1HashFind(terms, pToken, nTokenBytes);
    if( p==NULL ){
      p = docListNew(DL_POSITIONS_OFFSETS);
      docListAddDocid(p, iDocid);
      fts1HashInsert(terms, pToken, nTokenBytes, p);
    }
    if( iColumn>=0 ){
      docListAddPosOffset(p, iColumn, iPosition, iStartOffset, iEndOffset);
    }
  }

  /* TODO(shess) Check return?  Should this be able to cause errors at
  ** this point?  Actually, same question about sqlite3_finalize(),
  ** though one could argue that failure there means that the data is
  ** not durable.  *ponder*
  */

  }

 err:
  docListDestroy(&doclist);
  return rc;
}

/* Add doclists for all terms in [pValues] to the hash table [terms]. */
static int insertTerms(fulltext_vtab *v, fts1Hash *terms, sqlite_int64 iRowid,
                sqlite3_value **pValues){
  int i;
  for(i = 0; i < v->nColumn ; ++i){
    char *zText = (char*)sqlite3_value_text(pValues[i]);
    int rc = buildTerms(v, terms, iRowid, zText, i);
    if( rc!=SQLITE_OK ) return rc;
  }
  return SQLITE_OK;
}

/* Add empty doclists for all terms in the given row's content to the hash
 * table [pTerms]. */
static int deleteTerms(fulltext_vtab *v, fts1Hash *pTerms, sqlite_int64 iRowid){
  const char **pValues;
  int i;

  int rc = content_select(v, iRowid, &pValues);
  if( rc!=SQLITE_OK ) return rc;

  for(i = 0 ; i < v->nColumn; ++i) {
    rc = buildTerms(v, pTerms, iRowid, pValues[i], -1);
    if( rc!=SQLITE_OK ) break;
  }

  freeStringArray(v->nColumn, pValues);
  return SQLITE_OK;
}

/* Insert a row into the %_content table; set *piRowid to be the ID of the
 * new row.  Fill [pTerms] with new doclists for the %_term table. */
static int index_insert(fulltext_vtab *v, sqlite3_value *pRequestRowid,
                        sqlite3_value **pValues,
                        sqlite_int64 *piRowid, fts1Hash *pTerms){
  int rc;

  rc = content_insert(v, pRequestRowid, pValues);  /* execute an SQL INSERT */
  if( rc!=SQLITE_OK ) return rc;
  *piRowid = sqlite3_last_insert_rowid(v->db);
  return insertTerms(v, pTerms, *piRowid, pValues);
}

/* Delete a row from the %_content table; fill [pTerms] with empty doclists
 * to be written to the %_term table. */
static int index_delete(fulltext_vtab *v, sqlite_int64 iRow, fts1Hash *pTerms){
  int rc = deleteTerms(v, pTerms, iRow);
  if( rc!=SQLITE_OK ) return rc;
  return content_delete(v, iRow);  /* execute an SQL DELETE */
}

/* Update a row in the %_content table; fill [pTerms] with new doclists for the
 * %_term table. */
static int index_update(fulltext_vtab *v, sqlite_int64 iRow,
                        sqlite3_value **pValues, fts1Hash *pTerms){
  /* Generate an empty doclist for each term that previously appeared in this
   * row. */
  int rc = deleteTerms(v, pTerms, iRow);
  if( rc!=SQLITE_OK ) return rc;

  /* Now add positions for terms which appear in the updated row. */
  rc = insertTerms(v, pTerms, iRow, pValues);
  if( rc!=SQLITE_OK ) return rc;

  return content_update(v, pValues, iRow);  /* execute an SQL UPDATE */
}

/* This function implements the xUpdate callback; it's the top-level entry
 * point for inserting, deleting or updating a row in a full-text table. */
static int fulltextUpdate(sqlite3_vtab *pVtab, int nArg, sqlite3_value **ppArg,
                   sqlite_int64 *pRowid){
  fulltext_vtab *v = (fulltext_vtab *) pVtab;
  fts1Hash terms;   /* maps term string -> PosList */
  int rc;

  fts1HashElem *e;


  TRACE(("FTS1 Update %p\n", pVtab));


  
  fts1HashInit(&terms, FTS1_HASH_STRING, 1);

  if( nArg<2 ){
    rc = index_delete(v, sqlite3_value_int64(ppArg[0]), &terms);
  } else if( sqlite3_value_type(ppArg[0]) != SQLITE_NULL ){
    /* An update:
     * ppArg[0] = old rowid
     * ppArg[1] = new rowid
     * ppArg[2..2+v->nColumn-1] = values
     * ppArg[2+v->nColumn] = value for magic column (we ignore this)
     */
    sqlite_int64 rowid = sqlite3_value_int64(ppArg[0]);
    if( sqlite3_value_type(ppArg[1]) != SQLITE_INTEGER ||
      sqlite3_value_int64(ppArg[1]) != rowid ){
      rc = SQLITE_ERROR;  /* we don't allow changing the rowid */
    } else {
      assert( nArg==2+v->nColumn+1);
      rc = index_update(v, rowid, &ppArg[2], &terms);
    }
  } else {
    /* An insert:
     * ppArg[1] = requested rowid
     * ppArg[2..2+v->nColumn-1] = values
     * ppArg[2+v->nColumn] = value for magic column (we ignore this)
     */
    assert( nArg==2+v->nColumn+1);
    rc = index_insert(v, ppArg[1], &ppArg[2], pRowid, &terms);
  }

  if( rc==SQLITE_OK ){
    /* Write updated doclists to disk. */
    for(e=fts1HashFirst(&terms); e; e=fts1HashNext(e)){
      DocList *p = fts1HashData(e);
      rc = index_insert_term(v, fts1HashKey(e), fts1HashKeysize(e), p);
      if( rc!=SQLITE_OK ) break;
    }
  }








  /* clean up */































  for(e=fts1HashFirst(&terms); e; e=fts1HashNext(e)){
    DocList *p = fts1HashData(e);
    docListDelete(p);
  }
  fts1HashClear(&terms);


  return rc;
























}

/*
** Implementation of the snippet() function for FTS1
*/
static void snippetFunc(
  sqlite3_context *pContext,