fts1.c

sqlite 3.3.8 支持加密的版本

  return str;
}

/* Duplicate a string; the caller must free() the returned string.
 * (We don't use strdup() since it's not part of the standard C library and
 * may not be available everywhere.) */
static char *string_dup(const char *s){
  return string_dup_n(s, strlen(s));
}

/* Format a string, replacing each occurrence of the % character with
 * zName.  This may be more convenient than sqlite_mprintf()
 * when one string is used repeatedly in a format string.
 * The caller must free() the returned string. */
static char *string_format(const char *zFormat, const char *zName){
  const char *p;
  size_t len = 0;
  size_t nName = strlen(zName);
  char *result;
  char *r;

  /* first compute length needed */
  for(p = zFormat ; *p ; ++p){
    len += (*p=='%' ? nName : 1);
  }
  len += 1;  /* for null terminator */

  r = result = malloc(len);
  for(p = zFormat; *p; ++p){
    if( *p=='%' ){
      memcpy(r, zName, nName);
      r += nName;
    } else {
      *r++ = *p;
    }
  }
  *r++ = '\0';
  assert( r == result + len );
  return result;
}

static int sql_exec(sqlite3 *db, const char *zName, const char *zFormat){
  char *zCommand = string_format(zFormat, zName);
  int rc;
  TRACE(("FTS1 sql: %s\n", zCommand));
  rc = sqlite3_exec(db, zCommand, NULL, 0, NULL);
  free(zCommand);
  return rc;
}

static int sql_prepare(sqlite3 *db, const char *zName, sqlite3_stmt **ppStmt,
                const char *zFormat){
  char *zCommand = string_format(zFormat, zName);
  int rc;
  TRACE(("FTS1 prepare: %s\n", zCommand));
  rc = sqlite3_prepare(db, zCommand, -1, ppStmt, NULL);
  free(zCommand);
  return rc;
}

/* end utility functions */

/* Forward reference */
typedef struct fulltext_vtab fulltext_vtab;

/* A single term in a query is represented by an instances of
** the following structure.
*/
typedef struct QueryTerm {
  short int nPhrase; /* How many following terms are part of the same phrase */
  short int iPhrase; /* This is the i-th term of a phrase. */
  short int iColumn; /* Column of the index that must match this term */
  signed char isOr;  /* this term is preceded by "OR" */
  signed char isNot; /* this term is preceded by "-" */
  char *pTerm;       /* text of the term.  '\000' terminated.  malloced */
  int nTerm;         /* Number of bytes in pTerm[] */
} QueryTerm;


/* A query string is parsed into a Query structure.
 *
 * We could, in theory, allow query strings to be complicated
 * nested expressions with precedence determined by parentheses.
 * But none of the major search engines do this.  (Perhaps the
 * feeling is that an parenthesized expression is two complex of
 * an idea for the average user to grasp.)  Taking our lead from
 * the major search engines, we will allow queries to be a list
 * of terms (with an implied AND operator) or phrases in double-quotes,
 * with a single optional "-" before each non-phrase term to designate
 * negation and an optional OR connector.
 *
 * OR binds more tightly than the implied AND, which is what the
 * major search engines seem to do.  So, for example:
 * 
 *    [one two OR three]     ==>    one AND (two OR three)
 *    [one OR two three]     ==>    (one OR two) AND three
 *
 * A "-" before a term matches all entries that lack that term.
 * The "-" must occur immediately before the term with in intervening
 * space.  This is how the search engines do it.
 *
 * A NOT term cannot be the right-hand operand of an OR.  If this
 * occurs in the query string, the NOT is ignored:
 *
 *    [one OR -two]          ==>    one OR two
 *
 */
typedef struct Query {
  fulltext_vtab *pFts;  /* The full text index */
  int nTerms;           /* Number of terms in the query */
  QueryTerm *pTerms;    /* Array of terms.  Space obtained from malloc() */
  int nextIsOr;         /* Set the isOr flag on the next inserted term */
  int nextColumn;       /* Next word parsed must be in this column */
  int dfltColumn;       /* The default column */
} Query;


/*
** An instance of the following structure keeps track of generated
** matching-word offset information and snippets.
*/
typedef struct Snippet {
  int nMatch;     /* Total number of matches */
  int nAlloc;     /* Space allocated for aMatch[] */
  struct snippetMatch { /* One entry for each matching term */
    char snStatus;       /* Status flag for use while constructing snippets */
    short int iCol;      /* The column that contains the match */
    short int iTerm;     /* The index in Query.pTerms[] of the matching term */
    short int nByte;     /* Number of bytes in the term */
    int iStart;          /* The offset to the first character of the term */
  } *aMatch;      /* Points to space obtained from malloc */
  char *zOffset;  /* Text rendering of aMatch[] */
  int nOffset;    /* strlen(zOffset) */
  char *zSnippet; /* Snippet text */
  int nSnippet;   /* strlen(zSnippet) */
} Snippet;


typedef enum QueryType {
  QUERY_GENERIC,   /* table scan */
  QUERY_ROWID,     /* lookup by rowid */
  QUERY_FULLTEXT   /* QUERY_FULLTEXT + [i] is a full-text search for column i*/
} QueryType;

/* TODO(shess) CHUNK_MAX controls how much data we allow in segment 0
** before we start aggregating into larger segments.  Lower CHUNK_MAX
** means that for a given input we have more individual segments per
** term, which means more rows in the table and a bigger index (due to
** both more rows and bigger rowids).  But it also reduces the average
** cost of adding new elements to the segment 0 doclist, and it seems
** to reduce the number of pages read and written during inserts.  256
** was chosen by measuring insertion times for a certain input (first
** 10k documents of Enron corpus), though including query performance
** 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 */
  "insert into %_term (rowid, term, segment, doclist) values (?, ?, ?, ?)",
  /* TERM_UPDATE */ "update %_term set doclist = ? where rowid = ?",
  /* TERM_DELETE */ "delete from %_term where rowid = ?",
};

/*
** 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 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];
};

/*
** When the core wants to do a query, it create a cursor using a
** call to xOpen.  This structure is an instance of a cursor.  It
** is destroyed by xClose.
*/
typedef struct fulltext_cursor {
  sqlite3_vtab_cursor base;        /* Base class used by SQLite core */
  QueryType iCursorType;           /* Copy of sqlite3_index_info.idxNum */
  sqlite3_stmt *pStmt;             /* Prepared statement in use by the cursor */
  int eof;                         /* True if at End Of Results */
  Query q;                         /* Parsed query string */
  Snippet snippet;                 /* Cached snippet for the current row */
  int iColumn;                     /* Column being searched */
  DocListReader result;  /* used when iCursorType == QUERY_FULLTEXT */ 
} 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];
  return SQLITE_OK;
}

/* Step the indicated statement, handling errors SQLITE_BUSY (by
** retrying) and SQLITE_SCHEMA (by re-preparing and transferring
** bindings to the new statement).
** TODO(adam): We should extend this function so that it can work with
** statements declared locally, not only globally cached statements.
*/
static int sql_step_statement(fulltext_vtab *v, fulltext_statement iStmt,
                              sqlite3_stmt **ppStmt){
  int rc;
  sqlite3_stmt *s = *ppStmt;
  assert( iStmt<MAX_STMT );
  assert( s==v->pFulltextStatements[iStmt] );

  while( (rc=sqlite3_step(s))!=SQLITE_DONE && rc!=SQLITE_ROW ){
    sqlite3_stmt *pNewStmt;

    if( rc==SQLITE_BUSY ) continue;
    if( rc!=SQLITE_ERROR ) return rc;

    rc = sqlite3_reset(s);
    if( rc!=SQLITE_SCHEMA ) return SQLITE_ERROR;

    v->pFulltextStatements[iStmt] = NULL;   /* Still in s */
    rc = sql_get_statement(v, iStmt, &pNewStmt);
    if( rc!=SQLITE_OK ) goto err;
    *ppStmt = pNewStmt;

    rc = sqlite3_transfer_bindings(s, pNewStmt);
    if( rc!=SQLITE_OK ) goto err;

    rc = sqlite3_finalize(s);
    if( rc!=SQLITE_OK ) return rc;
    s = pNewStmt;
  }
  return rc;

 err:
  sqlite3_finalize(s);
  return rc;
}

/* Like sql_step_statement(), but convert SQLITE_DONE to SQLITE_OK.
** Useful for statements like UPDATE, where we expect no results.
*/
static int sql_single_step_statement(fulltext_vtab *v,
                                     fulltext_statement iStmt,
                                     sqlite3_stmt **ppStmt){
  int rc = sql_step_statement(v, iStmt, ppStmt);
  return (rc==SQLITE_DONE) ? SQLITE_OK : rc;
}

/* insert into %_content (rowid, ...) values ([rowid], [pValues]) */
static int content_insert(fulltext_vtab *v, sqlite3_value *rowid,
                          sqlite3_value **pValues){
  sqlite3_stmt *s;
  int i;
  int rc = sql_get_statement(v, CONTENT_INSERT_STMT, &s);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3_bind_value(s, 1, rowid);
  if( rc!=SQLITE_OK ) return rc;

  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]);
  }
  free((void *) pString);
}

/* select * from %_content where rowid = [iRow]
 * The caller must delete the returned array and all strings in it.
 *
 * TODO: Perhaps we should return pointer/length strings here for consistency