コード例 #1
0
ファイル: FTWords.java プロジェクト: jefferya/basex
  /**
   * Evaluates the full-text match.
   *
   * @param qc query context
   * @return number of tokens, used for scoring
   * @throws QueryException query exception
   */
  private int contains(final QueryContext qc) throws QueryException {
    first = true;
    final FTLexer lexer = ftt.lexer(qc.ftToken);

    // use faster evaluation for default options
    int num = 0;
    if (fast) {
      for (final byte[] t : tokens) {
        final FTTokens qtok = ftt.cache(t);
        num = Math.max(num, ftt.contains(qtok, lexer) * qtok.length());
      }
      return num;
    }

    // find and count all occurrences
    final boolean all = mode == FTMode.ALL || mode == FTMode.ALL_WORDS;
    int oc = 0;
    for (final byte[] w : unique(tokens(qc))) {
      final FTTokens qtok = ftt.cache(w);
      final int o = ftt.contains(qtok, lexer);
      if (all && o == 0) return 0;
      num = Math.max(num, o * qtok.length());
      oc += o;
    }

    // check if occurrences are in valid range. if yes, return number of tokens
    final long mn = occ != null ? toLong(occ[0], qc) : 1;
    final long mx = occ != null ? toLong(occ[1], qc) : Long.MAX_VALUE;
    if (mn == 0 && oc == 0) matches = FTNot.not(matches);
    return oc >= mn && oc <= mx ? Math.max(1, num) : 0;
  }
コード例 #2
0
ファイル: FTWords.java プロジェクト: jefferya/basex
  @Override
  public boolean indexAccessible(final IndexInfo ii) {
    /* If the following conditions yield true, the index is accessed:
     * - all query terms are statically available
     * - no FTTimes option is specified
     * - explicitly set case, diacritics and stemming match options do not
     *   conflict with index options. */
    data = ii.ic.data;
    final MetaData md = data.meta;
    final FTOpt fto = ftt.opt;

    /* Index will be applied if no explicit match options have been set
     * that conflict with the index options. As a consequence, though, index-
     * based querying might yield other results than sequential scanning. */
    if (occ != null
        || fto.cs != null && md.casesens == (fto.cs == FTCase.INSENSITIVE)
        || fto.isSet(DC) && md.diacritics != fto.is(DC)
        || fto.isSet(ST) && md.stemming != fto.is(ST)
        || fto.ln != null && !fto.ln.equals(md.language)) return false;

    // adopt database options to tokenizer
    fto.copy(md);

    // estimate costs if text is not known at compile time
    if (tokens == null) {
      ii.costs = Math.max(2, data.meta.size / 30);
      return true;
    }

    // summarize number of hits; break loop if no hits are expected
    final FTLexer ft = new FTLexer(fto);
    ii.costs = 0;
    for (byte[] t : tokens) {
      ft.init(t);
      while (ft.hasNext()) {
        final byte[] tok = ft.nextToken();
        if (fto.sw != null && fto.sw.contains(tok)) continue;

        if (fto.is(WC)) {
          // don't use index if one of the terms starts with a wildcard
          t = ft.get();
          if (t[0] == '.') return false;
          // don't use index if certain characters or more than 1 dot are found
          int d = 0;
          for (final byte w : t) {
            if (w == '{' || w == '\\' || w == '.' && ++d > 1) return false;
          }
        }
        // favor full-text index requests over exact queries
        final int costs = data.costs(ft);
        if (costs != 0) ii.costs += Math.max(2, costs / 100);
      }
    }
    return true;
  }
コード例 #3
0
ファイル: FTIndex.java プロジェクト: fpapai/basex
  /**
   * Performs a wildcard search for the specified token.
   *
   * @param token token to look for
   * @return iterator
   */
  private synchronized IndexIterator wc(final byte[] token) {
    final FTIndexIterator it = FTIndexIterator.FTEMPTY;
    final FTWildcard wc = new FTWildcard(token);
    if (!wc.parse()) return it;

    final IntList pr = new IntList();
    final IntList ps = new IntList();
    final byte[] pref = wc.prefix();
    final int pl = pref.length, tl = tp.length;
    final int l = Math.min(tl - 1, wc.max());
    for (int ti = pl; ti <= l; ti++) {
      int i = tp[ti];
      if (i == -1) continue;
      int c = ti + 1;
      int e = -1;
      while (c < tl && e == -1) e = tp[c++];
      i = find(pref, i, e, ti);

      while (i < e) {
        final byte[] t = inY.readBytes(i, ti);
        if (!startsWith(t, pref)) break;
        if (wc.match(t)) {
          inZ.cursor(pointer(i, ti));
          final int s = size(i, ti);
          for (int d = 0; d < s; d++) {
            pr.add(inZ.readNum());
            ps.add(inZ.readNum());
          }
        }
        i += ti + ENTRY;
      }
    }
    return iter(new FTCache(pr, ps), token);
  }
コード例 #4
0
ファイル: FTAnd.java プロジェクト: BaseXdb/basex
 /**
  * Merges two matches.
  *
  * @param i1 first item
  * @param i2 second item
  */
 private static void and(final FTNode i1, final FTNode i2) {
   final FTMatches all = new FTMatches((byte) Math.max(i1.matches().pos, i2.matches().pos));
   for (final FTMatch s1 : i1.matches()) {
     for (final FTMatch s2 : i2.matches()) {
       all.add(new FTMatch(s1.size() + s2.size()).add(s1).add(s2));
     }
   }
   i1.score(Scoring.avg(i1.score() + i2.score(), 2));
   i1.matches(all);
 }
コード例 #5
0
ファイル: FTIndex.java プロジェクト: fpapai/basex
  /**
   * Performs a fuzzy search for the specified token with a maximum number of errors.
   *
   * @param token token to look for
   * @param k number of errors allowed
   * @return iterator
   */
  private synchronized IndexIterator fuzzy(final byte[] token, final int k) {
    FTIndexIterator it = FTIndexIterator.FTEMPTY;
    final int tokl = token.length, tl = tp.length;
    final int e = Math.min(tl - 1, tokl + k);
    int s = Math.max(1, tokl - k) - 1;

    while (++s <= e) {
      int p = tp[s];
      if (p == -1) continue;
      int t = s + 1, r = -1;
      while (t < tl && r == -1) r = tp[t++];
      while (p < r) {
        if (ls.similar(inY.readBytes(p, s), token, k)) {
          it = FTIndexIterator.union(iter(pointer(p, s), size(p, s), inZ, token), it);
        }
        p += s + ENTRY;
      }
    }
    return it;
  }
コード例 #6
0
ファイル: FTIndex.java プロジェクト: fpapai/basex
  @Override
  public synchronized int costs(final IndexToken it) {
    final byte[] tok = it.get();
    if (tok.length > data.meta.maxlen) return Integer.MAX_VALUE;

    // estimate costs for queries which stretch over multiple index entries
    final FTOpt opt = ((FTLexer) it).ftOpt();
    if (opt.is(FZ) || opt.is(WC)) return Math.max(1, data.meta.size >> 4);

    return entry(tok).size;
  }