Beispiel #1
0
  /**
   * 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;
  }
Beispiel #2
0
  @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;
  }
Beispiel #3
0
 /**
  * Finds line and column for the specified query parser.
  *
  * @param parser parser
  */
 void pos(final InputParser parser) {
   markedCol = parser.mark;
   if (info != null) return;
   // check if line/column information has already been added
   parser.pos = Math.min(parser.mark, parser.length);
   info = new InputInfo(parser);
 }
Beispiel #4
0
 /**
  * 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);
 }
Beispiel #5
0
 @Override
 void calcSize(final long[] minMax) {
   minMax[0] = Math.min(minMax[0], 1);
 }
Beispiel #6
0
  /**
   * Formats the specified number and returns a string representation.
   *
   * @param item item
   * @param pics pictures
   * @param ii input info
   * @return picture variables
   * @throws QueryException query exception
   */
  private byte[] format(final ANum item, final Picture[] pics, final InputInfo ii)
      throws QueryException {

    // Rule 1: return results for NaN
    final double d = item.dbl(ii);
    if (Double.isNaN(d)) return nan;

    // Rule 2: check if value if negative (smaller than zero or -0)
    final boolean neg = d < 0 || d == 0 && Double.doubleToLongBits(d) == Long.MIN_VALUE;
    final Picture pic = pics[neg && pics.length == 2 ? 1 : 0];
    final IntList res = new IntList(), intgr = new IntList(), fract = new IntList();
    int exp = 0;

    // Rule 3: percent/permille
    ANum num = item;
    if (pic.pc) num = (ANum) Calc.MULT.ev(num, Int.get(100), ii);
    if (pic.pm) num = (ANum) Calc.MULT.ev(num, Int.get(1000), ii);

    if (Double.isInfinite(num.dbl(ii))) {
      // Rule 4: infinity
      intgr.add(new TokenParser(inf).toArray());
    } else {
      // Rule 5: exponent
      if (pic.minExp != 0 && d != 0) {
        BigDecimal dec = num.dec(ii).abs().stripTrailingZeros();
        int scl = 0;
        if (dec.compareTo(BigDecimal.ONE) >= 0) {
          scl = dec.setScale(0, RoundingMode.HALF_DOWN).precision();
        } else {
          while (dec.compareTo(BigDecimal.ONE) < 0) {
            dec = dec.multiply(BigDecimal.TEN);
            scl--;
          }
          scl++;
        }
        exp = scl - pic.min[0];
        if (exp != 0) {
          final BigDecimal n = BigDecimal.TEN.pow(Math.abs(exp));
          num = (ANum) Calc.MULT.ev(num, Dec.get(exp > 0 ? BigDecimal.ONE.divide(n) : n), ii);
        }
      }
      num = num.round(pic.maxFrac, true).abs();

      // convert positive number to string
      final String s =
          (num instanceof Dbl || num instanceof Flt
                  ? Dec.get(BigDecimal.valueOf(num.dbl(ii)))
                  : num)
              .toString();

      // integer/fractional separator
      final int sep = s.indexOf('.');

      // create integer part
      final int sl = s.length();
      final int il = sep == -1 ? sl : sep;
      for (int i = il; i < pic.min[0]; ++i) intgr.add(zero);
      // fractional number: skip leading 0
      if (!s.startsWith("0.") || pic.min[0] > 0) {
        for (int i = 0; i < il; i++) intgr.add(zero + s.charAt(i) - '0');
      }

      // squeeze in grouping separators
      if (pic.group[0].length == 1 && pic.group[0][0] > 0) {
        // regular pattern with repeating separators
        for (int p = intgr.size() - (neg ? 2 : 1); p > 0; --p) {
          if (p % pic.group[0][0] == 0) intgr.insert(intgr.size() - p, grouping);
        }
      } else {
        // irregular pattern, or no separators at all
        final int gl = pic.group[0].length;
        for (int g = 0; g < gl; ++g) {
          final int pos = intgr.size() - pic.group[0][g];
          if (pos > 0) intgr.insert(pos, grouping);
        }
      }

      // create fractional part
      final int fl = sep == -1 ? 0 : sl - il - 1;
      if (fl != 0) for (int i = sep + 1; i < sl; i++) fract.add(zero + s.charAt(i) - '0');
      for (int i = fl; i < pic.min[1]; ++i) fract.add(zero);

      // squeeze in grouping separators in a reverse manner
      final int ul = fract.size();
      for (int p = pic.group[1].length - 1; p >= 0; p--) {
        final int pos = pic.group[1][p];
        if (pos < ul) fract.insert(pos, grouping);
      }
    }

    // add minus sign
    if (neg && pics.length != 2) res.add(minus);
    // add prefix and integer part
    res.add(pic.prefSuf[0].toArray()).add(intgr.finish());
    // add fractional part
    if (!fract.isEmpty()) res.add(decimal).add(fract.finish());
    // add exponent
    if (pic.minExp != 0) {
      res.add(exponent);
      if (exp < 0) res.add(minus);
      final String s = Integer.toString(Math.abs(exp));
      final int sl = s.length();
      for (int i = sl; i < pic.minExp; i++) res.add(zero);
      for (int i = 0; i < sl; i++) res.add(zero + s.charAt(i) - '0');
    }
    // add suffix
    res.add(pic.prefSuf[1].toArray());
    return new TokenBuilder(res.finish()).finish();
  }
Beispiel #7
0
  /**
   * Analyzes the specified patterns.
   *
   * @param patterns patterns
   * @return picture variables
   */
  private Picture[] analyze(final byte[][] patterns) {
    // pictures
    final int picL = patterns.length;
    final Picture[] pics = new Picture[picL];

    // analyze patterns
    for (int p = 0; p < picL; p++) {
      final byte[] pt = patterns[p];
      final Picture pic = new Picture();

      // position (integer/fractional)
      int pos = 0;
      // active character found
      boolean act = false;
      // number of characters after exponent
      int exp = -1;
      // number of optional characters
      final int[] opt = new int[2];

      // loop through all characters
      final int pl = pt.length;
      for (int i = 0, cl; i < pl; i += cl) {
        final int ch = ch(pt, i);
        cl = cl(pt, i);
        boolean active = contains(actives, ch);

        if (ch == decimal) {
          ++pos;
          act = false;
        } else if (ch == optional) {
          opt[pos]++;
        } else if (ch == exponent) {
          if (act && containsActive(pt, i + cl)) {
            exp = 0;
          } else {
            active = false;
          }
        } else if (ch == grouping) {
          if (pos == 0) pic.group[pos] = Array.add(pic.group[pos], pic.min[pos] + opt[pos]);
        } else if (contains(digits, ch)) {
          if (exp == -1) pic.min[pos]++;
          else exp++;
        }

        if (active) {
          act = true;
        } else {
          // passive characters
          pic.pc |= ch == percent;
          pic.pm |= ch == permille;
          // prefixes/suffixes
          pic.prefSuf[pos == 0 && act ? pos + 1 : pos].add(ch);
        }
      }
      // finalize integer-part-grouping-positions
      final int[] igp = pic.group[0];
      final int igl = igp.length;
      for (int g = 0; g < igl; ++g) igp[g] = pic.min[0] + opt[0] - igp[g];

      // check if integer-part-grouping-positions are regular
      // if yes, they are replaced with a single position
      if (igl > 1) {
        boolean reg = true;
        final int i = igp[igl - 1];
        for (int g = igl - 2; g >= 0; --g) reg &= i * igl == igp[g];
        if (reg) pic.group[0] = new int[] {i};
      }

      pic.maxFrac = pic.min[1] + opt[1];
      pic.minExp = Math.max(0, exp);
      pics[p] = pic;
    }
    return pics;
  }