bidi.java

linux下建立JAVA虚拟机的源码KAFFE

            continue;
          }
        // No explicit embedding.
        boolean isLtoR = false;
        boolean isSpecial = true;
        switch (types[i])
          {
            case Character.DIRECTIONALITY_LEFT_TO_RIGHT_EMBEDDING:
            case Character.DIRECTIONALITY_LEFT_TO_RIGHT_OVERRIDE:
              isLtoR = true;
              // Fall through.
            case Character.DIRECTIONALITY_RIGHT_TO_LEFT_EMBEDDING:
            case Character.DIRECTIONALITY_RIGHT_TO_LEFT_OVERRIDE:
              {
                byte newEmbedding;
                if (isLtoR)
                  {
                    // Least greater even.
                    newEmbedding = (byte) ((currentEmbedding & ~1) + 2);
                  }
                else
                  {
                    // Least greater odd.
                    newEmbedding = (byte) ((currentEmbedding + 1) | 1);
                  }
                // FIXME: we don't properly handle invalid pushes.
                if (newEmbedding < MAX_DEPTH)
                  {
                    // The new level is valid.  Push the old value.
                    // See above for a comment on the encoding here.
                    if (directionalOverride != -1)
                      currentEmbedding |= Byte.MIN_VALUE;
                    embeddingStack[sp++] = currentEmbedding;
                    currentEmbedding = newEmbedding;
                    if (types[i] == Character.DIRECTIONALITY_LEFT_TO_RIGHT_OVERRIDE)
                      directionalOverride = Character.DIRECTIONALITY_LEFT_TO_RIGHT;
                    else if (types[i] == Character.DIRECTIONALITY_RIGHT_TO_LEFT_OVERRIDE)
                      directionalOverride = Character.DIRECTIONALITY_RIGHT_TO_LEFT;
                    else
                      directionalOverride = -1;
                  }
                }
              break;
            case Character.DIRECTIONALITY_POP_DIRECTIONAL_FORMAT:
              {
                // FIXME: we don't properly handle a pop with a corresponding
                // invalid push.
                if (sp == 0)
                  {
                    // We saw a pop without a push.  Just ignore it.
                    break;
                  }
                byte newEmbedding = embeddingStack[--sp];
                currentEmbedding = (byte) (newEmbedding & 0x7f);
                if (newEmbedding < 0)
                  directionalOverride
                  = (((newEmbedding & 1) == 0)
                      ? Character.DIRECTIONALITY_LEFT_TO_RIGHT
                      : Character.DIRECTIONALITY_RIGHT_TO_LEFT);
                else
                  directionalOverride = -1;
              }
              break;
            default:
              isSpecial = false;
              break;
          }
        levels[i] = currentEmbedding;
        if (isSpecial)
          {
            // Mark this character for removal.
            if (formatterIndices == null)
              formatterIndices = new ArrayList();
            formatterIndices.add(Integer.valueOf(i));
          }
        else if (directionalOverride != -1)
          types[i] = directionalOverride;
      }

    // Remove the formatting codes and update both the arrays
    // and 'length'.  It would be more efficient not to remove
    // these codes, but it is also more complicated.  Also, the
    // Unicode algorithm reference does not properly describe
    // how this is to be done -- from what I can tell, their suggestions
    // in this area will not yield the correct results.
    if (formatterIndices == null)
      return;
    int output = 0, input = 0;
    final int size = formatterIndices.size();
    for (int i = 0; i <= size; ++i)
      {
        int nextFmt;
        if (i == size)
          nextFmt = length;
        else
          nextFmt = ((Integer) formatterIndices.get(i)).intValue();
        // Non-formatter codes are from 'input' to 'nextFmt'.
        int len = nextFmt - input;
        System.arraycopy(levels, input, levels, output, len);
        System.arraycopy(types, input, types, output, len);
        output += len;
        input = nextFmt + 1;
      }
    length -= formatterIndices.size();
  }

  /**
   * An internal function to compute the boundaries of runs
   * in the text.  It isn't strictly necessary to do this, but
   * it lets us write some following passes in a less complicated
   * way.  Also it lets us efficiently implement some of the public
   * methods.  A run is simply a sequence of characters at the
   * same level.
   */
  private void computeRuns()
  {
    int runCount = 0;
    int currentEmbedding = baseEmbedding;
    for (int i = 0; i < length; ++i)
      {
        if (levels[i] != currentEmbedding)
          {
            currentEmbedding = levels[i];
            ++runCount;
          }
      }

    // This may be called multiple times.  If so, and if
    // the number of runs has not changed, then don't bother
    // allocating a new array.
    if (runs == null || runs.length != runCount + 1)
      runs = new int[runCount + 1];
    int where = 0;
    int lastRunStart = 0;
    currentEmbedding = baseEmbedding;
    for (int i = 0; i < length; ++i)
      {
        if (levels[i] != currentEmbedding)
          {
            runs[where++] = lastRunStart;
            lastRunStart = i;
            currentEmbedding = levels[i];
          }
      }
    runs[where++] = lastRunStart;
  }

  /**
   * An internal method to resolve weak types.  This implements
   * rules W1 through W7.
   */
  private void resolveWeakTypes()
  {
    final int runCount = getRunCount();
    
    int previousLevel = baseEmbedding;
    for (int run = 0; run < runCount; ++run)
      {
        int start = getRunStart(run);
        int end = getRunLimit(run);
        int level = getRunLevel(run);

        // These are the names used in the Bidi algorithm.
        byte sor = (((Math.max(previousLevel, level) % 2) == 0)
                      ? Character.DIRECTIONALITY_LEFT_TO_RIGHT
                      : Character.DIRECTIONALITY_RIGHT_TO_LEFT);
        int nextLevel;
        if (run == runCount - 1)
          nextLevel = baseEmbedding;
        else
          nextLevel = getRunLevel(run + 1);
        byte eor = (((Math.max(level, nextLevel) % 2) == 0)
                      ? Character.DIRECTIONALITY_LEFT_TO_RIGHT
                      : Character.DIRECTIONALITY_RIGHT_TO_LEFT);

        byte prevType = sor;
        byte prevStrongType = sor;
        for (int i = start; i < end; ++i)
          {
            final byte nextType = (i == end - 1) ? eor : types[i + 1];

            // Rule W1: change NSM to the prevailing direction.
            if (types[i] == Character.DIRECTIONALITY_NONSPACING_MARK)
              types[i] = prevType;
            else
              prevType = types[i];

            // Rule W2: change EN to AN in some cases.
            if (types[i] == Character.DIRECTIONALITY_EUROPEAN_NUMBER)
              {
                if (prevStrongType == Character.DIRECTIONALITY_RIGHT_TO_LEFT_ARABIC)
                  types[i] = Character.DIRECTIONALITY_ARABIC_NUMBER;
              }
            else if (types[i] == Character.DIRECTIONALITY_LEFT_TO_RIGHT
                     || types[i] == Character.DIRECTIONALITY_RIGHT_TO_LEFT
                     || types[i] == Character.DIRECTIONALITY_RIGHT_TO_LEFT_ARABIC)
              prevStrongType = types[i];

            // Rule W3: change AL to R.
            if (types[i] == Character.DIRECTIONALITY_RIGHT_TO_LEFT_ARABIC)
              types[i] = Character.DIRECTIONALITY_RIGHT_TO_LEFT;

            // Rule W4: handle separators between two numbers.
            if (prevType == Character.DIRECTIONALITY_EUROPEAN_NUMBER
                && nextType == Character.DIRECTIONALITY_EUROPEAN_NUMBER)
              {
                if (types[i] == Character.DIRECTIONALITY_EUROPEAN_NUMBER_SEPARATOR
                    || types[i] == Character.DIRECTIONALITY_COMMON_NUMBER_SEPARATOR)
                  types[i] = nextType;
              }
            else if (prevType == Character.DIRECTIONALITY_ARABIC_NUMBER
                     && nextType == Character.DIRECTIONALITY_ARABIC_NUMBER
                     && types[i] == Character.DIRECTIONALITY_COMMON_NUMBER_SEPARATOR)
              types[i] = nextType;

            // Rule W5: change a sequence of european terminators to
            // european numbers, if they are adjacent to european numbers.
            // We also include BN characters in this.
            if (types[i] == Character.DIRECTIONALITY_EUROPEAN_NUMBER_TERMINATOR
                || types[i] == Character.DIRECTIONALITY_BOUNDARY_NEUTRAL)
              {
                if (prevType == Character.DIRECTIONALITY_EUROPEAN_NUMBER)
                  types[i] = prevType;
                else
                  {
                    // Look ahead to see if there is an EN terminating this
                    // sequence of ETs.
                    int j = i + 1;
                    while (j < end
                           && (types[j] == Character.DIRECTIONALITY_EUROPEAN_NUMBER_TERMINATOR
                               || types[j] == Character.DIRECTIONALITY_BOUNDARY_NEUTRAL))
                      ++j;
                    if (j < end
                        && types[j] == Character.DIRECTIONALITY_EUROPEAN_NUMBER)
                      {
                        // Change them all to EN now.
                        for (int k = i; k < j; ++k)
                          types[k] = Character.DIRECTIONALITY_EUROPEAN_NUMBER;
                      }
                  }
              }

            // Rule W6: separators and terminators change to ON.
            // Again we include BN.
            if (types[i] == Character.DIRECTIONALITY_EUROPEAN_NUMBER_TERMINATOR
                || types[i] == Character.DIRECTIONALITY_EUROPEAN_NUMBER_TERMINATOR
                || types[i] == Character.DIRECTIONALITY_COMMON_NUMBER_SEPARATOR
                || types[i] == Character.DIRECTIONALITY_BOUNDARY_NEUTRAL)
              types[i] = Character.DIRECTIONALITY_OTHER_NEUTRALS;

            // Rule W7: change european number types.
            if (prevStrongType == Character.DIRECTIONALITY_LEFT_TO_RIGHT
                && types[i] == Character.DIRECTIONALITY_EUROPEAN_NUMBER)
              types[i] = prevStrongType;
          }

        previousLevel = level;
      }
  }

  /**
   * An internal method to resolve neutral types.  This implements
   * rules N1 and N2.
   */
  private void resolveNeutralTypes()
  {
    // This implements rules N1 and N2.
    final int runCount = getRunCount();
    
    int previousLevel = baseEmbedding;
    for (int run = 0; run < runCount; ++run)
      {
        int start = getRunStart(run);
        int end = getRunLimit(run);
        int level = getRunLevel(run);

        byte embeddingDirection
          = (((level % 2) == 0) ? Character.DIRECTIONALITY_LEFT_TO_RIGHT
                                : Character.DIRECTIONALITY_RIGHT_TO_LEFT);
        // These are the names used in the Bidi algorithm.
        byte sor = (((Math.max(previousLevel, level) % 2) == 0)
                      ? Character.DIRECTIONALITY_LEFT_TO_RIGHT
                      : Character.DIRECTIONALITY_RIGHT_TO_LEFT);
        int nextLevel;
        if (run == runCount - 1)
          nextLevel = baseEmbedding;
        else
          nextLevel = getRunLevel(run + 1);
        byte eor = (((Math.max(level, nextLevel) % 2) == 0)
                      ? Character.DIRECTIONALITY_LEFT_TO_RIGHT
                      : Character.DIRECTIONALITY_RIGHT_TO_LEFT);

        byte prevStrong = sor;
        int neutralStart = -1;
        for (int i = start; i <= end; ++i)
          {
            byte newStrong = -1;
            byte thisType = i == end ? eor : types[i];
            switch (thisType)
              {
              case Character.DIRECTIONALITY_LEFT_TO_RIGHT:
                newStrong = Character.DIRECTIONALITY_LEFT_TO_RIGHT;
                break;
              case Character.DIRECTIONALITY_RIGHT_TO_LEFT:
              case Character.DIRECTIONALITY_ARABIC_NUMBER:
              case Character.DIRECTIONALITY_EUROPEAN_NUMBER:
                newStrong = Character.DIRECTIONALITY_RIGHT_TO_LEFT;
                break;
              case Character.DIRECTIONALITY_BOUNDARY_NEUTRAL:
              case Character.DIRECTIONALITY_OTHER_NEUTRALS:
              case Character.DIRECTIONALITY_SEGMENT_SEPARATOR:
              case Character.DIRECTIONALITY_PARAGRAPH_SEPARATOR:
                if (neutralStart == -1)
                  neutralStart = i;
                break;
              }
            // If we see a strong character, update all the neutrals.
            if (newStrong != -1)
              {
                if (neutralStart != -1)
                  {
                    byte override = (prevStrong == newStrong
                                     ? prevStrong
                                     : embeddingDirection);
                    for (int j = neutralStart; j < i; ++j)
                      types[i] = override;
                  }
                prevStrong = newStrong;
                neutralStart = -1;
              }
          }

        previousLevel = level;
      }