Class UTF8Input

java/com/google/re2j/MachineInput.java:63–171

  // An implementation of MachineInput for UTF-8 byte arrays.
  // |pos| and |width| are byte indices.
  private static class UTF8Input extends MachineInput {

    final byte[] b;
    final int start;
    final int end;

    UTF8Input(byte[] b) {
      this.b = b;
      start = 0;
      end = b.length;
    }

    UTF8Input(byte[] b, int start, int end) {
      if (end > b.length) {
        throw new ArrayIndexOutOfBoundsException(
            "end is greater than length: " + end + " > " + b.length);
      }
      this.b = b;
      this.start = start;
      this.end = end;
    }

    @Override
    int step(int i) {
      i += start;
      if (i >= end) {
        return EOF;
      }

      // UTF-8.  RFC 3629 in five lines:
      //
      // Unicode code points            UTF-8 encoding (binary)
      //         00-7F  (7 bits)   0tuvwxyz
      //     0080-07FF (11 bits)   110pqrst 10uvwxyz
      //     0800-FFFF (16 bits)   1110jklm 10npqrst 10uvwxyz
      // 010000-10FFFF (21 bits)   11110efg 10hijklm 10npqrst 10uvwxyz
      int x = b[i++] & 0xff; // zero extend
      if ((x & 0x80) == 0) {
        return x << 3 | 1;
      } else if ((x & 0xE0) == 0xC0) { // 110xxxxx
        x = x & 0x1F;
        if (i >= end) {
          return EOF;
        }
        x = x << 6 | (b[i++] & 0x3F);
        return x << 3 | 2;
      } else if ((x & 0xF0) == 0xE0) { // 1110xxxx
        x = x & 0x0F;
        if (i + 1 >= end) {
          return EOF;
        }
        x = x << 6 | (b[i++] & 0x3F);
        x = x << 6 | (b[i++] & 0x3F);
        return x << 3 | 3;
      } else { // 11110xxx
        x = x & 0x07;
        if (i + 2 >= end) {
          return EOF;