src/Pure/General/utf8.scala
author wenzelm
Sun, 30 Apr 2017 09:23:03 +0200
changeset 65641 3b0110e25745
parent 65196 e8760a98db78
child 68224 1f7308050349
permissions -rw-r--r--
tuned message;
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
64639
bad5de3f9554 clarified directories;
wenzelm
parents: 64617
diff changeset
     1
/*  Title:      Pure/General/utf8.scala
50203
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
     2
    Author:     Makarius
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
     3
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
     4
Variations on UTF-8.
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
     5
*/
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
     6
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
     7
package isabelle
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
     8
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
     9
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    10
import java.nio.charset.Charset
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    11
import scala.io.Codec
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    12
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    13
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    14
object UTF8
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    15
{
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    16
  /* charset */
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    17
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    18
  val charset_name: String = "UTF-8"
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    19
  val charset: Charset = Charset.forName(charset_name)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    20
  def codec(): Codec = Codec(charset)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    21
62527
aae9a2a855e0 tuned signature;
wenzelm
parents: 54444
diff changeset
    22
  def bytes(s: String): Array[Byte] = s.getBytes(charset)
aae9a2a855e0 tuned signature;
wenzelm
parents: 54444
diff changeset
    23
50203
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    24
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    25
  /* permissive UTF-8 decoding */
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    26
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    27
  // see also http://en.wikipedia.org/wiki/UTF-8#Description
54444
a2290f36d1d6 prefer UTF8.decode_permissive;
wenzelm
parents: 54440
diff changeset
    28
  // overlong encodings enable byte-stuffing of low-ASCII
50203
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    29
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    30
  def decode_permissive(text: CharSequence): String =
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    31
  {
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    32
    val buf = new java.lang.StringBuilder(text.length)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    33
    var code = -1
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    34
    var rest = 0
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    35
    def flush()
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    36
    {
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    37
      if (code != -1) {
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    38
        if (rest == 0 && Character.isValidCodePoint(code))
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    39
          buf.appendCodePoint(code)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    40
        else buf.append('\uFFFD')
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    41
        code = -1
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    42
        rest = 0
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    43
      }
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    44
    }
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    45
    def init(x: Int, n: Int)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    46
    {
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    47
      flush()
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    48
      code = x
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    49
      rest = n
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    50
    }
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    51
    def push(x: Int)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    52
    {
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    53
      if (rest <= 0) init(x, -1)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    54
      else {
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    55
        code <<= 6
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    56
        code += x
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    57
        rest -= 1
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    58
      }
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    59
    }
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    60
    for (i <- 0 until text.length) {
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    61
      val c = text.charAt(i)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    62
      if (c < 128) { flush(); buf.append(c) }
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    63
      else if ((c & 0xC0) == 0x80) push(c & 0x3F)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    64
      else if ((c & 0xE0) == 0xC0) init(c & 0x1F, 1)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    65
      else if ((c & 0xF0) == 0xE0) init(c & 0x0F, 2)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    66
      else if ((c & 0xF8) == 0xF0) init(c & 0x07, 3)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    67
    }
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    68
    flush()
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    69
    buf.toString
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    70
  }
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    71
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    72
  private class Decode_Chars(decode: String => String,
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    73
    buffer: Array[Byte], start: Int, end: Int) extends CharSequence
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    74
  {
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    75
    def length: Int = end - start
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    76
    def charAt(i: Int): Char = (buffer(start + i).asInstanceOf[Int] & 0xFF).asInstanceOf[Char]
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    77
    def subSequence(i: Int, j: Int): CharSequence =
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    78
      new Decode_Chars(decode, buffer, start + i, start + j)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    79
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    80
    // toString with adhoc decoding: abuse of CharSequence interface
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    81
    override def toString: String = decode(decode_permissive(this))
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    82
  }
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    83
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    84
  def decode_chars(decode: String => String,
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    85
    buffer: Array[Byte], start: Int, end: Int): CharSequence =
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    86
  {
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    87
    require(0 <= start && start <= end && end <= buffer.length)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    88
    new Decode_Chars(decode, buffer, start, end)
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    89
  }
00d8ad713e32 explicit module UTF8;
wenzelm
parents:
diff changeset
    90
}