src/Pure/General/yxml.scala
author wenzelm
Mon, 25 Aug 2008 20:01:17 +0200
changeset 27993 6dd90ef9f927
parent 27971 57dc3bd6f841
child 29140 e7ac5bb20aed
permissions -rw-r--r--
simplified exceptions: use plain error function / RuntimeException;
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
27930
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
     1
/*  Title:      Pure/General/yxml.scala
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
     2
    ID:         $Id$
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
     3
    Author:     Makarius
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
     4
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
     5
Efficient text representation of XML trees.
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
     6
*/
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
     7
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
     8
package isabelle
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
     9
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    10
import java.util.regex.Pattern
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    11
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    12
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    13
object YXML {
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    14
27943
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    15
  /* chunk markers */
27930
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    16
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    17
  private val X = '\5'
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    18
  private val Y = '\6'
27960
65b10d8ef0c6 added parse_failsafe;
wenzelm
parents: 27946
diff changeset
    19
  private val X_string = X.toString
27945
d2dc5a1903e8 tuned parse performance: avoid splitting terminal Y chunk;
wenzelm
parents: 27944
diff changeset
    20
  private val Y_string = Y.toString
27930
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    21
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    22
  def detect(source: CharSequence) = {
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    23
    source.length >= 2 &&
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    24
    source.charAt(0) == X &&
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    25
    source.charAt(1) == Y
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    26
  }
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    27
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    28
27943
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    29
  /* iterate over chunks (resembles space_explode in ML) */
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    30
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    31
  private def chunks(sep: Char, source: CharSequence) = new Iterator[CharSequence] {
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    32
    private val end = source.length
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    33
    private var state = if (end == 0) None else get_chunk(-1)
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    34
    private def get_chunk(i: Int) = {
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    35
      if (i < end) {
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    36
        var j = i; do j += 1 while (j < end && source.charAt(j) != sep)
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    37
        Some((source.subSequence(i + 1, j), j))
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    38
      }
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    39
      else None
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    40
    }
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    41
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    42
    def hasNext() = state.isDefined
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    43
    def next() = state match {
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    44
      case Some((s, i)) => { state = get_chunk(i); s }
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    45
      case None => throw new NoSuchElementException("next on empty iterator")
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    46
    }
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    47
  }
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    48
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    49
27930
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    50
  /* parsing */
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    51
27993
6dd90ef9f927 simplified exceptions: use plain error function / RuntimeException;
wenzelm
parents: 27971
diff changeset
    52
  private def err(msg: String) = error("Malformed YXML: " + msg)
27930
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    53
  private def err_attribute() = err("bad attribute")
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    54
  private def err_element() = err("bad element")
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    55
  private def err_unbalanced(name: String) =
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    56
    if (name == "") err("unbalanced element")
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    57
    else err("unbalanced element \"" + name + "\"")
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    58
27944
2bf3f30558ed parse_attrib: more efficient due to indexOf('=');
wenzelm
parents: 27943
diff changeset
    59
  private def parse_attrib(source: CharSequence) = {
2bf3f30558ed parse_attrib: more efficient due to indexOf('=');
wenzelm
parents: 27943
diff changeset
    60
    val s = source.toString
2bf3f30558ed parse_attrib: more efficient due to indexOf('=');
wenzelm
parents: 27943
diff changeset
    61
    val i = s.indexOf('=')
2bf3f30558ed parse_attrib: more efficient due to indexOf('=');
wenzelm
parents: 27943
diff changeset
    62
    if (i <= 0) err_attribute()
27946
ec706ad37564 parse_attrib: proper index of name end!
wenzelm
parents: 27945
diff changeset
    63
    (s.substring(0, i), s.substring(i + 1))
27944
2bf3f30558ed parse_attrib: more efficient due to indexOf('=');
wenzelm
parents: 27943
diff changeset
    64
  }
27930
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    65
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    66
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    67
  def parse_body(source: CharSequence) = {
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    68
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    69
    /* stack operations */
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    70
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    71
    var stack: List[((String, XML.Attributes), List[XML.Tree])] = null
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    72
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    73
    def add(x: XML.Tree) = stack match {
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    74
      case ((elem, body) :: pending) => stack = (elem, x :: body) :: pending
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    75
    }
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    76
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    77
    def push(name: String, atts: XML.Attributes) =
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    78
      if (name == "") err_element()
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    79
      else stack = ((name, atts), Nil) :: stack
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    80
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    81
    def pop() = stack match {
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    82
      case ((("", _), _) :: _) => err_unbalanced("")
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    83
      case (((name, atts), body) :: pending) =>
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    84
        stack = pending; add(XML.Elem(name, atts, body.reverse))
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    85
    }
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    86
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    87
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    88
    /* parse chunks */
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    89
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    90
    stack = List((("", Nil), Nil))
27943
f34ff5e7728f replaced Pattern.split by chunks iterator (more efficient, resembles ML version more closely);
wenzelm
parents: 27930
diff changeset
    91
    for (chunk <- chunks(X, source) if chunk != "") {
27945
d2dc5a1903e8 tuned parse performance: avoid splitting terminal Y chunk;
wenzelm
parents: 27944
diff changeset
    92
      if (chunk == Y_string) pop()
d2dc5a1903e8 tuned parse performance: avoid splitting terminal Y chunk;
wenzelm
parents: 27944
diff changeset
    93
      else {
d2dc5a1903e8 tuned parse performance: avoid splitting terminal Y chunk;
wenzelm
parents: 27944
diff changeset
    94
        chunks(Y, chunk).toList match {
d2dc5a1903e8 tuned parse performance: avoid splitting terminal Y chunk;
wenzelm
parents: 27944
diff changeset
    95
          case "" :: name :: atts => push(name.toString, atts.map(parse_attrib))
d2dc5a1903e8 tuned parse performance: avoid splitting terminal Y chunk;
wenzelm
parents: 27944
diff changeset
    96
          case txts => for (txt <- txts) add(XML.Text(txt.toString))
d2dc5a1903e8 tuned parse performance: avoid splitting terminal Y chunk;
wenzelm
parents: 27944
diff changeset
    97
        }
27930
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    98
      }
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
    99
    }
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   100
    stack match {
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   101
      case List((("", _), result)) => result.reverse
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   102
      case ((name, _), _) :: _ => err_unbalanced(name)
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   103
    }
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   104
  }
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   105
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   106
  def parse(source: CharSequence) =
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   107
    parse_body(source) match {
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   108
      case List(result) => result
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   109
      case Nil => XML.Text("")
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   110
      case _ => err("multiple results")
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   111
    }
27960
65b10d8ef0c6 added parse_failsafe;
wenzelm
parents: 27946
diff changeset
   112
65b10d8ef0c6 added parse_failsafe;
wenzelm
parents: 27946
diff changeset
   113
  def parse_failsafe(source: CharSequence) = {
65b10d8ef0c6 added parse_failsafe;
wenzelm
parents: 27946
diff changeset
   114
    try { parse(source) }
65b10d8ef0c6 added parse_failsafe;
wenzelm
parents: 27946
diff changeset
   115
    catch {
27993
6dd90ef9f927 simplified exceptions: use plain error function / RuntimeException;
wenzelm
parents: 27971
diff changeset
   116
      case _: RuntimeException => XML.Elem (Markup.BAD, Nil,
27960
65b10d8ef0c6 added parse_failsafe;
wenzelm
parents: 27946
diff changeset
   117
        List(XML.Text(source.toString.replace(X_string, "\\<^X>").replace(Y_string, "\\<^Y>"))))
65b10d8ef0c6 added parse_failsafe;
wenzelm
parents: 27946
diff changeset
   118
    }
65b10d8ef0c6 added parse_failsafe;
wenzelm
parents: 27946
diff changeset
   119
  }
65b10d8ef0c6 added parse_failsafe;
wenzelm
parents: 27946
diff changeset
   120
27930
2b44df907cc2 Efficient text representation of XML trees.
wenzelm
parents:
diff changeset
   121
}