Xml Parsing with large embedded SVG and Data URI can be very slow

[Nanonid]

I could really use this performance fix. :-)

I don't have a file I can share just as yet, but this diagram.xml file has a large embedded SVG and images encoded as URI.

Unfortunately, those large strings are attributes and not text elements.

I realize I should probably change the XML model, but I can't anytime soon.

Diagram: 2420642 diagram_export_1.xml

void main(){
  group( "diagram", (){
    test( "parse visit", (){
      String xml = readFileTextSync("diagram_export_1.xml");
      Stopwatch stopwatch = new Stopwatch()..start();
      XmlDocument doc = parse(xml);
      print( "parse time ${stopwatch.elapsed}");
      stopwatch.reset();
      DiaTransformer xform = new DiaTransformer();
      DiaDoc dia = xform.visitDocument(doc);
      print( "visit ${stopwatch.elapsed}");
    });
  });
}

unittest-suite-wait-for-done
parse time 0:00:20.558550
visit 0:00:00.010319
PASS: diagram parse visit

All 1 tests passed.
unittest-suite-success

[Nanonid]

_JSSyntaxRegExp._ExecuteMatch 98% CPU time

[Nanonid]

Reduced file to just large icons. And left in a tag by mistake.
It still took 20 seconds for this error.
Expected </dia:Model>, but found </Icons>

[Nanonid]

Hacking the parser

[Nanonid]

Thought maybe quote scanning was an issue. So I hacked in a probably not correct anyExcept scanner. Still 20 seconds.

/**
 * Returns a parser that accepts any input element except element.
 *
 * For example, `anyExcept()` succeeds and consumes any given letter
 * except element. It only fails for an empty input.
 * Equivalent to any().starLazy(char(element.codeUnitAt(0))
 */
Parser anyExcept(element,[String message = 'input expected']) {
  return new AnyExceptParser(element,message);
}

class AnyExceptParser extends Parser {

  final String _message;
  final String _element;
  final int _code;

  AnyExceptParser(element, this._message) :
    _element = element,
    _code = element.codeUnitAt(0);

  @override
  Result parseOn(Context context) {
    var start = context.position;
    var buffer = context.buffer;
    if( start >= buffer.length ){
      return context.failure(_message);
    }
    int len = buffer.length;
    int stop = start;
    while( stop < len && !identical(buffer[stop],_element) ) {
      stop++;
    }
    var result = context.buffer is String
        ? context.buffer.substring(start, stop)
        : context.buffer.sublist(start, stop);
    return context.success(result, stop);
  }

  @override
  Parser copy() => new AnyExceptParser(_element,_message);

  @override
  bool equalProperties(AnyExceptParser other) {
    return super.equalProperties(other)
        && _code == other._code
        && _message == other._message;
  }
}

[Nanonid]

It is _decodeXml!

      .seq(any().starLazy(char(DOUBLE_QUOTE)).flatten())
//      .seq(any().starLazy(char(DOUBLE_QUOTE)).flatten().map(_decodeXml))

unittest-suite-wait-for-done
parse time 0:00:00.363150
PASS: diagram parse visit

[Nanonid]

Sent a pull request which ~fixes this issue.
If you want to continue improving performance and usability, I'll let the bounty stand.
If you care.

[Nanonid]

Final results after new decode.

unittest-suite-wait-for-done
parse time 0:00:00.591667
visit 0:00:00.009809
PASS: diagram parse visit

All 1 tests passed.
unittest-suite-success

[Nanonid]

Working for now.