Type 'ParseErrorToken' is not a subtype of type 'StartTagToken' in type cast

[ditman]

When attempting to access attribute values that contain an & (for example, in a src URL), the attributeValueSpans getter throws the following exception:

type 'ParseErrorToken' is not a subtype of type 'StartTagToken' in type cast
package:html/dom.dart 287:37          Node._ensureAttributeSpans
package:html/dom.dart 191:5           Node.attributeValueSpans
test/parser_feature_test.dart 499:21  _testElementSpans.<fn>.<fn>

I wrote a small parser_feature_test that reproduces the issue:

test('attribute values can contain ampersand', () {
  final expectedUrl = 'foo?key=value&key2=value2';
  final text = '<script src="$expectedUrl">';

  final doc = parse(text, generateSpans: true);
  final elem = doc.querySelector('script')!;

  expect(elem.attributeValueSpans!['src'], isA<FileSpan>()); // Throws
});

The effect is that I can't parse code like:

<script src="foo?key=value&key2=value2">

[ditman]

The problem seems to be that the tokenizer sees the &, thinks that it's going to be a HTML entity, or something special, and when it isn't, it emits a ParseErrorToken and keeps going.

• tools/pkgs/html/lib/src/tokenizer.dart

  Lines 342 to 346 in 9c53358

  	} else {
  	_addToken(ParseErrorToken('expected-named-entity'));
  	stream.unget(charStack.removeLast());
  	output = '&${charStack.join()}';
  	}

This ParseErrorToken is unexpected by _ensureAttributeSpans, that assumes that the first token it's going to see is a "StartTagToken":

• tools/pkgs/html/lib/dom.dart

  Lines 286 to 287 in 9c53358

  	tokenizer.moveNext();
  	final token = tokenizer.current as StartTagToken;

[devoncarew]

fixed by #2036