2010-08-09 Eric Seidel <eric@webkit.org>
Reviewed by Adam Barth.
Use the HTML5 TreeBuilder for Fragment Parsing
https://bugs.webkit.org/show_bug.cgi?id=43586
* html5lib/runner-expected.txt:
- We pass 2 more tests now that the tokenizer is set to
the right state before fragment parsing.
* html5lib/runner-expected-html5.txt:
- We pass 4 more tests now that HTML5 fragment parsing is
(partially) implemented behind the --html5-treebuilder flag.
2010-08-09 Eric Seidel <eric@webkit.org>
Reviewed by Adam Barth.
Use the HTML5 TreeBuilder for Fragment Parsing
https://bugs.webkit.org/show_bug.cgi?id=43586
The bulk of this change is just a transcription of the HTML5 spec:
http://www.whatwg.org/specs/web-apps/current-work/multipage/the-end.html#fragment-case
DocumentParser() now sets the right tokenizer state before handling
fragments. Since default code path uses the HTML5 tokenizer with the
legacy tree builder, this fixes 2 tokenizer-state tests for the html5lib.
This fixes 2 more tree-builder tests when the html5 treebuilder is enabled too.
* dom/DocumentFragment.cpp:
(WebCore::DocumentFragment::parseHTML):
- HTML5 implementation of fragment parsing requires the context element (parent)
pointer. The legacy implementation handles all parsing branches which require
the context element outside of the parser itself, which likely produces "wrong"
results. We'll remove those branches when we turn on the HMTL5 fragment parser.
(WebCore::DocumentFragment::parseXML):
- Rename "parent" to "contextElement" to match the HTML path (and the HTML5 spec).
* dom/DocumentFragment.h:
* dom/Element.cpp:
(WebCore::Element::createContextualFragment):
- Pass the contextElement (this).
* html/HTMLConstructionSite.cpp:
(WebCore::HTMLConstructionSite::setForm):
- Needed by for the fragment case.
(WebCore::HTMLConstructionSite::insertHTMLBodyStartTagInBody):
- I'm not sure this method should exist, it only has one caller.
but for now, I've moved the fragment case FIXME out to the
one caller and implemented it.
* html/HTMLConstructionSite.h:
* html/HTMLDocumentParser.cpp:
(WebCore::HTMLNames::tokenizerStateForContextElement):
- Part of the fragment case handling for the HTML5 parser.
(WebCore::HTMLDocumentParser::HTMLDocumentParser):
- Set the Tokenizer state based on the contextElement.
(WebCore::HTMLDocumentParser::parseDocumentFragment):
- Take a contextElement argument.
* html/HTMLDocumentParser.h:
* html/HTMLElement.cpp:
(WebCore::HTMLElement::setInnerHTML):
- Added a FIXME before the branches the old parser used to avoid ever needing
to pass a context element when parsing HTML fragments.
(WebCore::HTMLElement::insertAdjacentHTML):
- Pass the context element.
* html/HTMLTokenizer.h:
* html/HTMLTreeBuilder.cpp:
- I added a new helper class FragmentParsingContext to simplify the two
HTMLTreeBuilder constructors and hide the existence of m_dummyDocumentForFragmentParse.
- I also removed m_isParsingFragment and replaced it with an isParsingFragment() function.
(WebCore::HTMLTreeBuilder::HTMLTreeBuilder):
(WebCore::HTMLTreeBuilder::FragmentParsingContext::FragmentParsingContext):
(WebCore::HTMLTreeBuilder::FragmentParsingContext::document):
(WebCore::HTMLTreeBuilder::FragmentParsingContext::~FragmentParsingContext):
(WebCore::HTMLTreeBuilder::passTokenToLegacyParser):
(WebCore::HTMLTreeBuilder::processStartTagForInBody):
- I had to implement two notImplemented() states in order to get most of the
layout tests to pass when using the HTML5 parser fragment case.
(WebCore::HTMLTreeBuilder::processColgroupEndTagForInColumnGroup):
(WebCore::HTMLTreeBuilder::processStartTagForInTable):
(WebCore::HTMLTreeBuilder::processStartTag):
(WebCore::HTMLTreeBuilder::resetInsertionModeAppropriately):
(WebCore::HTMLTreeBuilder::processEndTagForInTableBody):
(WebCore::HTMLTreeBuilder::processEndTagForInRow):
(WebCore::HTMLTreeBuilder::processEndTagForInCell):
(WebCore::HTMLTreeBuilder::processCaptionEndTagForInCaption):
(WebCore::HTMLTreeBuilder::processTrEndTagForInRow):
(WebCore::HTMLTreeBuilder::processTableEndTagForInTable):
(WebCore::HTMLTreeBuilder::processEndTag):
(WebCore::HTMLTreeBuilder::processCharacterBuffer):
(WebCore::HTMLTreeBuilder::processEndOfFile):
(WebCore::HTMLTreeBuilder::finished):
- The HTML5 parsing algorithm fragment case uses a dummy document to handle all the parsing
and then moves all the children from the dummy document over into a DocumentFragment
when the parse is done. We could avoid this adopt step by making the TreeBuilder a bit
more complicated, but I'm not sure that's necessary (yet).
* html/HTMLTreeBuilder.h:
(WebCore::HTMLTreeBuilder::isParsingFragment):
(WebCore::HTMLTreeBuilder::FragmentParsingContext::fragment):
(WebCore::HTMLTreeBuilder::FragmentParsingContext::contextElement):
(WebCore::HTMLTreeBuilder::FragmentParsingContext::scriptingPermission):
* xml/XSLTProcessor.cpp:
(WebCore::createFragmentFromSource):
git-svn-id: http://svn.webkit.org/repository/webkit/trunk@65006 268f45cc-cd09-0410-ab3c-d52691b4dbfc
18 files changed
