- From: Laura Holmes <holmes@google.com>
- Date: Fri, 3 Aug 2007 09:39:39 -0400
- To: public-mobileok-checker <public-mobileok-checker@w3.org>
- Message-ID: <135a9f560708030639p89e91aci79daef936e10c411@mail.gmail.com>
> Miguel, sorry I didn't recognize the XML version of the CSS called from > the HTTPCSSResrouce constructor. Some of it was there in the moki, but i > failed to notice it because it did not contain all of the css styles I had > written into my test document. Primarily, @ import tags. However, this small > error raises a more important question: should we continue to refine the > CSS->XML parser to get it to the point where we can use it, or bypass this > method and complete the parser at a different time? Because right now, it > doesn't function well enough to perform the necessary CSS tests. > > I agree with Sean that we don't really need a serialization of the CSS at > this point. What we need are the following: > 1) validate correct CSS markup > 2) a way to identify the resources referenced in CSS (stylesheets and > images) > 3) a way to report the lines numbers of markup errors and references > > As I see it, all of these things can be accomplished with regular > expressions. The current CSS->XML handler would have to be changed > significantly in order to support these features, and as far as resource URL > identification, we would still need to perform some regular expression > matching. However, with regular expressions, I feel like we can do all of > these things (Miguel, would you elaborate what you mean by "more > fine-grained processing"?) > > Also, this method isn't as flexible for inline style elements. It looks > like we'd be making a separate stylesheet element for each inline style > reference, which can get really messy. I would much rather have unified > methods that identify elements in all external, embedded, and inline > stylesheets. > > For the scope of our task, and that fact that this issue is holding up > development on my side, I'm encouraging that we just get these tests done > rather than focus on perfecting our own implementation of a CSS parser. We > can always add in the CSS parser as a resource for developers later. Would > everyone be ok with moving forward with regular expressions and saving the > parser for later? > > Cheers, > Laura > > On 8/2/07, Sean Owen <srowen@google.com> wrote: > > > > What about inline style attributes, and embedded stylesheets? For > > consistency wouldn't we want to handle these too? I think it gets > > quite complex. > > > > Again, I am not sure why we need to construct a parse tree for CSS at > > all. Either way we need to just run some regexes on the property > > values, so, it is only more complex to add on this tree. > > > > The argument I heard for it is, well, wouldn't it be nice to define a > > CSS serialization in XML? and we'd rejected other parsers not because > > they didn't work but because we didn't like the XML. All that's fine, > > but seems like a side project. > > > > Right now we have an uneven approach to parsing CSS and I think that > > has to be addressed. I think we either define a more comprehensive > > serialization of CSS for all CSS, not just external stylesheets, or > > else remove the serialization. Right now I see the former being a lot > > of work for no gain. Is it meeting a requirement I am not thinking of? > > > > On 8/2/07, Miguel Garcia <miguel.garcia@fundacionctic.org> wrote: > > > > > > > I'm working on including resources mentioned in css, and I've come > > > across > the CSSResourceToDom method in CSSUtils. Can the author of > > this > > > method help me answer some questions? > > > > > > > 1) Why is this method used in EmbeddedCSSResource, but not present > > in > > > > HTTPCSSResource? > > > > > > HTTPCSSResource makes a call to CSSResourceToDom in its constructor. > > At > > > least in the revision 1.10 (which is the latest in CVS). > > > > > > > 2) What are the distinct advantages of this approach vs. just using > > > the > > > > text data of the css to get the information we need for the tests? > > > > > > Basically this approach allow us a more fine-grained processing than > > > just using regular expressions. Aditionally we can use some existing > > > tools (SAC parsers) to do part of the work. Besides I think it's > > easier > > > using SAC than regular expressions. > > > > > > > Thanks! > > > > > > > Cheers, > > > > Laura > > > > > > > > > >
Received on Friday, 3 August 2007 13:40:35 UTC