- From: Tab Atkins Jr. <jackalmage@gmail.com>
- Date: Thu, 12 Apr 2012 21:47:13 -0700
- To: Dirk Schulze <dschulze@adobe.com>
- Cc: Boris Zbarsky <bzbarsky@mit.edu>, "www-style@w3.org" <www-style@w3.org>
On Thu, Apr 12, 2012 at 8:39 PM, Dirk Schulze <dschulze@adobe.com> wrote: > I just want to make sure that specifying the parsing doesn't make current CSS parsers inflexible. It's exactly as flexible as a grammar-based parser. (And easier to reason about, imo.) > For instance the CSS parser in WebKit can easily be modified to support the SVG parsing rules on transform lists for CSS Transforms [1] on SVG presentation attributes. This includes whitespaces between function names and the first opening brace, comma or whitspace separation of arguments or functions and so on. I wouldn't be happy to write the parser for transform functions twice. (Nearly) all browsers engines support scientific notations for numbers on SVG presentation attributes and like I wrote before: I bet no one implemented the CSS parser a second time just for SVG presentation attributes! I didn't know about the whitespace-in-function-token bit. That's an easy change to make. As I said earlier, I'm going to add the ability to parse scientific-notation numbers. Separation of function arguments is a matter for the value's own grammar to cover. The grammar of individual properties and values is not the concern of the Syntax module. > The last part is the quirks mode. I am sure that no one wants to maintain a second parser for quirks mode. Even if just some properties differ. Therefore css3-syntax should be aware of all this. I already stated that I'll be integrating the necessary quirks. It looks like it's just two - the hashless hex color and the unitless length. I'll need to do some extra work, though - I think the hashless hex color algorithm is incorrect (it wouldn't convert "a00000" into a hash token). ~TJ
Received on Friday, 13 April 2012 04:48:01 UTC