- From: Daniel Glazman <daniel.glazman@libertysurf.fr>
- Date: Sun, 09 Apr 2000 19:06:15 +0200
- To: "Peter S. Linss" <peter@linss.com>
- CC: "L. David Baron" <dbaron@fas.harvard.edu>, www-style@w3.org
"Peter S. Linss" a écrit : > Actually a similar issue also exists in element selector parsing with > namespaces. Once you have parsed the first token, you don't know if it is an > element selector or a namespace prefix until you look at the next token (ie. > is a '|' following?). CSS1 & CSS2 generally don't require this kind of > look-ahead (although it does help some implementations in some places). > ... > Yes and no. It requires a one token look ahead during parsing which is > something new and is also something that has been historically avoided (I > believe to make it easier to use generated parsers to parse CSS). However, > the reality is that no major CSS parser implementation actually uses a > generated parser, so this functionality is trivial to add. The working group > has discussed this issue, and when I last heard about it, it was decided to > accept the parsing issue because it allowed a more logical "human readable" > namespace syntax (and it looks more like namespaces in XML). > ... CSS 1 and 2 were LL(1) (see spec). Adding namespaces turns it into a "LL(1) most of the time, and locally LL(2)". Implementing namespaces in CSSize (my STTS implem in Java) took me very little time, thanks to JavaCC which enables to modify the look-ahead in two seconds :-) </Daniel>
Received on Sunday, 9 April 2000 13:07:56 UTC