- From: William M. Perry <wmperry@aventail.com>
- Date: Wed, 11 Mar 1998 16:08:01 -0500 (EST)
- To: Bert Bos <Bert.Bos@sophia.inria.fr>
- Cc: <www-style@w3.org>
Bert Bos <Bert.Bos@sophia.inria.fr> writes: > Frank Boumphrey writes: > > >>Only the "advanced" users are going to be inventing DTDs for their XML yet > > this has been accepted by the W3C. For the more complicated features, the > > less advanced users will simply use authoring tools which write the regexp > > for them.<< > > > > This discussion has centered on the user, but remember the developer. > > One of the nice things in the XML spec was that it was stated "It > > shall be easy to write programs for XML" > > > > If regexp becomes part of the CSS specification, then a whole new > > layer of complexity is added to any program that is written for the > > standard, and automatically it means that only those companies that > > have the resources can support them. That means that prices go up > > etc. etc. > > On a related note, a programmer here noted that regexps could become > quite costly in execution time as well. For a very simple document (3 > paragraphs, 6 list items, 2 classes, and the sample style sheet from the > spec + 2 more rules), there were already some 70,000 comparisons involved > in matching selectors against elements. At the moment they are all > hashcodes (i.e., comparing one integer against another), so they are > fast, but you can imagine what will happen if some of the selectors > contain regexps. Regular expressions are not that expensive if you precompile them and just keep the regex_t hanging around instead of the much less efficient way of just recompiling the regexp every time. This is similar to what the perl 'study' command does for regexps I believe. And like someone pointed out before, regular expressions have been around for years, and highly efficient (and free :) implementations exist for just about any platform. I personally use regexps on 15 different platforms (mostly unix, but occasionally crosses over to NT/95 just fine) in my own programs. -Bill P.
Received on Wednesday, 11 March 1998 16:28:38 UTC