On 2/21/08, David Gerard dgerard@gmail.com wrote:
- Wikitext is literally defined as "whatever the present software
does." This is bad.
- unless whatever the present software does is wrong.
- There have been several attempts to write a grammar. The latest one
is looking promising for completeness (though ANTLR is slow and buggy).
- If I said that, I think I was wrong. ANTLRworks is slow and buggy. I think ANTLR itself is probably ok, particularly once the grammar itself is stabilised. It would be good to have something to benchmark against though - does mediawiki report how long it takes to *parse* a page?
- A replacement grammar can be used for third-party implementations
(WYSIWYG, XML, etc) with perfect fidelity.
Right.
- Any replacement grammar will only replace the present
implementation if it (a) covers present behaviour sufficiently (b) is fast enough.
Grammars don't replace parsers. Parsers replace parsers. There's a big gap between what I've done so far (write a grammar) and what is needed (write a parser and XHTML generator).
Current status of ANTLR-based parser: somewhere between promising vapourware and unreleased early alpha.
If you mean "parser" then definitely vapourware. If you mean "grammar" then, yes, early alpha.
Steve