[grammarians] Comparison to RubyFront grammar

[Terence Parr]

On Oct 24, 2006, at 2:52 PM, Xue Yong Zhi wrote:

>
>> Nick Sieger and I (well, mostly Nick) were looking at the XRuby  
>> grammar at RubyConf. We generated a Java parser and compared it  
>> with our current YACC (Jay) based version. It wasn't much faster,  
>> unfortunately, but it would be much easier to work with and use  
>> for other projects. In addition, it avoids our hand-written  
>> (ported from C) lexer, which is a nasty bit of code. We did not do  
>> anything to confirm it was parsing correctly, however, so that  
>> would probably be the next step.
>
> Antlr is not famous for its speed:) In fact, quite a few folks from  
> yacc backgound are disappointed by its speed. Antlr v3 seems to be  
> faster.

Yep, v3 is MUCH faster.  note that due to java's lack of static data  
init, start up cost is very high for java lexers, but then they race.

>> Nick made a few minor changes to the XRuby grammar and got it to  
>> work with AntlrWorks, which is Antlr v3-based. I do not believe he  
>> tried to generate a parser with v3.
>
> I would like to do a v3 parser, but won't start myself unless v3 is  
> out of beta and has documentation ready. Terence(ANTLR's dictator)  
> is very quiet since school started, so it does not look like he is  
> going to finish something soon. If everyone else want to do it I  
> would be very happy to help.

I'm now getting back to bug fixes but must work on book at same time.

Ter