[antlr-interest] open-ended speed question: order of magnitude comparison?

John D. Mitchell johnm-antlr at non.net
Mon Nov 24 20:04:26 PST 2003


>>>>> "Tripp" == Tripp Lilley <tlilley at perspex.com> writes:
[...]

> 	Does anyone have a rough, order of magnitude, back of the envelope,
> rule of thumb, etc. idea about the (runtime, lexing and parsing and
> what-not) speed difference between a parser generated with ANTLR, with
> one of the LALR(1) tools, or by hand (by someone competent but not
> necessarily wizardly)?

Nope.  The language, algorithmic, and then engineering particulars
dominate.  That's the whole point of my mini-rant -- transform the
ugly/ambiguous/etc. cruft into something that you can rock and roll on.

[...]
> 	- As John's post suggests, some languages optimized for easy
> parsing would be... parsed easily :-) which also implies that writing a
> "more" efficient parser would be eas{y,ier}.

Note that the density improvements can be synergistic (down to a base which
is related to the data compressibility limit on the true information that
needs to be communicated) and, IME, there are some large step functions in
there (see my next message :-).

[...]
> Finally, I'm hoping to be able to use the generated parsers to do
> 'realtime' (for some values of "realtime") recognition of instances of
> the languages.

Without control of the language and context, I can't see how you can ever,
a priori, do better than "best effort" (aka "soft-RT").

Hope this helps,
		John

 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/ 




More information about the antlr-interest mailing list