[antlr-interest] Antlr v4 - C++ target -- idea for Terrence ...

[Terence Parr]

On Jan 11, 2012, at 12:27 AM, Ruslan Zasukhin wrote:
> To Terrence ...
> 
> I see you make big progress in challenge of left-recursions,
> And before that in automates
> 
> I want to underline again, that C 3.4 was not faster actually of v2 C++
> 
> And ANTLR 3.4 still is slower for us than parser of SQLite - Lemon
> 
> I very recommend you make benches under profiles some your ANTLR grammar and
> its analog of YACC and that Lemon.
> 
> I believe (relative) slowness of ANTLR comes from the fact of long chains of
> function calls ... Yes this is strong side for debugging -- human style,,,
> But weak side for RELEASE and speed.
> 
> My point is -- may be you can invent some way to produce two branches of
> code 
> A) for debugging  - funcs style
> B) for release -- automate style

My focus for Honey Badger is not speed. It's ease of use.  Most people are fine with speed even if not optimal.  Most compiler writers don't use parser generators so I'm less worried about speed.

Ter