[antlr-interest] ANTLR 3 & hidden token management?

Sean Walton swalton at cs.utah.edu
Mon Aug 7 14:19:56 PDT 2006


Loring Craymer wrote:
> Kay--
>
> I tend to think that it is "early" rather than "late" in the design 
> process.  This is the first point at which people can comment from 
> experience with a working version of ANTLR 3.  ANTLR 3 is not yet 
> implemented in ANTLR 3; I consider that essential for a "mature" 
> implementation, and consequently see 3.1 as the first "real" release.
Wow.  Bootstrapping is an old concept.  Old and out of date.  I would 
focus on features and defects.
>   For that matter, the parsing engine has not yet been fully shaken 
> down (it's getting close).  Ter keeps making adjustments--mostly for 
> the sake of efficiency at this point--in response to user-detected 
> issues.  There are good reasons for beta testing!
Here is the foundational difference between the industry and academics.  
This is far from "beta" stability.  I regularly crash the thing, and 
functionality is less than consistent and stable.  This is what the 
industry calls "alpha-code".  It's "beta" when most of the major defects 
are fixed and most of the phased functionality is in place.  "Phased 
functionality" is a term that describes the incremental 
releases/deployments which include functional features a piece at a 
time.  This helps users/testers evaluate the sections and integration of 
the product.

After working with ANTLR, I was very surprised that this version was 
given a major version number (without even an 'a' or 'b'!), and "early 
release" is simply not descriptive enough.  (How would the Java 
community react if 1.6 crashed all the time! How would Microsoft 
react??).  I know that major version numbers indicate a complete change 
or overhaul to the tool, but it also indicates a certain level of stability.
> ...I expect to have a version out this week; I am pretty much down to 
> cleanup work with only a little more testing needed.
>
> --Loring
I like 3.0, because it is cleaner.  There are some really big show 
stoppers in it right now, but 1) I have committed too much time and 
effort to reverse gears now, 2) the promise of getting these show 
stoppers fixed is too attractive, and 3) using 3.0 for my work will get 
more attention than using 2.x.

-Sean

PS. Beating up a product because you have a "yet to come" bigger and 
better product -- especially when it appears that you have been in the 
ANTLR inner circle is bad form.  It also doesn't help motivate current 
testing.


More information about the antlr-interest mailing list