[antlr-interest] Support to Ter, questions

[Christian Boitet]

Hi form Grenoble,			2/7/07

To those "complaining" about this and that concerning ANTLR-3.0:
----------------------
- ANTLR-3 is certainly an "ongoing" progress on ANTLR-2.

- We should THANK Terence and HELP him continue 
this HUGE and FREE work. Maybe some knowledgeable 
users could help him with the doc, with examples, 
etc.

- Terence says and writes clearly that ANTLR-3.0 
is not yet as complete as ANTLR-2. That is why 
the version number ends in ".0". A huge work 
takes time.

- We still may use the previous versions while 
this new one is not yet complete as far as target 
languages are concerned.

   Our case: Jean-Claude Durand and I have been 
building C++ compilers with ANTLR and previously 
C compilers with PCCTS for some time: knowing 
ANTLR-3.0 cannot yet generate C++, we simply 
continue to use ANTLR-2.

I am afraid we did not contribute anything 
concrete yet beyond (what we think are) 1 or 2 
bug reports. We then hope Terence will take the 
above as some kind of (psychological) 
contribution or at least support! His 20-year 
dedication to that project is truly remarkable.



Questions (or remarks?):
---------
To recap what we suspect are bugs in ANTRL-2:

- using the "à" character (à) anywhere in 
a xxx.g file, even if only in a comment, causes 
ANTLR to abort.

- when trying to call a grammar not by its axiom, 
but by another non-terminal, for example to do 
some incremental parsing, we noticed an 
unexpected behavior. I guess this is because the 
FOLLOW sets are computed only with the first 
non-terminal as axiom. For example, with s: u+ 
END v+ END ; u:AA; v:BB; a call to u on input AA 
(ended by EOF) did not work because ANTLR expects 
to see AA or END after a u string. There may 
exist workarounds but we did not figure them out 
and went for another solution, namely write build 
a compier for each "increment".

That leads to the following question: is there a 
way in ANTLR-2 to include rules from files?

- #include is used by the C or C++ preprocessor, 
but apparently not by the ANTLR preprocessor.
- What we want is to build several compilers (for 
reengineering a variety of existing specialized 
programming languages sharing many parts of their 
syntax), and not to build one huge compiler with 
many options, and smaller compilers for 
increments.
- For instance, we have a compiler for a whole 
tree rewriting system, containing a block of 
procedures, another one for a "control graph", 
and one for a set of rules. An increment may be a 
whole such block, or a rule, or various parts of 
a rule (nome+parameters, lhs, rhs).

Best regards,

Xan


At 10:22 -0600 1/07/07, David Piepgrass wrote:
>>Getting started with ANTLR v3 is a frustrating experience for its lack of
>>coherent documentation and non-trivial example code.
>>
>>I've been using ANTLR since the good old times when it used to be called
>>PCCTS, and I always used to be fond of it, but I am disappointed with v3.
>Š (mind you, there is a lot of useful
>information despite these problems.)
>Š

-- 
-------------------------------------------------------------------------
Christian Boitet
(Pr. Universite' Joseph Fourier)
       ======= Notez svp /Please note ======= GETA --> GETALP ============
Groupe d'Etude pour la Traduction Automatique
                  et le Traitement Automatisé des Langues et de la Parole
G        E             T          A              L                P
GETALP, LIG, IMAG-campus, BP 53        (ex: GETA, CLIPS, IMAG-campus)        
Tel: +33 (0)4 76 51 43 55 / 51 48 17   Fax: +33 (0)4 76 44 66 75 / 51 44 05
385, rue de la Bibliothe`que           Mel: Christian.Boitet at imag.fr    
38041 Grenoble Cedex 9, France