Interest for Dylan-generating ANTLR ? 
Author Message
 Interest for Dylan-generating ANTLR ?

This message was originally sent through Usenet but does not appear
to have ever made it outside. So here is a resent through
dylan-info. Apologies if you happen to see it twice.

 Hi,

 For a long time I have wanted to get decent language tools
for Dylan but I always managed to do without.

 I have used the parsergen tool from the Gwydion distribution and
a manually written parser for some property-list like parsing
(quite fast with HD 1.2 on a 233Mhz Laptop: ~15ms for a 10k file,
 ~18s on a 5MB file for lexing/parsing/objects creation).
I had to hand-tune both the lexer and parser heavily to get at this
level of performance (and I would not dare to generate the parser
again).

 Now I would like to have a more powerful lexer/parser construction
tool. I also always wanted to look at ANTLR (formerly PCCTS).
What better way than to get ANTLR to generate Dylan code ?

 I downloaded ANTLR (and its source) last week (from www.antlr.org)
and started to see how I could get it to generate Dylan.

 While not a difficult task, it will still take me some time to get
anything working and I do not have that much spare time. Therefore
I would like to probe the Dylan community to see how much interest
there would be for such a tool.

 My questions are:
1) Are you interested in lexer/parser generation tools for Dylan ?

2) Are you interested in ANTLR or would you prefer to get another
tool (like JavaCC ?). Bear in mind that I can only help for
open-source tools.

3) ANTLR 2.5.0 is written in Java. Would you prefer:
   a) to get a Java tool that generates Dylan and can be
      updated as ANTLR evolves ?
   b) to get a Dylan tool that is a functionnal equivalent
      of ANTLR and evolves along its own path ?
 Note that a) is much much easier than b) in the short term.

4) Would you be willing to help in coding/testing such a program ?

 Note that I have only worked 4/5 hours on the ANTLR source at this
point and that a first version of a Dylan-generating version is
probably weeks if not months away (depending on my professional
engagements).

 I look forward to your feedback,

 Eric

Note: ANTLR 2.6.0 is just around the corner, so it will probably
be the basis for any Dylan work.

--

New Technologies Consultant                        Tel:   01.47.62.83.19
Oracle France                                      Fax:   01.47.62.28.49



Mon, 10 Sep 2001 03:00:00 GMT  
 Interest for Dylan-generating ANTLR ?

Quote:

> Now I would like to have a more powerful lexer/parser construction
>tool. I also always wanted to look at ANTLR (formerly PCCTS).
>What better way than to get ANTLR to generate Dylan code ?

I've been involved with PCCTS/ANTLR since 1.06.  Nice package.

Quote:
> I downloaded ANTLR (and its source) last week (from www.antlr.org)
>and started to see how I could get it to generate Dylan.
> While not a difficult task,

APPEARANCES CAN BE VERY DECEIVING HERE!  Watch it.  :-)

Quote:
> My questions are:
>1) Are you interested in lexer/parser generation tools for Dylan ?

Yes.

Quote:
>2) Are you interested in ANTLR or would you prefer to get another
>tool (like JavaCC ?). Bear in mind that I can only help for
>open-source tools.

ANTLR kicks butt.  :-)

Quote:
>3) ANTLR 2.5.0 is written in Java. Would you prefer:
>   a) to get a Java tool that generates Dylan and can be
>      updated as ANTLR evolves ?
>   b) to get a Dylan tool that is a functionnal equivalent
>      of ANTLR and evolves along its own path ?
> Note that a) is much much easier than b) in the short term.

I'd say a) in the short term and then use it to make b) in the long term.
(ANTLR is a bit slow right now IMO.  A fully-optimized Dylan version would
probably be faster.)

Quote:
>4) Would you be willing to help in coding/testing such a program ?

Yes to both.  It'd give me a good project to learn the ins and outs of
Dylan.  (I'm a Dylan newbie.)  Initially I'd only be good at testing, but as
I read and understood your Dylan code, I could probably start to help on
coding as well.


Mon, 10 Sep 2001 03:00:00 GMT  
 Interest for Dylan-generating ANTLR ?

Quote:

>  While not a difficult task, it will still take me some time to get
> anything working and I do not have that much spare time. Therefore I
> would like to probe the Dylan community to see how much interest there
> would be for such a tool.

If you want to work on ANTLR, we can provide resources for you--CVS, bug
tracking and even a mailing list. Of course, the ANTLR team may also be
willing to do this for you.

Quote:
> 2) Are you interested in ANTLR or would you prefer to get another tool
> (like JavaCC ?). Bear in mind that I can only help for open-source tools.

We're interested in as much Dylan tool support as possible. ;-) I won't be
using ANTLR myself (parsergen is adequate for my tasks), but somebody else
may be interested in a good compiler-generation tool.

Quote:
> 3) ANTLR 2.5.0 is written in Java. Would you prefer:
>    a) to get a Java tool that generates Dylan and can be
>       updated as ANTLR evolves ?

As much as we like Dylan, this is probably less work in both the short run
and the long run. Also, if you have any parsergen enhancements, we're
interested in folding them back into the standard version.

Cheers,
Eric
(Gwydion volunteer)



Mon, 10 Sep 2001 03:00:00 GMT  
 Interest for Dylan-generating ANTLR ?

Quote:
> -----Original Message-----


> Sent: Thursday, March 25, 1999 7:25 AM

> Subject: Interest for Dylan-generating ANTLR ?

[...]

Quote:
>  While not a difficult task, it will still take me some time to get
> anything working and I do not have that much spare time. Therefore
> I would like to probe the Dylan community to see how much interest
> there would be for such a tool.

>  My questions are:
> 1) Are you interested in lexer/parser generation tools for Dylan ?

Oh, yes! :-)

Quote:
> 2) Are you interested in ANTLR or would you prefer to get another
> tool (like JavaCC ?). Bear in mind that I can only help for
> open-source tools.

ANTLR is an excellent choice. JavaCC actually started as a spin-off of ANTLR
targeted at Java code generation (when ANTLR still did C/C++),  better
Unicode support, etc. I have used both, and while I consider JavaCC to have
a slight edge in usability/flexability, it is *not* open-source. :-/

Quote:
> 3) ANTLR 2.5.0 is written in Java. Would you prefer:
>    a) to get a Java tool that generates Dylan and can be
>       updated as ANTLR evolves ?
>    b) to get a Dylan tool that is a functionnal equivalent
>       of ANTLR and evolves along its own path ?
>  Note that a) is much much easier than b) in the short term.

My own preference would be b -- but either would be a significant
contribution to the Dylan community.

Quote:
> 4) Would you be willing to help in coding/testing such a program ?

Ah, so many worthey projects! Sigh. Realistically all I'd be able to offer
is feedback. T. Parr is a friend of a friend, but he's also pretty
responsive to posts to the PCCTS newsgroups and generally a helpful guy...

Perhaps we could revive the suggestion I made to him some years ago to
pursue making PCCTS language-neutral. (At the time he was quite frustrated
with C/C++ and considering inventing his own language to get around
problems -- this was before Java hit the scene.) Although he's quite the
Java disciple nowdays...

Quote:
>  Note that I have only worked 4/5 hours on the ANTLR source at this
> point and that a first version of a Dylan-generating version is
> probably weeks if not months away (depending on my professional
> engagements).

>  I look forward to your feedback,

>  Eric

> Note: ANTLR 2.6.0 is just around the corner, so it will probably
> be the basis for any Dylan work.

Great!


Mon, 10 Sep 2001 03:00:00 GMT  
 Interest for Dylan-generating ANTLR ?
...

Quote:

>  My questions are:
> 1) Are you interested in lexer/parser generation tools for Dylan ?

   Yes.  Why after all,  Dylan-lint needs to get built to solve the
   "style" problem.  :-)

Quote:
> 2) Are you interested in ANTLR or would you prefer to get another
> tool (like JavaCC ?). Bear in mind that I can only help for
> open-source tools.

  I've been considering ANTLR, which would be even nicer if it generated
Dylan
  code.  Parsegen, if I recall correctly, is LALR. It would be nice to
  have the "alternative", LL,  around also.

Quote:
> 3) ANTLR 2.5.0 is written in Java. Would you prefer:
>    a) to get a Java tool that generates Dylan and can be
>       updated as ANTLR evolves ?
>    b) to get a Dylan tool that is a functionnal equivalent
>       of ANTLR and evolves along its own path ?
>  Note that a) is much much easier than b) in the short term.

   As the initial step  a).

   besides  a) could be an interium step in creating b).  
   Aren't there some support libraries you have to construct
   that the generated lexer/parsers take advantage of?  That should
   be plenty in addition to getting the correct  classes, methods, and
   modules/library definition files generated.

   [ Besides it probably would be good "reality check" to generate
     Dylan, java, and C++ parsers and compare. Using the same
     "generator" would make it pretty close to a"apples to apples"
     comparison; barring quirks introduced by the different object
     models. ]

   I think I remember a posting to the "pccts" newsgroup about
   adding support for another language means dealing with a particularly
   hairy code section of code.  It wasn't very modularly.

Quote:
> 4) Would you be willing to help in coding/testing such a program ?

    coding? too many projects now...
    testing?  probably.

Quote:
> Note: ANTLR 2.6.0 is just around the corner, so it will probably
> be the basis for any Dylan work.

     Cool, for some reason I get the feeling that lexer as a token stream
     producer might make macro expansion easier. However, that may just
     be lack of sleep. ;-)

--

Lyman



Tue, 11 Sep 2001 03:00:00 GMT  
 Interest for Dylan-generating ANTLR ?

Quote:


> > [...]
> I've been involved with PCCTS/ANTLR since 1.06.  Nice package.

 Nice package indeed. I do not like the syntax that much, but it
is a minor annoyance.

Quote:
> > While not a difficult task,
> APPEARANCES CAN BE VERY DECEIVING HERE!  Watch it.  :-)

 IMHO the difficult part was to come up with the pseudo-LL(k)
theory and to write the first implementation. While certainly
a worthy goal, I do not consider a language port to be a major
achievement.

Quote:
> >1) Are you interested in lexer/parser generation tools for Dylan ?
> Yes.

 Good !

Quote:
> >   a) to get a Java tool that generates Dylan and can be
> >      updated as ANTLR evolves ?
> >   b) to get a Dylan tool that is a functionnal equivalent
> >      of ANTLR and evolves along its own path ?
> [...]
> I'd say a) in the short term and then use it to make b) in the long term.
> (ANTLR is a bit slow right now IMO.  A fully-optimized Dylan version would
> probably be faster.)

 Well plan A seems to be the only reasonnable one at this point.

Quote:
> >4) Would you be willing to help in coding/testing such a program ?

> Yes to both.  It'd give me a good project to learn the ins and outs of
> Dylan.  (I'm a Dylan newbie.)  Initially I'd only be good at testing, but as
> I read and understood your Dylan code, I could probably start to help on
> coding as well.

 I plan to work "alone" for a few weeks and then try to
organize some collaborative effort. My first post was really a trial
balloon to gauge the community interest.

 Thanks for the feedback.

 Cheers,
   Eric

--

New Technologies Consultant                        Tel:   01.47.62.83.19
Oracle France                                      Fax:   01.47.62.28.49



Tue, 11 Sep 2001 03:00:00 GMT  
 Interest for Dylan-generating ANTLR ?

Quote:


> > [...]
> I've been involved with PCCTS/ANTLR since 1.06.  Nice package.

 Nice package indeed. I do not like the syntax that much, but it
is a minor annoyance.

Quote:
> > While not a difficult task,
> APPEARANCES CAN BE VERY DECEIVING HERE!  Watch it.  :-)

 IMHO the difficult part was to come up with the pseudo-LL(k)
theory and to write the first implementation. While certainly
a worthy goal, I do not consider a language port to be a major
achievement.

Quote:
> >1) Are you interested in lexer/parser generation tools for Dylan ?
> Yes.

 Good !

Quote:
> >   a) to get a Java tool that generates Dylan and can be
> >      updated as ANTLR evolves ?
> >   b) to get a Dylan tool that is a functionnal equivalent
> >      of ANTLR and evolves along its own path ?
> [...]
> I'd say a) in the short term and then use it to make b) in the long term.
> (ANTLR is a bit slow right now IMO.  A fully-optimized Dylan version would
> probably be faster.)

 Well plan A seems to be the only reasonnable one at this point.

Quote:
> >4) Would you be willing to help in coding/testing such a program ?

> Yes to both.  It'd give me a good project to learn the ins and outs of
> Dylan.  (I'm a Dylan newbie.)  Initially I'd only be good at testing, but as
> I read and understood your Dylan code, I could probably start to help on
> coding as well.

 I plan to work "alone" for a few weeks and then try to
organize some collaborative effort. My first post was really a trial
balloon to gauge the community interest.

 Thanks for the feedback.

 Cheers,
   Eric

--

New Technologies Consultant                        Tel:   01.47.62.83.19
Oracle France                                      Fax:   01.47.62.28.49



Tue, 11 Sep 2001 03:00:00 GMT  
 Interest for Dylan-generating ANTLR ?

Quote:


> > ...
> > 3) ANTLR 2.5.0 is written in Java. Would you prefer:
> >    a) to get a Java tool that generates Dylan and can be
> >       updated as ANTLR evolves ?
> >    b) to get a Dylan tool that is a functionnal equivalent
> >       of ANTLR and evolves along its own path ?
> >  Note that a) is much much easier than b) in the short term.

>    As the initial step  a).

>    besides  a) could be an interium step in creating b).
>    Aren't there some support libraries you have to construct
>    that the generated lexer/parsers take advantage of?  That should
>    be plenty in addition to getting the correct  classes, methods, and
>    modules/library definition files generated.

 The Java runtime environment is made of around 40 classes. I do not
expect the Dylan environment to be much larger.

Quote:
>    I think I remember a posting to the "pccts" newsgroup about
>    adding support for another language means dealing with a particularly
>    hairy code section of code.  It wasn't very modularly.

 It could be better but it is not that bad (from my initial review).
Most of the code generation is in a single class called XXXCodeGenerator
where XXX is the generated language (XXX being Java/Cpp and soon Dylan).

Quote:
> > 4) Would you be willing to help in coding/testing such a program ?
>     coding? too many projects now...
>     testing?  probably.

 That would already be wonderful !

Quote:
> > Note: ANTLR 2.6.0 is just around the corner, so it will probably
> > be the basis for any Dylan work.

>      Cool, for some reason I get the feeling that lexer as a token stream
>      producer might make macro expansion easier. However, that may just
>      be lack of sleep. ;-)

 It might indeed, but I am not a good reference for Dylan parsing.

 Wondering if our Harlequin friends could shed some light on their lexing
/ parsing technology. Do you use any internal parser-generating tool ?
a macro-based parser ? hand-coded ?

 Thanks for the feedback anyway,

 Regards,
   Eric

--

New Technologies Consultant                        Tel:   01.47.62.83.19
Oracle France                                      Fax:   01.47.62.28.49



Tue, 11 Sep 2001 03:00:00 GMT  
 Interest for Dylan-generating ANTLR ?

Quote:


> >  While not a difficult task, it will still take me some time to get
> > anything working and I do not have that much spare time. Therefore I
> > would like to probe the Dylan community to see how much interest there
> > would be for such a tool.

> If you want to work on ANTLR, we can provide resources for you--CVS, bug
> tracking and even a mailing list. Of course, the ANTLR team may also be
> willing to do this for you.

 Thanks for the proposition. Having assessed that there is enough interest
I need to contact the ANTLR team and then do some initial work to organize
a project.

Quote:
> [...] Also, if you have any parsergen enhancements, we're
> interested in folding them back into the standard version.

 I'll see if I can fold my manual changes to the parsergen generator.

 Thanks for the feedback,

 Regards,
   Eric

--

New Technologies Consultant                        Tel:   01.47.62.83.19
Oracle France                                      Fax:   01.47.62.28.49



Tue, 11 Sep 2001 03:00:00 GMT  
 Interest for Dylan-generating ANTLR ?



Quote:
>  Wondering if our Harlequin friends could shed some light on their lexing
> / parsing technology. Do you use any internal parser-generating tool ?
> a macro-based parser ? hand-coded ?

We have a parser generator. I couldn't possibly comment on its future
availability.

__Jason



Tue, 11 Sep 2001 03:00:00 GMT  
 
 [ 10 post ] 

 Relevant Pages 

1. Sudden interest in Dylan/Dylan should be open.

2. Generating HTML with Dylan

3. Dylan Barriers (was: Sudden interest in Dylan/Dylan should be open.)

4. Announcing USQAGMS - research utility for PCCTS/Antlr

5. "Using Antlr for Ruby?" with ArgoUML

6. ANTLR rules for Ruby

7. ANTLR rules for Ruby

8. ANTLR grammar for Verilog?

9. ANTLR Ada grammar update

10. ANTLR grammar for Ada available

11. vhdl-93 parser with antlr

12. Parser/lexer generator ANTLR

 

 
Powered by phpBB® Forum Software