[antlr-interest] Code generation advice

Matt Benson gudnabrsam at gmail.com
Wed Jun 20 13:13:19 PDT 2012 

I'm using "E" to mean "Embedded."  "Embedded DSL" seems to turn up a
lot of hits on Google, but a rose by any other name... your
query-building example is in the neighborhood, yes.

Matt

On Wed, Jun 20, 2012 at 3:03 PM, Scott Stanchfield <scott at javadude.com> wrote:
> I think I see now what you're looking for (I think you're talking about an
> internal DSL though, assuming your use of EDSL means "external DSL").
>
> xtext won't do that... it's really for parsing external DSLs and loading the
> data into a model.
>
> So if I understand what you're looking for, it could be used to define
> something like SQLiteQueryBuilder:
>
>   Cursor c = new SQLiteQueryBuilder()
>                           .setTables(tables)
>                           .appendWhere(whereClause)
>                           .query(...);
>
> using a grammar kinda like
>
> SQLiteQueryBuilder
>   : setTables appendWhere query
>   ;
>
> setTables[String[] tables]
>   : {set tables property} ;
>
> appendWhere[String where]
>   : {set where clause} ;
>
> query[...]
>   : {run query}
>   ;
>
> Is that kinda on the track you're thinking?
>
> -- Scott
>
> ----------------------------------------
> Scott Stanchfield
> http://javadude.com
>
>
> On Wed, Jun 20, 2012 at 3:44 PM, Matt Benson <gudnabrsam at gmail.com> wrote:
>>
>> Hi, Scott--and thanks for ANTLREclipse, which (never having moved past
>> v2) I still use.  ;)  Actually, embedded in my original email was the
>> fact that I have looked at xtext and making it do what I'm looking for
>> is at least an endeavor whose HOWTO is non-obvious.  AFAICT, its focus
>> is on parsing a DSL into a Java model:  well and good, but I know how
>> to do that with ANTLR.  Xtext might well make this job easier, but in
>> my case, I want more or less to bypass the idea of parsing a DSL and
>> simply use the same meta-grammar to describe an embedded DSL whose
>> rules are actually invoked as Java methods.  Possibly xtext *can* do
>> this, but so far its documentation is quite obtuse to me, and I have
>> yet to see anything in its examples that suggests that it would give
>> me a leg up on this particular task.
>>
>> @Eric:  your question is answered ;)
>>
>> Regards,
>> Matt
>>
>> On Wed, Jun 20, 2012 at 2:31 PM, Scott Stanchfield <scott at javadude.com>
>> wrote:
>> > Have you taken a look at xtext ( http://www.eclipse.org/Xtext/) It uses
>> > a
>> > grammar similar to antlr to generate model interfaces/classes, a
>> > syntax-highlighting editor (optional, but very useful), persistence
>> > manager
>> > and other doodads. Under the covers it uses antlr to do the actual
>> > parsing.
>> >
>> > xtext is intended for DSL definition and is rather mature, too.
>> >
>> > As a really short example (from the xtext site):
>> >
>> > grammar org.xtext.example.mydsl.MyDsl
>> >     with org.eclipse.xtext.common.Terminals
>> >
>> > generate myDsl "http://www.xtext.org/example/mydsl/MyDsl"
>> >
>> > Model:
>> >     greetings+=Greeting*;
>> >
>> > Greeting:
>> >     'Hello' name=ID '!';
>> >
>> > would generate a Model class that has a property greetings which is a
>> > list
>> > of Greeting instances, and a Greeting class that has a name property.
>> >
>> > I don't know if it fits exactly what you're looking for, but it's worth
>> > taking a look.
>> > -- Scott
>> >
>> > ----------------------------------------
>> > Scott Stanchfield
>> > http://javadude.com
>> >
>> >
>> >
>> > On Wed, Jun 20, 2012 at 2:49 PM, Matt Benson <gudnabrsam at gmail.com>
>> > wrote:
>> >>
>> >> What I want to do is make it easier to write an EDSL by having it
>> >> specified like any other grammar, generating Java interfaces to
>> >> represent the grammar's various rules, as well as basic
>> >> implementations of these, which I expect will accumulate "tokens"
>> >> extracted from method arguments specified in fluent fashion.  In fact,
>> >> spelling it out here makes it sound as though it might be just as
>> >> reasonable for the end result of a series of calls upon the
>> >> implementation of this EDSL interface to create a (parse?) tree, which
>> >> could then be walked to resolve the final result of such a series of
>> >> method invocations.  This would have the benefit that such a tree
>> >> walker (parse, AST, what-have-you) would be equally applicable to a
>> >> tree generated from a "normal" ANTLR-generated Java Parser
>> >> implementation.
>> >>
>> >> Is that more clear?
>> >>
>> >> Thanks for your interest,
>> >> Matt
>> >>
>> >> On Wed, Jun 20, 2012 at 1:14 PM, Eric <researcher0x00 at gmail.com> wrote:
>> >> > Ok, I'm curious as to what you are trying to do. It sounds like
>> >> > something I
>> >> > once tinkered with but since I did not understand your exact goal I
>> >> > could be
>> >> > totally wrong.
>> >> >
>> >> > The only part that I understood as part of the goal was "generation
>> >> > of
>> >> > Java
>> >> > interfaces from EBNF-style grammars". Are you trying to generate Java
>> >> > interfaces from a grammar for other languages or interface languages
>> >> > such as
>> >> > C++, CORBA, etc.?
>> >> >
>> >> > Eric
>> >> >
>> >> >
>> >> >
>> >> > On Wed, Jun 20, 2012 at 12:08 PM, Matt Benson <gudnabrsam at gmail.com>
>> >> > wrote:
>> >> >>
>> >> >> Hi all,
>> >> >>  In the Java ecosystem embedded DSLs have become increasingly
>> >> >> popular
>> >> >> over the past several years.  In my opinion the primary drawback to
>> >> >> these is the painful task of their creation.  For this reason I want
>> >> >> to experiment with the generation of Java interfaces from EBNF-style
>> >> >> grammars to see what I can accomplish, in true Dr. Terence Parr "why
>> >> >> program by hand in 5 days..." fashion.  I have explored, for
>> >> >> example,
>> >> >> xtext, but the process of going from its meta-grammar to EDSL code
>> >> >> is
>> >> >> non-obvious.  I'm open to other suggestions if anyone has any, but
>> >> >> in
>> >> >> the meantime I'm looking at ANTLR 4 and/or 3-based approaches.  It
>> >> >> would seem most appropriate to use the ANTLR 4 meta-grammar, from
>> >> >> which point I would seem to have the following options:
>> >> >>
>> >> >> * Implement a custom code generation target for ANTLR 4
>> >> >> * Use ANTLRMorph with the ANTLR 3-based ANTLR 4 meta-grammar
>> >> >>
>> >> >> It might well be the case that these differ only in the manner in
>> >> >> which the output stringtemplates are provided.  Does anyone have any
>> >> >> opinions on which approach is to be preferred?  Failing that, I am
>> >> >> willing to entertain aspersions cast on, or less likely, advocations
>> >> >> in favor of, my sanity in conceiving this endeavor.
>> >> >>
>> >> >> Thanks in advance for any input,
>> >> >> Matt
>> >> >>
>> >> >> List: http://www.antlr.org/mailman/listinfo/antlr-interest
>> >> >> Unsubscribe:
>> >> >>
>> >> >> http://www.antlr.org/mailman/options/antlr-interest/your-email-address
>> >> >
>> >> >
>> >>
>> >> List: http://www.antlr.org/mailman/listinfo/antlr-interest
>> >> Unsubscribe:
>> >> http://www.antlr.org/mailman/options/antlr-interest/your-email-address
>> >
>> >
>
>

More information about the antlr-interest mailing list