Re: sablecc generating parsers for other languages

[Indrek Mandre <indrek@mare.ee>]

Ouch! And here is the famous Indrek happily generating grammars,
not parsers.

My grandchildren will crucify me when they read this. I hope you
can forgive me for my stateless mind.

Regards,
Indrek

Indrek Mandre wrote:

> I'm pondering here an idea on how simple would it be to make
> SableCC generate grammars for other languages. And foremost i'm having
> C++ in mind. C++ is rather close to Java so it shouldn't be very
> difficult.
>
> I'm not thinking yet on how the structure or object model of generated
> c++ parser should be but more how easy would it be generating it. As far as
> I've looked into sablecc the code generation is hardcoded into the Gen* files.
>
> There is a class named ResolveIds that contains a lot of information about the
> structure of parser. But it does not have everything in it to generate a parser.
>
> So it either needs improvement or a completely new data model for parser
> data has to be created. A data model that could be used by generators for
> generating parsers. Which to prefer?
>
> What I have in mind is something like:
>
>     interface ParserGenerator {
>         void generate (ResolveIds parserData);
>     };
>
> Anyways I'm not afraid to spend a weekend or two for making it happen.
> If you think my efforts are not pointless and give me your blessing :)
>
> Regards,
> Indrek
>
> PS: I found the following oddity from the SableCC jdk2 distribution at
>        generated lexer code:
>
>     private String getText(int acceptLength)
>     {
>         StringBuffer s = new StringBuffer(acceptLength);
>         for(int i = 0; i < acceptLength; i++)
>         {
>             s.append(text.charAt(i));
>         }
>
>         return s.toString();
>     }
>
> Here text is a StringBuffer. Wouldn't this be cleaner:
>
>     private String getText(int acceptLength)
>     {
>        return text.substring (0, acceptLength);
>     }
>
> This function is heavily used I think; so the new approach would benefit
> the performance.