[antlr-interest] Don't want to pervert reference grammar like Terence Parr one on Java 1.5

[Johannes Luber]

Freddy Mallet schrieb:
> Hi all,
> 
> I'd like to create my own Open Source library to calculate metrics on
> java and pl/sql source code. Those metrics could be lines of code,
> cyclomatic complexity, maintainability index (SEI), etc. I've studied
> javacc and antlr, then read the definitive ANTLR Reference.
> 
> At that point, I guess I've enough material to begin writing this
> library but ... there is a recurring question I can't evacuate from my
> brain.
> 
> Here is my issue : on the shelf, there is already two reference grammars
> available to parse java and pl/sql. Of course, those two reference
> grammars will evolve in time to follow java and pl/sql evolutions.
> 
> Tomorrow, beginning to touch those two grammars to build my own AST or
> to add my own actions on rules is equivalent to create a new branch on a
> source code. After few months, the differences between the grammars will
> be so important that I couldn't synchronize them. So after a while, I
> couldn't capitalize on ANTLR community work.
> 
> That's exactly what happens currently to JavaNCSS library. The library
> used javacc and all work has been done from a Java 1.1 grammar which has
> now great difficulty to evolve.
> 
> So, is there an abstraction mechanism which allow us to use on the shelf
> grammars without modifying them directly ?
> 
> Thanks for your help,
> 
> Freddy

AFAIK, there is a whitepaper on a tool named gdiff, which would be a
specialized diff for grammars. It would ignore actions and focus only on
the grammar structure. But I don't know how far the implementation has
been realized.

Johannes

> 
> 
> -- 
> Freddy Mallet,
> Sonar , Embrace Quality
> http://sonar.codehaus.org
> 
> 
> ------------------------------------------------------------------------
> 
> 
> List: http://www.antlr.org/mailman/listinfo/antlr-interest
> Unsubscribe: http://www.antlr.org/mailman/options/antlr-interest/your-email-address
>