Re: Re: SIDE NOTE: CIL Schema

On Sun, Jan 09, 2005 at 05:56:51PM +0000, Martin Klang wrote:
> It's going to require work. I'm currently playing around with the Antlr 
> parser generator, and its Java grammar in particular - I'm looking at 
> ways of making it generate XML output from its AST. That would take 
> care of Java -> MLML. Unfortunately I don't have much time to play 
> around with stuff...

I'm not sure if you have seen it already, so I'll just keep this
posting short: Take a look at the parser-java module of SrcML. The
java.g[1] and java.tree.g[2] are the enhanced Antlr java grammar 
(and tree parser) which produce our SrcML output. The generated 
parser relies on some classes of the parser-shared module and it 
currently only works for java 1.3 (I have not yet invested time 
into 1.4 compliance, but from what I know it should only be the 
assert keyword which is trivial to add). Also comment handling is
a bit on the magical side as it is of course impossible to tell
what a commment actually belongs to. Right now the parser assumes
a simple 'comment belongs to the next statement' heuristic most
of the times.

But other than that you should be able to quickly make the neccessary
changes from SrcML to MLML.

In case you have already looked into that and decided to go into a 
different solution I'd be pleased to hear your arguments on the decision.
Of course the process is also possible using a java.g grammar only,
however we found the use of a tree grammar to be very advantageous
as you don't run into the problem of when to open which tag. This is
rather difficult to decide in those cases were java.g is still in the
process of looking ahead to find out what actually happens there in the
code. Also the resulting tree grammar turned out to be nice for 
maintaining due to its simple structure (for a grammar at least ;)

[1] http://cvs.sourceforge.net/viewcvs.py/srcml/parser-java/java.g?view=markup
[2] http://cvs.sourceforge.net/viewcvs.py/srcml/parser-java/java.tree.g?view=markup

With kind regards,
-- 
Raiser, Frank
Student @ University of Ulm (www.uni-ulm.de)

The most likely way for the world to be destroyed, most experts agree,
is by accident. That's where we come in; we're computer professionals.
We cause accidents. (Nathaniel Borenstein)