Cobol AS/400 hard

Hi all,
because it's very hard to find a complete grammar definition for Cobol
AS/400, I wonder if there is a way for describing a grammar (BNF,
EBNF,...) that tokenize only some specific rules and ignore all the
rest.
Many Thanks
Matteo.

"matteo" <matteom@avanade.com> wrote in message
news:3d89db82.0407230340.45d8eb09@posting.google.com...
> because it's very hard to find a complete grammar definition for Cobol
> AS/400, I wonder if there is a way for describing a grammar (BNF,
> EBNF,...) that tokenize only some specific rules and ignore all the
> rest.

This is called "Island Parsing"  Check out comp.compilers.

The down side to island parsing is
a)  you still have to define
much of the complete languages (e.g., mostly you have
to lex everything, you still have to worry about COPYLIBs
and COPY REPLACING, etc.),
b) often you discover after you get the "islands" of interest
to parse, that to do the job you want done, you have
to go back and expand the islands.  Eventually you
end up with most of the full grammar anway.
c) If you need to know what a name means, then
you pretty much have to do the full grammar,
and you have to have the symbol table.
Most interesting analyses require C anyway.
So we don't see this as really a big win.

However, we've found it pretty effective to take a clean base grammar (e.g.,
ANSI COBOL 85),
and add the differences as defined by reference manuals.

Also in the not-for-free category, too, is our tool,
the DMS Software Reengineering Toolkit.
It has very strong parsers for many languages, including ANSI and VS II
COBOL,
with full symbol table construction.

Lastly, there's lots of tool work needed beyond parsing.
DMS provides means to define analyzers and source-to-source transformers
of the code.

--
Ira D. Baxter, Ph.D., CTO   512-250-1018
Semantic Designs, Inc.      www.semdesigns.com

Check www.corporola.com. There is a BNF definition for COBOL 74/85 for COBOL
to Java translation.
I think It will cover AS/400.

Thomas

"matteo" <matteom@avanade.com> wrote in message
news:3d89db82.0407230340.45d8eb09@posting.google.com...
> Hi all,
> because it's very hard to find a complete grammar definition for Cobol
> AS/400, I wonder if there is a way for describing a grammar (BNF,
> EBNF,...) that tokenize only some specific rules and ignore all the
> rest.
> Many Thanks
> Matteo.