[fpc-pascal] A complex trouble(at least for me)

[David W Noon]

On Fri, 2009-05-08 at 09:41 -0300, Arí Ricardo Ody wrote:

> It's out matter but, I'm working with legacy 
> modernization. The company I work deals with 
> legacy modernization. Mainly in mainframes

So, are you trying to convert COBOL code to some more modern language?

If not, I still don't understand what you are trying to do.

Moreover, very few modern languages implement COBOL semantics,
especially as regards decimal arithmetic. You will likely need to
generate assembler code for your target platform, and that is even more
of a "legacy" approach to software than coding in COBOL. [Note: double
precision floating point does *not* implement decimal semantics, in
spite of it seeming an easy substitute.] The "modern" language that is
most likely to provide you with decimal capability is Java; it also runs
even slower than COBOL.

Assuming you are trying to automate conversion of COBOL source to some
newer language, then you will largely need to write a parser for the
COBOL grammar. This is far from trivial. You will then need to write an
emitter for your chosen "modern" language, so that you rebuild the COBOL
semantics as exactly as possible.

>From your original message, you seem to be still working on the parser.
I would suggest you use a more compiler theoretical approach than just
generating a string list. If you encode the parsed COBOL source into
some internal meta-language, this will make your emitter much easier to
write. Better yet, if your encoding includes some semantic analysis of
the COBOL source, you can optimize the generated code so that it both
runs better and is easier maintained in your "modern" language.

Just my 2¢ worth.

-- 
Regards,

Dave  [RLU #314465]
=======================================================================
david.w.noon at ntlworld.com (David W Noon)
=======================================================================