Re: [photran] Hook for custom preprocessor

[Jeffrey Overbey <subscribe-photran@xxxxxxxxxxxx>]

Hi Mike,

> I'm working on a Fortran program which uses a homegrown preprocessor. It
> has directives of the form ##IF etc., which cause Photran to complain of
> unexpected extraneous characters. What are the most natural extension
> point(s) to enable analysis and refactoring on such code? Will this
> change much in version 5?

Preprocessing makes refactoring and source-level static analysis 
extremely difficult, particularly when conditional compilation is 
involved.  One of my colleagues (Alejandra Garrido) did her entire PhD 
dissertation on refactoring C-preprocessed code.

So, unfortunately, there isn't an extension point for custom preprocessors.

A big part of my research agenda for the Fall focuses on preprocessing. 
 Without going into the details, I hope to end up with something like 
that, where you could plug in an arbitrary preprocessor and it would 
heuristically try to map the syntactic and semantic models back to the 
unpreprocessed code.  But since that's an open research problem, I can't 
guarantee anything.  :)

At the moment, we treat C preprocessor lines as comments.  So macros 
(for example) can still kill the parser, but it works as a stopgap 
measure.  I could pretty easily modify the lexer/parser to do the same 
with other types of directives, if that would help.

On that note, what other preprocessors are people using? m4? m5? sed? 
others?

Jeff