How to get an AST of OCaml source?

[mulhern <mulhern@cs.wisc.edu>]

Hi,

I need to use OCaml source code as input to a tool that generates a 
related sort of output. It's not a source-to-source translator; the 
code generated will not do what the OCaml source code does. It's not a 
pretty printer either, the relationship isn't so direct. I'ld like to 
write the tool itself in OCaml and am hoping to use CamlTemplate.

What is the most direct way to get a useful, easily traversable, 
representation of OCaml source code in OCaml? Clearly, there is one 
embedded in the various ocaml tools.

I understand that Camlp4 will dump out a binary AST, which is then 
input to the OCaml compiler. If that is the best way to go, could 
somebody give me some pointers about how to traverse this AST? I have 
been unable to extract the information from the Camlp4 documentation.

I have also been looking at the OCaml src code distribution. I realize 
it's possible to pass the compiler a flag (dparsetree) that will cause 
a pretty-printed version of the parse tree to be dumped out. On 
examination of compiler.ml I can see how that ast that gets pretty 
printed is constructed. Is my best bet to write an ocaml application 
that just uses a a large chunk of the ocamlc source code modeling the 
application as best I can on the compiler.ml source?
Or would I be better off parsing the pretty-printed stuff that gets 
dumped out? Or, could I write my own printer that is not so pretty and 
dumps a textual representation that is very easily parsed so that the 
AST can be reconstructed and insert that into my local version of 
ocamlc?

I'm sure people have encountered similar problems before. Any advice 
based on your experience would be very much appreciated.

-mulhern