[Haskell-cafe] compilation to C, not via-C

[Jason Dagit]

On Fri, Apr 24, 2009 at 10:09 AM, Sam Martin <sam.martin at geomerics.com> wrote:
> Thanks Bulat and Miguel.
>
> Unless I've missed something, the FFI is an great description of how to
> bind C and Haskell - I think it's fine. I've read most of the content
> either on or referenced by the wiki - although I could well be missing
> something. I believe the problem is with the actual code compilation
> itself.
>
> I need a list of .c and .h files as an end result of the Haskell
> compilation stage. I expect these c files will need to include Haskell
> runtime C code to operate, and therefore have some dependencies in order
> to compile and link.
>
> Afaict, GHC as it stands does not allow me to do this, even though it
> presumably generates C in the process of compiling binary objects.
>
> Actually having C source as an end result is critical as I need control
> over exactly how the source is compiled and linked. For example:
> - I need to compile to different targets: either a static C lib, exe,
> dll or C++ lib.
> - I need to support multiple compilers.
> - I might want to produce a custom runtime.
>
> In short, I'd like to use Haskell as a code-generator.
>
> I can't see that this would be unachievable, particularly given it's
> generating C already. Have I missed something?

You might have better luck with a different Haskell compiler.  For
example, JHC/LHC try to generate ansi-C and then compile that.  YHC
has a very hackable backend, see for the example the YHC-javascript
backend.

Jason