Philippe Veber a écrit :
> You're right this is a linking issue and now the question is at which level
> you want to "link" your code. I do not see the point of including a cmo in
> another like you describe : i believe there are simpler and mainstream
> options. Maybe I miss some details about your problem ? Using findlib to
> help the linker is one way to do it, but if you insist on loading a single
> module then you have two other options :
> - code inclusion -> m4, camlmix, camlp4 or any preprocessor (not that ugly,
> but still)
> - the -pack option for combining several cmo in a single one (but then all
> your modules are included in a "toplevel" module)
> sorry if i still didn't get your problem ;o).
> ph.

You're getting my problem right. I'm working on the OCaml-R binding of 
Maxence Guesdon.

The thing is, I have a r.ml file, where all the binding work is done. 
However, in order to launch the shared library containing the code of 
the R interpreter, some environment variables must be set up first. 
Therefore I have a small R script, based on Dirk Eddelbuettel's littler 
software, that generates an .ml file, rstdenv.ml, containing the 
environment variables that need to be set up correctly.

Maybe it's trying to be a bit too pretty, but I do not like the name 
Rstdenv for a module. I'd rather have R.Standard.env...

So I'm looking for a way to include a module as a submodule, without 
having to link it to rstdenv.cmo, which I really do not want to be 
available from the r.cma archive.

That's why I was wondering if there is a clean, camlish way, to 
link-substitute code from one .cmo into another .cmo...

The -pack option is clearly overkill, camlp4 seems overkill too, m4 is 
ugly, do not know camlmix. There's also a package from Gerd Stolpmann, 
xstrp4 or something like that which could be useful. But as I want to 
minimise dependencies on extra stuff, I was wondering if there was an 
out-of-box obscure compiling option to do so.

All the best,

Guillaume Yziquel.

> 2009/11/12 Guillaume Yziquel <guillaume.yziquel@citycable.ch>
> 
>> Philippe Veber a écrit :
>>
>>  Hi
>>> maybe you can have a look at findlib and its #require statement. For
>>> instance, pxp (xml related library) depends on many cma, but everything
>>> loads automagically when invoking #require :
>>>
>> No, no, no... this is not the issue at all. My issue is not about loading
>> stuff with findlib, it's about including a .cmo into another .cmo. I'd like
>> to create a .cma with only a.ml, and not containing b.ml.
>>
>> It's not a toplevel issue, but a 'linking' issue.
>>
>> Thanks anyway.
>>
>> Guillaume.
>>
>>
>>
>>
>>  2009/11/12 Guillaume Yziquel <guillaume.yziquel@citycable.ch>
>>>  Hello.
>>>> Imagine I have a file named a.ml containing
>>>>
>>>>  module C = struct
>>>>
>>>>>  include B
>>>>> end
>>>>>
>>>>>  and a file named b.ml containing the code
>>>>  let f x = x + 1
>>>> When I compile everything to .cmo files, I cannot load a.cmo from the
>>>> toplevel without loading b.cmo beforehand.
>>>>
>>>> Is there a way to make the 'include B' statement to include the code of
>>>> the
>>>> B module in the C submodule directly so that it is not required to load
>>>> the
>>>> b.cmo file before loading the a.cmo file?
>>>>
>>>> That would be extremely useful to me...
>>>>
>>>> All the best,
>>>>
>>>> --
>>>>    Guillaume Yziquel
>>>> http://yziquel.homelinux.org/