From: Hugh Sasse Staff Elec Eng <hgs@...>
Date: 2002-01-16T20:12:52+09:00
Subject: Re: Swig Ruby documentation mods.

On Wed, 16 Jan 2002, Luigi Ballabio wrote:

> At 07:00 PM 1/16/02 +0900, Hugh Sasse Staff Elec Eng wrote:
> > > Have you thought about sending to the swig project?
> >
> >Of course.  No feedback so far from swig-dev on this.
>
> Well, Dave's feedback is here, I think:
> http://mailman.cs.uchicago.edu/pipermail/swig-dev/2002-January/016502.html

This was since then. I have joined the list, then started on the mods.
http://mailman.cs.uchicago.edu/pipermail/swig-dev/2002-January/016586.html

> However, it might be that he sent the mail to the list and not to you
> personally, so that you might have missed it.
>
> As for you questions:
>
> >* if you don't use -module in the swig -ruby command then swig
> >   complains.  This is probably always correct, but needs mentioning.
>
> Are you sure you are including the %module directive in your .i file?
> I'm not using -module and SWIG doesn't complain.

I'm not using a .i file -- I'm using the fact that swig -ruby example.c
will generally work, as described elsewhere in the docs.
>
> >* It is not clear what should be done about integrating several C
> >   functions into the same module.  Does this depend on one writing a .i
> >   file to connect them together?  How?
>
> What I usually do is:
> - somefuncs.i contains the declarations for some functions

OK,
> - someotherfuncs.i contains the declarations for some other functions
> - I write a file mymodule.i containing:
>
> %module MyModule
> %include somefuncs.i
> %include someotherfuncs.i
>
> - by running SWIG on the latter file, all functions from the two original
> ones will be exported in MyModule.

Will it produce one "%s_wrap.c" or two?
>
> >* Can you make swig roll the C functions into methods of a class, or
> >   must one do this via including a module.
>
> If the functions all refer to a struct, you can do it in the usual SWIG way
> (documented for other languages, so you can probably save yourself some
> work by referring to the relevant places):
	[...]
> After creating the extension you'll be able to use the struct as:
>
> s = MyModule::SomeStruct.new
> s.f()
> i = f.g()
> x = f.h(2)

OK, it always will come out as a module.  I was thinking it could look
like one had required a file just containing a class definition.
>
> If the functions are unrelated, You can declare an empty struct and do the
> same thing, but you'll end up with instance methods, not static methods. If
> the latter is what you want, you might be better off doing it on the Ruby
> side. Finally, if you are not forced to use C only and can use C++, you can
> create a class whose methods proxy the C functions and wrap such class with
> SWIG.

I think I'll leave that can of worms for another day, but thank you.
>
> >* Can extconf.rb be made to make other preliminary stuff, that is not
> >   part of the external interface of the class?  Eg, can one make it run
> >   swig if necessary?
>
> The roughest way would be to make a system call before creating the
> makefile, i.e., write extconf.rb as:
>
> require 'mkmf'
> system('swig -ruby MyModule.i')
> create_makefile('MyModule')

That would build it, but it would build it always -- I wanted to get

target: dependencies
	action

stuff in there.  I'll take that as a No for now.
>
> If you want to get fancy, you can compare the modification times of
> MyModule.i and mymodule_wrap.cpp and run swig only when needed, as in:
>
> require 'mkmf'
> require 'ftools'
> system('swig -ruby MyModule.i') unless File.exists? "mymodule_wrap.cpp" and
> File.mtime("mymodule_wrap.cpp") > File.mtime("MyModule.i")
> create_makefile('MyModule')
>
> I didn't actually run the above code, so it might need corrections.
> However, you get the drift.

Yes, that sort of thing would work, but since extconf is supposed to be
used to write a makefile, it seems a bit much to re-implement make
inside it! :-)
>
> Hope this helped,
>                          Luigi
>
>