From: Lyle Johnson <ljohnson2@...>
Date: 2002-03-29T11:32:16+09:00
Subject: Re: Ruby SWIG problems

>      One final thought.  I'm a little confused as to why you would
> want to wrap a C++ class, when Ruby already supports classes. Wouldn't
> it make more sense to declare the class in Ruby, and have each Ruby
> method do its work by calling a regular C routine?

SWIG's Ruby module looks at a C++ class declaration and generates wrapper
code to:

* define the class
* define all of the class instance methods (and singleton methods, as
appropriate)
* convert input arguments from their Ruby types (e.g. strings, fixnums,
floats) to the appropriate C++ types;
* call the C++ member function on the correct C++ instance
* convert the return value of that call from its C++ type to the appropriate
Ruby type

This is just the bare minimum of what SWIG does. A lot more is possible...

> may not be able to properly connect to your C++ code due to name-mangling.
> And, if you can figure out how to properly match the name mangling method
> used by a given compiler, then you'll have problems porting it to a
> different C++ compiler which may use a different name mangling scheme.

OK, it's becoming increasingly obvious that you have never actually tried
this. SWIG's output is straight C++ code. You compile that code with your
C++ compiler. SWIG doesn't ever get involved with your compiler's "name
mangling methods".

> It seems like too much work for too little payback, if you ask me.

For very small extensions, with a handful of methods, SWIG might indeed be
overkill. But for any nontrivial project, and most any project involving
wrapping C++ classes, I think most who have tried it would agree that it's
worth the learning investment.

> Course, that's just my take, and I don't know the "big picture" as you do.

No comment ;)