为什么要使用SWIG通过C接口包装C ++库？

Question

SWIG novice exploring the feasibility of wrapping a large C++ library primarily for Python access. Speaking with the developers working on the C++ one has proposed writing a C interop layer that is then wrapped in SWIG.

The two potential options are:

   Base|  Interop  | Scripting Access

============================================

1) C++ |    SWIG   | Supported Languages

2) C++ | C | SWIG  | Supported Languages.

Does #2 add some functionality or stability that I am missing? It looks like a layer of added complexity. Can anyone suggest why a C layer might be the better interface for wrapping in SWIG? (In general terms as you haven't seen the library, etc.)

Answer 1

SWIG is very powerful and will eliminate a lot of the laborious boilerplate code that you would have to write otherwise. Even with boost::python there is such boilerplate code, although far less than doing everything from scratch. The nice thing with SWIG is that it also makes it easy to integrate with additional languages. For instance I had a C++ library that I was exposing to C# so I could build the GUI in C#/WPF, and with the same SWIG input files, exposed to Lua so that the GUI could be scripted via the C++ layer. It was awesome!

About the C layer of your #2 option, I don't see the benefit, except if you want to integrate with a language that doesn't support C++, but I am not aware of such language (SWIG supports about 20 and I only know 4 so maybe one of the others has such restriction). SWIG is very capable, and can in fact wrap rather complex interfaces. With SWIG you can completely change the API if you wish using the %inline and %extend directives. I can't think of a reason that a C layer would be required.

Answer 2

看看直接使用C ++的Boost Python 。