From: Miloš J. <mil...@sk...> - 2015-04-22 19:22:06
|
Hi Vadim, thanks for your prompt reply! I'm not questioning the motivations for this, I think they are all valid as far as I can assess, especially if they are in long-term going to reduce the size of the C++ code. I'm just pointing out that this is a backward incompatible change that should be documented in bold in the CHANGES file (as other such changes are). As for: 2015-04-22 15:01 GMT+02:00 Vadim Zeitlin <vz...@ze...>: > 1. It is much more convenient to see the function argument names and, at > the very least, their number for Python programmers using the generated > API. Let's not forget that the goal is not to make SWIG just generate > something, the goal is to make the API usable from Python and losing all > information about arguments doesn't help at all with this. > and For me the answer is emphatically yes and I honestly don't see how could > it be argued otherwise if you think about it from the point of view of the > programmer using the Python API. > To be honest, in a group of 5 Python programmers and 2 C++ programmers being interfaced by SWIG;), those Python programmers have never ever read the module.py file (they use a doxygen documentation built from module.i) and we C++ programmers have almost never read module.py either since the important things that SWIG is doing were always in the module_py_wrap.cc file. That's just to add -- and why I see the main motivation being reducing and simplifying C++ code size rather than making .py files more readable (most of the file is just so much boilerplate for a regular Python programmer that I don't think you can anticipate they would actually read it). Anyway, back to my problem -- do our typemaps look all right to you? Thanks, Milos |