Cplusplus-sig

Download

cplusplus-sig@python.org

November 2018

1 participants
1 discussions

Distributing a simple boost.python package
by Gautham V Nov. 2, 2018

Nov. 2, 2018

Hi, I've written some code using boost.python (accessing numpy arrays) that I am able to use on my Ubuntu 16.04 system (Boost 1.68.0) . Using GNU make I am able to compile it into a .so file, and then I import it in Python to use. I would like to distribute this as a package. I was wondering what the easiest way to do this would be. After some time searching, I've found the following options: 1. Use bjam --- I could not get the tutorial to work 2. Use a CMakelists.txt or a Makefile --- I used a modification of a file on Github ( https://github.com/zpoint/Boost-Python-Examples/blob/master/Examples/hello_…) to build the .so file 3. Use setuptools/distutils --- I am trying out a sample package right now. Should I provide the C++ source, or just a binary? 4. Use conda manager --- It was suggested in the archives ( https://mail.python.org/pipermail/cplusplus-sig/2018-October/017499.html). 5. Use Scons --- this is an alternative that seems simpler than the others, I am traversing its documentation currently. My package (essentially) contains the following files: class1.h, class1.cpp class2.h, class2.cpp class3.h, class3.cpp (contains instance of class1 and class2) wrapper.cpp (contains boost includes, and a wrapper class of class3) helper.py (contains helper functions to deal with the wrapped objects) I would prefer to use boost.python as it is very similar to Rcpp (the same code works seamlessly on all platforms when converted into an R package via RStudio), but I have not understood the build process for a python package yet. Does anyone have any useful tips on how to use the above tools, or a checklist I can use to mark off steps in this process? Thank you, Gautham

1 0