Clean way to generate data file from setup.py stored at the root of the package

[VincentDary]

Hi,

My setup.py generates data files during the build of a package (via python -m build or pip install -e), and store these files at the root of the package in a directory named third_party.

To achieve this result, the project uses a MANIFEST.in with include directives which references all the future built files. And, the generation code is located in the setup.py in a subclasse of setuptools.command.egg_info.egg_info in the override method run(), the command is register via the cmdclass argument of setuptools.setup. I put my generation function in this method because this is the cleanest way I found to have my generated data file produced before the MANIFEST.in is analysed.

This work perfectly but there is a side effect, setuptools.command.egg_info.egg_info.run is called four times in: 1) sdist build dependency, 2) sdist build, 3) wheel build dependency and 4) wheel build. To avoid the data file generation to be relaunched the generation function check if a file artifact is present on the filesystem.

I think my hack is bit dirty, there is a better way to achieve my goal in order to have a generation step call only one time before MANIFEST.in analysis ? Any code sample ?

Of course, I have read the following discussions carefully before publishing my answer.

• Generating python files from setup.py?
• how to execute a script that will modify my package at build time?
• Recommended way to add custom build steps