article: Why you shouldn't invoke setup.py directly

[ana]

This is an article that would be interested to keep in mind in future evolutions of setup.py:
https://blog.ganssle.io/articles/2021/10/setup-py-deprecated.html

This is a summary in the end https://blog.ganssle.io/articles/2021/10/setup-py-deprecated.html#summary

[richtja]

Just to be clear, that I understand it correctly. Right now, we are making a lot of effort to move from Makefile to setup.py but this article is about going further and stop using the setup.py also. So if we will decide to use approach from this article, shouldn't we stop the work on Makefile and setup.py and start thinking how to move away form Makefile directly with this new approach, am I right?

[beraldoleal]

We are doing a lot of effort to remove the lack of standard with regards Makefile vs setup.py.

This does not mean that setuptools itself is deprecated, or that using setup.py to configure your package builds is going to be removed.

setup.py is still used, in fact multiple commands that are the "suggested way" reads from setup.py in order to execute (ie. pip install). The article says that we should avoid calling directly. But it is still relevant.

So, IMO we should keep working on that, but taking in account those suggestions to make sure that we are using those suggestions.

[beraldoleal]

Another interesting reading about this setuptools / disutils mess can be found at PEP517.

[ana]

For me, it doesn't change anything of what we're doing now. But we must keep this article in mind for the long term.
In the short term (~9-12 months) I only see two things we should work on:

• Stop using distutils. This means re-writing 'clean' command, and considering that we not longer build the package locally and you can clean with git, this might not even be necessary.
• A lot of of stuff from setup.py/setup() can be moved to a declarative setup.cfg file, see https://setuptools.pypa.io/en/latest/userguide/declarative_config.html

[beraldoleal]

I agree with you here @ana. Based on the current mess that is python dist vs setuptools I would give more time so we can see other early adopters and more community involvement. So yes, lets just keep doing what we are doing with the article in mind.

Stop using distutils. This means re-writing 'clean' command, and considering that we not longer build the package locally and you can clean with git, this might not even be necessary.

I started already an effort on that front removing old distutils code and the only left it was with the 'clean' command because of lack of better choice since I avoiding rewriting-it at the time.

About the need, I would say that it is still useful. A simple make develop (or pip install -e . as suggested in the article) would still produce some temporary artifacts that some might want to get rid of. I would leave the removal of this command only after we are sure no temporary command is generated.

A lot of of stuff from setup.py/setup() can be moved to a declarative setup.cfg file, see https://setuptools.pypa.io/en/latest/userguide/declarative_config.html

+1 for that.