Packaging files setup.cfg and setup.py have been removed

[GiuseppeDiGuglielmo]

Summary

The files setup.cfg and setup.py (in the top level directory) have been removed in this commit 9857da1

They were useful when running hls4ml with Catapult. I could have local working copies of the repository and point to them with PYTHONPATH.

Reproduce

git clone --recursive git@github.com:fastmachinelearning/hls4ml.git
cd hls4ml
git checkout 9857da1
PYTHONPATH=$PWD python -c "import hls4ml"

It will fail with importlib.metadata.PackageNotFoundError: No package metadata was found for hls4ml

At the previous commit (1ad1ad9), the setup files are still there and I can run:

PYTHONPATH=$PWD python -c "import hls4ml"

Can we bring them back? Is there an alternative solution so I can keep using PYTHONPATH to point to local hls4ml copies?

Thanks

[vloncar]

The official suggestion for packaging is to use the pyproject.toml file to set up the project and the dependencies so we switched to that. The old way is gone and not coming back. If your goal is to install any given repo/branch and make it visible to your scripts (so import hls4ml works) you can do it like:

git clone --recursive git@github.com:fastmachinelearning/hls4ml.git
cd hls4ml
pip install -e .
git checkout whichever_branch_or_commit_you_want

after you do pip install -e . (don't forget the dot) your environment will "link" to the installation folder, which is the git repo you cloned. Then all changes within that repo will be reflected in the install, so you can modify files, switch branches etc.

[GiuseppeDiGuglielmo]

This breaks a little how we were doing things with Catapult. Let me highlight this for @dgburnette and see if they offer an alternative.

pip install -e. is fine but I personally liked PYTHONPATH better.

[dgburnette]

The problem with using 'pip install -e' is that it requires the user to link their python environment to a specific installation of Catapult.
This limits the ability of a user to work with multiple different Catapult releases using the same python env.
It also prevents customers from using a centrally installed python env shared by all Catapult users.