Conda recipe integration with pyproject.toml

[michaelcdevin]

Conda-forge has jinja2 templating allowing it to read name, version number, and dependencies from the setup.py file of a Python package into the Conda recipe meta.yaml file, allowing it to auto-populate for each release. However, WecOptTool builds from a pyproject.toml file as per PEP518 instead of a setup.py file, and conda-forge does not yet have the jinja2 templating for this, requiring us to manually update the Conda recipe with each new WecOptTool release.

It appears that pyproject.toml support was in the works last year, but there is no evidence of progress on it since June 2022.

Other users have succeeded with some workarounds with getting meta.yaml and pyproject.toml to cooperate (conda-forge/conda-forge.github.io#1174), but I am yet to have success with these.

We need to investigate whether there’s an easy workaround to automate this. Alternatively, if/when support has been added to conda-forge, we need to update our Conda recipe with the new jinja2 templating so we do not have to manually update this in the future.

[michaelcdevin]

Conda now supports loading data from poetry using the load_file_data function (see the conda-build docs), which looks like it can go a long way to do what we're looking for here. Will investigate.

[michaelcdevin]

Will be fixed in conda-forge/wecopttool-feedstock#8. Updated meta.yaml file is written, but having some issues with the conda-forge linter.

[michaelcdevin]

As per discussion in conda-forge/wecopttool-feedstock#8, there is no easy way to implement this, and there still has not been any upstream movement on this almost two years later. Given that the Conda feedstock for WecOptTool is pretty automated at this point anyway and #388 documents what to do if you do need to a manual PR, I think we can close this.