Python Packages Project Generator VS summer

- python-package-template

Hey HN, this would be more of an early release (still planning on some tweaks before a release) -- would love to hear your thoughts on this!

For some back-story, this is more of a side-side-project (made this while working on another side-project).

When I switched to using Go for my projects (from Python), the lack of a template generator similar to python-package-template[1] was very annoying. I would copy the basic files (Makefile, Github actions, PR templates, etc) from the previous project only to realize I forgot to change some stuff, and now would need to rewrite git history.

By the third project, I decided to create a template generator for Go! I've tried to keep the generated project as flexible as possible - you can decide to skip the of it and go for a simple project, or take the bloat (pre-commit would need Python for one).

While making go-template, one of my side goals has been to keep the project beginner-friendly. I remember stumbling upon python-package-template[1] as a novice, and learning more than I had in a semester - Makefiles, linters, code-formatters, semantic versioning, pipelines, and so much more! With go-template, I hope to give that same experience to some other newbie who might stumble upon my repo (or a project generated using go-template).

As a fun fact, go-template has an option to remove Github-specific-features (pull request templates, workflows, etc). This was inspired by a comment on HN[2] pointing out that many open-source projects were on Github simply because of FOMO, which in-turn promoted Github's dominance!

[1]: https://github.com/TezRomacH/python-package-template

Oh, and if I am not mistaken, you have also used the python-package-template itself to generate goli structure 🔥

CookieCutter or Kedro are the winners. I still think we will stick to Kedro template, because it offers extra functionality, and I like to think of each project as a set of pipelines to be run. Anyway, some cookiecutter templates are very good, like this one. In case we use both Kedro and ClearML, we'll have to figure out how to integrate its pipelines with ClearML tasks. But in the slack channel of ClearML there are other teams doing the same, so at least it's possible.

It's set up so the docstrings are automatically read for public classes/functions/methods (ie no _ prefix) and inserted into the docs. I did a little bit of manual setup to get the "automodule" feature to read in the docstrings from the modules I wanted to expose, but after that it's all automatic.