On 03/07/2020 11:15, Benjamin Schubert wrote: > Hey everyone, > > TLDR: I propose that we allow adding types anywhere in code if it is useful. > > A few months ago, we moved towards annotating our public API with mypy > annotations [0]. > > I think it has overall gone well and also helped us catch a few problems. > > The initial thought was to use it only for the public API and what was needed > to have mypy pass the checks. > > One thing that I often end up doing for large refactors is adding types in > other places, and removing them > before I make the MR, which I think is not an ideal workflow and could > benefit others. > I would now propose that we move towards allowing adding types anywhere the > author think it is useful, > this would also allow us to stop documenting the types in the docstrings and > only leverage type hints. > > I do not mean by that that we should require type hints, rather that we > should accept them as long as they > are correct and not causing problems. > > Any thoughts on this? > > Ben > > [0] https://mail.gnome.org/archives/buildstream-list/2019-June/msg00043.html >
I've found myself doing similar things. I think improvements to documentation and additions of types can only be a positive change. Since the project is in a general state of development I can't see those sorts of changes as being controversial. Type annotations are very helpful for sanity checking api and saving time when reading code. -- Best Regards, Darius For Codethink's privacy-policy please see https://www.codethink.co.uk/privacy.html
