SciPy is discussing migrating to keyword-only arguments

This is maybe something we can start doing as well across the library.

https://mail.python.org/archives/list/scipy-dev@python.org/thread/MK7AWN5ARX7AOJXDANUPTFYHSJ5Y53ES/

I’ve been encouraging this in reviews for new functions for some time now, see e.g. https://github.com/scikit-image/scikit-image/blob/1b52bdf5689a59517e37d997f6ba2e1421e2abb6/skimage/restoration/j_invariant.py#L200, which is 4 years old. :smiley: For skimage2 I of course think it’s a no-brainer.

1 Like

Same. I think until we get around to skimage2, incrementally switching to keyword arguments were possible is the way to go. I encourage updating the whole signature at once though if a function is touched. E.g. Deprecate output and positional args in `filters.gaussian` by lagru · Pull Request #7225 · scikit-image/scikit-image · GitHub.