Josh Wilson

I guess lurking in the background there's also a bigger question of "what is the future of `numpy_financial`?" If the answer is "the package will be these 10 functions forever"...

Ach I only just saw https://github.com/dshahbaz/bazel-mypy-integration/wiki#2-confusion-around-py_library-targets-with-imports. So I guess it's between a rock and a hard place.

> the README reference to a mypy version should also be updated Done in https://github.com/thundergolfer/bazel-mypy-integration/pull/37/commits/87089096631fafa46f221ed4cca90218906496f9. Feel free to reject-I can work around the 3.9 issue.

> Is the "the 3.9 issue" Yeah, you need 0.800 to support 3.9 (ref https://mypy-lang.blogspot.com/2021/01/mypy-0800-released.html), but then that issue kicks in. (So I'd work around it by downgrading to Python...

> mpmath also has an implementation of powm1(x, y), so this is not an unusual or highly specialized function. Yup, seems good to have it!

After thinking about this for a while, I think it’s probably better to just explicitly generate the casted signatures at codegen time and add them to `signatures.py`. Keeps the runtime...

> I suppose the challenge is going to be when there's multiple matching implementations that might work. Yeah you’d have to make sure the signatures are sorted by specificity and...

I can do `scipy.special`. @Juanlu001 what parts of integrate are you most interested in seeing? Something like `quad` would be a royal pain, but if we’re talking e.g. Gaussian quadrature...

(When it comes to adaptive quadrature it would probably be good to do some work on the SciPy side first vs. reproducing quadpack here.)

> I was thinking more of accelerating the new Initial Value Problem solvers introduced in SciPy 1.0, which are much more feature-rich than odeint and ode but have a performance...