Opt Einsum v3.4.0 - Githubissues

dgasmith commented 2 years ago

We should go ahead and kick out a v3.4.0 release given our current cadence. However, thinking ahead to the next release we have a few changes we should make:

Based on NumPy's NEP: On Dec 26, 2021 drop support for Python 3.7 (initially released on Jun 27, 2018) we will drop Py3.7 support on the next release. Dataclasses and the walrus operator will help with code cleanliness.
Deprecate optimal for dp or fix #167. I would note that the dp algorithm is pretty dense and not particularly readable.
Make a decision on global preprocessing in #114.
The code base is grown fairly organically and is written in a Python 2 style, I would propose a general refactor for better readability and therefore extensibility.

@jcmgray Any features you would like to try and get in?

jcmgray commented 2 years ago

The only possible small thing from my side I've just raised in (https://github.com/dgasmith/opt_einsum/issues/193). The list of bigger things to consider for next release all look good too.

dgasmith commented 6 months ago

Nearly 2 (!) years later we should look into a 3.4 release after 4 (!!!) years after 2.3, especially for Py3.12 (#221). I'm a bit amazed that we appear to be stable and working for this duration without too many issues while maintaining 16M/month downloads.

I'm going to ask around and play with our typing a bit to make sure we do not make the situation worse with the current hints (although advantageous for internal dev). Anything else we should get in?

dgasmith commented 2 months ago

Done!

dgasmith / opt_einsum

Opt Einsum v3.4.0 #191