search

functime-org / functime

Time-series machine learning at scale. Built with Polars for embarrassingly parallel feature extraction and forecasts on panel data.
https://docs.functime.ai
Apache License 2.0
975 stars 52 forks source link

[BUILD] Bump minimum required Python version to 3.9? #112

Closed baggiponte closed 2 months ago

baggiponte commented 8 months ago

Dask, NumPy and SciPy only supports Python >=3.9. Does this mean we have to update the minimum Python version? We currently use >=3.8.

topher-lo commented 8 months ago

This makes sense 👍

baggiponte commented 8 months ago

Should I make the PR?

topher-lo commented 8 months ago

Let's do it. Also need to update the badges in the readme

baggiponte commented 7 months ago

Finally time to get back to this. I am preparing a PR. One more thing: since we remove py3.8, we can rework a bit the type hints - no need to import List from typing, we can annotate with bare list. Can I work on those too? Should I create a separate PR for that? ruff can do quick work of those.

baggiponte commented 7 months ago

@topher-lo I think I shall also open another PR to take care of the type annotations currently in place. As I was writing: "since we remove py3.8, we can rework a bit the type hints - no need to import List from typing, we can annotate with bare list". Do you agree with that?

topher-lo commented 7 months ago

This would be a breaking change, so let's bundle this PR with 0.10.

baggiponte commented 7 months ago

Why would you say it's a breaking change? It's the internal API. But sure!

baggiponte commented 7 months ago

Also I forgot to update the badges in the README 😱 How do I generate it from shields.io?

topher-lo commented 7 months ago

Users currently using 3.8 when they run pip upgrade functime, an error will occur

baggiponte commented 2 months ago

Will postpone this until Py38 actually becomes unsupported (i.e., once 3.13 comes out). I think it's good to support the same Pyversions as polars does.