Ts summary - polars backend

business-science / pytimetk

Time series easier, faster, more fun. Pytimetk.

https://business-science.github.io/pytimetk/

MIT License

696 stars 60 forks source link

Ts summary - polars backend #211

Closed GTimothee closed 1 year ago

GTimothee commented 1 year ago

partially solves #77

Done, tests working. Updated augment_fourier so that it uses this new version.

Part of it depends on the function "get_frequency_summary" so it is the next on my list :) In the meantime I computed this part like this: frequency_summary = pl.from_pandas(get_frequency_summary(date.to_pandas()))

mdancho84 commented 1 year ago

I'm seeing some failed tests on older versions of pandas. https://github.com/business-science/pytimetk/actions/runs/6739223865/job/18320315245?pr=211

GTimothee commented 1 year ago

Oh I see ! I will have a look

mdancho84 commented 1 year ago

Ok, thanks. Let me know what you find out.

GTimothee commented 1 year ago

Apparently 2 different errors:

pandas==2.0.0 : dtype error (not reproduced locally with python3.11, need to investigate further)
pandas==1.5.0 : bug in the 1.5.0 version it seems. It has been fixed later on but maybe for this version to work we should think of a workaround (bug reference: https://github.com/pandas-dev/pandas/issues/45725)

mdancho84 commented 1 year ago

Nice work! 1 test to go. Looks like datetime[ns] vs datetime[us]

GTimothee commented 1 year ago

yay 🎉

mdancho84 commented 1 year ago

Nice work!