arctic icon indicating copy to clipboard operation
arctic copied to clipboard

Deprecation warning in tickstore

Open bmoscon opened this issue 7 years ago • 6 comments
trafficstars

arctic/tickstore/tickstore.py:704: FutureWarning: The 'convert_datetime64' parameter is deprecated and will be removed in a future version
  recs = df.to_records(convert_datetime64=False)

bmoscon avatar Aug 25 '18 15:08 bmoscon

@bmoscon was taking a look at this, but I can't change this for now as internally we seem to be on pandas 0.22.0 which defaults to True and post 0.23.0 onwards (which we use in travis defaults to the latest - 0.23.2+). Was planning to pin pandas in setup.py and that should get rid of this warning as part of https://github.com/manahl/arctic/issues/685

shashank88 avatar Jan 09 '19 14:01 shashank88

please do not pin pandas. just like AHL uses a specific version, so do most other users of Arctic. They wont all be able to use the same version AHL uses. I can fix this with out pinning pandas. Personally I think pinning anything is bad for the same reason

bmoscon avatar Jan 09 '19 14:01 bmoscon

I used to follow this: https://www.kennethreitz.org/essays/a-better-pip-workflow in my previous projects where there is a separate requirements-to-freeze.txt file with pinned stuff. We used to do updates on the first business day of the month to keep it up to date. This prevents issues where automatic upgrades break the build and it's hard to know which one was updated as it's not version controlled (similar to package-lock.json.

shashank88 avatar Jan 09 '19 16:01 shashank88

that website currently is not loading for me, but I'll try it again later

bmoscon avatar Jan 09 '19 17:01 bmoscon

convert_datetime64 has actually been deprecated now, so this actually keeps TickStore from working for users of current versions of pandas.

jeffneuen avatar Feb 05 '20 08:02 jeffneuen

This commit from @bmoscon fixes the pandas >1.0 problems - it's possible to cherry-pick it to generate a new release? The current version available on conda-forge breaks with pandas 1.0 using the Tickstore.

tgcandido avatar Nov 29 '20 15:11 tgcandido