Repository navigation
Short term pandas 2.0 support: avoid pandas==2.0 installation. #159
Description
Task Title
Task: Short term pandas 2.0 support: avoid pandas==2.0 installation.
Task Description
On installing a fresh conda environment for a cobra experiment just now, I got this error during pandas preprocessing:
File ~/.conda/envs/cobra/lib/python3.9/site-packages/cobra/preprocessing/kbins_discretizer.py:330, in KBinsDiscretizer._transform_column(self, data, column_name, bins)
324 data.loc[:, column_name_bin] = (data[column_name_bin]
325 .cat.rename_categories(bin_labels))
327 if data[column_name_bin].isnull().sum() > 0:
328
329 # Add an additional bin for missing values
--> 330 data[column_name_bin].cat.add_categories(["Missing"], inplace=True)
332 # Replace NULL with "Missing"
333 # Otherwise these will be ignored in groupby
334 data[column_name_bin].fillna("Missing", inplace=True)
(...)
TypeError: add_categories() got an unexpected keyword argument 'inplace'
Pandas 2.0 removed inplace arguments: https://pandas.pydata.org/docs/dev/whatsnew/v2.0.0.html
Add in the requirements.txt and readme.md a <=2.0.0 clarification for short term support, to avoid many cobra users struggling.
Activity
sandervh14 commentedon Apr 5, 2023
This issue is for short term support. We can integrate pandas 2.0 on a longer term on issue #160.
ZlaTanskY commentedon Apr 5, 2023
If this also inflicts cobra users who install by using
pip install pythonpredictions-cobra, then the pandas version should be limited insetup.pyas well.requirements.txtandconda_env.ymlwill only be used by cobra developers