search

feature-engine / feature_engine

Feature engineering package with sklearn like functionality
https://feature-engine.trainindata.com/
BSD 3-Clause "New" or "Revised" License
1.8k stars 304 forks source link

future warnings #693

Open solegalli opened 9 months ago

solegalli commented 9 months ago

List of future pandas warnings

c:\users\sole\documents\repositories\feature_engine\feature_engine\creation\math_features.py:212: FutureWarning: The provided callable <function sum at 0x0000016E1CA6D090> is currently using Series.sum. In a future version of pandas, the provided callable will be used directly. To keep current behavior pass 'sum' instead.
glevv commented 9 months ago

Connected to this https://github.com/pandas-dev/pandas/issues/53425

I think we can easily suppress this warning or not so easily bump pandas version.

We should also add information to this transformer, that same numpy and pandas functions could lead to different results (this is why they raise this warning). For example, np.std has ddof=0 while Series.std() has ddof=1

olikra commented 1 week ago

I sorted out the > 3000 warning messages of a complete pytest-run. Most of them ('SkipTestWarning', 'UserWarning', 'ConvergenceWarning', 'RuntimeWarning') can be irgnored by now to get a better view whats relevant.

In the feature-engine code itself we have to take care for:

For details look into this file: pytest_warnings.csv

In #774 I already worked on math_features - so it makes sense for me taking care of it. The rest needs to get sorted out as well cause we saw last week what happens ignoring / not seeing this warnings(#772).

solegalli commented 3 days ago

Thank you @olikra That would be great!