ewm check for mutually exclusive arguments doesn't work when times argument is present

[trimcrae]

https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.ewm.html ewm is supposed to accept only one of com, span, halflife and alpha and throw a ValueError if multiple are submitted at once. However, when times is input as an argument, this check is bypassed.

[mroeschke]

Could you provide a minimal code example?

[trimcrae]

import pandas as pd
import numpy as np

times = pd.Series([np.datetime64(x) for x in ['2017-01-01', '2017-01-02', '2017-01-03', '2017-01-04']])
values = pd.Series([1, 2, 3, 4])
smoothed_value = values.ewm(alpha = 0.5, halflife = "1 day", times=times).mean()

print(smoothed_value)
print(pd.__version__)

Output:
0    1.000000
1    1.666667
2    2.428571
3    3.266667
dtype: float64
1.4.2

[mroeschke]

Actually, this is the intended behavior and the documentation just needs updating.

Within the code there's this comment:

            # Halflife is no longer applicable when calculating COM
            # But allow COM to still be calculated if the user passes other decay args

And I think the reason being is allowing the user to not have to reconstruct a new ExponentialMovingWindow if they want to run cov/var/std separately. This behavior is indirectly tested as well in test_online.py