Pandas problem, 0 transformed to NaN when use "where" method

Question

I'm facing a problem i don't understand :

>>> Y.isnull().values.any()
False
>>> Y.where(Y == 0).isnull().values.any()
True

I don't understand how NaN values can appeared in the second result. Y has a dtype = int64

Any idea ? Thank you very much!

Answer 1

Check out the behavior of pandas.DataFrame.where: https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.where.html

It substitutes the values that don't meet the condition with something. By default that something is not boolean values, it's 1 and nan. Which means by doing Y.where(Y == 0) you're creating nan values instead of all the values that weren't 0. Hence nan values in the second line.