Null values in GBQ not correctly converted to NaN in pandas

[snauhaus]

I'm experiencing an issue where the data frame returned by pd.read_gbq contains null values that are not recognized by Pandas. That is, df.loc[df.foobar.isnull()] or df.loc[df.foobar.isna()] returns rows where foobar is actually not null/na. And, vice versa, for notnull() and notna() the same happens.

The problem does not exist when first downloading from GBQ as csv and importing that file.

Attached two screenshots of the issue. Unfortunately I cannot provide a data example.

[tswast]

@snauhaus are there specific data types where you are noticing this issue?

[tswast]

Closing, as I'm not able to reproduce without more information.