masking empty DataFrame

[kdebrab]

In pandas 0.16.1, I get a ValueError when trying to mask an empty DataFrame:

>>> import pandas as pd
>>> df = pd.DataFrame()
>>> df[df>0]
Traceback (most recent call last):

  File "<ipython-input-3-efe84c9ebabc>", line 1, in <module>
    df[df>0]

  File "C:\Python27\lib\site-packages\pandas\core\frame.py", line 1787, in __getitem__
    return self._getitem_frame(key)

  File "C:\Python27\lib\site-packages\pandas\core\frame.py", line 1859, in _getitem_frame
    raise ValueError('Must pass DataFrame with boolean values only')

ValueError: Must pass DataFrame with boolean values only

Using 'where' works as expected:

>>> df.where(df>0)
Out[4]: 
Empty DataFrame
Columns: []
Index: []

Also, masking a Series works as expected:

>>> ts = pd.Series()
>>> ts[ts>0]
Out[6]: Series([], dtype: float64)

[jreback]

This can be trivial fixed by adding here
if len(key) and not is_bool_dtype(key.value):

[shoyer]

fixed by #10196

[zhaoguixu]

This issue seems not fixed completely. I still get the same error. I have looked up the code and found that _getitem_frame was not fixed yet.

[jreback]

how so? are u using 0.16.2 at least

[zhaoguixu]

Yes, I am using the latest version. I have checked code of the master branch. _getitem_frame(here) was not fixed as _setitem_frame.

[jreback]

I think you are right. must have slipped by somehow. can you open a new issues (and xref this one), which a copy-pastable example (you can use the above) one.

pull-requests are welcome as well.