AttributeError when using numpy.logical_and with Pandas Series object

Question

I'm confused about the difference between Pandas Series objects when using reindex_like and related features. For example, consider the following Series objects:

>>> import numpy
>>> import pandas
>>> series = pandas.Series([1, 2, 3])
>>> x = pandas.Series([True]).reindex_like(series).fillna(True)
>>> y = pandas.Series(True, index=series.index)
>>> x
0    True
1    True
2    True
>>> y
0    True
1    True
2    True

On the surface x and y appear to be identical in their contents and indexing. However, they must be different in some way because one of them causes an error when using numpy.logical_and() and the other does not.

>>> numpy.logical_and(series, y)
0    True
1    True
2    True
>>> numpy.logical_and(series, x)
Traceback (most recent call last):
  File "<ipython-input-10-e2050a2015bf>", line 1, in <module>
    numpy.logical_and(series, x)
AttributeError: logical_and

What is numpy.logical() and complaining about here? I don't see the difference between the two series, x and y. However, there must be some subtle difference.

The Pandas documentation says the Series object is a valid argument to "most NumPy functions." Clearly this is true somewhat in this case. Apparently the creation mechanism makes x unusable to this particular numpy function.

As a side-note, which of the two creation mechanisms, reindex_like() and the index argument are more efficient and idiomatic for this scenario? Maybe there is another/better way I haven't considered also.

Which pandas/numpy/python version are you using? numpy.logical_and(series, x) works without error for me in pandas-0.9.0-py2.7... — Andy Hayden
– Andy Hayden, Commented Nov 29, 2012 at 16:03
I would be surprised if this is a new bug in pandas 0.9.1, but perhaps worth upgrading to numpy version 1.8? (the version I am using which seems to work...) — Andy Hayden
– Andy Hayden, Commented Nov 29, 2012 at 16:19
The current stable release of numpy is 1.6.2 (via PyPi and numpy site). Are you using the github repo of numpy for version 1.8? I think I'll post a github issue for pandas and see what they say. — durden2.0
– durden2.0, Commented Nov 29, 2012 at 16:25
I posted an issue on github for Pandas. Maybe they will have more insight: github.com/pydata/pandas/issues/2388 — durden2.0
– durden2.0, Commented Nov 29, 2012 at 16:32

durden2.0 · Accepted Answer · 2012-11-29 21:29:18Z

0

It looks like this is not a bug and the subtle difference is due to the usage of the reindex_like() method. The call to reindex_like() inserts some NaN data into the series so the dtype of that series changes from bool to object.

>>> series = pandas.Series([1, 2, 3])
>>> x = pandas.Series([True])
>>> x.dtype
dtype('bool')
>>> x = pandas.Series([True]).reindex_like(series)
>>> x.dtype
dtype('object')

I posted an issue about this anomaly on the Pandas github page.

The full explanation/discussion is here. It looks like this behavior could potentially change in the future so watch the issue for more on-going details.

answered Nov 29, 2012 at 21:29

durden2.0

9,59212 gold badges49 silver badges58 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

AttributeError when using numpy.logical_and with Pandas Series object

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related