DataFrame.interpolate(axis=1, method="time", inplace=True)

[markrichardson]

In pandas 0.15.2, DataFrame.interpolate( ... ) doesn't work with the keyword combination axis=1, method="time", inplace=True. It just returns the original DataFrame.

import pandas as pd
import numpy as np

periods=5
idx = pd.date_range(start="2014-01-01", periods=periods)
data = np.random.rand(periods, periods)
data[data < 0.5] = np.nan
df0 = pd.DataFrame(index=idx, columns=idx, data=data)
df1 = df0.copy()

print "(1) Original df"
print df0

print "\n(2a) axis=0, inplace=False"
print df0.interpolate(axis=0, method="time")
print "\n(2b) axis=0, inplace=True"
df0.interpolate(axis=0, method="time", inplace=True)
print df0

print "\n(3a) axis=1, inplace=False"
print df1.interpolate(axis=1, method="time")
print "\n(3b) axis=1, inplace=True"
df1.interpolate(axis=1, method="time", inplace=True)
print df1

[jreback]

hmm, prob needs some more testing. pull-requests are welcome on this.
just a note though. inplace makes code IMHO harder to read and does not allow method chaining, so generally should be avoided.

[TomAugspurger]

The method parameter doesn't matter. All inplace ops fail w/ axis=1.

[TomAugspurger]

@jreback am I right that df.T returns a copy?

In [16]: df.T.T is df
Out[16]: False

If so, then my implementation w/ inplace and axis=1 is pretty broken.

[jreback]

you need to do something like this (as you are just overwriting the reference)

if inplace:
        if axis == 1:
                self._update_inplace(new_data)
                self._data = self.T._data
        else:
                self._update_inplace(new_data)

[mroeschke]

Looks to be fixed on master. Could use a test.