ENH: Add axis and level keywords to where, so that the other argument can now be an alignable pandas object.

[jreback]

So traditionally a fillna that does the means of the columns is an apply operation

In [1]: df = DataFrame(np.random.randn(10,3))

In [2]: df.iloc[3:5,0] = np.nan

In [3]: df.iloc[4:6,1] = np.nan

In [4]: df.iloc[5:8,2] = np.nan

In [5]: df
Out[5]: 
          0         1         2
0  0.096030  0.197451  1.645981
1 -0.443437  0.359204 -0.382563
2  0.613981  1.418754 -0.589935
3  0.000000  0.449953 -0.308414
4  0.000000  0.000000 -0.471054
5 -2.350309  0.000000  0.000000
6 -0.218522  0.498207  0.000000
7  0.478238  0.399154  0.000000
8  0.895854  0.230992  0.025799
9  0.085675  2.189373 -0.946990

The following currently fails in 0.12 as where is finicky about how it broadcasts

In [4]: df.where(df>0,df[0],axis='index')
ValueError: other must be the same shape as self when an ndarray

Adding axis and level arguments to where (which uses align under the hood), now enables the other object to be a Series/DataFrame (as well as a scalar) without a whole bunch of alignment/broadcasting. This also should be quite a bit faster.

IIn [6]: df.where(df>0,df[0],axis='index')
Out[6]: 
          0         1         2
0  0.096030  0.197451  1.645981
1 -0.443437  0.359204 -0.443437
2  0.613981  1.418754  0.613981
3  0.000000  0.449953  0.000000
4  0.000000  0.000000  0.000000
5 -2.350309 -2.350309 -2.350309
6 -0.218522  0.498207 -0.218522
7  0.478238  0.399154  0.478238
8  0.895854  0.230992  0.025799
9  0.085675  2.189373  0.085675

This works in 0.12.

In [7]: df.apply(lambda x, y: x.where(x>0,y), y=df[0])
Out[7]: 
          0         1         2
0  0.096030  0.197451  1.645981
1 -0.443437  0.359204 -0.443437
2  0.613981  1.418754  0.613981
3  0.000000  0.449953  0.000000
4  0.000000  0.000000  0.000000
5 -2.350309 -2.350309 -2.350309
6 -0.218522  0.498207 -0.218522
7  0.478238  0.399154  0.478238
8  0.895854  0.230992  0.025799
9  0.085675  2.189373  0.085675

[bmu]

fillna works with a Series, see my answer on SO.

[jreback]

@bmu right....I had enabled that for 0.12 (duh!)
(and I fixed the doc string)

should have had a better example one that is somewhat non-trivial

I have updated the top display