.ix() failure for mixed-integer index

[lodagro]

from mailinglist
cc @ruidc

With the following DataFrame:

df = pandas.DataFrame(index=[1, 10, 'C', 'E'], columns=[1, 2, 3])

df.ix[df.index[:-1]] works
df.ix[:-1] doesn't

df.ix[[1, 10]] does not work
df.ix[pandas.Index([1, 10], dtype=df.index.dtype)] works

df.ix[10] does not work

[lodagro]

Just spend a bit of time on fixing this.

In [9]: df
Out[9]: 
           1         2         3
1   0.476835 -0.541463  0.142869
10  2.027024  0.434206 -1.092403
C   1.108003  1.692173 -1.413435
E   0.688612 -1.082818  1.289867

In [10]: df.ix[10]
Out[10]: 
1    2.027024
2    0.434206
3   -1.092403
Name: 10

In [11]: df.ix[[1, 10]]
Out[11]: 
           1         2         3
1   0.476835 -0.541463  0.142869
10  2.027024  0.434206 -1.092403

When starting to fix the slices and df.ix[list], i noticed that slicing from label to position (or reversed) and selecting a range with both labels and positions on 0.8.1 has (at least for me) unexpected behavior (even for indexes without integers). So what can we expect from pandas to behave in these situations?

In [24]: df
Out[24]: 
          A         B         C         D         E
a  0.613189  0.689797 -0.234561  2.147688 -2.809169
b  0.203202  0.025470  0.862836  1.250575  1.166273
c  0.002628  0.413194 -1.367218  1.112502  0.059427
d -2.734778 -1.671669  0.029811 -1.469516 -0.043275
e  1.560336  0.845640  1.517610 -0.861052 -0.206302

In [25]: df.ix[[2, 'e']]
Out[25]: 
          A        B        C         D         E
2       NaN      NaN      NaN       NaN       NaN
e  1.560336  0.84564  1.51761 -0.861052 -0.206302

In [26]: df.ix[2:'e']
Out[26]: 
          A         B         C         D         E
a  0.613189  0.689797 -0.234561  2.147688 -2.809169
b  0.203202  0.025470  0.862836  1.250575  1.166273
c  0.002628  0.413194 -1.367218  1.112502  0.059427
d -2.734778 -1.671669  0.029811 -1.469516 -0.043275
e  1.560336  0.845640  1.517610 -0.861052 -0.206302

In [27]: df.ix[['b', 4]]
Out[27]: 
          A        B         C         D         E
b  0.203202  0.02547  0.862836  1.250575  1.166273
4       NaN      NaN       NaN       NaN       NaN

In [28]: df.ix['b':4]
Out[28]: 
Empty DataFrame
Columns: array([A, B, C, D, E], dtype=object)
Index: array([], dtype=object)

[wesm]

df.ix['b':4] was not intended to be a supported API. It only works by accident because:

1. the index is sorted
2. in python 2 integers are comparable with strings

I'll have to poke around at some point about raising a helpful error...

I'll look into fixing the cited bugs now

[wesm]

fixed in c32cc6e