read_stata ignores index parameter

[TomAugspurger]

Code Sample, a copy-pastable example if possible

import pandas as pd
import pandas.util.testing as tm

In [19]: df = tm.makeDataFrame()

In [20]: df.to_stata('foo.stata')

In [21]: pd.read_stata('foo.stata', index='index')
Out[21]:
         index         A         B         C         D
0   2wIkgwAbVm  0.310317 -0.753652 -0.509405 -0.309492
1   59guDFHo0l  0.404983 -0.098995 -0.330786 -0.548584
2   5sX3JbPQjk  0.520377  1.379943  2.375164  1.475831
3   UsF3izOnQy -1.413272  0.668192  0.859581 -0.646165
...

It should be like read_stata('foo.stata').set_index('index')

Side-note, I think elsewhere we use index_col instead of index.

Output of pd.show_versions()

master

[VincentLa]

What is tm.makeDataFrame()?

[TomAugspurger]

@VincentLa Sorry, it's import pandas.util.testing as tm I'll fix it.

[stanleyguan]

Attempting to take a stab at this issue.

[TomAugspurger]

FYI I think @mogillies is working on this? Is that right?

[VincentLa]

Awesome. Anyone at Pycon Sprints right now who is working on this issue?

[stanleyguan]

Looking at Gitter discussions I think @mogillies is working on it. I will look for something else :)

[VincentLa]

@mogillies let me know if you're at Pycon right now and would be up for pairing!

[jorisvandenbossche]

@TomAugspurger if the index param is actually not working right now, shouldn't we change it to index_col at the same time?

[TomAugspurger]

@jorisvandenbossche oh yeah, I guess we won't be breaking backwards compat :D

[mogillies]

Should we change to Index_col?

[mogillies]

Sure. I'm in the same room as you :)

[TomAugspurger]

@mogillies yeah, if you would index_col (lowercase i)

[mogillies]

Sure.

[mogillies]

Just to clarify index param index not working
do you mean that first index not working or the second index not working

1. index = index_cols
2. index_cols = index_cols

if index is not None:
data.set_index(index_cols)

or (I don't think this is the case):
do you mean to use the function index_cols() in class Table?

if index is not None:
data.index_cols()

[jorisvandenbossche]

Currently the keyword argument in read_stata to specify which colunm to set as the index is called index. And there are two issues:

• the fact that this parameter is not working
• the fact that it is a bit inconsistent with other read functions where we typically use index_col

You are trying to solve the first one. But what we suggest is to also fix the second one and rename the index keyword argument to index_col (but for now keeping index as an alias). The fact that the old keyword was not yet working just makes it a bit easier to do this change, as nobody could already rely on its behaviour.