passing index parameter shouldn't be necessary when applying pivot function

[scls19fr]

Hello,

I have a Pandas DataFrame df like this:

            name  value
2014-01-01  temp   10.0
2014-01-01     p  101.0
2014-01-02  temp   11.5
2014-01-02     p  100.0

I wanted to get a new DataFrame df2like:

            temp      p
2014-01-01  10.0  101.0
2014-01-02  11.5  100.0

I did this:

df2 = df.pivot(columns='name', values='value')

but it raises cannot label index with a null key

I need to do this

df2 = df.pivot(index=df.index, columns='name', values='value')

see http://stackoverflow.com/questions/26101246/spread-data-across-multiple-columns

In my mind, passing index parameter shouldn't be necessary when applying pivot function in such a case.

Kind regards

[jreback]

The canonical way of doing this is as follows:

In [25]: df.reset_index().pivot('index','name','value')
Out[25]: 
name          p  temp
index                
2014-01-01  101  10.0
2014-01-02  100  11.5

That said, I think it could simply use the index (though their IS a bit of ambiguity as even in this simple case (after the reset) its an integer index).

pull requests are welcome for an enhancement like this.

[TomAugspurger]

This seems to have a different behavior now.

In [38]: df.pivot_table(columns='name', values='value')
Out[38]:
name       p   temp
value  100.5  10.75

Closing, since we won't break compatibility over this.