Bug in DataFrame()?

[Richard Stanton]

When using DataFrame() with a dataframe as input, trying to set the column
names results in very odd behavior. Here's an example:

df = DataFrame({'A' : [1,2,3], 'B' : [1,2,3]})
DataFrame(df, columns=['C','D'])

Result:

C D
0 NaN NaN
1 NaN NaN
2 NaN NaN


Of course I can always reset the column names in a separate statement, but
pandas makes me want to do as much as possible in a single line.

[Jeff Reback]

this is equivalent of

df.reindex(columns=['C','D'])

so it is correct and not a bug

you are reindexing to columns which are empty

> --
> You received this message because you are subscribed to the Google Groups "PyData" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to pydata+un...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

[Paul Hobson]

On Tue, May 21, 2013 at 2:42 PM, Jeff Reback <jeffr...@gmail.com> wrote:

this is equivalent of

df.reindex(columns=['C','D'])

so it is correct and not a bug

you are reindexing to columns which are empty

On May 21, 2013, at 5:34 PM, Richard Stanton <sta...@haas.berkeley.edu> wrote:

> When using DataFrame() with a dataframe as input, trying to set the column
> names results in very odd behavior. Here's an example:
>
> df = DataFrame({'A' : [1,2,3], 'B' : [1,2,3]})
> DataFrame(df, columns=['C','D'])
>
> Result:
>
>    C   D
> 0 NaN NaN
> 1 NaN NaN
> 2 NaN NaN
>
>
> Of course I can always reset the column names in a separate statement, but
> pandas makes me want to do as much as possible in  a single line. 

 

So with version 0.11, I can do this:

import pandas

df = pandas.DataFrame({'A' : [1,2,3], 'B' : [1,2,3]}).rename(columns={'A':'C', 'B':'D'}) 

df

   C  D

0  1  1

1  2  2

2  3  3