从DataFrame提取的列具有不同的索引

Question

I'm encountering the following situation:

some_df.index                 #=> Int64Index([0, 1], dtype='int64')
some_df['some_column'].index  #=> Float64Index([7.0, 5.0], dtype='object')

Why is this happening? Does this mean there was something wrong in the way some_df was constructed? Finally, what's the best way to ensure that columns I extract from some_df all use the same index as some_df itself?

EDIT: I dove deeper into the code and apparently there's a line that simply reassigns the index: some_df['some_column].index = some_df['another_column'] . How broken is this?

Answer 1

It's unclear whether this is a bug, though perhaps assigning to the Series index should raise (it may be quite tricky to get this behaviour)... You should definitely not be doing this!

To confirm that this is indeed the case:

In [11]: df = pd.DataFrame([[1, 2], [3, 4]], columns=['A', 'B'])

In [12]: df
Out[12]:
   A  B
0  1  2
1  3  4

In [13]: df['A'].index
Out[13]: Int64Index([0, 1], dtype='int64')

In [14]: df['A'].index = [7., 8.]

In [15]: df['A'].index
Out[15]: Float64Index([7.0, 8.0], dtype='float64')

In [16]: df
Out[16]:
   A  B
0  1  2
1  3  4

So whilst this is apparently valid, you're going to get some surprising (potentially undefined) behaviour...

For example:

In [21]: df.groupby("A").sum()
Out[21]:
Empty DataFrame
Columns: [B]
Index: []