熊猫转置concat（）

Question

How to transpose a DataFrame returned by concat() ?

df = DataFrame([
            dict(a=1, b=10, c=41),
            dict(a=1, b=20, c=42),
            ])

concat([df, df]).T

I get :

AttributeError: 'DataFrame' object has no attribute 'dtypes' !

If I try :

concat([df, df]).T.to_dict()

I get :

RuntimeError: maximum recursion depth exceeded in cmp

I think this is related to the duplicates introduced by concat() in the index but didn't find a workaround.

Answer 1

You can specify hierarchical index using keys :

In [288]: concatenated = concat([df,df], keys=['first', 'second'])

In [289]: print concatenated.T
   first      second    
       0   1       0   1
a      1   1       1   1
b     10  20      10  20
c     41  42      41  42

In [290]: print concatenated.T.to_dict().values()
[{'a': 1, 'c': 41, 'b': 10}, {'a': 1, 'c': 41, 'b': 10}, {'a': 1, 'c': 42, 'b': 20}, {'a': 1, 'c': 42, 'b': 20}]