来自dicts的dicts列表的DataFrame

Question

I have a list of dicts where each key in the dict contains another dict :

In [256]: data_list
Out[256]: 
[{'1111': {'index': 602, 'prop_1': 0, 'prop_2': 1},
  '2222': {'index': 602, 'prop_1': 0, 'prop_2': 1}},
 {'1111': {'index': 603, 'prop_1': 0, 'prop_2': 0},
  '2222': {'index': 603, 'prop_1': 1, 'prop_2': 1}}]

In [257]: index = {i.pop('index') for x in data_list for i in x.values()}

In [258]: df = DataFrame(data_list, index=index)

In [259]: df
Out[259]: 
                             1111                          2222
602  {u'prop_1': 0, u'prop_2': 1}  {u'prop_1': 0, u'prop_2': 1}
603  {u'prop_1': 0, u'prop_2': 0}  {u'prop_1': 1, u'prop_2': 1}

How can I create the following or similar pandas.DataFrame ?

index1 index2    prop_1    prop_2
602    1111      0         1
       2222      0         1
603    1111      0         0
       2222      1         1 

Answer 1

Considering you can transform a series of dictionaries into a multiple columns data frame by doing .apply(pd.Series) , you can stack() your original data frame to a multiIndex Series and then use .apply(pd.Series) :

df.stack().apply(pd.Series)

#            prop_1  prop_2
# 602   1111      0       1
#       2222      0       1
# 603   1111      0       0
#       2222      1       1

Answer 2

Solution with concat , transpose by T and set_index :

df = pd.concat([pd.DataFrame(key) for key in data_list], axis=1)
       .T
       .set_index('index', append=True)

df.index = df.index.swaplevel(0,1)
df.index.names = ['index1','index2']
print (df)
               prop_1  prop_2
index1 index2                
602    1111         0       1
       2222         0       1
603    1111         0       0
       2222         1       1