combine two pandas dataframe into one dataframe “dict type cell” (pd.Panel deprecated)

问题

I'm trying to concat multiples pandas.DataFrame to be saved in a mongodb in just one collection, all the dataframes have the same index/columns and I wanted to save it, in just one document, using to_json() method. Having all the cells of the dataframe as dicts, its probably a good approach. To accomplish that I wanted to concat the dataframes like this:

df1:                
 index   A      B
 1     'A1'   'B1'
 2     'A2'   'B2'
 3     'A3'   'B3'

df2:
 index  A      B
 1    'a1'   'b1'
 2    'a2'   'b2'
 3    'a3'   'b3'

Expected solution:

df_sol:
 index    A                    B
 1        {d1:'A1', d2:'a1'}   {d1:'B1', d2:'b1'}
 2        {d1:'A2', d2:'a2'}   {d1:'B2', d2:'b2'}
 3        {d1:'A3', d2:'a3'}   {d1:'B3', d2:'b3'}

the aproach that im using is

pd.Panel(dict(d1=df1, d2=df2)).apply(pd.Series.to_dict, 0)

                              A                         B
index                                                    
1      {'d1': 'A1', 'd2': 'a1'}  {'d1': 'B1', 'd2': 'b1'}
2      {'d1': 'A2', 'd2': 'a2'}  {'d1': 'B2', 'd2': 'b2'}
3      {'d1': 'A3', 'd2': 'a3'}  {'d1': 'B3', 'd2': 'b3'}

but pd.Panel its deprecated DeprecationWarning : Panel is deprecated and will be removed in a future version.its there a workaround using just pandas? thanks!

Original Question

回答1:

Solution
pd.concat + other stuff

pd.Series(
    pd.concat([df1, df2], axis=1, keys=['d1', 'd2']).stack().to_dict('index')
).unstack()

                              A                             B
1  {'d1': ''A1'', 'd2': ''a1''}  {'d1': ''B1'', 'd2': ''b1''}
2  {'d1': ''A2'', 'd2': ''a2''}  {'d1': ''B2'', 'd2': ''b2''}
3  {'d1': ''A3'', 'd2': ''a3''}  {'d1': ''B3'', 'd2': ''b3''}

---

Explanation
I want to get [1, 2, 3] and ['A', 'B'] into the index and ['d1', 'd2'] as the columns.

I start with pd.concat

pd.concat([df1, df2], axis=1, keys=['d1', 'd2'])

         d1          d2      
          A     B     A     B
index                        
1      'A1'  'B1'  'a1'  'b1'
2      'A2'  'B2'  'a2'  'b2'
3      'A3'  'B3'  'a3'  'b3'

Which almost gets me there. If I follow that with a stack, it will drop the last level of the columns into the last level of the index:

pd.concat([df1, df2], axis=1, keys=['d1', 'd2']).stack()

           d1    d2
index              
1     A  'A1'  'a1'
      B  'B1'  'b1'
2     A  'A2'  'a2'
      B  'B2'  'b2'
3     A  'A3'  'a3'
      B  'B3'  'b3'

And this is what I want. From here I can use .to_dict('index')

pd.concat([df1, df2], axis=1, keys=['d1', 'd2']).stack().to_dict('index')

{(1, 'A'): {'d1': "'A1'", 'd2': "'a1'"},
 (1, 'B'): {'d1': "'B1'", 'd2': "'b1'"},
 (2, 'A'): {'d1': "'A2'", 'd2': "'a2'"},
 (2, 'B'): {'d1': "'B2'", 'd2': "'b2'"},
 (3, 'A'): {'d1': "'A3'", 'd2': "'a3'"},
 (3, 'B'): {'d1': "'B3'", 'd2': "'b3'"}}

And pass that back to the pd.Series constructor to get a series of dictionaries.

pd.Series(
    pd.concat([df1, df2], axis=1, keys=['d1', 'd2']).stack().to_dict('index')
)

1  A    {'d1': ''A1'', 'd2': ''a1''}
   B    {'d1': ''B1'', 'd2': ''b1''}
2  A    {'d1': ''A2'', 'd2': ''a2''}
   B    {'d1': ''B2'', 'd2': ''b2''}
3  A    {'d1': ''A3'', 'd2': ''a3''}
   B    {'d1': ''B3'', 'd2': ''b3''}
dtype: object

The only thing left to do is unstack which I show in the solution above.

回答2:

This is a completely different concept that I am having fun with.

---

You can create a subclass of dict where we define addition to be a dictionary merge.

from cytoolz.dicttoolz import merge

class mdict(dict):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

    def __add__(self, other):
        return(mdict(merge(self, other)))


df1.applymap(lambda x: mdict(d1=x)) + df2.applymap(lambda x: mdict(d2=x))

                                  A                             B
index                                                            
1      {'d1': ''A1'', 'd2': ''a1''}  {'d1': ''B1'', 'd2': ''b1''}
2      {'d1': ''A2'', 'd2': ''a2''}  {'d1': ''B2'', 'd2': ''b2''}
3      {'d1': ''A3'', 'd2': ''a3''}  {'d1': ''B3'', 'd2': ''b3''}

来源：https://stackoverflow.com/questions/46591824/combine-two-pandas-dataframe-into-one-dataframe-dict-type-cell-pd-panel-depre