连接后Pandas排序MultiIndex

Question

When I create a multi-indexed table in one shot, sortlevel() works as expected. However, if I concatenate multiple tables to create the same multi-indexed table, I cannot sortlevel() anymore. Full example below:

import pandas as pd
a=pd.DataFrame({'country':'Zimbabwe','name':'Fred'}, index=[1])
b=pd.DataFrame({'country':'Albania','name':'Jeff'}, index=[0])
not_working = pd.concat([a,b],keys=['second','first'])
working = pd.DataFrame({'country':['Zimbabwe','Albania'],'name':['Fred','Jeff']}, index=pd.MultiIndex.from_tuples([('second',1),('first',0)]))

not_working_sorted = not_working.sortlevel(0)
working_sorted = working.sortlevel(0)

I expect both of these to produce:

           country  name
first  0   Albania  Jeff
second 1  Zimbabwe  Fred

However, I only get that for "working". Anyone knows what I am doing wrong ? Using pandas 0.19.2

Answer 1

Try this ?

working.sort_index()
Out[702]: 
           country  name
first  0   Albania  Jeff
second 1  Zimbabwe  Fred

or to be more specific

working.sort_index(level=[0, 1], ascending=[True, False])

EDIT: Your multiple index labels show as below.

not_working.index
Out[565]: 
MultiIndex(levels=[['second', 'first'], [0, 1]],
           labels=[[0, 1], [1, 0]])
working.index
Out[566]: 
MultiIndex(levels=[['first', 'second'], [0, 1]],
           labels=[[1, 0], [1, 0]])

So if you want the not_working sort work :

not_working.sort_index(level=[0, 1], ascending=[False, False])
Out[567]: 
           country  name
first  0   Albania  Jeff
second 1  Zimbabwe  Fred

Answer 2

sortlevel() is deprecated so try to use sort_index()

not_working.sort_index(level = 1)

And

working.sort_index(level = 1)

You get

            country     name
first   0   Albania     Jeff
second  1   Zimbabwe    Fred