groupby中的值之间的差异

Question

I have 2 data frames, one for export and one for import. I concatenated the export and import figure in a single data frame using pd.concat() .

table3 = pd.concat([table1,table2],keys=['table1','table2'])

The output is :

SRI LANKA DSR
                  count   sumavlue
table1 194  SRI LANKA DSR  139571409
table2 185  SRI LANKA DSR   1803152


ST HELENA
                count  sumavlue
table1 195  ST HELENA        24
table2 186  ST HELENA         0


ST KITT N A
                  count  sumavlue
table1 196  ST KITT N A         0
table2 187  ST KITT N A         0

Now I need to calculate the difference between first and second row of each country and get a new column (rename version). How can I get this?

I need to have (export- import) as diff (new column name ) for each country.

• For Srilanka it will be 139571409-1803152 =XXXXXX
• For ST HELENa it will be 24-0 =24
• and so on for other countries

In the above example

Answer 1

You could try using a combination of groupby, unstack, and stack. I'm not sure what your column names are so I assumed some liberties. Here is my work:

# Make DataFrame
df = pd.DataFrame({'country' : ['Sri Lanka DSR']*2 + ['St Helena']*2 + ['St Kitt']*2,
                   'table' : ['table1', 'table2']*3,
                   'ID' : [194, 185, 195, 186, 196, 187],
                   'sumvalue' : [139571409, 1803152, 24, 0, 0, 0]})

# Groupby 'country', 'table', 'ID';
# unstack 'table', 'ID' and take reverse difference on the columns;
# stack 'table', 'ID' and rename 'sumvalue' to 'diff'
df.groupby(['country',
            'table',
            'ID']).agg({'sumvalue' : 'sum'}).unstack(['table',
                                                      'ID']).\
            diff(periods = -1,
                 axis = 1).stack(['table',
                                  'ID']).rename(columns = {'sumvalue' : 'diff'})