python pandas sorting by group

Question

Each row in my DataFrame is a user vote entry for a restaurant. The data look like

id   cuisine    
91   american   
3    american   
91   american   
233  cuban      
233  cuban      
2    cuban      

where id refers to the restaurant.

I want to get something like the following

american  91   100
          3    30
          12   10
cuban     233  80
          2    33
mexican   22   99
          8    98
          21   82

where the 2nd column is the id , and the 3rd column is the number of rows in the DataFrame for that id . In other words, sort by the number of rows, but group by cuisine. I tried

g = df.groupby(['cuisine', 'id'])
c = g.size().sort_values(ascending=False)

But the order of the cuisines is mixed.

Answer 1

is that what you want?

In [2]: df
Out[2]:
    id   cuisine
0   91  american
1    3  american
2   91  american
3  233     cuban
4  233     cuban
5    2     cuban

In [3]: df.groupby(['cuisine', 'id']).size()
Out[3]:
cuisine   id
american  3      1
          91     2
cuban     2      1
          233    2
dtype: int64

or as a data frame:

In [10]: df.groupby(['cuisine', 'id']).size().reset_index(name='count').sort_values(['cuisine', 'count'], ascending=[1,0])
Out[10]:
    cuisine   id  count
1  american   91      2
0  american    3      1
3     cuban  233      2
2     cuban    2      1

Answer 2

use value_counts after group_by followed by sort_index

# ascending=[1, 0] says True for level[0], False for level[1]
df.groupby('cuisine').id.value_counts().sort_index(ascending=[1, 0])

cuisine   id 
american  91     2
          3      1
cuban     233    2
          2      1
Name: id, dtype: int64