sort_values() in Pandas behaves contrary to documentation

Question

I am puzzled with the behavior of sort_values() in Pandas which does not seem to respond appropriately to the axis argument.

For a toy example:

toy.to_json()
'{"labels":{"0":7,"1":4,"2":7,"3":1,"4":5,"5":0,"6":3,"7":1,"8":4,"9":9},"companies":{"0":"Apple","1":"AIG","2":"Amazon","3":"American express","4":"Boeing","5":"Bank of America","6":"British American Tobacco","7":"Canon","8":"Caterpillar","9":"Colgate-Palmolive"}}'

toy.sort_values('labels') # this works alright
labels  companies
5   0   Bank of America
3   1   American express
7   1   Canon
6   3   British American Tobacco
1   4   AIG
8   4   Caterpillar
4   5   Boeing
0   7   Apple
2   7   Amazon
9   9   Colgate-Palmolive

toy.sort_values(by = 'labels', axis = 1) # Returns an exception
KeyError: 'labels'

Answer 1

这是因为在示例中，轴0为“下”，轴为“右”（即跨列）。如果查看sort_values的文档，则会看到第一个参数的确是by ，而默认值是axis为0。因此，重复您的第一个示例，您需要执行toy.sort_values(by='labels', axis=0)

Answer 2

Adding on an example to the above comments and answers:

Lets assume you had a dataframe as below:

df = pd.DataFrame(data={"labels":{"0":7,"1":4,"2":7,"3":1,"4":5},"companies":{"0":9,"1":1,"2":6,"3":1,"4":8}})
>>df
    labels  companies
0   7       9
1   4       1
2   7       6
3   1       1
4   5       8

For axis=0 , it would sort when you pass a index levels and/or column labels as:

df.sort_values(by='labels')

which gives you a sorted label column (ascending by default).

   labels   companies
3   1       1
1   4       1
4   5       8
0   7       9
2   7       6

Coming to axis=1 , refer to the below code:

df.sort_values('4',axis=1)

This will sort the columns in a way the index 4 is sorted. Here it wont change anything since for index 4 since 5 is less than 8 and by default the sorting is ascending . However if you execute df.sort_values('1',axis=1) where the value under label is more than companies , you will see that the position of labels and companies has been exchanged.

    companies   labels
0   9           7
1   1           4
2   6           7
3   1           1
4   8           5

Hope this clarifies.

Answer 3

Just to get understanding around axis and rows to clear when we choose axis=1 or axis=0 .

df.shape[0]  # gives number of row count
df.shape[1]  # gives number of col count

Let's assume a dataFrame as follow:

>>> df = pd.DataFrame({
...     'col1' : ['A', 'A', 'B', np.nan, 'D', 'C'],
...     'col2' : [2, 1, 9, 8, 7, 4],
...     'col3': [0, 1, 9, 4, 2, 3],
... })

>>> df
  col1  col2  col3
0    A     2     0
1    A     1     1
2    B     9     9
3  NaN     8     4
4    D     7     2
5    C     4     3

So, applying the df.shape and see how it turns around the columns & rows:

>>> df.shape[0]
6            <-- Here, we have six row into the dataFrame
>>> df.shape[1]
3            <-- Here, we have three columns into the dataFrame

Now if you are just sorting the value by column name hence you don't need to specify axis=1 because column name already been specified, you can do simply :

>>> df.sort_values(by=['col1'])
  col1  col2  col3
0    A     2     0
1    A     1     1
2    B     9     9
5    C     4     3
4    D     7     2
3  NaN     8     4

or, you can pass multiple column names as a list with by :

>>> df.sort_values(by=['col1', 'col2'])
  col1  col2  col3
1    A     1     1
0    A     2     0
2    B     9     9
5    C     4     3
4    D     7     2
3  NaN     8     4