对每组的pandas数据帧进行排序并保持所需的顺序

Question

我有一个数据框,如下所示

df = pd.DataFrame({
    "Junk":list("aaaaaabbbcccc"),
    "Region":['West','West','West','West','East','East','East','South','South','South','North','North','North'],
    "Sales":[1, 3, 4, 2, 4, 2, 5, 7, 9, 7, 5, 9, 5]
})

+------+--------+-------+
| Junk | Region | Sales |
+------+--------+-------+
| a    | West   |     1 |
| a    | West   |     3 |
| a    | West   |     4 |
| a    | West   |     2 |
| a    | East   |     4 |
| a    | East   |     2 |
| b    | East   |     5 |
| b    | South  |     7 |
| b    | South  |     9 |
| c    | South  |     7 |
| c    | North  |     5 |
| c    | North  |     9 |
| c    | North  |     5 |
+------+--------+-------+

我想做两件事

1. 根据每个Region对数据帧进行排序

我可以用下面的代码实现它

df.sort_values(by = ['Region','Sales'])


+------+--------+-------+
| Junk | Region | Sales |
+------+--------+-------+
| a    | East   |     2 |
| a    | East   |     4 |
| b    | East   |     5 |
| c    | North  |     5 |
| c    | North  |     5 |
| c    | North  |     9 |
| b    | South  |     7 |
| c    | South  |     7 |
| b    | South  |     9 |
| a    | West   |     1 |
| a    | West   |     2 |
| a    | West   |     3 |
| a    | West   |     4 |
+------+--------+-------+

但我想保留Region列的顺序.West应该是第一个,然后East,然后South,然后North

期望的输出

+--------+----------+---------+
|  Junk  |  Region  |  Sales  |
+--------+----------+---------+
|  a     | West     |       1 |
|  a     | West     |       2 |
|  a     | West     |       3 |
|  a     | West     |       4 |
|  a     | East     |       2 |
|  a     | East     |       4 |
|  b     | East     |       5 |
|  b     | South    |       7 |
|  c     | South    |       7 |
|  b     | South    |       9 |
|  c     | North    |       5 |
|  c     | North    |       5 |
|  c     | North    |       9 |
+--------+----------+---------+

2. 我只是要排序的Region = East和Region = North其余地区的应该是他们的方式

期望的输出:

+--------+----------+---------+
|  Junk  |  Region  |  Sales  |
+--------+----------+---------+
|  a     | West     |       1 |
|  a     | West     |       3 |
|  a     | West     |       4 |
|  a     | West     |       2 |
|  a     | East     |       2 |
|  a     | East     |       4 |
|  b     | East     |       5 |
|  b     | South    |       7 |
|  b     | South    |       9 |
|  c     | South    |       7 |
|  c     | North    |       5 |
|  c     | North    |       5 |
|  c     | North    |       9 |
+--------+----------+---------+

Answer 1

首先创建有序的分类列,然后排序:

order = ['West', 'East', 'South', 'North']
df['Region'] = pd.CategoricalIndex(df['Region'], ordered=True, categories=order)

df = df.sort_values(by = ['Region','Sales'])
print (df)
   Junk Region  Sales
0     a   West      1
3     a   West      2
1     a   West      3
2     a   West      4
5     a   East      2
4     a   East      4
6     b   East      5
7     b  South      7
9     c  South      7
8     b  South      9
10    c  North      5
12    c  North      5
11    c  North      9

使用map字典创建新列,顺序然后删除帮助列的解决方案:

order = {'West':1, 'East':2, 'South':3, 'North':4}

df = df.assign(tmp=df['Region'].map(order)).sort_values(by = ['tmp','Sales']).drop('tmp', 1)
print (df)
   Junk Region  Sales
6     a   West      1
0     a   West      2
7     a   West      3
8     a   West      4
2     a   East      2
1     a   East      4
3     b   East      5
4     b  South      7
9     c  South      7
5     b  South      9
10    c  North      5
12    c  North      5
11    c  North      9

第二个是必须按筛选行排序,但是指定numpy数组以防止数据对齐:

order = ['West', 'East', 'South', 'North']
df['Region'] = pd.CategoricalIndex(df['Region'], ordered=True, categories=order)

mask = df['Region'].isin(['North', 'East'])
df[mask] = df[mask].sort_values(['Region','Sales']).values
print (df)
   Junk Region  Sales
0     a   West      1
1     a   West      3
2     a   West      4
3     a   West      2
4     a   East      2
5     a   East      4
6     b   East      5
7     b  South      7
8     b  South      9
9     c  South      7
10    c  North      5
11    c  North      5
12    c  North      9

map 替代方案:

order = {'East':1, 'North':2}
df = df.assign(tmp=df['Region'].map(order))

mask = df['Region'].isin(['North', 'East'])
df[mask] = df[mask].sort_values(['tmp','Sales']).values
df = df.drop('tmp', axis=1)