Question

我正在尝试找到一种基于列对数据排序的方法。我下面的当前代码非常接近，但我最终希望将Joe移到顶部 - 并将所有行保持在一起 - 因为他的总数最多。

更新1： 'Total'并不总是最大值 - 因此需要使用'Yes'指定 - （部分美元金额可能为负数）。

更新2 ：我的代码和所需的输出已更新，以显示'Total'行可能小于组中另一个'Dollar'的位置（由于负元），但它应该仍然是'Dude'组的第一行。

我的代码使分组正确，但最终不会对'Dude'组进行排序。

import pandas as pd

headers = ['Date','Dude','Dollar', 'Total']
df = pd.DataFrame({
    'Dude':['Bob','Bob','Sam','Bob','Joe','Joe','Joe','Bob','Sam','Sam','Joe','Sam'], 
    'Dollar':[4,1,-2,1,5,12,3,2,7,1,4,8],
    'Total':['Yes','No','No','No','No','Yes','No','No','Yes','No','No','No'],
    'Date':['1/1/2016','1/1/2016','1/1/2016','3/1/2016','3/1/2016','1/1/2016','1/1/2016','5/1/2016','1/1/2016','3/1/2016','5/1/2016','5/1/2016']
    }, columns = headers)

df['Date'] = pd.to_datetime(df['Date'])

df.sort_values(by = ['Dude','Total','Date'], ascending = [True, False, True], inplace = True)

输出：

         Date Dude  Dollar Total
0  2016-01-01  Bob       4   Yes
1  2016-01-01  Bob       1    No
3  2016-03-01  Bob       1    No
7  2016-05-01  Bob       2    No
5  2016-01-01  Joe      12   Yes
6  2016-01-01  Joe       3    No
4  2016-03-01  Joe       5    No
10 2016-05-01  Joe       4    No
8  2016-01-01  Sam       7   Yes
2  2016-01-01  Sam      -2    No
9  2016-03-01  Sam       1    No
11 2016-05-01  Sam       8    No

期望的输出：

         Date Dude  Dollar Total
5  2016-01-01  Joe      12   Yes
6  2016-01-01  Joe       3    No
4  2016-03-01  Joe       5    No
10 2016-05-01  Joe       4    No
8  2016-01-01  Sam       7   Yes
2  2016-01-01  Sam      -2    No
9  2016-03-01  Sam       1    No
11 2016-05-01  Sam       8    No
0  2016-01-01  Bob       4   Yes
1  2016-01-01  Bob       1    No
3  2016-03-01  Bob       1    No
7  2016-05-01  Bob       2    No

Answer 1

你可以设置＆＃39; Dude＆＃39;列作为具有所需排序的分类数据类型，然后按照之前的排序进行排序。这也可以让你拥有“老兄”的其他好处。列为分类。

# Get the ordering of Dudes based on max dollar.
dude_order = df[df['Total'] == 'Yes'].sort_values(by='Dollar', ascending=False)

# Set dude as categorical with the previously determined ordering.
df['Dude'] = df['Dude'].astype('category', categories=dude_order['Dude'], ordered=True)

# Sort the dataframe.
df = df.sort_values(by=['Dude', 'Total', 'Date'], ascending=[True, False, True])

结果输出：

         Date Dude  Dollar Total
5  2016-01-01  Joe      12   Yes
6  2016-01-01  Joe       3    No
4  2016-03-01  Joe       5    No
10 2016-05-01  Joe       4    No
8  2016-01-01  Sam       7   Yes
2  2016-01-01  Sam      -2    No
9  2016-03-01  Sam       1    No
11 2016-05-01  Sam       8    No
0  2016-01-01  Bob       4   Yes
1  2016-01-01  Bob       1    No
3  2016-03-01  Bob       1    No
7  2016-05-01  Bob       2    No

Answer 2

<强>更新

In [162]: m = df.loc[df.Total=='Yes'].set_index('Dude')['Dollar']

In [163]: m
Out[163]:
Dude
Bob     4
Joe    12
Sam     7
Name: Dollar, dtype: int64

In [164]: df.assign(x=df.Dude.map(m)) \
     ...:   .sort_values(['x','Dude','Total','Date'], ascending=[0,1,0,1]) \
     ...:   .drop('x', 1)
Out[164]:
         Date Dude  Dollar Total
5  2016-01-01  Joe      12   Yes
6  2016-01-01  Joe       3    No
4  2016-03-01  Joe       5    No
10 2016-05-01  Joe       4    No
8  2016-01-01  Sam       7   Yes
2  2016-01-01  Sam      -2    No
9  2016-03-01  Sam       1    No
11 2016-05-01  Sam       8    No
0  2016-01-01  Bob       4   Yes
1  2016-01-01  Bob       1    No
3  2016-03-01  Bob       1    No
7  2016-05-01  Bob       2    No

旧回答：

In [96]: df.assign(x=df.groupby('Dude').Dollar.transform('max')) \
    ...:   .sort_values(['x','Dude','Dollar','Date'], ascending=[0,1,0,1]) \
    ...:   .drop('x',1)
Out[96]:
         Date Dude  Dollar Total
5  2016-01-01  Joe      12   Yes
4  2016-03-01  Joe       5    No
10 2016-05-01  Joe       4    No
6  2016-01-01  Joe       3    No
8  2016-01-01  Sam       8   Yes
11 2016-05-01  Sam       5    No
2  2016-01-01  Sam       2    No
9  2016-03-01  Sam       1    No
0  2016-01-01  Bob       4   Yes
7  2016-05-01  Bob       2    No
1  2016-01-01  Bob       1    No
3  2016-03-01  Bob       1    No

Answer 3

我的解决方案......它首先找到所有“是”行，将它们合并回原始数据帧，然后先对它们进行排序。

import pandas as pd

headers = ['Date','Dude','Dollar', 'Total']
df = pd.DataFrame({
    'Dude':['Bob','Bob','Sam','Bob','Joe','Joe','Joe','Bob','Sam','Sam','Joe','Sam'], 
    'Dollar':[4,1,-2,1,5,12,3,2,7,1,4,8],
    'Total':['Yes','No','No','No','No','Yes','No','No','Yes','No','No','No'],
    'Date':['1/1/2016','1/1/2016','1/1/2016','3/1/2016','3/1/2016','1/1/2016','1/1/2016','5/1/2016','1/1/2016','3/1/2016','5/1/2016','5/1/2016']
    }, columns = headers)

df['Date'] = pd.to_datetime(df['Date'])

# Just the Total = Yes row for each dude, with dollar renamed to total_dollar
totals = df.loc[df['Total'] == 'Yes', ['Dude', 'Dollar']]
totals.columns = ['Dude', 'Total_Dollar']

# Merge back on dude, sort by total dollars before sorting by everything else 
df = df.merge(totals, on='Dude').sort_values(by = ['Total_Dollar', 'Dude', 'Total', 'Date'], ascending = [False, True, False, True])
del df['Total_Dollar']

输出：

         Date Dude  Dollar Total
9  2016-01-01  Joe      12   Yes
10 2016-01-01  Joe       3    No
8  2016-03-01  Joe       5    No
11 2016-05-01  Joe       4    No
5  2016-01-01  Sam       7   Yes
4  2016-01-01  Sam      -2    No
6  2016-03-01  Sam       1    No
7  2016-05-01  Sam       8    No
0  2016-01-01  Bob       4   Yes
1  2016-01-01  Bob       1    No
2  2016-03-01  Bob       1    No
3  2016-05-01  Bob       2    No

Pandas：对行组进行排序，而不是对各行进行排序

3 个答案: