Python pandas 数据框 sort_values 不适用于第二学期答案

【问题标题】：Python pandas dataframe sort_values does not work for second termPython pandas 数据框 sort_values 不适用于第二学期
【发布时间】：2017-10-29 20:23:14
【问题描述】：

我有一个包含销售数据的数据框

Order ID Order Date Order Priority  Order Quantity     Sales  
  928.0   1/1/2009           High            32.0    180.36 
10369.0   1/2/2009            Low            43.0  4,083.19
10144.0   1/2/2009       Critical            16.0    137.63             
32323.0   1/1/2009  Not Specified             9.0    872.48          
48353.0   1/2/2009       Critical             3.0    124.81         
51008.0   1/3/2009       Critical            15.0     85.56       
26756.0   1/2/2009       Critical            43.0     614.8         
18144.0   1/2/2009            Low             4.0  1,239.06         
22912.0   1/2/2009            Low            32.0  4,902.38
...

我想按日期（从旧到新）和销售额（从大到小）对值进行排序。我在 PyCharm Edu 3.5.1 (python 2.7) 中编写了这段代码：

df = pd.read_csv('sales.csv', header=0)
df['Order Date'] = pd.to_datetime(df['Order Date'])
df = df.sort_values(by=['Order Date', 'Sales'], ascending=[True, False])
print df.head(10)

输出：

    Order ID Order Date Order Priority  Order Quantity      Sales  
    32323.0 2009-01-01  Not Specified             9.0     872.48          
    928.0   2009-01-01           High            32.0     180.36          
    26756.0 2009-01-02       Critical            43.0      614.8         
    22912.0 2009-01-02            Low            32.0   4,902.38         
    10369.0 2009-01-02            Low            43.0   4,083.19          
    10144.0 2009-01-02       Critical            16.0     137.63          
    48353.0 2009-01-02       Critical             3.0     124.81          
    18144.0 2009-01-02            Low             4.0   1,239.06         
    29376.0 2009-01-03  Not Specified             4.0     896.49
...

“订单日期”列已正确排序，但“销售”列未按预期排序。 PyCharm 似乎忽略了 1000 分隔符的 , 值。我在这里错过了什么吗？

【问题讨论】：

似乎需要df = pd.read_csv('sales.csv', header=0, thousands=',') 将列Sales 转换为floats。

标签： python-2.7 sorting pandas dataframe

【解决方案1】：

使用read_csv 和参数thousands 删除浮点数中的,，使用parse_dates 将列转换为日期时间，因为Sales 列的值被读取为strings：

df = pd.read_csv('sales.csv', thousands=',', parse_dates=['Order Date'])
print (df)
   Order ID Order Date Order Priority  Order Quantity    Sales
0     928.0 2009-01-01           High            32.0   180.36
1   10369.0 2009-01-02            Low            43.0  4083.19
2   10144.0 2009-01-02       Critical            16.0   137.63
3   32323.0 2009-01-01  Not Specified             9.0   872.48
4   48353.0 2009-01-02       Critical             3.0   124.81
5   51008.0 2009-01-03       Critical            15.0    85.56
6   26756.0 2009-01-02       Critical            43.0   614.80
7   18144.0 2009-01-02            Low             4.0  1239.06
8   22912.0 2009-01-02            Low            32.0  4902.38

df = df.sort_values(by=['Order Date', 'Sales'], ascending=[True, False])
print (df)
   Order ID Order Date Order Priority  Order Quantity    Sales
3   32323.0 2009-01-01  Not Specified             9.0   872.48
0     928.0 2009-01-01           High            32.0   180.36
8   22912.0 2009-01-02            Low            32.0  4902.38
1   10369.0 2009-01-02            Low            43.0  4083.19
7   18144.0 2009-01-02            Low             4.0  1239.06
6   26756.0 2009-01-02       Critical            43.0   614.80
2   10144.0 2009-01-02       Critical            16.0   137.63
4   48353.0 2009-01-02       Critical             3.0   124.81
5   51008.0 2009-01-03       Critical            15.0    85.56

另一种解决方案是使用replace + astype 或to_numeric：

df['Order Date'] = pd.to_datetime(df['Order Date'])
df['Sales'] = df['Sales'].replace(',', '', regex=True).astype(float)
#if astype does not work because bad data
#df['Sales'] = pd.to_numeric(df['Sales'].replace(',', '', regex=True), errors='coerce')
df = df.sort_values(by=['Order Date', 'Sales'], ascending=[True, False])
print (df)
   Order ID Order Date Order Priority  Order Quantity    Sales
3   32323.0 2009-01-01  Not Specified             9.0   872.48
0     928.0 2009-01-01           High            32.0   180.36
8   22912.0 2009-01-02            Low            32.0  4902.38
1   10369.0 2009-01-02            Low            43.0  4083.19
7   18144.0 2009-01-02            Low             4.0  1239.06
6   26756.0 2009-01-02       Critical            43.0   614.80
2   10144.0 2009-01-02       Critical            16.0   137.63
4   48353.0 2009-01-02       Critical             3.0   124.81
5   51008.0 2009-01-03       Critical            15.0    85.56

【讨论】：