在熊猫中合并“一对一”数据框

Question

我有两个数据框：

df1 = pd.DataFrame(data = 
{'Invoice' : [1, 2, 3, 4, 5], 'Value' : [10, 25, 40, 10, 15]}) 
df2 = pd.DataFrame(data =
{'Invoice' : [2, 3, 5, 2], 'Value' : [25, 11, 15,25], 'TestData':["A",'B','C','D']})  

我已经合并了它们并得到df3：

df3=pd.merge(df1,df2, left_on=["Invoice","Value"], right_on=["Invoice","Value"])

Df3输出：

   Invoice    Value   TestData
0      2    25        A
1      2    25        D
2      5    15        C

我的问题是如何以“一对一”方式合并数据框（我的意思是-当2号发票中的2号发票仅出现一次（或通常较少）时，然后不要在其中创建2号发票的另一行合并的数据框）。我想得到这样的东西：

   Invoice    Value   TestData
0      2    25        A
1      5    15        C

或这个：

   Invoice    Value   TestData
0      2    25        D
1      5    15        C

我只尝试左右合并，但这不起作用-总是有两行发票编号为2。

谢谢你
Jarek

Answer 1

使用drop_duplicates与指定的列名，参数keep='last'是最后一行重复：

df2 = df2.drop_duplicates(["Invoice","Value"])
#same as
#df2 = df2.drop_duplicates(["Invoice","Value"], keep='first')
df3=pd.merge(df1,df2, on=["Invoice","Value"])
print (df3)
   Invoice  Value TestData
0        2     25        A
1        5     15        C

df2 = df2.drop_duplicates(["Invoice","Value"], keep='last')
df3=pd.merge(df1,df2, on=["Invoice","Value"])
print (df3)
   Invoice  Value TestData
0        2     25        D
1        5     15        C

编辑：

如果需要按所有行分组，则必须添加新列以确保唯一性：

df1['g'] = df1.groupby(['Invoice','Value']).cumcount()
df2['g'] = df2.groupby(['Invoice','Value']).cumcount()

print (df1)
   Invoice  Value  g
0        1     10  0
1        2     25  0
2        3     40  0
3        4     10  0
4        5     15  0

print (df2)
   Invoice TestData  Value  g
0        2        A     25  0
1        3        B     11  0
2        5        C     15  0
3        2        D     25  1

df3=pd.merge(df1,df2, on=["Invoice","Value", "g"]).drop('g', axis=1)
print (df3)
   Invoice  Value TestData
0        2     25        A
1        5     15        C