问题
I'm trying to merge two data frames based on a column present in both, keeping only the intersection of the two sets.
The desired result is:
foo bar foobar
x y z x j i x y z j i
a 1 2 a 9 0 a 1 2 9 0
b 3 4 b 9 0 b 3 4 9 0
c 5 6 c 9 0 c 5 6 9 0
d 7 8 e 9 0
f 9 0
My code that does not produce the desired result is:
pd.merge(foo, bar, how='inner', on='x')
Instead, the code seems to return:
foo bar foobar
x y z x j i x y z j i
a 1 2 a 9 0 a 1 2 9 0
b 3 4 b 9 0 b 3 4 9 0
c 5 6 c 9 0 c 5 6 9 0
d 7 8 e 9 0 e * * 9 0
f 9 0 f * * 9 0
(where * represents an NaN)
Where am I going wrong? I've already reached the third Google page trying to fix this an nothing works. Whatever I do I get an outer join, with all rows in both sets.
回答1:
Usually it means that you have duplicates in the column(s) used for joining, resulting in cartesian product.
Demo:
In [35]: foo
Out[35]:
x y z
0 a 1 2
1 b 3 4
2 c 5 6
3 d 7 8
In [36]: bar
Out[36]:
x j i
0 a 9 0
1 b 9 0
2 a 9 0
3 a 9 0
4 b 9 0
In [37]: pd.merge(foo, bar)
Out[37]:
x y z j i
0 a 1 2 9 0
1 a 1 2 9 0
2 a 1 2 9 0
3 b 3 4 9 0
4 b 3 4 9 0
来源：https://stackoverflow.com/questions/45471466/pandas-inner-merge-join-returning-all-rows