可以使用map
通過dict
:
d = df2.set_index('ID')['F'].to_dict()
print (d)
{1: 'c1', 2: 'c2', 3: 'c3'}
df1['F'] = df1['ID'].map(d)
print (df1)
ID A B C F
0 1 x y z c1
1 1 x y z c1
2 2 x y z c2
3 2 x y z c2
4 2 x y z c2
5 3 x y z c3
另一種解決方案是map
通過Series
:
s = df2.set_index('ID')['F']
print (s)
ID
1 c1
2 c2
3 c3
Name: F, dtype: object
df1['F'] = df1['ID'].map(s)
print (df1)
ID A B C F
0 1 x y z c1
1 1 x y z c1
2 2 x y z c2
3 2 x y z c2
4 2 x y z c2
5 3 x y z c3
時序:
#[60000 rows x 5 columns]
df1 = pd.concat([df1]*10000).reset_index(drop=True)
In [115]: %timeit pd.merge(df1, df2[['ID', 'F']],how='left')
100 loops, best of 3: 11.1 ms per loop
In [116]: %timeit df1['ID'].map(df2.set_index('ID')['F'])
100 loops, best of 3: 3.18 ms per loop
In [117]: %timeit df1['ID'].map(df2.set_index('ID')['F'].to_dict())
100 loops, best of 3: 3.36 ms per loop
In [118]: %timeit df1['ID'].map({k:v for k, v in df2[['ID', 'F']].as_matrix()})
100 loops, best of 3: 3.44 ms per loop
In [119]: %%timeit
...: df2.index = df2['ID']
...: df1['F1'] = df1['ID'].map(df2['F'])
...:
100 loops, best of 3: 3.33 ms per loop
你可能會看到我的問題有見地的問題的方法:http://stackoverflow.com/questions/43311266/merging-two-dataframe-on-column-and-index –