使用apply（）合并熊猫DataFrame以仅在两列的部分匹配中进行合并

Question

I need to merge two pandas DataFrames but not only on exact column values, but also on approximate ones.

For example, I have these two DataFrames:

import pandas as pd
d = {'col1': ["a", "b", "c", "d"], 'col2': [3, 4, 66, 120]}
df = pd.DataFrame(data=d)

    col1    col2
0   a       3
1   b       4
2   c       66
3   d       120

d2 = {'col1a': ["aa", "bb", "cc", "dd"], 'col2b': [3, 4, 67, 100]}
df2 = pd.DataFrame(data=d2)
    col1a   col2b
0   aa      3
1   bb      4
2   cc      67
3   dd      100

Now, if I simply join them on col2 and col2b columns, I will only get two rows where the column values are exactly the same.

pd.merge(df, df2, how='inner', left_on='col2', right_on='col2b')
    col1    col2    col1a   col2b
0   a       3       aa      3
1   b       4       bb      4

Now, say for the simplicity of an example, I also want to merge column values based on the integer that is either +1 or -1 of the integer value from the left DataFrame. In our example in the left DataFrame the value 66 should be matched to 67 to the value from the right DataFrame in addition to the rows with values 3 and 4 :

        col1    col2    col1a   col2b
    0   a       3       aa      3
    1   b       4       bb      4
    2   c       66      cc      67

Not sure how to approach this problem, somehow would need to merge based on the approximated column values using apply() ?

Answer 1

Here is one way from merge_asof

pd.merge_asof(df,df2,left_on='col2',right_on='col2b',tolerance = 1,direction ='nearest').dropna()
Out[7]: 
  col1  col2 col1a  col2b
0    a     3    aa    3.0
1    b     4    bb    4.0
2    c    66    cc   67.0