如何通过pandas在复杂条件下合并2个数据集

Question

I am trying to merge two datasets by using pandas.

this is master dataset:

id  num1   num2
0   5      8
1   2      9
2   8      7
3   9      6

this is another one:

id2  num1_min  num1_max  num2_min  num2_max
0    1         3         8         10       
1    3         6         6         10
2    7         9         6         9

the output that I expect:

id  num1   num2  id2
0   5      8     1
1   2      9     0  
2   8      7     2
3   9      6     2

I want to give id2 to the master. At that time, I want to left join them under the condition that the value of num1 is between num1_min / max and the value of num2 is between num2_min / max. Master dataset has only one id2 or null. So id2 will not be joined in duplicate.

Please advise me.

Answer 1

It is possible to accomplish this using boolean masking. ie find the index id2 in df that satisfy the join condition.

In [1]: import pandas as pd
In [2]: df
Out[2]: 
   id  num1  num2
0   0     5     8
1   1     2     9
2   2     8     7
3   3     9     6

In [3]: df1
Out[3]: 
   id2  num1_min  num1_max  num2_min  num2_max
0    0         1         3         8        10
1    1         3         6         6        10
2    2         7         9         6         9

#find id2 based on conditions
In [4]: df['id2'] = df.apply(lambda row: (((row['num1'] >= df1['num1_min']) &
                           (row['num1'] <= df1['num1_max'])) &
                          ((row['num2'] >= df1['num2_min']) &
                           (row['num2'] <= df1['num2_max']))).idxmax(), axis=1)

In [5]: df
Out[5]: 
   id  num1  num2  id2
0   0     5     8    1
1   1     2     9    0
2   2     8     7    2
3   3     9     6    2

above, I used apply to go through df rows, check each row against the condition then find the index in df1 satisfying the condition.

EDIT

Another way to find id2

df['id2'] = df.apply(lambda row: df1.loc[(((row['num1'] >= df1['num1_min']) &
                                           (row['num1'] <= df1['num1_max'])) &
                                          ((row['num2'] >= df1['num2_min']) &
                                           (row['num2'] <= df1['num2_max']))),
                                         'id2'].values[0], axis=1)