同一数据帧 Pandas Python 中的交叉引用列

Question

** I have edited the sample df for two columns to be tuples instead of integers to illustrate the problem I have with the solution once I change the data from integers to tuples **

I am trying to create a new column in Pandas whose value will depend value of a specific column being present on a different row on a separate columns, and where a match is found, use the value of a third columns.

To illustrate, see the below example.

I am using a lambda function in df.apply() to do the following: in the first row, it will filter for rows where the value of column 'two' equals the value of column 'zero', and where it does, it takes the value of column 'one' and copies it into new column 'three'.

df = pd.DataFrame([[(0,9),(1,9),(2,9),(3,9),(4,9)],['a','b','c','d','e'],[(2,9),(3,9),(4,9),(5,9),(6,9)]]).transpose()

df.columns = ['zero','one','two']

df['three] = df.apply(lambda x : df[df['zero'] == x['two']].loc[:,'one'], axis=1)

Note, column 'two' and column 'zero' are unique, so the filter result will one ever have one row.

In theory, the result of column 'three' should be : 'c', 'd', 'e', 'nan', 'nan'.

Thank you

Answer 1

Just set row zero as the index for convenient lookup of column one .

Update: the solution now works for tuple indexes.

import pandas as pd
import numpy as np

df = pd.DataFrame([[0,1,2,3,4],['a','b','c','d','e'],[2,3,4,5,6]]).transpose()
df.columns = ['zero','one','two']

# set index for quick lookup    
df_indexed = df.set_index("zero")

# the indexed dataset look like this
df_indexed
Out[21]: 
     one two
zero        
0      a   2
1      b   3
2      c   4
3      d   5
4      e   6

# apply the mapping logic, taking df_indexed from outside the function
def f(el):
    return df_indexed.at[el, "one"] if el in df_indexed.index else np.nan

df["three"] = df["two"].apply(f)

print(df)
Out[18]: 
  zero one two three
0    0   a   2     c
1    1   b   3     d
2    2   c   4     e
3    3   d   5   NaN
4    4   e   6   NaN

# On the updated dataset
df
Out[71]: 
     zero one     two three
0  (0, 9)   a  (2, 9)     c
1  (1, 9)   b  (3, 9)     d
2  (2, 9)   c  (4, 9)     e
3  (3, 9)   d  (5, 9)   NaN
4  (4, 9)   e  (6, 9)   NaN