How to join DataFrames in Pandas with consideration of RegEx?-CodePudding

I have two pandas DataFrames, df1 and df2.

As shown in the first DataFrame df1, the entries in the column "Box", which show the assignment of different boxes to a hand of people, can contain RegEx-values:

df1:

   Person        Box
0    Alex      Box 1
1   Linda      Box 3
2   David     Box .*
3  Rachel  Box [1-2]

df2:

     Box       Item  Qty.
0  Box 1      Apple     4
1  Box 1  Blueberry    12
2  Box 2      Lemon     1
3  Box 2     Papaya     2
4  Box 3      Apple     2

How do I join the DataFrames on the common column "Box" in a way that these regEx-values are evaluated correctly (as regex are not supported in pandas' join / merge-functions), so that I receive the following DataFrame as a result:

    Person    Box       Item  Qty.
0     Alex  Box 1      Apple     4
1     Alex  Box 1  Blueberry    12
2    Linda  Box 3      Apple     2
3    David  Box 1      Apple     4
4    David  Box 1  Blueberry    12
5    David  Box 2      Lemon     1
6    David  Box 2     Papaya     2
7    David  Box 3      Apple     2
8   Rachel  Box 1      Apple     4
9   Rachel  Box 1  Blueberry    12
10  Rachel  Box 2      Lemon     1
11  Rachel  Box 2     Papaya     2

I already tried to achieve this via list comprehensions, which leads to correct results, but drops columns of the left DataFrame.

def joinWithRegEx(left: pd.DataFrame, right: pd.DataFrame, left_on: str, right_on: str):

    df = pd.DataFrame
    df = pd.concat([right[right[right_on].str.match(entry)] for entry in left[left_on]], ignore_index=True)
    '''
    Left-Join of two DataFrame with considered Rege
    '''
    return df

I would have actually thought that this would be a more common usecase, or is Pandas simply not the best choice for these kind if tasks?

CodePudding user response：

Using df.iterrows it is possible to store the remaining columns of the left dataframe and reassign them after applying the REGEX filter. See verbose explanation in the commentaries of the updated function.

def joinWithRegEx(left: pd.DataFrame, right: pd.DataFrame, left_on: str, right_on: str):

    df = pd.DataFrame()
    # iterate over the left dataframe
    for idx, row in left.iterrows():
        # store REGEX pattern
        regex = row[left_on]
        # store remaining columns
        preserved_values = row.drop(left_on)
        # filter the right dataframe
        match = right[right[right_on].str.match(regex)]
        # add the preserved remaining columns
        match = match.assign(**preserved_values)
        # append to the final dataframe
        df = pd.concat([df, match], ignore_index=True)

    '''
    Left-Join of two DataFrame with considered Regex
    '''
    return df

CodePudding user response：

There are several other answers for this use case. Borrowing from: Can I perform a left join/merge between two dataframes using regular expressions with pandas?

import pandas as pd

df1 = pd.DataFrame([
    ['Alex','Box 1'],
    ['Linda','Box 3'],
    ['David','Box .*'],
    ['Rachel','Box [1-2]'], 
], columns=['Person','Box'])

df2 = pd.DataFrame([
    ['Box 1','Apple',4],
    ['Box 1','Blueberry',12],
    ['Box 2','Lemon',1],
    ['Box 2','Papaya',2],
    ['Box 3','Apple',2]
],columns=['Box','Item','Qty.'])

import re
def merge_regex(df1, df2,left_on,right_on):
    idx = [(i,j) for i,r in enumerate(left_on) for j,v in enumerate(right_on) if re.match(r,v)]
    df1_idx, df2_idx = zip(*idx)
    t = df1.iloc[list(df1_idx)].reset_index(drop=True)
    t1 = df2.iloc[list(df2_idx)].reset_index(drop=True)
    return pd.concat([t,t1],axis=1)

df = merge_regex(df1,df2,df1.Box,df2.Box)

#remove duplicate column
df = df.T.groupby(level=0).last().T

#reorder to match desired output
df = df[['Person','Box','Item','Qty.']]

df
    Person  Box Item    Qty.
0   Alex    Box 1   Apple   4
1   Alex    Box 1   Blueberry   12
2   Linda   Box 3   Apple   2
3   David   Box 1   Apple   4
4   David   Box 1   Blueberry   12
5   David   Box 2   Lemon   1
6   David   Box 2   Papaya  2
7   David   Box 3   Apple   2
8   Rachel  Box 1   Apple   4
9   Rachel  Box 1   Blueberry   12
10  Rachel  Box 2   Lemon   1
11  Rachel  Box 2   Papaya  2