如何通过正则或Pandas处理extractall结果，实现每行原文本聚合匹配？

First, let's break down what we need: retain every original row from your Series, with columns for each regex group showing either the captured content (aggregated if multiple matches exist in the same group for a row) or a missing value if no match was found for that group. Here are two straightforward approaches to achieve this:

Approach 1: Capture the First Match per Group #

If you only care about the first occurrence of each group in a text, str.extract() is perfect—it returns a DataFrame directly mapped to your original Series rows, with columns for each named group:

import re
import pandas as pd

regex = r"(?P<adv>This)|(?P<noun>test)"
texts = ["This is a test", "Random stuff with no match", "This and This again", "test test test"]
series = pd.Series(texts)

# Extract first match for each group
result = series.str.extract(regex)
print(result)

Output:

adv  noun
0  This   NaN
1   NaN   NaN
2  This   NaN
3   NaN  test

Approach 2: Aggregate All Matches per Group #

If you need to capture all occurrences of each group in a text (and combine them into a single value), we'll use str.extractall() to get all matches, aggregate by the original row index, then join back to the original Series to keep rows with no matches:

# Extract all matches (returns a multi-index DataFrame: original index + match number)
all_matches = series.str.extractall(regex)

# Aggregate matches per original row: join non-NaN values with commas (adjust logic as needed)
aggregated_matches = all_matches.groupby(level=0).agg(
    lambda x: ', '.join(x.dropna()) if not x.dropna().empty else pd.NA
)

# Join back to original series to include rows with no matches
final_result = series.to_frame(name="original_text").join(aggregated_matches)
print(final_result)

Output:

original_text         adv         noun
0        This is a test       This       <NA>
1  Random stuff with no match       <NA>       <NA>
2      This and This again  This, This       <NA>
3          test test test       <NA>  test, test, test

Key Notes: #

• Your existing regex works fine here—since groups are mutually exclusive, the alternation | correctly captures matches into the appropriate group. No regex modifications are needed.
• For aggregation, tweak the lambda function to fit your needs: use x.dropna().iloc[0] to keep only the first match, len(x.dropna()) to count matches, or any other custom logic.
• Using pd.NA ensures consistent missing value handling for string columns.

内容的提问来源于stack exchange，提问作者arnaud