如何合并行数不同的DataFrame，提取LikeColumn值不等的行？

Hey Matt, great question! Since you're working with two DataFrames that share the same column names and count (but different row counts), here are a couple of straightforward ways to build your df3 with rows where the "LikeColumn" values don't overlap between the two dfs:

方法1：基于集合交集的筛选（简洁高效） #

This is my go-to for this kind of task—it's quick and easy to follow:

1. First, find all the "LikeColumn" values that exist in both DataFrames:

import pandas as pd

# Calculate common values in LikeColumn across both dfs
common_likes = set(df1['LikeColumn']).intersection(set(df2['LikeColumn']))

2. Filter each DataFrame to keep only rows where "LikeColumn" isn't in that common set:

# Filter df1: keep rows where LikeColumn is NOT in common values
df1_filtered = df1[~df1['LikeColumn'].isin(common_likes)]
# Filter df2 the same way
df2_filtered = df2[~df2['LikeColumn'].isin(common_likes)]

3. Combine the filtered results into your final df3:

df3 = pd.concat([df1_filtered, df2_filtered]).reset_index(drop=True)

The reset_index(drop=True) just cleans up the index so it's sequential instead of keeping the original indices from df1 and df2.

方法2：使用Outer Merge（适合追踪行来源） #

If you also want to know which original DataFrame each row in df3 came from, this method is perfect:

1. Do an outer merge on "LikeColumn" and add an indicator column to track source:

merged_df = pd.merge(
    df1, 
    df2, 
    on='LikeColumn', 
    how='outer', 
    indicator=True  # Adds a column showing which df the row came from
)

2. Filter rows that only exist in df1 (left_only) or only exist in df2 (right_only):

filtered_rows = merged_df[(merged_df['_merge'] == 'left_only') | (merged_df['_merge'] == 'right_only')]

3. Restore the original column names (merge adds _x/_y suffixes):

df3 = pd.DataFrame()
for col in df1.columns:
    # Fill missing values from one df with values from the other
    df3[col] = filtered_rows[f'{col}_x'].fillna(filtered_rows[f'{col}_y'])

# Optional: Keep the _merge column if you want to track row origin
# df3['source_df'] = filtered_rows['_merge'].replace({'left_only': 'df1', 'right_only': 'df2'})

Both methods will give you exactly what you need: a df3 containing all rows where the "LikeColumn" value doesn't appear in the other DataFrame. Pick whichever fits your workflow better!

内容的提问来源于stack exchange，提问作者Matt