def merge_similar_rows()

in placement.py [0:0]

• 29 lines of code
• 10 McCabe index (conditional complexity)

def merge_similar_rows(df):
    df = df.copy()
    initial_length = len(df)

    df['db_name'] = df['table'].apply(lambda x: x.split('.')[0])
    df['table_name'] = df['table'].apply(lambda x: x.split('.')[1])

    removed_rows = []  # for debug

    for db_name in df['db_name'].unique():
        group_row = df[df['table'] == f'{db_name}.group']

        if group_row.empty or len(group_row) > 1:
            assert len(group_row) <= 1, "Multiple conflicting groups found"
            # print(f"[Merge Rows] {db_name}.group not found")
            continue  # Skip if no group or multiple conflicting groups

        group_z, group_w = group_row['z'].values[0], group_row['w'].values[0]
        if group_z == 0 and group_w == 0:
            print(f"[Merge Rows] {db_name}.group is replicated (z=w=0), skipped")
            continue

        similar_rows = df[(df['db_name'] == db_name) &
                          (df['z'] == group_z) & (df['w'] == group_w) &
                          (df['table'] != f'{db_name}.group')]
        if not similar_rows.empty:
            # Store removed rows before dropping
            removed_rows.append(similar_rows.copy())

            # update group size to add up
            group_size = similar_rows['size'].sum()
            df.loc[group_row.index, 'size'] += group_size
            df = df.drop(similar_rows.index)

    print(f"Merging rows reduces rows from {initial_length} to {len(df)}.")

    second_length = len(df)

    removed_rows.append(df[df['size'] == 0].copy())

    # Drop rows where size is 0
    df = df[df['size'] > 0]

    print(f"Removing rows with size 0 reduces rows from {second_length} to {len(df)}.")
    return df, pd.concat(removed_rows)