Python 在Pandas 数据框中堆叠多列

您可以使用带有临时MultiIndex的自定义整形:

out = (df
   .set_index(['date', 'other_col'])
   .pipe(lambda x: x.set_axis(x.columns.str.split('_', expand=True), axis=1))
   .rename_axis(columns=['side', None])
   .stack('side').reset_index()
)

或melt+pivot:

tmp = df.melt(['date', 'other_col'], var_name='side')
tmp[['side', 'col']] = tmp['side'].str.split('_', n=1, expand=True)

out = (tmp.pivot(index=['date', 'other_col', 'side'],
                 columns='col', values='value')
          .reset_index().rename_axis(columns=None)
      )

输出:

         date other_col   side  count  sum
0  2023-12-01         a   left      1    0
1  2023-12-01         a  right      4    2
2  2023-12-05         b   left      8    8
3  2023-12-05         b  right      7    3
4  2023-12-07         c   left      5    4
5  2023-12-07         c  right      9    5

或者，使用janitor库和pivot_longer库要容易得多:

# pip install pyjanitor
import janitor

out = df.pivot_longer(index=['date', 'other_col'],
                      names_to=('side', '.value'),
                      names_pattern=r'([^_]+)_([^_]+)')

输出:

         date other_col   side  count  sum
0  2023-12-01         a  right      4    2
1  2023-12-05         b  right      7    3
2  2023-12-07         c  right      9    5
3  2023-12-01         a   left      1    0
4  2023-12-05         b   left      8    8
5  2023-12-07         c   left      5    4