python - 使用 Pandas 从两列获取值范围的优雅方法

Question

我有一个如下所示的数据框(运行下面的完整代码)

df1 = pd.DataFrame({'person_id': [11,21,31,41,51],
                        'date_birth': ['05/29/1967', '01/21/1957', '7/27/1959','01/01/1961','12/31/1961']})
df1 = df1.melt('person_id', value_name='date_birth')
df1['birth_dates'] = pd.to_datetime(df1['date_birth'])
df_ranges = df1.assign(until_prev_year_days=(df1['birth_dates'].dt.dayofyear - 1),
     until_next_year_days=((df1['birth_dates'] + pd.offsets.YearEnd(0)) - df1['birth_dates']).dt.days)
f = {'until_prev_year_days': 'min', 'until_next_year_days': 'min'}

min_days = df_ranges.groupby('person_id',as_index=False).agg(f)
min_days.columns = ['person_id','no_days_to_prev_year','no_days_to_next_year']
df_offset = pd.merge(df_ranges[['person_id','birth_dates']], min_days, on='person_id',how='inner')

请参阅下面关于我试图获得范围的内容

df_offset['range_to_shift'] = "[" + (-1 * df_offset['no_days_to_prev_year']).map(str) + "," + df_offset['no_days_to_next_year'].map(str) + "]"

虽然我的方法有效，但我想有什么更好、更优雅的方法来做同样的事情
请注意，对于来自 no_days_to_prev_year 的值, 我们必须加上前缀 minus标志
我希望我的输出如下所示

Answer 1

使用 DataFrame.mul 连同 DataFrame.to_numpy :

cols = ['no_days_to_prev_year', 'no_days_to_next_year']
df_offset['range_to_shift'] = df_offset[cols].mul([-1, 1]).to_numpy().tolist()

结果:

# print(df_offset)

   person_id birth_dates  no_days_to_prev_year  no_days_to_next_year range_to_shift
0         11  1967-05-29                   148                   216    [-148, 216]
1         21  1957-01-21                    20                   344     [-20, 344]
2         31  1959-07-27                   207                   157    [-207, 157]
3         41  1961-01-01                     0                   364       [0, 364]
4         51  1961-12-31                   364                     0      [-364, 0]

timeit性能结果:

df_offset.shape
(50000, 5)

%%timeit -n100
cols = ['no_days_to_prev_year', 'no_days_to_next_year']
df_offset['range_to_shift'] = df_offset[cols].mul([-1, 1]).to_numpy().tolist()

15.5 ms ± 464 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)