IIUC，这应该可以实现你想要的：

new_df = pd.DataFrame({'Date': daily_df['Date'], 'PWH': None, 'PWL': None, 'CWH': None, 'CWL': None})

new_df['Date'] = pd.to_datetime(new_df['Date'])
new_df['Week'] = new_df['Date'].dt.week

# Compute the cumulative high and low for the previous week, and fill in the appropriate values
for week in new_df['Week'].unique():
    if week == new_df['Week'].min():
        # Skip the first week, since there is no previous week
        continue
    prev_week_mask = new_df['Week'] == (week - 1)
    curr_week_mask = new_df['Week'] == week
    prev_week_high = daily_df.loc[prev_week_mask, 'High'].max()
    prev_week_low = daily_df.loc[prev_week_mask, 'Low'].min()
    new_df.loc[curr_week_mask, 'PWH'] = prev_week_high
    new_df.loc[curr_week_mask, 'PWL'] = prev_week_low

# Compute the current week high and low, and fill in the appropriate values
for week in new_df['Week'].unique():
    week_mask = new_df['Week'] == week
    start_date = new_df.loc[week_mask, 'Date'].min()
    for i, row in new_df.loc[week_mask].iterrows():
        if row['Date'] == start_date:
            # For the starting date of the week, set the current week high and low to None
            new_df.loc[i, 'CWH'] = None
            new_df.loc[i, 'CWL'] = None
        else:
            # For other dates in the week, set the current week high and low to the max/min of the previous values and the current values
            prev_high = new_df.loc[i-1, 'CWH']
            prev_low = new_df.loc[i-1, 'CWL']
            curr_high = daily_df.loc[daily_df['Date'] == row['Date'], 'High'].values[0]
            curr_low = daily_df.loc[daily_df['Date'] == row['Date'], 'Low'].values[0]
            new_df.loc[i, 'CWH'] = max(prev_high, curr_high)
            new_df.loc[i, 'CWL'] = min(prev_low, curr_low)

# Drop the Week column to match expected output
new_df.drop(columns=['Week'], inplace=True)

这是我的解决方案

df['Date'] = pd.to_datetime(df['Date'])
df = df.set_index('Date').sort_index()

#compute weekly min and max
week_max = df.resample('W')['High'].max()
week_min = df.resample('W')['Low'].min()

#compute weekly min and max
max_min = (df.resample('W')[['High', 'Low']]
           .agg({'High': 'max', 'Low': 'min'})
           .rename({'Low': 'PWL', 'High':'PWH'}, axis=1)
           .shift(1)
           )
max_min.index = max_min.index - timedelta(weeks=1)

#merge_asof will perform a join based on the closest date, in the backward direction
df = pd.merge_asof(df, max_min, left_index=True, right_index=True, direction='backward')

#compute cumulative min and max in each week
df[['year','week', 'day']] = df.index.isocalendar()
#we also group by year in case we have multi-year data

#set `CWH` and `CWL` to the `High` and `Low` value of the first day of the week
df['CWH'] = df.groupby(['year','week'])['High'].transform('first')
df['CWL'] = df.groupby(['year','week'])['Low'].transform('first')

#perform max and min between the first day and current day
df['CWH'] = df[['High','CWH']].max(axis=1)
df['CWL'] = df[['Low','CWL']].min(axis=1)

#removing the first day of each week (weekdays were already sorted for merge_asof)
df.loc[df.groupby(['year','week']).head(1).index, ['CWH', 'CWL']] = None
df = df.drop(['High', 'Low', 'year', 'week', 'day', 'Week Day'], axis=1)