We have this df:

df = pd.DataFrame({
        'date': [pd.Timestamp('2020-12-26'), # week 52 of year 2020
                 pd.Timestamp('2020-12-27'), # last day of week 52 of year 2020
                 pd.Timestamp('2021-03-10'), # week 10 of year 2021
                 pd.Timestamp('2022-01-03'), # first day of week 1 of year 2022
                 pd.Timestamp('2022-01-09')], # last day of week 1 of year 2022
        'value' : [15, 15.5, 26, 36, 36.15]

})

We want a new df that looks so:

      date  value
0   202052  15.50
1   202201  36.15

In other words we need to:

1. convert 'date' to format year/week number (and store result as an object)
2. select only rows which date correspond to the last day of the week

Note both (1) and (2) need to be done following ISO-8601 definition of weeks. Actual dataset has thousands of rows.

How do we do this?

CodePudding user response：

You can work directly on the series by using the dt call on the column to transform the format of the date. To find if it is the last day of the week, Sunday corresponds to 7 so we can do an equality check.

iso = df.date.dt.isocalendar()
mask = iso.day == 7
df.loc[mask].assign(date=iso.year.astype(str)   iso.week.astype(str).str.rjust(2, "0"))

     date  value
1  202052  15.50
4  202201  36.15