Change `dt.week()` to be more consistent with `dt.year()` (Gregorian Year)

[Queilow]

Description

Checks

• [x] I've verified this issue has not been raised before

Description

Currently Series.dt.week() returns the ISO 8601 definition of a week. Although this is a very consistent definition of a week, it could potentially lead users to accidentally pair it with dt.year() instead of dt.iso_year() which can cause difficult to catch mistakes. For example:

import polars as pl
from datetime import date

df = pl.DataFrame(
    {'date': [date(2022, 1, 1), date(2022, 12, 31)]}
)

df.with_columns(
    [
        pl.col('date').dt.year().alias('Gregorian year'),
        pl.col('date').dt.iso_year().alias('ISO year'),
        pl.col('date').dt.week().alias('ISO week'),
        pl.col('date').dt.to_string('%W').alias('Proposed Week format')
    ]
)

We can see both dates have the same (Gregorian year, isoweek) tuple despite being 12 months apart. I believe a week method compatible with Gregorian year is practical enough to be an explicit Method

Proposal

• Per the example above, I propose changing the definition of dt.week() to correspond with the %W format from chrono's strftime. This format is identical to ISO Week except all week 52/53 in January are turned to 0.
• Additionally I propose creating an explicit dt.iso_week() method to keep access to the current functionality

Discussion

Pros:

• If a user wants to group by a 'Gregorian week' they can use (year, week), rather than (year, iso_year, iso_week) currently (or create the column via string casting)
• Improves readability as users are explicitly using (year, week) or (iso_year, iso_week), and any situations where users are mix-matching the two formats are done consciously.
• The week value will match only on 7 days of the Gregorian year (e.g 2022 has 8 days for iso-week 52)
• The %W format is aligned with the current dt.weekday() method as well, as opposed to other week definition such as the one from lubridate in R

  Cons:

• This is a breaking change in the behaviour of dt.week()
• Would need to also change the output of DATEPART(date, 'week') in SQLContext, currently both 'week' and 'isoweek' return isoweek format

[Julian-J-S]

Yes, I like it and imo good idea to bring with/before 1.0!

Date logic is a very common use case and as you said current design is a bit confusing

• year -> calendar year ✅
• iso_year -> iso year ✅
• week -> iso week 🤔
• iso_week -> missing 😞