Refactoring calendars to BaseCalendar class and subclasses

[BSchilperoort]

This PR refactors the current calendars into a BaseCalendar class, where most methods are implemented, and subclasses like AdventCalendar or MonthlyCalendar.

As an example I implemented two extra calendars; MonthlyCalendar: where a user inputs an anchor month, and a freq in the form 1M WeeklyCalendar: where a user inputs an anchor week (as in, week of year), and a freq in the form 1W

Additionally, the implementation of resample has been moved to a separate module, as s2spy.time had gotten too large. Resample now also works as discussed in #50

---

List of changes:

• Resample has also been refactored to a separate function, with its implementations in a separate module.
• Resample will require a mapped calendar as input along with the data that should be resampled.
• Calendars are based of a BaseCalendar class.
• a MonthlyCalendar and a WeeklyCalendar have been implemented, using calendar months and weeks numbers respectively.
• Intervals are not stored in the calendar object, but generated upon request, either with .show() or .get_intervals().
• Input data checks that had to be used both in s2spy.time and s2spy._base_calendar have been moved to s2spy.utils.

---

Demo of functionality

>>> from s2spy.time import MonthlyCalendar
>>> cal = MonthlyCalendar(anchor='Dec', freq='6M')
>>> cal
MonthlyCalendar(month=12, freq=6M, n_targets=1, max_lag=None)

>>> cal.map_years(2020, 2021)
>>> cal.show()
i_interval             (target) 0                     1
anchor_year                                            
2021         (2021 Jun, 2021 Dec]  (2020 Dec, 2021 Jun]
2020         (2020 Jun, 2020 Dec]  (2019 Dec, 2020 Jun]

>>> cal.get_intervals()
i_interval                          0                         1
anchor_year                                                    
2021         (2021-06-30, 2021-12-31]  (2020-12-31, 2021-06-30]
2020         (2020-06-30, 2020-12-31]  (2019-12-31, 2020-06-30]

>>> from s2spy.time import resample
>>> ds_resampled = resample(cal, ds) # Resampled to the intervals shown above

---

Still to do:

• [x] Update all docstrings
• [x] Add check to see if the data can fully cover the calendar, and warn the user if it cannot Out of scope for current PR see #67
• [x] Ensure that if the data is insufficient to cover the calendar, NaNs will be used for the empty intervals Works for xarray data, not for pandas data. Out of scope for this PR. See #67

Also:

• decide on setter/getter methods #69

---

This PR: Closes #33 Closes #41 Closes #50

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[sonarcloud[bot]]

Kudos, SonarCloud Quality Gate passed!   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
0 Code Smells

92.8% Coverage
0.0% Duplication

[Peter9192]

Thanks for addressing. Indeed, I think it's fine to merge this and restructure in another PR. One more comment that I forgot to write before: with the addition of a weekly and monthly calendar, I think the adventcalendar feels a bit awkward now. Shall we consider changing it to daily calendar? And/or shouldn't they be called NDayCalendar, NWeekCalendar, NMonthCalendar?

[BSchilperoort]

Hm. An alternative could also to have the users interface with a single calendar object, e.g. AdventCalendar (or CountdownCalendar or any other suitable name). Where the following would all be valid inputs;

daycalendar = AdventCalendar(anchor=(31, 12), freq='4d')
weekcalendar = AdventCalendar(anchor=40, freq='2W')
monthcalendar = AdventCalendar(anchor='Jan', freq='1M')