Python core - add additional unit tests

[stijnvanhoey]

@Sachagobeyn, as I do not have the user rights to adjust the open PR directly (it is not my original PR anymore), this PR adds a number of unit tests to the open PR python-core:

• test_process, unit tests to test the different refactored process functions. I did some minor adjustments on the functionality itself (additional error checks, align rain load functions,...) as well.
• test_rfactor, add a benchmark check unit test with two unit tests: 1/ for a number of years/stations, the data is compared to the reference data and 2/ a full benchmark run, marked as 'slow' ( as this one takes a while).

The second part is the shortest in implementation, but the most important one. However, based on https://github.com/cn-ws/rfactor/pull/26#issue-712335495 it is for me not clear which of the files tests/data/testdata_maximum_intensity.csv and/or tests/data/testdata_maximum_intensity_matlab_clone.csv were created by which scripts/routine, which is the Matlab benchmark output to compare with to check the correct implementation and what should be the difference between both reference files?

Apart from that, there were some elements in the example notebook, e.g. cwd = Path().resolve() that should not be there, please verify that the adjustments were correct and appropriate.

[stijnvanhoey]

With respect to the current benchmark implementation, we still have a difference in the unit test for ("P05_039", 2017) which is due to the inclusion/exclusion of the event_rain_cum equal to 1.27 with some float approximation difference. I did added the round(rain_events["event_rain_cum"], 2) > event_threshold - round - to the comparison to solve a similar issue in ("P01_003", 2020) where this value was added in my calculation, but not in the benchmark data set. However, in this case it is vice-versa and there should be a clear way of handling these boundary-values (triggered by numerical approximation).

I left the round in the code, but I leave it to you to decide on how to handle this.

Note both 'troubleshooters' are part of the parameterized version of the benchmark for easier comparison, see test test_rfactor_benchmark_single_year

[Sachagobeyn]

Hi Stijn,

Thanks for the help,

Things I'll tackle before going to release

• [ ] Check test data with rounding of total event depth.
• [ ] Clean-up jupyter notebook
  • [ ] check joblib (in environment.yml)?
  • [ ] define rfactor_all in notebook.
  • [ ] clean-up last section of notebook.
• Docs build / folder
  • [ ] Read docs.
  • [ ] Restructure matlab scripts to src/rfactor/matlab, add info on location on code_legacy.
  • [ ] Add new results to code_legacy for Flanders.
• Release
  • [ ] Tag/release.
  • [ ] Adjust CHANGELOG.rst in master.