include xr-ed versions of properscorings crps_gaussian, crps_ensemble

[aaronspring]

• probabilistic.py: xr.apply_ufunc on properscoring.crps_gaussian and properscoring.crps_ensemble and properscoring.threshold_brier_score
• test_probabilistic.py:
  • test whether same numerical result as inside properscoring
  • test whether xr_crps_x returns chunked arrays when input dask
  • no testing of properscoring itself (aka. no np_probabilistic)

[raybellwaves]

LGTM.

Did you mean to leave the doc string for xr_crps_gaussian and xr_crps_ensemble empty?

Mind padding them out like https://github.com/raybellwaves/xskillscore/blob/master/xskillscore/core/deterministic.py#L10L33 ?

[aaronspring]

Stephan Hoyer suggested to have this wrapper in xskillscore.

[aaronspring]

also added threshold_brier_score. unfortunately I didnt get apply_ufunc for the different cases of threshold (scalar works, but for an 1d array threshold I get chunking errors.) However, the user can bypass this by looping over different threshold levels if thats needed. Ready to merge from my side.

[raybellwaves]

Thanks a lot. The only thing I can see is the long line here re. PEP8
https://github.com/aaronspring/xskillscore/blob/AS_probabilistic_properscoring/xskillscore/core/probabilistic.py#L70

Run flake8 on the codebase as there may be other lint in there

FYI I learnt this here: dask/dask-jobqueue#78 (comment)

[aaronspring]

sorry. forget to beautify.

[raybellwaves]

Cheers @aaronspring!