Checking that we have all years

[juliettelavoie]

Generic Issue

simulation/raw/CMIP6/ScenarioMIP/global/CAS/FGOALS-g3/ssp126/r1i1p1f1/day/tasmax is missing years 2015-2019 but xscen didn't warn me of this or fail at any point. I figured it out because the diagnostics looked weird.

Should search/extract verify that all years are present in the period given or do we want miranda to do that check ?

[Zeitsperre]

I think it should be within xscen that a check like this happens.

There are cases where there are more years for some models/runs than others. Expectations would need to be coded for each case in miranda, whereas within xscen, ensuring that a timeseries is complete for the period of interest seems a lot more fitting (and simpler).

[RondeauG]

What did you do, exactly ? _subset_file_coverage should have caught that, but you might have found a bug.

EDIT: Oh, actually _subset_file_coverage is only called with search_data_catalog. extract should indeed have a similar check.

[juliettelavoie]

cat_sim = xs.search_data_catalogs(
    data_catalogs=['dataCat_cmip6_2022-09-02.json'],
    variables_and_freqs= {'tasmax':'D'},
          match_hist_and_fut= True,
          allow_conversion= True,
          allow_resampling= False,
          periods= ['1956', '2100'],
          other_search_criteria={
            'experiment': 'ssp126', 'source':'FGOALS-g3'}
            )

ds_sim = xs.extract_dataset(catalog=cat_sim['CMIP6_ScenarioMIP_CAS_FGOALS-g3_ssp126_r1i1p1f1_global'],
                    periods= ['1956', '2100'])['D']

This returns a dataset with a whole.