Variable names in describe_data('reanalysis-era5-land') different than what can be used in a request about xcube-cds

[AliceBalfanz]

And once I get the dataset from the cds store
The variables are again called differently than in the request! Quite confusing and incosistent.

[pont-us]

Isn't this a duplicate of #15 ?

[AliceBalfanz]

Isn't this a duplicate of #15 ?

Looks like it, but we do not use the UI at all anymore. So how about closing 15 and adress this issue here?

[pont-us]

Looks like it, but we do not use the UI at all anymore. So how about closing 15 and adress this issue here?

OK, I can copy-paste the relevant content from #15 (there isn't much).

[pont-us]

Some background, adapted from comments at #15 :

The CDS API doesn't have unified variable names between the request and the result -- in some cases, e.g. in the soil moisture data set, there isn't even a one-to-one mapping between the two sets of variables. the CDS store plugin generally tries to provide a thin interface to the CDS API (i.e. manipulates the requests and results as little as possible). Of course it would be possible to expand the scope of the plugin to rename variables and unify the two schemes, but that's quite a lot of implementation and maintenance effort. We should discuss whether this ought to be a design goal for the plugin (and for other data store plugins).

Comment from 2021-02-21: After discussion today: still no clear resolution on how we should proceed with this, so I'm leaving the bug open but not actively working on it.

[pont-us]

To expand on the earlier comments above: data_vars in the return value of describe_data for a data store describes the values you can expect to see in the returned dataset. The parameters you can use in the request are not described under data_vars but under open_params_schema (which you can also get by calling get_open_data_params_schema), and there is no guarantee that the parameters schema will contain a list of variable names which corresponds exactly with those in the returned dataset. Indeed, there isn't even a guarantee that there will be a list of variable names in the parameter schema at all (e.g. it might be a one-variable dataset that therefore doesn't need to to explicitly specify the variables).

In the CDS store, the parameter schema is taken as directly as possible from the parameter schema of the CDS backend itself, so e.g. for reanalysis-era5-land, the user can browse around at https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-land-monthly-means?tab=form and use the same parameter names in their xcube request.

[pont-us]

Plan, after internal discussion: expand the parameter schema and implementation to include, as an alternative to variable_names, an additional parameter era5_variable_names, where the caller can specify the requested variable names using the ERA5 identifiers (matching the names in the returned dataset) instead of the CDS identifiers. Implementation should be fairly straightforward, since the store already has a table to map between these identifiers.