Distinguish between out_name and var_name

[jvegreg]

@sloosvel (#333) found that some tables contain variables with the 'out_name' different from the 'var_name'. We are currently assuming that both are the same, this pull request fixes this.

Closes #333

[sloosvel]

Thanks @jvegasbsc , it looks like it works well. Although, the variable needs to be called in the recipe by the var_name, not the out_name. Maybe this could become confusing if someone is trying to work with 6hourly files named zg_... .nc , because in the recipe they would need to call zg7h (or zg27) to be able to load the files properly.

[valeriupredoi]

yeah am with Saskia on this one

[jvegreg]

Yes, that is inconvenient. But i have no idea on how I can distinguish between them, as they belong to the same table

[bouweandela]

I think the choice var_name is not very clear, because this is also the name of the iris cube attribute that contains the short_name. Would it be possible to use something else? Maybe cmor_name instead of var_name?

Note that this pull request breaks existing recipes in which short_name is defined, i.e.:
examples/recipe_variable_groups.yml
examples/recipe_check_obs.yml
recipe_flato13ipcc.yml
recipe_tcr.yml
recipe_ecs.yml
recipe_anav13jclim.yml

[jvegreg]

I think the choice var_name is not very clear, because this is also the name of the iris cube attribute that contains the short_name. Would it be possible to use something else? Maybe cmor_name instead of var_name?

Good idea. I will change it

Note that this pull request breaks existing recipes in which short_name is defined, i.e.:
examples/recipe_variable_groups.yml
examples/recipe_check_obs.yml
recipe_flato13ipcc.yml
recipe_tcr.yml
recipe_ecs.yml
recipe_anav13jclim.yml

Ok. I will create a companion branch in ESMValTool to fix them

[bouweandela]

Can you also check what name is used in the drs? The short name or the cmor name?

[jvegreg]

It is the short_name. It is not a problem because the variablse that share table and out_name are never present together in the data request for any activity. We need to distinguish them to be able to check against the correct definition, though

[bouweandela]

@jvegasbsc Could you have a look at #391 (comment)? Apart from that, I think this pull request looks fine.

[bouweandela]

@jvegasbsc Could you a have a look at the merge conflicts and failing tests?

[bouweandela]

Looks good to me now. @mattiarighi Could you please test, using the branch from ESMValGroup/ESMValTool#1493?

[mattiarighi]

I tested successfully a few recipes and cmorizers, other needs still to be adjusted (I reported here).

Outdated

[bouweandela]

@jvegasbsc Instead of the massive renaming approach, I would be more in favour of solving this with minimal changes, similar to how it was done for the variables with different names between CMIP5 and CMIP6 in #595. Do you think that would be possible?

[jvegreg]

@jvegasbsc Instead of the massive renaming approach, I would be more in favour of solving this with minimal changes, similar to how it was done for the variables with different names between CMIP5 and CMIP6 in #595. Do you think that would be possible?

Not, because it is not a massive rename but a splitting one name in two. Currently we are treating two different names as if they are only one.

I kept short_name for the out_name, as it is the most widely used (specially in diagnostics) and only use the new cmor_name when working with the tables, which is mostly internal stuff.

[bouweandela]

It will lead to a massive rename in the ESMValTool repository as it is now (partly implemented in ESMValGroup/ESMValTool#1493), because we will need to adjust the recipes/diagnostics/cmorizers to the fact that one name is split into two.

Would it lead to less renaming if we keep short_name for the variable name in the table and call out_name var_name so it aligns with iris?

[jvegreg]

Would it lead to less renaming if we keep short_name for the variable name in the table and call out_name var_name so it aligns with iris?

In the recipes, yes. In ESMValCore code I think we will have more, but involve less people...

[bouweandela]

Shall we plan a short meeting at the workshop to discuss how to go forward with this @jvegasbsc @sloosvel?

[sloosvel]

Fine for me!

[bouweandela]

Summary of the discussion: instead of the approach here, an extra name, e.g. cmor_name will become available in the variable definition section of the recipe to specify the name of the variable in the cmor table if it differs from short_name. This name will default to short_name if it is not provided. This ensures backward compatiblity.

[jvegreg]

Replaced by #1099