search

NCAS-CMS / cfa-conventions

NetCDF Climate and Forecast Aggregation (CFA) Conventions
https://github.com/NCAS-CMS/cfa-conventions/blob/main/source/cfa.md
1 stars 1 forks source link

Clarify canonical units #15

Closed davidhassell closed 3 years ago

davidhassell commented 3 years ago

The generic form of a fragment should not mention "canonical" units.

nmassey001 commented 3 years ago

I agree.

What is the mechanism for determining whether a fragment has "equivalent, but not equal" units to the aggregated data? Will we have to rely on udunits? Who does the conversion, the user, or the CFA library? I think I'd prefer that we enforce that all fragments have the same units, and we require the user to do the conversion when aggregating the file.

davidhassell commented 3 years ago

Hi @nmassey00,

Will we have to rely on udunits?

Good question. My current thoughts are:

Happy to think about and discuss this more ...

I think I'd prefer that we enforce that all fragments have the same units, and we require the user to do the conversion when aggregating the file.

I'm afraid that, nice as that would be!, we have the use case of, e.g. needing to aggregate timeseries where the time coordinate variables have different reference time units.

davidhassell commented 3 years ago

Hi @nmassey001, I've put in a minor rewording, borrowed from the CF conventions, that I hope makes sense in plain language and leaves the implementation of assessing equivalence a free choice. What do you think?

davidhassell commented 3 years ago

Hi @nmassey001 - are you OK with the new text here? The key bit that might get over looked is, perhaps, "This manipulation of the fragment is carried out by the application program that is managing the aggregation", which applies all manipulations required to coerce the fragment to its canonical form.

davidhassell commented 3 years ago

Hi @nmassey001 - is this PR OK to merge, now? Thanks.

nmassey001 commented 3 years ago

Hi @davidhassell Yes, this is fine now. Thanks,