Open simone-silvestri opened 4 days ago
Don't we want a different backend for repeat year vs multi year forcing?
So we don't need to refactor, we just need to create a new backend + change the name of the existing?
But having a single "metadata" abstraction for all data is more ambitious. It does seem possible. Less sure about the backend. I think each data could be idiosyntratic, motivating different backends for different datatypes.
If you sketch your idea here we can have a more concrete discussion
The JRA55 requires a refactor because it assumes that all the data for a variable is contained in one file. This is true for only for repeat year forcing; in reality, multi-year forcing is contained in different files. This requires us to restructure
1) the downloading procedure 2) the JRA55NetCDFbackend 3) the setting of the data in memory
We also need a design that allows us to possibly add different versions efficiently. I propose to generalize
ECCOMetadata
to a generalMetadata
type and structure the JRA55 data handling in the same way we do ECCO at the moment and provide aJRA55RepeatYear
and aJRA55MultiYear
version.@glwagner