Purpose of register_implemented_routines and when should it be used.

[BeatrizArellano]

Hi, I have a question about diagnostics in FABM.
I noticed that some modules use register_implemented_routines to register source_do, source_do_surface, or source_do_bottom, used to indicate which routine computes a diagnostic variable. However, I am not fully clear on the purpose of register_implemented_routines, or on when it is actually needed.

Also, when registering a diagnostic variable, should the source argument always be specified, or only in some cases?

I would appreciate any clarification on these features.

[jornbr]

Hi! That one fortunately has an easy answer - register_implemented_routines was originally introduced as performance optimization. It allowed FABM to skip calls (and associated preparation of variables) to routines that were not implemented. However, as of version 3, FABM finds out itself what routines are implemented, so you will not need to call register_implemented_routines, ever.

As for source: that is needed only if its default value is not suitable for your use case. It defaults to source_do, that is, "the diagnostic will be computed in the do subroutine", for diagnostic identifiers of type type_diagnostic_variable_id, to source_do_bottom for type_bottom_diagnostic_variable_id and to source_do_surface for type_surface_diagnostic_variable_id.

In practice, the typical reason to explicitly provide source is for diagnostics that are computed in do_column, i.e., source=source_do_column. It is also required for diagnostics of type type_horizontal_diagnostic_variable_id (for which source has no default), but that type is deprecated - type_bottom_diagnostic_variable_id or type_surface_diagnostic_variable_id should be used instead.

[BeatrizArellano]

Thank you very much! That's very clear.