Should we create a DP obect? about climpred HOT 4 CLOSED

Some more thoughts on the benefit:

• Currently, it's somewhat meticulous to rename DP datasets to have the right dimensions. Also, we have "time_dim" in a few places still, complicating things. On object creation, we could have a 'CESM' and 'MPI' flag that ingest raw output and rename as would be needed.

• We could have the DP object also be able to hold hindcast/control, observation, uninitialized information. Either as distinct variables in the datasets or some companion DataArrays that get carried along.

from climpred.

Also when we approach the predictability limit we need always an upper limit for ACC and a lower limit for RMSE to check for significance. This could also be built-in.

I never wrote any code for classes so far. Would be a good learning. But in the future some time.

from climpred.

Yeah I think we can get by with the current setup. Though if we decide to do classes, it should be sooner than later because I think it would take a decent rewrite of code to get it to work.

I think we're fine with the functional programming for now.

from climpred.

I think we're fine to stick with functional programming. It just doesn't seem like this would enhance the project a lot.

Although I was thinking we should add a check similar to _check_xarray() that tests for the proper dimension names.

from climpred.