add a splitdims?

[tiemvanderdeure]

Would it be worth having something that is effectively the opposite of mergedims - to split dims with Tuple eltype into two separate dims?

Just to come with an example from the documentation:

julia> A = rand(X(1:0.01:2), Ti(tempo))
julia> minimum.(groupby(A, Ti=>yearmonth))

This now leaves us with a Sampled{Tuple{Int64, Int64}} Ti dimension. I think it'd be neat to be able to do splitdims(B, Ti => (:year, :month))

[tiemvanderdeure]

Or maybe even splitdims(A, Ti => (year, month) => (:year, :month) could be possible?

[rafaqz]

We have unmergedims already, is that not enough?

[tiemvanderdeure]

I had a look, but it only works for MergedLookup. Although I guess in this case one could wrap the Ti dimension in a MergedLookup

[tiemvanderdeure]

Okay I tried this approach and this is the best I could do with unmergedims.

A = rand(X(1:0.01:2), Ti(tempo))
B = minimum.(groupby(A, Ti=>yearmonth))
C = rebuild(B, dims = (Ti(Rasters.DD.MergedLookup(val(lookup(dims(B, Ti))), (Dim{:year}(1:2), Dim{:month}(1:12)))),))
unmergedims(C, (:year, :month))

set for some reason didn't work so I had to use rebuild. Probably there is a better way to do it's not very obvious. But maybe the way is to improve unmergedims rather than add another function.

[rafaqz]

Yeah that's too hard. let's make unmergedims work on anything.

[tiemvanderdeure]

I realized that this is in some ways similar to #739 - maybe we can use some of the internals? The dims to be split are kind of like table columns (and they could also miss some values etc).