Anshul Singhvi
Anshul Singhvi
I am clauding this up, PR incoming
#1048 makes sparsearrays a weakdep
We would also need an error hinter in Rasters telling users to use Planar. But the broader question still remains ... how can we detect if a crs is not...
We could do a basic parser from ProjJSON I guess...at least checking ellipsoid parameters and the nature of the CRS
Reopening because I have a thought about what the error should be - will make a PR this week
I forgot the thought 😢 But I think we could do a ```julia struct ProjTransformationError source destination error_text end ``` or so
Could you try reproducing this on the original data using AccurateArithmetic.jl? Or convert the values to Float64 before summing?
Hmm it's hard to do this with lazy iteration without essentially copying the Base implementation. One can always define `Base.sum(sm::Rasters.SkipMissing) = Base.sum(collect(sm))`, but that materializes `sm` which may not be...
KahanSummation.jl is also pretty light and I have a PR up that causes it to use mapreduce so presumably disklike things will be supported well. We can point users to...
Could we also make it a weak dependency of GeoInterface, to get wrapper geometries enabled by default? It would just add `GeoInterfaceMakieExt` to GeoInterface in addition to the regular GeoInterfaceMakie....