Rafael Schouten

Thanks, I wasn't aware of `nomissing`. But I was hoping this could happen without both conversion steps and the extra memory allocations that requires. We could use a keyword argument...

Yes something like that. I forgot it was `load` for NCDatasets. ArchGDAL and GeoData use `read`, as does HDF5, and Base julia. But some other packages seem to use `load`...

Any luck with this @ali-ramadhan? netcdf could also be a sink for parallel DiskArrays/Dagger.jl processing, so this would be widely useful.

There are a few obscure things that dont convert from wkt to netcdf grid mapping. But otherwise they're interchangeable. This is a relevant CF thread about it: https://github.com/cf-convention/cf-conventions/issues/222 Writing grid...

Just thinking it would also be good to also read `grid_mapping` and add support for `GeoInterface.crs` for a `Dataset` and `CFVariable`. This would return `nothing` if it wasn't found, and...

> For example, do you think that for a disk array a of size 10x10 this a[:,:] = ones(4,4) should be equivalent to a[1:4,1:4] = ones(4,4), i.e. let the RHS...

I totally agree that the functionality is good and useful. But using `Colon`/`setindex!` syntax like that is not Julian - you can't do that with `Array`. It will create problems...

@meggart > However, I still think this syntax would not be very Julia-like, so was holding back merging this until there is a bit more feedback. This syntax means GeoData.jl...

In the current case we at least know the size from the indices passed to `setindex!`. In the proposed changes we need to check the array bounds. But I also...

Fixing type stability and adding precompilination runs brings the `CSV.File` load time down to 7 seconds, and `using CSV` is 1 second. There is another second available by removing the...