Anshul Singhvi

Thanks Felix, that seems interesting! Are those actually DimArrays though, or just convenient wrapper structs? I went over the code briefly but am not able to see how this is...

I just wrote up a more interesting vector data cube example with polygons. See the following: ```julia using DimensionalData using DimensionalData.Lookups using DimensionalData.Dimensions import DimensionalData as DD # load data...

yep! I'm building out a PolygonLookup now that would support that, but we could change that to be a more generic GeometricLookup at some point that works for any geometries

btw the big issue here is probably going to be file I/O. Any suggestions on that? NetCDF has some support for this but I have no clue how we would...

Yeah...Ideally we would want a sliced dimtable from the stack, with a given index column

> Have you looked at the python/R vector data cubes ? If we can use something already being standardised... > > (The CF standard is likely pretty arcane) They also...

`DD._dims2indices` is where we would implement an `Unaligned` like thing that would let us harvest X and Y from getindex and move it into geometry lookups

There should always be some selector for X and Y, i.e `X(10)` should never work with a geometric lookup unless there is another X dim

Thanks for the write up, that clarifies a lot! @rafaqz we can indeed do a dimtree, but there is one issue - we may need to implement s2 indexing 😅...

The only potential hiccup here is that DGGS are, at this point, more like matrix (actually 3d array) lookups than like a single vector flat geometry lookup.