Given we have t_eval vectors already, and if #35 is implemented, we basically have all the components of a lazy interpolating AbstractArray, but still make users do that manually.
It would be cleaner to provide an InterpArray wrapper that to the rest of the ecosystem behaves just like any other array over the t_eval of each dimension in the NDInterpolator. This would allow e.g. larger than memory interpolation grids to be treated as a materialised array.
I will have to do this for DimensionalData.jl anyway, so it may as well live here.
Given we have
t_evalvectors already, and if #35 is implemented, we basically have all the components of a lazy interpolatingAbstractArray, but still make users do that manually.It would be cleaner to provide an
InterpArraywrapper that to the rest of the ecosystem behaves just like any other array over thet_evalof each dimension in theNDInterpolator. This would allow e.g. larger than memory interpolation grids to be treated as a materialised array.I will have to do this for DimensionalData.jl anyway, so it may as well live here.