So Query.jl should be quite performant? Do you happen to have a benchmark suite that compares several query frameworks, similar to your csv file reader benchmark? That should be a fairly useful tool.
I don’t know the details, but isn’t this required since vectors in R are restricted to a few element types? In Julia, this is not the case, so most tabular implementations just work with generic elements out of the box.
For sure, you could get basically the same functionality with a data frame with geo-data attached.
But there still might be good reasons for the own type, overloading plotting / mapping functions, or storing a geo-referencing system.
I just spent two weeks writing
sf code and wanted to highlight that it basically represents the concept we are working towards.