Open krlmlr opened 7 years ago
(Me too!) I'm also very interested in generalization here. I'm still ironing this out, but there's a kind of a marriage of NetCDF / raster / and tbl_cube in this project https://github.com/hypertidy/tidync that gets around the geographic-assumptions in raster, the need for curvlinear coordinates for grids, and the critical need to delay the generation of coordinates for geom_raster.
Here's another realish example for ocean model output: https://rpubs.com/cyclemumner/281801
None of the existing "gridded abstractions" are sufficiently general IMO, and I'm keen to discuss where it should go.
The definition of a "wide" dataset looks very similarly to what
tbl_cube
in dplyr does. (It generalizes to more than two dimensions, though.) I wonder if and how these concepts can be integrated.