Right now I have this weird, inefficient process where I load in a list of sf objects (one for each sentence), each of which has the same attached geometry (the hexagonal grid). Then I extract just one element to get the grid and its initial values. And then I use lapply to remove the geometry column from all of the list elements.
I think I initially did this so I wouldn't be redundantly storing the geometry column over and over, but the way I'm handling the processing here actually isn't improving efficiency at all.
Should move this processing to a preprocessing script. Save a "medium grid" object that's just the geometries, and then load in a list of the predicted values only. That actually might translate to some improvements in performance.
Right now I have this weird, inefficient process where I load in a list of sf objects (one for each sentence), each of which has the same attached geometry (the hexagonal grid). Then I extract just one element to get the grid and its initial values. And then I use lapply to remove the geometry column from all of the list elements.
I think I initially did this so I wouldn't be redundantly storing the geometry column over and over, but the way I'm handling the processing here actually isn't improving efficiency at all.
Should move this processing to a preprocessing script. Save a "medium grid" object that's just the geometries, and then load in a list of the predicted values only. That actually might translate to some improvements in performance.