fasterize_extract.R

fast_extract <- function(x, y, ...) {
  ## BEWARE no extract options are respected
  raster::extract(x, sfpoly_cells(x, y))
}

sfpoly_cells <- function(rast, sfpoly) {
  ## BEWARE, could be a big-data, this is not
  ## sparsely specified, so it's wasteful on memory, but fast (if you have the mem)
  which(!is.na(fasterize::fasterize(sfpoly, rast)[]))
}


data("wrld_simpl", package = "maptools")
w <- raster(wrld_simpl, nrows = 150, ncols = 300)
w[] <- rnorm(prod(dim(w)))

p <- sf::st_as_sf(wrld_simpl)
system.time(fast_extract(w, p))
#user  system elapsed 
#0.041   0.000   0.042
system.time(raster::extract(w, wrld_simpl))
# user  system elapsed 
# 14.673   0.002  14.679 

If you are interested, this is WIP. It needs polygon grouping but that is trivial. Direct first use will be in tabularaster