lines

[mdsumner]

this example

l <- sf::st_cast(silicate::inlandwaters, "MULTILINESTRING")
library(vaster)
## define a raster (xmin, xmax, ymin, ymax), (ncol, nrow)
v <- silicate::sc_coord(l)
ext <- unlist(lapply(v, range)) #+ c(-1, 1, -1, 1) * .1

dm <- c(10000, 40000)
#ext <- c(0, dm[1], 0, dm[2])/dm

## less than a second
system.time({
r <- controlledburn:::burn_line(l, extent = ext,dimension = dm)
})
pp <- function(listindex, dimension, extent = NULL) {
  if (is.null(extent)) extent <- c(0, dimension[1], 0, dimension[2])
  idx <- do.call(rbind, listindex) + 1
  vaster::xy_from_cell(dimension, extent, vaster::cell_from_row_col(dimension, idx[,2], idx[,1]))
}
plot(pp(r, dm, ext), pch = ".")


# tr <- terra::rast(terra::ext(ext), ncols = dm[1], nrows = dm[2])
# vct <- terra::vect(l)
# tr <- terra::rasterize(vct, tr)

now committed.

• not correct yet, it's got a hack for the initial y discretization (I've had enough getting confused for the moment)
• _polygon and _edge versions of functions are now established
• there is burn_polygon and burn_line unexported functions
• currently we only allow sf lines for burn_line, but that could accept polygons trivially
• consider, move to a model of storing cell numbers not yline,xcol ?

[mdsumner]

check out Robin's benchmark https://github.com/geocompx/geocompx.org/blob/10-blog-on-downloading-splitting-and-processing-roads-data/post/2023/linesplit/index.qmd