Support for new spatial data classes

[rhijmans]

It would be great if addRasterImage could support for SpatRaster objects from the terra package (a replacement for the raster package). It should perform better as terra::project is much faster than raster::projectRaster. It also should be easy to do, the interface of terra is very similar to that of raster. It would also be great if the leaflet method could support vector data from sf and SpatVector from terra.

[jcheng5]

Hi @rhijmans! That would be really great indeed.

Can you clarify what you mean by "vector data from sf"? I thought we already had pretty good sf support, though I haven't been paying much attention.

[rhijmans]

My bad, I missed that as I was reading the docs --- but it is clearly there.

[rhijmans]

And, by the way, I think you set me on the path towards terra (C++ with some R around it) with your contributions to make raster faster for leaflet.

[gilbertocamara]

I second the request from @rhijmans: supporting "terra" objects in leaflet would help all those that work with geospatial data.

[AMBarbosa]

Dear colleagues, any news on implementing "SpatRaster" input for addRasterImage, and "SpatVector" input for addPolygons, addPolylines, etc.? This is becoming more and more urgent, given the evolution in spatial data users. The "SpatVector" implementation could be as simple as coercing to the already-implemented 'sf' format at the beginning of the functions:

if (inherits(data, "SpatVector"))  data <- sf::st_as_sf(data)

I guess the analogous could provisionally be done for "SpatRaster" too (if (inherits(x, "SpatRaster")) x <- raster::raster(x)), though given the differences in performance, it is (speaking from experience) totally worth the trouble to do the opposite, i.e. make the minimal changes in code to accomodate "SpatRaster" instead, and eventually coerce "Raster" inputs to it (if (inherits(x, "RasterLayer")) x <- terra::rast(x)). Thanks and kudos for the great package!

[rhijmans]

@jcheng5: I would happily create a PR if you are interested.

[jcheng5]

@rhijmans That would be fantastic, thank you

[rhijmans]

With my fork, that you can install with:

remotes::install_github("https://github.com/rhijmans/leaflet")

I can now do

library(terra)
library(leaflet)

# SpatRaster
r <- rast(xmin=-2.8, xmax=-2.79, ymin=54.04, ymax = 54.05, nrow=30, ncol=30, crs="epsg:4326", vals=1:900)
leaflet() |> addTiles() |> addRasterImage(r, colors = "Spectral", opacity = 0.8)

# SpatVector
v <- geodata::gadm("Uganda", level = 1, path =".") |> simplifyGeom(.01) |> makeValid()
leaflet() |> addTiles() |> addPolygons(data = v)
#or
leaflet(v) |> addTiles() |> addPolygons()

I would appreciate it if any of you could test this a bit more, and let me know.

[AMBarbosa]

I've tested it with a few different 'leaflet' scripts, combining a variety of layers of different classes (including "Spatial*", "SpatVector", "RasterLayer", "SpatRaster"), and it seems to work seamlessly! Also, 'terra' layers plot visibly faster than their "Spatial*" versions (cf. Uganda in the above example). Thank you so much, Robert!

[AMBarbosa]

Forgot to mention I'm on Linux Mint 19.3 Cinnamon. I also deployed a Shiny app with 'leaflet' maps at https://ambiogeo.shinyapps.io/atlasmampor/, using @rhijmans's fork and a few 'SpatVector' layers, with no problems so far.

[rhijmans]

This has now been merged into the main branch. A follow-up PR (currently pending, #808) adds support for color-tables and RGB rasters. See my fork at with these examples

library(terra)
library(leaflet)

# RGB
s <- rast(system.file("ex/logo.tif", package="terra"))  
crs(s) <- "+proj=longlat" 
leaflet() |> addTiles() |> addRasterImage(s)

# color table
r <- rast("/vsicurl/https://geodata.ucdavis.edu/test/pr_nlcd.tif")
r[550:750, ] <- NA
leaflet() |> addTiles() |> addRasterImage(r, opacity = 0.75)

# difference with a continuous raster
x <- r * 1
leaflet() |> addTiles() |> addRasterImage(x, opacity = 0.75)