Bing Roads as $R$ sf digital files

Bing Maps has released in late 2022, through the Microsoft GitHub account, a whole bunch of mined roads around the world, covering in all ~48 millions km over 202 Countries.

I've downloaded all zipped GeoJSON files, partitioned by Country (see the 3-chars iso3 column in the countries.csv file) and US State (see the 2-chars code column in the states.csv file), and converted each as digital vector files to be easily loaded into $R$ using the sf package. The coordinate reference system is WGS84 or EPSG:4326.

You can find all the files at this location, for 202 Countries and 49 US States (the number of lines in each country/state file can be found in the column nlines in the two csv files mentioned above).

The files have been serialized using the qs package.

I suggest you use the leaflet package to visualize the roads, together with the leafgl add-on:

library(qs)
library(sf)
library(leaflet)
library(leafgl)
y <- qread('/path/to/downloads/XXX', nthreads = 8)
leaflet() |> addTiles() |> addGlPolylines(y)

The biggest file is for India, with more than 13 millions lines, needing at least 7GB RAM only to keep the spatial data in memory. Acting on it requires obviously more memory for operations. For example, filtering out the roads for the Capital city only, ~120K, requires more than 16GB:

y <- qread('/path/to/downloads/IND', nthreads = 8)
yx <- st_read('/path/to/downloads/delhi_administrative.shp')
ynd <- y |> st_filter(yx)
leaflet() |> 
    addTiles() |>
    addPolylines(data = yx |> st_cast('LINESTRING'), color = 'black', fillOpacity = 0) |> 
    addGlPolylines(ynd)

You can see the result with the attached leaflet html map New_Delhi.html (besides the downloads, the process took a bit over 5 minutes to complete on an AMD Ryzen 9 5900X machine).

Moreover, if you need more than one country just use the rbind command, listing all the countries you need (be careful here because of the overhead you need more memory than the total combined for the involved countries).

Finally, notice that ~450K US roads (out of more than 54 mlns) have been deleted because of either being single points or dropped in the validation process. This problem was not reported in any of the other Country's files.