s2 much slower than sf with st_union in epsg 4326

[arthurgailes]

Hello,

Writing the following operation, I get much slower runtime using s2 using the 4326 projection

library(sf)
#> Linking to GEOS 3.9.0, GDAL 3.2.1, PROJ 7.2.1
sf_use_s2(T)
shp <- read_sf(system.file("shape/nc.shp", package = "sf")) 
shp2 <- st_transform(shp,3857)
system.time(st_union(shp))
#>    user  system elapsed 
#>    0.16    0.00    0.15
system.time(st_union(shp2))
#>    user  system elapsed 
#>    0.04    0.00    0.05


sf_use_s2(F)
#> Spherical geometry (s2) switched off
system.time(st_union(shp))
#> although coordinates are longitude/latitude, st_union assumes that they are planar
#>    user  system elapsed 
#>    0.03    0.00    0.03
system.time(st_union(shp2))
#>    user  system elapsed 
#>    0.05    0.00    0.05

^{Created on 2021-07-27 by the reprex package (v2.0.0)}

I don't have a reprex for it, but the difference in performance increases severely at scale, and sometimes breaks altogether. Also seems to hold true for st_intersects. Perhaps the difference here deserves a warning, such as is given in the non-sf workflow?

[arthurgailes]

Showing the difference at scale:

library(sf)

nc = st_read(system.file("shape/nc.shp", package="sf"))
#> Reading layer `nc' from data source 
#>   `C:\Users\Arthur.Gailes\Documents\R\R-4.1.0\library\sf\shape\nc.shp' 
#>   using driver `ESRI Shapefile'
#> Simple feature collection with 100 features and 14 fields
#> Geometry type: MULTIPOLYGON
#> Dimension:     XY
#> Bounding box:  xmin: -84.32385 ymin: 33.88199 xmax: -75.45698 ymax: 36.58965
#> Geodetic CRS:  NAD27

grid <- st_make_grid(nc, cellsize = 1e-1)

system.time(st_union(st_transform(grid, 4326)))
#>    user  system elapsed 
#>    2.41    0.00    2.41

system.time(st_union(st_transform(grid, 3857)))
#>    user  system elapsed 
#>    0.33    0.02    0.35

grid <- st_make_grid(nc, cellsize = 1e-2)

system.time(st_union(st_transform(grid, 4326)))
#>    user  system elapsed 
#> 1826.68    0.28 1827.17

system.time(st_union(st_transform(grid, 3857)))
#>    user  system elapsed 
#>   18.72    0.17   18.91