st_as_sfc.SpatialPoints relatively slow

[bart1]

It seems this function is relatively slow compared to creating a sf object from a data.frame. It takes about 6 seconds to convert 300000 points. The conversion through a data frame is considerably faster. I think this could be easily resolved by using points_cpp here:

sf/R/sp.R

Line 94 in 4f550e5

lst = lapply(seq_len(nrow(cc)), function(x) st_point(cc[x,]))

. I don't seem to be able to label this issue as an enhancement but it is very much meant like that and not a bug. If desirable I can create a pull request.

require(sf)
#> Loading required package: sf
#> Linking to GEOS 3.8.0, GDAL 3.0.4, PROJ 6.3.1; sf_use_s2() is TRUE
require(sp)
#> Loading required package: sp


a<-SpatialPoints(cbind(runif(300000),runif(300000)), proj4string =CRS( "+proj=longlat"))
bench::mark(st_as_sfc(a),
st_as_sfc(st_as_sf(as.data.frame(a), coords = 1:2, crs=st_crs(a))))[,1:3]
#> Warning: Some expressions had a GC in every iteration; so filtering is disabled.
#> # A tibble: 2 × 3
#>   expression                                                                min
#>   <bch:expr>                                                           <bch:tm>
#> 1 st_as_sfc(a)                                                            5.61s
#> 2 st_as_sfc(st_as_sf(as.data.frame(a), coords = 1:2, crs = st_crs(a)))  173.8ms
#> # … with 1 more variable: median <bch:tm>

^{Created on 2022-12-14 with reprex v2.0.2}

[edzer]

A (tested) PR would be great; as you can see #2059 has an alternative approach to this (making sfc_POINT much more like a SpatialPoint), but it's currently unclear whether we want to merge that.

[bart1]

I'll keep an eye out for #2059 and submit an pull request depending on what happens there