These benchmarks only account for the time spent in R (i.e., processing data), and not in javascript (i.e., plotting the map).

Mapdeck supports

data.frame
sf
sfencoded
sfencodedLite
geojson / json / character (as GeoJSON)

In all cases, plotting raw geojson is the quickest way to get your data to javascript as it doesn’t go through any pre-processing. See the geojson section for how you can use colorus on the raw GeoJSON

However, if you specify one of the properties as one of the colour variables, it then has to get converted to sf, processed, and back again to GeoJSON. This incurs a relatively big penalty.

The sfencoded objects were created before I created the spatialwidget library, and getting the coordinates and data from sf objects was slow. Therefore, I wrote the library googlePolylines to quickly convert sf objects into encoded polylines. The benefit of this library is the speed in getting the coordinates, and the size of the object gets reduced.

However, since spatialwidget, the processing time of sf objects has reduced significantly. So now the benefit is less-so the speed, and more-so the reduced size.

library(microbenchmark)
library(mapdeck)
library(sf)
library(geojsonsf)
library(googlePolylines)

Points / Scatterplot

set_token( "MAPBOX" )

## scatterplot
set.seed(20181209)
n <- 1e5
lons <- seq(-180,180, by = 0.001)
lats <- seq(-90,90, by = 0.001)
df <- data.frame(
    lon = sample(lons, size = n, replace = T)
    , lat = sample(lats, size = n, replace = T )
)
df$id <- 1:nrow(df)
df$val <- rnorm(n=n)
sf <- sf::st_as_sf( df, coords = c("lon","lat"))
enc <- googlePolylines::encode( sf )
encLite <- googlePolylines::encode( sf, strip = T )

sf$fill_colour <- colourvalues::colour_values(sf$val)
geo <- geojsonsf::sf_geojson( sf, simplify = FALSE )

microbenchmark(
    df = {
        mapdeck() %>%
          add_scatterplot( data = df, lon = "lon", lat = "lat", fill_colour = "val")
    },
    sf = {
        mapdeck() %>%
          add_scatterplot( data = sf, fill_colour = "val" )
    },
    enc = {
        mapdeck() %>%
          add_scatterplot( data = enc, fill_colour = "val" )
    },
    enclite = {
        mapdeck() %>%
          add_scatterplot( data = encLite, fill_colour = "val" )
    },
    geo_sf = {
        mapdeck() %>%
          add_geojson( data = sf, fill_colour = "val" )
    },
    geojson = {
        mapdeck() %>%
          add_geojson( data = geo )
    },
    times = 5
)

# Unit: milliseconds
#    expr        min         lq       mean     median          uq         max neval
# df      456.205270 466.227301 505.972537 521.141831  535.952496  550.335787     5
# sf      424.461500 426.184777 546.094275 433.743690  466.887704  979.193704     5
# enc     783.477064 806.466145 933.792187 885.506362  903.656713 1289.854652     5
# enclite 681.644942 730.634827 769.643424 740.264817  750.807022  944.865514     5
# geo_sf  636.163284 781.685261 975.163050 885.819292 1022.844405 1549.303007     5
# geojson   1.267405   1.348774   1.387614   1.401142    1.440508    1.480242     5

obj <- c("df","sf","enc","encLite","geo")
res <- vapply( mget( obj ), function(x) object.size(x), 1.0 )
res <- sort( res )
as.data.frame( res )

#              res
# df       2801088
# encLite 14001376
# geo     16785296
# sf      45237912
# enc     50003384

Lines / Path

sf <- roads
enc <- googlePolylines::encode( sf )
encLite <- googlePolylines::encode( sf, strip = T )
sf$stroke_colour <- colourvalues::color_values( sf$RIGHT_LOC )
geo <- geojsonsf::sf_geojson( sf, simplify = FALSE )


microbenchmark(
    sf = {
        mapdeck() %>% add_path( data = sf, stroke_colour = "RIGHT_LOC" )
    },
    enc = {
        mapdeck() %>% add_path( data = enc, stroke_colour = "RIGHT_LOC" )
    },
    enclite = {
        mapdeck() %>% add_path( data = encLite, stroke_colour = "RIGHT_LOC" )
    },
    geo_sf = {
        mapdeck() %>% add_geojson( data = sf, stroke_colour = "RIGHT_LOC" )
    },
    geojson_raw = {
        ## uses the fill_colour property
        mapdeck() %>% add_geojson( data = geo )
    },
    geojson = {
        ## converts to sf internally
        mapdeck() %>% add_geojson( data = geo, stroke_colour = "RIGHT_LOC" )
    },
    times = 5
)

# Unit: milliseconds
#        expr         min          lq        mean      median          uq         max neval
# sf           115.869173  127.285125  130.419934  127.974642  132.759235  148.211496     5
# enc          131.021713  133.304998  137.932747  134.803647  139.380252  151.153124     5
# enclite      125.009099  133.982827  163.172389  139.418675  147.436373  270.014970     5
# geo_sf       157.998692  168.252376  176.258981  172.229096  173.356248  209.458494     5
# geojson_raw    1.367818    1.403125    1.877565    1.504032    1.608655    3.504197     5
# geojson     1303.502648 1326.171179 1359.787277 1360.266253 1389.298148 1419.698157     5


obj <- c("sf","enc","encLite","geo")
res <- vapply( mget( obj ), function(x) object.size(x), 1.0 )
res <- sort( res )
as.data.frame( res )

#              res
# encLite  5008024
# geo      8771736
# enc     11739384
# sf      14516848

Polygon

sf <- sf::st_read( system.file("shape/nc.shp",package="sf"))
## make a bit bigger
sf <- rbind(sf, sf, sf, sf, sf, sf, sf, sf)
sf <- rbind(sf, sf, sf, sf, sf)
enc <- googlePolylines::encode( sf )
encLite <- googlePolylines::encode( sf, strip = T )
sf$fill_colour <- colourvalues::color_values( sf$AREA )
geo <- geojsonsf::sf_geojson( sf, simplify = FALSE )


microbenchmark(
    sf = {
        mapdeck() %>% add_polygon( data = sf, fill_colour = "AREA" )
    },
    enc = {
        mapdeck() %>% add_polygon( data = enc, fill_colour = "AREA" )
    },
    enclite = {
        mapdeck() %>% add_polygon( data = encLite, fill_colour = "AREA" )
    },
    geo_sf = {
        mapdeck() %>% add_geojson( data = sf, fill_colour = "AREA" )
    },
    geojson_raw = {
        ## uses the fill_colour property
        mapdeck() %>% add_geojson( data = geo )
    },
    geojson = {
        ## converts to sf internally
        mapdeck() %>% add_geojson( data = geo, fill_colour = "AREA" )
    },
    times = 5
)


# Unit: milliseconds
# expr               min         lq       mean     median         uq        max neval
# sf           83.684579  92.341509 103.123123  92.442575 106.142269 141.004682     5
# enc          17.203230  17.334162  17.925281  17.353199  17.978978  19.756838     5
# enclite      15.780515  17.205796  23.409552  19.847965  21.481832  42.731652     5
# geo_sf       91.054227  92.663365 107.123839  93.328851 103.012873 155.559878     5
# geojson_raw   1.167508   1.307513   1.376268   1.345209   1.492888   1.568223     5
# geojson     477.168586 492.928279 560.850102 508.497863 565.509485 760.146295     5


obj <- c("sf","enc","encLite","geo")
res <- vapply( mget( obj ), function(x) object.size(x), 1.0 )
res <- sort( res )
as.data.frame( res )

#             res
# encLite 1556432
# enc     3030544
# sf      5030176
# geo     5199600

Benchmarks

David Cooley

2020-08-17

Points / Scatterplot

Lines / Path

Polygon