plot(SpatialPolygonsDataFrame) is insanely slow for RStudioGD

[MichaelChirico]

I wanted to plot the shapefile for Myanmar found here:

https://gadm.org/download_country_v3.html

library(rgdal)
library(sp)
tdir = tempdir()

get_poly = function() {
  tmp = tempfile(tmpdir = tdir)
  download.file(
    'https://biogeo.ucdavis.edu/data/gadm3.6/shp/gadm36_MMR_shp.zip',
    tmp
  )
  
  unzip(tmp, exdir = tdir)
  
  readOGR(tdir, 'gadm36_MMR_0', stringsAsFactors = FALSE)
}

Plotting this with RStudioGD is much, much slower than to e.g. pdf:

mmr = get_poly()
system.time(plot(mmr))
#    user  system elapsed 
# 170.430   0.434 171.086 
unlink(tdir, recursive = TRUE)

Restart R to clear cache/overhead and run again:

mmr = get_poly()
tpdf = tempfile(tmpdir = tdir, fileext = 'pdf')
system.time({
  pdf(tpdf)
  plot(mmr)
  dev.off()
})
#    user  system elapsed 
#   0.513   0.022   0.539 
unlink(tdir, recursive = TRUE)

So using the external device is about 300x faster... any idea?

png also takes < 1 second

RStudio: Version 1.1.456

R version 3.5.1 (2018-07-02)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS High Sierra 10.13.6

Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.5/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] grab_0.1.8

loaded via a namespace (and not attached):
 [1] OpenStreetMap_0.3.3 Rcpp_1.0.0          pillar_1.3.0        compiler_3.5.1      plyr_1.8.4         
 [6] bindr_0.1.1         RJDBC_0.2-7.1       tools_3.5.1         jsonlite_1.5.9000   tibble_1.4.2       
[11] gtable_0.2.0        lattice_0.20-38     pkgconfig_2.0.2     rlang_0.3.0.1       DBI_1.0.0          
[16] rstudioapi_0.8      yaml_2.2.0          rgdal_1.3-6         bindrcpp_0.2.2      rJava_0.9-10       
[21] dplyr_0.7.8         raster_2.8-4        rgeos_0.4-2         grid_3.5.1          tidyselect_0.2.5   
[26] glue_1.3.0          data.table_1.11.9   R6_2.3.0            geohash_0.3.0       sp_1.3-1           
[31] SparkR_2.4.0        purrr_0.2.5         ggplot2_3.1.0       magrittr_1.5        scales_1.0.0       
[36] codetools_0.2-15    assertthat_0.2.0    colorspace_1.3-2    lazyeval_0.2.1      munsell_0.5.0      
[41] crayon_1.3.4

[pascaloettli]

Could kindly provide information about the RStudio version, the OS and the R version? I tried your code and plot(mmr) is very fast. (RStudio 1.2.1090, Ubuntu 18.04, R 3.5.1 and Rstudio Server, MacOS 10.14.1, R 3.5.1)

[MichaelChirico]

@pascaloettli added

[pascaloettli]

@MichaelChirico Thank you. I tried also in RStudio 1.2.1099 on MacOS 10.14.1 but, while a bit slower compared to Ubuntu, it is still fast. (For some unknown reason, it is very slow in R console running on iTerm2, ~40x slower.)

[ronblum]

I am able to reproduce the problem with RStudio Desktop 1.2.1184-1 on MacOS 10.14.2 Mojave, nd when done, nothing is displayed. It works fine with RStudio Server 1.2.1186-1 on Ubuntu 16.04.5. On RStudio Desktop 1.1.463, the process never completed—it continued for at least 10 minute.

@MichaelChirico Can you clarify: on Ubuntu 18, were you using the desktop or server product? And for MacOS, did you mean desktop, or are you referring to using a Mac to connect to a remote server?

[pascaloettli]

Strangely, I still cannot reproduce with RStudio Desktop 1.2.1184 on macOS Mojave 10.14.1. It takes about 2 seconds to plot and display the map.

[MichaelChirico]

@ronblum I think only @pascaloettli was trying this on Ubuntu.

For me, I only ran this on my local machine (see <details> section of original post for all details)

[dmontecino]

Same here... MAC OS mojave
any solution?

[ronblum]

This is still an issue in RStudio Desktop 1.3.799 on MacOS 10.15.3 (Catalina). We'll review the issue as we continue development of RStudio, but we unfortunately don't have a timeframe for a solution.

[jakelawlor]

Hi, did anyone ever find an answer to this? I just got a new macbook pro (13 inch, 2020) and am having the same problem.

RstudioGD:

#user  system elapsed 
#128.162   0.510 129.271 

pdf:

# user  system elapsed 
# 0.423   0.027   0.460 

I am on rstudio version 1.3.1093 and mac big sur 11.1

[singhtanya88]

Hi, did anyone ever find an answer to this? I just got a new macbook pro (13 inch, 2020) and am having the same problem.

RstudioGD:

#user  system elapsed 
#128.162   0.510 129.271 

pdf:

# user  system elapsed 
# 0.423   0.027   0.460 

I am on rstudio version 1.3.1093 and mac big sur 11.1

I have exactly the same problem, but only on a Macbook Air 2019. I was hoping that an upgrade from Catalina to Mac Big Sur 11.1 would help. But it didn't. If you found the solution for this @jakelawlor I'd appreciate you sharing it. Thanks

[kevinushey]

I've reproduced this locally as well. Here's a snippet of the call stack:

For some reason, the calls are much more expensive through our shadow graphics device versus the "default" graphics device, but I'm not sure why.

[Polkas]

The same extremely slow plotting for me on my Mac Pro. Solution for others.

plot_2png <- function(x, filename) {
  x <- substitute(x)
  stopifnot(is.call(x))
  filename_path <- file.path(getwd(), filename)
  file.create(filename_path)
  png(filename_path)
  eval(x)
  dev.off()
}

usage for one or many plots.

plot_2png({
plot(pl)
plot(voi, add = TRUE)
}, "PATH/FILE.png")

Still not enough if sb want to use it in e.g. Rmd.

EDIT

I have a new observation that the same slow computation occurs from R base GUI.
Thus it looks to not be only a RStudio problem.
https://stat.ethz.ch/pipermail/r-sig-mac/2017-July/012446.html

[rosie-williams]

I'm having the same issue. Occasionally it will run as normal but most of the time Rstudio crashes and can't produce the plot. It definitely appears to be a graphics issue for me but I haven't worked out the differences yet between when it runs as normal or when it crashes Rstudio. However, I've found that if I run X11(type = "cairo") before the query then the performance time is normal in the graphics box.

[RaulMartinezRincon]

I had the same problem in a mac os. My solution was to install XQuartz (https://www.xquartz.org/), then use x11() to plot the spatial object, however you will need to set "usePolypath = FALSE". Example:

library(raster)
fraL1 <- getData('GADM', country='FRA', level=1) 
x11()
plot(fraL1, usePolypath = FALSE)

[pascaloettli]

Don't use the "(default)" Backend under General > Graphics.

With (default):

system.time(plot(mmr))
   user  system elapsed 
 79.262   0.248  79.713 

With Cairo:

system.time(plot(mmr))
   user  system elapsed 
  0.301   0.025   0.329 

With AGG:

system.time(plot(mmr))
   user  system elapsed 
  0.231   0.012   0.253 

RStudio 2021.09.1+372 "Ghost Orchid" Release (8b9ced1, 2021-11-08) for macOS
Mozilla/5.0 (Macintosh; Intel Mac OS X 12_1_0) AppleWebKit/537.36 (KHTML, like Gecko) QtWebEngine/5.12.10 Chrome/69.0.3497.128 Safari/537.36