Parallel Processing With Magick image_read() with Furrr

Is there a way to use parallel processing with image_read()? The below illustrates the issue.

library(purrr)
library(magick)
#> Linking to ImageMagick 6.9.12.3
#> Enabled features: cairo, freetype, fftw, ghostscript, heic, lcms, pango, raw, rsvg, webp
#> Disabled features: fontconfig, x11
library(furrr)
#> Loading required package: future

links <- 
  c(
    'https://images.google.com/images/branding/googlelogo/1x/googlelogo_color_272x92dp.png', 
    'https://th.bing.com/th/id/OIP.v7IDxAcM1Dw7ANwK2-MpBgHaJW?w=123&h=180&c=7&r=0&o=5&pid=1.7'
    )

# this works
map(links, image_read)
#> [[1]]
#>   format width height colorspace matte filesize density
#> 1    PNG   272     92       sRGB  TRUE     5969   72x72
#> 
#> [[2]]
#>   format width height colorspace matte filesize density
#> 1   JPEG   123    180       sRGB FALSE     4354   72x72

# this doesn't
plan(multisession, workers = 2)
future_map(links, image_read)
#> [[1]]
#> Error: Image pointer is dead. You cannot save or cache image objects between R sessions.

^{Created on 2023-12-04 with reprex v2.0.2}

Solution

Instead of returning the image as read by image_read, extract the bitmap first and return that bitmap. Then, reread the bitmap as an image.

library(purrr)
library(magick)
#> Linking to ImageMagick 6.9.12.93
#> Enabled features: cairo, freetype, fftw, ghostscript, heic, lcms, pango, raw, rsvg, webp
#> Disabled features: fontconfig, x11
library(furrr)
#> Loading required package: future

links <- 
  c(
    'https://images.google.com/images/branding/googlelogo/1x/googlelogo_color_272x92dp.png', 
    'https://th.bing.com/th/id/OIP.v7IDxAcM1Dw7ANwK2-MpBgHaJW?w=123&h=180&c=7&r=0&o=5&pid=1.7'
  )

# this works
map(links, image_read)
#> [[1]]
#>   format width height colorspace matte filesize density
#> 1    PNG   272     92       sRGB  TRUE     5969   72x72
#> 
#> [[2]]
#>   format width height colorspace matte filesize density
#> 1   JPEG   123    180       sRGB FALSE     4354   72x72

# this also works
plan(multisession, workers = 2L)
img_list <- future_map(links, \(l) {
  # read the file
  img <- image_read(l)
  # extract the raw bitmap and return it to caller
  img[[1L]]
})

lapply(img_list, image_read)
#> [[1]]
#>   format width height colorspace matte filesize density
#> 1    PNG   272     92       sRGB  TRUE        0   72x72
#> 
#> [[2]]
#>   format width height colorspace matte filesize density
#> 1    PNG   123    180       sRGB FALSE        0   72x72

^{Created on 2023-12-04}