temporary files

[PointingerMartin]

I'm currently cropping and concerting lots of pix/files (over 20.000) and magick creates lots of big temporary files in a folder in C:\Users\xxx\AppData\Local\Temp (Folder e.g. RtmpopLpl0).
The files are named like magick-1532XblcXdSEDs_Q
As I work on lots of files this takes lot of disk space (over 20 GB) and the process gets slower and slower ...
Is this a known issue and is there a way/setting to avoid this?
Thanks! Martin

[jeroen]

It is known that imagemagick starts swapping to disk when it runs out of memory. But afaik it should be cleaning up those files, once your image objects in R have gone out of scope.

You can use gc() in R to force cleaning the temp files after the image objects are no longer in your R session. Can you check if this resolves the issue?

If not, you can also help us by creating an example script that we can use to test this problem.

The version of imagemagick we use on Windows is about two years old (6.9.9-9). @dlemstra do you know if there have been fixes on cleaning temp files?

[dlemstra]

There have been fixes for cleaning temp files but I don't know for sure when this happened. I guess it did happen in the last 2 years but I am not sure.

[gsimchoni]

I was having the exact same issue, @jeroen's suggestion saved me about a week's CPU time. I'm chunking my thousands of images and at the end of processing each chunk I do gc(). Without it R is really slowing down.

process_images <- function(paths) {
  images_list <- image_read(paths) %>%
    image_crop(geometry=str_c(wi_hi, "x", wi_hi, "+330+80")) %>%
    image_resize(geometry = "299x299")
  for (i in seq_along(paths)) {
    image_write(images_list[i], str_c("out/", paths[i]))
  }
}

# processing 1K images in chunks of 10s
ns <- 100
begins <- (seq.int(ns) - 1) * 10 + 1
ends <- seq.int(ns) * 10 + 0

for (i in seq.int(ns)) {
  cat(begins[i], " ", ends[i], "\n")
  process_images(images_paths[begins[i]:ends[i]])
  gc()
}

Thank you for this amazing package. There should be something in the docs about it though 😲