Optimal way of saving .pdf as .png pages

[gmile]

Hi! Thanks for a fantastic library!

What's a recommended way of saving PDF pages as .png files?

I have arrived at the following code, it's working, but it seems like reading pdf pages in a loop is only sub-optimal: I need to read the PDF file N+1 times, where N is the number of pages in PDF. After all, each time I call Vix.Vips.Operation.pdfload/2, the whole image is loaded in the memory, right?

I'm not familiar with vips API, hence the question.

Mix.install [:vix]

[path] = System.argv()

{:ok, img} = Vix.Vips.Image.new_from_file(path)
{:ok, pages} = Vix.Vips.Image.header_value(img, "n-pages")

Enum.each(1..pages, fn page ->
  {:ok, {img, _flags}} = Vix.Vips.Operation.pdfload(path, dpi: 300, page: page)
  :ok = Vix.Vips.Operation.pngsave(img, "/tmp/page-#{page}.png")
end)

For context, the above vix is using a system libvips with pdfium support compiled in.

[akash-akya]

Hi @gmile! Thanks for trying Vix.

libvips lazy loads wherever it is possible, so it does not loads the whole image when you ask for a specific page with page, and n (# pages) with value 1 (default). So it is already optimal in that sense. If you have large number of pages then you can write multiple pages concurrently using Task.async_stream at the cost of more memory.

If you see abnormal memory usage then please share more details.