Converting postscript with repetitive image calls, extremely large PDF files are generated and extreme conversion times are experienced. After dumping the output PDF if appears that ps2pdf is creating multiple XObjects per input EPS file. Converting a 20,000 page file with 15 unique images took 15 hours to complete, with an extremely large output PDF file. After removing all the image calls from the postscript, the conversion took 12 minutes. For testing I am disabling compression to dump the PDF. Stripping the PS to demonstrate the issue: (TESTFOO.EPS = 240,684 bytes) ps2pdf -dCompressPages=false -dMaxInlineImageSize=0 testb.ps testb.pdf (output size = 19,497 bytes) %!PS 217.25 141 moveto (TESTFOO.EPS) run adding another image call: (output size = 36,971 bytes) %!PS 217.25 141 moveto (TESTFOO.EPS) run 317.25 141 moveto (TESTFOO.EPS) run adding a 3rd call to the EPS image raises the filesize to 54445 bytes.
(In reply to comment #0) > Converting postscript with repetitive image calls, extremely large PDF files are > generated and extreme conversion times are experienced. After dumping the output > PDF if appears that ps2pdf is creating multiple XObjects per input EPS file. Realistically, its extremely hard for pdfwrite to know that two images presented separately are the same image. We would need to maintain a hash of every image presented so far, temporarily store each new image, hash it and compare the hash with every previous image in order to determine this. Frankly that would slow down 'normal' processing quite considerably, especially for people handling large images. While it would benefit your workflow, it would penalise everyone. Ideally the way to handle this is to construct the images as Form Resources in PostScript. You then use each Form as required. Because each form is unique there is no need to compare the images, and so a Distiller application can reuse the forms. Unfortunately at present pdfwrite does not properly honour PostScript forms, so this approach would not help you yet. We will be adding this functionality in the future (see #687561 and #691202) I believe you will find that Adobe Acrobat Distiller behaves in the same fashion with the workflow you describe, though in this case using forms *would* reduce the output size and processing time considerably, as Distiller is able to recognise forms. However we won't address the case where the same image is imply thrown through multiple times.