pypdf makes larger output files than gs

[johnkw]

pypdf makes larger output files than gs

I trimmed it down to a simple page with small files. In some cases the bloat is much more severe.

gs-output.pdf
pypdf-output.pdf
test-input.pdf

import pypdf ; pdfwrite = pypdf.PdfWriter() ; page = pdfwrite.add_page(pypdf.PdfReader(open('test-input.pdf','rb')).pages[0]) ; page.compress_content_streams(level=9) ; pdfwrite.write(open('pypdf-output.pdf','wb'))

followed by

gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.6 -dPDFSETTINGS=/prepress -dNOPAUSE -dQUIET -dBATCH -sOutputFile=gs-output.pdf - < pypdf-output.pdf

14078 Jun 20 13:31 gs-output.pdf
16722 Jun 20 13:31 pypdf-output.pdf
16722 Jun 20 13:29 test-input.pdf

[johnkw]

See also bug #1910

[stefan6419846]

test-input.pdf and pypdf-output.pdf are identical. Ghostscript rewrites the complete PDF file, while pypdf tries to preserve the structure as much as possible. Additionally, as mentioned in the referenced issue - your file contains quite some images which are modified by Ghostscript as well.

[johnkw]

Why was this bug split off into a discussion?

[stefan6419846]

Yes they're intentionally identical

This contradicts your comment on the initial discussion:

Just loading any random PDF and re-writing with pypdf causes it to bloat in size with much more text instead of compressed binary. For example this adds a few 100k in size

With the example you have provided in the new issue, this cannot be verified.

Why was this bug split off into a discussion?

Because I do not consider this a bug given the provided details and example files from running your input file through pypdf. If the output size of the file would grow unexpectedly, this might be a bug, but for now it just looks like a question about pypdf in general and how to use it.

We might reconsider this if you are able to provide a proper example reproducing the pypdf issue or some actual analysis where you see room for improvement (without modifying any images).

[johnkw]

The output is from pypdf. It already bloated the output. I just took a page from https://www.primasoft.com/librarymanager/library-files/library-barcode-labels-20-avery-5161-2000.pdf and saved it so you could see the bloat in isolation on a smaller file. But certainly you can do the entire https://www.primasoft.com/librarymanager/library-files/library-barcode-labels-20-avery-5161-2000.pdf if you want.

Regardless it is a bad thing for pypdf to bloat output. Whether it's a bug or a feature request to fix it, it belongs in the issue list not this discussion thing.

[stefan6419846]

I am sorry if there are any misunderstandings here, but for me your initial report does not show an issue with pypdf - thus I decided to convert this into a discussion as it seemed like there has been a misconception somewhere which better fits a discussion.

A proper issue report would include a reproducing example which is as easy/small as possible regarding the PDF file and clearly show your issue. Additionally, a proper indication on what exact aspect of the compression you are missing is always appreciated.

[johnkw]

I've never dived into how the PDF standard works, so I can't really comment much on why pypdf produces bloated files. But I do know when there's lots of plain text ASCII in a PDF that looks like postscript text then that's bad, pypdf produces files with that. gs then turns all of that ASCII into Flate gzipped bytes or whatever. But https://www.primasoft.com/librarymanager/library-files/library-barcode-labels-20-avery-5161-2000.pdf is an example. I turned it into a single page to try to make things as easy as possible, but if you want to use the full file, go ahead.

Either way this is either a bug, or a missing feature to not produce bloat.