pikepdf VS PyPDF2

pikepdf – A Python library for reading and writing PDF, powered by qpdf

Not off the shelf but here are some tools. I have no experience with them.

Wolf binarization - I think it makes the text more clear before OCR.

https://github.com/chriswolfvision/local_adaptive_binarizati...

This thing OCRs the pdf using Tesseract OCR

https://github.com/ocrmypdf/OCRmyPDF/

Two other pdf tools

https://github.com/qpdf/qpdf

https://github.com/pikepdf/pikepdf

i recently transitioned a large PDF processing pipeline[1] away from PyPDF3 to use the pikepdf[2] library instead. pikepdf is based on the C++ `qpdf` library, and this switch has cut the necessary special-case manual checking + error handling in half compared to pypdf for these sorts of tasks.

[1]: i'm helping out with the CV Open Access archive, https://openaccess.thecvf.com/menu. broadly, our pipeline needs to ingest tons of author-provided PDFs of varying quality and output a canonical PDF with corrected page numbers, proper metadata, and a banner stamped on the first page. it's a lot of work, and it's not uncommon for pypdf to simply fail for slightly invalid input or give garbled results. this isn't optimal since we can't review 2,000 papers per conference release lol. pypdf has been nice, but pikepdf has handled everything i've thrown at it.

[2] https://pikepdf.readthedocs.io

I still haven't needed to do that part in Rust yet, unfortunately. My mother is still using the pypdf-based concatenator I wrote for her years ago.

I did an extremely quick search and am linking this without knowing anything about it.

As other posters have said, if you're just generating PDFs, that's doable with minor library support. If you want to open existing PDFs and do anything non-trivial with them, you'll want a mature, powerful PDF parsing library. PDFKit, which is part of macOS is pretty amazing. If you need portability, something like Python's PyPDF2 is probably the best bet. Knowing the Rust community, though, we'll probably get a library at least as good in Rust in a surprisingly short amount of time.

Feel free. I poked around the PyPDF2 code and it seems like reading signatures is something it supports via xfa_form property (GitHub issue where this was discussed; seems to have been closed v. recently) . would probably be a very simple PR provided you knew where to look for that property.

This tool was built to fill a gap in the PDF assessment landscape. Didier Stevens's pdfid.py and pdf-parser.py are still the best game in town when it comes to PDF analysis tools but they lack in the visualization department and also don't give you much to work with as far as giving you a data model you can write your own code around. Peepdf seemed promising but turned out to be in a buggy, out of date, and more or less unfixable state. And neither of them offered much in the way of tooling for embedded binary analysis. Thus I felt the world might be slightly improved if I strung together a couple of more stable/well known/actively maintained open source projects (AnyTree, PyPDF2, and Rich) into this tool.

PyPDF2 – A pure-python PDF library capable of splitting, merging, cropping, and transforming the pages of PDF files

Maybe PyPDF2 https://pythonhosted.org/PyPDF2/

Many of our PDFs were buggy, and many python libraries to work with PDFs are buggy. The library PyPDF2 works well on many PDFs, but hangs in this loop on others. Attempts to use most python PDF libraries resulted in endless log files of:

Look into the metadata of the .pdf file (extracted via PyPDF2) and see if any string matches the pattern of a DOI or an arXiv ID. Priority is given to the metadata which contains the word 'doi' in its label.

I also think Python with the requests library is a pretty good way to go (possibly along with this PyPDF library, but let me give you a bit more context for what you'll need to do to solve this problem.