PDF Splitting about pressurecooker

[kollivier]

One issue with this functionality is that PDF splitting can break accessibility. Unfortunately, my research so far has failed to find a quick fix for this, for a couple reasons:

1. The easiest workable solution would be to modify the PDF to remove pages, rather than creating a new PDF, so that the accessibility data remains intact. Unfortunately, all the Python PDF APIs I've found only open a document for reading or for writing, and the writing APIs always create a new PDF rather than taking an existing one. This is likely due to the technical complexity of modifying an existing PDF.

2. None of the Python APIs I've explored so far offer the ability to both read and write accessibility info to PDFs. A vast majority of the Python libraries I've explored are concerned with converting to and from PDF, and don't consider PDF accessibility as a feature at all.

For reference, here's some of the libraries I've explored so far:

• PyPDF2 - The library we currently use. Can read and write, but does not handle accessibility data at all.
• pdfminer - Does provide read access to at least some of the accessibility data we need. The problem is that it's read-only, so we lack a library to write that data into the split PDF.
• ReportLab - Write only, does not appear to have accessibility writing APIs, though documentation is somewhat sparse so it's possible it's in the source somewhere.

I will keep exploring, but if I cannot find a library that supports adding accessibility data into PDFs, we will either need to extend an existing library or seek other solutions to splitting the PDF in order to keep accessibility. With pdfminer, we can at least warn about the loss of accessibility support when it happens.

[kollivier]

This task is on hold as we will probably just implement the ability to have ContentNodes display specific a PDF page range instead, in order to keep accessibility.

[jayoshih]

Here is the PDF-splitting logic used in Open Stax chef:

https://github.com/learningequality/sushi-chef-openstax/blob/master/utils/pdf.py#L93-L112

Recommended commercial product to check:

https://www.pdflib.com/download/pdflib-product-family/

recommendation on stack overflow:

[ivanistheone]

PDF splitting utils are now provided in ricecooker, see https://ricecooker.readthedocs.io/en/latest/pdfutils.html#pdf-splitter