ENH: consider images inside PDF made with onlyoffice

[0xNath]

closes #2613

Added code to detect patterns in "_get_ids_image".
To avoid any conflicts with images that could be located directly in a page or images using the same ID in different patterns, images ids under patterns are returned in this form : "/Pattern/patternNameHere/imageNameHere"

Added code to deal with Pattern images in "_get_image".

Added a new test to verify ids and image data returned related to the modifications done above.

[0xNath]

I made a PR to add the sample files :
py-pdf/sample-files#30

Instead of putting the files directly into this repository, should I host the files on github or somewhere else so they get downloaded ?

[stefan6419846]

Instead of putting the files directly into this repository, should I host the files on github or somewhere else so they get downloaded?

In general, files should go into the sample files if you own all copyrights and are okay with the license terms there. Unfortunately, merging permissions are more restricted over there, thus I am not able to merge anything there myself.

The usual approach we take apart from this is to upload to an issue or PR and download the corresponding files from the GitHub URLs. Unfortunately, this will not work for image files any more as far as I know as GitHub started to add JWTs to their URLs which expire accordingly and are user-specific.

[pubpub-zz]

Can you rename the title of your PR : this is not a a STYle but an ENHancement.
You should also edit your first thread to set you "closes " info to ease closing

you should be able to "factorize" your code to use the code for image extraction. the current code looking for "/Resources/Xobjects/'Forms'" should be very similar to "/Resources/Patterns/" adjusting x_object

You should also confirm your code works with patterns within forms

Also, It might be interesting to extract the thumbnail and confirm that no other images are ignored

[0xNath]

Instead of putting the files directly into this repository, should I host the files on github or somewhere else so they get downloaded?

In general, files should go into the sample files if you own all copyrights and are okay with the license terms there. Unfortunately, merging permissions are more restricted over there, thus I am not able to merge anything there myself.

The usual approach we take apart from this is to upload to an issue or PR and download the corresponding files from the GitHub URLs. Unfortunately, this will not work for image files any more as far as I know as GitHub started to add JWTs to their URLs which expire accordingly and are user-specific.

I reused images that were already present in the repo so it should be good in term of rights. The issue is that only office convert images to jpeg when included inside a document, so I had to re-upload them once converted to pass content verification in the test unit.

Let's wait to see if my PR on the other repo get merged then, thanks.

[0xNath]

Can you rename the title of your PR : this is not a a STYle but an ENHancement. You should also edit your first thread to set you "closes " info to ease closing

I was not sure which one to use, thanks.

you should be able to "factorize" your code to use the code for image extraction. the current code looking for "/Resources/Xobjects/'Forms'" should be very similar to "/Resources/Patterns/" adjusting x_object

You're right, I didn't thought about it, let's do that.

You should also confirm your code works with patterns within forms

It will probably not in the current state, I'll check what are forms, thank you.

Also, It might be interesting to extract the thumbnail and confirm that no other images are ignored

Yes I agree. I'll need to read a bit more the PDF standard, currently I've only read the part about patterns and image objects so I'm lacking a lot about how images object are used/located. Maybe it would be nicer to have an issue for each case where images are missing and have their matching PR ?

[0xNath]

I have tested onlyoffice forms and images inside it are not taken into account.

Images inside PDF from libreoffice without and with forms are taken into account, as well as images from word forms.

[pubpub-zz]

@0xNath
Can you please save your test files in this thread and modify your PR to get them from the URLs. for the name, use iss2613a.pdf, iss2613b.pdf, ...

[0xNath]

I'll add the code for extracting the images from the form PDF and the corresponding test. I'm going to follow the same logic as for the patterns, but with "/Annots/.../..." as image identifiers

iss2613-onlyoffice-form.pdf
iss2613-onlyoffice-standardImages.pdf

[0xNath]

Images to test against

[codecov]

Codecov Report

Attention: Patch coverage is 85.18519% with 8 lines in your changes are missing coverage. Please review.

Project coverage is 94.89%. Comparing base (c227b0c) to head (24218c8).

Files	Patch %	Lines
pypdf/_page.py	85.18%	2 Missing and 6 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2637      +/-   ##
==========================================
- Coverage   94.97%   94.89%   -0.08%     
==========================================
  Files          50       50              
  Lines        8331     8371      +40     
  Branches     1669     1689      +20     
==========================================
+ Hits         7912     7944      +32     
- Misses        260      262       +2     
- Partials      159      165       +6     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.