BUG: Title sometimes is bytes and not str.

[reformy]

closes #2929

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Please upload report for BASE (main@5b50f47). Learn more about missing BASE report.
Report is 113 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2930   +/-   ##
=======================================
  Coverage        ?   96.40%           
=======================================
  Files           ?       52           
  Lines           ?     8744           
  Branches        ?     1591           
=======================================
  Hits            ?     8430           
  Misses          ?      186           
  Partials        ?      128           

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

[stefan6419846]

Thanks for the PR. I assume that you own all the necessary copyrights for the test file?

For me this mostly looks okay, apart from the review comment left above about the position after the comment. Nevertheless, I am going to wait for @pubpub-zz here to have another look.

[pubpub-zz]

Personnaly I would have prefer to add the str to the BytesStringObject() with trying diffrerent encoding(see §3.8.1) .

we should use CHARSET (

pypdf/pypdf/generic/_base.py

Line 841 in 98aa974

CHARSETS = ("utf-8", "gbk", "latin1")

) adding as first format UTF-16 (based on the spec)

It would then just require to add a call to str in the diffrerent properties of metadata in https://github.com/py-pdf/pypdf/blob/main/pypdf/_doc_common.py

[reformy]

Thank you for your comments.
I've moved the utf-8 to be after the pdfdocencoding.
@pubpub-zz I did not understand your comment. If the whole str-decoding is not in the right place, that a different issue. I've just added another option to decode it.

[stefan6419846]

If the whole str-decoding is not in the right place, that a different issue.

No, this is related to this issue. The suggestion is to revert the current changes and provide ByteStringObject.__str__() accordingly, which would then test the different charsets. At the same time, DocumentInformation._get_text() should be rewritten to check for StringObject and ByteStringObject and - instead of returning the pure value - return str(retval).

[reformy]

I've changed the code as suggested, adding __str__ and calling it from _get_text. This works well.

However, adding utf-16 to the CHARSETS list breaks other things. I don't understand why do we need this and how is this related to my fix.

[stefan6419846]

However, adding utf-16 to the CHARSETS list breaks other things. I don't understand why do we need this and how is this related to my fix.

This is another encoding to consider. You should only consider this in __str__, not add it to the list permanently.