pdf:charsPerPage is broken for nested PDFs #486

Activity

1. chrismattmann

   changed the title [-]`pdf:charsPerPage` is fuckedup on nested PDFs[/-] [+]`pdf:charsPerPage` is broken for nested PDFs[/+] on Apr 26, 2026
2. chrismattmann commented on Apr 26, 2026

   

   chrismattmann

   on Apr 26, 2026

   Owner

   More actions

   Please see my comments on #487
3. chrismattmann

   added this to the tika-next milestone on Apr 26, 2026
4. chrismattmann

   self-assigned this

   on Apr 26, 2026
5. chrismattmann

   added

   bug

   help wanted

   question

   on Apr 26, 2026
6. ExtReMLapin commented on Apr 26, 2026

   

   ExtReMLapin

   on Apr 26, 2026

   Author

   More actions

   I'm pretty sure the title was fine
7. chrismattmann commented on Apr 26, 2026

   

   chrismattmann

   on Apr 26, 2026

   Owner

   More actions

   I'm pretty sure the title was fine

   And I'm not sure, actually, which is why I changed it. Please don't use profanity in issue titles or anywhere in this repo.

   😕React with 😕1Reacted by ExtReMLapin
8. chrismattmann commented on Jun 8, 2026

   

   chrismattmann

   on Jun 8, 2026

   Owner

   More actions

   I don't think this is a bug and in fact changing this would change the default behavior that people have been using for a while. Thanks anyways!
9. chrismattmann

   closed this as completedon Jun 8, 2026
10. ExtReMLapin commented on Jun 8, 2026

    

    ExtReMLapin

    on Jun 8, 2026

    Author

    More actions

    I don't think this is a bug

    Have you tried running it with Tika rest api and using the Tika python lib, it's clear to us at the office it's a bug
11. ExtReMLapin commented on Jun 8, 2026

    

    ExtReMLapin

    on Jun 8, 2026

    Author

    More actions

    Alright now re-reading the whole thing I feel like I forgot to show a before-after delta and compare it to tika rest behavior.

    (Out of bandwidth with everything tbh)

    ---

    Follow-up with actual output. Clean install, stock tika==3.1.0 + tika-server 3.1.0, the PDF above (very first post), no patches:

    from tika import parser
    v = parser.from_file("d0305b35-7e69-4e1d-a9ed-8a59178d1ff8.pdf")["metadata"]["pdf:charsPerPage"]
    print(len(v), sorted({type(e).__name__ for e in v}))
    print(type(v[0]).__name__, type(v[-1]).__name__)
    sum(int(x) for x in v)

    Before (current main):

    193 ['list', 'str']
    str list
    TypeError: int() argument must be a string ... not 'list'
    

    After #487:

    381 ['str']
    str str
    1151171
    

    This PDF is 3 nested PDFs. /rmeta returns charsPerPage as a list for each (lengths 191, 17, 173). _parse keeps the first one flat, then appends each later list as a single element instead of concatenating, so you end up with 191 strings followed by 2 lists-of-strings in the same value. Normal iteration over it breaks (the TypeError above). #487 flattens it to 381 strings and it works.
12. chrismattmann

    modified the milestones: tika-next, 3.3.2 on Aug 1, 2026