KeyError: '/Root' due to invalid start of xref

[owurman]

I was trying to get the pages for the attached PDF but received a KeyError: '/Root'. The file appears to be encrypted to me, but pdf.is_encrypted is False.

Environment

Which environment were you using when you encountered the problem?

$ python -m platform
macOS-10.16-x86_64-i386-64bit

$ python -c "import pypdf;print(pypdf.__version__)"
3.7.0

Code + PDF

import pypdf

reader = pypdf.PdfReader("641-Attachment-B-Pediatric-Cardiac-Arrest-8-1-2019.pdf")
assert (not reader.is_encrypted)
len(reader.pages)

Share here the PDF file(s) that cause the issue. The smaller they are, the
better. Let us know if we may add them to our tests!
641-Attachment-B-Pediatric-Cardiac-Arrest-8-1-2019.pdf

It's a public document so it should be fine to add to your tests.

Traceback

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "pypdf/_page.py", line 2155, in __len__
    return self.length_function()
           ^^^^^^^^^^^^^^^^^^^^^^
  File "pypdf/_reader.py", line 452, in _get_num_pages
    self._flatten()
  File "pypdf/_reader.py", line 1186, in _flatten
    catalog = self.trailer[TK.ROOT].get_object()
              ~~~~~~~~~~~~^^^^^^^^^
  File "pypdf/generic/_data_structures.py", line 291, in __getitem__
    return dict.__getitem__(self, key).get_object()
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
KeyError: '/Root'

[MartinThoma]

A PDF file should look like this:

with:

A trailer giving the location of the cross-reference table and of certain special objects within the body of the
file

The trailer of that file is empty, thus the error.

[MartinThoma]

This command fixes it:

gs -o repaired.pdf -sDEVICE=pdfwrite -dPDFSETTINGS=/prepress 641-Attachment-B-Pediatric-Cardiac-Arrest-8-1-2019.pdf

See https://superuser.com/q/278562/64857

[owurman]

Thanks. Is it a reasonable ask that a better message be given if the trailer is missing? I'm guessing that actually repairing the PDF as ghostscript does is beyond the scope of what you want pypdf to do...

[MartinThoma]

The problem is not this specific case. Sure, we can (and regularly do) add robustness improvements. It's just a never ending story. There is an infinite number of ways the standard can be broken

[MartinThoma]

I was hoping that we could use similar techniques as web browsers / beautiful soup does for HTML for that problem. I just didn't have the time to look into it so far.

[pubpub-zz]

@owurman
I've prepared a PR to improve robustness if you want to try it.

[MartinThoma]

The robustness improvement was just added to main and will be released this weekend with pypdf>3.7.1.