Assert fails when getting the mediabox property for certain PDFs

[Paethon]

Hi

We are processing quite a lot of PDFs, and from time to time we see the following assert fail on specific PDFs when trying to get the mediabox property of a page.

pypdf/pypdf/generic/_rectangle.py

Line 24 in 27edc06

assert len(arr) == 4

Here is the content of page:

{'/Contents': [IndirectObject(34, 0, 131870331607200)],
 '/CropBox': [0, 0, 595, 841, 0, 0, 595, 841],
 '/MediaBox': [0, 0, 595, 841, 0, 0, 595, 841],
 '/Parent': IndirectObject(1, 0, 131870331607200),
 '/Resources': IndirectObject(5, 0, 131870331607200),
 '/Type': '/Page'}

Is this "just" a malformed PDF (it opens without problem in a wide range of pdf readers)? Unfortunately, I can't share the PDF, since it contains sensitive customer information.

[stefan6419846]

I just had a look at the PDF specification (ISO 32000-2:2020/PDF 2.0): According to table 31 in section 7.7.3.3, the MediaBox is a rectangle:

(Required; inheritable) A rectangle (see 7.9.5, "Rectangles"), expressed in default user space units, that shall define the boundaries of the physical medium on which the page shall be displayed or printed (see 14.11.2, "Page boundaries").

Looking into section 7.9.5, we can read:

A rectangle shall be written as an array of four numbers giving the coordinates of a pair of diagonally opposite corners.

The PDF 1.7 specification (ISO 32000-1:2008) states the same here.

For this reason, your PDF files are indeed broken, although we might consider truncating rectangles to their first four elements and issue a warning instead. Feel free to submit a corresponding PR.

Just out of curiosity: Do you have any information on the origin/software generating these faulty files?

[Paethon]

Yes, it says: Producer: "pypdf" 😂

Although I have the feeling that they have a different origin and were just processed afterwards using pypdf (but not 100% sure)

[stefan6419846]

Ouch. This might have been an issue from our side, but does not necessarily have to be ;)

[Paethon]

I'll see if I can open a PR in the next few days 👍

[sjudd]

I've run into this as well, my media box looks slightly different:

/CropBox: [
0,
0,
612,
792,
0,
0,
612,
792,
0,
0
],

/MediaBox: [
0,
0,
612,
792,
0,
0,
612,
792,
0,
0
],

Sadly I also can't directly share the file.

[sjudd]

I sent a PR for the read issue reported here. I also realized that pypdf is the source for the corrupt boxes and I should be able to reproduce it reliably locally, so I'll look at a fix for the write side issue too.

[sjudd]

Here's the how I'm reproducing. I have 5 input files (again all with customer data :/):

pdf_writer = PdfWriter()
for i in range(1, 6):
    path = <path_to_file_i>
    with PdfReader(path) as pdf_reader:
        for page in range(len(pdf_reader.pages)):
            pdf_writer.add_page(pdf_reader.pages[page])
output_path = <anywhere>
with open(output_path, "wb") as f:
    pdf_writer.write(f)

reader = PdfReader(output_path)
for i, page in enumerate(reader.pages, 1):
    page.mediabox # this eventually throws

One fun thing, if I read the mediabox value from each page before I add it to the pdf_writer, the merged pdf mediabox is no longer corrupt. E.g. this never throws:

pdf_writer = PdfWriter()
for i in range(1, 6):
    path = <path_to_file_i>
    with PdfReader(path) as pdf_reader:
        for page in range(len(pdf_reader.pages)):
            print(i, pdf_reader.pages[page].mediabox) # Added!!
            pdf_writer.add_page(pdf_reader.pages[page])
output_path = <anywhere>
with open(output_path, "wb") as f:
    pdf_writer.write(f)

reader = PdfReader(output_path)
for i, page in enumerate(reader.pages, 1):
    page.mediabox # this now never throws!

[sjudd]

I've narrowed this down to the get_object call:

pypdf/pypdf/_writer.py

Line 487 in 27edc06

"PageObject", page_org.clone(self, False, excluded_keys).get_object()

. Here's the page instance printed before and after it (I removed the clone part):

27 {'/Type': '/Page', '/Parent': IndirectObject(2, 0, 4371288464), '/Resources': IndirectObject(16, 0, 4371288464), '/Contents': [IndirectObject(15, 0, 4371288464)], '/MediaBox': [0, 0, 612, 792], '/CropBox': [0, 0, 612, 792]} [0, 0, 612, 792]
27 {'/Type': '/Page', '/Resources': IndirectObject(242, 0, 4367985936), '/Contents': [IndirectObject(250, 0, 4367985936)], '/MediaBox': [0, 0, 612, 792, 0, 0, 612, 792], '/CropBox': [0, 0, 612, 792, 0, 0, 612, 792]} [0, 0, 612, 792, 0, 0, 612, 792]

I think I'm going to go back to my normal job for now, but let me know if I can help reproduce later.