PdfWriter as context manager while passing existing file not working in 5.2.0

[AlbertUnruh]

After upgrading pypdf from 5.1.0 to 5.2.0 my function to merge multiple pdfs didn't merge my files anymore.

I wasn't able to find anything in the change-log about any breaking changes.

Environment

Which environment were you using when you encountered the problem?

$ python -m platform
Windows-10-10.0.19045-SP0

$ python -c "import pypdf;print(pypdf._debug_versions)"  # 5.1.0
pypdf==5.1.0, crypt_provider=('local_crypt_fallback', '0.0.0'), PIL=11.1.0

$ python -c "import pypdf;print(pypdf._debug_versions)"  # 5.2.0
pypdf==5.2.0, crypt_provider=('local_crypt_fallback', '0.0.0'), PIL=11.1.0

Code + PDF

This is a minimal, complete example that shows the issue:

def merge_pdfs(*pdfs: Path, destination: Path):
    """Merge multiple pdfs into one big pdf at the given destination."""
    with PdfWriter(destination) as merger:
        merger.add_metadata(PdfReader(pdfs[0]).metadata)
        for pdf in pdfs:
            merger.append(pdf)

# merge_pdfs() works in 5.1.0 just fine
# merge_pdfs() doesn't work in 5.2.0 anymore

I don't think that any specific files are required.
I tested it with multiple files and it didn't work.

[stefan6419846]

Thanks for the report. Unfortunately, I have not been able to reproduce this with a simple setup:

from pypdf import PdfReader, PdfWriter


pdfs = ['file1.pdf', 'file2.pdf']

with PdfWriter('out.pdf') as merger:
    merger.add_metadata(PdfReader(pdfs[0]).metadata)
    for pdf in pdfs:
        merger.append(pdf)

#2909 might be related.

Please provide some further details and the PDF files which show this behavior for you to further analyze this issue.

[AlbertUnruh]

I'm merging all pdfs into the first one.

For my setup I would make a call like merge_pdfs(pdf1, pdf2, destination=pdf1).

Is it possible that it has something to do with the order of closing and writing to pdf1?

I'm not home atm, otherwise I would check it myself.

[stefan6419846]

I have been able to reproduce this with the output file already existing - it becomes even more obvious once you set the output file to an existing non-PDF file. As mentioned before, this is related to #2909 and #2913, although only #2909 is relevant as part of the release.

The underlying issue is more or less bad legacy API design which requires guessing the intent of the user. Ideally, the parameters to PdfWriter would be keyword-only to completely eliminate the need of guessing, but this requires a proper deprecation process with some more complex intermediary handling: #2913 (comment)

I have not yet decided how to go forward with this specific case, although input from you, @alexaryn and other interested parties is greatly appreciated. As a general rule of thumb, using clone_from or a non-existing output file usually is the safest approach.