PdfFileMerger generates blank pages on Python 3 when given file handlers

[mluszczyk]

I'm using the merger like in the snippet below. Whatever input PDF file I provide, I get a PDF with the same number of pages, but all of them are blank (even if in the input file they are not blank).

from PyPDF2 import PdfFileMerger

merger = PdfFileMerger()

INPUT_PATH = "/tmp/a-pdf-file.pdf"
OUTPUT_PATH = "/tmp/output.pdf"

with open(INPUT_PATH, 'rb') as input_file:
    merger.append(input_file)

with open(OUTPUT_PATH, 'wb') as out:
    merger.write(out)

Does not happen on python 2.7, but happens in exactly the same environment on python 3.5. Tested on PyPDF2==1.26.0.

One finding I have is that the if branch on line 116 is never entered in Python 3, but it is entered on Python 2. The reason is that isinstance(open("/tmp/x"), file) is evaluated to True on Python 2, but isinstance(open("/tmp/x"), io.FileIO) is evaluated to False on Python 3 (note from io import FileIO as file at the top of the file).

[keliomer]

@mstamy2
Seems the fix (#294 ) hasn't been made available on the version available through PyPI.
Experiencing the same issue as #293 with Python 3.6.3.

I uninstalled and reinstalled through pip, and still ended up having to manually changing the line @mluszczyk did in their PR.

Could we reopen this issue until the fix is live and pip-able?

[cgarbin]

Passing along a solution I got from https://stackoverflow.com/a/49927541/336802: keep the input files open while writing the output file.

This works with the currently released version of PyPDF2.

[kislyuk]

Thank you @cgarbin for mentioning this!

FYI @mstamy2 this is still an issue. Pages turn out blank if the underlying fd was closed, instead of throwing an error.