Passing a string as a page number raises IndexError instead of TypeError.

[bwagner]

Description of the bug

import fitz
doc = fitz.open("some.pdf")
doc["0"]

raises IndexError:

IndexError: page 0 not in document

Which is misleading, since the page 0 happens to be in the document, it's just being addressed wrongly.

How to reproduce the bug

import fitz
doc = fitz.open("some.pdf")
doc["0"]

raises IndexError:

IndexError: page 0 not in document

Which is misleading, since the page 0 happens to be in the document, it's just being addressed wrongly.

PyMuPDF version

1.23.9rc1

Operating system

MacOS

Python version

3.12

[bwagner]

I tried to fix it, but my changes don't come through when I install the fixed package. When patching the installed file directly, however, my changes are effective. If anyone could recommend what I need to do to have my fix to come through when installing, please let me know.

[julian-smith-artifex-com]

PyMuPDF in git changed a week or two ago to default to installing the rebased implementation as module fitz, not the classic implementation.

So you should probably be patching src/__init__.py, not src_classic/fitz_old.i.

[Apart from that, i'm wondering whether it might be better to simply do assert isinstance(i, int), because this feels more appropriate for what is a programming error, not a normal runtime error?]

[bwagner]

Thank you!

[bwagner]

I signed the CLA. Is anything else required for this PR to be applied?

[JorjMcKie]

I haven't looked at the proposed code yet. But as a reminder:
Integers are not the only allowed item types for loading a page!
Always valid is also the form page = doc[(chapter, pno)] with a tuple of two integers.
Therefore checking for integer only is incorrect.

[julian-smith-artifex-com]

[Reopening until the fix is released.]

[julian-smith-artifex-com]

This is fixed in release candidate 1.23.9rc2; see: #2996

[julian-smith-artifex-com]

Fixed in 1.23.9.