Fix out-of-bound access on some PDFs.

[Flameeyes]

Replace the non-emptiness check with a minimum length check — you can't get the second to last item in a list of less than two items.

[pietermarsman]

Hi @Flameeyes, thanks for the suggestion. Can you create an issue first? Maybe also add an example PDF that shows the problematic behavior.

I'm closing this PR since it might introduce new problems. E.g. if x == None the len(x) fails.

[Flameeyes]

Unfortunately, I cannot provide an example PDF because the PDF having an issue is a bank statement, which is one of the strongest forms of PII. That's the reason why I didn't file an issue — there's nothing I can provide beside the stack trace.

I'd like to point out that closing a first-contribution pull request is not a very friendly presentation. GitHub has a review system where you can request changes, instead of closing a pull request. That's significantly less adversarial than closing.

Furthermore, your assertion regarding the change's ability to introduce new problem is actually incorrect. x is assigned an empty list ([]) at the beginning of the block the if is in, so it can never be None. So len(x) is always valid.

You're welcome to reopen and merge this pull request if that made you change your mind.

[pietermarsman]

I was indeed a bit quick in closing this PR. By looking a bit more to the context of your change I now see that the len(x) check is indeed a much better alternative than checking for an empty list. Sorry for my impatience!

[pietermarsman]

To finish up this PR, can you add a line to the changelog?

[Flameeyes]

To finish up this PR, can you add a line to the changelog?

Yup, done!

Thank you for circling back on this one :)

[pietermarsman]

Thanks for the work (and the patience)!