Request: Have .extract_text() return an empty string ('') instead of None in the case of no text found in a PDF

[tungph]

pdfplumber/pdfplumber/utils.py

Lines 372 to 374 in 002500a

	chars = to_list(chars)
	if len(chars) == 0:
	return None

I would like to make a small suggestion here to avoid for None checking in the calling code, extract_text utils must return an empty string '' instead of None in case of no text found in a pdf.

[jsvine]

Thank you for this proposal, @tungph. I think it makes sense, but I want to be careful that we're not overlooking important use-cases. Especially: Are there instances where it would be important to distinguish between '' and None? Any thoughts on this, @samkit-jain?

[samkit-jain]

My preference would be to keep the existing None behaviour because None and '' can mean 2 different things. Also, this should be considered as a breaking change as there will be existing workflows that might be relying on the method returning a None.

An important question should also be whether pdfplumber can even return an empty string? I think so yes as there exists a Unicode code for an empty string and if a PDF consists of that single character, an empty string will be returned. It would be ambiguous as it can mean both that no character exists or a null character exists.

[jsvine]

Thanks @samkit-jain. That (needing to distinguish between '' and None) had also been my initial instinct — and, if I recall correctly, why I had initially implemented it that way. I'm starting, however, to change my mind, for a couple of reasons:

• Is there a meaningful, end-user-relevant difference between a page (or table cell, etc.) that contains no characters vs. one that contains only null/'' characters? In that vast majority of cases, I think not; what's relevant to the user is just that there is no text there.

• There may be some instances where that distinction is important, but I think that a simple examination of page.chars is better suited to the detection of null/'' characters that .extract_text(). One reason: .extract_text() (in its current implementation) still condenses multiple adjacent null/'' characters into a single ''.

Still, I'm not quite 100% there. Is '' a more Pythonic/philosophically-accurate way than None of representing a page/cell/etc. with no characters in it? I'd be interested for more pdfplumber users/watchers to weigh in.

And, as you point out: This, if implemented, would constitute a breaking change for many workflows. It would probably have to wait for 0.6.0.

[jsvine]

Hi @tungph, and thanks again for this suggestion. This is now the default behavior for .extract_text(...) as of v0.6.0. (See cb9900b and CHANGELOG.md, with an h/t to you.)