How to do visual debugging of a PDF

[mlissner]

I saw in a post a few days ago (sorry, I lost it), a PDF that had all of the elements outlined in different colors so you could see the edges of each piece of text.

Is there a trick for doing that? I don't think I've seen it in the docs, but it would help with a project I'm doing related to the intersection of rectangles with words.

Thank you!

[JorjMcKie]

did you mean this: https://github.com/pymupdf/PyMuPDF-Utilities/blob/master/text-extraction/layout-demo1.pdf

[mlissner]

I actually just found it. It's this, but same idea:

#318 (comment)

The actual rectangle I'm trying to understand is the first redaction in the attached:

I think the rectangle's bbox spans all three lines and thus includes the text you can see in the screenshot, but I can't be sure. For example, the text on the first and third lines is getting flagged as part of the redaction when it's not under the black part of the rectangle.

Here's the doc if you're curious:

partial_intersection_no_bad_redaction_no_hits.pdf

[JorjMcKie]

This and similar examples are handled in folder https://github.com/pymupdf/PyMuPDF-Utilities/tree/master/word%26line-marking.

The arked text here has probably been found via text search. The returned single rectangles of this method jointly make up the marked area. The highlight annotation's rect (if it actually is a highight annot) is the union of these rectangles and in this case would cover all three complete lines.

[mlissner]

Thanks. That wasn't quite what I was looking for — I'm trying to find text under bad redactions — but I did find this, which sort of helped: https://github.com/pymupdf/PyMuPDF-Utilities/blob/master/text-extraction/layout-analyzer.py. It gave me some of the outlines of the boxes that are in the PDF, but I'm still trying to figure out how to do this properly. I'll keep digging, thank you.

[mlissner]

In the end, the issue I had was that I was doing:

drawings = some_page.get_drawings()

And then:

rect = drawings[0].rect

That totally doesn't work for multi-line redactions like I had. In those cases, you have to read the docs, which explain that the drawing["items"] key is where the real shapes are. Once I started using those, I was able to handle the multi-line redactions no problem.

[JorjMcKie]

you have to read the docs

I could not agree more 😎👍😉