apply_redactions() erases neighboring text which is not redacted

[tigrankh]

Hi,

I'm facing a strange issue when i apply redaction on a PDF.
I have the following text:

in my experiment i redact the word "von", which overlaps with bbox of the word "Niederschrift" below it and erases some of its letters, as shown below:

In one of your github issues I saw you recommended using fitz.TOOLS.set_small_glyph_heights(True), which updated the bboxes and they are not overlapping anymore.

However, when i apply redaction to the word "von" it still erases the word below.
I'm suspecting that apply_redactions() isn't just operating on the bbox rectangle, but can't really say for sure. bbox coords of those words do not overlap at all

Could you please help with this ?

[JorjMcKie]

There is a solution to this, but let me first move this to the "Discussions" tab.

[JorjMcKie]

I assume you have used the lower rectangle produced after fitz.TOOLS.set_small_glyph_heights(True) for the redaction as well?

The reason for this effect is the logic behind the redaction algorithm (inside MuPDF code): it is based on characters overlapping redaction rectangles - as opposed to characters being contained. If the PDF creator was unfortunate enough choosing line distances much smaller than the font's natural one, there will always be pesky situations like yours.

Shrinking your redaction rect (height) even more will however do the trick. Your redaction rect can be as small as you want, it doesn't need to cover the word "von". For example, reduce its height by 20% from its top and its bottom:

delta = rect.height * 0,2
page.add_redact_annot(rect + (0,delta,0,-delta))

[JorjMcKie]

the small_glyph_heights, i enable right before loading the binary into fitz Document.

If you do that, every word is encircled like this:

instead of like this:

You must have a very unfortunate situation in your doc if you still see inter-line overlaps.

[tigrankh]

I guess i indeed found a bad scenario.
When i use draw_rect() with the new values affected by small_glyph_heights, i see that the rectangles don't overlap, just like in your red rectangles. but as soon as i apply redaction on the word it erases the word.

in my experiment i had to use the font size to avoid overlaps

[JorjMcKie]

If you used a red rectangle as the redaction rectangle and still got unwanted erasures, your PDF must have really extremely compacted text.
If you want, send me an example page - I still wonder was is going on.
Also don't understand your mention of font size ...

[tigrankh]

so the height of rectangle computed by small_glyph_heights was 7.92, which was still causing erasing of the word below.
I used page.get_fonts() and saw that there are 4 fonts used throughout the page and one of them has fontsize of 7. so i assumed that its this same word.
I have set the height of redaction rectangle to 7 and it wasn't erasing the word below it anymore.

I'm sorry, but i cant share the PDF.
If you have ideas in mind, i'm happy to do whatever experiments needed to debug this.

Thank you very much for help !

[JorjMcKie]

eight of redaction rectangle to 7

this is a variation of my 20% suggestion further up. If that works, fine.