apply_redactions() deleting text outside of annoted box

[rnik12]

Description of the bug

Check the attached pdf Original -> Annotated -> Redacted.

You can see the insult word is getting redacted -

Original -
"The woman whom it was intended to insult or whose privacy was intruded upon."

Redacted -
"The woman whom it was intended to ult or whose privacy was intruded upon."

How to reproduce the bug

Code

import fitz
import json

fitz.TOOLS.set_small_glyph_heights(True)

filepath = f"original.pdf"

doc = fitz.open(filepath)


def get_redact_box(bbox):
    return [round(x, 2) for x in bbox]


LEFT_START = 115
RIGHT_END = 482


def redact_text_outside_bounds(page_number):
    """Prints text and line numbers for a specific page of a PDF."""

    if page_number < 1 or page_number > len(doc):
        print("Invalid page number.")
        return

    page = doc[page_number - 1]  # Get the page object (zero-based indexing)
    blocks = page.get_text("dict", sort=True)["blocks"]

    for block in blocks:
        bbox = block["bbox"]
        x0, y0, x1, y1 = bbox

        for line in block["lines"]:
            bbox = line["bbox"]
            x0, y0, x1, y1 = bbox

            if x0 <= x1 and (x1 <= LEFT_START or x0 >= RIGHT_END):
                page.add_redact_annot(
                    get_redact_box(bbox),
                    fontname="helv",
                    fontsize=8,
                    align=fitz.TEXT_ALIGN_CENTER,
                )

        # text = ""
        # for line in block["lines"]:
        #     for d in line["spans"]:
        #         text += " " + d["text"]
        # print()
        # print(bbox)
        # print(text)

    page.apply_redactions()


if __name__ == "__main__":
    pages = range(1, doc.page_count+1)
    # pages = range(128, 129)
    for page_no in pages:
        redact_text_outside_bounds(page_no)

    doc.save("test.pdf")

original.pdf
annotated.pdf
redacted.pdf

PyMuPDF version

1.23.26

Operating system

MacOS

Python version

3.12

[JorjMcKie]

Cannot reproduce.
Tried several of my own rectangles and there was no problem.
To reduce my effort to understand your code, please provide an example where you simply draw a rectangle and then demonstrate that unintended text was deleted.

[rnik12]

@JorjMcKie I've updated the code. Made page range 1 to 2. It is reproducing. Can you check ?

[JorjMcKie]

This is the result after recent code version - all the number are removed and nothing else.
I suppose this is intended?
test.pdf

[rnik12]

@JorjMcKie that's strange. Attaching my Macbook M1 screenshot. Using python 3.12 and latest 1.23.26.

Maybe M1 architecture is causing this issue ? dependency / floats not getting compiled properly.

[JorjMcKie]

let me it on Linux again - don't have a Mac to cross check.

[JorjMcKie]

Got it!
The problem is the base library: that problem is fixed in MuPDF. The next version of PyMuPDF will be linked to it, so the error will be fixed.

[julian-smith-artifex-com]

Fixed in 1.24.0.