apply_redactions() does not work as expected

[nsklei]

Description of the bug

When using apply_redactions(images=pymupdf.PDF_REDACT_IMAGE_NONE) I get several "MuPDF error: syntax error: cannot find XObject resource" errors and as well there are some pages which are completely empty, altough all pages originally contain images.

How to reproduce the bug

import pymupdf
from io import BytesIO
from pathlib import Path

file_path = "path\to\Example_PDF.pdf"
output_path = "path\to\Example_PDF_redacted.pdf"

new_doc = pymupdf.open(file_path)

for num, page in enumerate(new_doc):
    print(f"Page {num + 1} - {page.rect}:")
    
    for image in page.get_images(full=True):
        print(f"  - Image: {image}")

    redact_rect = page.rect

    if page.rotation in {90, 270}:
        redact_rect = pymupdf.Rect(0, 0, page.rect.height, page.rect.width)

    page.add_redact_annot(redact_rect)
    page.apply_redactions(images=pymupdf.PDF_REDACT_IMAGE_NONE)

byte_stream = BytesIO()
new_doc.save(byte_stream)
byte_stream.seek(0)

Path(output_path).write_bytes(byte_stream.getvalue())

The code above prints the following information:

Page 1 - Rect(0.0, 0.0, 598.3200073242188, 813.5999755859375):
  - Image: (22, 0, 554, 754, 8, 'ICCBased', '', 'Im0', 'DCTDecode', 0)
  - Image: (23, 43, 554, 754, 8, 'ICCBased', '', 'Im1', 'DCTDecode', 0)
Page 2 - Rect(0.0, 0.0, 598.3200073242188, 816.47998046875):
  - Image: (25, 0, 554, 756, 8, 'ICCBased', '', 'Im001', 'DCTDecode', 0)
  - Image: (26, 44, 554, 756, 8, 'ICCBased', '', 'Im002', 'DCTDecode', 0)
Page 3 - Rect(0.0, 0.0, 815.760009765625, 596.8800048828125):
  - Image: (28, 0, 553, 756, 8, 'ICCBased', '', 'Im001', 'DCTDecode', 0)
  - Image: (29, 45, 553, 756, 8, 'ICCBased', '', 'Im002', 'DCTDecode', 0)
Page 4 - Rect(0.0, 0.0, 815.760009765625, 597.5999755859375):
  - Image: (31, 0, 554, 756, 8, 'ICCBased', '', 'Im001', 'DCTDecode', 0)
  - Image: (32, 46, 554, 756, 8, 'ICCBased', '', 'Im002', 'DCTDecode', 0)
Page 5 - Rect(0.0, 0.0, 815.0399780273438, 597.5999755859375):
  - Image: (34, 0, 554, 755, 8, 'ICCBased', '', 'Im001', 'DCTDecode', 0)
  - Image: (35, 47, 554, 755, 8, 'ICCBased', '', 'Im002', 'DCTDecode', 0)
Page 6 - Rect(0.0, 0.0, 806.4000244140625, 598.3200073242188):
  - Image: (37, 0, 554, 747, 8, 'ICCBased', '', 'Im001', 'DCTDecode', 0)
  - Image: (38, 48, 554, 747, 8, 'ICCBased', '', 'Im002', 'DCTDecode', 0)
Page 7 - Rect(0.0, 0.0, 815.0399780273438, 597.5999755859375):
  - Image: (39, 0, 554, 755, 8, 'ICCBased', '', 'Im001', 'DCTDecode', 0)
  - Image: (40, 49, 554, 755, 8, 'ICCBased', '', 'Im002', 'DCTDecode', 0)
MuPDF error: syntax error: cannot find XObject resource 'Im1'

MuPDF error: syntax error: cannot find XObject resource 'Im2'

Page 8 - Rect(0.0, 0.0, 815.760009765625, 596.8800048828125):
  - Image: (41, 0, 553, 756, 8, 'ICCBased', '', 'Im001', 'DCTDecode', 0)
  - Image: (42, 50, 553, 756, 8, 'ICCBased', '', 'Im002', 'DCTDecode', 0)
MuPDF error: syntax error: cannot find XObject resource 'Im1'

MuPDF error: syntax error: cannot find XObject resource 'Im2'

As you can see, each page contains two images. The function should remove all content from the PDF file except the images.
But when saving the byte_stream there are some pages completely empy.

PyMuPDF version

1.24.10

Operating system

Windows

Python version

3.12

[JorjMcKie]

This post cannot be accepted as a bug report because no reproducer file is provided.

[JorjMcKie]

test2.pdf

MuPDF bug report: https://bugs.ghostscript.com/show_bug.cgi?id=708032.

[JorjMcKie]

@nsklei - You are aware that all pages only contain images - no text, no vector graphics.
So your redactions effectively are no-ops!

[nsklei]

Thank you for reviewing my issue and creating a bug report.
The described behaviour in your bug report is correct. I am aware, that all pages only contain images and nothing else, so the redactions should indeed be no-ops in this case.

[JorjMcKie]

I found that removing page rotation avoids the problem:

for page in doc:
    page.add_redact_annot(page.rect * page.derotation_matrix)
    page.remove_rotation()
    page.apply_redactions(images=pymupdf.PDF_REDACT_IMAGE_NONE)

Works without problem.

[nsklei]

Thank you for providing a solution to my problem. I tested your suggestion and it works perfectly :)

[JorjMcKie]

Thanks for the feedback!
I am going to re-open this until the fix itself is publicly available. This is our policy for dealing with issue resolutions.

[sebras]

@JorjMcKie This appears to have been fixed upstream, so can be marked "fix developed"?

[julian-smith-artifex-com]

Fixed in 1.24.11.