Piximap program crash

[JordanGarske]

Description of the bug

What is happening is that when I read from the PDF, I use the rectangle information to collect color data. Recently, however, I encountered an issue with one of the PDF collections: while collecting the color information, the program crashes unexpectedly without any error messages. I traced the problem to the color_topusage function. Specifically, the error appears to originate from the call to this function, where the issue seems to be related to color_count -> JM_color_count -> oldpix = read_sample(pm, s, n).

How to reproduce the bug

I traced the problem to the color_topusage function. Specifically, the error appears to originate from the call to this function, where the issue seems to be related to color_count -> JM_color_count -> oldpix = read_sample(pm, s, n).

PyMuPDF version

1.24.10

Operating system

Windows

Python version

3.12

[JorjMcKie]

Please do provide reproducing file and code.

[JordanGarske]

case.pdf
import pymupdf
from pymupdf import Page, table, Rect , Pixmap, sRGB_to_rgb

from pymupdf.utils import get_pixmap, get_page_pixmap
#Similar file with error; it loops on a 4-page document and crashes

on page 2 of 4.
pdf = pymupdf.open(f'./case.pdf')
for i in range(0, len(pdf) ):
page = pdf.load_page(i)
print(page)
for annot in page.get_drawings():
if page.get_textbox(annot['rect'] ) :
rect = annot['rect']
pix: Pixmap = get_pixmap(page, clip= rect)
colorBytes = pix.color_topusage()
print('done')

[JorjMcKie]

This script works:

import pymupdf

doc = pymupdf.open("case.pdf")
page = doc[1]
paths = page.get_drawings()
for i, p in enumerate(paths):
    rect = p["rect"]  # best check for emptiness already here!
    text = page.get_textbox(rect)
    if text:
        pix = page.get_pixmap(clip=rect)
        if pymupdf.IRect(pix.irect).is_empty:
            print(i, "pixmap has empty area - skipping")
            continue
        print("path", i, pix.color_topusage())

So the reason for the crash is a pixmap that covers no area, and the color_topusage() method does not protext itself against nonsense input.
Which must be fixed of course: we will raise an exception in such cases.

Otherwise please take note of the following comments:

• You are iterating over vector graphics which can consist of just a line - which obviously cannot contain anything, including text. You could / should ignore those right away. Lines may be not parallel to either the x- or the y-axis. In that case, the p["rect"]is not empty, so everything appears to work ... but still the underlying line contains no text, while p["rect"] may have text inside. Can't imagine you want those false alarms.
• Depending on the circumstances (please read the relevant documentation about Rect / IRect objects and their relationship) a non-empty rectangle may be wrapped by an empty IRect (!).
• Your use of the term "annotation" may lead to misconceptions: this is a PDF term reserved for something entirely different. Best stay with line-art, vector graphic or drawing.