very slow when use 'get_pixmap' for some page

[Aloha0424]

Description of the bug

When I convert a PDF to images, some pages (containing charts like histograms or line graphs, possibly generated by matplotlib or Excel) are particularly slow.

How to reproduce the bug

Code snippet

with fitz.open(file_path) as doc:
            zoom = 1.0
            mat = fitz.Matrix(zoom, zoom)
            for page_no in range(doc.page_count):
                try:
                    page = doc.load_page(page_no)
                    pix = page.get_pixmap(matrix=mat)

The test PDF has been attached.

3RCHLDD2YCPDNLHEV4AVKEPBYJP5UBZB.pdf

PyMuPDF version

1.26.7

Operating system

MacOS

Python version

3.11

[Aloha0424]

Another function call is also very slow on these kinds of pages.
text_dict = page.get_text("rawdict")

[JorjMcKie]

The file has problems: multiple Structure Tree issues.
You should have seen respective MuPDF messages! E.g.

for page in doc:
    pix=page.get_pixmap()
    print(f"{page.number=}, {pymupdf.TOOLS.mupdf_warnings()=}")

    
MuPDF error: format error: No common ancestor in structure tree

page.number=0, pymupdf.TOOLS.mupdf_warnings()='format error: No common ancestor in structure tree\nstructure tree broken, assume tree is missing'
page.number=1, pymupdf.TOOLS.mupdf_warnings()=''
MuPDF error: format error: No common ancestor in structure tree
...

This script removes the structure tree information and voilà all pages are processed in fractions of a second:

import pymupdf, time

doc = pymupdf.open("test.pdf")
cat = doc.pdf_catalog()
# remove the broken StructTreeRoot entry
doc.xref_set_key(cat, "StructTreeRoot", "null")
for page in doc:
    t0 = time.time()
    pix = page.get_pixmap()
    t1 = time.time()
    print(f"Page {page.number}: Rendered in {t1-t0:.3f} seconds")
    pix.save("output_page_%d.png" % page.number)
doc.close()

Log:

Page 0: Rendered in 0.008 seconds
Page 1: Rendered in 0.006 seconds
Page 2: Rendered in 0.004 seconds
Page 3: Rendered in 0.003 seconds
Page 4: Rendered in 0.002 seconds
Page 5: Rendered in 0.002 seconds
Page 6: Rendered in 0.002 seconds
Page 7: Rendered in 0.002 seconds
Page 8: Rendered in 0.003 seconds
Page 9: Rendered in 0.002 seconds
Page 10: Rendered in 0.020 seconds
Page 11: Rendered in 0.002 seconds
Page 12: Rendered in 0.003 seconds
Page 13: Rendered in 0.002 seconds
Page 14: Rendered in 0.002 seconds
Page 15: Rendered in 0.001 seconds
Page 16: Rendered in 0.002 seconds
Page 17: Rendered in 0.002 seconds
Page 18: Rendered in 0.002 seconds
Page 19: Rendered in 0.002 seconds
Page 20: Rendered in 0.002 seconds
Page 21: Rendered in 0.002 seconds
Page 22: Rendered in 0.002 seconds
Page 23: Rendered in 0.001 seconds
Page 24: Rendered in 0.001 seconds
Page 25: Rendered in 0.002 seconds
Page 26: Rendered in 0.002 seconds
Page 27: Rendered in 0.001 seconds
Page 28: Rendered in 0.004 seconds
Page 29: Rendered in 0.002 seconds
Page 30: Rendered in 0.005 seconds
Page 31: Rendered in 0.004 seconds
Page 32: Rendered in 0.003 seconds
Page 33: Rendered in 0.003 seconds
Page 34: Rendered in 0.002 seconds
Page 35: Rendered in 0.008 seconds
Page 36: Rendered in 0.007 seconds
Page 37: Rendered in 0.001 seconds
Page 38: Rendered in 0.002 seconds
Page 39: Rendered in 0.002 seconds
Page 40: Rendered in 0.003 seconds
Page 41: Rendered in 0.002 seconds
Page 42: Rendered in 0.002 seconds
Page 43: Rendered in 0.002 seconds
Page 44: Rendered in 0.001 seconds
Page 45: Rendered in 0.002 seconds
Page 46: Rendered in 0.002 seconds
Page 47: Rendered in 0.003 seconds
Page 48: Rendered in 0.001 seconds
Page 49: Rendered in 0.002 seconds
Page 50: Rendered in 0.003 seconds
Page 51: Rendered in 0.002 seconds
Page 52: Rendered in 0.002 seconds
Page 53: Rendered in 0.001 seconds
Page 54: Rendered in 0.001 seconds
Page 55: Rendered in 0.001 seconds
Page 56: Rendered in 0.001 seconds
Page 90: Rendered in 0.006 seconds
Page 91: Rendered in 0.012 seconds
Page 92: Rendered in 0.001 seconds
Page 93: Rendered in 0.012 seconds
Page 94: Rendered in 0.003 seconds
Page 95: Rendered in 0.009 seconds
Page 96: Rendered in 0.004 seconds
Page 97: Rendered in 0.008 seconds
Page 98: Rendered in 0.004 seconds
Page 99: Rendered in 0.011 seconds
Page 100: Rendered in 0.005 seconds

[Aloha0424]

I didn’t see any error messages when I run these code below：

for page in doc:
    pix=page.get_pixmap()
    print(f"{page.number=}, {pymupdf.TOOLS.mupdf_warnings()=}")

When I add these code lines, it really works!!!

cat = doc.pdf_catalog()
# remove the broken StructTreeRoot entry
doc.xref_set_key(cat, "StructTreeRoot", "null")

Thanks a lot.
By the way, should I always add these code lines before reading a PDF?

[JorjMcKie]

By the way, should I always add these code lines before reading a PDF?

Hm, not easy to answer. Some background:
A PDF structure tree is a hierarchical map of all meaningful elements in a PDF - like headings, paragraphs, lists, tables, and figures - used to convey the document's logical reading order and semantics. It is optional.
Like everything else in PDF, it can be broken or misused causing more damage than benefit ... like in your case.
PyMuPDF's current / normal extraction and rendering methods never use it ... as per today!
But we are planning to change this and use this extra information for an easier and more focused access to higher level information.
If a sensible structure tree is present, you can access all of that without rolling out heavy artillery like artificial intelligence or layout analysis.

So I cannot recommend to remove this item from now on without looking.