How does pymupdf set the value of rotation for one page?

[monchin]

Hello, I found that one of my pdf document, consists of some 90-rotation pages, cannot be recognized as rotation=90. Precisely, 2 of 9 pages (the 0th and the 6th, index from 0) and are regarded as rotation=90, while others are regarded as rotation=0, although from human being's view, all the pages are rotation=90. What is more strange is that if I use

new_doc = pymupdf.Document()
new_doc.insert_pdf(doc, from_page=1, to_page=1) # that page is regarded as rotation=0 when in the whole document
new_doc.save("xxx.pdf")

the page of the 1-page new document can be regarded as rotation=90 although it is rotation=0 in the whole document.

For confidentiality reasons I can't provide my pdf document, and I tried to make a rotated pdf myself, but that pdf is correctly recognized as all pages are rotation=90. So maybe it is hard to give the example. I'm a little confusing and curious about the phenomena above. Is rotation the raw information in the pdf or is it calculated by pymupdf? Why is the page in the whole document regarded as rotation=0 while the same page in the 1-page pdf regarded as rotation=90?

I'm using pymupdf 1.24.11 on windows 11, but I also tried the newest version 1.26.3 and it seems it behaves exactly the same on this issue.

[JorjMcKie]

Whether or not a page is rotated depends exclusively on the /Rotate parameter in the page's PDF object definition.
This value has nothing to do with how text and other content on that page is being shown! IAW The complete text may be rotated by 90° while the page still is declared /Rotate 0.

[monchin]

Thank you!