Insert_pdf different result on same file

[yn2415]

I was testing the new annots and widgets options in insert_pdf, but got different result when operating on the same file twice. Following are the codes:

with open('test_annotation.pdf', 'rb') as f:
    docs = f.read()
doc_pdf = fitz.open(filetype='pdf', stream=docs)
print("Page number", doc_pdf.page_count)

stacked_file = fitz.open()
stacked_file_2 = fitz.open()

for page_num in range(doc_pdf.page_count):
    stacked_file.insert_pdf(doc_pdf, from_page=page_num, to_page=page_num, annots=True, widgets=True)
    stacked_file_2.insert_pdf(doc_pdf, from_page=page_num, to_page=page_num, annots=True, widgets=True)

cleaned_file = stacked_file.tobytes()
cleaned_file_2 = stacked_file_2.tobytes()

with open('document_1.pdf', 'wb') as f:
    f.write(cleaned_file)

with open('document_2.pdf', 'wb') as f:
    f.write(cleaned_file_2)

with open('original_1.pdf', 'wb') as f:
    f.write(doc_pdf.tobytes())

document_1 and original_1 is the same while document_2 is different.
Version I'm using is 1.25.3.
Here is a test file:
test_annotation.pdf

I have a work around that I can do fitz.open twice, and insert the second file against the newly opened file. But still would like to understand what is happening here.

[JorjMcKie]

You are right, support for widget-copy modifies the source file unintendedly. I will open a bug issue for this.

Apart from that, you shouldn't be doing things the way you did:

1. There is absolutely no need to open pymupdf.Documents from memory every single time.
2. Never do loops like this one:

   for page_num in range(doc_pdf.page_count):
       stacked_file.insert_pdf(doc_pdf, from_page=page_num, to_page=page_num, annots=True, widgets=True)

   This effectively does its best to achieve worst-possible performance! Always use the maximum available page range.

Here is a version that does things correctly.

import pymupdf

doc_pdf = pymupdf.open("test_annotation.pdf")
print("Page number", doc_pdf.page_count)

stacked_file = pymupdf.open()
stacked_file.insert_pdf(doc_pdf)
stacked_file.ez_save("document_1.pdf")

doc_pdf = pymupdf.open(doc_pdf.name)  # <== this should not be necessary
stacked_file_2 = pymupdf.open()
stacked_file_2.insert_pdf(doc_pdf)
stacked_file_2.ez_save("document_2.pdf")

Note: the version Document.ez_save() is simply .save() with a number of different defaults: garbage=3, deflate=True, use_objstms=True. This is the second-most radical garbage collection combined with object source compression.
If you want to achieve the same effect using method .tobytes() you must explicitly specify those parameters.

[yn2415]

Thank you!! This is very helpful!