merge_translated_page() on page where its cropbox does not start at x1=0, y1=0

[jassler]

For a pdf file, I'm trying to duplicate the first page, rotate it by 180º and merge it with the first page such that they appear below each other. I've come up with the following solution.

from pypdf import PdfWriter

def stitch_files(filename, outfile="output.pdf"):
    merger = PdfWriter()

    with open(filename, 'rb') as file:
        merger.append(file)
        merger.append(file)

    merger.pages[1].rotate(180)
    merger.pages[1].transfer_rotation_to_content()

    insertion_point = merger.pages[0].mediabox.height
    merger.pages[0].merge_translated_page(merger.pages[1], tx=0, ty=-insertion_point, expand=True)
    del merger.pages[1]

    with open(outfile, 'wb') as fp:
        merger.write(fp)

stitch_files('01.pdf', '01_stitched.pdf')
stitch_files('02.pdf', '02_stitched.pdf')
stitch_files('03.pdf', '03_stitched.pdf')

I tried it on the following three files:

• 01.pdf, an ordinary name tag;
• 02.pdf, the file I got after stitching 01.pdf;
• 03.pdf, the final version of the file I'd like to stitch.

Unfortunately, the script only seems to work for 01.pdf. In 02.pdf, it seems to translate the rotated page down by the old mediabox height. In 03.pdf, the file size is doubled, but nothing appears to happen visibly.

Digging around, I found this answer to a discussion thread on rotating pages: #903 (comment)

In the code example, Martin asserts that the x1 and y1 coordinates of the page's cropbox start at 0. I suppose that in my code, I'd have to check the following before inserting the translated page.

x1, y1, x2, y2 = merger.pages[0].cropbox
assert x1 == 0
assert y1 == 0

This assertion fails in 02.pdf and 03.pdf. I assume this is also the reason, why merger.pages[0].mediabox.height does not, in fact, correspond to the ty coordinate where I want to stitch the page to.

I've tried juggling around with variables and expressions to come up with the correct translation value, but to no avail. Does anyone know how to fix this? Also, what causes the cropbox of a page not to start at 0, 0?

[stefan6419846]

Before digging deeper here or doing a more extensive analysis, to answer the easy part:

Also, what causes the cropbox of a page not to start at 0, 0?

The crop box is the region to display or print (see section 14.11.2.1 of the PDF 2.0 specification). As all corresponding page boxes, it can basically be whatever rectangle you want it to be.

[jassler]

Ah, I see. I just started trying to translate the page such that the origin of the content starts at 0, 0 – I suppose, I'd then have to move the cropbox to 0, 0 then, too, right?

This was my try so far (without adjusting the cropbox).

page = merger.pages[0]
x1, y1, x2, y2 = page.cropbox
if x1 != 0 or y1 != 0:
    op = Transformation().translate(tx=-x1, ty=-y1)
    page.add_transformation(op, expand=False)

[stefan6419846]

Without having actually tested it myself, this looks right.

Otherwise, you would have to manually fix the resulting crop box as pypdf only adapts the media box on merging. My short testing has shown that this might require some further adaption as well.