About pdf_clip_page

[Syntamin]

---

MuPDF’s C API recently added pdf_clip_page, which appears to filter page contents (e.g., for selective rendering or cropping). However, it’s unclear whether this function performs physical page clipping (modifying the PDF’s content stream to permanently remove clipped regions) or merely a visual clip (temporary for rendering).
need a way to programmatically crop PDF pages permanently (similar to CropBox but applied to content). If pdf_clip_page does this, I’d like to request Python bindings in PyMuPDF.

---

Clarify pdf_clip_page’s behavior:

Does it alter the PDF’s content stream to discard data outside the clip region?

Or is it a rendering-time optimization (like setMediaBox/setCropBox)?
Python binding timeline:

If pdf_clip_page supports permanent clipping, when might PyMuPDF expose this function?

[JorjMcKie]

This function is a "hard clip", i.e. it will remove content outside the clip rectangle - roughly similar to what redactions do.
In contrast to redactions however, images are left untouched here. I.e. images will not be blanked out for the parts not overlapping the clip and neither be deleted if completely outside.
We are in the process to expose this function in PyMuPDF.

[JorjMcKie]

You can already today use this function because most of MuPDF is available by the mupdf submodule of PyMuPDF.
Here is an example script that removes page content except for the bottom right quarter page.

import pymupdf

mupdf = pymupdf.mupdf


def clip_page(page, clip):
    clip *= page.transformation_matrix  # need PDF coordinates
    ppage = pymupdf._as_pdf_page(page)  # need underlying PDF page
    pclip = pymupdf.JM_rect_from_py(clip)  # need a MuPDF rectangle
    mupdf.pdf_clip_page(ppage, pclip)


doc = pymupdf.open("v110-changes.pdf")
page = doc[0]
clip = page.rect
clip.x0 = page.rect.width / 2
clip.y0 = page.rect.height / 2
clip_page(page, clip)
page.draw_rect(clip, color=(1, 0, 0), overlay=True)
doc.ez_save("clipped.pdf")

Here are images of the original page and the resulting one. The red rectangle indicates the clip rectangle.
You will notice that clipping for vector graphics is limited to those graphics that are completely outside the clip. Similar is true for text, too.
Original:

Result: