First off, let's clarify a key limitation: python-docx doesn't have a native concept of "pages" because Word uses dynamic pagination (it calculates page breaks based on content, font size, margins, etc.). So we need to handle this in two scenarios, depending on how your document is paginated.

Scenario 1: Document uses manual page breaks #

If your document relies on explicit manual page breaks (inserted via Word's "Page Break" command), we can detect these breaks and insert our PAGEEND marker right before each one. We'll also add a marker at the very end for the final page (which won't have a trailing page break).

Here's the modified code built on your existing snippet:

from docx.api import Document
from docx.enum.text import WD_BREAK

inputfile = 'test.docx'
document = Document(inputfile)
page_number = 1

for idx, paragraph in enumerate(document.paragraphs):
    # Check each run in the paragraph for a manual page break
    for run in paragraph.runs:
        # Use XPath to detect page breaks in the underlying XML (python-docx doesn't expose this directly)
        if WD_BREAK.PAGE in run._element.xpath('.//w:br[@w:type="page"]'):
            # Insert the PAGEEND marker right before the page break paragraph
            page_end_paragraph = document.add_paragraph(f'PAGEEND_<<{page_number}>>')
            document._body.insert(idx, page_end_paragraph._element)
            page_number += 1
            break  # No need to check other runs in this paragraph

# Add marker for the last page (no trailing page break)
document.add_paragraph(f'PAGEEND_<<{page_number}>>')

# Save the modified document
document.save('test_with_pageends.docx')

How this works: #

• We loop through every paragraph and its individual text runs to spot manual page breaks using XPath (since python-docx doesn't have a built-in method for this).
• When a page break is found, we insert the PAGEEND marker immediately before that paragraph, increment the page counter, and move on.
• Finally, we append a marker to the end of the document for the last page.

Scenario 2: Document uses automatic pagination #

If your document lets Word handle pagination automatically (no manual breaks), python-docx can't help here—it can't render the document to calculate dynamic page boundaries. For this case, we can use pywin32 (Windows-only) to interact directly with Word's COM object, which has access to rendered page data.

First, install pywin32 if you haven't:

pip install pywin32

Then use this code:

import win32com.client as win32

inputfile = 'test.docx'
outputfile = 'test_with_pageends.docx'

# Launch Word in background mode
word = win32.gencache.EnsureDispatch('Word.Application')
word.Visible = False  # Set to True if you want to see Word working in real-time

doc = word.Documents.Open(inputfile)
total_pages = doc.ComputeStatistics(2)  # 2 = wdStatisticPages, gets total page count

for page_num in range(1, total_pages + 1):
    # Navigate to the end of the current page
    doc.GoTo(What=win32.constants.wdGoToPage, Which=win32.constants.wdGoToAbsolute, Count=page_num)
    doc.GoTo(What=win32.constants.wdGoToLine, Which=win32.constants.wdGoToLast)
    
    # Insert the PAGEEND marker
    doc.Range().InsertAfter(f'PAGEEND_<<{page_num}>>')
    # Add a line break after the marker for better formatting
    doc.Range().InsertAfter('\n')

# Save changes and clean up
doc.SaveAs(outputfile)
doc.Close()
word.Quit()

Notes for this method: #

• This only works on Windows, and requires Microsoft Word to be installed on your machine.
• It uses Word's own rendering engine, so it accurately detects automatic page breaks.
• Adjust the insertion logic (like adding a line break) if you want the marker to fit your document's formatting.

Quick Tips #

• If your document uses a mix of manual and automatic pagination, the COM method is more reliable—it handles both cases seamlessly.
• Always test with a copy of your document first to avoid accidental data loss!

内容的提问来源于stack exchange，提问作者Bonson