manipulating Microsoft Word DOCX files that have links and track changes using Python

Question

I have been using the excellent python-docx package to read, modify, and write Microsoft Word files. The package supports extracting the text from each paragraph. It also allows accessing a paragraph a "run" at a time, where the run is a set of characters that have the same font information. Unfortunately, when you access a paragraph by runs, you lose the links, because the package does not support links. The package also does not support accessing change tracking information.

My problem is that I need to access change tracking information. Or, more specifically, I need to copy paragraphs that have change tracking indicated from one document to another.

I've tried doing this at the XML level. For example, this code snippet appends the contents of file1.docx to file2.docx:

from docx import Document
doc1 = Document("file1.docx")
doc2 = Document("file2.docx")
doc2.element.body.append(doc1.element.body)
doc2.save("file2-appended.docx")

When I try to open the file on my Mac for complicated files, I get this error:

But if I click OK, the contents are there. The manipulation also works without problem for very simple files.

What am I missing?

Answer 1

The .element attribute is really an "internal" interface and should be named ._element. In most other places I have named it that. What you're getting there is the root element of the document part. You can see what it is by calling:

print(doc2.element.xml)

That element has one and only one w:body element below it, which is what you get when with doc2.element.body (.xml will work on that too, btw, if you want to inspect that element).

What your code is doing is appending one body element at the end of another w:body element and thereby forming invalid XML. The WordprocessingML vocabulary is quite strict about what element can follow another and how many and so forth. The only surprise for me is that it actually sometimes works for you, I take it :)

If you want to manipulate the XML directly, which is what the ._element attribute is there for, you need to do it carefully, in view of the (complex) WordprocessingML XML Schema.

Unlike when you stick to the published API, there's no safety net once ._element (or .element) appears in your code.

Inside the body XML can be relationships to external document parts, like images and hyperlinks. These will only be valid within the document in which they appear. This might explain why some files can be repaired.