revisions/track changes

[igptrs]

Request for implementation of revision/track changes. This should allow to generate documents from other tools, such as databases highlighting the differences into the resulting MSWord document.

[kaidoloor]

Ability to read original and amended text would be great, to start with. Maybe I am able to contribute.

[jfthuong]

@kaidoloor

To get text with accepted changes, we could add something like that in class Paragraph of paragraph.py:

from xml.etree.ElementTree import XML

...
class Paragraph(Parented):
...
    @property
    def accepted_text(self):
        """
        String formed by accepting all changes and concatenating
        the text of each run in the paragraph.
        Tabs and line breaks in the XML are mapped to ``\\t`` and ``\\n``
        characters respectively.
        """
        W_NAMESPACE = "{http://schemas.openxmlformats.org/wordprocessingml/2006/main}"
        TEXT = W_NAMESPACE + "t"
        xml = self._p.xml
        if "w:del" in xml or "w:ins" in xml:
            tree = XML(xml)
            runs = (node.text for node in tree.iter(TEXT) if node.text)
            return "".join(runs)
        else:
            return self.text

We would use as follows:

for p in Document(path).paragraphs:
    print(p.accepted_text)

Note: Tested and working on Python 3.7 and Python 2.7.

I actually added a better and simpler version in a PR #734

[sterbon]

@igptrs did you find any solution for this?

[alanlivio]

I needed to do some similar task and created the following python code using python-openxml and xml.etree:
   https://github.com/alanlivio/docx_reviews_to_txt
Hope it can be used as inspiration.

[caramdache]

I had to handle additional cases that were not catered for by @jfthuong or @semanticz0, so I came up with the following solution. It handles inserted, moved and deleted text, as well as figure text annotations (AlternateContent). It also reuses the already parsed XML structure rather than reparsing from string.

from docx.oxml.ns import qn, nsmap

def accepted_text(p):
    def _accepted_text(p):
        text = ''
        for run in p.xpath('w:r[not(w:pPr/w:rPr/w:moveFrom)] | w:ins/w:r'):
            for child in run:
                if child.tag == qn('w:t'):
                    text += child.text or ''
                elif child.tag == qn('w:tab'):
                    text += '\t'
                elif child.tag in (qn('w:br'), qn('w:cr')):
                    text += '\n'
                elif child.tag == qn('mc:AlternateContent'):
                    for nested_p in child.xpath('mc:Choice[1]//w:p', namespaces=nsmap):
                        text += _accepted_text(nested_p)
                        text += '\n'
        return text

    nsmap['mc'] = 'http://schemas.openxmlformats.org/markup-compatibility/2006'
    return _accepted_text(p._p)

[MoritzImendoerffer]

Hi, this issue is rather old. I found this issue while browsing the issues of the unstructured library: issue 1821

Is there a specific reason why this feature is not yet implemented?

[jfthuong]

@caramdache Great addition. Maybe we should work together to present a strong PR with associated tests :)

[amithkk]

A few years late but I've taken a shot at this with #1534

[caramdache]

Excellent! I hope it gets merged soon.