Add custom properties support + unit-tests, feature-tests

[michael-koeller]

Follow-up to MR #580, which I closed unintendedly.

Sorry for any inconveniences.

[michael-koeller]

Hi @scanny,

I did the rebase, as you suggested.

Is there anything more you need to be done, to have this long-running MR merged?

Best regards,
Michael

[Ricyteach]

Still looking forward to this since 2018! ;)

Thanks for all the hard work!!!!

[ssmoliarchuk]

Hi, first of all thanks for your work.
It's exactly what i'm looking for and it works perfect from your branch.

There is only one issue for me:
I modified custom properties, saved file, opened file. Custom properties have been changed. But changes are not applied in the document (i'm using inserting custom properties values inside doc).
It's a way to trigger it manually inside Word, but i'm looking how to make it automatically.
May be i need additionally trigger smthg or change some metadata? (I understand that it relates more to Word but may be you know how to do it).

[michael-koeller]

There is only one issue for me: I modified custom properties, saved file, opened file. Custom properties have been changed. But changes are not applied in the document (i'm using inserting custom properties values inside doc). It's a way to trigger it manually inside Word, but i'm looking how to make it automatically. May be i need additionally trigger smthg or change some metadata? (I understand that it relates more to Word but may be you know how to do it).

@ssmoliarchuk: Yes I know about this issue. My workaround is to select the area in the document and have Word refresh all dynmic fields. Interestingly, the issue does not occur in LibreOffice.

[chrisopenhab]

Hi, thanks for your work! Sad that this feature is still not merged :-(
Would be cool it the feature will be available in future releases.

[ryanamannion]

Hi all,

I have been using this change for a few months now and have enjoyed it, thank you for your work on this @michael-koeller.

My team and I have noticed, however, that when opening documents created with python-docx with this change, MS Word warns of "Unreadable content"

We have been able to trace this back to the namespace prefix used in this PR for custom-properties, namely cup. I haven't been able to find references to cup in any ooxml specs I have seen.

To recreate the above custom.xml you can run the following

from docx import Document
d = Document()
d.add_paragraph("Test text.)
d.custom_properties["test_prop"] = "foo"
d.save("/path/to/document.docx")

Microsoft's DocumentFormat.OpenXml documentation says the prefix is op, and this article on adding custom properties via the SDK also show the prefix as op, yet when I changed the prefix from cup -> op locally I still got the unreadable content warning with Word.

It looks like creating a new custom.xml in Word by adding a custom property to a blank document yields a Properties tag, no namespace prefix.

To recreate the above document in MS Word:

1. Open blank word document.
2. Optionally add some paragraph text to document.
3. File > Info > Click the word "Properties" top right to reveal "Advanced Properties" > Custom > Add a custom property

To test a crude potential fix, I changed the definition of the _customProperties_tmpl attribute of CT_CustomProperties in oxml/customprops.py to be

xmlns = 'xmlns="http://schemas.openxmlformats.org/officeDocument/2006/custom-properties"'
_customProperties_tmpl = "<Properties %s/>\n" % (xmlns + " " + nsdecls("vt"))

and changed definitions of cup to be op instead, namely in docx/oxml/ns.py and in docx/oxml/__init__.py for register_element_cls. Doing this seems to keep custom property functionality, and work with all editors I have tested with (LibreOffice, OnlyOffice, MS Word).

Any thoughts? I'd love to contribute by cleaning up my changes and opening a PR into your branch @michael-koeller

[michael-koeller]

@ryanamannion: Thanks for the detailed input. I'll soon have a look into it.

[ryanamannion]

I have been looking into this more to make a proper fix for BlackBoiler's fork of python-docx. Using dotnet's Open XML Productivity Tool, validation revealed that the actual issue was the missing namespace tag on property children of <op:Properties>. Adding the namespace prefix for property e.g. <op:property> in CustomProperties.__setitem__ seems to have fixed the issue, and is compliant with the spec using the correct namespace tags.

python-docx now generates a custom.xml that looks like the following (using the same code snippet from above):

The resulting document does not raise "Unreadable Content" warnings when opening with MS Word, and is validated successfully with the Open XML Productivity Tool

Code changes to achieve this can be viewed at a glace here: BlackBoiler#25