v1.0.0: ImportError: cannot import name 'parse_xml' from 'docx.oxml'

[matthuisman]

between v0.8.11 and v1.0.0.0 this seems to have been moved to: from docx.oxml.parser import parse_xml

I cant recall seeing any warning about this and cant see any migration notes for the 1.0.0 release.
Maybe this was a mistake?

Would it be possible to add the above to docx.oxml.init to allow backwards compatibility with existing imports?

[scanny]

Hmm, interesting, thanks for taking a look at this and reporting :)

Let me have a look here ...

Ah, right. So this actually moved to docx.oxml.parser.parse_xml(). There were some import cycles that led me to extract the parser into its own module. OxmlElement() also moved to that module.

Those are not part of the API per se, so we reserve the right to move those around as needed for the implementation.

What is the problem that it causes?

btw, docx.opc.oxml.parse_xml() is strictly for parsing packaging items like relationships (.rels) package parts, so wouldn't work for things like document elements, paragraphs, runs, etc.

[matthuisman]

i guess the problem is we have a lot of legacy code using the old import
And if we change to new import - it breaks any clients using an older version.

And yes, your right. The new equivalent is
from docx.oxml.parser import parse_xml

[hf-kklein]

same goes for OxmlElement

ImportError: cannot import name 'OxmlElement' from 'docx.oxml'

It's now available as:

from docx.oxml.parser import OxmlElement

[scanny]

k, I think I can "re-export" those from docx.oxml(.__init__) without running into any cycles, let me have a look ...

[patou01]

Hi, this broke a number of packages indeed as far as I can tell. Broke my CI this morning, as a dependency of https://github.com/elapouya/python-docx-template/blob/master/requirements.txt and https://github.com/4teamwork/docxcompose/blob/master/setup.py, which installed 1.0.0 instead of previously 0.8.11 and broke after that.

src\base_items\test_reports\control\test_report.py:8: in <module>
    from docxtpl import DocxTemplate
C:\hostedtoolcache\windows\Python\3.9.13\x86\lib\site-packages\docxtpl\__init__.py:13: in <module>
    from .subdoc import Subdoc
C:\hostedtoolcache\windows\Python\3.9.13\x86\lib\site-packages\docxtpl\subdoc.py:11: in <module>
    from docxcompose.properties import CustomProperties
C:\hostedtoolcache\windows\Python\3.9.13\x86\lib\site-packages\docxcompose\properties.py:9: in <module>
    from docx.oxml import parse_xml
E   ImportError: cannot import name 'parse_xml' from 'docx.oxml' (C:\hostedtoolcache\windows\Python\3.9.13\x86\lib\site-packages\docx\oxml\__init__.py)

Because of the way their requirements are defined, I guess both of these packages are broken by the release. It's unclear to me how or if these are maintained though, and I don't think it's your job to check on each of your users.

Just putting this here as info in case it could help.

From my side, I'm probably just going to pin to the older release because of my environment (32 bits), which will eventually not be supported anymore and will just bring me back here to look for somebody who had the same issue :) .

(Thanks for the work!)

[scanny]

Okay, I can live with these two being "de facto published API". It doesn't appear to cause me any real problems to republish them in their original location.

I'll include these changes in v1.0.1. That won't be long, within two weeks I expect, just want to see what else might need to go in there.

In the meantime, pinning to python-docx<1.0.0 should do the trick. Btw the same will be necessary for anyone on a "legacy" version of Python, meaning 3.6 or before.

[dsanmarting]

is the version, install:

pip install python-docx==0.8.11

that worked for me

[scanny]

Okay, v1.0.1 is live on PyPI and should remedy this. Please confirm that fixes this for you, meantime I'll close this issue :)

[matthuisman]

i cant see any code changes on this repo relating to 1.0.1?

Also, should these be considered deprecated? Is it advised to change imports?

[scanny]

The last four commits on master are these, or you can look at tag "v1.0.1".

I looked into adding a deprecation warning, but that looked like more trouble that it's worth. In my own code I would import them from where they're defined rather than somewhere they're republished, but I don't see an immediate reason to remove them from docx/oxml/__init__.py. I'll leave it up to you.