Constructor throw "File is not a zip file" on file created using Word

[swarmttied]

Hi. I'm trying to parse docx using the BytesIO overload of the Document ctor. when I parse any docx created using python-docx, the codes snippet below works just fine. However, with those saved from Word, I get the "File not valid zip" error. I notice that the Word-created files are indeed larger than those created using python-docx, indicative of it not being a zip file. Is there a way to parse a non-zipped stream in this case? I'm on Python 3.11

[scanny]

I'm not seeing any doc.save() call. What's the traceback?

I don't know what the type of stream is, but it will need to be IO[bytes], meaning either an open file or an instance of io.BytesIO, not just bytes.

[swarmttied]

@scanny , The exception is thrown at the constructor so I still don't have the chance to save the file. Just curious, if I'm just parsing why would I call save()?

The other files created using python-docx work. Only those created in Microsoft Word fail. I got the stream using this function

and is based on Microsoft suggestion here

And lastly, the trace

I hope these make things clear. Thanks for looking into this

[scanny]

It looks like the problem is getting from Azure to BytesIO. I expect if you just save the BytesIO to a file then Word will not open it and probably not zip either. Maybe the .readinto() mechanism defaults to text (str) rather than bytes, I'm not familiar. But the problem occurs before you get to python-docx.

[swarmttied]

They do open after saving to Azure. I'm ruling out ByteIO as the culrpirt because docx created using python-docx just works.
I suspect it's the effect of saving thru MS Word. I'll try to experiment with other versions of the lib and python for the time-being.

Thank you for your time @scanny.

[swarmttied]

My colleague figured this out. It was the cp037 encoding used by Word.