docx created with word online

[burbma]

https://github.com/ankushshah89/python-docx2txt/blob/f71c423c3562b7c3f5bfcec822e384273d0034f2/docx2txt/docx2txt.py#L87

If I create a docx in SharePoint it takes me to Word Online. I add some text and it saves automatically. Then I download the file.

Now I do the following:

import zipfile
zip = zipfile.ZipFile('path/to/file.docx')
xml = zip.read('word/document.xml')

This fails with KeyError: "There is no item named 'word/document.xml' in the archive"

There is, however, a 'word/document2.xml' which contains (at least for my one trial case) the same as 'word/document.xml'. I discovered this by opening 'path/to/file.docx' in actual Microsoft Word on my local machine and then saving the file. NOW when I do zip.read('word/document.xml') the xml file is there as expected.

I really don't know much about this stuff or why creating a file with Word Online appears to create something different then local Word. Thus I don't know what the best solution is. It seems hack-ish to just put a line in the code that says if you can't find 'word/document.xml' look for 'word/document2.xml' but maybe that's all we need. Let me know.

[SamMorrowDrums]

@mmb90 the only alternative I can see is to look for an item in the list of documents in the zipfile that matches word/\d?document.xml but I think that frankly what I've done is enough. Stupid inconsistent format. There's probably some spec defined way too, but I really don't care enough to put more time in.

[fjouault]

Actually, it seems that _rels/.rels (an XML file zipped in the docx) contains the name of the document file as the Target attribute of the Relationship element with Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/officeDocument".

For instance: <Relationship Id="rId1" Type="http://schemas.openxmlformats.org/officeDocument/2006/relationships/officeDocument" Target="/word/document2.xml"/>

[SamMorrowDrums]

Yep, @fjouault I've updated my PR to reflect this. As far as I can tell this solution is robust. I'd appreciate any help with manually testing it, but hopefully this is ready to merge (or at least very close).

[ShayHill]

For what it's worth, I've come across a document where the rels file is wrong.

_rels/.rels correctly identifies officeDocument as "word/document.xml"

HOWEVER

word/_rels/document.xml.rels identifies header and footer as "word/header.xml" and "word/footer.xml" (not "header.xml" and "footer.xml" as they should be). I don't know the history of this file, and it's only 1 of 6000 I'm testing with, but the potential appears to be there. I would like to post the file, but removing proprietary information then re-saving corrects the problem.

[wendywangwwt]

I'm getting this issue and a temporary fix I created is as follows:

path = '/opt/conda/envs/Python-3.6-WMLCE/lib/python3.6/site-packages/docx2txt/docx2txt.py'
with open(path,'r') as f:
    script = f.readlines()

with open(path,'w') as f:
    for i,line in enumerate(script):
        if i == 86:
            line = "    doc_xml = [re.findall('(word\/document.*)',fn)[0] for fn in filelist if len(re.findall('(word\/document.*)',fn)) > 0][0]\n"
        f.write(line)

Basically I replace the hard coded xml (line 87) with a regular expression search as highlighted above. We are doing it in this way because the environment is containerized so every time we need to reinstall the package and change this line. For those who run this in their own long-lasting environment, simply replace line 87 with the following:

doc_xml = [re.findall('(word\/document.*)',fn)[0] for fn in filelist if len(re.findall('(word\/document.*)',fn)) > 0][0]\n

It's definitely not perfect and can be improved.. but anyway it solves my problem :)