fix: accommodate NULL relationship (by skipping)

[LightningSpace]

when i using "docx.Document("a.docx")" want to open that document, had make 'keyError',what is happend?

the file i unzip, idont know what is that?Does this file make the keyError?

[LightningSpace]

ps: the word file was created or saved by MSWORD2016,however,rels always has relationship target is null.

[scanny]

We've seen this problem before, I think it comes from a Word Add-in like a "management-pack" or something that improperly leaves dangling relationships. There is some more information about it here: scanny/python-pptx#206 where it came up in the sister project python-pptx. It looks like if you delete that relationship from the .rels XML file it solves the problem. Otherwise you can patch the loading code.

[LightningSpace]

when i skipping ,another problem arises,the relationship have in the rels

[LightningSpace]

RootGcc.docx

that's my document , i try to find where the error position

[scanny]

Looks like you corrupted the main rels file somehow.

[LightningSpace]

I have solved this problem by deleting the NULL relationship each time, and then packaged it into docx. thank you very much

[whybeyoung]

The same problem i have . So how to avoid this in code

[deiferni]

Just to bump this issue a bit, we have also encountered the problem mentioned here in production. I have manually fixed the affected word-document but of course it would be more elegant if python-docx could handle such incorrect relationships gracefully 😉.

[zhangjx53]

It seems that python-docx can't handle such problem. Can anyone tell me how to avoid this in code ,or how to delete the null relationship in word 2016? thanks a lot~

[ngc7023]

I fix this problem by the following steps:

1. Modify the file extension from .docx to .zip
2. Unzip this file
3. Find the rel file and delete the with 'NULL'
4. Zip the file and modify the file extension to .docx

[scanny]

@ngc7023 how do you "delete the with 'NULL'"? Do you just do that by hand or do you have some code you use?

[ngc7023]

@scanny
I just do that by hand.
The step 3 should be modified as 'Open the rel file(.xml) and delete tag of relationship with 'NULL' by hand'.
I cannot promise the docx file will be totally repaired, because some codes are deleted. At least the keyError will not be raised.

[yingang]

bad_file.docx

Just encountered the same situation (as attached), just wondering if this kind of bad file would be handled gracefully by python-docx? Or prefer not to fix it?

[wonzer]

@yingang @scanny @deiferni @whybeyoung @zhangjx53 You can find a solution here->
#1105 (comment)