partition_msg is unable to process attachments

[MthwRobinson]

To reproduce

from unstructured.partition.auto import partition
from unstructured.partition.msg import partition_msg
import traceback

filename = "example-docs/fake-email-multiple-attachments.msg"
try:
  elements = partition_msg(
    filename=filename, process_attachments=True, attachment_partitioner=partition
  )
except RuntimeError as e:
  print(e)
except:
    print("ERROR about attachments")
    traceback.print_exc()

output_filename = "msg_mul_attach.json"
elements_to_json(elements, filename=output_filename)

The error tha arise is:
ValueError: Invalid file /tmp/tmpo99fe8l4/Engineering Onboarding.pptx. The FileType.ZIP file type is not supported in partition.

[MthwRobinson]

OS: Windows 11 in my PC but I´m running the code in Databricks Notebooks or in Google Colab with the same result.
Python Version: Python 3.10.6
Unstructured Version: lasta (0.13.7) because I´ve applied %pip install -q --upgrade unstructured but I don´t know the comment to know directly the version.

[scanny]

@MthwRobinson This appears to be either a corrupted message or a defect or limitation in msg_parser in how it extracts attachments. I'd say next step is either to chalk this up to a fluke and wait for recurrence or to try another .msg file that is "known good" (shown to work with Outlook). In any case, this probably shouldn't be a file in example-docs/ unless it's to demonstrate how we handle this sort of failure and named accordingly.

---

Diagnostics

I'm able to reproduce this error.

I was able to detach the attachment using msg_parser, as partition_msg() does. However the attachment is not a valid PPTX file and cannot be opened with PowerPoint.

On attempt to open it with PowerPoint it signals a "repair error":

PowerPoint found a problem with content in Engineering Onboarding.pptx.
PowerPoint can attempt to repair the presentation.
If you trust the source of this presentation, click Repair.

When clicking "Repair" it states:

Sorry, PowerPoint can't read Engineering Onboarding.pptx.

On inspection, the attachment binary appears to be a zip archive (first two bytes of file are "PK"). However it cannot be unzipped and fails with this message:

$ unzip Engineering\ Onboarding.pptx
Archive:  Engineering Onboarding.pptx
  End-of-central-directory signature not found.  Either this file is not
  a zipfile, or it constitutes one disk of a multi-part archive.  In the
  latter case the central directory and zipfile comment will be found on
  the last disk(s) of this archive.
unzip:  cannot find zipfile directory in one of Engineering Onboarding.pptx or
        Engineering Onboarding.pptx.zip, and cannot find Engineering Onboarding.pptx.ZIP, period.

[MthwRobinson]

Sounds good, I'll update this issue to reflect removing that document from example-docs, and we can keep an eye out for other examples of corrupted files.

[scanny]

This turned out to be a defect in msg_parser. When replaced with python-oxmsg for parsing MSG files this and other attachments are extracted fine.