bug/ImportError-unstructured.partition.pdf

[mtb-beta]

Describe the bug

There is a bug that causes an ImportError when trying to import unstructured.partition.pdf.

To Reproduce

The issue can be reproduced by running the following in a Python shell:

>>> import unstructured.partition.pdf
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/mypath/venv/lib/python3.9/site-packages/unstructured/partition/pdf.py", line 6, in <module>
    from pdfminer.utils import open_filename
ImportError: cannot import name 'open_filename' from 'pdfminer.utils' (/mypath/venv/lib/python3.9/site-packages/pdfminer/utils.py)

Expected behavior

At a minimum, it would be preferable to have no code that results in an ImportError.

Screenshots

None

Desktop (please complete the following information):

• OS: Mac
• Python version: 3.9.16

Additional context

This issue occurs in version 0.6.2 of the unstructured
It does not occur in version 0.6.1. Also, pdfminer has been in maintenance mode since 2020, and the latest version does not have open_filename in pdfminer.utils.

euske/pdfminer@7ed7d4e

It appears that this issue surfaced due to the import statement being defined in the module global scope in this commit:

894a190#diff-cefa2d296ae7ffcf5c28b5734d5c7d506fbdb225c05a0bc27c6b755d5424ffda

[MthwRobinson]

@mtb-beta - Thanks a lot for reporting this. We'll pick this up as soon as we can.

[MthwRobinson]

@mtb-beta - You may have installed pdfminer instead of pdfminer.six. pdfminer.six still has open_filename in utils.py. pip uninstall pdfminer && pip install pdfminer-six may solve this for you.

We can add pdfminer.six as an explicit requirement for unstructured to help mitigate this. Currently it inherits that requirement from unstructured-inference.