Words in `DocumentLine` store polygons as lists of floats instead of lists of `Point`s

[dylwil3]

• Package Name: azure.ai.formrecognizer
• Package Version: 3.3.3
• Operating System: MacOS
• Python Version: 3.12

Describe the bug

The words in a DocumentLine seem to store their polygons as a list of floats rather than as a list of Point objects.

from typing import reveal_type

line: DocumentLine
for w in line.get_words():
    for p in w.polygon:
       reveal_type(p) # this reveals as `Point` but it turns out to be a `float` at runtime

This is in contrast to iterating over words from the words field in a DocumentPage, where the polygons are indeed sequences of Point objects.

To Reproduce
Steps to reproduce the behavior:

1. Use a prebuilt-read model on pdf bytes to run OCR
2. Examine the words in a document line in the result

Tangentially related to #39031

[xiangyan99]

Thanks for the feedback, we’ll investigate asap.

[github-actions]

Thanks for the feedback! We are routing this to the appropriate team for follow-up. cc @bojunehsu @vkurpad.