Table extraction - vertical text not handled correctly

[kyliemsauter]

Description of the bug

Vertical text in table columns isn't handled correctly anymore, it used to work before the recent switch to use the "word extractor" methods.

How to reproduce the bug

Call extract() on any table that has vertical text
Example file: the two cells that have vertical text "Text" and "Numbers" are returned in reverse order as "txeT" and "srebmuN"
table_examples_vertical_text.pdf

PyMuPDF version

1.23.21

Operating system

Windows

Python version

3.11

[JorjMcKie]

The fix includes support for all rotations by multiples of 90°.

You may be aware that line breaks in a cell are preserved. This makes some sense as a last resort to identify additional table rows, and in cases with complex cell content ... but only for rotation 0 (horizontal) text. In other cases (rotations by 90°, 180°, 270°) I am replacing line breaks by spaces.

Forgot to mention, that the rotation of all of the cell content is determined by its first character / word. IAW, there cannot be different rotations of content in one single cell. But the following perfectly works:

A slightly modified version of your nice example

will be extracted as

['', 'Column 1', 'Column 2', 'Column 3']
['Text', 'Text A', 'Text B', 'Text C']
[None, 'Text D', 'Text E', 'Text F']
[None, 'Text G', 'Text H', 'Text I']
['First Set of Numbers', '1', '1', '1']
[None, '2', '2', '2']
[None, '3', '3', '3']
[None, '4', '4', '4']
[None, '5', '5', '5']
['Second Set of Numbers', '6', '6', '6']
[None, '7', '7', '7']
[None, '8', '8', '8']
[None, '9', '9', '9']
[None, '10', '10', '10']

[kyliemsauter]

great thank you!!!

[julian-smith-artifex-com]

Fixed in 1.23.24.