Fixed the way right to left language (Hebrew) text is inserted to pdf

[smijo149]

Issue Description:

The pdf file generated using hocr-pdf has Hebrew text printed in the opposite direction.

Steps I followed:

1. I used Google cloud vision to get the OCR
2. Used gcv2hocr to generate hocr.
3. Used hocr-pdf --savefile output.pdf actual-file.jpg to generate pdf file.

The pdf file has Hebrew text inserted in it but in the reverse order.

Actual image:

This is how hocr file looks:

Text in pdf file: (I have set text visibility mode to 0 so that the inserted text is visible)

Solution:

Use python package for bidi algorithm ( Bi-Directional Algorithm ) to transform text before drawing it into the pdf file.

Solution is based on the suggestion: ocropus#163

Tests:

Tested both Hebrew text image and English text image and they give expected results.

For Issue:

https://github.com/StarfishCo/api.raven.com/issues/3740

[RedEnchilada]

I think technically, the proper solution for Python would be something like from bidi import algorithm and then it would just be referenced as algorithm, but I don't know that for sure and if this works then it's good to me 👍

[smijo149]

@RedEnchilada That's a good suggestion. So basically took this as reference, https://github.com/tesseract-ocr/tesstrain/blob/master/generate_wordstr_box.py but then in the actual package documentation, https://pypi.org/project/python-bidi/ I found this in their example, from bidi.algorithm import get_display so I made the change accordingly.

[elgranchuy]

LGTM

[elgranchuy]

Im surprised by the I am afraid that hocr-pdf was never tested with RTL text.