convert more exceptions

[bertsky]

To be used in conjunction with tesseract-ocr/tesseract#4420 – which hopefully will get merged in some form, ultimately.

There may be more functions that need exception conversion.

Example result:

15:36:15.156 ERROR ocrd.processor.base - Failure on page PHYS_0067: ELIST_ITERATOR::forward:Error:List would have returned a nullptr data pointer
Traceback (most recent call last):
  File "/data/ocr-d/ocrd_all/core/src/ocrd/processor/base.py", line 710, in process_workspace_handle_page_task
    task.result()
  File "/data/ocr-d/ocrd_all/core/src/ocrd/processor/base.py", line 124, in result
    return self.fn(*self.args, **self.kwargs)
  File "/data/ocr-d/ocrd_all/core/src/ocrd/processor/base.py", line 1157, in _page_worker
    _page_worker_processor.process_page_file(*input_files)
  File "/data/ocr-d/ocrd_all/core/src/ocrd/processor/base.py", line 809, in process_page_file
    result = self.process_page_pcgts(*input_pcgts, page_id=page_id)
  File "/data/ocr-d/ocrd_all/ocrd_tesserocr/ocrd_tesserocr/recognize.py", line 512, in process_page_pcgts
    self._process_existing_regions(regions, page_image, page_coords, pcgts.mapping)
  File "/data/ocr-d/ocrd_all/ocrd_tesserocr/ocrd_tesserocr/recognize.py", line 995, in _process_existing_regions
    self._process_existing_lines(textlines, region_image, region_coords, mapping)
  File "/data/ocr-d/ocrd_all/ocrd_tesserocr/ocrd_tesserocr/recognize.py", line 1059, in _process_existing_lines
    self._process_existing_words(words, line_image, line_coords, mapping)
  File "/data/ocr-d/ocrd_all/ocrd_tesserocr/ocrd_tesserocr/recognize.py", line 1095, in _process_existing_words
    word_conf = self.tessapi.AllWordConfidences()
  File "tesserocr.pyx", line 2386, in tesserocr.PyTessBaseAPI.AllWordConfidences
RuntimeError: ELIST_ITERATOR::forward:Error:List would have returned a nullptr data pointer

That is, there was a failed assertion in libtesseract, which now throws a C++ exception instead of hard abort(), so in Cython we could convert it to Python exception, which we can then catch and act on.

[sirfz]

This won't break older versions right?

[bertsky]

This won't break older versions right?

I cannot imagine how. Adding except + will just convert C++ exceptions (instead of having them be handled by the C++ runtime). If libtesseract does not throw them, nothing should change. (And the difference of sacrificing the const specifier should also be negligable.)

Sry for not creating a minimal PR in the first place, btw. Do you want me to rebase to current master?

[sirfz]

Yes please rebase

[sirfz]

Hello @bertsky, is this good to move forward? You implied in your original post that this change is depending on a tesseract issue (or is it not a hard dependency?)

[bertsky]

Hi @sirfz, yes, I can see no reason not to include this in tesserocr. (Like I said, it's a soft dependency – it only acts if libtesseract throws C++ exceptions, which so far it never did.) However, it will not be much use until the PR in Tesseract gets merged. Unfortunately, the discussion there seems to be stalled.