Creating the scratch-files in `_get_grayscale` using a random integer is
for one inherently unsafe and can cause a collision. On the other hand,
it should be unnecessary given that the files will be cleaned up after
the OCR run.
Since we don't know if OCR runs might be parallel in the future, this
commit implements thread-safe and deterministic directory-creation.
Additionally it fixes the call to `_cleanup` by `consume`. In the
current implementation `_cleanup` will not be called if the last
consumed document failed with an `OCRError`, this commit fixes this.
To detect the language currently the entire document gets processed. If
a different language has been detected than the default one, the entire
document will be processed again for the new language.
This PR analyzes the middle page for its language and either processes
the remaining pages with the default language if it didn't differ, or
processes all pages for the new guessed language.
The amount of processed pages comes down from the worst case `2n` to
worst case `n+1`.
At the moment, every page in a PDF will be processed one by one using
tesseract. Since the processing of a single page is independent from every
other page, one can make use of multi-core machines.
This PR introduces a multiprocessing pool to process multiple pages
simultaneously. The amount of threads to use can be specified in the
environment variable `PAPERLESS_OCR_THREADS`. This will default to the
number of cores/hyperthreads Python detects for your system.
Brian Martin
Daniel Quinn
Daniel Quinn
Tikitu de Jager
Pit Kleyersburg