Run OCR on image after saving (synchronously!) (#20)

[danieloeh]

This is a first implementation for running OCR whenever an image is newly created or changed. One major downside of this implementation is that saving an Item where the image has been changed now takes much longer because the OCR is running on django's main thread.
A better approach would probably be to use something like Celery to run OCR tasks in the background. The latter approach would also allow us to run periodic tasks. However this would add more dependencies and background services to the project.
Just let me know whether you would prefer the solution submitted in this PR or the one i have just described :)

[cod3monk]

Have you benchmarked how long it takes to upload and process 3 photos? If this is <1s, I wouldn't mind. Uploads from smartphone already take longer on most phones (with high resolution images).
I'm not opposed to having celery in the stack, as long as the development version can still run standalone (with synchronous or no OCR). I guess it would be more sound implementation.

[danieloeh]

I just did a quick test and it took a little less than 6 seconds to upload 3 smartphone photos. So, in the long run, i think that moving such tasks in the background would make sense.

[cod3monk]

I agree, background seems like the way to go.

[danieloeh]

I have changed the code so that celery is used if available. Celery requires a redis instance and a separate worker process. If the config variable CELERY_TASK_ALWAYS_EAGER is set to True in local_settings.py, then tasks will be executed in the foreground. If you want to run celery in the dev environment, use the command ./manage.py celery_worker run to start the worker and e.g. docker to run a redis instance.
I have added celery and redis to the prod environment as well, but wasn't able to test this fully because i had some issues with the web interface there. However, i have checked that the redis container and the celery worker are running in the docker compose cluster.

[cod3monk]

Cool! Will try this out (hopefully) soon ;)

BTW: on Monday there is an "Open Source Developers Meetup" at ZAM, where @unaimed, who is interested in this project, will attend. See here for details.

[danieloeh]

Thank you for the info! I am planning to be there as well this time, hopefully around 18:30.

[cod3monk]

Unfortunately, I won't be able to attend. 😢