Current approach to restarting data collection?

[ShaheedHaque]

Have you asked elsewhere?

The thread in https://stackoverflow.com/a/40518553/6332554 describes a way to "restart" data collection.

Describe your situation

Celery workers do not honour atexit() handlers, see celery/celery#8923. I'm trying a couple of different approaches to dealing with this, one being to use the Coverage API to dump the collected data as the worker runs. Since the worker runs my tasks, I can "easily" add the code to the end of my task function.

I therefore need incremental collection. I don't see an obvious solution using the current public API, but I do see the auto_load option on the constructor. Can it or the load() method be used to restart collection?

If not, then is the method in the stackoverflow.com exchange still current?

[ShaheedHaque]

I was able to close #1454 by implementing on-going dumping and restarting data collection. My current approach, in lieu of using the internal APIs used by the stackoverflow thread, is something like this:

    ctx.coverage.stop()
    ctx.coverage.save()
    data_file = <some PID-based name with a monotonically increasing count as a suffix>
    ctx.coverage = _coverage.Coverage(data_file=data_file, data_suffix='cov', cover_pylib=False, config_file=whatever)
    ctx.coverage.start()

This assumes that the garbage collector will clean up after the old instance of ctx.coverage, and has the potential to create a lot of files. Hence the interest in whether auto_load or load() (or indeed some other API) could be useful. Ideally, something like:

    ctx.coverage.stop()
    ctx.coverage.save()
    ctx.coverage.resume()