Helpers for using Django from threads
When handling requests django manages database connection lifecycles for
you. By default closing them after each request or keeping
them alive allowing re-use for a specified amount of time when CONN_MAX_AGE
is set so long as no errors are encountered.
Sometimes you want to do work outside of a web request. When the work is large and you would like to distribute it, likely to places other than where the web requests are
served, there are systems like Celery. There are other cases though were you're just trying
to do work outside of the request & response cycle even and it's lightweight enough that running a full-blown job queuing system and it's associated data store is too
involved. You just want Threads or even better a
ThreadPoolExecutor.
If you just jump straight in and use either of those you'll run into slow to manifest and tough to track down problems with broken database connections. This is because
Django automatically opens a connection per-thread when a database is accessed and leaves them lying around in that thread indefinitely. This results in errors when the
database has timed out the connection or it has otherwise encountered problems.
django_thread
provides a this problem by implementing a Thread
class that mimics Django's request connection handling and provides a ThreadPoolExecutor
that does so
around the invocations of submitted calls.
Usage
Thread
django_thread.Thread
is a 100% drop-in replacement for threading.Thread
. See threading.Thread for
usage and documentation.
from django_thread import Thread
class ExampleThread(Thread):
def run(self):
for some_model in SomeModel.objects.filter(...):
...
thread = ExampleThread()
thread.start()
thread.join()
ThreadPoolExecutor
django_thread.ThreadPoolExecutor
is a 100% drop-in replacement for concurrent.futures.ThreadPoolExecutor
. See
concurrent.futures.ThreadPoolExecutor for usage and documentation.
from concurrent.futures import as_completed
from django_thread import ThreadPoolExecutor
def update_or_create_thing(name):
thing, _ = Thing.objects.update_or_create(name=name)
return thing.id
executor = ThreadPoolExecutor()
futures = []
for i in range(5):
future = executor.submit(update_or_create_thing, f'Name i')
futures.append(future)
ids = [f.result() for f in futures]