Details

It is often desirable to process computations asynchronously. Until there was lovely.remotetask this was not so easy to achieve in a Zope 3 application due to transaction integration issues.

gocept.async makes the task even easier:

>>> import gocept.async
>>> @gocept.async.function(service='events')
... def heavy_computing(a, b):
...     print "Computing", a, "+", b, "=", a + b

The decorator gocept.async.function takes exactly one argument, the name of a lovely.remotetask.interfaces.ITaskService utility. Note that gocept.async does not define any task service by itself.

Test Setup

Note that the decorated function must have an importable module to be usable:

>>> import gocept.async.tests
>>> heavy_computing.undecorated.__module__ = 'gocept.async.tests'
>>> gocept.async.tests.heavy_computing = heavy_computing

We defined task-service called events in this test:

>>> import zope.component
>>> import lovely.remotetask
>>> import lovely.remotetask.interfaces
>>> import lovely.remotetask.processor
>>> sm = zope.component.getSiteManager()
>>> getRootFolder()['tasks'] = tasks = lovely.remotetask.TaskService()
>>> tasks.processorFactory = lovely.remotetask.processor.MultiProcessor
>>> tasks.processorArguments = {'maxThreads': 1}
>>> sm.registerUtility(
...     tasks, lovely.remotetask.interfaces.ITaskService, name='events')

Basics

When the decorated function is called it returns nothing:

>>> heavy_computing(2, 7)

When we start the processing of the task service, the function is called:

>>> gocept.async.tests.process()
Computing 2 + 7 = 9

When the function is called while a user is logged in, the function will be called as that user (Note that it might be necessary to manually create security proxies to enable security in the async function.):

>>> @gocept.async.function('events')
... def who_am_i():
...     print gocept.async.task.TaskDescription.get_principal()
...
>>> who_am_i.undecorated.__module__ = 'gocept.async.tests'
>>> gocept.async.tests.who_am_i = who_am_i
>>> who_am_i()
>>> gocept.async.tests.process()

Now login:

>>> gocept.async.tests.login('zope.user')
>>> who_am_i()
>>> gocept.async.tests.process()
zope.user
>>> gocept.async.tests.logout()

If an async function is called while the process is already async'ed the function is called immediately:

>>> @gocept.async.function(service='events')
... def call_another():
...     print "Before"
...     heavy_computing(1, 2)
...     print "After"
>>> call_another.undecorated.__module__ = 'gocept.async.tests'
>>> gocept.async.tests.call_another = call_another
>>> call_another()
>>> gocept.async.tests.process()
Before
Computing 1 + 2 = 3
After

There is a helper to test if currenlty async is in progress:

>>> gocept.async.is_async()
False
>>> @gocept.async.function(service='events')
... def is_async_test():
...     print gocept.async.is_async()
>>> is_async_test.undecorated.__module__ = 'gocept.async.tests'
>>> gocept.async.tests.is_async_test = is_async_test
>>> is_async_test()
>>> gocept.async.tests.process()
True

Teardown

>>> sm.registerUtility(
...     tasks, lovely.remotetask.interfaces.ITaskService, name='events')
>>> del gocept.async.tests.heavy_computing
>>> del gocept.async.tests.who_am_i
>>> del gocept.async.tests.call_another
>>> del gocept.async.tests.is_async_test

Changes