Unreleased Release Notes

3.4.0-3

New Features

• Adds DelayedExecutorMixin and GreenDelayedExecutorMixin mixin classes that extend any Executor subclass with a submit_after(delay, fn, *args, **kwargs) method.

  A single background scheduler thread maintains a min-heap of pending tasks ordered by their deadline. When a task’s deadline arrives the scheduler submits it to the underlying executor. Tasks can be cancelled while waiting for their deadline via the returned Future.

  Use DelayedExecutorMixin with native-thread executors (e.g. ThreadPoolExecutor) and GreenDelayedExecutorMixin with eventlet-based executors (e.g. GreenThreadPoolExecutor). Each submit_after call accepts its own delay value, so different tasks submitted to the same executor can have different delays.

  Example:

  class DelayedThreadExecutor(
      futurist.DelayedExecutorMixin,
      futurist.ThreadPoolExecutor,
  ):
      pass
  
  executor = DelayedThreadExecutor(max_workers=4)
  fut = executor.submit_after(0.5, my_func, arg1, kwarg=val)
  result = fut.result()
  executor.shutdown()
  

3.4.0

New Features

• The ProcessPoolExecutor class now accepts the mp_context. Refer to docs for concurrent.futures.ProcessPoolExector for more information.

Upgrade Notes

• Support for Python 3.10 has been dropped. The minimum version of Python now supported is 3.11.

3.3.0

Upgrade Notes

• Support for Python 3.9 has been removed. Now the minimum python version supported is 3.10.

3.2.1

Bug Fixes

• shutdown(wait=True) abandoned queued work. This fix ensures it blocks until all submitted tasks (running and queued) finishes, consistent with Python’s concurrent.futures.ThreadPoolExecutor

3.2.0

New Features

• Adds a new DynamicThreadPoolExecutor that can resize (grow or shrink) its pool based on demand. As new work is scheduled on the executor, it will try to keep the proportion of busy threads within the provided range (between 40% and 80% by default).

  The motivation is to provide a scalable alternative to GreenThreadPoolExecutor.

3.1.0

Upgrade Notes

• Support for Python 3.8 has been removed. Now the minimum python version supported is 3.9 .

Deprecation Notes

• Eventlet usages are deprecated and the removal of Eventlet from OpenStack is planned, for this reason the executors based on Eventlet are deprecated. Start migrating your stack to the threading executor. Please also start considering removing your internal Eventlet usages.