Unbounded Thread Pool

Implementation of python's concurrent.futures.Executor that creates new threads on demand and stops them if they are not needed anymore.

It is designed to allow infinite recursive submitting of tasks, so the following code works properly, despite creating of 10k threads:

from unbounded_thread_pool import UnboundedThreadPoolExecutor

with UnboundedThreadPoolExecutor() as executor:
    def factorial(n: int):
        if n == 0 or n == 1:
            return 1
        else:
            return executor.submit(factorial, n - 1).result() * n
    print(factorial(10000))

Installation

pip install unbounded_thread_pool

Requirements

• Python (3.6, 3.7, 3.8)

Usage

UnboundedThreadPoolExecutor supports the following constructor parameters:

• name: str: Name of executor, all thread names are prefixed with it. Default is UnboundedThreadPoolExecutor.
• max_thread_idle_time: float: How many seconds idling worker thread should live. Default is 30 seconds.

For more details about methods check official python docs about concurrent.futures.Executor.

Usage with asyncio

When you write a lot of sync code that calls async code and vice versa (for example, if you use asgiref sync_to_async and async_to_sync) default asyncio executor can stuck. It happens in case you have the following code flow:

async code ---> sync code (1) ---> async code ---> sync code(2)

Here the second (2) sync code requires free worker to be available in thread pool, but thread pool can be already exhausted by the first (1) sync code. This causes deadlock, as (1) waits for (2), while (2) waits for (1) to free thread pool. To eliminate this issue, you can use UnboundedThreadPoolExecutor:

loop = asyncio.get_running_loop()
loop.set_default_executor(UnboundedThreadPoolExecutor(name='AsyncioExecutor'))