A map function that uses multiple processes to map, but does it more efficiently than the `multiprocessing` library by using the magic of Unix forking

mp_map.py

from typing import *
import multiprocessing as mp

def mp_map(fn, input_sequence: Iterable) -> Iterable:
    input_queue = mp.Queue()
    output_queue = mp.Queue()

    def process_items():
        while True:
            item = input_queue.get()
            if item is None:    # sentinel value
                break
            item_index, item = item

            try:
                processed_item = fn(item)
            except Exception as e:
                output_queue.put((None, e))
            else:
                output_queue.put((processed_item, None))

        output_queue.close()
        output_queue.join_thread()

    processes = []
    try:
        for cpu_number in range(mp.cpu_count()):
            process = mp.Process(target=process_items)
            processes.append(process)
            process.start()

        item_count = 0
        for i, item in enumerate(input_sequence):
            input_queue.put((i, item))
            item_count += 1

        for _ in range(mp.cpu_count()):
            input_queue.put(None)

        while item_count > 0:
            processed_item, error = output_queue.get()
            if error is not None:
                raise error
            yield processed_item
            item_count -= 1
    finally:
        for process in processes:
            process.terminate()