When running some scripts on Windows, a following RuntimeError is raised:

raise RuntimeError('''
RuntimeError:
    An attempt has been made to start a new process before the
    current process has finished its bootstrapping phase.
...

This can be fixed by introducing the main guard above the first top-level statement that is not a function or an import and indenting everything by one level after that. Here are the locations and potential solutions:

Chapter01/example1.py

if __name__ == '__main__':
    input = [i for i in range(10 ** 13, 10 ** 13 + 500)]


    # sequential
    ...

Chapter02/example1.py

if __name__ == '__main__':
    input = [i for i in range(10 ** 13, 10 ** 13 + 1000)]

    for n_workers in range(1, multiprocessing.cpu_count() + 1):
        ...

Chapter07/example1.py

if __name__ == '__main__':
    my_array = [i for i in range(20)]

    result = reduce_sum(my_array)
    print('Final result: %i.' % result)

Pycaches are built for specific Python version so it doesn't make much sense to commit them if the end-user will be using a different Python version.

pycache in chapter 3
pycache in chapter 16

In example 2 of chapter 1. When I run the file I get

Result is very large. Only printing the last 5 digits: 35443
Sequential took: 0.05 seconds.
Result is very large. Only printing the last 5 digits: 45807
Concurrent took: 0.02 seconds.

but 35433 is not equal to 45807. The results should be equal to 35443. What is the issue here ?

Hi There,

I tried out examples 05 and 06 from Chapter08 of the book. Both of them run error-free. However, I don't observer any speed-up in Example-06. In fact example06 is slower. Outputs from both examples are below. I am using a Spyder4.0.1 IDE with Python version 3.7.6 on a windows-10 machine with 12 cores. Would appreciate any pointers on how this issue can be resolved.

============Example05.py output====================
Took 1.2080 seconds with 1 process(es).
Took 1.1229 seconds with 2 process(es).
Took 0.9995 seconds with 3 process(es).
Took 1.0282 seconds with 4 process(es).
Took 1.0535 seconds with 5 process(es).
Took 1.0578 seconds with 6 process(es).
Done.

============Example06.py output====================
Took 1.4141 seconds with 1 process(es).
Took 1.1089 seconds with 2 process(es).
Took 1.0429 seconds with 3 process(es).
Took 1.0382 seconds with 4 process(es).
Took 1.0399 seconds with 5 process(es).
Took 1.0465 seconds with 6 process(es).
Done.

Problem

The code in the consumer class doesn't guarantee that although the queue has more than two elements, a single consumer can get all the two element. Because other consumers will compete for these elements too, which will often result in a situation that the two elements are got by two different consumers, therefore making the control variable result_size work in an unexpected manner.

Solution

Add a lock when get element from queue.

lock = multiprocessing.Lock()

class ReductionConsumer(multiprocessing.Process):
    def __init__(self, task_queue, result_queue):
         ...

    def run(self):
        pname = self.name
        print("Using process %s..." % pname)

        while True:
            with lock:
                num1 = self.task_queue.get()
                ...
                self.task_queue.task_done()
                self.result_queue.put(num1 + num2)

Package concurrencytest that is imported in Chapter19/example7.py can't be used on Windows, because the default interpreter is missing implementation of the os.fork function. When running the example under Windows the following error is generated:

AttributeError: module 'os' has no attribute 'fork'

Also the package is missing from software list in README.

When running script Chapter19/example3.py a NameError is raised:

scheduler.shutdown()
NameError: name 'scheduler' is not defined

The problem can be fixed by indenting the line 27 under the main guard.

Current version

if __name__ == '__main__':
    scheduler = BackgroundScheduler()
    ...
scheduler.shutdown()

Solution

if __name__ == '__main__':
    scheduler = BackgroundScheduler()
    ...
    scheduler.shutdown()

When running the file Chapter05/example6.py in Python 3.9 or newer, this error is raised:

AttributeError: 'MyThread' object has no attribute 'isAlive'. Did you mean: 'is_alive'?

This can be fixed by renaming the method at line 21.

alive = [1 if thread.isAlive() else 0 for thread in threads] # original
alive = [1 if thread.is_alive() else 0 for thread in threads] # solution

Also when running the script under Python 3.10 or newer, a warning is emitted:

DeprecationWarning: setDaemon() is deprecated, set the daemon attribute instead

The issue can be fixed by changing the line 41 like so:

thread.setDaemon(True) # original
thread.daemon = True # solution

The method threading.currentThread was also deprecated in favor of threading.current_thread. When running scripts Chapter12/example3.py and Chapter12/example4.py a warning will be emitted:

DeprecationWarning: currentThread() is deprecated, use current_thread() instead

Interestingly, in chapter 13, the correct method threading.current_thread is used.

When running the echo server example from chapter 10 under Python 3.10 or newer a TypeError is raised:

TypeError: BaseEventLoop.create_server() got an unexpected keyword argument 'loop'

The solution is to simply remove the 'loop' argument from line 17.

coro = asyncio.start_server(handle_echo, '127.0.0.1', 8888, loop=loop) # original
coro = asyncio.start_server(handle_echo, '127.0.0.1', 8888) # solution

This can be done safely because the function asyncio.get_event_loop that is used internally by asyncio.start_server always returns the currently running loop from Python 3.6 onwards.