multithreading - Communication between threads in Python (without using Global Variables) -

let's if have main thread launches 2 threads test modules - " test_a" , " test_b". both test module threads maintain state whether done performing test or if encountered error, warning or if want update other information.

how main thread can access information , act accordingly. example, if " test_a" raised error flag; how "main" know , stop rest of tests before existing error ?

one way using global variables gets ugly.. soon.

the obvious solution share kind of mutable variable, passing in thread objects/functions @ constructor/start.

the clean way build class appropriate instance attributes. if you're using threading.thread subclass, instead of thread function, can use subclass place stick attributes. i'll show list because it's shorter:

def test_a_func(thread_state):     # ...     thread_state[0] = my_error_state     # ...  def main_thread():     test_states = [none]     test_a = threading.thread(target=test_a_func, args=(test_a_state,))     test_a.start() 

you can (and want to) pack lock or condition mutable state object, can synchronize between main_thread , test_a.

(another option use queue.queue, os.pipe, etc. pass information around, still need queue or pipe child thread—which in exact same way above.)

---

however, it's worth considering whether need this. if think of test_a , test_b "jobs", rather "thread functions", can execute jobs on pool, , let pool handle passing results or errors back.

for example:

try:     concurrent.futures.threadpoolexecutor(workers=2) executor:         tests = [executor.submit(job) job in (test_a, test_b)]         test in concurrent.futures.as_completed(tests):             result = test.result() except exception e:     # stuff 

now, if test_a function raises exception, main thread exception—and, because means exiting with block, , of other jobs cancelled , thrown away, , worker threads shut down.

if you're using 2.5-3.1, don't have concurrent.futures built in, can install the backport off pypi, or can rewrite things around multiprocessing.dummy.pool. (it's slightly more complicated way, because have create sequence of jobs , call map_async iterator on asyncresult objects… that's still pretty simple.)