Python multiprocessing: set() in dict() doesn't work - python-3.x

Question

When there is a set() in dict(), the .add() function doesn't work.
manager = multiprocessing.Manager()
shared_dict = manager.dict()
def worker1(d, key):
if key not in shared_dict:
d[key] = {'0': set(), '1': set()}
def worker2(d, key):
if key not in shared_dict:
d[key] = {'0': set(), '1': set()}
d[key]['0'].add(1)
d[key]['1'].add(2)
process1 = multiprocessing.Process(
target=worker1, args=[shared_dict, 'a'])
process2 = multiprocessing.Process(
target=worker2, args=[shared_dict, 'b'])
process1.start()
process2.start()
process1.join()
process2.join()
I expected the following output:
{'a': {'1': set([]), '0': set([])}, 'b': {'1': (2), '0': (1)}}
instead of:
{'a': {'1': set([]), '0': set([])}, 'b': {'1': set([]), '0': set([])}}

Answer 1

You can read about your problem in Python documentation, which says:
If standard (non-proxy) list or dict objects are contained in a
referent, modifications to those mutable values will not be propagated
through the manager because the proxy has no way of knowing when the
values contained within are modified.
So, under "normal" circumstances if you create a new reference to an object and modify it, the modification is applied to the object no matter which reference you use to modify it:
a = set([1])
b = a
b.add(2)
print(a, b) # {1, 2} {1, 2}
In the manager, however, the modifications are not applied to the object for the quoted reason. Nevertheless, you can create a new reference to the object, change the value form there and then reassign the modified version to the dict.
import multiprocessing
manager = multiprocessing.Manager()
shared_dict = manager.dict()
def worker1(d, key):
d.setdefault(key, {'0': set(), '1': set()})
def worker2(d, key):
d.setdefault(key, {'0': set(), '1': set()})
buffer = d[key]
for i, (k, v) in enumerate(buffer.items()):
buffer[k].add(i)
d[key] = buffer
process1 = multiprocessing.Process(
target=worker1, args=[shared_dict, 'a'])
process2 = multiprocessing.Process(
target=worker2, args=[shared_dict, 'b'])
process1.start()
process2.start()
process1.join()
process2.join()
Btw, use dict.setdefault refactor these if statements.