An observable-ish pattern with dask’s pub sub model.

https://docs.dask.org/en/latest/futures.html#actors

Use threads on the in-process client

</tr> </table> The `Observable` class is built to comply with the `distributed` client actor model. ```python class Observable: value = munch.Munch() def dlink(self, source, target=None, callable=None): source = id(source[0]), source[1] if target and callable is None: callable = toolz.partial(setattr, *target) if target: target = id(target[0]), target[1] self.value.update({source:{target:callable}}) def link(self, source, target=None, callable=None): self.dlink(source, target, callable), self.dlink(target, source, callable) def trigger(self, source): toolz.juxt(*self.value[id(source[0]), source[1]].values())(getattr(*source)) ``` Create our `pubsub` client. ```python pubsub = client.submit(Observable, actor=True).result() ``` Demonstrate the efficacy with `Thing`. ```python @dataclasses.dataclass(unsafe_hash=True, eq=True) class Thing: a: int = 10 ``` ```python t = Thing(a=10) s = Thing(20) assert s != t ``` Link the traits. ```python pubsub.link((t, 'a'), (s, 'a')) t.a = 30 assert s != t ``` `s` has not changed yet because we have no asked to update it yet. We should likely over the `__setattr__` method for usability. ```python pubsub.trigger((t, 'a')) __import__('time').sleep(.1) # it seems now there race conditions! assert s.a == t.a == 30 ``` go back the other way. ```python s.a = 40 pubsub.trigger((s, 'a')) __import__('time').sleep(.1) assert t.a == 40 ```

Client Scheduler: inproc://10.8.0.40/15634/1 Dashboard: http://localhost:8787/status </ul> </td>

Cluster Workers: 1 Cores: 8 Memory: 17.18 GB