David Sutherland

Ok I've added more features and taken some ideas from #568 .. Turns out you can subscribe/unsubscribe from topics without restarting the connection: https://github.com/cylc/cylc-uiserver/issues/568 If you print the `topic` near...

Not sure the reason for this `mypy` failure:  `new_data[field.name].CopyFrom(value)` has always been there... and passes locally Also, I get a lot of this: ``` ERROR cylc/uiserver/tests/test_workflows_mgr_data_store.py::test_data_store_changes[active-active-actions8] - AttributeError: 'CylcUIServer'...

> provided we don't need to change much of the cylc-flow code It can go in now, and not break anything.. So if we can at least do that for...

Have a solution for queries, hopefully post it up tomorrow..

Query sync level change and expire implemented, and functional:  The expiry is set to 60sec, with a periodic callback every 30sec to check and demote. (~30sec expiry in the...

> > provided we don't need to change much of the cylc-flow code > > It can go in now, and not break anything.. So if we can at least...

I think I may have found an easier way to hook into the query/(subscriptions) before execution .. Actually, at other points of the execution too: https://github.com/graphql-python/graphene-tornado/blob/master/graphene_tornado/graphql_extension.py So, probably a good...

I'm probably with Hilary, and as he hinted, this wouldn't be an issue if we didn't spawn partially satisfied tasks. If we only spawned parentless and prereq satisfied tasks the...

> If we only spawned parentless and prereq satisfied tasks Obviously, this would require a different strategy for knowing when a task can run.. Something like, given a task output...

> I'm in danger of confusing the issue by going off topic here I was merely reducing our spawning strategy down to it's idealized limit to illustrate that stalling doesn't...