Manage background tasks in BaseService

What is wrong?

BaseService handles background tasks awkwardly right now. For example:

• have to await asyncio.sleep(0) in cleanup to make sure that all tasks cancel
• often need to catch and drop OperationCancelled exceptions

This is what kicked off the thought: https://github.com/ethereum/py-evm/pull/1097/files#diff-3942908a5c806895979f03283db24df3R206

An toy example with these quirks:

class MyService(BaseService):
  async def check_connection(self):
    while self.is_running:
      if not is_connected():
        self.wait(reconnect())
      try:
        await self.sleep(60.0)
      except OperationCancelled:
        break

  async def _run(self):
    asyncio.ensure_future(self.check_connection())

  async def _cleanup(self):
    # We don't need to cancel() anything, but we yield control just so that the coroutines we
    # run in the background notice the cancel token has been triggered and return.
    # Also, don't use self.sleep() here as the cancel token will be triggered and that will
    # raise OperationCancelled.
    await asyncio.sleep(0)

How can it be fixed

Add a BaseService.loop_in_background(my_task) method that handles these things for you. Then services can define something like:

class MyService(BaseService):
  async def check_connection(self):
    if not is_connected():
      self.wait(reconnect())
    await self.sleep(60.0)

  async def _run(self):
    self.loop_in_background(self.check_connection())

[cburgdorf]

so I think it’d make sense to have this new helper simply run the coroutine with ensure_future(), and then if necessary we could have a run_periodict_task() on top of that

Second that. How about:

self.run_task(self._sync())
self.run_periodic_task(self.check_connection(), restart_sec=60.0)

[pipermerriam]

Ah, how about run_periodic_task and you can specify the period.

while True:
    ... # run the task
    self.sleep(period)