:Version: 1.10.2
:Web: http://mode.readthedocs.org/
:Download: http://pypi.python.org/pypi/mode
:Source: http://github.com/fauststream/mode
:Keywords: async, service, framework, actors, bootsteps, graph

What is Mode?
=============

Mode is a library for Python AsyncIO, using the new ``async/await`` syntax
in Python 3.6 to define your program as a set of services.

When starting a larger project using ``asyncio``, it immediately became
apparent that we needed a way to manage the different services running in the
program.  Questions such as "how do we shutdown the event loop" is frequently
answered by telling users to "wait for all coroutines in
asyncio.Task.all_tasks", but we needed more control over what services
where stopped, in what order and what services can we safely shutdown without
waiting for current operations to complete.

The result is Mode and its generic ``Service`` class.

Services
    Programs in Mode are built out of one or more services, where the
    services organizes themselves at runtime into a dependency graph.

    In practice this means services may start other services, that
    when told to stop, does so in the correct order::

        class App(Service):

            def on_init_dependencies(self) -> None:
                return [
                    self.websockets,
                    self.webserver,
                    self.user_cache,
                ]

            async def on_start(self) -> None:
                print('App is starting')

Graph
    If we fill out the rest of this code to implement the additional
    services.

    First the Websocket server service could look like this::

        class Websockets(Service):

            def __init__(self, port: int = 8081, **kwargs: Any) -> None:
                self.port = 8081
                self._server = None
                super().__init__(**kwargs)

            async def on_start(self) -> None:
                self._server = websockets.run()

            async def on_stop(self) -> None:
                if self._server is not None:
                    self._server.close()

    Second, the web server we want to run in a separate thread
    so we use ``ServiceThread`` instead::

        from aiohttp.web import Application
        from mode.threads import ServiceThread

        class Webserver(ServiceThread):

            def __init__(self,
                         port: int = 8000,
                         bind: str = None,
                         **kwargs: Any) -> None:
                self._app = Application()
                self.port = port
                self.bind = bind
                self._handler = None
                self._srv = None
                super().__init__(**kwargs)

            async def on_start(self) -> None:
                handler = self._handler = self._app.make_handler()
                # self.loop is the event loop in this thread
                #   self.parent_loop is the loop that created this thread.
                self._srv = await self.loop.create_server(
                    handler, self.bind, self.port)
                self.log.info('Serving on port %s', self.port)

            async def on_thread_stop(self) -> None:
                # see examples/tutorial.py for an actual example
                self._srv.stop()

    Third, our user cache, which has a background coroutine used to
    remove old expired items from the cache::

        class UserCache(Service):
            _cache: MutableMapping[str, User]

            def on_init(self):
                self._cache = {}

            async def lookup(self, user_id: str) -> User:
                try:
                    return self._cache[user_id]
                except KeyError:
                    user = self._cache[user_id] = await User.objects.get(user_id)
                    return user

            @Service.timer(10)  # execute every 10 seconds.
            def _remove_expired(self):
                remove_expired_users(self._cache)

Proxy
    Now we just need to create these services in our "App" class.

    In our little tutorial example the "app" is the entrypoint for
    our program.  Mode does not have a concept of apps, so we don't
    subclass anything, but we want the app to be reusable in projects
    and keep it possible to start multiple apps at the same time.

    If we create apps at module scope, for example::

        # example/app.py
        from our_library import App
        app = App(web_port=6066)

    It is very important to instantiate services lazily, otherwise
    the :mod:`asyncio` event loop is created too early.

    For services that are defined at module level we can create a
    ``ServiceProxy``::

        from typing import Any

        from mode import Service, ServiceProxy, ServiceT
        from mode.utils.objects import cached_property

        class AppService(Service):
            # the "real" service that App.start() will run

            def __init__(self, app: 'App', **kwargs: Any) -> None:
                self.app = app
                super().__init__(**kwargs)

            def on_init_dependencies(self) -> None:
                return [
                    self.app.websockets,
                    self.app.webserver,
                    self.app.user_cache,
                ]

            async def on_start(self) -> None:
                print('App is starting')

        class App(ServiceProxy):

            def __init__(self,
                         web_port: int = 8000,
                         web_bind: str = None,
                         websocket_port: int = 8001,
                         **kwargs: Any) -> None:
                self.web_port = web_port
                self.web_bind = web_bind
                self.websocket_port = websocket_port

            @cached_property
            def _service(self) -> ServiceT:
                return AppService(self)

            @cached_property
            def websockets(self) -> Websockets:
                return Websockets(
                    port=self.websocket_port,
                    loop=self.loop,
                    beacon=self.beacon,
                )

            @cached_property
            def webserver(self) -> Webserver:
                return Webserver(
                    port=self.web_port,
                    bind=self.web_bind,
                    loop=self.loop,
                    beacon=self.beacon,
                )

            @cached_property
            def user_cache(self) -> UserCache:
                return UserCache(loop=self.loop, beacon=self.beacon)

Worker
    To start your service on the command-line, add an
    entrypoint for a ``Worker`` to start it::

        app = App()

        if __name__ == '__main__':
            from mode import Worker
            Worker(app, loglevel="info").execute_from_commandline()

    Then execute your program to start the worker::

        $ python examples/tutorial.py
        [2018-03-27 15:47:12,159: INFO]: [^Worker]: Starting...
        [2018-03-27 15:47:12,160: INFO]: [^-AppService]: Starting...
        [2018-03-27 15:47:12,160: INFO]: [^--Websockets]: Starting...
        STARTING WEBSOCKET SERVER
        [2018-03-27 15:47:12,161: INFO]: [^--UserCache]: Starting...
        [2018-03-27 15:47:12,161: INFO]: [^--Webserver]: Starting...
        [2018-03-27 15:47:12,164: INFO]: [^--Webserver]: Serving on port 8000
        REMOVING EXPIRED USERS
        REMOVING EXPIRED USERS

    To stop it hit :kbd:`Control-c`::

        [2018-03-27 15:55:08,084: INFO]: [^Worker]: Stopping on signal received...
        [2018-03-27 15:55:08,084: INFO]: [^Worker]: Stopping...
        [2018-03-27 15:55:08,084: INFO]: [^-AppService]: Stopping...
        [2018-03-27 15:55:08,084: INFO]: [^--UserCache]: Stopping...
        REMOVING EXPIRED USERS
        [2018-03-27 15:55:08,085: INFO]: [^Worker]: Gathering service tasks...
        [2018-03-27 15:55:08,085: INFO]: [^--UserCache]: -Stopped!
        [2018-03-27 15:55:08,085: INFO]: [^--Webserver]: Stopping...
        [2018-03-27 15:55:08,085: INFO]: [^Worker]: Gathering all futures...
        [2018-03-27 15:55:08,085: INFO]: [^--Webserver]: Closing server
        [2018-03-27 15:55:08,086: INFO]: [^--Webserver]: Waiting for server to close handle
        [2018-03-27 15:55:08,086: INFO]: [^--Webserver]: Shutting down web application
        [2018-03-27 15:55:08,086: INFO]: [^--Webserver]: Waiting for handler to shut down
        [2018-03-27 15:55:08,086: INFO]: [^--Webserver]: Cleanup
        [2018-03-27 15:55:08,086: INFO]: [^--Webserver]: -Stopped!
        [2018-03-27 15:55:08,086: INFO]: [^--Websockets]: Stopping...
        [2018-03-27 15:55:08,086: INFO]: [^--Websockets]: -Stopped!
        [2018-03-27 15:55:08,087: INFO]: [^-AppService]: -Stopped!
        [2018-03-27 15:55:08,087: INFO]: [^Worker]: -Stopped!

Beacons
    The ``beacon`` object that we pass to services keeps track of the services
    in a graph.

    They are not stricly required, but can be used to visualize a running
    system, for example we can render it as a pretty graph.

    This requires you to have the ``pydot`` library and GraphViz
    installed::

        $ pip install pydot

    Let's change the app service class to dump the graph to an image
    at startup.

        class AppService(Service):

            async def on_start(self) -> None:
                print('APP STARTING')
                import pydot
                import io
                o = io.StringIO()
                beacon = self.app.beacon.root or self.app.beacon
                beacon.as_graph().to_dot(o)
                graph, = pydot.graph_from_dot_data(o.getvalue())
                print('WRITING GRAPH TO image.png')
                with open('image.png', 'wb') as fh:
                    fh.write(graph.create_png())


Creating a Service
==================

To define a service, simply subclass and fill in the methods
to do stuff as the service is started/stopped etc.::

    class MyService(Service):

        async def on_start(self) -> None:
            print('Im starting now')

        async def on_started(self) -> None:
            print('Im ready')

        async def on_stop(self) -> None:
            print('Im stopping now')

To start the service, call ``await service.start()``::

    await service.start()

Or you can use ``mode.Worker`` (or a subclass of this) to start your
services-based asyncio program from the console::

    if __name__ == '__main__':
        import mode
        worker = mode.Worker(MyService(), loglevel='INFO', logfile=None)
        worker.execute_from_commandline()

It's a Graph!
=============

Services can start other services, coroutines, and background tasks.

1) Starting other services using ``add_depenency``::

    class MyService(Service):

        def on_init(self) -> None:
           self.add_dependency(OtherService(loop=self.loop))

2) Start a list of services using ``on_init_dependencies``::

    class MyService(Service):

        def on_init_dependencies(self) -> None:
            return [
                ServiceA(loop=self.loop),
                ServiceB(loop=self.loop),
                ServiceC(loop=self.loop),
            ]

3) Start a future/coroutine (that will be waited on to complete on stop)::

    class MyService(Service):

        async def on_start(self) -> None:
            self.add_future(self.my_coro())

        async def my_coro(self) -> None:
            print('Executing coroutine')

4) Start a background task::

    class MyService(Service):

        @Service.task
        async def _my_coro(self) -> None:
            print('Executing coroutine')


5) Start a background task that keeps running::

    class MyService(Service):

        @Service.task
        async def _my_coro(self) -> None:
            while not self.should_stop:
                # NOTE: self.sleep will wait for one second, or
                #       until service stopped/crashed.
                await self.sleep(1.0)
                print('Background thread waking up')
