Version: | 0.6.1 |
---|---|
Web: | http://mode-ng.readthedocs.org/ |
Download: | https://pypi.org/project/mode-ng |
Source: | https://github.com/lqhuang/mode-ng |
Keywords: | async, service, framework, actors, bootsteps, graph |
Original mode
homepage: https://github.com/ask/mode. Thanks to its creator Ask Solem (@ask).
Project homepage: https://github.com/faust-streaming/mode
We have decided to fork the original Mode project because there is a critical process of releasing new versions which causes uncertainty in the community. Everybody is welcome to contribute to this fork, and you can be added as a manitainer.
We want to:
- Ensure continues release
- Code quality
- Support latest Python versions
- Update the documentation
and more...
WIP notice. Welcome to try it in your programs and feedback!
Well, here is an another fork for mode
. Generally, my target is to
create an individual and separated repository to develop mode
for next
stage/generation and keep rolling forward fastly.
mode
could be a very potential and powerful framework for various
applications. So I very care about how faust-streaming
and mode-streaming
goes in future. Currently the most important thing in developing mode-streaming
is to fix bugs and keep back compatibility for faust-streaming
, it would be
uncertain or not-willing to add new features. For now, one big problem is if I
try to continue working on current mode-streaming
branch, it's hard to me
to know its consequences in faust-streaming
. I don't want to introduce
break changes and inconsistent behaviors.
Hence, mode-ng
provides a new package to make some aggressive improvements, do some experiments,
and do not consider compatible problems from faust-streaming
. At least,
mode-ng
can be quickly used by more users with more advanced features to
build their own applications. In the future, if this fork could be ported back
to mode-streaming
or used as base framework of faust-streaming
, that
would be really great!
Here are some thoughts from practical experiences and what I want to do next step:
- Bug fixes: yeah, why not.
- Use standard library implementations: When mode was first developed,
many features haven't exist, so there are many hacks and tricky solutions in
mode
codes. Likecached_property
,AsyncMock
,loop arguments
, evenObject
class (missing some inner__xxx__
attrs after redefinition). - Port some features from faust:
web
module in faust is really useful for building application. With web part,mode
is able to expose, control, monitor and more from outside api. - Improve or complete left part of
signal
module: Some modules likeSignal
are unfinished. It will be useful for some observer patterns in programming. - Add some message commuting behaviors like real actors? (for thread serices?)
- More documents and more examples
It's more like a personal flavor for now. But if we say py3.6
bring us stable async/await
syntax firstly to introduce wonderful coroutine concurrency, I thought py3.10
would be the next major popular minimal supported version with a more matured asyncio
api interfaces. From feedbacks from many real cases, asyncio
interfaces have changed a lot. Happy news is it stabilized by development of std library. And Python 3.10 is faster than before, though Python is not a language which cares about speed, seldom people don't want to higher performance.
This is not absolute, it aslo depends how many people want back compatibility for older versions.
You can install Mode either via the Python Package Index (PyPI) or from source.
To install using pip
:
$ pip install -U mode-ng
Download the latest version of Mode from http://pypi.org/project/mode-ng
You can install it by doing the following:
$ tar xvfz mode-ng-0.6.1.tar.gz
$ cd mode-0.6.1
$ python setup.py build
$ python setup.py install
The last command must be executed as a privileged user if you are not currently using a virtualenv.
You can install the latest snapshot of Mode using the following pip command:
$ pip install https://github.com/lqhuang/mode-ng/zipball/master#egg=mode-ng
Mode is a very minimal Python library built-on top of AsyncIO that makes it much easier to use.
In Mode your program is built out of services that you can start, stop, restart and supervise.
A service is just a class:
class PageViewCache(Service):
redis: Redis = None
async def on_start(self) -> None:
self.redis = connect_to_redis()
async def update(self, url: str, n: int = 1) -> int:
return await self.redis.incr(url, n)
async def get(self, url: str) -> int:
return await self.redis.get(url)
Services are started, stopped and restarted and have callbacks for those actions.
It can start another service:
class App(Service):
page_view_cache: PageViewCache = None
async def on_start(self) -> None:
await self.add_runtime_dependency(self.page_view_cache)
@cached_property
def page_view_cache(self) -> PageViewCache:
return PageViewCache()
It can include background tasks:
class PageViewCache(Service):
@Service.timer(1.0)
async def _update_cache(self) -> None:
self.data = await cache.get('key')
Services that depends on other services actually form a graph that you can visualize.
Mode optionally provides a worker that you can use to start the program, with support for logging, blocking detection, remote debugging and more.
To start a worker add this to your program:
if __name__ == "__main__":
from mode import Worker
Worker(Service(), 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 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!
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())
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,
daemon=False,
)
worker.execute_from_commandline()
Services can start other services, coroutines, and background tasks.
- Starting other services using
add_depenency
:
class MyService(Service):
def __post_init__(self) -> None:
self.add_dependency(OtherService(loop=self.loop))
- 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),
]
- 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')
- Start a background task:
class MyService(Service):
@Service.task
async def _my_coro(self) -> None:
print('Executing coroutine')
- 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')
Yes! Use gevent/eventlet as a bridge to integrate with asyncio.
This works with any blocking Python library that can work with gevent.
Using gevent requires you to install the aiogevent
module,
and you can install this as a bundle with Mode:
$ pip install -U mode-ng[gevent]
Then to actually use gevent as the event loop you have to execute the following in your entrypoint module (usually where you start the worker), before any other third party libraries are imported:
#!/usr/bin/env python3
import mode.loop
mode.loop.use('gevent')
# execute program
REMEMBER: This must be located at the very top of the module, in such a way that it executes before you import other libraries.
This works with any blocking Python library that can work with eventlet.
Using eventlet requires you to install the aioeventlet
module,
and you can install this as a bundle with Mode:
$ pip install -U mode-ng[eventlet]
Then to actually use eventlet as the event loop you have to execute the following in your entrypoint module (usually where you start the worker), before any other third party libraries are imported:
#!/usr/bin/env python3
import mode.loop
mode.loop.use('eventlet')
# execute program
REMEMBER: It's very important this is at the very top of the module, and that it executes before you import libraries.
Yes! Use the tornado.platform.asyncio
bridge:
http://www.tornadoweb.org/en/stable/asyncio.html
Yes! Use the asyncio reactor implementation: https://twistedmatrix.com/documents/17.1.0/api/twisted.internet.asyncioreactor.html
If you get warnings such as this at shutdown:
Task was destroyed but it is pending!
task: <Task pending coro=<Service._execute_task() running at /opt/devel/mode/mode/services.py:643> wait_for=<Future pending cb=[<TaskWakeupMethWrapper object at 0x1100a7468>()]>>
Task was destroyed but it is pending!
task: <Task pending coro=<Service._execute_task() running at /opt/devel/mode/mode/services.py:643> wait_for=<Future pending cb=[<TaskWakeupMethWrapper object at 0x1100a72e8>()]>>
Task was destroyed but it is pending!
task: <Task pending coro=<Service._execute_task() running at /opt/devel/mode/mode/services.py:643> wait_for=<Future pending cb=[<TaskWakeupMethWrapper object at 0x1100a7678>()]>>
Task was destroyed but it is pending!
task: <Task pending coro=<Event.wait() running at /Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/asyncio/locks.py:269> cb=[_release_waiter(<Future pendi...1100a7468>()]>)() at /Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/asyncio/tasks.py:316]>
Task was destroyed but it is pending!
task: <Task pending coro=<Event.wait() running at /Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/asyncio/locks.py:269> cb=[_release_waiter(<Future pendi...1100a7678>()]>)() at /Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/asyncio/tasks.py:316]>
It usually means you forgot to stop a service before the process exited.
For guidance on setting up a development environment and how to make a contribution to mode-ng, see the contributing guidelines.
Check code of conduct for recommended or discouraged behaviors while communicating.