os-aio-pod

A container of AIO components.

This project is a framework for combining multiple AIO components into one. For example, you can easily extend your TCP server with a HTTP server to offer HTTP API. Usage is simple, write regular coroutine and config file, start with the command line tool, the framework will take care of the work loop.

Each of the coroutine is a magic bean, this framework is a magic pod.

Conception

Custom coroutine is packed as a asyncio.Task, we call it: bean. Thanks to the Python asyncio, all beans can work together. A runner named pod maintains the beans and the whole work loop life cycle.

Each bean has a unique id. label can also be used for identifying one or more beans.

You can access other beans by using a context object with id or label.

Signals can be dispatched to each beans which registered callback function.

Install

pip install os-aio-pod

There are some extra packages can be installed for more features.

subpackage	install command	enables
uvloop	pip install os-aio-pod[uvloop]	enable uvloop
uvicorn	pip install os-aio-pod[uvicorn]	enable uvicorn http server adapter
aiohttp	pip install os-aio-pod[aiohttp]	enable aiohttp http server adapter
aiomonitor	pip install os-aio-pod[aiomonitor]	enable aiomonitor adapter

Usage

Quick start

1. Write your coroutine function

   # example.py
   import asyncio
   
   async def helloworld(**kwargs):
       await asyncio.sleep(1)
       print("hello world!", kwargs)
   

2. run with os-aio-pod

   os-aio-pod run example.helloworld:hi="Ozzy"
   

APIs

Custom coroutine

Actually, there are three types of coroutine code for your choice:

1. Regular coroutine, the hello object of the following example.

   import asyncio
   
   async def hello_world(**kwargs):
       print('hello world!')
       await asyncio.sleep(1)
   
   hello = hello_world()
   

2. Regular coroutine fucntion, the hello_world function of the above example. Keyword arguments can be set in the config file.

3. A class with async def __call__(self) method and have context as init arguments:

   class HelloWorld(object):
   
       def __init__(self, context):
           self.context = context
           
       async def __call__(self, **kwargs):
           print('hello world!')
           await asyncio.sleep(1)
   

Context

When you use class type coroutine, you can use the context to communicate with the framework and other beans.

Since v0.1.27, pass_context decorator can be used for passing context to function as the first argument if it is invoked by the framework. lable can also be specified as argument @pass_context(label="app")

from os_aio_pod.decorators import pass_context

@pass_context
async def hello_world(context, **kwargs):
    print(context, kwargs)

Signals

Thanks to asyncio_dispatch, we easily can register and deliver signals.

Typically, you should only use context APIs to process signals.

The system SIGINT,SIGTERM are caught by the framework to handler shutdown stuff after dispatch to each registered callback.

Other system signals are not supported yet.

Configure

Config file is a regular Python file, all upper case variables will pass to the frame work which can be accessed later. The reserved key words:

• BEANS: a list of bean config dict, the reserved key words of each bean config are: core, label, other keyword arguments will pass to your function

  core: string path of your coroutine

  label: optional, can be used to trace your bean

• LOG_LEVEL: logger level, default INFO

• LOOP_TYPE: default is asyncio, can be uvloop when you install uvloop

• DEBUG: enable debug mode, default False

• STOP_WAIT_TIME: the wait time when recieve signal(SIGINT, SIGTERM). Once timeout, all unfinished bean will be cancelled. Default is None, indicate wait until all beans done

Example:

config.py

BEANS = [
    {
        'core' : 'hello_world.HelloWorld',
        'label': 'first-bean',
        'key1' : 'value1',
    }
]

LOG_LEVEL      = 'debug'
LOOP_TYPE      = 'asyncio'
DEBUG          = False
STOP_WAIT_TIME = 10

Command line

os-aio-pod command can be used to start the whole framework, the typical usage:

$ os-aio-pod run -c config.py

or quick start

$ os-aio-pod run [awaitable-func1:k1=v1,k2=v2] [awaitable-func2:k3=v3,k4=v4]

The reserved config key words(exclude BEANS) can be set by passing command line options.

$ os-aio-pod run --help

Built-In Components

There are some built-in adapters can be used for convenient:

• built-in simple server class os_aio_pod.contrib.simple.Server

  It is a scaffold base class for simple server

  from os_aio_pod.contrib.simple import Server
  
  class YourServer(Server):
  
      # can be async/sync 
      async def startup(self, **kwargs):
          pass
  
      # can be async/sync 
      async def cleanup(self, **kwargs):
          pass
  
      async def run(self, **kwargs):
          print(self.config)
  
      # on kill(Ctrl+C)
      def on_stop(self, **kwargs):
          pass
  

• ptpython, python REPL

  pip install ptpython contextvars
  

  BEANS = [
      {
          'core': 'os_aio_pod.contrib.ptpython.TelnetServerAdapter',
      }
  ]
  

  you can connect this server with telnet

• uvicorn, a lightning-fast ASGI server

  pip install uvicorn
  

  BEANS = [
      {
          'core': 'os_aio_pod.contrib.uvicorn.UvicornAdapter',
          'app' : 'your.app.object.path'
      }
  ]
  

  a context object named aio_pod_context will attached to the app object

• aiohttp, a well known aio http server

  pip install aiohttp
  

  BEANS = [
      {
          'core': 'os_aio_pod.contrib.aiohttp.WebAdapter',
          'app' : 'your.app.object.path'
      }
  ]
  

  a context object named aio_pod_context will attached to the app object

• aiomonitor, adds monitor and python REPL capabilities for asyncio application

  pip install aiomonitor
  

  BEANS = [
      {
          'core': 'os_aio_pod.contrib.aiomonitor.AioMonitorAdapter',
      }
  ]
  

• built-in tcp server

  An event driven server can be inherited from os_aio_pod.contrib.tcp_server.Server(default server).

  If protocol is configured, low-level networking protocol interface will be used instead of the server's on_connect method. The server instance can be accessed with your_protocol.server

  BEANS = [
      {
          'core': 'os_aio_pod.contrib.tcp_server.TCPServerAdapter',
          # 'protocol': 'your.asyncio.Protocol.path'
          # 'server': 'your.event.driven.server'
      }
  ]
  

• built-in producer-consumer model

  One producer and multi-consumers is a common model. You can inherit from os_aio_pod.contrib.pcflow.Server(which is inherit from built-in simple server) and implement produce and consume methods to run as this model.

  import asyncio
  
  from os_aio_pod.contrib.pcflow import Server
  
  
  class YourProducerConsumerServer(Server):
      def startup(self, **kwargs):
          self.stopping = False
  
      def on_stop(self, **kwargs):
          self.stopping = True
  
      async def produce(self, **kwargs):
          while not self.stopping:
              await asyncio.sleep(1)
              yield 1
  
      async def consume(self, obj, **kwargs):
          print(obj) 
  

  BEANS = [
      {
          'core': 'YourProducerConsumerServer',
          'consumer_num': 10,
      }
  ]
  

Unit Tests

tox

License

MIT licensed.