dirt
: a comprehensive framework for building Python applications which are part of a service oriented architecture
dirt
provides all the tools required to build long-running Python
applications ("services") which can communicate over RPC.
Specifically, it provides:
- Tools for exposing methods over RPC (using either a dirt custom protocol or ZeroRPC, including nicities like iPython tab completion).
- A framework for creating long-running applications, and exposing their methods over RPC.
- Tools for running multiple applications in one terminal, either for development or in production.
- A simple syntax for defining applications (currently Django-style
settings.py
files are best supported, but.ini
could easily be supported too).
An application can be as simple as:
$ cat app.py import gevent import logging from dirt import DirtApp, runloop log = logging.getLogger(__name__) class PingAPI(object): def ping(self): return "pong" class PingApp(DirtApp): def get_api(self, edge, call): return PingAPI() class LongRunningApp(DirtApp): @runloop(log) def serve(self): ping_app = self.settings.get_api("ping") while True: result = ping_app.ping() log.info("ping: %r", result) gevent.sleep(1) $ cat settings.py from dirt import logging_default USE_RELOADER = False DIRT_APP_PIDFILE = "/tmp/dirt-example-{app_name}.pid" LOGGING = logging_default("/tmp/dirt-example-{app_name}.log", root_level="INFO") class PING: app_class = "app.PingApp" bind_url = "zrpc+tcp://127.0.0.1:9990" remote_url = bind_url class LONG_RUNNING: app_class = "app.LongRunningApp" $ ./run ping long_running 23:20:21.289 ping INFO dirt.app: binding to zrpc+tcp://127.0.0.1:9990... 23:20:21.477 long_running INFO app: ping: 'pong' 23:20:22.380 long_running INFO app: ping: 'pong'
Some notes:
dirt
plays well with others, integrating easily into existing projects. See theexample_project/
directory for a Django project which usesdirt
.dirt
depends ongevent==1.0
. In theory it should be possible for applications to use operating system threads, but this hasn't been tested.
dirt
is used by Luminautics in a production environment, and has been
incredibly stable for the last six months.
That said, there will likely be some undocumented assumptions which could affect other users, so some amount of testing should be done before trusting it in your production environment.