There is a lot of boilerplate and copypasta associated with building services (or any other software that runs continuously and does things)
sparts is a python library that aims to eliminate as much of the boiler plate as possible, making it as dead simple to write new services with little to no code.
- Be able to implement services using sparts with as little code as possible
- Support as many RPC transports as possible (thrift, http, dbus, etc)
- Make it painless to integrate services that require custom IO loops (twisted, tornado, glib, etc)
A sparts service typically consists of two parts, the "service" definition, and its tasks.
sparts.vtask.VService - This is the meat of any service implemented on sparts.
Simply put, just subclass VService, and run initFromCLI() and you are done.
For example, myservice.py:
from sparts.vservice import VService class MyService(VService): pass MyService.initFromCLI()
Now, you can run this file with python -h (to see the available options), or run with:
You should the following:
DEBUG:VService:All tasks started
And pressing ^C will emit:
^CINFO:VService:KeyboardInterrupt Received! Stopping Tasks... INFO:VService:Instance shut down gracefully
This is pretty damn useless. That's where Tasks come into play
sparts.vtask.VTask - This is the base class for all tasks
Tasks are what trigger your program to take action. This action can be processing periodic events, handling HTTP requests, handling thrift requests, working on items from a queue, waking up on an event, operating some ioloop, or whatever.
Here's a simple example of a service with tasks (requires tornado installed):
from sparts.vservice import VService from sparts.tasks.tornado import TornadoIOLoopTask, TornadoHTTPTask class MyService(VService): ALL_TASKS=[TornadoIOLoopTask, TornadoHTTPTask] DEFAULT_PORT=8000 MyService.initFromCLI()
Now running it emits:
> python myservice.py --http-port 8000 INFO:MyService.TornadoHTTPTask:TornadoHTTPTask Server Started on 0.0.0.0:8000 DEBUG:MyService:All tasks started
And as you can see, you can curl the webserver:
> curl localhost:8000 Hello, world