torpc 0.0.1

A tiny async tcp and duplex rpc implementation using Tornado IOLoop.

ToRPC(Tornado + RPC) is a tiny tcp and duplex RPC implementation in Python based on Tornado IOLoop. It’s very lightweight and high performance(especially on PyPy).

Notice: ToRPC was only tested on CPython 2.7+ and PyPy 2.5+ until now.

## Examples

RPC server

from tornado import ioloop
from torpc import RPCServer
server = RPCServer(('127.0.0.1', 5000))

@server.service.register()
def echo(x):
    return x

server.start()
ioloop.IOLoop.instance().start()

RPC client

from tornado import ioloop, gen
from torpc import RPCClient

def result_callback(f):
    print(f.result())

@gen.coroutine
def using_gen_style():
    want_to_say = 'way to explore'
    ret = yield rc.call('echo', want_to_say)
    assert ret == want_to_say
    print('gen_style complete')

rc = RPCClient(('127.0.0.1', 5000))
rc.call('echo', 'hello world', callback=result_callback)
future = rc.call('echo', 'code for fun')
future.add_done_callback(result_callback)
using_gen_style()
ioloop.IOLoop.instance().start()

See more in examples.

Performance

OS: CentOS 6.6 x64 CPU: Intel i5-3470 3.20GHz Memory: 8 GB 1600 MHz DDR3 Python: 2.7.10 PyPy: 4.0.0

environment	call coroutine(qps)	callback(qps)
Python(with timeout)	9842	11614
Python	13192	16638
PyPy(with timeout)	40486	41225
PyPy	53252	59151
PyPy(unix domain)	67100	74362

In this benchmark, Python loops 100k times and PyPy loops 1000k times, then run 3 times of each, the result is on gist:benchmark_result.txt

Document

中文