Very simple RPCs built with sockets & pickle in python. Support numpy as input and output arguments.
pip install picklerpcServer
import picklerpc
import numpy as np
# function
def func_numpy(data:np.ndarray) -> np.ndarray:
return data + np.random.random(size=data.shape)
def func_str(name: None) -> str:
return "Hello, %s" % name
def func_none() -> None:
return None
def func_number(a, b):
return a+b
class HelloRPC(object):
def func_obj(self, name) -> str:
return "Hello, %s" % name
# register
server = picklerpc.PickleRPCServer(('localhost', 9102))
server.register_function(func_numpy)
server.register_function(func_str)
server.register_function(func_none)
server.register_function(func_number)
server.register_function(lambda x,y: x-y, 'func_lambda')
server.register_instance(HelloRPC()) #from instance
server.serve_forever()Client
import picklerpc
import numpy as np
client = picklerpc.PickleRPCClient(('localhost', 9102))
print(client.func_numpy(np.arange(10), dtype=np.float32))
print(client.func_str('Pickle RPC'))
print(client.func_none())
print(client.func_number(np.range(10), 3))
print(client.func_lambda(np.range(10), 3))
print(client.func_obj('Pickle RPC'))The picklerpc is most 10x faster than zerorpc and xmlrpc.
| zerorpc | xmlrpc | picklerpc (mine) | |
|---|---|---|---|
| Protocal | tcp | http | tcp |
| Data Stream | - | xml | pickle |
| Multiple Client | Yes | Yes | Yes |
| Data Types | Natives | Natives | Natives + Numpy |
| Speed (local loop) | 683 µs | 462 µs | 59 µs |
| Speed (remote) | 2.76 ms | 4.64 ms | 0.54 ms |