buffered-pipe 0.1.0

Buffered pipe through shared memory.

buffered_pipe

Buffered pipe through shared memory.

core features

• Pipe: alias for Generic_Pipe
• Generic_Pipe: free-length data, bytes or picklable
• Static_Pipe: fixed length data, only bytes supported

install

---

pip install buffered-pipe

usage

---

• Generic_Pipe

import multiprocessing
import threading

from buffered_pipe import Generic_Pipe


def foo(barrier, pipe_send):
    pipe_send.register()
    barrier.wait()
    pipe_send.send("Hello, world!")


def bar(barrier, pipe_recv):
    pipe_recv.register()
    barrier.wait()
    print(pipe_recv.recv())


if __name__ == "__main__":
    barrier = multiprocessing.Barrier(3)
    pipe_recv, pipe_send = Generic_Pipe(
        buffer_size=2048, duplex=False
    )
    P = multiprocessing.Process(target=foo, args=(barrier, pipe_send))
    T = threading.Thread(target=bar, args=(barrier, pipe_recv))
    P.start()
    T.start()
    barrier.wait()
    P.join()
    T.join()

• Static_Pipe

import multiprocessing
import threading

from buffered_pipe import Static_Pipe


def foo(barrier, pipe_send):
    pipe_send.register()
    barrier.wait()
    ret_string = "Hello, world!"
    ret_string = ret_string + " " * (32 - len(ret_string))
    ret_string = ret_string.encode()
    pipe_send.send(ret_string)


def bar(barrier, pipe_recv):
    pipe_recv.register()
    barrier.wait()
    ret_string = pipe_recv.recv()
    print(ret_string.decode())


if __name__ == "__main__":
    barrier = multiprocessing.Barrier(3)
    pipe_recv, pipe_send = Static_Pipe(object_size=32, object_count=16, duplex=False)
    P = multiprocessing.Process(target=foo, args=(barrier, pipe_send))
    T = threading.Thread(target=bar, args=(barrier, pipe_recv))
    P.start()
    T.start()
    barrier.wait()
    P.join()
    T.join()

• When you create a process or thread, call register of Pipe
• The buffer_size of Generic_Pipe is in bytes.
• The object_size of Static_Pipe is in bytes.
• object_count in Static_Pipe is the maximum number of objects placed in the buffer.
• duplex is equivalent to multiprocessing.Pipe.

tests

---

> benchmark_result.txt
for f in benchmarks/*.py; do python3 "$f" &>> benchmark_result.txt; done
> test_result.txt
for f in tests/*.py; do python3 "$f" &>> test_result.txt; done