kcp.py

Python bindings and networking for the KCP protocol.

此README同时也有中文版本

What is KCP?

KCP is a protocol focusing on low latency data delivery with a guarantee of data delivery. It serves as an alternative to the TCP protocol.

How to install?

kcp.py is available on PyPi, meaning installing is as simple as running

pip install kcp

Examples

Just the raw connection

While kcp.py features a diverse set of pre-implemented uses of KCP (see below), it also allows you to directly manage your KCP connections. Here is an example using two independent connections locally.

from kcp import KCP

# Create two connections using the same conversation ID.
kcp1 = KCP(
    conv_id=1,
)

kcp2 = KCP(
    conv_id=1,
)

# Update their timing information.
kcp1.update()
kcp2.update()


# Set each connection to send data to the other one (usually this would go through some network layer, but
# for the purpose of the example we do this).
@kcp1.outbound_handler
def send_kcp1(_, data: bytes) -> None:
    kcp2.receive(data)


@kcp2.outbound_handler
def send_kcp2(_, data: bytes) -> None:
    kcp1.receive(data)

# Enqueue data to be sent and send it off.
kcp1.enqueue(b"Hello, world!")
kcp1.flush()

print(kcp2.get_received()) # b"Hello, world!"

Asynchronous Server

kcp.py features an implementation of an asynchronous server using the event loop protocol API.

from kcp.server import Connection
from kcp.server import KCPServerAsync

# Create the initial server instance.
server = KCPServerAsync(
    "127.0.0.1",
    9999,
    conv_id=1,
    no_delay=True,
)

# Ability to set performance options after initialisation.
server.set_performance_options(
    update_interval=10,
)


# Ran when the server starts.
@server.on_start
async def on_start() -> None:
    print("Server started!")


# Ran when a connection is made.
@server.on_data
async def on_data(connection: Connection, data: bytes) -> None:
    print(f"Received data from {connection.address}: {data}")


server.start()

Client

kcp.py also implements a KCP client using Python's sockets and threads.

from kcp import KCPClientSync

client = KCPClientSync(
    "127.0.0.1",
    9999,
    conv_id=1,
)


@client.on_data
def handle_data(data: bytes) -> None:
    print(data)


@client.on_start
def on_start() -> None:
    print("Connected to server!")

    while True:
        client.send(b"Data!")


client.start()

You may find more examples in the examples directory within the repo.

Features

• Bindings to the C implementation of KCP
• Pythonic API over said C bindings
• Asynchronous KCP Client
• Synchronous KCP Client
• Asynchronous KCP Server
• Full support for installation through pip

Credit

kcp.py uses the official KCP implementation behind the scenes.