Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
240 changes: 240 additions & 0 deletions Doc/howto/asyncio-chat-server.rst
Original file line number Diff line number Diff line change
@@ -0,0 +1,240 @@
.. _asyncio-chat-server-howto:

***********************************************
Building a TCP chat server with :mod:`!asyncio`
***********************************************

This guide walks you through building a TCP chat server where multiple users
can connect and exchange messages in real time. Along the way, you will learn
how to use :ref:`asyncio streams <asyncio-streams>` for network programming.

The guide assumes basic Python knowledge --- functions, classes, and context
managers --- and a general understanding of async/await.

.. seealso::

:ref:`a-conceptual-overview-of-asyncio`
An introduction to the fundamentals of asyncio.

:mod:`asyncio` reference documentation
The complete API reference.


.. _asyncio-chat-server-echo:

Starting with an echo server
============================

Before building the chat server, let's start with something simpler: an echo
server that sends back whatever a client sends.

The core of any asyncio network server is :func:`asyncio.start_server`. You
give it a callback function, a host, and a port. When a client connects,
asyncio calls your callback with two arguments: a
:class:`~asyncio.StreamReader` for receiving data and a
:class:`~asyncio.StreamWriter` for sending data back. Each connection runs
as its own coroutine, so multiple clients are handled concurrently.
Comment on lines +28 to +36
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It would be nice if we provided a small example of this before diving into the full echo server. We should start with something small, and then slowly build on that as we introduce concepts to the user.


The :meth:`~asyncio.StreamWriter.write` method buffers data without sending
it immediately. Awaiting :meth:`~asyncio.StreamWriter.drain` flushes the
buffer and applies back-pressure if the client is slow to read. Similarly,
:meth:`~asyncio.StreamWriter.close` initiates shutdown, and awaiting
:meth:`~asyncio.StreamWriter.wait_closed` waits until the connection is
fully closed.

Using the server as an async context manager (``async with server``) ensures
it is properly cleaned up when done. Calling
:meth:`~asyncio.Server.serve_forever` keeps the server running until the
program is interrupted. Finally, :func:`asyncio.run` starts the event loop
and runs the top-level coroutine.
Comment on lines +38 to +49
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same here -- let's introduce both these concepts with their own examples.


Here is a complete echo server::

import asyncio

async def handle_client(reader, writer):
addr = writer.get_extra_info('peername')
print(f'New connection from {addr}')

while True:
data = await reader.readline()
if not data:
break
writer.write(data)
await writer.drain()

print(f'Connection from {addr} closed')
writer.close()
await writer.wait_closed()

async def main():
server = await asyncio.start_server(
handle_client, '127.0.0.1', 8888)
addr = server.sockets[0].getsockname()
print(f'Serving on {addr}')

async with server:
await server.serve_forever()

asyncio.run(main())

To test, run the server in one terminal and connect from another using ``nc``
(or ``telnet``):
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why is this in parentheses?

Suggested change
(or ``telnet``):
or ``telnet``:


.. code-block:: none
$ nc 127.0.0.1 8888
.. _asyncio-chat-server-building:

Building the chat server
========================

The chat server extends the echo server with two additions: tracking connected
clients and broadcasting messages to everyone.

Client tracking
---------------

We store each connected client's name and :class:`~asyncio.StreamWriter` in a
module-level dictionary. When a client connects, ``handle_client`` prompts for
a name and adds the writer to the dictionary. A ``finally`` block ensures the
client is always removed on disconnect, even if the connection drops
unexpectedly.
Comment on lines +100 to +104
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is introduced as a solution, but we didn't explain the problem. Why do we need to store writers? Why do we need to remove them?


Broadcasting messages
---------------------

To send a message to all clients, we define a ``broadcast`` function.
:class:`asyncio.TaskGroup` sends to all recipients concurrently rather than
one at a time. :func:`contextlib.suppress` silently handles any
:exc:`ConnectionError` from clients that have already disconnected::
Comment on lines +111 to +112
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think we need to explain contextlib.suppress -- that's not part of asyncio and should fall under the umbrella of "basic Python knowledge". If you want, we can add a comment in the example explaining what it does.


async def broadcast(message, *, sender=None):
"""Send a message to all connected clients except the sender."""
async def send(writer):
with contextlib.suppress(ConnectionError):
writer.write(message.encode())
await writer.drain()

async with asyncio.TaskGroup() as tg:
# Iterate over a copy: clients may leave during the broadcast.
for name, writer in list(connected_clients.items()):
if name != sender:
tg.create_task(send(writer))

The complete chat server
------------------------

Putting it all together::

import asyncio
import contextlib

connected_clients: dict[str, asyncio.StreamWriter] = {}

async def broadcast(message, *, sender=None):
"""Send a message to all connected clients except the sender."""
async def send(writer):
with contextlib.suppress(ConnectionError):
writer.write(message.encode())
await writer.drain()

async with asyncio.TaskGroup() as tg:
# Iterate over a copy: clients may leave during the broadcast.
for name, writer in list(connected_clients.items()):
if name != sender:
tg.create_task(send(writer))

async def handle_client(reader, writer):
addr = writer.get_extra_info('peername')

writer.write(b'Enter your name: ')
await writer.drain()
data = await reader.readline()
if not data:
writer.close()
await writer.wait_closed()
return

name = data.decode().strip()
connected_clients[name] = writer
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You may mention that the code doesn't handle two clients with the same name properly, it's left as an exercice to the reader :-)

print(f'{name} ({addr}) has joined')
await broadcast(f'*** {name} has joined the chat ***\n', sender=name)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe move this code in the try/finally block to cleanup the client if broadcast() raises ConnectionError?


try:
while True:
data = await reader.readline()
if not data:
break
message = data.decode().strip()
if message:
print(f'{name}: {message}')
await broadcast(f'{name}: {message}\n', sender=name)
except ConnectionError:
pass
finally:
# Ensure cleanup even if the client disconnects unexpectedly.
del connected_clients[name]
print(f'{name} ({addr}) has left')
await broadcast(f'*** {name} has left the chat ***\n')
writer.close()
await writer.wait_closed()

async def main():
server = await asyncio.start_server(
handle_client, '127.0.0.1', 8888)
addr = server.sockets[0].getsockname()
print(f'Chat server running on {addr}')

async with server:
await server.serve_forever()

asyncio.run(main())

To test, start the server and connect from two or more terminals using ``nc``
(or ``telnet``):

.. code-block:: none
$ nc 127.0.0.1 8888
Enter your name: Alice
*** Bob has joined the chat ***
Bob: Hi Alice!
Hello Bob!
Each message you type is broadcast to all other connected users.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
Each message you type is broadcast to all other connected users.
Each message you type is broadcasted to all other connected users.



.. _asyncio-chat-server-timeout:

Adding an idle timeout
======================

To disconnect clients who have been idle for too long, wrap the read call in
:func:`asyncio.timeout`. This async context manager takes a duration in
seconds. If the enclosed ``await`` does not complete within that time, the
operation is cancelled and :exc:`TimeoutError` is raised. This frees server
resources when clients connect but stop sending data.

Replace the message loop in ``handle_client`` with::
Comment on lines +212 to +221
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This section should go above the complete example, and then the timeout code should be in the full code.


try:
while True:
try:
async with asyncio.timeout(300): # 5-minute timeout
data = await reader.readline()
except TimeoutError:
writer.write(b'Disconnected: idle timeout.\n')
await writer.drain()
break
if not data:
break
message = data.decode().strip()
if message:
await broadcast(f'{name}: {message}\n', sender=name)
except ConnectionError:
pass
finally:
# ... (cleanup as before) ...
2 changes: 2 additions & 0 deletions Doc/howto/index.rst
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@ Python Library Reference.
:hidden:

a-conceptual-overview-of-asyncio.rst
asyncio-chat-server.rst
cporting.rst
curses.rst
descriptor.rst
Expand Down Expand Up @@ -42,6 +43,7 @@ Python Library Reference.
General:

* :ref:`a-conceptual-overview-of-asyncio`
* :ref:`asyncio-chat-server-howto`
* :ref:`annotations-howto`
* :ref:`argparse-tutorial`
* :ref:`descriptorhowto`
Expand Down
3 changes: 3 additions & 0 deletions Doc/library/asyncio.rst
Original file line number Diff line number Diff line change
Expand Up @@ -34,6 +34,9 @@ asyncio is often a perfect fit for IO-bound and high-level
:ref:`a-conceptual-overview-of-asyncio`
Explanation of the fundamentals of asyncio.

:ref:`asyncio-chat-server-howto`
Build a TCP chat server with asyncio streams.

asyncio provides a set of **high-level** APIs to:

* :ref:`run Python coroutines <coroutine>` concurrently and
Expand Down
Loading