Aiochan is a library written to bring the wonderful idiom of CSP-style concurrency to python. The implementation is based on the battle-tested Clojure library core.async, while the API is carefully crafted to feel as pythonic as possible.
- Doing concurrency in Python was painful
- asyncio sometimes feels too low-level
- I am constantly missing capabilities from golang and core.async
- It is much easier to port core.async to Python than to port all those wonderful python packages to some other language.
What am I getting?
- Pythonic API that includes everything you'd need for CSP-style concurrency programming
- Works seamlessly with existing asyncio-based libraries
- Fully tested
- Fully documented
- Guaranteed to work with Python 3.5.2 or above and PyPy 3.5 or above
- Depends only on python's core libraries, zero external dependencies
- Proven, efficient implementation based on Clojure's battle-tested core.async
- Familiar semantics for users of golang's channels and Clojure's core.async channels
- Flexible implementation that does not depend on the inner workings of asyncio at all
- Permissively licensed
- A beginner-friendly tutorial to get newcomers onboard as quickly as possible
How to install?
pip3 install aiochan
How to use?
Read the beginner-friendly tutorial that starts from the basics. Or if you are already experienced with golang or Clojure's core.async, start with the quick introduction and then dive into the API documentation.
I want to try it first
In addition to the introduction and the tutorial, we have the complete set of examples from Rob Pike's Go concurrency patterns translated into aiochan. Also, here is a solution to the classical dining philosophers problem.
I still don't know how to use it
We are just starting out, but we will try to answer aiochan-related questions on stackoverflow as quickly as possible.
I found a bug
File an issue, or if you think you can solve it, a pull request is even better.
Do you use it in production? For what use cases?
aiochan is definitely not a toy and we do use it in production, mainly in the two following scenarios:
- Complex data-flow in routing. We integrate aiochan with an asyncio-based web server. This should be easy to understand.
- Data-preparation piplelines. We prepare and pre-process data to feed into our machine learning algorithms as fast as possible so that our algorithms spend no time waiting for data to come in, but no faster than necessary so that we don't have a memory explosion due to data coming in faster than they can be consumed. For this we make heavy use of parallel_pipe and parallel_pipe_unordered. Currently we are not aware of any other library that can completely satisfy this need of ours.
What's up with the logo?
It is our 'hello world' example:
import aiochan as ac async def blue_python(c): while True: # do some hard work product = "a product made by the blue python" await c.put(product) async def yellow_python(c): while True: result = await c.get() # use result to do amazing things print("A yellow python has received", result) async def main(): c = ac.Chan() for _ in range(3): ac.go(blue_python(c)) for _ in range(3): ac.go(yellow_python(c))
in other words, it is a 3-fan-in on top of a 3-fan-out. If you run it, you will have an endless stream of
A yellow python has received a product made by the blue python.
If you have no idea what this is, read the tutorial.