async-redis-rate-limiters 0.0.6

Rock solid async python generic distributed rate limiters (concurrency and time) backed by Redis.

async-redis-rate-limiters

Rock solid async python generic distributed rate limiters (concurrency and time) backed by Redis.

[!WARNING]
This is a very preliminary version of the library and only concurrency limiters are available for now.

Features

• ✅ Support very high concurrency (>100K), keep a reasonable number of connections to Redis (default: 300)
• ✅ Rock solid with Redis/Network failures (multiple attempts, exponential backoff, etc.)
  • you can restart the Redis server during the execution without any exception or losing any semaphore! (of course, if persistence is setup in the redis instance)
• ✅ Very high performances with almost no polling at all
• ✅ Memory backend (for testing)

Non-features

• ❌ No time based rate limiters (yet)
• ❌ No blocking support, only async Python

Installation

pip install async-redis-rate-limiters

(or same with your favorite package manager)

Usage

import asyncio
from typing import AsyncContextManager
from async_redis_rate_limiters import DistributedSemaphoreManager


async def worker(semaphore: AsyncContextManager):
    async with semaphore:
        # concurrency limit enforced here
        pass


async def main():
    manager = DistributedSemaphoreManager(
        redis_url="redis://localhost:6379",
        redis_max_connections=100,
    )
    # Limit the concurrency to 10 concurrent tasks for the key "test"
    semaphore = manager.get_semaphore("test", 10)
    tasks = [asyncio.create_task(worker(semaphore)) for _ in range(1000)]
    await asyncio.gather(*tasks)


if __name__ == "__main__":
    asyncio.run(main())

What about if you want to use the memory backend?

WARNING: the memory backend is just a wrapper on a classic asyncio.Semaphore, it is not "distributed" at all!

manager = DistributedSemaphoreManager(
    backend = "memory"
)

# and use it classically

Dev

• Lint the code:

make lint

• Run the tests:

make test