Throttler

Zero-dependency Python package for easy throttling with asyncio support.

🎒 Install

Just

pip install throttler

🧪 Development

uv sync --locked --all-extras --dev
uv run pytest -vl --cov=./ --cov-report=xml
uv run ruff check .
uv run ruff format .
uv run python -m build

📌 API Overview

This package exposes four context managers and six decorators:

• Throttler / throttle: rate-limits how often a block or async function can be entered.
• ThrottlerSimultaneous / throttle_simultaneous: caps how many async calls can run at once.
• ExecutionTimer / execution_timer: enforces a minimum period between entries by sleeping.
• Timer / timer: prints timing information, including average duration across iterations.

Throttler and ThrottlerSimultaneous are async context managers and must be used within an event loop. ExecutionTimer and Timer support both sync and async usage.

✅ Choosing the Right Tool

• Use Throttler when you need rate limiting (e.g., 10 requests per 1 second).
• Use ThrottlerSimultaneous when you need concurrency limiting (e.g., at most 5 in-flight tasks).
• Use ExecutionTimer when you want fixed spacing between iterations (e.g., run a job once per minute).
• Use Timer when you want timing output for repeated operations.

It is common to combine Throttler with ThrottlerSimultaneous for APIs that enforce both rate and concurrency.

🛠 Usage Examples

All run-ready examples are here.

Throttler and ThrottlerSimultaneous

Both are async context managers. Use them inside an async function or event loop.

Throttler:

Async context manager that limits how often the block can be entered.

from throttler import Throttler

# Limit to three calls per second
t = Throttler(rate_limit=3, period=1.0)
async with t:
    pass

Or

import asyncio

from throttler import throttle

# Limit to three calls per second
@throttle(rate_limit=3, period=1.0)
async def task():
    return await asyncio.sleep(0.1)

ThrottlerSimultaneous:

Async context manager that limits concurrent access to a block.

from throttler import ThrottlerSimultaneous

# Limit to five simultaneous calls
t = ThrottlerSimultaneous(count=5)
async with t:
    pass

Or

import asyncio

from throttler import throttle_simultaneous

# Limit to five simultaneous calls
@throttle_simultaneous(count=5)
async def task():
    return await asyncio.sleep(0.1)

Simple Example

import asyncio
import time

from throttler import throttle


# Limit to two calls per second
@throttle(rate_limit=2, period=1.0)
async def task():
    return await asyncio.sleep(0.1)


async def many_tasks(count: int):
    coros = [task() for _ in range(count)]
    for coro in asyncio.as_completed(coros):
        _ = await coro
        print(f'Timestamp: {time.time()}')

asyncio.run(many_tasks(10))

Result output:

Timestamp: 1585183394.8141203
Timestamp: 1585183394.8141203
Timestamp: 1585183395.830335
Timestamp: 1585183395.830335
Timestamp: 1585183396.8460555
Timestamp: 1585183396.8460555
...

API Example

import asyncio
import time

import aiohttp

from throttler import Throttler, ThrottlerSimultaneous


class SomeAPI:
    api_url = 'https://example.com'

    def __init__(self, throttler):
        self.throttler = throttler

    async def request(self, session: aiohttp.ClientSession):
        async with self.throttler:
            async with session.get(self.api_url) as resp:
                return resp

    async def many_requests(self, count: int):
        async with aiohttp.ClientSession() as session:
            coros = [self.request(session) for _ in range(count)]
            for coro in asyncio.as_completed(coros):
                response = await coro
                print(f'{int(time.time())} | Result: {response.status}')


async def run():
    # Throttler can be of any type
    t = ThrottlerSimultaneous(count=5)        # Five simultaneous requests
    t = Throttler(rate_limit=10, period=3.0)  # Ten requests in three seconds

    api = SomeAPI(t)
    await api.many_requests(100)

asyncio.run(run())

Result output:

1585182908 | Result: 200
1585182908 | Result: 200
1585182908 | Result: 200
1585182909 | Result: 200
1585182909 | Result: 200
1585182909 | Result: 200
1585182910 | Result: 200
1585182910 | Result: 200
1585182910 | Result: 200
...

ExecutionTimer

Context manager that enforces a minimum period between entries by sleeping. It is not a rate limiter like Throttler. With align_sleep=True, it aligns to wall-clock boundaries (e.g. each minute).

import time

from throttler import ExecutionTimer

et = ExecutionTimer(60, align_sleep=True)

while True:
    with et:
        print(time.asctime(), '|', time.time())

Or

import time

from throttler import execution_timer

@execution_timer(60, align_sleep=True)
def f():
    print(time.asctime(), '|', time.time())

while True:
    f()

Result output:

Thu Mar 26 00:56:17 2020 | 1585173377.1203406
Thu Mar 26 00:57:00 2020 | 1585173420.0006166
Thu Mar 26 00:58:00 2020 | 1585173480.002517
Thu Mar 26 00:59:00 2020 | 1585173540.001494

Timer

Context manager for pretty printing start, end, elapsed and average times. Elapsed timing uses a monotonic clock, while start/end timestamps are wall-clock time.

import random
import time

from throttler import Timer

timer = Timer('My Timer', verbose=True)

for _ in range(3):
    with timer:
        time.sleep(random.random())

Or

import random
import time

from throttler import timer

@timer('My Timer', verbose=True)
def f():
    time.sleep(random.random())

for _ in range(3):
    f()

Result output:

#1 | My Timer | begin: 2020-03-26 01:46:07.648661
#1 | My Timer |   end: 2020-03-26 01:46:08.382135, elapsed: 0.73 sec, average: 0.73 sec
#2 | My Timer | begin: 2020-03-26 01:46:08.382135
#2 | My Timer |   end: 2020-03-26 01:46:08.599919, elapsed: 0.22 sec, average: 0.48 sec
#3 | My Timer | begin: 2020-03-26 01:46:08.599919
#3 | My Timer |   end: 2020-03-26 01:46:09.083370, elapsed: 0.48 sec, average: 0.48 sec

🧠 Behavior Notes

• Throttler uses a sliding window. Each entry may sleep until the oldest recorded entry exits the time window.
• ThrottlerSimultaneous uses an async semaphore under the hood.
• ExecutionTimer uses a monotonic clock for waiting; when align_sleep=True it aligns to wall time.
• Timer prints elapsed duration per entry and the running average across all entries.

⚠️ Error Handling

• Throttler(rate_limit, period) requires a positive integer rate and a positive float period.
• ThrottlerSimultaneous(count) requires a positive integer count.

These validations raise ValueError early so invalid configuration fails fast.

💬 Contributing

Contributions, issues and feature requests are welcome!

📝 License

This project is MIT licensed.