Skip to main content

darq

[ ~ Dependencies scanned by PyUp.io ~ ]

https://github.com/seedofjoy/darq/workflows/Lint%20&%20test/badge.svg?branch=master https://codecov.io/gh/seedofjoy/darq/branch/master/graph/badge.svg

Async task manager with Celery-like features. Fork of arq.

Features

  • Celery-like @task decorator, adds .delay() to enqueue job

  • Proper mypy type checking: all arguments passed to .delay() will be checked against the original function signature

  • Graceful shutdown: waits until running tasks are finished

Installation

Darq uses aioredis 1.x as Redis client. Unfortunately, this library has been abandoned, and does not support Python 3.11. I made a fork with compatability fixes: evo-aioredis (https://github.com/evo-company/aioredis-py).

Because of this, aioredis is not currently added as Darq dependency, and you must install it yourself:

For Python<3.11 you can use:

pip install aioredis<2.0.0

For Python 3.11 (and older versions too) you can use fork:

pip install evo-aioredis<2.0.0

Quick start

# some_project/darq_app.py
import asyncio
import darq

darq = darq.Darq(redis_settings=darq.RedisSettings(host='redis'))


@darq.task
async def add_to_42(a: int) -> int:
    return 42 + a


async def main():
    # Before adding tasks to queue we should connect darq instance to redis
    await darq.connect()

    # Direct call job as function:
    result = await add_to_42(5)  # result == 47

    # Celery-like add task to queue:
    await add_to_42.delay(a=5)

    await darq.disconnect()


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

And start worker:

python3 -m darq.cli -A some_project.darq_app.darq worker

Worker output:

15:24:42: Starting worker for 1 functions: some_project.darq_app.add_to_42
15:24:42: redis_version=5.0.7 mem_usage=834.87K clients_connected=1 db_keys=2
15:25:08:   0.22s → 1315f27608e9408392bf5d3310bca38c:darq_app.add_to_42(a=5)
15:25:08:   0.00s ← 1315f27608e9408392bf5d3310bca38c:darq_app.add_to_42 ● 47

Changelog

0.11.3 (2026-08-27)

  • Handle asyncio.CancelledError separately during job execution

  • Keep cancelled jobs queued for retry when retry_jobs is enabled

0.11.2 (2024-01-18)

  • Fix reconnecting after disconnect (thanks to @heimtathurs)

0.11.1 (2022-11-30)

  • Add Python 3.11 support (with evo-aioredis dependency instead of aioredis)

  • Remove pydantic dependency

  • Remove aioredis from dependencies to allow choose between aioredis and evo-aioredis - fork with Python 3.11 compatability

0.11.0 (2022-08-03)

  • Added ability to optionally pass ctx to the task, like this:

@task(with_ctx=True)
def foobar(ctx):
    log.info('Foobar try %s', ctx['job_try'])

ctx contains: job_id, job_try, enqueue_time, score, metadata + all worker’s ctx (including custom context which can be passed via on_startup). Thanks to @kindermax (https://github.com/seedofjoy/darq/pull/426) !

0.10.2 (2022-02-03)

  • Add proper typing for functions wrapped with the @task decorator. Mypy will now check that parameters are passed correctly when calling func() and func.delay()

0.10.1 (2021-07-29)

  • Add sentinel_timeout (defaults to 0.2) param to RedisSettings

0.10.0 (2021-07-09)

  • Breaking change: Rename darq.worker.Function to darq.worker.Task

  • Made job to task naming migration

  • Add max_jobs parameter to CLI (thanks to @antonmyronyuk)

  • Fixed bug with expires argument: default_job_expires could not be replaced with None in @task or .apply_async

0.9.0 (2020-06-24)

  • Breaking change: Add scheduler_ctx param to on_scheduler_startup and on_scheduler_shutdown to share data between this callbacks. It already has ctx['redis'] - instance of ArqRedis

0.8.0 (2020-06-22)

  • Breaking change: Changed CLI command format. Before: darq some_project.darq_app.darq. Now: darq -A some_project.darq_app.darq worker

  • Breaking change: Scheduler (cron jobs) now run’s seperate from worker (see darq scheduler command)

  • Breaking change: Changed some function signatures (rename arguments)

  • Breaking change: Remove redis_pool param from Darq app

  • Add on_scheduler_startup and on_scheduler_shutdown callbacks

0.7.2 (2020-06-18)

  • Fix some types (cron, OnJobPrepublishType)

  • on_job_prerun now runs before “task started” log and on_job_postrun now runs after “task finished” log

0.7.1 (2020-05-25)

  • .apply_async: Make args and kwargs arguments optional

0.7.0 (2020-05-25)

  • Fork arq to project and merge it with darq (It was easier to rewrite arq than to write a wrapper)

  • Breaking change: Remove “magic” params from .delay. For enqueue job with special params added .apply_async.

  • Add watch-mode to CLI.

  • Fix: Now worker will not run cronjob if it’s functions queue not match with worker’s

0.6.0 (2020-03-08)

  • Breaking change: Changed Darq constructor from single config param to separate params.

  • arq_function.coroutine now has .delay method.

0.5.0 (2020-03-03)

  • Add on_job_prepublish(metadata, arq_function, args, kwargs) callback. metadata is mutable dict, which will be available at ctx['metadata'].

0.4.0 (2020-03-03)

  • Add default_job_expires param to Darq (if the job still hasn’t started after this duration, do not run it). Default - 1 day

  • Add expires param to @task (if set - overwrites default_job_expires)

0.3.1 (2020-03-02)

  • Rewrite warm shutdown: now during warm shutdown cron is disabled, on second signal the warm shutdown will be canceled

0.3.0 (2020-02-27)

  • Breaking change: on_job_prerun and on_job_postrun now accepts arq.worker.Function instead of the original function (it can still be accessed at arq_function.coroutine)

0.2.1 (2020-02-26)

  • Fix add_cron_jobs method. Tests added.

0.2.0 (2020-02-26)

  • Add on_job_prerun(ctx, function, args, kwargs) and on_job_postrun(ctx, function, args, kwargs, result) callbacks.

0.1.0 (2020-02-26)

  • Breaking change: Jobs no longer explicitly get JobCtx as the first argument, as in 99.9% cases it doesn’t need it. In future release will be possible to optionally pass JobCtx in some way.

  • Breaking change: All cron jobs should be wrapped in @task decorator

  • Directly pass functions to arq.Worker, not names.

0.0.3 (2020-02-25)

  • .delay() now returns arq_redis.enqueue_job result (Optional[Job])

  • Add py.typed file

  • Fixed add_cron_jobs typing

0.0.2 (2020-02-24)

  • Add add_cron_jobs method

0.0.1 (2020-02-21)

First release

Metadata

Release files for darq 0.11.3

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for darq 0.11.3
File Size Uploaded Provenance
darq-0.11.3.tar.gz 40.6 kB Without provenance Details

Built distribution (wheel)

Table of built distributions (wheels) for darq 0.11.3
File Interpreter ABI Platform
darq-0.11.3-py3-none-any.whl Python 3 none any Details

Total release size: 73.1 kB

Release files / darq-0.11.3.tar.gz

Download URL darq-0.11.3.tar.gz
Size 40.6 kB
Tags Source
SHA-256 checksum
How to use checksums
979d888e026bc10443046ebf3d05987ba1b3c31eca49ecd695741239a4007b90
BLAKE2b-256 checksum
How to use checksums
8652bbca3db4cf098fa2b74fc0eade050afc2ec2ef137da4a24e8a4c0a726022
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.14.7

Release files / darq-0.11.3-py3-none-any.whl

Download URL darq-0.11.3-py3-none-any.whl
Size 32.6 kB
Tags Python 3
SHA-256 checksum
How to use checksums
42058d36918df3b6f8bfabf1237e0b67c68e585b0f8f94c10af6e53c73fcad6b
BLAKE2b-256 checksum
How to use checksums
7286b1b121f26a79324b678e11bc8c18ba5d069a51ba093ebb1267559575cb3d
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.14.7

Security for darq 0.11.3

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page