async-timeout

async-timeout provides an async with timeout(delay): context manager for asyncio code, cancelling the wrapped block if it runs longer than the given delay and raising TimeoutError outside the block. It predates and directly inspired Python 3.11’s built-in asyncio.timeout().

The project is now explicitly deprecated by its own maintainers: as of Python 3.11, its functionality was upstreamed into the standard library, and the repository has moved under the aio-libs-abandoned GitHub organization and is archived. It remains widely installed as a transitive dependency of packages that still support Python versions older than 3.11, and version 5.0+ provides a compatibility shim that dual-mode delegates to asyncio.Timeout on 3.11+.

What You Get

• timeout(delay) - async context manager that cancels its block after delay seconds, faster than asyncio.wait_for() since it doesn’t spawn a new task
• timeout_at(when) - schedule a timeout at an absolute event-loop time instead of a relative delay
• .expired() / .when() - inspect whether the block was cancelled by the timeout and what its deadline was
• .reschedule() / .shift() - adjust or cancel a pending timeout while still inside the context manager
• Python 3.11+ compatibility shim that delegates to stdlib asyncio.Timeout under the hood

Common Use Cases

• Bounding the runtime of an asyncio operation (network call, DB query) without the task-creation overhead of asyncio.wait_for()
• Supporting Python versions older than 3.11 with the same timeout semantics that later became stdlib
• Legacy dependency chain: many older asyncio libraries (older aiohttp/redis-py versions) still pin this as a transitive dependency
• Migration bridge: codebases upgrading to Python 3.11+ can swap the import for stdlib asyncio.timeout with no API changes

Under The Hood

Architecture - The entire implementation lives in a single 276-line module, async_timeout/__init__.py, defining a Timeout class that wraps loop.call_at() to schedule a cancellation callback and exposes it via two factory functions, timeout() and timeout_at(); the callback throws asyncio.CancelledError into the current task, which the context manager’s __aexit__ intercepts and converts to TimeoutError if triggered by expiry rather than user code. Tech Stack - Pure Python with zero runtime dependencies, targeting the asyncio event loop API directly (loop.time(), loop.call_at()); packaged with a standard setup.py/pyproject.toml and tested via a single tests/test_timeout.py file. Code Quality - Small, single-file, single-purpose codebase with focused test coverage of expiry, rescheduling, and nested-timeout edge cases; CI badges (GitHub Actions, Codecov) in the README indicate coverage was tracked, though the project’s own README now states it is deprecated and no longer actively supported. API Design - The API is minimal by design — one context manager, one absolute-time variant, and a handful of inspection/reschedule methods — which made it easy to port verbatim into Python 3.11’s stdlib asyncio.timeout; new projects on 3.11+ are explicitly directed by the README to prefer the stdlib version and use this package only for backward-compatible dual support.