Timeout function decorator

Timeout decorator for synchronous and asynchronous functions.

Dependencies

Written in pure Python and has no dependencies other than the base libraries.

Installation

From source code:

pip install .

From PyPI:

pip install timeout-function-decorator

Import

There are three ways to import the decorator:

1. Import directly from the package

from timeout_function_decorator import timeout

2. Import directly from the module

from timeout_function_decorator.timeout_decorator import timeout

3. Import the module

from timeout_function_decorator import timeout_decorator

This last case is useful if a timeout function is already present in your namespace.

Usage

Using the decorator is as simple as:

import time
from timeout_function_decorator import timeout


@timeout()
def i_will_never_time_out(value):
    while True:
        time.sleep(1)

        
@timeout(None)
def i_will_never_time_out(value):
    while True:
        time.sleep(1)
        

@timeout(1)
def i_will_not_time_out(value):
    return value


@timeout(1)
def i_will_time_out(value):
    time.sleep(2)
    return value


@timeout(1, RuntimeError)
def i_will_raise_runtime_error(value):
    time.sleep(2)
    return value

As you may have noticed, the decorator requires the brackets even when no parameters are passed.

The same result could be obtained for asynchronous functions:

import asyncio
from timeout_function_decorator import timeout


@timeout()
async def i_will_never_time_out(value):
    while True:
        await asyncio.sleep(1)

        
@timeout(None)
async def i_will_never_time_out(value):
    while True:
        await asyncio.sleep(1)
        

@timeout(1)
async def i_will_not_time_out(value):
    return value


@timeout(1)
async def i_will_time_out(value):
    await asyncio.sleep(2)
    return value


@timeout(1, RuntimeError)
async def i_will_raise_runtime_error(value):
    await asyncio.sleep(2)
    return value

If you already have a timeout function in your namespace, you could as easily use the decorator with the module namespace:

import time
from timeout_function_decorator import timeout_decorator


@timeout_decorator.timeout(1)
def i_still_time_out(value):
    time.sleep(2)
    return value

:warning: Warning: When a function times out, an exception is raised but cancellation is not guaranteed. This decorator only notifies the user when enough time has passed since a function call. Handling of the situation and ensuring cancellation is up to the user.

Signature

In general, the decorator accepts two parameters:

• timeout_duration: a float specifying the timeout time in seconds. If None, the function will not time out. Defaults to None
• exception_to_raise: the Exception type to be raised. Defaults to TimeoutError.

They can be passed as positional or keyword arguments.

Intended use cases

The wrapper uses asyncio and threads to keep track of timeouts. This adds a non-trivial overhead on the wrapped functions, in particular on the synchronous ones as they need a thread to be created.

As such, common use cases would include test suites where control on what happens when they time out is important, as with packages like pytest-timeout you can not have tests fail with a specified exception.

Use cases in production code, on the other hand, would be limited to long functions with a low call rate as the overhead is linear with the number of calls but independent of the size of the function or its execution time.

A notebook outlining the impact of the overhead on sync and async functions can be found here.

Plots from the notebook are presented below for a quick performance evaluation.

Performance evaluation

General performance

Sync performance

Quick function performance

Medium function performance

Long function performance

Huge function performance

Async performance

Quick function performance

Medium function performance

Long function performance

Huge function performance