GitHub | PyPI | Issues | Changelog

pytest-fail-slow is a pytest plugin for treating tests as failed if they took too long to run. It adds markers for failing tests if they or their setup stages run for longer than a given duration, along with command-line options for applying the same cutoff to all tests.

Note that slow tests will still be run to completion; if you want them to instead be stopped early, use pytest-timeout.

Installation

pytest-fail-slow requires Python 3.8 or higher and pytest 7.0 or higher. Just use pip for Python 3 (You have pip, right?) to install it:

python3 -m pip install pytest-fail-slow

Usage

Failing Slow Tests

To cause a specific test to fail if it takes too long to run, apply the fail_slow marker to it, with the desired cutoff time as the argument:

import pytest

@pytest.mark.fail_slow("5s")
def test_something_sluggish():
    ...

If a test fails due to being slow, pytest's output will include the test's duration and the duration threshold, like so:

________________________________ test_func ________________________________
Test passed but took too long to run: Duration 123.0s > 5.0s

(New in version 0.6.0) If you only want a given test to fail for being slow under certain conditions — say, when running under CI or on a certain platform — supply the enabled keyword argument to the marker. The value of enabled can be either a boolean expression or a condition string. When the enabled value evaluates to True (the default), the test will fail if its runtime exceeds the given duration; if it evaluates to False, a lengthy runtime will not cause the test to fail. Example usage:

import os
import pytest

@pytest.mark.fail_slow("5s", enabled="CI" in os.environ)
def test_something_that_needs_to_be_fast_in_ci():
    ...

An an alternative or in addition to the marker, the --fail-slow DURATION option can be passed to the pytest command to, in essence, apply the fail_slow marker with the given cutoff to all tests that don't already have the marker. (As far as pytest is concerned, the option does not actually cause any markers to be added to any tests, in case your code cares about that.) If a test already has the fail_slow marker, the --fail-slow option will have no effect on it.

If you want a test to fail for being slow only if the --fail-slow option is given, but you also want a different cutoff for the test than that passed to the option, you can give the test a fail_slow marker that sets the desired cutoff and also sets enabled to a condition string that checks whether --fail-slow has been given, like so:

import pytest

@pytest.mark.fail_slow(3, enabled="config.getoption('--fail-slow') is not None")
def test_something_sometimes_sluggish():
    ...

Note: This feature only takes the durations for tests themselves into consideration. If a test passes in less than the specified duration, but one or more fixture setups/teardowns take longer than the duration, the test will still be marked as passing. To fail a test if the setup takes too long, see below.

Failing Slow Setups

New in version 0.4.0

To cause a specific test to fail if the setup steps for all of its fixtures combined take too long to run, apply the fail_slow_setup marker to it, with the desired cutoff time as the argument:

import pytest

@pytest.mark.fail_slow_setup("5s")
def test_costly_resource(slow_to_create):
    ...

Do not apply the marker to the test's fixtures; markers have no effect on fixtures.

If the setup for a test takes too long to run, the test will be marked as "errored," the test itself will not be run, and pytest's output will include the setup stage's duration and the duration threshold, like so:

_______________________ ERROR at setup of test_func _______________________
Setup passed but took too long to run: Duration 123.0s > 5.0s

Like fail_slow, the fail_slow_setup marker takes an optional enabled keyword argument that can be used to conditionally enable or disable failure for slow setups. There is also a --fail-slow-setup DURATION option that can be passed to pytest to, in essence, apply the marker to all tests that don't already have it.

Note: If a test depends on multiple fixtures and just one of them exceeds the given duration on its own, the remaining fixtures will still have their setup steps run. Also, all fixture teardowns will still be run after the test would have run.

Specifying Durations

A duration passed to a marker or command-line option can be either a bare number of seconds or else a floating-point number followed by one of the following units (case insensitive):

• h, hour, hours
• m, min, mins, minute, minutes
• s, sec, secs, second, seconds
• ms, milli, millisec, milliseconds
• us, μs, micro, microsec, microseconds