Why is Python's IsolatedAsyncioTestCase so slow? - Stack Overflow

I'm working on writing test cases for some Python code that uses asyncio. I'm noticing a significant performance degradation when my test classes inherit from unittest.IsolatedAsyncioTestCase vs. using the non-async UnitTest and manually calling a coroutine using asyncio.run

I've tested this on Python 3.13.1 and 3.10.10 on my personal desktop (Mac Studio M1 Max) and on a Linode Nanode VPS. I'm noticing around 10x slowdowns across testing platforms.

Below is a simplified example, my actual use case is far more complicated.

# test.py
import asyncio

from unittest import TestCase
from unittest import IsolatedAsyncioTestCase

async def f():
    for i in range(100000):
        await asyncio.sleep(0)

class TestManualAsync(TestCase):
    def test_async(self):
        asyncio.run(f())

class TestIsolatedAsync(IsolatedAsyncioTestCase):
    async def test_async(self):
        await f()


$ python3 -m unittest test.py -k TestManualAsync
.
----------------------------------------------------------------------
Ran 1 test in 0.614s

$ python3 -m unittest test.py -k TestIsolatedAsync
.
----------------------------------------------------------------------
Ran 1 test in 5.115s

You can see from the above it's much faster to just trigger a coroutine in a traditional test class. I don't see why this should be the case.

I'm working on writing test cases for some Python code that uses asyncio. I'm noticing a significant performance degradation when my test classes inherit from unittest.IsolatedAsyncioTestCase vs. using the non-async UnitTest and manually calling a coroutine using asyncio.run

I've tested this on Python 3.13.1 and 3.10.10 on my personal desktop (Mac Studio M1 Max) and on a Linode Nanode VPS. I'm noticing around 10x slowdowns across testing platforms.

Below is a simplified example, my actual use case is far more complicated.

# test.py
import asyncio

from unittest import TestCase
from unittest import IsolatedAsyncioTestCase

async def f():
    for i in range(100000):
        await asyncio.sleep(0)

class TestManualAsync(TestCase):
    def test_async(self):
        asyncio.run(f())

class TestIsolatedAsync(IsolatedAsyncioTestCase):
    async def test_async(self):
        await f()


$ python3 -m unittest test.py -k TestManualAsync
.
----------------------------------------------------------------------
Ran 1 test in 0.614s

$ python3 -m unittest test.py -k TestIsolatedAsync
.
----------------------------------------------------------------------
Ran 1 test in 5.115s

You can see from the above it's much faster to just trigger a coroutine in a traditional test class. I don't see why this should be the case.

• python
• python-3.x
• python-asyncio
• python-unittest

Share Improve this question Follow edited Jan 2 at 15:05 TylerH 21.1k7979 gold badges7979 silver badges114114 bronze badges asked Jan 2 at 14:56 MikeMike 61k7878 gold badges183183 silver badges228228 bronze badges

Add a comment  | 

1 Answer 1

Sorted by: Reset to default 3

It's just because it runs the event loop in debug mode:

runner = asyncio.Runner(debug=True, loop_factory=self.loop_factory)

which adds a bunch of overhead. Try using asyncio.run(f(), debug=True) in TestManualAsync, and it'll take about as much time as TestIsolatedAsync does.