I have some tests with assertions in loop, each assertion is kinda like separate test and I don't want previous assertions' output to pollute error logs of current failed assertion.
def test_foos(captured): foos = [] # some data for foo, bar in foos: captured.clear() assert logic(foo) == bar I have found caplog.clear but it doesn't seem to work.
Parametrize your test. Pass the foos as parameter and pytest will run the test assert line multiple times, recording success/failure as if each were a separate test.
import pytest testdata = [ (3,9), (4,16), (5,25) ] @pytest.mark.parametrize("x,expected", testdata) def test_foos(x, expected): assert logic(foo) == bar # in the example, say logic squares the input 
I agree with the other answer that in your case, using pytest.mark.parametrize is the best approach. However, for completeness, you are able to clear captured output with the capsys fixture.
Example:
def logic(letter): print("Handling letter", letter) return letter.isupper() def test_foo(): letters = ["A", "b", "c"] for letter in letters: assert logic(letter) The test will fail at "b" and print:
----------------------------- Captured stdout call ----------------------------- Handling letter A Handling letter b To prevent the output from (successfully) handling "A", we can add capsys to the arguments and use capsys.readouterr to clear the buffer in between:
def test_foo(capsys): letters = ["A", "b", "c"] for letter in letters: assert logic(letter) capsys.readouterr() Now, the tests still fail at "b", but only print:
----------------------------- Captured stdout call ----------------------------- Handling letter b
pytestwill finish test execution after first failed assertion so you will see only one error log.logiccalls because they are in the same scope as failed assertlogiccalls... It wasn't clear. Glad that you got an answer :)