It would be good if benchmark library requested garbage collection before or after each test. In our case, we generate random strings in our text related tests, and after a while the tests accumulate many big strings in memory which impacts tests that follow and might be causing discrepancy in numbers.
Description
In our case, we generate random strings in our text related tests, and after a while the tests accumulate many big strings in memory which impacts tests that follow and might be causing discrepancy in numbers.