This is a more systematic solution than our previous homemade logging mechanism. I was worrying that the overhead might be too large, but the performance impact is only on a per-callsite basis. Even adding a few spans in the hottest code in fp doesn't measurably affect runtime.
This is a more systematic solution than our previous homemade logging mechanism. I was worrying that the overhead might be too large, but the performance impact is only on a per-callsite basis. Even adding a few spans in the hottest code in
fpdoesn't measurably affect runtime.