>That would double the function's runtime on average, without changing
>the results at all in most cases (it could only be different if the
>given operator has different semantics from the equality operator used
>while building the statistics list).
So I agree with Tom’s reasoning.