Sure, good point; I generally think of evaluation, not stringification, as the likely dominant expense. But maybe it’s sometimes the other way around?