Fastest way to determine if a DataFrame's length exceeds a threshold

Question

count() is expensive, and countApprox() is no faster; maybe I am using it incorrectly. Does anyone have a better idea?