The article is correct. double(pi) is exactly 884279719003555/281474976710656, which is a perfectly valid number to apply the sin function to. sin(884279719003555/281474976710656) is a transcendental number close to 1.2246467991473532e-16. In fact, that is the closest 64-bit float value to the true value of sin(884279719003555/281474976710656) and thus the best answer that the sin function can give. There is no way to ask the sin(double x) function what the sine of the transcendental number π is since the argument must be a double and there's no double value for π. Since the sin function is not in the business of guessing which of the uncountably many real numbers you really wanted to take the sine of, it can only give you the best answer for the value you actually passed it.