question

Although the uniform() command documentation claims to be inclusive, if you actually test it, it isn't. It might theoretically be inclusive, but there are theoretically an infinite number of values between any two decimal values so the probably of actually being the extremes is infinitesimally small (roughly 1 in 2^53 for a double-precision value) such that in actual practice, I don't think it ever returns the extremes.

So you can use use the uniform() distribution to get an exclusive distribution of floating point values, as the example code below shows:

int extremes = 0;
for (int i = 1; i <= 10000000; i++) {
	double num = uniform(0,100);
	if (num == 0 || num == 100) {
		extremes++; // this code will never get hit
	}
}
return extremes; // this will always return 0

If you are using duniform(), then you can add 1 to the beginning or subtract 1 from the end to make it exclusive.

The random values uniform() gives back are unlikely to ever actually equal any specific value in the range, including the extremes, unless that value was explicitly obtained using that pseudo-random number generation and then repeated.