To support the answer by Alexander Vogt, I'll generalize.
The intrinsic random_number(u)
returns a real number u
(or an array of such) from the uniform distribution over the interval [0,1). [That is, it includes 0 but not 1.]
To have a discrete uniform distribution on the integers {n, n+1, ..., m-1, m} carve the continuous distribution up into m+1-n equal sized chunks, mapping each chunk to an integer. One way could be:
call random_number(u)
j = n + FLOOR((m+1-n)*u) ! We want to choose one from m-n+1 integers
As you can see, for the initial question for {0, 1, 2, 3, 4, 5} this reduces to
call random_number(u)
j = FLOOR(6*u) ! n=0 and m=5
and for the other case in your comment {-1, 0, 1}
call random_number(u)
j = -1 + FLOOR(3*u) ! n=-1 and m=1
Of course, other transformations will be required for sets of non-contiguous integers, and one should pay attention to numerical issues.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…