RANDP - pseudorandom integers from a specified discrete distribution

R = randp(P, N) returns an N-by-N matrix containing pseudorandom
integers drawn from a specified discrete distribution on 1:numel(P).
The distribution is specified by the relative values of P so that a
value K is present approximately "P(K)/sum(P) times in the matrix R.
All values of P should => 0, NaNs are set to 0.

The other arguments to randp specify the size of R in the same way as
matlab's own functions do: randp(P, N) returns an N-by-N matrix,
randp(P,M,N) and randp(P, [M N]) return M-by-N arrays, etc.

Examples:
% random values from [1 2 4] and a bias for 2
R = randp([1 2 0 1], 1, 100) ; % 100 values
histc(R, 1:4) % -> ~25 ~50 0 ~25

% create a random, but biased DNA sequence
C ='AGCT', P = [4 1 1 2]
DNA = C(randp(P, 1, 50))