The method Matthias suggests would have the problem that both 1 and 100 have half the probability of the other numbers appearing. (The float numbers are equally probable between 1 and 100; but only 1.0 to 1.5 will give 1, then 1.5 to 2.5 gives 2, ... and finally 99.5 to 100.0 gives 100)
I would multiply by the random number by 100.0, giving 0.0 to 100.0, (excluding 0 and 100) then round to + infinity, ie the nearest integer above, giving 1 to 100.
You might possibly want to check that rounding errors on the multiply haven't resulted in the rounding up giving 101.
Rod.
