Comment by kortex

If you are willing to accept lookup tables or approximations, then yeah, arbitrary distributions are trivial. However for certain domains, you may want a more closed form solution for mapping uniform to your distribution, which may not be obvious. E.g. I would say generating pink noise is not "trivial" because there is no closed form solution (c.f. generating gaussian distribution via Box-Muller, that is trivial), so you need to pick a method which may have tradeoffs.