It's a very useful distribution in Bayesian statistics, for various reasons. It can be sampled from efficiently and all of its moments can be evaluated in closed form. So that gives a lot of functions…
The [HDP](http://www.cs.berkeley.edu/~jordan/papers/hdp.pdf) paper contains some discussion (below) of posterior sampling the Dirichlet process concentration parameters (`\alpha_0`/`\gamma` or `\dish_…