mostly-small-nonempty-subset.clj

(defn mostly-small-nonempty-subset
  "Returns a subset of the given collection, with a logarithmically decreasing
  probability of selecting more elements. Always selects at least one element.

      (->> #(mostly-small-nonempty-subset [1 2 3 4 5])
           repeatedly
           (map count)
           (take 10000)
           frequencies
           sort)
      ; => ([1 3824] [2 2340] [3 1595] [4 1266] [5 975])"
  [xs]
  (-> xs
      count
      inc
      Math/log
      rand
      Math/exp
      long
      (take (shuffle xs))))