sublist generator for clojure test.check... which has better shrinking properites?

sublist.clj

;; my first version taking advantage of the new shuffle generator introduced in 0.6...
(defn sublist1
  [coll]
  (-> (gen/tuple (gen/shuffle coll) (gen/choose 0 (count coll)))
      (gen/bind (fn [[coll* how-many]]
                  (gen/return (take how-many coll*))))))

;; version inspired by http://roberto-aloi.com/erlang/notes-on-erlang-quickcheck/
;; sublist(L0) ->
;;   ?LET(L, [{E, eqc_gen:bool()} || E <- L0], [X || {X, true} <- L]).
(defn sublist2 [coll]
  (-> (gen/vector gen/boolean 0 (count coll))
      (gen/bind (fn [bools]
                  (->> bools
                       (keep-indexed (fn [i b] (when b (coll i))))
                       gen/return)))))

(gen/sample (g/sublist2 [:a :b :c :d :e :f :g :h :i :j :k])) ; => ((:a :b) (:a) () (:b) () (:a :b :c :d :f :j) (:a :c) (:c) (:a :f :g) (:a :d))

I'm inclined to think that sublist2 would be better since it does less work but I haven't dug into how shrinking actually works so that is a total guess.

sublist2 should shrink better, yes. You can simplify it's definition a little bit too, since your gen/bind call ends with gen/return, you can simply use gen/fmap and remove the call to gen/return.

Thanks for the feedback @reiddraper!

Are you suggesting that I use fmap in place of bind too (since fmap takes a generator)? I don't see how to do that since I also need an index for collection lookup as well. So I would either need a fmap-indexed or fmap be variadic so I could zip an index in. What am I missing?

Words are hard, code is easier:

(defn sublist [coll]
  (gen/fmap
    (fn [bools] (keep-indexed (fn [i b] (when b (coll i))) bools))
    (gen/vector gen/boolean (count coll))))