Clojurians Log v2

> Can't say I understand the question to be honest... Do you have an example? -- @pavlos Something like this (where gen and prop come from test.check), but with a lot more possible transformations, and ideally able to provide a meaningful selection of functions for any value generated by gen/any.

(def int-fns [inc
              (partial * 3)
              str])

(defn gen-transform [x]
  (cond (int? x) (gen/one-of (map gen/return int-fns))
        :else :not-implemented))

(gen/sample (gen-transform 2))

(defspec foo-equiv-to-bar
  (prop/for-all [x gen/small-integer]
    (gen/let [f (gen-transform x)]
      (= (foo f x) (bar f x)))))

This sounds like a solution to a problem. Do you have a goal or is the question philosophical in a sense? 🙂 (Both are fine but you might get better answers for a more specific problem) Perhaps you're trying to do a kind of coercion, in which case https://github.com/exoscale/coax might help

I am trying to test higher order functions, including such things as this functor law.

fmap (f . g)  ==  fmap f . fmap g

This law says that for any function f and any function g, mapping f over the functor and g over the result is the same as mapping (comp f g) over the functor. And I would like to generate a reasonable and varied set of functions to represent "any function" in testing that.

I see... That's an interesting puzzle 🙂 I don't think I can help, sorry... I can't imagine how you'd pick a compatible function on the fly based on randomly generated data without having a curated set of functions + metadata specs

If you have a set of functions and they are specced, you could use their input specs to generate the data after you've picked the function

But I suspect that won't cut it for you

Yeah, I think v1 would be a collection of curated functions mapped to acceptable input types, and randomly picking from matches, very similar to my example but lots more code. But Clojure and test.check definitely supports going a lot farther than that.

Reading through this additional context, it sounds like exactly the type of property that Typed Clojure could help verify; it has the ability to instrument codebases without requiring those codebases to depend on Typed Clojure. https://github.com/typedclojure/typedclojure/tree/main/example-projects/zero-deps

Interesting direction to go. 🤔