but as I expected, you lose conform destructuring and explain can't explain the problem anymore

...so different solutions would be welcome

Clojurescript is failing to compile my file if my fdefs are located above their corresponding defns. Is this the way things are intended?

I.E., you can't spec your function before you define it?

@george.w.singer: Yes. The var is not defined yet. You could declare the var if you must write the spec above the function. I have just been writing my specs below my function definitions.

Ok, thanks.

Macro sugar needs to be added to spec IMO

Something like

(fn-spec
 fn-name ::arg1, ::arg2 -> ::ret-spec 
                             :such-that    ::ret-value-greater-than-max-of-its-inputs)

@george.w.singer: you could write that macro 🙂

yea 😄

@george.w.singer: I think that’s a CLJS-impl bug, I’d report it

in CLJ, you can (s/fdef new-fn) ahead of the var

@george.w.singer: Does it work if you do

(s/fdef 'your-sym
        :args ...
        :ret ...)

It does sometimes but then the whole thing becomes buggy. Sometimes it doesn't compile. All tests pass

Really strange behavior is exhibited

The strange behavior goes away if put them before

I seem to be getting a strange error when creating a spec for a function with a type hinted argument. Error:

java.lang.ClassCastException: clojure.spec$spec_checking_fn$fn__11414 cannot be cast to clojure.lang.IFn$DO

(defn bar
  [f]
  (f 0.0))

(s/fdef bar
        :args (s/cat :f fn?)
        :ret number?)

(defn foo
  [^double value]
  (bar (fn [_] value)))

(s/fdef foo
        :args (s/cat :v number?)
        :ret number?)

(defn foo2
  [value]
  (bar (fn [_] value)))

(s/fdef foo2
        :args (s/cat :v number?)
        :ret number?)

(foo 1.0) => error
(foo2 1.0) => 1.0

(s/instrument-all)

Problem exists in 1.9.0-alpha4.

Looks like primitive hinting encodes the argument and return type to be encoded in the underlying type (if you add ^double as a return type hint, the cast is to IFn$DD, for example).

Very surprised to see that calling an instrumented function can actually load and call clojure.test.check to do generative testing as part of the conformance test, without explicitly calling clojure.spec.test/run-all-tests etc.

http://dev.clojure.org/jira/browse/CLJ-1936

is this a bug?

spec> (gen/sample (s/gen #{1 0}) 10)
(0 0 1 0 0 0 1 0 1 0)
spec> (gen/sample (s/gen #{false true}) 10)
(true true true true true true true true true true)

s/gen takes a spec, which is a predicate

#{false true} will only return true, as a predicate, for true

@hiredman: thanks for your explanation, but I still don't get how I would generate true/false randomly

I’ve defined these specs and vars

(s/def ::point (s/cat :x number? :y number?))
(s/def ::body (s/+ ::point))
(s/conform ::point [2 3])
(s/conform ::body [[2 3]])
(s/explain ::body [[2 3]])

::point conforms nicely, but ::body does not conform, and the explanation is:

In: [0] val: [2 3] fails spec: :snake-game.utils/point at: [:x] predicate: number?

I would have thought that (s/+ ::point) would mean one or more points?

@slipset: (s/def ::point (s/spec (s/cat :x number? :y number?)))

The combination of s/+ and s/cat means that body is defined as a series of :xs and :ys without nesting

thanks 🙂

for the record, I ended up with this for booleans:

(defn- boolean? [x] (instance? Boolean x))
(s/def ::boolean (s/with-gen boolean? #(gen/boolean)))

is there any way to make s/cat specs generate a vector?

What @stathissideris said :)

This is a pretty minor note, but this appears to be one of the only channels with _ separation instead of -. Any chance an admin could change that?

and as we all know, clojure names should prefer - to _

I see no stinkin’ underscores here...