type should return something else than PersistentMap for defrecord?

[borkdude]

(defrecord Foo []) (type (Foo.))

In Clojure a Class is returned:

user=> (defrecord Foo [])
user.Foo
user=> (type (type (Foo.)))
java.lang.Class

In ClojureScript a function is returned:

cljs.user=> (defrecord Foo [])
cljs.user/Foo
cljs.user=> (type (type (Foo.)))
#object[Function]

I'm not sure what to return in sci, because it does not define classes like Clojure.

[mk]

For our use case, it would be sufficient, if we had some way, of getting to the fully-qualified class name.

In Clojure this works:

(ns user)
(defrecord R [foo])
(-> "bar" ->R type .getCanonicalName) ;; => "user.R" 

In sci:

(ns user)
(defrecord R [foo])
(-> "bar" ->R meta :sci.impl/type str) ;; => "R"

[mk]

I've added a failing test in #493 but haven't been able to get it working yet.

[borkdude]

@mk Thanks. I'll have a look during the weekend, hopefully.

[borkdude]

@mk How would you tackle this for normal CLJS?

[borkdude]

@mk In branch record-type, commit a40d4a7, this works:

$ lein run -m sci.impl.main "(ns foo) (defrecord Foo []) (type (Foo.))"
foo/Foo

Note that type returns a symbol here and not a class, like in Clojure. Also it's namespaced, not dotted. Not sure if that matters?

[borkdude]

@mk I found in clara rules how they solve this for CLJS:

(ns foo)

(defrecord Foo [])

(derive Foo ::system-type)

(isa? (type (Foo.)) ::system-type)

I think that could be a useful way to not rely on the specifics of the return value of type but still get the information you need out of it. I wrote a test for this in sci now (in the branch).

[borkdude]

@mk With the changes just merged to master it should now be possible to do what you want.