Is it expected that compiler’s env doesn’t include an anonymous function in :fn-scope? Here :fn-scope is only going to have fn-1

(defn fn-1 []
  (fn [a]
    a))

:fn-scope is about shadowing

shadowing isn't a problem with anonymous fns

@roman01la concrete example (fn console [] ...)

w/o :fn-scope this will create a local named fn with the exact JS name console

which breaks expectations

fwiw it's sometimes useful to use the git history to see what test was associated w/ a change - the test hopefully makes it clear what was broken

definitely true in this case

to clarify further, we use each entry in :fn-scope to produce a qualified name that can be used for recursion

something like fn_1$console

I’m trying to detect if a local var is an argument of the top most function, not sure what’s the general algorithm here?

we could of course generate random names - but the above has the added benefit of aligning w/ something you see in Clojure as well as helping w/ debugging

@roman01la I wouldn't mess w/ any of this

or attempt to re-use anything in ClojureScript

beyond calling analyze, just walk the AST yourself and sort it out

yes, that’s what I’m doing

then I don't see why you care about :fn-scope

should just be able to look at bindings and args

Ok, perhaps I missed something, I have to look more into it. I’ll get back if I need more help. Thanks!

as you walk down the AST w/ the original fns args, just check to see if you got shadowed via a new binding or fn arg

is it normal to get errors when using ^boolean type hints in cljc code? eg

Error in phase :compile-syntax-check
IllegalArgumentException: Unable to resolve classname: clojure.core$boolean@6fc1e52d

I don’t get an error if I type something like (defn ^boolean x []) in a clojure repl

@mhuebert the return value tag is placed incorrectly. so likely it just isn't interpreted? this fails

(defn x ^boolean [])
CompilerException java.lang.IllegalArgumentException: Only long and double primitives are supported, 

https://github.com/Day8/re-frame/blob/master/src/re_frame/trace.cljc#L20

an example. i thought the typehint went on the var

pretty sure for defn the hint goes on the args vector (for :ret-tag)

hmm

@mhuebert pretty sure ^boolean won't work in Clojure

so you need to conditionalize

also I'm sure you're doing something very performance sensitive to bother here at all

yeah it’s in a tight loop where the ^boolean in cljs makes a big difference

k, for some history ^boolean was introduced when we started to doing persistent datastructures

wasn't really important for anything else

@mhuebert It is also possible in many cases for the compiler to infer that a function returns a Boolean value. This depends on the body of the function.

This raises an interesting question on inference / hinting around goog-define values

I think I considered a flag for :warn-on-cljs-truth but I think that was before file-scoped dynamic vars

was too noisy, could revisit for people doing perf work

(Sorry about the goog-define noise in my previous comment; thought that was the code in question.)

could it infer from the last fn passed to comp?

no

there is no higher order inference

ok. After using the analyzer to do a bit of digging, hints were propagating up from function bodies more than I realized

right a lot code users write isn't higher-order (this isn't Haskell afterall), so the surface area that our simple inference covers is surprisingly effective

and lazy seq ops return seq or nil, so even that's covered

but we can't do too much about partial, comp, apply etc. right now - but I think not really a big deal, and those things would require a lot more effort

Hello cljs-dev! I am interested in seeing this one addressed: https://dev.clojure.org/jira/browse/CLJS-3074 - do you think it would be a good first issue for me to tackle?

@lee absolutely!

doesn't require getting too deep into compiler know-how - but enough to get familiar with the process

make sure to read over this https://clojurescript.org/community/contributing

I am perusing it now, thanks! Have already submitted my Contributor Agreement.