Do you have that many private vars that you need to suppress a whole lot of them? :)

Yes! :)

Tell me more, I want to see them :)

but seriously, I think that would be fine

issue / PR welcome

haha, it's not easy to convince you 🙂

I use some general patterns for private functions to later generate public functions

The way I usually solve this is move those to an foo.internal namespace and make them public

hmmm... it's possible in this case.

actually I can make them public

anyway, I think that every linter which can have :exclude could have :exclude-patterns as well, what do you think?

it will enable a lot of possiblities and simplification in some cases

yes, I'm not opposed to it

I'll leave the issue then

@U1EP3BZ3Q Note that you can also disable the unused-private-var linter in one specific namespace only

Yes, I know it. I usually prefer to narrow exclusion as much as possible. I solved this particular case by making these functions public.

Do you have an example?

The example from ClojureDocs (https://clojuredocs.org/clojure.core/amap)

(def an-array (int-array 25000 (int 0)))

(time (amap ^ints an-array 
            idx 
            ret 
            (+ (int 1) 
               (aget ^ints an-array idx))))

clj-kondo v2023.04.14

yes, underscore seems fine there if ret is not used?

or am I missing something?

ret is internal name used by amap which is a macro. It expands as:

(let*
  [a__6552__auto__
   an-array
   l__6553__auto__
   (alength a__6552__auto__)
   ret
   (aclone a__6552__auto__)]
  (loop*
    [idx 0]
    (if (< idx l__6553__auto__)
      (do
        (aset ret idx (+ (int 1) (aget an-array idx)))
        (recur (unchecked-inc idx)))
      ret)))

In 99% you do not use this name.

why does amap make you choose a name for this?

It doesn't. But this is stated in documentation >

Maps an expression across an array a, using an index named idx, and
> return value named ret, initialized to a clone of a, then setting 
> each element of ret to the evaluation of expr, returning the new 
> array ret.

So probably everyone just leave it as ret

you can also name it _

Of course.

I'm fine with suppressing that lint warning

it raises the question: is there ever a situation when someone does want to have a warning about that being unused? seems unlikely right

Seem unlikely I think. It's kind of pattern introduced by authors and documented as "use ret"

areduce is another case

sure issue welcome, although myself I've always just underscored the binding

Without knowing the internals, most people - probably - will use ret because it's stated in the doc and clojuredocs examples. I agree that maybe there is someone who will complain about switching off this warning, who knows... Anyway, leaving the issue.

areduce is the wrong example, so it applies only for amap

ret in areduce is and should be used by user (it's the accumulator)

it feels a bit arbitrary to make an exception just for amap - I think most clj-kondo users will know that when you're not referring to ret, they can underscore it

Yes, it's arbitrary. I agree with that. Well. I've never hit the case when ret was referred. But it should be named.

I'm starting to have doubts 🙂