Repro please. A screenshot and example lint error is not a repro we can do much with. Are you able to make an example file we can lint locally?

Ah, your answer indicates this is not a known, or common user error, or some such thing, which is mostly what I had hoped for. I'll try to figure it out myself for now, and will reach out with repro if it persists. Thanks for responding so quickly and being down to help troubleshoot as usual!!!

Indeed, it doesn't ring a bell and should not happen with normal Clojure code

In case you were curious, strangely, this is the result of something within the intersection of Calva, c-lsp, kondo and whatever peculiarities of our codebase working correctly now when it hadn't before. The & error was in fact not an false positive for rest/spread like kwarg args passing, but actually the usage of the special & "operator" within lilactown/helix. I ended up writing a couple hooks for helix to get everything working as expected. Thanks for everything again!

cool! Maybe you can contribute the hooks to helix for everybody to enjoy?

https://github.com/lilactown/helix/pull/93

fantastic, good work @U037TPXKBGS!

3 questions about hooks if you don't mind: 1. what's the diff between reg-finding and simply throwing? your hooks.md doc doesn't differentiate I think 2. looking thru the rewrite-clj docs and noticing all the talk of the zipper API, it does look nifty. Would the kondo use-case make sense for providing the zipper API? Or have I maybe misinterpreted its intended use case? 3. I wanted to also catch and provide error diagnostics for cases of, for example, passing maps as children to be rendered in helix components. I can do that most trivially by checking if there's a map-node as a child somewhere, but that doesn't work if the map was bound to a symbol/var elsewhere and then the latter is instead passed, making it a token-node. Is there anything I can do within a linting hook to maybe "resolve" references "1 or 2 levels deep" for error purposes?

1. simply throwing doesn't allow you to configure the linter name and level

gosh you're responsive

2. the zipper API was exposed in the past, but I ended up removing it because I didn't find it very useful for hooks, but might add it back if people want it

3. not yet

Do you mind if I submit a PR for adding some of this stuff/my learnings to the hooks docs?

Please do! 🙏

@borkdude would love help knowing how we can enable similar linting for the generated symbols in https://github.com/lilactown/helix/blob/master/src/helix/dom.cljc. That is, users typically import/require syms from that ns and use em like (d/div {:some 'props} "some children"). It has similar linting rules as $ , but I don't know how to tell kondo to lint every sym as such. Any ideas?

@U04V5V0V4 you could do this using :lint-as {helix.dom/a helix.foo.$}

you'd have to repeat it for every tag in the config

Ah, and could I do that as part of a helix PR, such that when consumers update to whatever helix version includes the new hooks/lint configs, they don't have to copy over every dom/* :list-as? Also dragging in @U4YGF4NGM for his take > btw I think you've tagged the wrong fellow

yes

Would you be open to a special lint-as* type key in config.edn that would accept something like:

:lint-as* { helix.dom/* helix.core/$ }

I've solved this for namespaces like this: https://github.com/clj-kondo/clj-kondo/blob/master/doc/config.md#namespace-groups So you can define a group using a regex and then use the group name in the config. We could do the same thing for vars

Oh cool, yeah that looks promising! It's not the biggest personal priority, and I'll see how @U4YGF4NGM feels about the upcoming big-ol-list of :lint-as that I'm opening, but if there's a way I can help out in your suggestion, lmk!

So something like:

{:var-groups [{:regex ".*" :name all}]}
{:lint-as {helix.dom/all helix.core/$}}

Yep. Any clue what an exclusion override might look like with that? Eg in this helix case, almost all exported syms are applicable to $-like linting, except one or two of em.

I think you would have to take care of that in the regex

sure, fair enough

Maybe it should look like:

{:var-groups [{:regex "helix.dom/.*" :name helix.dom/all}]}

not sure

didn't think hard enough about this specific case yet

so it would probably take me some more time

Another question, in this helix case again, there are some platform reader tag-based syms exported, eg

#?(:clj (defmacro some-macro ...))

to properly do that

clj-kondo lints a .cljc file twice, once for :clj and once for :cljs

I think the latter more explicit/specific version might make more sense

Ah right I recall now

@U037TPXKBGS 👋 "dumb as usual" not at all! :) The hook expands into code and then clj-kondo takes over as if the user had written that code. Perhaps you could just try to make such a fully written out example, without hooks, and then see how that's working.

hmm, I'm guessing you mean by throwing in a (prn (api/sexpr my-hook-return-val)) in my hooks and paste it in from cli?

no, I mean, write out an example that would normally be returned from the hook code

just normal code as how the user would write it

then I expect you would also see the unresolved var

Well, that's assuming my hook is written correctly, right? That it accurately outputs what I would have written by hand?

But ok I'll try that

Thanks

Okay... I didn't do anything, but a manual cli lint + couple of clojure lsp restarts later, and it's no longer false-positive-ing the symbol references... wtf lol

DIdn't change any config or anything, and already restarted lsp server tons of times in the past 30 minutes

Well, in any case, thanks again!!!

it could be that the other ns wasn't linted yet

or not refreshed

you're welcome

It depends on the syntax of those things :)

examples would be helpful

example meaning a link to the source of these things?

(defmacro defhtml
  "Define a function, but wrap its output in an implicit html macro."
  [name & fdecl]
  (let [[fhead fbody] (split-with #(not (or (list? %) (vector? %))) fdecl)
        wrap-html     (fn [[args & body]] `(~args (html ~@body)))]
    `(defn ~name
       ~@fhead
       ~@(if (vector? (first fbody))
           (wrap-html fbody)
           (map wrap-html fbody)))))

(defmacro defelem
  "Defines a function that will return a element vector. If the first argument
  passed to the resulting function is a map, it merges it with the attribute
  map of the returned element value."
  [name & fdecl]
  `(do (defn ~name ~@fdecl)
       (alter-meta! (var ~name) update-in [:arglists] #'update-arglists)
       (alter-var-root (var ~name) wrap-attrs)))

(defelem text-field
  "Creates a new text input field."
  ([name] (text-field name nil))
  ([name value] (input-field "text" name value)))

I'm more interested in the usages of those things

oh sure, sorry

as that is what matters for linting

(defn foo
  []
  (h.form/text-field {:class "hello"}))

oh I see

you should do lint-as defelem clojure.core/defn and then re-lint your dependency

if you are using clojure-lsp, that means lsp-workspace-restart

defelem without it’s namespace?

no, with, I was just lazy in typing

i did a restart workspace, and have

{:list-as {hiccup.def/defelem clojure.core/defn
           hiccup.def/defhtml clojure.core/defn}}

let me do it one more time with feeling just to be sure

going to restart my emacs and manually delete the .cache dir

define, "the cache dir"

.clj-kondo/.cache

sounds good

clears all of them up except for:

(h.def/defhtml bar
  [& content]
  [:body content])

where bar and content are unresolved symbols

right

that's weird btw, I would expect that syntax to be equivalent to core defn

what does your lint as look like?

what i posted above, both being treated as clojure.core/defn — for kicks i tried to use the def-catch-all for defhtml, but same thing

can you post your exact config please?

oh I see

omg

sorry, I'm doing 10 things at once

I can take a look at this tomorrow

config posted above was :list-as, not :lint-as

hehe ok then

sorry for the bother!

no problem