err that's probably just a vector?

what is a clj-nil then?

i guess it's not important, in that it's not a number...

Github search is not great. But looking through https://github.com/clojure/clojurescript/blob/9562ae11422243e0648a12c39e7c990ef3f94260/src/main/clojure/cljs/analyzer.cljc, I get the impression maybe clj-nil is a symbol that internally represents the combination of null and undefined? Found this, which definitely seems relevant to your query. Seems that since 1.10.741, it is in fact "not a number". 😄 > from Changelog for v1.10.741 > • CLJS-3086: Switch clj-nil to not be considered a numeric type

Looks like you have (a vector of (a set of nil and clj-nil) and a number). Don't know if spelling that out explicitly helps you any with figuring what went wrong.

thanks, yeah that helps at least i know what i'm looking at.

[#{nil clj-nil} number] is the types of the arguments, so you gave a #{nil clj-nil} as first argument (probably after calling a function) and a number as of second.

And it comes and goes. Now, I cannot reproduce the error anymore… Probably shadow listens to Slack or something. 😛

And back again… wtf

My component looks like this if it can help:

(defn search-bar
  "Search bar where the user can input its origin."
  []
  (let [origin (reagent/atom nil)]
    (fn []
      [:div {:style {:padding "9px"}}
       [:input {:type         "search"
                :style        {:border "none" :width "100%"}
                :placeholder  "Paris..."
                :value        @origin
                :on-change    #(reset! origin (-> % .-target .-value))
                :on-key-press #(when (-> % .-key (= "Enter"))
                                 (>evt [:eve.events.api/search @origin {:clear true}]))}]])))

>evt is just re-frame/dispatch.

in general when looking at stacktraces do not look at the first line

look for the first reference of your code

Yeah, there is none

what is the rest of the stacktrace then?

could be a problem in the event handler?

the error you get when calling nil as a function

so f being nil and (f 1 2 3) or so

maybe you call something that isn't yet set on load

There are three logs, apparently the same. The traces are then: 1. cljs.core, reagent.ratom, day8.reagent.impl.batching 2. cljs.core, reagent.ratom, day8.reagent.impl.batching 3. day8.re-frame_10x.inlined_deps.re-frame….re-frame.router

hmm so its re-frame-10x, not even your code?

then I don't know. could be timing issue I guess

That’s a dependency I have to have a tracing interface in the browser.

This https://github.com/day8/re-frame-10x.

My event handler is quite straightforward, I don’t see what could not be set:

(re-frame/reg-event-fx
 ::search
 (fn [{:keys [db]} [_ stop-name & [opts]]]
   (let [connections-request (query {:method     :get
                                     :uri        (url (str "/connections?stop_name=" stop-name))
                                     :on-success [::set-connections]
                                     :on-failure [::raise-http]})
         origin-request      (query {:method     :get
                                     :uri        (url (str "/stops?name=" stop-name))
                                     :on-success [::set-origin]
                                     :on-failure [::raise-http]})]
     {:http-xhrio [connections-request
                   origin-request]

      :db         (let [with-search (assoc db :search/origin stop-name)]
                    (if (:clear opts)
                      (assoc with-search :origins (list))
                      with-search))})))

Most likely it has to do with this extension/plugin. Thanks for your answer, I was afraid I was missing something obvious. 🙂

I think I made some progress. Actually the problem occurs after page reload too, it just appears random. Since there are no «obvious» mistakes after investigation, I decided to poke where the more impure stuff happens, and it’s in this handler above. Two requests are made concurrently. I tried disabling the origin-request one, and the problem never occurs. I re-enabled it and disabled connections-request and now the problem always occurs. I suspect something is messy in my subscriptions and it works only in a given order of the responses.

The stack-trace was not very helpful 😅 .

Also, I noticed that the 10x reference only appeared when the panel was open on the page. So, it could not be the cause.

And here it is! I was «expecting» the «origin» response to come after the other one which is essentially building a map. I then used this map as a function in a filter. 💥 💥 💥

make sure you are consistent with file names and namespaces

_ in filenames become - in namespace names

my_project.core-test is breaking that rule and should be my-project.core-test

@U05224H0W thanks for the reply, actually is correct named on my local, I just write wrongly here…