I think I was over thinking it. I should be able to get the behavior I need just by allowing the user to nest my component in their hiccup forms

`ws` package must be installed to use `react-devtools`.

Hmm after thinking about it, probably more of a general node issue than cljs specific.

I thought there's a cljs flag to do that

:process-shim

https://github.com/clojure/clojurescript-site/blob/master/content/reference/compiler-options.adoc#process-shim

I guess not :)

After upgrading to ClojureScript "1.10.516" I get a warning when figwheel tries to reload my code: Can't take value of macro cljs.core/and, when the code looks like this:

(when (and beg end) ,,,)

(let [both? (and beg end)]
      (when both? ,,,))

looks like it. you should create a minimal repro and report it to Jira

will do

this only happens when figwheel tries to reload my code tho, not when it initially builds everything - so there's something weird going on for sure.

The Day8/re-frame-debux lib mentions I should replace day8.re-frame/tracing with day8.re-frame/tracing-stubs when doing production builds (as the /tracing one might hurt dead-code elimination), but doesn't elaborate how to achieve this. Is there any tricks for replacing a lib just when doing a uberjar build but not otherwise?

Yes, here’s what’s in our project.clj file:

:profiles
  {:uberjar {:dependencies [[day8.re-frame/tracing-stubs "0.5.1"]]
             :omit-source true
            ,,, ;; etc
   :dev [:project/dev :profiles/dev]
   :project/dev {:dependencies [[day8.re-frame/re-frame-10x "0.3.3"]
                                [day8.re-frame/tracing "0.5.1"
                                ,,, ;; etc  

@manutter51 many thanks for sharing that!

anybody using semantic-ui with reagent ?

yogthos did something with it recently https://github.com/yogthos/semantic-ui-example

it wants you to initialize certain components on the page with jQuery calls, which I can't get working

i had the bright idea to just add the call I needed to :component-did-mount like so ^{:component-did-mount #(.dropdown (js/$ ".ui.dropdown"))} but apparently that's not correct either

isn't it simpler to use semantic-ui-react instead? it doesn't require jQuery, so should have less footprint

and it looks like there .dropdown probably just want's an element, so you could use a vanilla js .querySelectorAll, or goog.dom/getElementsByClass

@dj942 yeah i've pretty much arrived at the same conclusion

semantic-ui-react is jQuery free and there appears to be several straightforward wrappers for it on clojars, first and foremost cljsjs.semantic-ui-react

semantic-ui-react works pretty well with reagent/re-frame

Just hit up against https://dev.clojure.org/jira/browse/CRRBV-15 when trying to run some doo tests in one of our cljs projects. Does anyone happen to know a workaround?

if someone can supply a patch, happy to help get it applied/released

Can a ClojureScript defmulti be queried? VM1169:1 Uncaught TypeError: cljs.core.methods is not a function

It can, the name is munged, my bad cljs.core.methods$

is it possible to access javascript equality in cljs?

I think identical? will work.

okay, which leads me onto my next question. Is returning js/undefined the only way to emulate a function without a return?

I seem to have found a bug, but I'm really not sure where to even start (and I really don't have time right now), but does anyone ever lose global vars inside a deftest block? The following code:

(prn 1 validate!)
(deftest test-example
  (prn 2 validate!)
  ...)

1 #object[consigncloud$policy$pos$validate$validate_BANG_]

Testing my.test.ns
2 nil

I'm using shadow-cljs with a custom test-runner namespace if that matters

Although I'm really just swapping out the reporter, and I haven't had a problem like this in a while. Odds are I've made some mistake

try (js/console.log ...) if it really goes away or your printer/reporter maybe just gets confused

Ha it gives me:

nil
undefined

but if I add (def validate! validate!) at the root level, it works

that is not valid

that will cause a self reference

I take that back, (def v! validate!) works

x = x;

I'm importing validate! from another namespace, where it's just defined using defn

then use the alias/validate! 😉

When I (prn v!) inside the test it gives me

#object[consigncloud$policy$pos$validate$validate_BANG_]
[Function: consigncloud$policy$pos$validate$validate_BANG_]

What's alias?

Oh you mean v!?

(:require [some.thing :as alias])

ah

(prn v/validate!)

gives me nil still

oh wait I misread your require

nevermind, that still didn't work

is there a builtin I'm nuking?

I'm lost and don't know what you are doing anymore 😛

ha sorry, let me recap

(ns x (:require [y :refer [validate!]])
(def v! validate!)
(prn v! validate!) ;; => function, function
(deftest test-x
  (prn v! validate!) ;; => function, nil
  )

edited for succinctness

works perfectly fine for me in a test build using exactly that setup

My test case is very much tied together with other code, if I continue to have the problem I'll try to put together an isolated test case. Unfortunately I've got a time crunch, so I'll use the redefinition trick for now

do you may run other tests that with with-redefs or other stuff?

something that may mess up the tests when they actually execute? since the initial prints all run when the code is loaded

so if something messes with the namespace objects after they have been loaded that will break stuff

Not much, I'm not super good at namespace/var manipulation. I do use binding with a different var from the same namespace, and elsewhere in the calling namespace I use ns-publics for an unrelated ns

That is, I bind a var from the validate! namespace

That's all done at load time

I think before the tests are registered

after the tests are done

try checking the state of the code in the JS console

something likely just got renamed somehow

Ok, I'll give it a shot, thanks for your help

And I'll try to put a test case together too

try running only that test ns and nothing else

@thheller I tracked down the problem; I was in the midst of a refactor, and a namespace was clashing with a var named the same as the last segment of that namespace exported from another namespace with the same first few segments. To make that concrete, the warning I got eventually was: Namespace consigncloud.policy.pos.validate clashes with var consigncloud.policy.pos/validate. So it was a weird thing I was doing. But I did experience some buggy behavior, first, I wasn't getting that warning until I had removed half my source code, so I'm not sure why but something was suppressing it. Second, I only experienced the bug because of what I assume are internals of dependency resolution order; I removed a completely unrelated file and my code started working. Anyhow, I'm good to go, is there anywhere I might report that but? I'm not sure I can produce a test case for it since it involves a lot of my code. Definitely one of those hard-earned pieces of wisdom that I'll remember though 😅

hehe yeah those things can be tricky. I wish namespace weren't built like that but its too late to change that now 😞

not sure why you wouldn't get the warning immediately but I suppose compilation order matters for that

Yeah, it's very weird that they would clash like that

so I finally succeeded in getting Clojurescript to build under Bazel, with an external Closure library and so on.

but I'm getting double-inclusion, and I'm possibly grabbing the wrong files. I want cljs to output everything but the Closure core bits

I can find the files with just my own code, but they depend on the cljs core library.

oh, cljs.core is a Closure library; I can probably include those separately.

is that a question? CLJS compiled code always depends on cljs.core since the collections and stuff live there

I discovered more files than I thought were generated; I think I grabbed a "big" file instead of the "small" one.

cljs/core.js is pretty big 1.3mb or so unoptimized

yeah. the objective here is to let Bazel do the combining and compilation; I want just my code as a Closure library with goog.require statements.

linking in the goog.base and cljs.core libraries is done externally, here.

I realize this is against the grain of the usual CLJS build style; but it's more important to fit with the much larger grain of Google's build system.

you shouldn't combine output from different compiles that may cause issues when they came from different versions

but otherwise you can take the :none output just fine and include it

not sure about bazel but the source files act like any other closure library source

yeah, it's Bazel that's causing difficulty here

I've got it grabbing the right, standalone files now.

the cljs.main run produces both the simple JS files and bigger ones that include cljs.core and goog.base already.

you only need to run :optimizations :none if you are going to process them anyways. no need to run :simple

it was several weeks ago I first took a crack at this; I thought the :none output didn't have the goog.requires?

I guess I can try it

:simple doesn't have them, :none does.