@thheller When using :target npm-module and wanting to use a repl via cider, I know I have to start a node process for the repl to work, but given that npm-module outputs so many files, which one do I need to run with node to get the repl to work? Or does the repl not work with this target?

hello, I’m trying to eval using (in a Chrome extension):

(defonce env-state (cljs.js/empty-state))
(defn eval-str [s]
  (cljs.js/eval env-state
                (reader/read-string s)
                {:eval       cljs.js/js-eval
                 :source-map true
                 :context    :expr}
                (fn [result] result)))

likely that the current ns wasn't created properly. eg. trying to run cljs.user.a = 1; when cljs.user is not defined properly.

@thheller thanks, that’s a useful pointer

chrome extensions have very weird scoping rules so could be that things are just eval'd in the wrong scope

I’m in uncharted waters (for me!) because I’m learning about cljs.js and chrome extensions at the same time

I'm not sure if this is really a "#clojurescript" question or more of a general web dev question, but does anyone have a good way to invalidate browser cache when you push out a new recompiled app.js to production? Right now I've been putting version names, aka, app-v1.1.1.js, app-v1.1.2.js, etc, to get browser to DL new code, but it's pretty embarrassing when I forget to bump the version.

@goomba usually you’d append a hash to the file name

you mean that in the literal sense, yes? like app.<hash-value>.js?

yep

Gotcha. I really need to figure out my CI/CD better >.<

there’s actually compiler option to include hash in the name of out file :fingerprint true

interesting. I don't suppose you also have a neat trick for updating the HTML without having to manually change the name over? 😛

(I'm sure I could figure it out ... some combination of directory scanning + templating, but curious if there's a better way)

Yeah this would require fiddling with html

I use this, which is also pretty hacky

;; Use fingerprinted (cache-busting) app.js when created for production builds.
(let [js-str "resources/public/js/compiled/app.js"
      subm (some->> "public/js/compiled/manifest.edn"
                    io/resource
                    slurp
                    edn/read-string)]
  (def js-path (-> (cond-> js-str (map? subm) subm)
                   (s/replace #"resources/public/" ""))))

basically, manifest.edn will have the data on the specific hash used

Since it’s for production build only, you can add hash manually and update the html as well

slurp js bundle -> hash -> rename js bundle -> update html

but this really depends on your pipeline and how you serve a page

given someone adept with secretary for managing routes as of circa ~2 yrs ago, is there any innovation/development in CLJS land which would warrant switching to some other framework for managing routes?

[assuming a greenfield project]

or should one in that situation merely proceed with secretary?

I haven’t used this but have heard great things about it https://www.metosin.fi/blog/reitit/

I use https://github.com/juxt/bidi for all my cljs routing and it has been meeting all our needs so far.

Another question about self hosted CLJS and 'ambient' name spaces. Let's say we have 3 such namespaces A, B, C. Further, they have their caches loaded at init time. Now, we try to evaluate "(ns my.test (:require [A] [B] [C]))". This throws an error about *load-fn* not being set (true, it is not being set by the code). Despite already having A,B,C cache loaded in the compiler state it appears that ns forms with requires need the *load-fn* set to do something. For things like A,B,C (and core things), would this basically be a NoOp?? @mfikes

@jsa-aerial Yes, to create a NoOp *load-fn* have it call back with {:lang :js, :source ""}.

You may also be able to avoid attempts to load things by priming cljs.js/*loaded*

Specifically in your case @jsa-aerial you might be able to get away with doing this at init time: (swap! cljs.js/*loaded* into '[A B C])

@mfikes Many thanks for the information. Now, just to take the other extreme - if D is not ambient but is required, then the *load-fn* needs to be able to resolve the actual source for this and hand it off, correct?

Yep

OK, I think I am finally getting the idea how this stuff works 😬

In that last case (assuming cljs) it should callback with {:lang :cljs, :source the-resolved-and-fetched-source)}, ccrrect?

The :lang would actually be :clj (even though it is the ClojureScript dialect 🙂 ) And the-resolved-and-fetched-source would be something like "(ns D) ,,,"

Ah, OK. So the source is basically the "typical" name space at top source file for D

Nah, sorry, that's not what I meant by that example.

Perhaps a better example would be "(ns D) (def x 3)". In other words, the entire source file.

When you feed back that source, the self-hosted compiler is going to march through it, compiling each form.

OK, right - that I think is still what I'm thinking - it's just that that string would be fetched (from a url, say)

Further - anything D requires would also need to be so resolved?

The self-hosted compiler will take care of that as well, calling *load-fn* again when needed.

Be sure to read through https://github.com/clojure/clojurescript/blob/master/src/main/cljs/cljs/js.cljs#L75-L99

Right - but the *load-fn* would need to be able to fetch the source for those things just like for D (assuming it isn't an ambient already cached)

Right, it is up to *load-fn* to somehow produce the value for :source

Thanks so much @mfikes - this has helped a lot. Still a fair amount of futzing to do, but at least I think I know what the futzing is!!

Cool!

If you are at all interested in where / what this is all about, you might check out https://clojurians.zulipchat.com/#narrow/stream/197746-datavis-gatherings/topic/hanami.2Bsaite.20meeting

@mfikes Wow 😮 - Works! 😂