Is it possible to compile a single cljs file to a single js file that does not require goog.base?

@iku000888 I believe compiling in :advanced mode, and perhaps :whitespace mode, will produce a single file, which inlines any goog.base deps into a single js file.

But I believe you'll need to specify your main ns with :main in your compiler options.

:main sounds like the trick! Will give it a try. (I did :advance and I was still seeing the goog.* stuff showing up). Thanks!

No prob. See here for more info: https://clojurescript.org/guides/quick-start#less-boilerplate

and note that in project.clj you do not need to prepend the main namespace with an apostrophe

Hrm Somehow this gets prepended to the output...?

I see maven central doesn't have any releases for org.clojure/google-closure-library since june 2016; will that be updated eventually? There's a pretty serious bug in the goog.math.Integer class.

@iku000888 I believe that's normal. Always there.

@gfredericks I would advise to bring it up in #cljs-dev or JIRA

@anmonteiro thanks, didn't know about #cljs-dev

@john thx!

@iku000888 np! Have fun 🙂

What is the best way to introspect a

#object[HTMLCollection [object HTMLCollection]]

(js/console.log the-coll) or (js/console.dir the-coll)

I tried js->clj, but seems to not change the output...

those are native objects that have no equivalent in clj

hrm...

So would that imply I only get the dom api's to speak to them and nothing else?

Oh and @thheller thanks for the tips!

Hey, how can I make the functions say_hi and say_hello js functions that I can call in node?

(ns my-ns.core
  (:require [cljs.nodejs :as nodejs]))

(nodejs/enable-util-print!)

(defn main [& [a name]]
  (case a
    "hi" (say_hi name)
    "hey" (say_hello name)
    (println "Nothing"))
  {"a" 1})

(defn ^:export say_hi [name]
  (println "Hello World!"))

(defn  ^:export say_hello []
  (println "I said"))

(set! *main-cli-fn* main)
;;(set! (.-say_hi main) 'say_hi)

node -e 'var my-package = require("./target/js/milia.js"); my-package.say_hello()'

or better yet how can I compile a cljs package into a node js one

The methods people advocate are just not covering exporting multiple functions

@urbanslug that looks right to me

Weird, it says "my-package.say_hello is not a function"

lol why did I even bother leaving out the name if I was going to leak it on the require. OpSec fail

var my-package = ... is not valid JavaScript

note the - illegal character in the identifier

OK in this case the name is actually different and doesn't have the -

OK since hiding the name was for nothing let me be clearer

then other problem is

goog.exportSymbol(...) has nothing to do with Node.js packages

that’s all ^:export does

you need (set! js/module.exports ...)

or whatever the convention is for Node.js (I always get it wrong)

@urbanslug this might work for you https://github.com/thheller/shadow-devtools/wiki/ClojureScript-for-node.js-libraries

@thheller That doesn't export any functions

see the configuration, you configure the exports there

:exports {:say_hi my-ns.core/say_hi}

no need for anything else in the code

hmmm trying to figure out which approach is better

the issue with the (set! js/module.exports ...) hack is that it doesn't work in :none. so it will be slower during development

OK one sec, let me play around with them

OK forgive my total newbiness here but I can't figure out in the set! way of doing things shouldn't this work?

(ns milia-wrapper.milia
  (:require [cljs.nodejs :as nodejs]))

(nodejs/enable-util-print!)

(defn main [& [a name]]
  (case a
    "hi" (say_hi name)
    "hey" (say_hello name)
    (println "Nothing"))
  {"a" 1})

(defn ^:export say_hi [name]
  (println "Hello World!"))

(defn  ^:export say_hello []
  (println "I said"))

(set! js/milia-wrapper.milia.exports {:say_hello milia-wrapper.milia/say_hello})

node -e 'var milia = require("./target/js/milia.js"); milia.say_hello()'

Oh it doesn't work in node

*nonde

*nonde

*none

@urbanslug only module.exports is recognized by node when using require

My optimization level is none

hmm

So node -e 'var milia = require("./target/js/milia.js"); milia.exports.say_hello()'

?

no, none of this works in :none sorry

What did you mean by only module.exports is recognized?

Did you actually mean exports should be included?

(set! js/milia-wrapper.milia.exports {:say_hello milia-wrapper.milia/say_hello}) that part

only works as js/module.exports not js/milia-wrapper.milia.exports

but even then it won't work in :none, that is because node expects one file per require but CLJS is split into many files in :none

@thheller hmmm

(set! js/module.exports {:say_hello milia-wrapper.milia/say_hello})

node -e 'var milia = require("./target/js/milia.js"); milia.say_hello()'

ah :advanced doesn't work because you are exporting a CLJS map

(set! js/module.exports #js {:say_hello say_hello}) that should do it for :advanced

@thheller hmmm so I have to use your build tool to get/use the :exports in my project.clj

yep, others don't support this feature

Thanks a ton

I upgraded my project to [org.clojure/clojurescript "1.9.521" :scope "provided"]. Now a (let [{:keys [a b] :as full} (my-own-function)] (prn full)) prints {[:a "a's value"] [:b "b's value]} rather than {:a "a's value" :b "b's value"}. Did destructuring change?

I checked the final value of (my-own-function) by printing it before returning it, and it looks just fine, that is, {:a "a's value" :b "b's value"}, and is then magically converted by the destructuring or let or something.

@reefersleep what version did you upgrade from?

Here's a simplified example:

(let [{:keys [hey girl] :as full} {:hey "yo" :girl "dude!"}]
               (prn hey)
               (prn girl)
               (prn full))



nil
nil
{[:hey "yo"] [:girl "dude!"]}
nil

cljs.user=> (let [{:keys [hey girl] :as full} {:hey "yo" :girl "dude!"}]
       #_=>                (prn hey)
       #_=>                (prn girl)
       #_=>                (prn full))
"yo"
"dude!"
{:hey "yo", :girl "dude!"}
nil

^ in Lumo, which is running 1.9.521

@anmonteiro 1.9.93

right, my suspicion is you have an old analysis cache that you need to clean

I've no idea what's going on 🙂

I guess the first thing to try is to restart Emacs, I'll give it a go.

that’s not gonna do anything

@reefersleep this is unrelated to Emacs. Try do delete your output directory and restart your build

I'll try that. (I had restarted the repl, which runs figwheel, but not deleted the output dir, and I can see why you suggest that 🙂 )

Yeah, the point of the analysis cache is to make your compilation times faster even if you restart the REPL

I'm unsure of what I'm looking for. Tried mv'ing public/js/out to out.bak, which prompted the build process to create a new out folder, but this did not make any difference.

the typical issue is that the cached partially compiled code mixes with the new code in an unexpected way that creates nonsense results; the fix to this is to shut down any auto-builder, run lein clean (or whatever else it takes to completely erase all partially compiled artifacts), and the nrestart the builder

this is commonly a problem after changing dep versions (in my experience)

I'll try that, @noisesmith

note that lein clean is not gonna do what you expect it to do

@anmonteiro it does for me- because I set up my clean targets to do so, I guess I can’t generalize that

😕

but yeah, the main thing is to destroy all compilation artifacts

having a default config where clean fails to remove an output of a compiler driven by lein seems like a misconfiguration to me

lein cleandid not do the trick in my case.

rm some of them artifact output folders

I've tried to remove all of those that I could find, lein clean and re-run figwheel, nothing has helped so far.

Ah, just found a hidden folder I missed

Yes! Finally managed it. /.cljs_rhino_repl/ was the (last?) folder that I needed to remove. Thank you very much, @anmonteiro, @noisesmith and @john !

if that was my project I would add that dir to lein’s clean targets

I don’t like having caches of compiled code that lein doesn’t obliterate when you run clean

I'll run with that! 🙂

wonder how the folder was generated to begin with, it wasn't re-generated when I re-ran figwheel

Hi all, I'm messing with the new :npm-deps feature and I keep getting errors about transitive NPM dependencies. Currently running with a deps.cljs that looks like this:

{:npm-deps {:draft-js                 "0.10.0"
            :draft-js-import-markdown "0.2.1"
            :draft-js-export-markdown "0.2.2"}}

My project also uses reagent, which is where I'd like to pull the react dependency from. Do I need to declare it as a foreign-lib or something?

@timgilbert you probably just need to specify react too in :npm-deps

if it’s an NPM peer dependency

currently no way to make it play nice with Reagent if there’s no lib support for it

Ok. I was hoping I could actually just give it :draft-js-import-markdown and have it pull in draft transitively too

it’ll pull transitive dependencies

Hmm, looks like the NPM dependency tree is broken though

but not peer dependencies

Ah, ok, that makes sense

Is there a similar mechanism to the lein/clojure :exclusions thing, so I can tell NPM "this will be loaded from elsewhere, don't worry about it"?

if some dep will be provided by other software at runtime, that’s what the :provided scope in project.clj is for

easy fix for someone who has a moment, lein help :sample doesn’t mention the :provided profile, maybe it should?

https://github.com/technomancy/leiningen/blob/master/doc/PROFILES.md it’s described in this document

oh - maybe what you are describing is different though

what's the correct way to optionally handle whether someone passed something into & args? Is it (if (first args)... or (if (empty? args) ... or is something with (apply afn stuff args) supposed to always work? I worry about a nil being passed in.

I’d test if args is empty - it always wraps all the args in a collection, so if they passed nil you would get [nil] which is not empty

Which is why I like the (first args) method, since it returns nil if there is a nil or if it is empty.

that’s your problem though - if something calls (fn [a & args] ...) with (a 1 nil 2) calling first misleads you

checking if args is empty is reliable though

that's why I felt so uneasy

Sage advice

Thanks @noisesmith

I'm seeing a Delay deftype and related documentation in the source that references a delay function, which doesn't appear to exist. Is delay not yet implemented?

https://github.com/clojure/clojurescript/blob/master/src/main/clojure/cljs/core.cljc#L2064

whats with diff between that and: https://github.com/clojure/clojurescript/blob/master/src/main/cljs/cljs/core.cljs ?

That's the one that the docs bring me to

hmm

that file is for core functions

the file I linked to are core macros

gotcha

What's the right way to require/import a closure class (e.g. goog.math.Integer) so that I can call static methods, refer to static fields, and extend the class to protocols?

I used to have (:require [goog.math.Integer :as int]) and then use int/ for the former cases and goog.math.Integer fully qualified for the latter

but that feels terrible and has started giving warnings in more recent versions of cljs so I assume there's a better way to do this

clojurescript complains: Use of undeclared Var cljs.core/and at line 13

wat?

hit, I have an endpoint where I can upload a text file with curl like this: curl -X POST -H "Content-Type: multipart/form-data" -F "file=@/resources/speciesDiffusion.tree" now I need to send a similar request from a browser, but

(ajax/ajax-request
      {:uri (str "" "/continuous/tree")
       :method :post
       :params {:treefile content}
       :handler #(println %1)
       :format (ajax/text-request-format)
       :response-format (ajax/json-response-format {:keywords? true})})

I’m sort of surprised this is hard to find, but does anyone know of a good way to configure routing (like via secretary) for a single-page app (SPA) in a way that can deal with browser refreshes, as well as someone just putting a full URL path into the location bar?

@mikerod I don’t know if there’s something fancier, but what my app does is make sure that any state that should persist across a refresh (or bookmarking, or following a link) needs to be injected into the fragment data of the page

since the data is in the fragment, it doesn’t cause a new load of the page just because you changed the url

because it’s in the fragment, the back button and link sharing etc. are less likely to break

I made a small library that takes edn data and translates to base64 and puts it in the fragment (with a decoder that goes with, of course)

I haven’t even had luck doing a page refresh while attempting to have things after the fragment

I’ve actually been seeing it just load the page ignoring the fragment

@mikerod weird -I have never seen a refresh alter the fragment

I’ve noticed that in an angular app I was able to do refreshes against a fragment and also to directly plug the fragment into the location bar and go

wait, who are you expecting to act on the fragment?

So I wonder what is happening there

the fragment is for your app, you have to do something with it (in my case I parse it and inject it into the app state)

e.g. for that one location is localhost/thing/#/stuff

If I refresh there It seems to ignore #/stuff

It doesn’t even call the navigate listener, so I must be doing something wrong

I just wish there were some clear docs on this issue

I guess it is just tricky or something

OK - when my fragment is changed, goog.history definitely gets invoked

that’s what drives my router

I’ve seen accountant which seems to try to address the overall problem (I think). However, it doesn’t work well for someone to manually put the location in the location bar and go (or I don’t know how to use it right if it does)

I probably have more than 1 probleem

I just hook the my router into goog.history and then parse data out of the fragment… I don’t feel like I did anything that unusual

so I need to isolate things

I followed the basic template I’m seeing other places.

The main thing I don’t see a clear path forward on is browser refreshes behaving sanely

I think I can get back, forward, and fragment changes to do what they are supposed to

for refresh, I had to make sure that app load propagated state via the same code that the router does

and that no page implicitly relied on state initialized from another page

interesting

that plus ensuring that all “bookmarkable” state and state that should be altered by forward / back areas was reified into the fragment data

then you get a mostly-not-broken app

(in terms of url / navigation / refresh)

I should really make a blog post about how we did this

would be interesting to see. I haven’t seen blogs really answering these points for me so far. Been tryign to search quite a bit.

I’ve done this once with secretary but I don’t remember the details (warning this code is 2+ years old): https://github.com/darwin/faceboard/blob/master/frontend/src/faceboard/router.cljs

the trick might be in using goog.History and https://github.com/darwin/faceboard/blob/master/frontend/src/faceboard/router.cljs#L32

interesting, thanks for the links