when using :exclusions it can also be a good idea to explicitly add a dep for the version of the thing you want at the top level, before anything that depends on it

yea, that's how I fixed it, had to add guava 19 at the top

its weird tho, I hadn't made any dependency changes in the past 2 weeks then all the sudden today it stopped working, really confusing

Behold the classpath 🎉

@mrchance I can't tell if that emoji i sa piece of pizza or a kazoo, or what

Little bit of both, I guess ^^

it made me smile

Then it already served its purpose 🙂

Now, if someone can tell me how syntax quote works when expanding clojurescript macros, my day is made 🙂

What does (foo ...) expand to when I use the same ns trick for macro and the matching cljs namespace has a foo function?

Can I rely on it being the foo from that ns, or should I always fully specify the ns in the macro?

sorry, I meant

`(foo ...)

Also, it didn't seem to work for a record in that namespace :thinking_face:

In Clojure that's not an issue, because the macro usually lives and is evaluated in the same ns that houses the functions it references...

has anyone ever had an issue that appears only when code is optimized? I’m running into a weird issue with https://github.com/benhowell/reagent-quill that only happens when the code is optimized.

@mrchance think of how many clojure.core functions you use though - ` namespace qualifies things such that your code works even if the code using the macro has renamed or hidden parts of the other ns

@lumpy Typically things having to do with code using (.foo x) and (.-foo x).

@noisesmith yes, I understand why that is needed in general, I just want to know how it works in cljs. Not sure how your reply answers that, maybe I'm being daft 🙂

@lumpy Are you using either?

So you're basically saying I should fully qualify all symbols?

That's a bit... unwieldy

` qualifies the symbols, you shouldn't have to do it by hand

just be sure the cljs is requiring the right namespaces somewhere at runtime

Hm, so I had a foo.cljs and a foo.clj file, the cljs file required the clj one and referred the macro, which had a reference to a record in foo.cljs in it. This didn't work. For the functions in foo.cljs however, it did. That's what confused me

Something like

(defmacro pipe [x]
 `(Pipe. ~x))

(Little more complicated, I know this doesn't warrant a macro :)

Figwheel handled it pretty weirdly too, I got a very short popup that looked like an undefined var warning, which immediately disappeared and was replaced with the success badge Oo

@jeaye yes that seems to be the problem

is there a good way to fix those issues

@lumpy Ok, use cljs-oops for all member access. It's a superb library.

https://github.com/binaryage/cljs-oops

awesome, I’m so mad I didn’t ask here sooner

is there a known reason this happens?

@noisesmith I still don't understand how that works, if a macro does

`(foo x#)

it's the defining namespace that determines the expansion

ok, so it checks in the cljs namespace that matches its own, always?

oh, - good point, that means that auto-expanding to the right cljs namespace would be tricky in some cases

it's a decent reason to have stub clj namespaces with the same path as the cljs namespace you want to emit calls to

maybe

ah, that's not a problem, I usually have both anyway, since I like to be able to omit :require-macros and treat them transparently

But in theory it should work similarly for records, correct? Nothing different there?

sure - you can refer to records via the ns level vars that get created rather than the raw classes / methods sometimes as well

(that might make things simpler)

(ns foo.awesome)

(defmacro to-bar [x]
 `(Bar. x))

Ah, good point

Yeah, I might just use the helper fns

and the answer is yes, unless Bar is in scope referenced from some other ns

ok, thanks

I should still qualify calls to other libraries though, and hope they have been required somewhere? e.g. cljs.core.async/chan or something like that?

that seems reasonable

ok, I think I got it, thanks again!

@lumpy Names are mangled during :advanced optimizations, to be shorter.

@lumpy http://squirrel.pl/blog/2013/03/28/two-ways-to-access-properties-in-clojurescript/

Of course, cljs-oops supercedes aget and is much more preferable.

Thanks @jeaye this was a lifesaver

Is there something like criterium for cljs?

js/console.time ?

@jfntn See cljs.core/simple-benchmark

Or, even simpler, cljs.core/time

@mfikes simple-benchmark is what I’m using, but I can’t explain why I’m seeing a lot more variability accross runs than with the same code on the jvm

@jfntn I've seen JavaScriptCore's FTL optimizer create mysterious variations https://github.com/mfikes/planck/issues/417

yeah seeing similar things with node, though it’s not quite as dramatic

I'm trying to make a cljsjs package for react-dom/test-utils

Does anyone know where the repo is ?

I packaged React 16 beta 2 to Cljsjs: https://github.com/cljsjs/packages/commit/c49c89169b9bb7391508c5f5e19d35514feee283, seems to break some tests at Reagent at least

What would be the fastest implementation of a global key-value cache? Current status: (def cache (volatile! (transient {})))

@jfntn why not an atom?

Thought volatile! would be slightly faster due to lack of validators, watches etc?

volatile! does not have any atomicity guarantees though, so it's not thread safe. I believe it's intended for smaller scope changes in a single thread (like in stateful transducers).

there are no threads in JS so volatile! is fine (and slightly faster)

@dnolen fwiw the issue from yesterday was a REPL environment one, a bug in a version of bignumber (which web3 relied on) hangs in Safari / similar envs (React Native on iOS) but not in a Node REPL, fixed by depending on a separate version of bignumber and doing arithmetic there

@oskarth ok good to know 🙂

quite an interesting bug actually https://github.com/MikeMcl/bignumber.js/issues/120

Hello all, a silly question -> how to create <ul uk-accordion> from [:ul {?}]?

@fabrao {:uk-accordion true}?

well, I did this but didn´t work, that because I asked here. Is that the correct way?

@anmonteiro

Hi. I’m following https://clojurescript.org/guides/javascript-modules and trying to load a vendored JavaScript project with CommonJS modules. Requiring the namespace fails (“No such namespace”). Re: this sentence: > Notice that a :foreign-libs entry may now specify a directory for :file. In this case, the ClojureScript compiler will recursively search this directory for .js files and automatically create the :foreign-libs entries for you with the supplied options. Is there a way to get the cljs compiler to list the foreign-libs entries it has created to aid in debugging? Currently I have the following foreign-libs entry: :foreign-libs [{:file "vendor" :module-type :commonjs}]; where “vendor” contains one directory, csp-report-norm, the contents of which are a local copy of this git repo: https://github.com/tollmanz/csp-report-norm

:verbose true is already on so that’s not it 🙂

It seems that I can only use {:data-stuff "something"}

@lvh currently there is no way to do debug, though that’s a probably a good feature to add to :verbose

@lvh no one might have the time to look at this right now, so best thing to do is make a minimal case - zero tooling except Quick Start uberjar

if you can repro then make a ticket

do not link outside the ticket

add all steps to repro to the ticket

(also hi @dnolen thanks for all the work you’ve been doing; I know it’s a team effort but so much cool stuff since I last touched CLJS :))

okiedokie, I will try to reproduce

if you like add separate issue about dumping foreign libs, probably something we could add to cljs.build.api

I’ll look into it right now

@lvh we’ve been a lot better about adding test cases for all these different combos - so if there’s an issue we’ll try to make sure it doesn’t regress

@anmonteiro thanks

oh actually

I know what’s wrong

@lvh there’s no :provides entry in your foreign-libs definition

hrm, that might not be needed

Ah. From that instructions page I liked, just after the part I quoted: > The :provides namespace for each entry will be automatically computed from the directory structure. … so I was also expecting that to work for my vendored repo.

so how are you requiring it?

(require [vendor.index]) should be how you require it

sorry

csp-report-norm.index

[:ul {:data-uk-accordion true}] -> <ul data-reactroot="" data-uk-accordion="true" class="uk-accordion"></ul> WTF !!!?

Ah; I was requiring it as [csp-reports-norm]; not with index, I thought that was implicit. That made the compiler happy, but it still doesn’t appear to be copied into out/; notably, when I try to run it with node, I get:

cspshovel/out/goog/base.js:1357
              throw Error('Undefined nameToPath for ' + requireName);
              ^

Error: Undefined nameToPath for vendor.csp_reports_norm.index
... (stacktrace omitted for brevity)

@lvh works for me

the only problem I’m seeing is that the package has a validator dependency

which you’re not honoring

I’ll try and create a minimal reproducer. Thanks @anmonteiro 🙂

@lvh is this with :target :nodejs?

yep

that might be your problem

1 sec

oh yeah

here’s the problem: we stopped converting modules entirely when target is nodejs cc @dnolen

but this is not a good assumption, and we should only do that for node modules

yeah just came to the same conclusion here

🙂

I can put together a patch today

k

should be an easy fix

welp, I’m glad my problem was interesting at least — does that mean you no longer need an issue from me @anmonteiro ?

no, I can create one

I’ll make sure to include a test so we don’t regress 🙂

cool; thanks 🙂 If I understand what you’re saying correctly, this wouldn’t be a problem if the dep was added in package.json instead of being vendored?

yes

if you use npm-deps you’ll be able to require it

@lvh also simply because we don’t even attempt to run those through Closure if your target is Node

there’s a native require, we use that

I didn’t know about :npm-deps — is that equivalent to a package.json with the appropriate devRequirements for cljs and the dependency?

(if so that’s really convenient)

@lvh https://clojurescript.org/news/2017-07-12-clojurescript-is-not-an-island-integrating-node-modules

thanks!

let me know if you have further questions after reading that

I also wrote something even before we launched it officially: https://anmonteiro.com/2017/03/requiring-node-js-modules-from-clojurescript-namespaces/

@lvh @dnolen ticket + patch here https://dev.clojure.org/jira/browse/CLJS-2296

is there a ssl https client library ?

for clojurescript ?

how do I forbid the compiler from mangling things like window.speechSynthesis ?

@whou I think you can use ajax-request directly to https

@fabrao i am writing a clojurescript + planck script to ping a server that requires two way ssl, without the help from the browser, i think i will need to establish the trust somehow myself. Do you happen to know an example how to do this on clojurescript? or am i misunderstanding something here ?

something like http-client + openssl package in chicken scheme, but i really want to stay clojurescript

@whou pink @mfikes whether planck has a https client, otherwise you might want to try lumo since that is node which has https support

Yes, Planck has https support. You can, for example (planck.http/get "") It is based on libcurl

I’ve been using :npm-deps for two days now, and I’ve already encountered two google closure compiler bugs. Do you have any work-around technique for those bugs, or do I have to wait for upstream fix? I’ve been trying to fork and patch affecting npm modules, but that’s dubious… (FYI: I’m referring to https://github.com/google/closure-compiler/issues/2377 and https://github.com/google/closure-compiler/issues/2450)

@pawel.kapala 2560 has been fixed in master

@pawel.kapala I started keeping track of packages and their problems: https://github.com/cljsjs/packages/wiki/Closure-npm-Module-processing-status

@juhoteperi Thanks, I’ll have a look at the list. Regarding 2560, if it’s fixed in master. Silly question: how can I use it in lein project. Special dependency? SNAPSHOT?

run mvn -DskipTests at Closure-compiler dir and replace closure dependency with version 1.0-SNAPSHOT

Anyone know the best library to draw on files?

@juhoteperi the @atlaskit/* packages are fun too, eg @atlaskit/navigation references a whole bunch of packages with fun wrappers

all webpack’d packages with externals

@thheller https://unpkg.com/@atlaskit/[email protected]/dist/bundle-cjs.js ? Closure has a test case for one webpack UMD wrapper, so mayber this works?

--- node_modules/@atlaskit/avatar/dist/bundle.js:2
The define function must be called as a top-level statement.
--- node_modules/@atlaskit/blanket/dist/bundle.js:2
The define function must be called as a top-level statement.
--- node_modules/@atlaskit/field-base/dist/bundle.js:2
The define function must be called as a top-level statement.

I got a bunch of not working things

@juhoteperi @pawel.kapala FWIW the Closure Compiler team started publishing nightlies a while ago

so you can just add 1.0-SNAPSHOT and it’ll download the nightly snapshot from Maven

no need to build it yourself

@anmonteiro Perfect! Thanks!

@thheller dist/bundle.js looks like the browserified file, package.json webpack property points to dist/bundle-cljs.js

^ the errors above are because you’re setting AMD -> CJS conversion to true

disabling that will either 1. produce a better error, 2. convert the file successfully (no guarantees about working at runtime)

wait now there is a webpack property as well?

didn’t notice that

I think JS build tools just add those properties when ever they feel like it

"main": "dist/bundle.js",
"webpack": "dist/bundle-cjs.js",
"jsnext:experimental:main": "dist/es/index.js",
"ak:webpack:raw": "src/index.jsx",
"types": "src/index.d.ts",

hehe yeah it has a bunch of them

there's also a module property you can add that will have webpack automatically pull in the ES6 version of a dependency if its defined

@anmonteiro's patch adds option to Closure to configure this, so it will be possible to set it to ["webpack", "browser", "main"] or something in future

which is better for tree-shaking/dead code elimination

@chrisbroome yep, it’s awaiting internal submission: https://github.com/google/closure-compiler/pull/2598

Hmph, though ClojureScript side only uses browser, and main so not all fields will work automatically

nice

@anmonteiro seems almost signed off?

yeah, I think it’s going in if it passes Google internal testing

I don’t think I’ll have time tomorrow to sort out our own Closure artifact, but it’s still high on my queue

AFAICT they run every commit through their distributed build of every internal project

might be completely off-topic, but has the clojure/clojurescript team thought about targeting WebAssembly as a platform?

@chrisbroome people ask us that all the time 🙂

hehe

short answer is not anytime soon

yeah i haven't seen any mention of it

WebAssembly is suboptimal for GCed languages

in it’s current incarnation anyway

interesting

I was reading some about it the other day after webpack said that they got funding to make it a first class module target

it's also kinda cool that the textual representation uses s-expressions 🙂

yeah

I mean supporting WA for modules, seems fine, narrower scope

but I don’t think we need to even do anything

somebody write a WA DSL in ClojureScript go have fun

lol

good point