@thheller do you have any plans to support requring CSS in JS, like webpack allows? what do you use for this functionality?

I do have plans for that yes, nothing actually usuable yet though. I just use node-sass currently.

So after realizing that the browser test wouldn't do for me I'm exploring node-test

I think this is what react uses with jest when you create a react app from the scaffold

basically testing even my node modules

seems there isn't much literature about this

========= Running Tests =======================
SHADOW import error /home/user/project/.shadow-cljs/builds/test/dev/out/cljs-runtime/shadow.js.shim.module$semantic_ui_react$dist$es$index.js
/home/user/project/node_modules/semantic-ui-react/dist/es/index.js:2
import _Confirm from './addons/Confirm';
       ^^^^^^^^

SyntaxError: Unexpected identifier
    at new Script (vm.js:74:7)
    at createScript (vm.js:246:10)
    at Object.runInThisContext (vm.js:298:10)
    at Module._compile (internal/modules/cjs/loader.js:657:28)
    at Object.Module._extensions..js (internal/modules/cjs/loader.js:700:10)
    at Module.load (internal/modules/cjs/loader.js:599:32)
    at tryModuleLoad (internal/modules/cjs/loader.js:538:12)
    at Function.Module._load (internal/modules/cjs/loader.js:530:3)
    at Module.require (internal/modules/cjs/loader.js:637:17)
    at require (internal/modules/cjs/helpers.js:20:18)

@urbanslug thats ES6 code which doesn't work natively in node. there is some option to enable it but I forgot.

you are also probably testing browser stuff so :browser-test or :karma may be better suited

@thheller I want it to work on CI

:karma then

maybe a headless browser in a docker container idk

ok let me check out karma

it seemed to only talk about angular

and I'm doing react

reagent

Which linter do you use?

karma is generic, not tied to any framework. its just a headless browser basically.

I don't use a linter

Ah > karma is generic, not tied to any framework. its just a headless browser basically.

thanks m8

hmm I have a problem with this args: ["shadow.test.karma.init"], in my karma.conf.js

seems to be an arg to karma-run

I'm basically having it be app.core-test

Uncaught ReferenceError: app is not defined...

no thats wrong

you leave it as is

that is the main test runner

https://shadow-cljs.github.io/docs/UsersGuide.html#target-karma

this works. no need to adjust anything

like all test targets your tests are included via the :ns-regexp

ALL included tests are then executed by shadow.test.karma/init

https://github.com/thheller/shadow-cljs/blob/master/src/main/shadow/test/karma.cljs#L113

basically that just makes sure that some output is emitted that karma can understand

unless you write your own runner ns you do not touch args

same thing though Uncaught ReferenceError: shadow is not defined...

that probably means that your paths are wrong

check that basePath: 'out/demo-karma' and files match your paths

tree is like so

├── src
├── target
├── test

:output-to "target/ci.js"

I'm using your basepath target

what is your :output-to?

basePath: 'target',
        // The file itself
        files: ['ci.js'],

Ok I know where I messed up

basically the two combined must match whatever :output-to is

I had ::output-to "target/cli.js"

cli instead of ci

I don't use karma myself so its a wonder any of this works at all

docs could certainly be improved

Do you test your code?

some of it yes

in clojure though

ha! the test ran successfully

and without chromium

how? > in clojure though

I mean with chromium not chrome

lein test 😉

I don't have lein on my system 😄

I should write more tests but meh testing 😉

True! Tests are hell and nobody appreciates the effort that goes into them

which sometimes is a ton

yep its often more code to write the test than the actual code

especially UI testing

thats why I don't do it

basically I only test critical functions that work with pure data

I also installed what was here https://github.com/Day8/re-frame-test#running-the-cljs-tests-with-karma but not sure that it helped

this is my aim too but I literally needed a way to run the tests somehow in CI > basically I only test critical functions that work with pure data

and I didn't have one

is there a reason you chose a different dir for your karma tests output? than the usual resources/public...

When I figure out what I'm doing I'll try improve the docs. They could do with some clarification. Comments like don't mess with karma client args

Also the need to set CHROME_BIN=chromium or whatever it is I wonder how nicely this will play in CI

seems people install the whole damn thing https://github.com/Zenika/alpine-chrome/blob/master/Dockerfile#L10

slow tests are tests that people will refuse to run :thinking_face:

I can't make any recommendations regarding karma. I don't use it. I don't know how it works.

even less so in a CI env

well I'll let you know what I learn anyway

when I learn it...

90% there

you have this https://github.com/thheller/shadow-cljs/blob/master/karma.conf.js

could it be that hot-reloading multimethods doesn't work as expected?

depends on what you expect?

probably true. it didn't seem to execute the updated code when i changed the source of the dispatch function

until i Ctrl + C'd and restarted

and i expected it to execute it

without having to restart the build

ah yeah defmulti is basically defonce. you can only change the dispatch fn if its a reference to some other fn

(defmulti foo bar) works but (defmulti foo (fn [x] ...)) doesn't

hm... i had it as a reference

has anybody have this before with deps.edn?

shadow-cljs - starting via "clojure"
Error building classpath. nil
java.lang.NullPointerException

ok got it

...almost

strange

now I got it, crazy maps 😄

it's a clojure issue

error reporting is not helpful when there are mistakes in the deps.edn file

@urbanslug @thheller we currently have Karma running tests under shadow-cljs in a CI env. happy to help if you need it

are folks here using shadow-cljs in Cursive? I have a question 😄

I do

@thheller how do you setup the REPL? If the port is random, can you still use the nrepl file automagically for that?

(writing a post btw, so this will all be documented :D)

using .nrepl-port in non-lein projects

uhm, which Run configuration do you select?

clojure repl -> remote -> use leiningen port

and how do you lauch the nrepl server? within Cursive?

always in the terminal

just running npx shadow-cljs server in the build-in terminal and leaving that running

ok I though there was a way to run it inside, if external then I am already good 😄

thank you

@biscuitpants hmmm do you have a way to watch your karma tests?

@thheller question, is there a function for unwatching a build?

I see shadow/stop-worker

yep

but start-worker is priivate

watch starts the worker

ok, what about making it consistent? adding unwatch or something? even just an alias

unwatch isn't exactly clear what it means

could just mean stop getting log messages but leave worker running

right

will think about it more, don't even know if there is some usability issue there, just going through what I have written it felt ....itchy 😄

I think its fine. just need to document what a worker is since people technically only ever see watch

yeah

but is a (shadow/watch :the-build {:autobuild false}) still a watch? 😉

well, I was thinking if there is a use case where you don't want to watch the logging

just compile...

but maybe not

because logging gives you the errors

well thats just compile which doesn't start a worker 😉

worker is just more generic term for "process responsible for managing the build compilation"

watching anything is actually optional

yeah...if we had protocol server on the shadow server, one wouldn't need to check the terminal for errors I guess, but we are not there yet

so you actually have to watch

yep

we'll the pure REPLs (node-repl, browser-repl) don't watch at all so it is possible working without watch

just need to manually load stuff which is totally fine too

yep for browser you actually see errors in there

true, for now node, you will see evaluatiion errors

I like this actually, I think I will try this workflow

I also added shadow/watch-compile-all! for compile-on-key-press vs compile-on-file-save

ok nice

that is very cool actually

fits very well into my workflow

forgot who I added it for but it makes total sense

watch without autobuild

and bind the trigger to a key. so you can compile all changed files when pressing that key instead of all the time a file is changed

that's awesome

watch often is a bit over eager

agree

that key just send a REPL command?

yep I guess 😄

oh the above command, right

yeah just send (shadow.cljs.devtools.api/watch-compile-all!) to the REPL on keypress

simple setup with REPL commands in cursive

nice

but yeah I really like the run in terminal + remote REPL connect in cursive

when I get tired of nrepl I sometimes use clojure.main REPL with -m shadow.cursive-repl

that autoconnects to the socket repl instead of nrepl

oh

nice

I was trying just now clojure -A:dev:tree -m shadow.cljs.devtools.cli server

how you start the server doesn't really matter

can be embedded too

yeah I am documenting both..for nrepl this is also valid clojure -A:dev:test -m nrepl.cmdline (does not end up in shadow.user though

and the you need also --middleware I guess, so I am not documenting that one 😄

how do you connect to socket REPL?

does not seem to work here

nrepl is horrible to setup manually

Exception starting REPL: java.lang.ArithmeticException: integer overflow

hmm?

clj -m shadow.cursive-repl should just work (+ whatever aliases you need of course)

ok will try that

still haven't played with the cursive deps.edn support. not on EAP at the moment.

kk yeah well, I don't see much here actually

@thheller is there a difference between repl and nrepl-select?

repl delegates to nrepl-select when connected via nrepl.

nrepl-select just skips the check if you are in nrepl if you know you are in nrepl

I recommend using repl since it also works for socket repl and just looks better 😉

Ok cool then

@thheller so watch-compile-all! does not work if I am not watching?

because I want a worker and do not want to watch

but want watch-compile-all! 😄

yeah it still needs the worker. its meant to be combined with (shadow/watch :the-build {:autobuild false}) for now

you do want to watch though since it otherwise wouldn't pickup file changes

oh so in that case wouldn't start-worker make sense?

so triggering a compile wouldn't do anything

just look at the impl of watch

it also hooks up the output printing and such

I guess you could argue that naming of all of this isn't great

but I suck at naming and couldn't come up with something useful

yeah I think that is my main objection

watch-compile-all! is also meant for force-triggering a compilation for builds that may be stuck

yep I think I got it now

say a missing npm dep currently leaves the build stuck

just was missing (shadow/watch :script {:autobuild false})