Seeing the HUD crash with this:

failed to parse websocket message {:type :build-failure, :build-config {:target :browser-test, :test-dir "resources/public/js/test", :ns-regexp "-test$", :runner-ns gondola.client.client-test-main, :devtools {:http-port 8022, :http-resource-root "public", :http-root "resources/public/js/test"}, :id :test, :build-id :test}, :report "------ ERROR -------------------------------------------------------------------\n File: /Users/tonykay/owsy/gondola-client/src/cljs/gondola/client/smartystreets.cljs:13:1\n--------------------------------------------------------------------------------\n  10 |   ([id] [::address-result id])\n  11 |   ([id field] [::address-result id field]))\n  12 | \n  13 | (defsc AddressResult [this {:keys [id]}]\n-------^------------------------------------------------------------------------\nSyntax Error at gondola.client.smartystreets:13\n{:cause clojure.lang.ExceptionInfo: Invalid arity for query {:expected 0, :got 1}}\nExceptionInfo: Syntax Error at gondola.client.smartystreets:13\n\tclojure.core/ex-info (core.clj:4739)\n\tclojure.core/ex-info (core.clj:4739)\n\tfulcro.client.primitives/defsc (primitives.cljc:3203)\n\tfulcro.client.primitives/defsc (primitives.cljc:3142)\n\tclojure.core/apply (core.clj:661)\n\tclojure.core/apply (core.clj:652)\n\tcljs.analyzer/macroexpand-1*/fn--3565 (analyzer.cljc:3286)\n\tcljs.analyzer/macroexpand-1* (analyzer.cljc:3285)\n\n--------------------------------------------------------------------------------\n  14 |   {:ident (fn [] (address-result-path id))\n  15 |    :query '[*]})\n  16 | \n  17 | (defmutation get-street-address [{:keys [id] :as params}]\n--------------------------------------------------------------------------------\n", :e clojure.lang.ExceptionInfo: failed to compile resource: [:shadow.build.classpath/resource "gondola/client/smartystreets.cljs"] {:tag :shadow.build.compiler/compile-cljs, :source-id [:shadow.build.classpath/resource "gondola/client/smartystreets.cljs"], :url #object[java.net.URL 0x71843d1d "file:/Users/tonykay/owsy/gondola-client/src/cljs/gondola/client/smartystreets.cljs"], :file #object[java.io.File 0x1367c8aa "/Users/tonykay/owsy/gondola-client/src/cljs/gondola/client/smartystreets.cljs"], :line 13, :column 1, :source-excerpt {:start-idx 9, :before ["  ([id] [::address-result id])" "  ([id field] [::address-result id field]))" ""], :line "(defsc AddressResult [this {:keys [id]}]", :after ["  {:ident (fn [] (address-result-path id))" "   :query '[*]})" "" "(defmutation get-street-address [{:keys [id] :as params}]"]}}} 

this is an exception thrown from a macro

The REPL experience is also less than ideal (using 2.1.11). I get things like Error handling response - class java.lang.ClassCastException: java.lang.Long cannot be cast to java.lang.CharSequence for expressions like (+ 1 1)

noob question 😞 Is it technically really hard to have figwheel with shadow-cljs ? In the latest defn-podcast think Bruce mentioned that he's planning on working on the internals of figwheel to make it more flexible.

@claudiu The hot code reloading is done by shadow-cljs, as is the compiling. Figwheel is both of those, but with the stock compiler. So, the short answer is “no, you don’t use figwheel with shadow-cljs”

@claudiu the question would be "why would you want to use figwheel with shadow-cljs?". what does figwheel provide that you do not get yet?

@thheller They seem to be two things that do the same thing. I get why it does not work out of the box, was just curious, it could theoretically work, with some updates to figwheel that made it more modular . 🙂

shadow-cljs is much more low level than figwheel so it really wouldn't be practical to use figwheel for anything

but again .. why would you?

if figwheel does something shadow doesn't I can take a look at adding it

it may look like they do the same thing but they really don't

figwheel is for development only and doesn't do anything for release builds

shadow-cljs started with release builds and later added the dev stuff

Ahh oky got it 🙂

I have never used figwheel so I might totally be missing something obvious ... so if you have something that shadow is missing over figwheel .. tell me.

So far so good. More of a noob question 🙂 since figwheel does hot reloading and is the 'default' in the clojurescript community, was just wondering if there technical aspects to why not work on integrating vs making the hot-reload from scratch 🙂 . Was listening to the defn podcast, Bruce said something about figwheel updates (can't remember exactly what), I immediately thought about shadow-cljs 🙂

he also said

> As a UI to ClojureScript the codebase is complex and complected.

shadow-cljs is a lot more complex due to being much more low level

integrating figwheel would probably be more work than just building it from scratch

hot reload is a tiny fraction of the code

the config validation stuff is nice but not usable as a lib

also shadow-cljs config is probably too different to use it

figwheel has had much more work done on the usability front and yes I'd love to use some of that

but I really can't

makes sense. Thank you for explaining the details 🙂

the rebel-readline stuff he is doing I can probably integrate at some point

that would be pretty awesome 🙂

@thheller following Tony comments, the only thing that I miss from Figwheel at this point is the REPL, it works really well there, but in shadow I've had problems with timeouts and sometimes not loading files at all

please report those things don't just ignore it

first time I hear of that 🙂

other thing that I miss, but I think it's not directly a shadow concern, is been able to start my REPL directly from Cursive, with Figwheel I can create a run with clojure.main pointing to a tiny script that starts figwheel, this boots up the compilation and with a REPL ready to use, with shadow I have to do it in 2 steps (open the process in a terminal and connect via nREPL on Cursive)

can you paste that figwheel startup file?

other than that, only love 🙂 and a special congrats on allowing external source paths, this is so good to work while developing multiple libs (I have a project where I'm having to use devbuilds for pathom, inspect and fulcro, without shadow it would be very painful)

sure, one moment

figwheel.clj

(use 'figwheel-sidecar.repl-api)
(start-figwheel! "devcards" "test")
(cljs-repl)

(require '[shadow.cljs.devtools.server :as server])
(require '[shadow.cljs.devtools.api :as shadow])
(require '[shadow.cljs.devtools.cli])

(server/start!)
(shadow/watch :devcards)
(shadow/watch :test)

put that into the figwheel.clj instead. change nothing else. done.

I guess I don't really know what (cljs-repl) does

since you are starting two builds which repl do you get?

(shadow/repl :devcards) would drop you into the REPL for devcards

figwheel seems to choose it kind by random (probably a (first) on a map), but you can type :cljs/quit, it takes you to a place where you can select which REPL to connect

what I personally do is just launch a lein repl

just tried running the script, got this:

https://github.com/thheller/shadow-cljs/blob/master/src/dev/repl.clj

objc[32799]: Class JavaLaunchHelper is implemented in both /Library/Java/JavaVirtualMachines/jdk1.8.0_111.jdk/Contents/Home/bin/java (0x10ce484c0) and /Library/Java/JavaVirtualMachines/jdk1.8.0_111.jdk/Contents/Home/jre/lib/libinstrument.dylib (0x10ced44e0). One of the two will be used. Which one is undefined.
Exception in thread "main" java.lang.RuntimeException: Unable to resolve symbol: es5>= in this context, compiling:(shadow/build/cljs_hacks.cljc:439:22)

(require 'repl) (repl/go)

uhm which clojurescript version are you using?

oh, sorry, so the project.clj version matters on this case, because of cursive classpath, correct?

yes

the lein file on this project was just a dump one for cursive, the real deps were in shadow, let me update and try again

problem here really is that cursive only supports lein

if there was a shadow-cljs startup option we wouldn't have those issues

I hope this gets better with deps.edn does shadow supports deps.edn?

yes

cool, is there something on the guide about it? I tried to find these days but I failed

its not really finished since clj isn't available on windows

but you can set {:deps {:aliaes [:foo :bar]}} in shadow-cljs.edn

shadow-cljs will then run the jvm via clj

cool 🙂

the script now ran

during the start it generated a lot of those:

Feb 16, 2018 11:48:15 AM clojure.tools.logging$eval27442$fn__27445 invoke
INFO: duplicate resource goog/locale/scriptToLanguages.js on classpath, using jar:file:/Users/wilkerlucio/.m2/repository/org/clojure/google-closure-library/0.0-20170809-b9c14c6b/google-closure-library-0.0-20170809-b9c14c6b.jar!/goog/locale/scriptToLanguages.js over jar:file:/Users/wilkerlucio/.m2/repository/org/clojure/google-closure-library/0.0-20170519-fa0499ef/google-closure-library-0.0-20170519-fa0499ef.jar!/goog/locale/scriptToLanguages.js
Feb 16, 2018 11:48:15 AM clojure.tools.logging$eval27442$fn__27445 invoke
INFO: duplicate resource goog/positioning/anchoredviewportposition.js on classpath, using jar:file:/Users/wilkerlucio/.m2/repository/org/clojure/google-closure-library/0.0-20170809-b9c14c6b/google-closure-library-0.0-20170809-b9c14c6b.jar!/goog/positioning/anchoredviewportposition.js over jar:file:/Users/wilkerlucio/.m2/repository/org/clojure/google-closure-library/0.0-20170519-fa0499ef/google-closure-library-0.0-20170519-fa0499ef.jar!/goog/positioning/anchoredviewportposition.js

classpath conflicts are fun arent they? 🙂

no, they are not 🙂

and the REPL experience seems limited, it seems to don't understand some REPL send commands:

[0:0]~cljs.user=> 3
3
[0:0]~cljs.user=> 4
4
[0:0]~cljs.user=> (in-ns 'com.wsscode.pathom.book.graphql.fulcro-network.demos)
Feb 16, 2018 11:51:36 AM clojure.tools.logging$eval27442$fn__27445 invoke
WARNING: client sent unknown msg {:type :repl/set-ns-complete, :id 7, :ns com.wsscode.pathom.book.graphql.fulcro-network.demos}
Timeout while waiting for REPL result.
[0:0]~com.wsscode.pathom.book.graphql.fulcro-network.demos=> (css/upsert-css "token-demo" RequireTokenApp)
#object[HTMLStyleElement [object HTMLStyleElement]]
[0:0]~com.wsscode.pathom.book.graphql.fulcro-network.demos=> {:map "with content"}
{:map "with content"}
[0:0]~com.wsscode.pathom.book.graphql.fulcro-network.demos=> [:book] Compiling ...
[:book] Build completed. (317 files, 2 compiled, 0 warnings, 0.78s)

the output is not been formatted, seems like something is missing to cursive to understand

it was better on the nREPL version

well its mixing repl output with log output

fighweel probably just doesn't log anything?

no, just an information at start, then no logs

(only responses)

WARNING: client sent unknown msg {:type :repl/set-ns-complete will be gone in next version.

cool 🙂

there is another one like this:

Feb 16, 2018 11:58:03 AM clojure.tools.logging$eval27442$fn__27445 invoke
WARNING: client sent unknown msg {:type :repl/require-complete, :id 10}

problem with the logs is that I really want them

could add an option though

makes sense

I like the logs too, would be nice if they could live together

this message seems to come now and then: Timeout while waiting for REPL result.

also fixed. the responses from the client weren't routed correctly in same cases

nice 🙂

[2018-02-16 22:42:59 - WARNING] TCP Port 9630 in use.
BindException: Address already in use (Bind failed)
	java.net.PlainSocketImpl.socketBind (PlainSocketImpl.java:-2)
	java.net.AbstractPlainSocketImpl.bind (AbstractPlainSocketImpl.java:387)
	java.net.ServerSocket.bind (ServerSocket.java:375)
	java.net.ServerSocket.<init> (ServerSocket.java:237)
	shadow.cljs.devtools.server.socket-repl/start (socket_repl.clj:161)
	shadow.cljs.devtools.server.socket-repl/start (socket_repl.clj:153)
	shadow.cljs.devtools.server/start-system (server.clj:157)
	shadow.cljs.devtools.server/start-system (server.clj:132)

error Command failed with exit code 1.
info Visit  for documentation about this

it exits when 9630 is used now?

shadow.cljs.devtools.server.socket-repl/start

socket repl port conflict

like I said .. fixed ports are annoying

agreed...

how about auto increment?

what is the point then? just use a random port

it defaults to a random port?

say you want a fixed port but that is taken by something else so you increment

your tool now connects to something else

ie. whatever claimed the port first

that seems strictly worse

I can offer you two things

create a shadow-cljs command that dumps the currently active ports

which doesn't seem all that useful given that you can just look at the files directly

or I can add a http://localhost:9630/ports or so page

either way ... whenever you use fixed ports you are going to run into conflicts

which is why I said to not use them or avoid them wherever possible

or you simly RUN shadow-cljs cljs-repl foo instead of connecting to any socket at all

not sure how it interacts with my toolchains, but at current I can accept random ports, as long as I can find the ports easily.

whats easier than looking at target/shadow-cljs/socket-repl.port?

like :socket-repl {:port 0}, it would be fine also.

it's printed in the logs

I can just see them.

seriously? you want to look at logs every time you start up a new server?

old habit from Webpack, did that a lot

look at the file and you'll never look for some random port

that seems like a horrible way to find out about a port but I can add that if you want

how do you use these ports currently?

reading them with scripts? and do what?

the shadow-cljs script looks at the cli-repl.port and connects to that

I made a tiny little helper for cursive to connect to the repl

https://github.com/thheller/shadow-cljs/blob/master/src/main/shadow/cursive_repl.clj

just remembered nREPL did that a lot...

I didn't understand why before. now it is...

nrepl always looks at .nrepl-port yes. but that belongs to lein and I don't want to conflict with that

is that common...solution in Java or in Clojure?

I haven't seen that before in JavaScript world

what is? writing ports to files?

yeah, ports in files

Unix systems?

seems pretty common. definitely not a java/clojure thing. probably a unix thing.

pretty similar to .pid files I guess

sounds quite similar.

well, just new stuff to JavaScript, we don't read files from browsers before. not seen in Node.js also.

anyway, I barely read those files, it's not in my toolchains yet. Reading them from logs is the most natural way for me.

does it need an update in the docs? https://shadow-cljs.github.io/docs/UsersGuide.html#http

what does?

oh, mistake...

no need to update, it's not socker REPL.

just they have different strategies of picking ports.

no they don't? they are all random by default unless you configure them

only the http port is sort of fixed

which was sort of a bad idea and should be removed to be honest

okay, only http port special.

consistency is preferred.

humans should never need to remember ports

thats just annoying to work with

but I still use localhost:8080 in browser history to develop my apps everyday.

unfortunately it will require a proper UI to make working with those easier

sounds like a groups of toolchains to me...

its certainly enough work for a couple of years 😉

can't wait that long~

somehow I want to see how you use shadow-cljs in you browser apps... just want to find out how mach difference out there

most of my habits are from Webpack.

might be different with many other tools.

I thought about writing that down somewhere. Probably would make it easier to understand why certain things are the way they are

that would be good.

can't use my work projects as examples though since I'm not allowed to post names

I need to rethink some of my toolchains also. Habits brought from JavaScript to Clojure, not all of them are good.

"work" means your current "job"?

demo projects are fine... todoMVC app?

I haven't set up a full demo project

well this is https://github.com/shadow-cljs/quickstart-browser

the workflow is simple shadow-cljs clj-repl. connect to that from Cursive.

launch the build in the cursive repl. edit code.

then you still need to demonstrate how to use inside the REPL

(shadow/watch :the-build)?

I think shadow variable is not in REPL by default?

maybe I'm wrong...

looks somewhat like that

REPL always on the side with build log output and eval results

browser on a 2nd monitor when building a website

I started REPL in iTerm3

Sublime Text for exiting raw code, that browser for Calcit Editor.

I do have the shadow-cljs clj-repl process running in a terminal

but pretty much never need to look at it

also need another browser tab for my real app.

I'm haveing a situation where I have two builds :admin and :public, with shadow-cljs watch admin it starts the http server for :public as well as :admin. Is it possible to prevent this?

no. but it only starts the http server not the build.

it starts the http server because I wanted to be able to use them with compile and release

ok then it's harmless, the ports were conflicting with other processes I was running

cool, thanks for clarifying!

yeah fixed ports are a PITA to work with

wish there was a proper cross-platform way to just use virtual hosts

ie. DNS not ports

yes, dns's would be neat.

off-topic, you seem the have mastered the urban dictionary more than anyone here on the slack, but nice to learn abbreviations like PITA along clojurescript 😄

hehe

I used Google to look up. English->Chinese dict not enough.

P.E.M.D.A.S

@wilkerlucio 2.1.13 should fix the REPL issues you had

also fixed an issue where the build would get stuff if a macro had errors, changes to that would not trigger a reload.

@thheller I don’t know if you did something specific to fix my protorepl issue, but if not, whatever you did in 2.1.13 fixed it

its nice and stable now

good to know

oops no i spoke too soon. there is definitely something weird going on here.

i’ll do a bit more investigation. its going to be annoying because things aren’t erroring, they are timing out

@thheller confirmed that changing ns and loading file are working now 🙂

is this expected behavior? when i connect to my app using shadow-cljs cljs-repl app, I get a timeout trying to switch to a library (and I confirmed that I’m connected to the browser by logging there first)

[3:1]~com.rpl.specter=> (js/console.log "***")
nil
[3:1]~com.rpl.specter=> (in-ns 'com.rpl.specter)
Timeout while waiting for REPL result.

the timeout causes bad things to happen in atom protorepl (a spinner shows up in the buffer and never goes away), but i wonder why i’m getting timeouts to begin with

is it possible that the namespace isn't loaded yet?

does it work if you (require 'com.rpl.specter) first?

oh wait you are already in that ns

but I can confirm that I get a timeout when switching to non existing namespaces

I tried to go back to terminal usage, now I'm getting:

shadow-cljs - config: /Users/wilkerlucio/Development/pathom-book/shadow-cljs.edn version: 2.1.13
shadow-cljs - socket connect failed, server process dead?

did you kill the REPL process?

yeah

hmm its supposed to clean up properly

delete the target/shadow-cljs/cli-repl.port file

is there a way to force the clean up, or check for it?

dunno I use a JVM shutdown hook to delete the port files

but maybe those don't trigger if you kill the jvm?

entering a meeting now, will check a bit later

@thheller other random question: it looks like you used asciidoctor to compose the user guide? i started using gitbook to do something similar but it doesn’t understand clojure. are you happy with asciidoctor?

kind of. I recently found slate which look a bit better but no idea if it supports clojure. https://lord.github.io/slate/

asciidoctor works well though. no complaints so far.

https://github.com/lord/slate

asciidoctor is probably better all around but slate definitely looks neat for API docs

gtg, bbiab

I just upgraded to the latest version, and the REPL experience in IntelliJ is better, but still gives some errors on output. If I use the “send to REPL” for something in a comment block like (defn g [a] (js/console.log :HI a)), I see the error Failed to parse client result: No dispatch macro for: '

Guessing someone’s parsing doesn’t like the return of def #'the-ns/g

@thheller would that be in shadow or Cursive?

actually, this is perhaps a better representation of the output:

(def h (f/debounce g 1000))
Failed to parse client result: No dispatch macro for: '
=> "#'the-ns/h"

so it does give back a string of what the def returned…everything “works”…it’s just a bit misleading/annoying

@thheller just FYI, running shadow-cljs stop fixed the start issue, after that it ran fine

omg .. I finally figured out how nrepl is supposed to be used ...

that took way too long to figure out

interesting, how much different it is from the current impl?

:printed-value true

I couldn't figure out why results from the shadow-cljs were always displayed as a string

so I though other tools must be reading the response value

but that causes all kinda of issues apparently

but instead I'm supposed to set :printed-value on the message and just pass the string along

https://github.com/clojure/tools.nrepl/blob/master/src/main/clojure/clojure/tools/nrepl/middleware/pr_values.clj

never knew that this middleware is used by default ...

yep ... appears to resolve all issues ...

so 2.1.14 on the way? 🙂

available now yes

I tested atom and cursive. both appear to be working correctly now

@thheller just tried, but got some problems, the nrepl is not responding, and I see this on the shadow-cljs terminal:

meh .. I guess that explains why its :printed-value 1 in piggieback .. no true

Some things are better, but I’m also seeing an issue

If I run something like a react DOM render by accident it causes an exception on the client, then I get no more REPL output. If this helps, it is the js console message:

encoding of result failed TypeError: Cannot convert a Symbol value to a string
    at Array.join (native)
    at Function.cljs.core.str.cljs$core$IFn$_invoke$arity$1 ()
    at Object.cljs$core$pr_writer_impl [as pr_writer_impl] ()
    at cljs$core$pr_writer ()
    at 
    at Object.cljs$core$pr_sequential_writer [as pr_sequential_writer] ()
    at Object.cljs$core$print_prefix_map [as print_prefix_map] ()
    at cljs$core$print_map ()
    at Object.cljs$core$pr_writer_impl [as pr_writer_impl] ()
    at Object.cljs$core$pr_writer [as pr_writer] () {$$typeof: Symbol(react.element), type: "div", key: null, ref: null, props: {…}, …}

I noticed any command I type on the REPL generate one of those

and :code is whatever I sent

I’m not seeing Wilker’s issue, though

that the error doesn't show up in the REPL is definitely not intended

mine seems to got totally borked, nothing that I send to the REPL is working

I fixed that I think

2.1.15?

coming in a sec

@tony.kay the symbol issue is this one https://dev.clojure.org/jira/browse/CLJS-1628

2.1.15 prints errors and should fix the Boolean issue

errors aren't yet source mapped, need to hook that up first

cool, sends are back on nREPL here

one more thing: when using figwheel repl Intellij is able to recognize it's cljs and switch the repl for it (there is combobox to switch), but on shadow it can't and I have to manually change it

hmm wonder what magic achieves that

hehe its just the printed message

To quit, type: :cljs/quit

but I printed To quit, type: :repl/quit

cursive seems to only detect the first one

I think worth asking on #cursive, colin surely knows how it happens 🙂

and Colin is really open to that sort of thing

I mean it works now

ah, you changed for :cljs/quit and it did the trick?

changing the message was all it requred

nice

super nice

you rock Thomas!

2.1.16 seems to work properly with Cursive now

yes, confirmed, thanks for such quick fixes 💚

if you find anything else let me know. I barely ever use the CLJS REPL so I never noticed these things

yeah, I guess after years of really bad REPL's on CLJS people just got used to live without it, but for me it's very important part of my workflow

I just rely on live reloading for CLJS. For CLJ I use the REPL all the time.

I can understand that. what I do is that while writing functions, I just keep doing small changes and running on the REPL, that's the fastest feedback loop I know about, specially when writing data transformations

Sooo... When are you going to support clojure? 😉

https://clojureverse.org/t/share-the-nitty-gritty-details-of-your-clojure-workflow/1208/10?u=thheller

the deftest workflow unfortunately doesn't work for CLJS in Cursive

otherwise I'd probably use that more there as well

@theasp probably never. enough tooling for that exists.