Getting some errors when with doing multiple :release builds at the same time. Does not happen if I release builds individually: https://gist.github.com/mhuebert/66fe536fa2f6c96b5e09d33deb6b8617

attempting to use google’s material-components-web, getting errors like:

Oct 16, 2017 4:51:23 PM com.google.javascript.jscomp.LoggerErrorManager println
SEVERE: node_modules/@material/animation/index.js:143: ERROR - Parse error. '}' expected
export {transformStyleProperties, getCorrectEventName, getCorrectPropertyName};
^

Oct 16, 2017 4:51:23 PM com.google.javascript.jscomp.LoggerErrorManager println
SEVERE: node_modules/@material/base/component.js:19: ERROR - Parse error. '}' expected
import MDCFoundation from './foundation';
^

@mhuebert the release errors: is it possible you are building the same build twice? shadow-cljs release browser browser?

I only saw this error before when 2 compiles where trying to use the same cache

(:require ["@material/animation"]) works fine for me?

oh no wait, got the same error

checking

@thheller there are some overlapping namespaces, but different builds. it’s the :live, :bootstrap and :trusted builds here: https://github.com/mhuebert/maria/blob/master/editor/shadow-cljs.edn#L36

ah doh, yeah @material is ES6. I haven’t finished that work yet

I still need all the checkouts to compile Maria?

shouldn’t need the checkouts, they are just there to facilitate reloading during dev

however i was working on updating deps earlier today and it is possible there are omissions. i think i do indeed have too many independent projects

Could not find artifact cljs-live:cljs-live:jar:0.2.9-SNAPSHOT in clojars ()

can I drop down to a non-SNAPSHOT version?

deploying that now

done

0.2.9?

-SNAPSHOT

hmm can’t be found

or do you mean having a snapshot version is an issue

oh to be able to have git branch dependencies

I can just add the checkouts

i’ll try a 0.2.9 release

just released 0.2.9

there is a bunch more -SNAPSHOT versions in there

do those in general cause problems?

weird, ok those work it seems

The required namespace "pack.prosemirror-markdown" is not available, it was required by "re_view_prosemirror/markdown.cljs".

shadow-cljs release live bootstrap trusted

ok I think I identified the issue

it works if you run shadow-cljs server first

but that should not be a requirement, just an unhappy race condition. threads again. 😉

is there not an error if you run shadow-cljs server separately, and then shadow-cljs watch?

the server is in a bit of a weird state at the moment. there are several command that act as the server themselves if none is running

otherwise they all connect to the server

generally I have one server or clj-repl instance running

and all other commands run through it (automatically)

ok, now that seems to work

is doc/shadow-repl.md intended to explain how to set up a repl? it seems to go in a lot of paths

i wasn’t sure the most straightforward way to connect a repl with cursive

just configuring a remote nrepl server didn’t seem to work

in cursive just connect to “Remote NREPL` at localhost:7888

or whichever port you have entered in shadow-cljs.edn

:nrepl {:port 7888}

no need for further setup

assuming that one server instance is running, which can either be server or watch

doc/shadow-repl.md is not about setup no.

you maybe didn’t try this before you you can drive all of shadow-cljs via the REPL

shadow-cljs clj-repl
[1:0]~shadow.user=> (shadow/release :live)
[:live] Compiling ...

its the same as shadow-cljs release live

see (help) at the repl

geez thats outdated .. but still

shadow-cljs clj-repl is basically the same as lein repl

you get a REPL prompt but can connect to it remotely as well

nice

to get a CLJS repl for your build

shadow-cljs cljs-repl live (while watch is running)

in cursive (shadow/repl :live)

looks like you are connected? try to eval something

(help) or so

I think its just missing the prompt, connect worked

https://github.com/thheller/shadow-cljs/wiki/Cursive

CompilerException java.lang.RuntimeException: Unable to resolve symbol: help in this context, compiling:(null:1:1)

doh 😛

(shadow.cljs.devtools.api/help)?

I don’t use nrepl much these days

ah, the wiki.

ok, now i understand. ’tis working.

(shadow.cljs.devtools.api/repl :live)

this switches the current REPL to CLJS, so every eval after that should happen in CLJS

the UI does not properly detect this so you might need to switch manually in Cursive clj/cljs

the dev command looks handy

so for es6 - i should keep doing rollup/webpack for now?

I did not expect people to put ES6 on npm

I thought they all re-packaged it first

I have a few ideas how to handle that but didn’t implement them yet

maybe this particular package is better handled directly with Google Closure as it is a google thing

yah

IIRC in shadow there is an option for that, it would then process all deps through closure?

the issue with that is that the closure thing is all or nothing kind of deal

right.

but not all npm deps survive closure

:js-options {:js-provider :closure} but that basically never works 😛

I think material was one I tested which worked though

i think anything made by google is fully annotated

firebase probably is too

yeah material-components-web works

could closure be a :resolve option?

not really as it rewrites the code quite heavily

it might work for a few cases

really depends on if you rewrite the dependencies as well or not

eg. if material-components-web doesn’t depend on anything you’ll be fine

but if it depends on something and rewrites that

but something else is not rewritten but also attempts to use that dep

boom

ah i see

the plan currently is to use babel to transform ES6 code

since that is what most other libs on npm use

ie. if you use material-ui you get ES3 which was transformed by babel and includes some of the babel-runtime

if I use closure to rewrite that you have the babel runtime from some package and the closure runtime for others

thats not so great

babel seems to be the most common ground and is actually fairly good

since I also get jsx support for free

I already did a few basic tests just didn’t get to implement it yet

in the short term, is there a way for me to expose my rollup’d pack of material-components-web in a way that would let me require things like how cljsjs/global-exports work?

:js-options {:resolve {"material-components-web" {:target :global :global "MDC"}}}

should work?

:file might also work if they still ship a bundle?

{"material-components-web" {:target :npm :require "material-components-web/dist/material-components-web.js"}}

there is also a file for :require-min

:js-options do not flow through upstream though, do they? so then whatever consumes re-view-material also has to.. i’m not sure, copy over a JS file somewhere?

i would rather not include their whole bundle

as i don’t use a lot of it

hmm yeah its tricky

I really don’t want to support cljsjs style packages because the chance of conflicts is so damn high

well its only an issue until I sort out the ES6 issue … I will do that tomorrow 😉

😉

also - minor thing - getting the server to emit the URL would avoid some confusion others had trying to get maria running for dev

the http url? yes I forgot about that

HTTP server for build :browser available at: 
HTTP server for build :bootstrap-host available at: 

suggestions? or good enough?

I think that’s great

sometimes you really have to wonder what google is doing

const el = windowObj['document']['createElement']('div');

don’t they know how to use externs? 😛

hah

finished the ES6 support 😉

it might not work at all [email protected]

its only 50% of the implementation I had planned initially but this should work, tested with "@material/animation" and it compiles

oh doh, require('@material/checkbox/dist/mdc.checkbox'); they do ship commonjs as well

well .. screw that 😉

how do you use this stuff from JS anyways?

haha doh, stay away from 19

running everything through babel was not the intent

material-components-web is designed primarily as a low-level framework which requires ‘adapters’ to actually use, so i have a lot of ugly code which requires these namespaces and then maps them to react components appropriately

react?

why not use material-ui instead then?

or there was another react version for material I think

hmm yeah as expected the babel transform is gonna be dog slow the way I did it now

last i checked, which was not recently, material-ui did not use the latest version of material-components-web and was also very very heavy

looking now, yes, material-ui is also not made by google

whereas material-components-web is actually maintained/developed by google and intended to be a production quality reference implementation

try [email protected]

as expected shelling out to babel per file is way too slow

gonna get that fixed next

but it should works as a test

material-ui also did this inline-styles thing which was simultaneously slow and difficult to integrate with a project. however again i haven’t looked at it in months 🙂.

ok

I haven’t used either myself. just thought it was weird to use a web-components lib with react

but if it works why not

I added a test and it seems to work, just missing the styles

(if (contains? #{"es3 es5"} js-language) spot the bug 😉

it actually seems to work way better than expected, the babel transform that is

just need to fix the performance problem

let me know if it works … afk for a bit

oh it’s not web-components per se

they did a bunch of the hard css/js stuff, and then you have to write an ‘adapter’ that maps it to whatever, can be web components or can be react

when i wrote re-view-material, material-components-web was at version 0.14 vs today 0.22, i don’t think they did the cjs dist back then

2.0.20 worked 🙂

slow but worked

1st build 57s, quite and built again, 38s

advanced compilation 68sec

got to do some externs debugging

really? damn shouldn’t need externs

38sec? what are you building?

68sec advanced is pretty darn slow

where do you need externs?

I think I found one thing I need to add to the externs generator

yeah … dammit …

_createClass(MDCCheckboxFoundation, [{
    key: 'init',
    value: function init() {
      this.adapter_.addClass(_constants.cssClasses.UPGRADED);
      this.adapter_.registerChangeHandler(this.changeHandler_);
      this.installPropertyChangeHooks_();
    }
  }, {
    key: 'destroy',
    value: function destroy() {
      this.adapter_.deregisterChangeHandler(this.changeHandler_);
      this.uninstallPropertyChangeHooks_();
    }

    /** @return {boolean} */

  }

babel rewrites class MDCCheckboxFoundation this way

it doesn’t detect that key: 'init' is the property we need

need to think about how to fix that

I really did underestimate how much dynamic stuff might generate properties

maybe I should also extern all calls done in the js? not just the assigns

may have spoken too soon. this is really a tangle of stuff that has JS deps and I have to sort some out before i know if there are any real externs issues

i don’t think 100% should be the goal, we can’t get around the dynamic stuff

i’ve thought about that before, if one could detect any access of globals that aren’t part of your build

make that stuff string requires

don’t know

yeah starting to believe its better to properly use :infer-externs and rely on that instead

the dynamic generation is ok but just doesn’t work in many cases

what kind of errors do you see?

all externs issues should be pretty easy to track down using shadow-cljs release build --pseudo-names

so i am building the http://re-view.io website, which contains examples of all the re-view-material components as well as re-view-prosemirror. i had forgotten that re-view-prosemirror also has a deps.clj foreign-lib that i have to include separately as a script tag

foreign-lib because its not on npm or custom build?

custom. but this is just another lib that was written in the old style (webpack of a .js file) before we could require stuff from node_modules

so my goal now is rewriting all of this stuff to not require these old bundles.

ah right. this thing https://github.com/braintripping/re-view/blob/master/re_view_prosemirror/src/js/pm.js

yeah, there is that and also one pmMarkdown

which depends on it

going to get rid of those now

yeah thats the reason why foreign-libs is such a bad idea

well not super bad since prosemirror doesn’t have many dependencies but still

is support for checkouts something that will be hard to add?

just add the :source-paths, should be enough

Is it possible to have different dependencies in different build. In fact I create two application : node + js. And some sources are common to both.

https://github.com/braintripping/re-view/blob/master/re_view_prosemirror/example/re_view_prosemirror/example.cljs#L10-L12

I see you already did externs annotations

Electron, in fact...

@ivanpierre put it all in one source-path, the build config will determine what goes into a build

I put together an example for electron a while ago

https://github.com/thheller/shadow-cljs-examples/tree/master/electron

Yes, That's what I do. But I had some conflicts with reagent and node.

you might need to use the latest alpha release of reagent

I did ,)

what problems did you see?

Well, perhaps a bug from mine... Il grab a little more... If it doesn't work I'll give you precise behaviour.... 😉

you might run into trouble if you include react-dom on the node side of things

would need to split that into a separate namespace that is only included for the web stuff

@thheller yeah I did a lot of those actually. but they were all based on the idea that my module wrote all of this stuff to a couple global objects, pm and pmMarkdown

in my version of :infer-externs you currently only need to do ^js thing

@thheller i just got

failed to start service
{:id :dev-http}
ExceptionInfo: failed to start service
	clojure.core/ex-info (core.clj:4725)
	clojure.core/ex-info (core.clj:4725)
	shadow.runtime.services/start-many/fn--194 (services.clj:135)
...
Caused by:
BindException: Address already in use

@thheller that sounds much simpler. so just ^js some-symbol-from-javascript-land

yes, skip the “typing”

if it works, i like this much better

well it would still be better to use the fully qualified thing but who wants to type annotate other people’s code

doh … will fix http

hmm wait

no that should not be an issue

what did shadow-cljs watch thing say? connected to server?

or did it start a new one?

second line should say connect

shadow-cljs compile browser
shadow-cljs - config: /Users/zilence/code/shadow-cljs/shadow-cljs.edn version: 2.0.18
shadow-cljs - connected to server

if it didn’t connect the target/shadow-cljs files may been deleted after starting the server

lein tends to do that

https://github.com/thheller/shadow-cljs/blob/master/shadow-cljs.edn#L4-L5

earlier today, it worked in another project. i have :lein true in this one

somehow now it is connecting

am not sure what changed.

regardless, starting the http server should not be a critical error

I’ll just print a warning

if website requires prosemirror, which has {:npm-deps {...}} in deps.cljs, should shadow-cljs npm-deps inside the website project have the effect of installing those npm modules?

hehe … maybe? it should yes

can do a quick test

I didn’t find any package that had deps.cljs when testing so its basically untested

I can test with re-view-prosemirror now that I know one

well, re-view-prosemirror may not be the best example at the moment as it is in transition

tested and it worked yes

Ok : /Users/ivan/.m2/repository/cljsjs/create-react-class/15.6.0-1/create-react-class-15.6.0-1.jar deps.cljs refers to file not in jar: cljsjs/create-react-class/production/create-react-class.min.inc.js {:tag :shadow.build.classpath/deps-cljs, :source-path #object[java.io.File 0x5da7cee2 "/Users/ivan/.m2/repository/cljsjs/create-react-class/15.6.0-1/create-react-class-15.6.0-1.jar"], :name "cljsjs/create-react-class/production/create-react-class.min.inc.js"} ExceptionInfo: /Users/ivan/.m2/repository/cljsjs/create-react-class/15.6.0-1/create-react-class-15.6.0-1.jar deps.cljs refers to file not in jar: cljsjs/create-react-class/production/create-react-class.min.inc.js problem with create-react-class. it search the jar in production and in .m2 there is only common and developement... But it's the last release. I'll go and see.

ah damn thats still not fixed

^^

[reagent "0.8.0-alpha1" :exclusions [cljsjs/create-react-class]]

use that

Thanks

I now realize that I’m far too strict in validating the deps.cljs files … 🙂

a warning should be enough

@mhuebert one thing missing for the npm-deps things is support for yarn, its currently just calling npm install --save thing@version

is there a way to add an annotation for a function call? eg. (.serialize this) where serialize shouldn’t be rewritten

or… does (.serialize ^js this) have that effect

yes, well it should. did not test.

confirmed, works

@mhuebert btw I cannot guarantee that CLJS will accept the plain ^js tag, so that might not actually work with anything but shadow-cljs

plain ^js has always been allowed, just not in combination with :infer-externs I think

weird Closure error only in :advanced -

-> Closure - Optimizing ...
IllegalStateException:  [GETPROP 14 [length: 31] [source_file: re_view_prosemirror.core.js], EXPR_RESULT 14 [length: 66] [source_file: re_view_prosemirror.core.js]]
	com.google.common.base.Preconditions.checkState (Preconditions.java:721)
	com.google.javascript.rhino.Node.replaceChild (Node.java:852)
	com.google.javascript.jscomp.AggressiveInlineAliases.inlineGlobalAliasIfPossible (AggressiveInlineAliases.java:309)
	com.google.javascript.jscomp.AggressiveInlineAliases.inlineAliases (AggressiveInlineAliases.java:162)
	com.google.javascript.jscomp.AggressiveInlineAliases.process (AggressiveInlineAliases.java:111)
	com.google.javascript.jscomp.PhaseOptimizer$NamedPass.process (PhaseOptimizer.java:304)
	com.google.javascript.jscomp.PhaseOptimizer.process (PhaseOptimizer.java:230)
	com.google.javascript.jscomp.Compiler.performOptimizations (Compiler.java:2457)
	com.google.javascript.jscomp.Compiler$3.call (Compiler.java:834)
	com.google.javascript.jscomp.Compiler$3.call (Compiler.java:830)
	com.google.javascript.jscomp.CompilerExecutor.runInCompilerThread (CompilerExecutor.java:128)
	com.google.javascript.jscomp.Compiler.runInCompilerThread (Compiler.java:862)
	com.google.javascript.jscomp.Compiler.stage2Passes (Compiler.java:829)
	com.google.javascript.jscomp.Compiler.compileModules (Compiler.java:778)
	sun.reflect.NativeMethodAccessorImpl.invoke0 (NativeMethodAccessorImpl.java:-2)
	sun.reflect.NativeMethodAccessorImpl.invoke (NativeMethodAccessorImpl.java:62)
	sun.reflect.DelegatingMethodAccessorImpl.invoke (DelegatingMethodAccessorImpl.java:43)
	java.lang.reflect.Method.invoke (Method.java:497)
	clojure.lang.Reflector.invokeMatchingMethod (Reflector.java:93)
	clojure.lang.Reflector.invokeInstanceMethod (Reflector.java:28)
	shadow.build.closure/compile-js-modules (closure.clj:874)
	shadow.build.closure/compile-js-modules (closure.clj:865)
	shadow.build.closure/optimize/fn--24464 (closure.clj:1138)
	shadow.build.closure/optimize (closure.clj:1124)
	shadow.build.closure/optimize (closure.clj:1115)
	shadow.build/optimize (build.clj:288)
	shadow.build/optimize (build.clj:280)
	shadow.cljs.devtools.api/release* (api.clj:197)
	shadow.cljs.devtools.api/release* (api.clj:183)
	shadow.cljs.devtools.cli/do-build-command (cli.clj:29)
	shadow.cljs.devtools.cli/do-build-command (cli.clj:25)
	shadow.cljs.devtools.cli/do-build-commands/fn--30494/fn--30495 (cli.clj:66)
	clojure.core/binding-conveyor-fn/fn--5297 (core.clj:2027)
	java.util.concurrent.FutureTask.run (FutureTask.java:266)
	java.util.concurrent.ThreadPoolExecutor.runWorker (ThreadPoolExecutor.java:1142)
	java.util.concurrent.ThreadPoolExecutor$Worker.run (ThreadPoolExecutor.java:617)
	java.lang.Thread.run (Thread.java:745)

(oops i thought that would collapse into a snippet)

what the heck

AggressiveInlineAliases good name

this is the file. but i don’t see a line number in the error? https://github.com/braintripping/re-view/blob/master/re_view_prosemirror/src/re_view_prosemirror/core.cljs

nothing immediately obvious, checking

uhm there is no shadow-cljs.edn?

i’m only using it from re-view/website

can add one

thats empty?

https://github.com/braintripping/re-view/blob/master/website/shadow-cljs.edn

how come you def everything? (def EditorState prosemirror-state/EditorState)?

could also do :refer in ns instead

just to not break other libs that were referencing things the old way

previously I did all the messy JS object reading here, so that my other code wouldn’t have to do all kinds of (.-someProp js/pmMarkdown) junk

ah ok, shouldn’t be an issue. just checking

i will be able to rewrite a lot of stuff in a nicer now that i can require directly from node_modules

yeah some of the interop code before was really messy

be back in a bit

i can clean this up into an independently compiling thing if you want first

hard to debug as it is now

just need something I can call to reproduce

the config in website should be enough but its empty

re_view_prosemirror.core.keymap = re_view_prosemirror.core.keymap;
re_view_prosemirror.core.commands = re_view_prosemirror.core.commands;
re_view_prosemirror.core.model = re_view_prosemirror.core.model;
re_view_prosemirror.core.history = re_view_prosemirror.core.history;

this is odd

(def keymap keymap)
(def commands commands)
(def model model)
(def history history)

yeah, those are causing the issue

resolves to itself

probably need to rename the :as in the ns if you want to keep those def

I see

seems strange that that would cause an error

x = x

(let [x x])

anyway i can get rid of those and change the references to use the node_modules stuff directly.

weird. somehow it is not picking up componentDidUpdate

$map__37705__$1_p__37704$$.$componentDidUpdate$ is not a function

despite having react externs in one of the dependencies (re-view-hiccup)

and thought that componentDidUpdate would have been in the generated externs, but didn’t see it there either

weird indeed, ShadowJS.prototype.componentDidUpdate it is in my test file for sure