@juhoteperi: oh man, i totally didn't see that

oh, shit...

it looks a lot cleaner than the previous version...

we should have involved the expert here but then again, we started out with looking into hoplon first...

Any idea why is this happening? Im in the https://github.com/exicon/hoplon-multipage-example repo and trying [adzerk/boot-cljs "1.0-SNAPSHOT"]

onetom@retinatom ~/e/hoplon-multipage-example> boot -u
Retrieving boot-2.2.0.jar from 
#
#Sun Aug 09 13:59:55 HKT 2015
BOOT_CLOJURE_VERSION=1.7.0
BOOT_VERSION=2.2.0
#App version: 2.2.0
WARNING: update already refers to: #'clojure.core/update in namespace: clj-http.client, being replaced by: #'clj-http.client/update
onetom@retinatom ~/e/hoplon-multipage-example>

If I got a directory which doesn't have a build boot, I still get the warning, but it downloads more files:

onetom@retinatom ~/exicon> boot -u
#
#Sun Aug 09 14:07:19 HKT 2015
BOOT_CLOJURE_VERSION=1.7.0
BOOT_VERSION=2.2.0
#App version: 2.2.0
Retrieving worker-2.2.0.jar from 
Retrieving tools.namespace-0.2.11.jar from 
Retrieving aether-2.2.0.jar from 
WARNING: update already refers to: #'clojure.core/update in namespace: clj-http.client, being replaced by: #'clj-http.client/update

@juhoteperi, im trying the examples from the boot-cljs/next/README:

onetom@retinatom /t/x> mkdir src
onetom@retinatom /t/x> echo -e '(ns foop)\n(.log js/console "hello world")' > src/foop.cljs

onetom@retinatom /t/x> boot -s src -d adzerk/boot-cljs cljs   # which is understandable
org.sonatype.aether.resolution.DependencyResolutionException: Failed to collect dependencies for [#object[org.sonatype.aether.graph.Dependency 0x6b57696f "adzerk:boot-cljs:jar:RELEASE (compile)"]]
...

onetom@retinatom /t/x> boot -s src -d adzerk/boot-cljs:1.0-SNAPSHOT cljs
ERROR: No ClojureScript dependency.
Writing main.cljs.edn...
             clojure.lang.ExceptionInfo: java.lang.NullPointerException

i was installing the boot-cljs next branch using boot pom jar install maybe it's not the correct way to do so?

@onetom: You'll need to add clojurescript dependency to the project (1.7.28)

running it now:

onetom@retinatom /t/x> boot -s src -d adzerk/boot-cljs:1.0-SNAPSHOT -d org.clojure/clojurescript:1.7.48 cljs
Retrieving google-closure-library-third-party-0.0-20150805-acd8b553.jar from 
Retrieving closure-compiler-externs-v20150729.jar from 
Retrieving clojurescript-1.7.48.jar from 
Retrieving closure-compiler-v20150729.jar from 
Retrieving google-closure-library-0.0-20150805-acd8b553.jar from 
Writing main.cljs.edn...
             clojure.lang.ExceptionInfo: java.lang.NullPointerException
    data: {:file
           "/var/folders/9d/9j3pmgps4y54_dj8bwyngks40000gn/T/boot.user2769719586904406429.clj",
           :line 15}
java.util.concurrent.ExecutionException: java.lang.NullPointerException
         java.lang.NullPointerException:
             io.clj:  404
                         io.clj:  416
                         io.clj:  418
                                ...
   adzerk.boot-cljs.middleware/main  middleware.clj:   44
  adzerk.boot-cljs/eval363/fn/fn/fn   boot_cljs.clj:  107
  adzerk.boot-cljs/eval335/fn/fn/fn   boot_cljs.clj:   61
                boot.core/run-tasks        core.clj:  695
                  boot.core/boot/fn        core.clj:  705
clojure.core/binding-conveyor-fn/fn        core.clj: 1916
                                ...

is it a temporary requirement or shall we incorporate it into the README?

Not sure yet

im getting an error still:

onetom@retinatom /t/x> boot -s src -d adzerk/boot-cljs:1.0-SNAPSHOT -d org.clojure/clojurescript:1.7.48 cljs
Writing main.cljs.edn...
             clojure.lang.ExceptionInfo: java.lang.NullPointerException
    data: {:file
           "/var/folders/9d/9j3pmgps4y54_dj8bwyngks40000gn/T/boot.user2769719586904406429.clj",
           :line 15}
java.util.concurrent.ExecutionException: java.lang.NullPointerException
         java.lang.NullPointerException:
             io.clj:  404
                         io.clj:  416
                         io.clj:  418
                                ...
   adzerk.boot-cljs.middleware/main  middleware.clj:   44
  adzerk.boot-cljs/eval363/fn/fn/fn   boot_cljs.clj:  107
  adzerk.boot-cljs/eval335/fn/fn/fn   boot_cljs.clj:   61
                boot.core/run-tasks        core.clj:  695
                  boot.core/boot/fn        core.clj:  705
clojure.core/binding-conveyor-fn/fn        core.clj: 1916
                                ...

Looks like you don't have main.cljs.edn created, could be something broken in that case

I think I see the problem

i dont, because the examples doesn't tell me to do so...

The problem is in the implementation

ok, "happy to hear"

Oh, I had a old Boot binary. Doesn't -u take care of updating the binary?

no

thats the whole point of the separation between the libs and the bin, so they can evolve independently

@onetom: Pushed fix to next branch

since updating the binary is a lot trickier (from file system permissions point of view)

and it also changes a lot less often, so it makes sense to deploy it independently

I had still binary from rc1 😄

that proves the point

so, is this the right way to install the lib:

onetom@retinatom ~/r/boot-cljs> time boot pom jar install

Should be okay, boot build-jar is the proper way but it does exactly same

it worked; trying my "multipage" repo now

would u mind deploying this snapshot version to clojars, just for the sake of convenience?

Sure

Ah, bootlaces doen't like that the branch is not master

the multipage repo compiled, but i got

WARNING: Multiple .cljs.edn files found, you should use `id` option to select one.Elapsed time: 9.478 sec

Yeah, you need one cljs task per different build

i see

why don't u merge this next branch back to master?

just branch master into a release branch

i usually call it based on the version number and an x in to suggest which part of the version can change on that branch

so it would be v0.0-x in this case

with that i mean u can put it effectively into maintenance mode

and let's use this new implementation which relies on the newer cljs compiler the master

(need to pick up some friends now, bbl)

Hmh, does boot-hoplon create separate cljs edn for each .cljs.hl file?

I think separate envs for multiple builds will make that significantly slower

might make it slower, but it doesnt matter. it should work 1st, then we will optimize

Is there a way to check if a ns is already loaded?

Ohhhhh... Figwheel does just (load changed-clj-file) for changed clj files which contain string (defmacro

That can't work when macro uses something from other ns which changes

Or maybe it's reloading all changed clj files

I got macro reloading working with boot-cljs: https://github.com/adzerk-oss/boot-cljs/compare/macro-reload

@micha: Why are we manually dependency ordering the files 😄

Closure could already take care of that

goog.net.jsloader.loadMany(uris, opt_options) Loads and evaluates the JavaScript files at the specified URIs, guaranteeing the order of script loads.

Or not...

But I'm thinking it should be possible as if cljs_deps.js is loaded first, Closure knows the dependency graph

@juhoteperi: if goog can do it that would be awesome

i think it's still useful to add analysis info to the files in the fileset though, no?

i mean metadata about the relationship between js files and the corresponding cljs namespaces, etc

it would make it much easier to make more powerful tooling later in the pipeline after cljs

also dude, 💥 macro reloading

that's incredible news

@juhoteperi, @onetom i was thinking about the multiple build thing

i think we can run multiple builds in parallel in separate pods

we'd need to make the output-dir unique for each cljs.edn file though

we were thinking yesterday like foo/bar.cljs.edn --> foo/bar.out/ (`:output-dir`)

perhaps in boot-cljs we loop over cljs.edn files, creating a pod and output-dir for each

and compile in parallel

!m deraen

@micha: Perhaps the parallel running could be done as a special task

clone input files and merge output files

or well no cloning really required

why not just run in a loop with futures in a single cljs task?

It keeps cljs task cleaner when there is separate tmp-dir for each build... though then user needs to sift the files to keep them from colliding... hmph

@juhoteperi: i was thinking unique output-dir for each cljs.edn file

separate pods also, I think

like foo/bar.cljs.edn --> foo/bar.out/

definitely separate pods

so they can't collide

Btw. other thing, what do you think about boot-cljs versioning and clojurescript dependency? Even though boot-cljs is now using Cljs API it looks like for a while it will require the latest Cljs version because API changes and fixes

and those would all be in the tmp-result tmpdir

So boot-cljs should still follow cljs versioning

maybe

yeah we should just override their cljs version

maybe

@martinklepsch: I'm thinking Weißbier is the best recovery drink after sports

😄 I’m not so much into Weißbier, sometimes really enjoy cold Radler (Sprite + Beer) on hot days & after sports

i'm listening to Weezer at the moment

if they were a drink i bet it would be great

@martinklepsch: Yeah Radler is definitely nice on hot days

@juhoteperi: good work on the macro reloading

that's going to be really great to use

I don't like that it must depend on ns-tracker but it's simplest way I can see

new pod would be another but I fear it would be really slow

can we persist the compiler env analysis to disk?

you can use a separate pod for ns tracker?

i think we can

I don't think

(I probably don’t see the complete picture)

It contains lots of objects

they can't be persisted

david told me we could, i think

i think we only need one of the keys in there

the analysis part

i noticed cljs is writing cache files now, too

to the output-dir

and the compiler state is not even the part I'm worrying, just loading all namespaces would take some time

you could preload a lot of them, like the cljs compiler and such

but there is the unfortunate class leaking bug

so we need to fix that first

Also it would be hard to see when we have to refresh the pod

we don't want to do that when backend code is changed

btw, in boot-garden there has been an seemingly caused by :init when the task is being used multiple times: https://github.com/martinklepsch/boot-garden/pull/8/files

:init was triggering the class leaking?

If only clojure require had :reload-recursively option...

(require 'my.namespace :reload :all)

is that not what you want?

Is there such thing?

i believe so

Oh, :reload-all. I wonder what it does.

:reload-all implies :reload and also forces loading of all libs that the
  identified libs directly or indirectly load via require or use

Looks like it would be what I want. Maybe there are some problems with it or why do ns-tracker and c.t.n.refresh exist?

@micha: I’m not sure, it seemed to cause the exception pasted here: https://github.com/martinklepsch/boot-garden/issues/7#issuecomment-128425504

https://github.com/clojure/tools.namespace#reloading-code-motivation but I don't think ns-tracker does take care of those problems either

martinklepsch: yeah i'm not sure what the :init fn was trying to do there

deraen: i think ns-tracker tries not to reload all

@micha: you mean this? :init (fn [pod] (pod/require-in pod 'garden.core))

i think it tries to prune the things that get reloaded to just what's needed

martinklepsch: yeah

oh i guess it takes a long time to require the namespace

👍

that Stream closed exception is really strange

@juhoteperi: i may take a stab at the parallel compiler thing, is that cool?

@micha: Sure. What solution are you thinking?

i was thinking to first generate the output-dir path from the cljs.edn file path, like foo/bar.cljs.edn --> :output-dir "foo/bar.out"

this would ensure that all compiled js have their own output dir in the proper relative path

the output dir would be craeted in a single tempdir shared by all js builds

then loop through .cljs.edn files and create a separate pod for each one

then run the compiler in the pod in a future

Yeah sounds good. Check the new master and create new branch.

i see you merged next into master :thumbsup:

I'll probably revert Clojurescript dependency changes sometime

i did some TDD with onetom yesterday and i want to see how it goes, so i'll add tests too

Yeah tests are good

cool!

I did add some myself... but I think they didn't make it into the current version :d

@martinklepsch: Radler sounds amazing, i will try it

as a child i made and drank a lot of spezi

I did test :reload-all vs. ns-tracker

if the macro ns requires e.g. schema.core, compojure.core and other heavy namespaces, :reload-all takes about 3 seconds 😄 with ns-tracker the compilation takes only 0.2 seconds.

Are there any problems with ns-tracker?

Not that I can see

I guess it's fine. I just wanted to be sure if it's necessary.

You can put it in a pod if you want to right?

It's in pod yeah

@juhoteperi: should it be removed from build.boot then?

@martinklepsch: No, it's test scoped there. It's for testing the project.

It's not transitive.

testing the project as in real tests?

as in repl

😄

ah 😄

Helps a lot that I can load the impl ns in repl. For autocompletion etc.

anyone have recommendations for fixtures in clojure tests?

like useful libraries etc

i'm going to be writing files and stuff and compiling them with cljs

i guess the built-in boot stuff will be pretty good for that

Excited to see what you come up with regarding tests, Must be awesome with boot/filesets

going to try htmlunit to render html for testing shims etc

not sure how to test source maps though

I'm not sure if it's that useful to test on that level

i would like to be able to test for regressions there

because the configuration for the shim and source maps is pretty convoluted

Source map configuration is not that bad I think

But yeah, shim configuration is quite complex

i remember a lot of workarounds needed to be done with source maps

because cljs compiler was not expecting tempdirs

so it would compute paths wrong

Nothing special there anymore

otg

I'm thinking shim could be satisfactorily tested just unittests, fileset + params -> should set these options

or well, it shold also write the shim file. But no compilation required, I think.

we were using phantomjs with a dump.js script to debug that yesterday, it was pretty good

compiling cljs and loading the page is nice because it exercises all the things, like cljs_deps.js, goog/base.js, etc

any of which can end up in the wrong place if you touch the configuration code

or be overwritten with the wrong thing

if the paths are not correct

@colin.yates: Have you tested https://github.com/adzerk-oss/boot-cljs/issues/75 ?

@juhoteperi: https://github.com/adzerk-oss/boot-cljs/compare/builds-in-parallel?expand=1

it's alive