@oahner @thheller I was stringifying the data further down the stream. Sorry for the noise.

Pretty sure it's working. Transfer speed hasn't improved much, but at least the main thread doesn't hang trying to convert the results to the array.

So I'm building a cellular automaton to show tradeoffs of using web workers in this agent way. I've got the automaton processing on both the main thread and the web worker, showing the difference between the two. Pretty fun little exercise.

Next bit to implement, parallelization of a 1D CA. Not looking forward to that 🙂

I've got an idea to process the whole thing in triangle patches, which will allow parallelization for a particular number of consecutive rows.

In two passes

how would i achieve this in clojurescript?

(defrecord ReversibleTie [a b])

(extend-protocol Reversible
  ReversibleTie
  (my-reverse [tie] (->ReversibleTie (:b tie) (:a tie)))
  js/String
  (my-reverse [s] (->> s
                    reverse
                    clojure.string/join "x")))

who defines the Reversible protocol?

@noisesmith I was sort of hoping clojurescript. But i wouldn’t be suprised to learn that it doesn’t exist. Part of what i was confused about was how to figure out what protocols current existed in clojure/clojurescript.

so, when implementing Reversible, you can only implement methods that are actually defined on that protocol

if it even exists...

even if clojurescript defined a Reversible protocol, it wouldn't have a method called my-reverse

but you are free to define Reversible, and call a method my-reverse

ouch, yea. That makes sense. How do you go about finding out what protocols exist? are they listed somewhere.

I don't know if there's good documentation, but there's a namespace that has them in cljs - one moment

https://github.com/clojure/clojurescript/blob/master/src/main/cljs/cljs/core.cljs#L449 - these are the main ones at least

@drewverlee here's the one https://github.com/clojure/clojurescript/blob/master/src/main/cljs/cljs/core.cljs#L643

awesome! I think i get it, lets see if i extend the actual protocal then…

hi. any schedule on porting the spec.alpha for cljs too?

is there perhaps a way to build multiple disparate :main cljs builds in parallel with cljsbuild? right now we’re doing cljsbuild id id2 id3 which does them in serial, which is much faster than do cljsbuild id, cljsbuild id2. but i wonder if perhaps we can build multiple entrypoints in parallel in a single cljsbuild call somehow? cc @dnolen

@robert-stuttaford in theory with :parallel-build true doing multiple builds in serial should be fastest since each successive build can re-use the analyzer data from the previous

if they are isolated each must compile everything on its own which is more work

but FWIW I think it is safer to not share analyzer data between builds, so do should be "safer"

cool, we’re doing :parallel-build true + one process already, so i guess we’ll just add cores until it’s as fast as it can be for now 🙂

thanks @thheller

hmm you are probably stuck waiting for closure? that part could definitely be done in parallel

almost certainly, yes

we’re building about 6 apps, and some of them have an :advanced + :pseudo-names true variant

hey! o/

is there a way to use apply with .methodName interop in ClojureScript?

(apply .methodName object args) doesn't compile

(apply (.-methodName object) args) ?

@savelichalex ah, of course! thanks

@robert-stuttaford I’m sure you could do it manually with your own script - cljsbuild used to have that feature but it I believe the problem was that it tried to share the compiler state atom across very different builds which isn’t a good idea

I'm aware of libraries like Om / Reagent / Re-Frame and why I should use them. However, assume I'm bent on directly manipulating the DOM. I want to keep hiccup style notation for creating DOMs (instead of directly using strings myself). Is there a library that will let me directly modify the DOM with hiccup like notation ?

@robert-stuttaford still @thheller might be right here and you’d have to do some testing to see

does hiccup work in cljs, or do I need some other library like crate or hiccups ?

http://respo.site ?

Hi, problem with building -

Circular dependency detected, cljs.core -> cljs.core

[org.clojure/clojurescript "1.9.521" :scope "provided"]

until you remove build directory

in build directory there are two files core.cljs and core.cljc (which requires cljs.core.cljs)

@nicola is it possible that the build directory is also a source path?

@qqq https://github.com/teropa/hiccups

I have never used it but it might be useful to u

@theller

{:source-paths ["src/cljs" "src/clj"]
                    :cljsbuild
                    {:builds
                     {:ui {:source-paths ["srcs" "env/prod/cljs"]
                           :compiler
                           {:main "ui.prod"
                            ;; :foreign-libs []

                            ;; :externs []

                            :output-to "build/js/ui.js"
                            :output-dir "build/js/out"
                            :optimizations :advanced
                            :pretty-print  false}}}}

@qqq you could try https://github.com/thheller/shadow specifically the shadow.dom namespace. I have been using that quite some time. not well documented but pretty straightforward.

(dom/append [:h1 "hello world"]) (append to js/document.body)

that looks correct although "srcs" looks a bit suspicious since you already have src/clj and src/cljs?

This src/clj and src/cljs is obsolete - will try to remove

did not help - second build always cljs.core -> cljs.core

There is some conf/task to lein copy my Index from resources to target?

cljs.core.cljc
(:require clojure.walk
            clojure.set
            [clojure.string :as string]
            [cljs.compiler :as comp]
            [cljs.env :as env]
            #?(:clj [cljs.support :refer [assert-args]])
            #?(:cljs [cljs.core :as core]) <- HERE
            #?(:cljs [cljs.analyzer :as ana]))

thanks @dnolen !

Are there any common pitfalls which break Closure compilers cross-module code motion? All namespaces and dependencies I don’t explicitly include in some module’s :entries end up in the :cljs-base module, though the clearly could be.

separated :output-dir and :output-to - works!

@kommen the :modules implementation is buggy IMHO. https://dev.clojure.org/jira/browse/CLJS-2011

you can try https://github.com/thheller/shadow-devtools/wiki/ClojureScript-for-the-browser if you want 😉

but if you want to rely on code motion: any "side-effect" won't be moved

that includes hash-map construction, defmethod, anything that swaps into an atom, etc

@thheller ok, thanks. good to know…

that explains things for us I guess

@thheller but does a side-effect than cause the whole namepsace not being moved?

no each def/side-effect individually

unless you are talking to something that wasn't moved

(def x (atom {})) nothing accessing x will be moved

ok, that’s definitely not the case for all of our code

do you have an example of something you'd expect to be moved?

for example my :editor module has editor.core in its :entries. only editor.core or namespaces required by it require reagent. but reagent still ends up in the base module

yeah things like that shouldn't rely on code motion at all as we can figure that out just by looking at the requires

I think defn with multiple arities are never moved as well

@thheller maybe I can make a sample repository demonstrating the issue tomorrow

this pretty much covers everything already https://github.com/thheller/cljs-issues/tree/master/modules 😉

suppose you were writing an app, where all it is is 100,000 circles (in svg) moving around randomly; for an app like this, would it be more appropriate to (1) use react / om / re-frame / some type of diffing or (2) just manipulate the dom directly?

@qqq I suppose with that many circles you want canvas (not svg) but react would not be ideal I think

@thheller ok, thanks!

@qqq I would strictly break such app into two parts: 1) “engine” 2) “logic” => I would write the engine in plain js with direct DOM manipulation (performance sensitive code), and the logic in clojurescript (as a scripting language, not performance sensitive code) - the engine api should be suitable to expected usage, e.g. submitting js array of data for each frame, or provide just selective updates, it depends on usage

darwin: but you also would use canvas rather than svg, and "render each frame from data" rather than "just update fieds/properties of what changed" ?

cannot tell, this depends on how many objects will be changing on average

if you are going update most of them, then go with canvas

if you are going to do some ridiculous amount of circles all moving every frame then I would consider webgl instead

you can then offload that work to GPU and upload only changed positions each frame

easy task for GPU

okay, so I lied a bit; I'm not rendering circles; I'm rendering lots of pieces of text

I considered GPU, but then I run into problem of text rendering

in particular, pure textures = blurry, signed distance fields = slightly better,

and vector / rendering bezier curves = complicated

otherwise, I'd happily jump to WebGL

Distance fields look ace

imho

some guy just wrote a live coding env in a web browser and replaced all of the UI components with shader blobs, it sounds mad (it is) but the text looks ace zoomed in

let me find it

all of this is web gl https://youtu.be/yOiuJ_Gll_E

great presenation

and you can try it out here https://makepad.github.io/makepad

That's pretty bad ass

in cljs, after I create a new pop up window with js/window open ... , I seem to have race conditions dealint with accessing divs of the window

do I need to somehow use a on-window-load of thepop up window as well ?

a JS lib is passing back an Object of data to my cljs. Is there a cljs method to show me the content of that data?

in Ruby-land I just had to object.inspect

(comp pr-str js->clj)

thx

hmm, so the value being sent to me by JS logs to the console as

{:source [{source /dzi/madonna.dzi, eventSource #object[Object [object Object]], userData nil}]}

I want that 'madonna.dzi' value but attempts at destructuring it are unsuccessful

time to get figwheel and cursive working together so that I can actually have a REPL and see values, I guess.

do popup windows inherit css from parent

or do they need to have their own css includes ?

#off-topic? lots of css slack channels.

@mobileink : fair enough, not everything done in cljs is on topic for #cljs

just bustin your chops. 😉

I had to go. But I think text rendering can be done with webgl without blur. This experimental project is state of the art IMO: https://github.com/evanw/sky

wow, loving cljs. interop with js is so easy.

@darwin: I think that's the same person behind http://thetamath.com/app/abs(x%5E(2)-1)=abs(y%5E(2)-1)

and the author of https://medium.com/@evanwallace/easy-scalable-text-rendering-on-the-gpu-c3f4d782c5ac

yes, and he also wrote very nice blog article(s) about text rendeirng

right

I almost attempted that until (1) found his code was in Skew (???) and (2) each char gets blown up to like 100 triangles (in particular, # of control points)

each char is just a single quad

he uses pre-rendered textures in this case

and IMO skew is not a problem if you want to re-implement the method

it reads pretty well

I’m talking about sky editor (not about his article or theta)

ah

i'm very interested in this, but maybe #off-topic is better? since we do not have #typography.

how does everyone load cljs namespaces in a lein project into the repl?

Is there an html generation lib à la hiccup that works across different targets? i.e. jvm, JS

@skeuomorf there's https://github.com/teropa/hiccups

not sure how closely that aligns with the clj version.

@john Yeah, I’ve seen that one, not cross-target though -I think-

@john What I mean is, something that’s mostly *cljc

*.cljc, maybe with minor cljs and clj code

It may be possible to take the hiccups version and make a cljc version, by pulling in bits from the clj version.

@john Yeah, took a quick glance and was thinking the same

Could potentially do server-side rendering with the hiccups lib

Though that's a different workflow, potentially negating the benefit you were looking for in a cross-platform solution

Sorry, I am not following

something like: https://github.com/reagent-project/reagent-cookbook/tree/master/recipes/reagent-server-rendering

oh, I am familiar with server-side rendering, was wondering how that ties into the conversation is all

prepending “fun fact:” to “Could potentially do server-side rendering with the hiccups lib” 😄

oh, I guess I meant the "hiccup lib"

I've never actually messed with server side rendering.

Yeah, I guess I got the rendering backwards 🙂 hiccup is already "rendering" the html server side.

I think

yeah, I just call it generation

cause rendering has all the “display” connotations attached to it

cause essentially, we’re just generating strings

sending it to a renderer e.g. (a rendering engine inside a browser) gets it rendered

Do both implementations generate strings? or does one make some kind of object?

both generate strings

huh

There’s https://github.com/ibdknox/crate

which is mentioned in the hiccups readme, which returns dom nodes instead

Seems like the hiccups version should be eligible for a cljc upgrade then.

If both versions generate strings

the same strings

Yeah, but I think hiccups is behind regarding some features, at least based on the TODO