what's a good introductory book / guide for clojurescript web dev?

@dnolen: https://github.com/clojure/core.async/commit/a833f6262cdaf92c6b16dd201d1876e0de424e14

@borkdude: nice!

I'm working on a core.async example for the wiki if anyone wants to help with it: https://github.com/clojure/core.async/wiki/Sieve-of-Eratosthenes

I'm getting stack overflow errors when I try to consume 7000 primes in a repl using (time (consume 7000 (chan-of-primes)))

@meow: What if you change consume to:

(defn consume [n ch] (take n (repeatedly #(<!! ch))))

any tips for why an uberwar-deployed app would barf with "[Error] ReferenceError: Can't find variable: goog (anonymous function) (app.js, line 1)" ? It looks to me like advanced compilation didn't run, but this is my first time trying to deploy to tomcat.

@angusiguess: To get a timing value I need to eliminate the default lazy behavior of clojure, that's why I defined it as:

(defn consume [n ch]
  (dorun (repeatedly n #(<!! ch))))

there's even a js/out directory, which shouldn't be there if it was picking up the :uberjar profile settings. Or at least that's my weak understanding of it so far.

yikes... If I try and run lein uberjar I get: Compiling "resources/public/js/app.js" failed. Exception in thread "main" java.lang.NoSuchMethodError: com.google.common.io.ByteStreams.limit

this is a mostly-ootb luminus project, fwiw

lein deps :tree to the rescue... something was pulling in a really old version of guava, so at least now it compiles in advanced mode.

but ring-uberwar doesn't seem to be picking that profile up.

confirmed... lein uberjar runs advanced compilation, which gets bundled by ring uberwar, but uberwar won't do the advanced if it's run clean.

@dnolen: When you have time could you look into applying the patch for CLJS 836? http://dev.clojure.org/jira/browse/CLJS-836

@spinningtopsofdoom: hrm weird have you ever submitted a CA?

ah yes

yep

@spinningtopsofdoom: applied thanks for the bump

@spinningtopsofdoom: there’s also a #C07UQ678E chan

ah did not know that thanks!

based on http://dev.clojure.org/jira/browse/CLJS-1363, we can pass a namespace to compiler options in cljs.js. given the default of ’cljs.user, guessed I could write something like :ns ‘my-app.core, but then I see Assert failed: (ana/ast? sym). Should this just work out-of-the-box or is there something special to do to prepare?

@mhuebert: need more info what precisely did you try + stack trace

@dnolen: trying to def into a namespace - https://gist.github.com/mhuebert/315aeaf81307275d35d4

this might be a stupid JS question, but does ‘undefined? work on js variables?

I’m trying to use google analytics, and (when (not (undefined? js/ga)) …) is giving me: Uncaught ReferenceError: ga is not defined

@arohner: you want exists? to do that

feature detection

@mhuebert: thanks for the report http://dev.clojure.org/jira/browse/CLJS-1383

will look into it

@dnolen: thanks

@arohner: compiles to typeof x != “undefined” which is safe from blowing up if x doesn’t exist

which means exists? has to be a macro btw

@mhuebert: took a second to test that, cannot reproduce your issue, if you can provide complete steps to repro with master I will take a look.

did anyone here have any luck doing some live development with github's electron. Figwheel style?

fluke777: also curious

martinklepsch: Just tried to run the flappy bird demo and wonders. Works like a charm. I will experiment with this a little more I have some node dependencies I would like to have. Will let you know here.

martinklepsch: If you are interested

fluke777: totally, please do!

is aget the proper (and only?) way to access non-CLJS JS in advanced compilation?

i.e. (aget js/window “foo”)

this is 3rd party JS that I don’t own

@arohner: you should use goog.object/get - aget is only for arrays - see: https://github.com/cljsinfo/cljs-api-docs/blob/catalog/refs/cljs.core_aget.md

@meow: thanks. For my own understanding, why should I use goog.object/get?

@arohner: tl;dr - because @dnolen says so 😉

seriously, though - aget is meant to mimic the functionality of Clojure, where it only works on java arrays

the fact that it works on other types now in cljs is not by design and might go away in the future for compatibility with the concept of Clojure and for type-based performance gains

the goog namespace is always available since cljs uses the goog closure compiler so it is a good choice for things like this

goog.object is for interacting with js objects, and has methods like get and getValueByKeys that will perform well and work across all browsers

@arohner: ask away if that doesn't answer all your questions

@fluke777 and @martinklepsch: someone recently posted a link to something for electron based on work done by @shaunlebron and @chrisoakman on cuttle or something.

@meow: awesome, thanks

I can't find it with a google search but I know I just saw something about it like maybe yesterday

@arohner: using aget incorrectly in the future will likely result in a type warning similar to arithmetic

there’s really no reason to abuse aget when you have goog.object/get and a lot more

Google Closure namespaces everyone should know about

goog.string, goog.array, goog.object

@meow: Maybe one of these? https://github.com/Gonzih/cljs-electron, https://github.com/karad/lein_template_descjop

@fluke777: found it - it wasn't here, it was on the mailing list: https://groups.google.com/forum/#!topic/clojurescript/nlqZMzgByjI

its a lein template: https://github.com/ducky427/electron-template

@meow: Thank you for the effort. Will check it out

@fluke777: np - let me know what you think as electron is something I'd like to get into at some point

aset and aget remain the simplest way to access js properties. using goog.object/get and getValueByKeys and set do the same thing in a more verbose way

and no multi arity set in goog.object

We need to update Cuttle.

I'll try to get that this weekend or next week sometime.

goog.object/get is not exactly the same — it adds an extra in check, as its real purpose to exist is for the optional arg — i think it’s worth considering something like oget in core which skips the extra check and signals intent clearly

agreed

^ @nullptr

another vote for oget @dnolen

not going to happen no matter how many people ask for it

goog.object exists, Google uses this namespace for everything

I find it hard to believe that Google didn’t get their core namespaces right

I've always used aget for JS object property access; is this not recommended?

I always liked that it compiled down to bracket notation using strings.

it’s possible that the compiler could optimize it in the absence of opt_val, at which point they are equivalent and oget makes no sense

@dnolen just saying that parts of cljs core library already wrap goog functionality for convenience to cljs users

JS strings are always safe for the Google Closure Compiler in any compilation mode (I believe)

yes, it’s quite commonly used for non-externed json stuff

in js terms, foo.bar vs foo[‘bar’]

remember to separate the problem from any particular solution

aget is not a solution to the extern problem, it’s just a hack that work

agreed, aget is a hack

people keep asking for oget and seem completely unaware that goog.object is a thing which has more

people need to start looking very seriously at Closure provides

weird JS interop helpers just keeps that from happening

plus I hate oget

worst name ever

goog.object.get solves the problem, but it exists to solve a different problem — it’s a subtle point :)

it should clearly be named: get-javascript-object-property 😉

@nullptr: the real problem is the externs problem, and I’ve already said multiple times we’re working on that

but that’s larger effort which isn’t going to be band-aided by some short-term thing

dnolen: i hear you — we have tens of thousands of closure (js) code here, and there’s tons of foo[‘bar’] that’s not externed

(def goget goog.object/get) -->> Just Kidding!!!!!

all im thinking about is UX for new cljs users. having a dead simple function for a common interop reqt

it’s nice to have the option — if you just need access to one thing, externs can be a lot of extra overhead — again goog.object/get works — just seems mildly hacky (as is aget)

(hacky because it is only there for the opt_val support)

anyways I don’t have anything more official to say about it

use goog.object/get

so yeah, externs solve the problem, but not always the most convenient thing

in the future property access on foreign values will work you won’t need all this stuff

we will generate externs for you

sweet

likely there will need to be a corresponding ^js if people are doing fancy higher order things

arbitrary json objects though

arbitrary json … this is what goog.object is for

totally different from the foreign library thing

which is way more painful

ux for new users is all im arguing for. simple multi-arity functions

I’m just not concerned about this problem at all

doc this in all the right places

and new users will know what to do

it’s also the case that in watching what developers actually do w/ ClojureScript, they convert JS data immediately

few people want to manipulate JS data directly (except maybe arrays) because the tools are so weak compared to ClojureScript data

investing in things only new users are going to use

no way

i mean, i guess this is a bigger discussion about convenience replaced by complete documentation

but i was working with people who just wanted to manipulate js options map as fast as possible

“as fast as possible"

for what purpose?

are they actually in an inner loop with 10+ million iterations

if not waste of time

if they are and it’s that important then writing an extern is the right answer and takes no time at all

if this is “JS-think” then I’m not really sympathetic to that

i never saw the code actually. there were just noticeable performance hits in creating cljs data structures and converting to js data. i think with clj->js

http://jsperf.com is a cesspool

what I mean is was the perf hit like

“let me measure this part of my program that doesn’t matter"

oh the part of the my program that doesn’t matter is 10X slower than the JS

right i get that

let me optimize this thing that doesn’t matter

it’s bonkers

agreed

not going to support that mentality at all

this was just a side point. the only thing i care about is not telling new users that to do such a simple interop thing by importing three functions from goog.object

this is just a non-goal

ok

new comers should know about goog.object

if the talking about the awesomess of Google Closure isn’t the first thing covered something is not right w/ the docs

we will have to add that to cljs-api-docs in some capacity

the fact that new users never have to include an additional dep for basic functionality is something to celebrate not hide

@dnolen: as a newcomer, I heard the “Closure is awesome” part, but for some reason the “you are expected to know Closure’s standard lib the way you are expected to know Java’s when using CLJ” was not obvious to me

I thought “Closure is awesome” referred to the compiler, not the stdlib

@arohner: if you rewatch the original ClojureScript announce that’s not what Rich Hickey said

but I agree the messaging over all should be improved and more clear

https://github.com/clojure/clojurescript/wiki/Dependencies#using-string-names

yes if people see docs where this isn’t be done correctly please fix

I’ve started switching my own libs away from the old style

at least we can all agree that aget is terrible! i can’t tell you how long it took me to figure out that it was “the way” to do that before it wasn’t… exceedingly non-obvious

@dnolen: I’ve put two examples in a clean repo- https://github.com/mhuebert/def-in-ns/blob/master/src/cljs/foo/eval.cljs evaling ‘(def x 1) with :ns set to ‘foo.other, an existing ns, compiles to “.x = 1” evaling ‘(do (ns foo.totally-new) (def x 1)) compiles to "(function (){\ngoog.provide('foo.totally_new');\ngoog.require('cljs.core');\n\ncljs.user.x = 1;\n})()\n”

@mhuebert: neither of those example are actually valid though. you never created the namespace first

sorry rather the first one is never going to work

we’re not going to magically create a namespace for you.

@dnolen so it doesn’t see that I created the other namespace elsewhere? (`(ns foo.other)` in cljs/foo/other.cljs)

@mhuebert: if you already made it yes that will work

but your example doesn’t demonstrate this clearly at all

I have an example in ClojureScript master right now that works

sorry, I linked to the eval file.

it’s in the dir with core & other

https://github.com/mhuebert/def-in-ns/tree/master/src/cljs/foo

@mhuebert: ah so I didn’t really understand what you are trying to do

it’s important to be very specific about what you’re attempting

you want to build a project and then eval inside that project’s namespace at runtime

yes

I’m having success as long as I only want to eval in cljs.user,

moderate success at passing in some helper functions there that eval can access

@mhuebert: also figwheel is a really undesirable thing

I’m not going to look at anything that isn’t actually minimal

massive time sink for me

I only want to look at something with zero deps

understandable, I can take that out

does leiningen cause problems?

@mhuebert: anything that isn’t minimal causes problems

ClojureScript is plenty fancy enough as it is

introducing additional variables is a productivity killer

@mhuebert: in anycase, make a real minimal example ping me and I will look at it

ok thanks