@dnolen a little late, but re: my question about profiling...the code i'm comparing uses createFactory rather abusively and the generating functions aren't identical, map vs. for loop not a great comparison, but the fwiw the om version is actually lagging behind the further i push it whereas the js one seems bound to 40% cpu usage. again, i'd just chalk that up to the functions employed and say i greatly prefer the simplicity. but i guess it shows i was always managing the react lifecycle well, plus good thing i didn't try to sell that client based on benchmarks (i seem sold anyway) since i may have had to hack through tons of things manually to beat what he had that way. although he wasn't using any comparable framework previously so perhaps not. anyway, grain of salt... but seems to back up what you told me

simple question - how to I swap a css selector value programatically in cljs?

directly without an atom

or jayq

@jcb Can you make an example? A css selector selects DOM elements...

I just need to set some css to already rendered reagent components

I'm used to clojure just not the frontend

@jcb - i think js/document.querySelector is a thing, or may js/document.querySelectorAll, google until you fin dosmething that looks like that

play with the .-style attribute of the objects it returns

that should give you what you want

what you are describing is a bit unusual though - usually you’d want to epxress your components’ css using reagent

thanks .-style is returning null

but if you have a specific reason why you want to do this, i think that that function is what i’d reach for

hm

try console.logging the output of calling that function i mentioned

you should be able to expand it and inspect it with a good UI

hm maybe not

sorry, it’s late and i’m on pain meds 😕

hm, (js/console.log (.-style (js/document.querySelector “canvas”))) works fine if i paste it into one of the snippet boxes on http://blog.jrheard.com/procedural-dungeon-generation-drunkards-walk-in-clojurescript

It's to get around an odd z-index problem with a frontend that has been written by someone else

gotcha

you should be able to manipulate this style object with aget and aset, if memory serves

thanks jrheard!

(aset (.-style (js/document.querySelector "canvas")) "background-color" “pink”) seems to work well, for instnace

np, gl

So I’m pretty much a total CLJS newb, and I’m having problems pretty printing an object in Node on AWS lambda.

I’m trying to print the environment variables out, so I’m doing this: (println (pr-str (js->clj (.. js/process -env))))

This just results in #object[Object [object Object]], and I can’t figure out why.

Similarly, (get (js->clj (.. js/process -env)) "AWS_ACCESS_KEY_ID”) results in nil, but (aget (.. js/process -env) "AWS_ACCESS_KEY_ID”) produces the value I want.

What am I missing?

@cfleming I know nothing about AWS but most likely js/process.env is not a normal javascript object but something else which js->clj does not support

@thheller That might explain it, thanks.

@cfleming https://github.com/Gozala/node.core/blob/6e646f8b482b1defbb095537135fa84e2e64df40/src/node/utils.cljs

Never used node myself. But note the comment.

@rauh: I see, that’s interesting, thanks. That looks like the problem.

Kind of awful solution.

hehe I cringed when I saw the solution 😛

Right now, if it works it’s good enough for me 🙂

if the object is shallow you can probably (-> js/process.env (js/goog.object.clone) (js->clj))

I think I’ll just aget it right now, and write to the author of cljs-lambda to ask about it.

js/goog.object.get is the preferred getter

aget is meant to arrays and you are not working with an array

might break in the future (though) unlikely

Oh, ok, thanks.

I’ll do that.

Long shot: Has anyone used multiple textures with http://thi.ng geom?

have an interesting issue with :advanced mode build: https://gist.github.com/darwin/34b2eaf97f1823b1180eb14ad64af9fd initial-config is not used so it should be elided, 01 does not elide it, 02 does note the comment there

it looks like I’m hitting some arbitrary limit on how deep closure compiler goes analysing my code, but this depth is pretty shallow IMO

@darwin that’s interesting but this probably a better conversation for #cljs-dev

ok, will continue there

Is it possible to access the current compiler profile, so as to use one app-state for dev and another for prod?

@wielderofmjolnir this can be done via a macro

macros can see the compiler environment via &env

Oh, sweet! Thanks!

https://twitter.com/anmonteiro90/status/796439950489223168

Check Lumo ^ out. It is really as good as it looks. 🙂

Cool stuff

In order to get consistent performance benchmarks, does it make sense to run the code you are trying to benchmark in several tabs (say 100) then aggregate the time results once all the tabs have finished running the code? Obviously I would prefer to not have 100 tabs open and close, is there a better way to run cljs code (targeting the browser) many times in an isolated environment?

@kenny running multiple benchmarks in multiple tabs is probably not a good idea

you generally want to run one benchmark at a time

if you’re serious about benchmarking you also want to make sure you’re not running anything else but the benchmark

So you're saying the better solution would be to run it 100 times in 1 tab?

absolutely

benchmarking is a painful slow process

but it also depends on what you’re trying to achieve

Will the results across all tabs really be so biased that it is not worth using?

yes

browsers are multiprocess

unfocused tabs get penalized

you are subject to OS scheduling

plus your benchmark would be “how fast does this code run… while being also run 99 other times."

don’t you want to know “how fast does this code run?"

In this case, I am really just looking for relative performance, not absolute. So, if all tabs are unfocused would they get penalized in an equal amount, making the results consistent but inaccurate?

@kenny I just do not recommend what you are suggesting

it won’t tell you anything interesting

if you want to know anything at all - run them in a focused tab, one after the other

if it’s not browser specific - just run it against Node or even better an unadorned JS VM like JavaScriptCore

if you’re not looking for particularly accurate results - then no the methodology doesn’t matter - but I wouldn’t put any real trust in whatever information you collect

You wouldn't trust the results if the benchmark was ran in the same way and then compared? It seems like the consistency is trustworthy (if it is consistent)

@kenny i use https://github.com/ptaoussanis/tufte for relative benchmarking fwiw

you don’t seem to want the answer to the question you asked 🙂

hee

Well, I guess that is part of the question 🙂

If there is no consistency then of course you can't trust the results

when optimizing a fucntion, i just run the function under test like 250 times or whatever, in one tab, and then make a change and then run it again in the same way

seems to work fine

if i see too much variation then i bump up the number of times i run it and that usually solves the problem

i am.. not a scientist

@jrheard Right. That normally works great. Also, thanks for the Tufte suggestion. I will take a look at it 🙂 The only problem with that method is when you are not testing pure functions (`memoize`ed functions).

ah interesting

@kenny no I would not trust the results

if it’s a function you control, perhaps you could add a flag that turns off memoization?

just briefly while you’re poking at it

I’ve done too much benchmarking around ClojureScript to believe anything that doesn’t follow the procedures I outlined above

there is just so much that can be going on with with the browser and the OS and threads and scheduling and stuff right? you’d have to be an expert at so many things to know what you could count on and what is random system jitter and how many threads make your machine start to spend a ton of time worrying about overhead vs. running your tests

@dnolen @cschep @jrheard thanks for the input! 🙂 I will change the direction I was headed down.

glgl

good luck I’m sure you’ll find some good data

😄

profiling/optimizing is super fun

lots of immediate gratification

"immediate"

🙂

😄

haa

@cschep fancy seeing you here

@noisesmith hey man! long time!