Its been a long time (like maybe last was in 2015?) since I had an opportunity to use CLJS @ work

Om was I think pretty new then, and I was pretty happy with it compared to react/js

I'm finding the frontend landscape a little harder to track than in 2015.... has om next stalled? it doesn't seem particularly further along than in early 2016

I'm basically writing highly interactive DOM widgets without a ton of server side state

I'm probably committing for a few months of dev at this point, so not a huge pain if I don't like what I choose, but yeah, all else being equal.....

Curious what both the exotic and stolid choices for highly dynamic frontend UX in cljs are in 2020?

Probably unless I see something very neato, I'll just go with re-frame (or mayyybe fulcro?) but I really have no idea what the tradeoffs are here

All else being equal, easy output of material UI style is a bonus

I think a react based framework is your best bet, seems to be the way to go for now in ClojureScript. Re-agent being the most popular one, with re-frame for state management.

I use rum personally, but I'm only making small toy projects, so I can't speak to more complicated projects. Though I quite like rum personally. I havn't tried re-agent as much to compare though.

And Fulcro is the new craze, but its value is with client/server interaction, so if you don't have the need for a server, I don't think Fulcro would add much benefit, might actually be tough to use it client only, but I know nothing of it, so don't take my word for it

@didibus I really appreciate your thoughts! Fulcro seems cool, but looking it over with your comments in mind, it doesn't seem applicable to my domain

It seems none of the vue-based CLJS frameworks have gotten significant traction

Ya, I have never heard of any non react based cljs lib. I heard of like one or two which tried to not use a JS lib, and do their own virtual dom, or of some using the react clones like preact, but I never heard of say using angular, vue, svelte, etc.

FWIW Fulcro works just as well for client-side only stuff, no need for a server for it to make sense. it is basically the continuation of om.next and then some

vue, svelte, etc and others don't map well to CLJS since they are compiler-based frameworks

I stand corrected on Fulcro then.

Is there something specific about ClojureScript that prevents it from a compilation based approach?

think about it ... it just doesn't map at all. in vue you write .vue files which mix html/css/js. the html directly uses the local JS.

but the vue compiler has no understanding of CLJS so you can't just drop that in there

isn’t cljs a “compile to javascript” language already? we also have macros for dsls

I've never used any of them, so I'm asking very naive questions here. I think I can see how they'd be hard to wrap in a CLJS shell. But I think there could be a pure CLJS library that took a similar approach no?

sure. I wrote one.

but the benefit of all the react wrappers is that you get access to the react ecosystem components

but you can't really get access to the vue ecosystem without the vue compiler/runtime

Well there you go 😛 What happened with it? (I realize that maintaining such a lib requires ton of effort and it might not be realistic for the small cljs community to do so and keep up with the web)

nothing happened with it. I'm using it and working on it. Nowhere near ready for public use but works well for my purposes.

Neat

Hum, this makes me wonder. React docs mention that one motivation to move to hooks is maybe to eventually be able to compile components ahead of time like svelte and vue does as well.

Do you think that would mean exclusively for JSX ? And thus it would leave CLJS in the dust?

no clue what kind of optimizations they have planned. I already don't like hooks so I left react behind about a year ago

What would be your recommendation then, assuming one didn't care abut leveraging anything from the existing react ecosystem, like say I just plan on writting my own UX toolkit from scratch?

I wrote this a while ago https://github.com/thheller/shadow-arborist

it evolved a bit since then but the basic idea is still the same

in practice this looks very much like fulcro/hiccup but just doesn't use react

but yeah .. not ready for public use yet. so if you want to build something from scratch you can maybe steal some ideas but you'd have to start from scratch

not sure I'll ever get this ready for public use. quite swamped already with all the shadow-cljs stuff.

Ah ya, I don't think I meant that much from scratch 😛, just meant like where I don't mind writing my own components 😛

But that I think is one of the project I remembered when I said I've seen a few that tried a pure cljs approach

yeah there aren't that many options for pure CLJS that don't use react

Unrelated question. Say I had a nice little rum component. And I might want to publish this component as a lib, so other projects could reuse it. What would be the approach? Should I go NPM? Clojars? Also, since its a rum component? What would it look like to use it from JS, or from re-agent? Or other react frameworks?

for CLJS users you publish to clojars just like any other CLJ library. no compiled code, just source files.

for JS the story is a bit more complicated and not something a lot of people do

Hum, ya maybe I'll pass publishing it for JS consumption then. Clojars seems easy enough.

And what about the fact it uses rum? Should I somehow just have a function that return the react component? Would that be the easiest to use from any react based framework?

I’m using hooks right now, and it’s basically a closure with memoized functions

Can someone please suggest what (set! js/COMPILED true) may do? why is it needed? :thinking_face: In here https://github.com/status-im/status-fiddle/blob/master/src/prod/app/compile.cljs

who does it? it isn't needed?

Hi people, I have a quick question about js/cljs interop, how do I coerce a js #Number object to be an int please ?

this is how I use it, the id in the following snippet is then a js number

(rf/dispatch [:group/delete id])

which when I dispatch for an http delete gives me a DELETE /api/group/[object%20Object]

sounds like less of a type coercion thing and more of a is is really a number kind of thing

probably want to inspect the value of i’d

I'm 99% certain I've seen an online editor somewhere that had an editor on the left side, with dom output on the right, which had built in support for reagent. But I cannot for the life of me find it anymore. Anyone have any ideas?

found it, was actually a re-frame playground: https://day8.github.io/re-playground/

Please help! I am debugging for few hours. Why this always evaluates nil ?

(require '[cljs.js :refer [eval-str empty-state js-eval]])
(eval-str (empty-state)
          (str "(ns cljs.user
          (:refer-clojure :exclude [atom])
          (:require [reagent.core :as reagent])
          (def atom reagent.core/atom)"
          "I think this will be evaluation result")
          'dummy-symbol
          {:ns            'cljs.user
           :eval          js-eval
           :static-fns    true
           :def-emits-var false
           :load          (fn [name cb]
                            (.log js/console "---name " name)
                            (cb {:lang :clj :source ""}))
           :context       :statement}
          (fn [{:keys [error value] :as x}]
            (.log js/console "-----cb" x)
            (if error
              (do
                (def *er x)
                (js/console.error "Error: " (str error)))
              value)))

@stebokas are you expecting to get the value of the last expression?

I expecting to see this "I think this will be evaluation result" I guess

yes the value of the last expression, in general that's not possible because JavaScript

so we don't support that

we could probably make some special cases - but this is very minor - patch welcome for bootstrapped if you want to work on it

Howdy folks, so we just recently switched over from figwheel to shadow (yay!) and I have had some issues with getting cider-connect-cljs / cider-jack-in-cljs to eval code from my emacs. After running clj -A:server from iterm the server launches and I can build/compile the webpage successfully. The problem arises when I try to cider-connect-cljs from my emacs to eval my code. I choose localhost but when I try to enter the correct nrepl port number I get this weird error

Error running timer: (error "In 'cider-connect-cljs' source: '#[0 \\307\\300\\301\\305\\306\\304\\203\\303\\204
\\202\\310%\\311\\302\"\\207 [#[771 \\306 \\307\\300\\301G\\303%\\310!\\311=\\205\\312\\303\\313\"\\206\\314\\315A\\203)\\315\\241\\210\\205.\\316\\317!\\262\\320\\302\\205C\\321>\\205C\\302C\\322\\203V	G\\323V\\203V!\\202W\\304\\305#\")\\207 [((learner:59489 learner 59489)) nil learner:59489 (metadata) nil nil helm-set-case-fold-search completion-all-completions last emacs completion-metadata-get display-sort-function #[257 \\300\\301\"\\207 [sort helm-generic-sort-fn] 4 

(fn CANDIDATES)] nil t copy-sequence append (helm helm-fuzzy) helm-completion-in-region--initial-filter 0 completion-ignore-case helm-completion-style helm-completion--sorting-done] 13 

(fn STR PREDICATE ACTION)] nil learner:59489 nil t nil t helm-comp-read-get-candidates  helm-cr-default helm-pattern] 6]' 
 (user-error \"No linked CIDER sessions\")")

^^ the correct nrepl port I was trying to connect to was 58833 .. the learner:59489 is a previous nrepl port

Is it unreliable to really on side-effects during cljs macro expansion? I’ve read only one place that it is - due to caching and parallel builds. Specifically, contemplating how to share state during macro expansion time across possible different expansions.

(defmacro step-1 [x]
  (swap! state :x x)
 `(some-code-stuff))

(defmacro step-2 [x]
  (let [x (:x @state) y (do-something x)]
  `(more-code-stuff y)))

As far as “no side-effects” warning. I find very little on the subject, but : https://code.thheller.com/blog/shadow-cljs/2019/10/12/clojurescript-macros.html “Gotcha #2: Caching and :parallel-build”

Does this also mean that a macro cannot rely on analyzer state to have happened “sequentially”?

yeah thats it pretty much

such as looking at :cljs.analyzer/namespaces

in the compiler state

can I not rely on :defs and stuff to be there?

sequentially no, only guarantee is that the namespaces that were required are compiled before the current ns

if they occurred “before my macro expands”

so it has to be a separate ns

to be sure it has happened

not sure what you mean. in the current ns things happen sequentially

but with caching enabled the macros won't be called so nothing happens

eg.

(defmacro my-macro [sym] (lookup-analyzer-state sym) `(whatever))

(def foo 5)

(my-macro foo)

so here the “state” is that def affects the analyzer compilation state

(def foo 5) is guaranteed to happen before (my-macro foo) yes and the analyzer state updated accordingly

ok one step further

(defmacro my-macro [sym] (lookup-analyzer-state sym) `(whatever))
(defmacro defs-thing [sym v] `(def ~sym ~v))

(defs-thing foo 5)

(my-macro foo)

what about when the def is itself behindd a macro? 😛

same thing

I thought this was the case

so I sort of consider the def affecting the analyzer/compiler state to be a side-effect

but in a sense here, it’s more “special”

than my example where I swapped into an atom I guess?

yes but :cljs.analyzer/namespaces 'the.ns is the cache

it will be restored, your state atom won't be

Yeah, I think that makes sense now that I look at it and see what you’re saying

you state effects all happen sequentially if only looking at the current file

with multiple files they might be in parallel, depending on :require setups

if I require another ns

via :require

with caching macro-expansion might be skipped, so no side-effects

will it definitley have happened before?

yes

I’m trying to understand how a parallel build would even work

it’d have to only do things in parallel when they weren’t dependent on each other right?

yes

I guess that does happen a fair amount

depends on the project but generally yes

Thanks for feedback. And good blog post by the way. It seems actually pretty hard to find many details on some of these deeper topics with compilation/macros

I always end up reading your “shadow” blogs and/or some of MFikes stuff it seems.

What's the best way to have pprint go to stderr instead stdout?

which env? node?

Correct, in node

I guess (js/process.stderr.write (with-out-str (pprint foo)))

(let [pperr (with-out-str (pprint (ex-data err))] 
  (js/console.error pperr))

ah yeah that might work too if console.error goes to stderr

Oops forgot to wrap with pprint but yes, we're on the same page. console.error does go to stderr if I recall, but I'll double check.

Following the source through https://github.com/nodejs/node/blob/master/lib/internal/console/constructor.js#L614 suggests it writes to stderr

is it advisable to play around with npm modules inside the clojure script REPL?

I was trying to run the normal node REPL and it's a bit of a pain to load modules

Super advisable! Well... maybe not advisable, but I like to do it

ok, I will try it... CLJS is incredible

!(^ ^)! it is!

the thing you gotta look out for is .bind if you're not familiar with that

I keep getting a window undefined error

ahhhhhh that's because window IS undefined!

because there is no window

ya, just want to run it headless

headless browser but not node?

well, I mean, just test stuff

not tied to a window object

there's a window npm

I'm not liking this require vs import stuff either

I really hope Deno fixes that

workflow you're going for is writing stuff in node (javascript) then dynamically loading to cljs compiled to node running on REPL?

ya

I'm going to end up running node on the JVM

GraalVM

you're a space explorer, sir! As in, there are probably like, 6 people doing that

We need to get a clojurescript conj going

I know I know. Just a proof of concept

I'm not saying it's a bad thing! space exploration is awesome!

just... not a lot of astronauts

this is my fun stuff

I've hacked other langs together before doing this

Jruby

sweeet

I am doing WebRTC

Decentral Video, want to run the websockets on the Java stack

WebRTC shouldn't need websockets, I wouldn't think. At least, not directly. Unless that's how the signal server is setup

Doesn't mean you shouldn't use that stack tho. I think it's a great stack

the Signaling server can use WebSockets

I'm either going to put the chat on the webrtc, or the websockets

haven't decided which yet

are you doing group chat or point to point?

well, webrtc also supports groups

need to elect a leader for that I think, right?

I like the p2p encryption idea

me too

oh neat. I haven’t seen webrtc with groups

it's saying it uses a mesh

it creates a mesh network, p2p, for groups

and yes, it can appoint a main node

so it's fault tolerant

that must be brand new. I think for practical groups above a certain point a leader is pretty useful otherwise it turns into a lot of traffic

just from a quick google: > Mesh: For smaller groups you can establish a connection from every peer to every other one.

ya, that would become crazy for large groups

someone gave me a repo, 300 forks

I am fixing it up

I'm looking into using serviceworkers to handle that kind of abstraction

just... having a hard time convincing work that this effort is worth it lol

My argument is "c'monnnn, it'd be awwwwesome"

async workers in the background to handle the mesh?

yuuuup

yikes

now that is interesting

service workers can do client side network proxy transparently

ummm, you should get into this then

I was playing around with Peer.js

not sure if I should use that or roll our own

I'll open source the node stuff. But the commercial level stays with Java.

It's really not toooooooo bad. I mean it's horrible. But not bad.

https://github.com/DiddoJS/video-peer

I changed the name. There were like 300 forks. Lots of interest.

I lost my github history too... re-organizing

I will happily make empty promises of contributing

well meaning empty promises though!

😭

nah, only do it if its in your interest

See that's the problem, it IS in my interest. I just get swallowed up by work all the time 😞

If I could start finishing my work DURING work then maybe I could do other things AFTER work

does your work need any of it?

chat, AI, video?

I'll DM you so as to not overly hijack this chat

k

Is it possible to wrap all functions call in ClojureScript? I would like to log/save all arguments and outputs from function calls.

I thought of hacking specs and the instrument function to achieve it, but not sure it is the best solution.

this may offer some clues: https://github.com/philoskim/debux