I just fixed an unnecessary dependency on lein in the shadow-cljs support of Calva. I guess most people have lein installed. But anyway, now the integrated REPLs are started using the shadow-cljs command.

@pez neat. its still not possible to do custom UI in VSCode right?

No, sadly it isn't. And sadly they seem to have decided to let the REPL needs be served by those integrated terminals: https://github.com/Microsoft/vscode/issues/547

yeah that sucks

I am still hungry for feedback from anyone revving Calva up for a spin for shadow-cljs projects.

I didn't have time to test it again yet

Well, anyone, I do appreciate that you have tons of other things on your plate. 😃

If I could interview some people about their shadow-cljs workflows, that would be awesome.

I can tell you mine briefly. run shadow-cljs server or lein repl (when working on shadow-cljs itself). connect to nREPL remotely. run everything else through the REPL.

only using command line for shadow-cljs release ...

one neat feature of Cursive is that it allows you to define snippets of code to send to the REPL on keypress

so I have snippets to start my server and stuff via keybinds

So you switch between builds using the REPL?

yep

Using snippets?

no, that I do manually

Ok. There is a concept of snippets in Code. I will look at what that is.

don't think thats the same

see https://cursive-ide.com/userguide/repl.html REPL Commands

Me neither. :face_with_cowboy_hat: It is probably more about boilerplate

Thanks for that link. Very helpful!

could it be that shadow-cljs watch isn't shutting down cleanly when stopped via CTRL + C?

it happened to me for the second time that i ended up with two ghost proccesses (one in node and one in java); i only realized when i started another instance and experienced very weird bugs

like objects not satisfying interfaces anymore even though they were when testing it on the repl

do you have another watch running?

different build?

no, just one build

it also doesn't happen always (of course not, that would maybe even make it reproducible)

which OS?

current macOS

10.13.4

I have no idea how they would interfere with each other on the client side though

the server leaving ghost processes has happened before

still not sure what causes it

ok. i'll see if i can collect more info if this happens again

great tool though! thanks a lot

@thheller - about the thing we talked about yesterday - different dependencies in dev and release mode - I actually think this may not be worth looking into at all. The two projects I mentioned are actually one, because re-frame/tracing is a fork of debux and I'm not aware of any other projects that do this. Since there's a workaround for the current situation and it's not really a great idea to do it this way in general, I'd say no reason to waste time/energy on this, unless there's another important use case.

I am making a quick editor helper with . Said library uses (js/require ... ) to include the acorn-jsx js parser. I have forked it to patch it, what would be the correct approach to make libs that use js/require universally compatible?

Hi! Could someone point me at good explanation about the advantages of shadow-cljs? I wonder is it worth and is it possible to replace standard pipeline and figwheel with shadow?

https://code.thheller.com/blog/shadow-cljs/2017/09/15/shadow-cljs-introduction.html

It's possible and it works great.

(and so far I think it's worth it)

btw: because it's not quite clear in the docs and you said you needed help, cider works via cider-connect, and then opening a cljs-repl as described (`(s/nrepl-select :build-id)`)

which is really really neat 🙂

what doesn't work is refactoring. but i don't know how good refactoring is with cljs anyways, with jvm-clj it's quite good

@achikin that's actually what i did. it's nice because interaction with the npm ecosystem feels very painless, also configuration is quite easy and mostly just works

also it feels like compilation is a bit quicker? maybe that's just subjective. biggest pluspoint for me is embracing npm, because i work in a team of frontend-webdevs and if it's on npm it's a bit less alien

This is what I'm actually interested in - having better npm dependency management.

works. you just require npm modules as you would with clojurescript modules. only surprise i had was when working with d3 i was requiring a module (d3-requests i think?) and it gave me the nodejs version of the module whereas i was expecting the browser version.

I'm looking through the examples, listed in the README.

For example here https://github.com/mhuebert/shadow-re-frame/blob/8d1bc1057e5b28dd8df6b34872a95a12a39a735a/package.json#L9 react is included in package.json dependencies. Why?

shadow-cljs pulls in npm modules directly from the node_modules, so what you do is install them normally using npm or yarn and a package.json.

note that cljsjs libraries (and some other libraries, like promesa) put the javascript in the jar and wrap it with a :foreign-libs declaration. but shadow-cljs doesn’t support foreign-libs, so it just fakes the namespace and continues to rely on node_modules

I think this business of reading the javascript directly is why the externs inference works so well

hey, how do I run multiple builds at the same time?

shadow-cljs watch app1,app2

right, just space between works