Why doesn't this work?

$ cat lumo-test.cljs
(clojure.string/split "one\ntwo\nthree" #"\n")

$ lumo lumo-test.cljs
WARNING: No such namespace: clojure.string, could not locate clojure/string.cljs, clojure/string.cljc, or Closure namespace "" at line 1

$ lumo

cljs.user=> (clojure.string/split "one\ntwo\nthree" #"\n")
["one" "two" "three"]

@whitecoop rule of thumb is you should always require the stuff that you’re using

it happens to work at the REPL because the REPL namespace probably requires clojure.string

@anmonteiro ah. hadn't thought of that possibility. that makes sense.

you need to require the stuff you use so that the CLJS compiler can figure out the dependency tree

I've mostly thought of require as primarily a convenience to not have to write our the full name (`clojure.string/split` vs split, etc.) but, now that I'm thinking about it, I suppose (require '[some.lib]) is about building the dependency tree and using :refer, :as, and such inside of the require are the convenience part of the require.

that was helpful, thanks

@whitecoop yeah, requires really need to be explicit

to be clear, this is a Google Closure limitation

we can’t have dynamic requires in ClojureScript like we can in Clojure

but even in Clojure you need to require stuff that you’re gonna use at some point

@anmonteiro ok. I was wondering about the CLJ vs CLJS part of this. so is needing to be explicit because of Google Closure is the case even without dead code elimination? obviously, we'd want to write with that in mind, but thus far, running stuff with lumo some.cljs isn't using dead-code elimination, right?

@whitecoop yeah, this is not about dead code elimination

it’s about generating deps.js, which Closure needs to know which namespaces to load