I was crashing into http://dev.clojure.org/jira/browse/CLJS-536 but per dnolen's note there, I learned about extend-type and got the following snippet to get clj->js to retain namespaces in keywords. Is this a reasonable way to do so?

(extend-type Keyword
  IEncodeJS
  (-clj->js
    [kw]
    (str (namespace kw) "/" (name kw))))

@fasiha: in apps sure that’s fine - just don’t do that kind of thing in a lib

Ah. Understood @dnolen. I'm stuffing EAVT tuples into a "regular" database (RethinkDB) as JSON and was confused why my attributes were coming out naked and un-namespaced.

right, converting to JS is inherently lossly - so that’s why we have knobs

no way to know what anyone is actually going to want to do

what's the proper way to specify an extern for (let [c (.-Constructor (js/require "module")), o (new c)]) ?

the simplest way

Object.Constructor = function() {};

ah, thanks!

then when you call (.someFn o), you need Constructor.prototype.someFn = function() {};, right?

if Constructor is a type

you need

/* @constructor */ Object.Constructor = function() {};

or something like this, really the Google Closure compiler docs are the place to look

externs files are just Google Closure annotated files without any code in the bodies

ok, I see

thanks!

and yes, if you’ve defined Constructor, then you define the methods as you’ve described

Idea: Slackify http://clojuredocs.org, so any library can get its own version of clojuredocs with its functions & docstrings and where people can post examples and add "see also"s.

I want to extend a javascript react class with a IFn interface so I can invoke it

can't quite figure out how to extend this javascript class with IFn, tried a few things

e.g.

(specify FooT
  IFn
  (-invoke [this a] :yay))

@jasonjckn: have a look here: http://blog.klipse.tech/clojure/2016/04/07/ifn.html

thank you @rohit

@jasonjckn: i think you are almost there. it should be extend-type instead of specify.

Okay, taking a shot in the dark here.

When starting up a cljs repl, I'm getting goog is not defined and clojure is not defined.

The namespace has loaded up properly, (e.g. I can (js/console.log "HELLO") in the cljs file and see it show up in my console), but the initial payload of the repl results in goog is not defined.

@potetm: sounds you haven’t configured :asset-path for whatever makes for your web server setup

looking at the Network panel of any browser devtools should show you that no sources could be loaded

So, I'm getting 200s for all of my js files. It's the repl eval that causes the error. (I actually have only 200s in my network panel.)

Do I need to set :asset-path on the repl or repl-env?

@rohit: i got it working but then the google closure compiler emitted warnings about invalid arity

@jasonjckn: could you share some code?

@potetm Can you create a minimal example?

@rohit:

(defui Foo 

  static IFn
  (-invoke ([this a b] :yay))

  Object
  (render [T]
          (div (str "Foo props: " (prn-str (om/props T))))))

(js/console.log (Foo 3 4))

wrong number of arguments passed to Foo

Figwheel: Compile Warning - Wrong number of args (2) passed to admin.ui/Foo at line 23 /Users/jason/d/play/admin/src/client/admin/ui.cljs

@meowy: Working on it now.

anyone have success loading figwheel to an already loaded page? I’m running into ‘can’t document.write after the page is fully loaded' errors

alternatively, getting the plain browser repl to work over SSL?

@jasonjckn: i gotta call it a night. its kinda late here. will check it in the morning. sorry about that

no sweat, thanks for your help @rohit night

cheers!

Is there any chance that a repl working dir can get into a bad state and cause those goog is not defined errors? cc: meowy

Yes.

Running (browser/repl-env :optimizations :none) does it for me.

Once that's been run, it doesn't appear to ever recover either.

It appears that all that matters is what's run first. If the first run does not include :optimizations :none, then it connects successfully and will connect successfully on subsequent runs, not matter what optimizations are later specified.

If the first run uses :optimizations :none it will never successfully connect.

Is this the expected behavior, or should I file a ticket for this?

Oh I forgot, the whole thing is loaded in an iframe, so it has to reload all of the code.

So :optimizations :none can only work if you've specified a :main, because otherwise it expects you to load yourself goog/base.js first.

But, unfortunately, the browser repl env only allows you to pass in :optimizations and :output-dir into the initial compile.

Ah lookie what I found: http://dev.clojure.org/jira/browse/CLJS-1502