does anyone know whether figwheel compiles macros (on the first build, not autobuilds)? i would think so, but i'm getting that "wrong number of args" error that would point to no...

@sophiago the ClojureScript compiler loads your macros

i was pointing to the issue here about how they're compiled in a separate stage and that's why in self-hosted cljs it exposes form and env args: http://blog.fikesfarm.com/posts/2015-09-07-messing-with-macros-at-the-repl.html

(i rarely use macros, so this must be my first time using one in cljs)

@anmonteiro i guess you're saying i'd experience that behavior regardless of using figwheel's local server?

@sophiago hrm you need to clarify what you’re trying to do

macros in ClojureScript need to be written in Clojure and required with :require-macros

I’m guessing you’re not doing that and running into weird errors

yes, exactly. but i assumed that was only an issue of running the files locally and that the figwheel server would compile the macros, which must be incorrect

anyway, wouldn't it make sense to use cljc these days instead of :require-macros or would that not do it?

you can write the macros in a .cljc file. that is orthogonal to what I’m describing

because the compiler effectively runs in Clojure (on the JVM), your macros are expanded during analysis

that’s why they need to be written in Clojure

right, but you're saying with cljc i'd still need to use :require-macros?

the macros would be written in the :clj branch, so yes

i meant putting everything in a .cljc file

there’s absolutely no difference to the CLJS compiler if your macros are in a foo.cljc or foo.clj file

ah ok. i understand what you're saying now: they're still compiled to clj so i need to require them for the cljs compiler

yes

thanks

@anmonteiro sorry, one more thing. isn't it possible to use refer-macros or :include-macros true on the same namespace? the github wiki shows how to do this for required namespaces, but not for the current one

yeah you can self-require

so like: (ns foo.core (:require [foo.core :refer :all :include-macros true]))?

(ns foo.core
  (:require-macros [foo.core :refer [macro-you-wanna-refer]]))

there’s no :refer :all in ClojureScript

ah so that was the problem (it's been a while)

hmm, i'm still getting a weird classpath error

figwheel is telling me Could not locate Foo/core__init.class or Foo/core.clj on classpath. Please check that namespaces with dashes use underscores in the Clojure file name.

@sophiago can you make a minimal repro I can look at?

should be fairly easy

let me see. the library is heavily intertwined 😕

something with just the requires should be enough

lol, well i created a whole figwheel project file with just the requires and an empty macro and still get the error

cool

can you put it on GitHub or in a gist?

is that fine or do you need the whole project file to diagnose?

so is that a .cljc file?

nope. .cljs

your macro needs to be defined in Clojure…

you need to create a core.clj file where you define your macro

ah i should have figured that when self-requiring it needed to be .cljc

😛

now i'm trying to call them from the aliased clj namespace and getting clojure.lang.PersistentList cannot be cast to clojure.lang.IFn. does that make any sense?

@tiensonqin I hit the error: [exp] goog.require could not find: figwheel.connect after a recent upgrade, so thought I'd pass on the solution. (I think we need to get a handle on exp/clojuresript upgrades somehow - they are not easy to do.) Anyway, if you hit this error, you need to change figwheel-bridge.js like so: https://github.com/WintonCentre/unspun/commit/a030d5596e9c046ae3329b3bf543aa0392a44e17#diff-a8aaf4a8ec45fa1d7054daff091c9365L228

@tiensonqin See also, https://github.com/drapanjanas/re-natal/issues/80 and the similar fix in re-natal at https://github.com/drapanjanas/re-natal/commit/999615e921555e0d42728e14a6f91d80eed63cc4#diff-a27f34062e9e57b68e6346717e0db9e5L223

Can anyone give me some tips on debugging macros in cljs? A common technique is to put (println) statements everywhere, which obviously doesn't work when evaluating macros (not to the js console, anyway).

sometimes what works is macroexpanding (probably in a clj repl), and then debugging the expanded code, finally converting back to fixed macro

A little tricky because it is generating cljs code which can't be evaluated in a clj repl but should give me some more visibility at least.

macroexpand doesn't care if the code it shows you is valid

you can copy paste to cljs

@danielstockton Connect to the NREPL before you start figwheel (manually).

Then you can REPL just like you'd do in Clojure.

What's the easiest way to embed a cljs interpreter inside a cljs app ?

in cljs, what is the most efficient way of going from "a string representing a PNG" to "a core.matrix reprensetatino of the same iamge" ?

is http://thi.ng/ndarray still maintined? why is it pulling in clojure 1.7

@danielstockton At a regular ClojureScript REPL, if your macro is defined in code on disk, you can use macroexpand, and if it doesn’t look right, change the code, and then (require-macros 'foo.core :reload) and expand again and/or try evaluating a form. If your macro is fairly self-contained (relies only on ClojureScript), you can fire up Lumo or Planck or any other self-hosted ClojureScript environment, and try different variations on the macro right in the REPL by abusing the $macros pseudo-namespace as described in http://blog.fikesfarm.com/posts/2015-09-07-messing-with-macros-at-the-repl.html

What is a good straightforward lein template for a ClojureScript utility project? (That is, ClojureScript code that will be used only as a dependency of other projects)

@mfikes Thanks, I'll give that a look.

@deg I’d normally just do lein new mylib and simply add ClojureScript as a dep in project.clj and go from there. You’d probably want to also set up for unit testing your lib using some other community developed thing (like doo, or even Tach, if you are also targeting self-host)

I suppose adding :scope "provided" to the ClojureScript dep ^ makes sense

The main idea is that a ClojureScript lib is really just JAR dep with some ClojureScript source in it

Thanks! Meanwhile, the simplest one I found on Clojars was mies, which looks pretty close to that.

The curse of the vibrant ecosystem is that there are so many templates available, most of which are ancient or irrelevant.

Btw, I've never quite understood what :scope "provided" does.

@deg see scope: https://maven.apache.org/guides/introduction/introduction-to-dependency-mechanism.html

Thanks!

@deg Let's say you want to make a ClojureScript library. For maximum usability, you want to make your library code only depend on some very old version of ClojureScript. So, in your project.clj you might have [org.clojure/clojurescript "1.7.28" :scope "provided"]. But, you don't want downstream clients of your library to accidentally have that old version transitively pulled in and placed on their classpath. Instead, you want the client to be able to specify any newer version of ClojureScript. The "provided" bit communicates that the consuming environment will provide that dependency. Ideally, the Maven system would choose the latest on its own, but the dependency resolution rules are not that straightforward. So, by using "provided" you work around that particular issue.

Is there a way to have query params in cljs-ajax automatically url-encoded?

How much do you tend to use reduce in you code to update state?

@danielstockton one thing that no one mentioned is the println technique you said you can't use

Turns out you actually can, but the output will be directed to your terminal instead

The trick is wrapping your println statements in binding [*out* *err*]

It works at the clj repl but I think I need to play with my setup to see that when compiling the cljs. I'm running figwheel and it doesn't print to the cljs repl.

#?(:clj ~(binding [*out* *err*]
                     (println "test")))

It shouldn't print into the CLJS REPL

That'll print when the analyzer expands the forms

Yeah, I didn't think it should. Just trying to work out where I can get it to print, except for evaluating it in the clj repl. If I evaluate it in the clj repl, I can see the expanded form anyway.

trying to use the latest org.clojure/clojurescript "1.9.542" and [org.clojure/clojure "1.9.0-alpha17"] but getting error when compiling via figwheel

Failed to compile "resources/public/compiled/cards.js" in 5.656 seconds.
----  Could not Analyze  resources/public/compiled/cards/cljs/pprint.cljs  ----

  Invalid :refer, macro cljs.pprint/deftype does not exist 

hm.. this is not the latest, tried with cljs 1.9.562 but getting this same compilation error 😕

Can you paste some code?

just trying to start with empty project and latest deps.. lein project with cljsbuild :main to a cljs.user file

(ns cljs.user
  (:require
    [om.next :as om]))

(enable-console-print!) 

Have you tried to clean the output dir? ie rm -rf ...?

@pedroteixeira ^

I removed :cljsbuild :compiler :preloads [devtools.preload] and got it to compile with lein cljsbuild but still fails via figwheel with Invalid :refer, macro cljs.pprint/deftype does not exist I notice that with lein cljsbuild I get some warnings:

Compiling "resources/public/compiled/cards.js" from ["src/clj/client" "dev/cards"]...
WARNING: Use of undeclared Var cljs.pprint/pprint at line 15 resources/public/compiled/cards/devcards/util/utils.cljs
WARNING: Use of undeclared Var cljs.pprint/with-pprint-dispatch at line 23 resources/public/compiled/cards/devcards/util/utils.cljs
WARNING: Use of undeclared Var cljs.pprint/code-dispatch at line 23 resources/public/compiled/cards/devcards/util/utils.cljs
WARNING: Wrong number of args (1) passed to UUID at line 564 resources/public/compiled/cards/cljs/reader.cljs
WARNING: Use of undeclared Var clojure.string/starts-with? at line 822 resources/public/compiled/cards/cljs/analyzer.cljc
WARNING: Use of undeclared Var clojure.string/ends-with? at line 972 resources/public/compiled/cards/cljs/analyzer.cljc
WARNING: Use of undeclared Var clojure.string/starts-with? at line 2643 resources/public/compiled/cards/cljs/analyzer.cljc 

perpahs some mixed clj/cljs versions? will probably consider downgrading from these latest versions for now

thank you all for being such a wonderful community ❤️

@sophiago Is there a single quote instead of backtick in the macro bar?

no, but i do use a let statement outside of the quasiquoted part (a hack to evaluate gensyms)

that works fine in clj

Can you post that macro bar?

Looks maybe as if (macros/bar ..) macroexpands to ('() ...)

well, yes given the let. but that's never been an issue when in the same clj namespace

(defmacro square-limit [image n]
  (let [img (image)]
    `(do
       (corner-split (flip-horiz (below-top (beside-left ~img))) ~n)
       (corner-split (below-top (beside-right ~img)) ~n)
       (corner-split (flip-vert (flip-horiz (below-bottom (beside-left ~img)))) ~n)
       (corner-split (flip-vert (below-bottom (beside-right ~img))) ~n))))

^ the actual macro i'm using

oy, now i'm wondering if i need to qualify all those functional calls or if the expansion just occurs in the other ns and it's fine once required?

Maybe that (image) is the problem, you could move the backtick one level up and gensym the img.

sorry, i forgot earlier. usually i use that hack for gensym. in this case i'm just passing a function in userland

Yes (image) is just in the reader, isnt treated as function. So it isnt called. Youd need eval bit it doesnt exist in cls

but i don't see how that could be the problem when it expands fine when i don't need to require it to a cljs ns

unless it's not being expanded because the function calls are in the cljs namespace, which i haven't required to the clj one...etc. etc.?

do you know the answer to that last question? seems wise to rule out first

No, shouldnt matter. You can require .clj macros and use then in .cljs, ive only encounter problem with that using lumo, but then id need to clean the cash. Maybe recompiling would be needed in your case. So you should be able to rule that out.

Delete /out?

Is the first argument being passed to bar a form that looks like a list?

@mfikes it's a function literal

@hlolli i mean when your macros call functions from the cljs ns...whether you need to require that ns to your clj one and then qualify the function calls?

AFK, but it might be like ('(fn []))

@mfikes where are you going with this? it does seem to match the error message exactly...

I think it fits the message. Dont know what qualifying a function means.

Just mentally evaluating (image). Need to check at a computer :)

@mfikes thanks! testing my other hypothesis now

@hlolli as in (macros/bar ...) vs. (bar ...)

Try adding the backtick to front of let and give img# a hashtag at the end, i think it will work.

welp, my hypothesis about the function calls obviously made no sense since you cannot require a cljs ns into a cljs one!

what does (image) return?

a keyword stored in a global map

@hlolli to be clear, you're saying (let [img# (image)] ...)` will cause the call to image to be evaluated?

Yes, then the macro returns an expression that knows what to do with (image) otherwise its just in a reader, so that you could for example apply functions to the argumwnts of (image) before evaluation.

i think he's saying the call to image will not be evaluated at macroexpansion

and will instead be evaluated at runtime

as-written, your macro evaluates (image) at macroexpansion

oh...i think i'm starting to understand this

btw, you should call macroexpand-1 on your macro and see what you get

that hack with let works when it's in the same ns, but when requiring a macro from another you don't want anything actually evaluated at expansion

so in this case, i actually do need runtime evaluation for that call to make any sense. hence why it's being interpreted as a list

yea so image might be bound to nil

and you're trying to evaluate ('())

at macroexpansion

so it might make more sense for you to forego the let entirely and replace your ~img with (~image)

if what you want is to call image at runtime

well, the let still makes sense for efficiency, it just needs to be inside the quasiquote

true, and then you want img to be a gensym

no, not in this case

anyway, now i'm running into a problem where all the functions i'm calling in that macro are undeclared. i assume i need to unquote them?

(the same for the img var even though it's now inside the quasiquoted body fwiw)

symbols inside a syntax-quoted form are automatically prefixed with the namespace in which you're expanding the macro

so i think it's more likely that the functions actually haven't been declared

they haven't in the clj namespace. i asked that question several times above

sorry, i'm a latecomer

np

but if that's the issue i'm not sure how to resolve it considering you can't import a cljs ns into a clj one

i thought the answer before was that since nothing is being evaluated in the clj ns i'm fine

ugh this is more complicated than common lisp 😅

seems maybe i just need to qualify those function calls with the cljs ns?

I suspect you could write

(let [img ((eval image))] ...

@mfikes that would work as well. but now i'm on to another problem 😕

it's telling me all the function calls in the macro (which are from the cljs ns) are undeclared. i'm confused since that's the ns the macro is actually being evaluated in

@mfikes actually your way is the way i should have been doing it...not sure why appending gensym to the local var was suggested

@sophiago Is square-limit defined in a Clojure namespace with a name matching the one where the ClojureScript function corner-split is defined?

no, they're different

Syntax quote will end up qualifying the corner-split symbol with the namespace where the macro is defined.

ah! that's what i was asking all along

what's the way around that, seeing as i can't import the cljs ns to the clj one?

Just qualify the symbol with the ClojureScript namespace.

See foo.core/add* here: http://blog.fikesfarm.com/posts/2016-01-05-clojurescript-macros-calling-functions.html

that should have been obvious 😛

but it's also telling me the local var img is invalid

wow, this is so verbose since i can't alias the cljs ns

but the local var img is still giving me trouble. i suppose i need to qualify that with the cljs ns as well?

Try macroexpand and see what you are getting

do you mean expand it without args? in that case i get "no such var." expanding it as i'm calling it just qualifies the function i'm passing

Just wondering if the result of macroexpand on the form results in valid ClojureScript.

on just the form it says it can't find it in its own ns... not sure why that is.

Oh, (macroexpand '(macros/bar ...)) where you replace ... with the arguments

That fails?

oh no, that just gave me back exactly how i called it except qualifying the function passed

Whatever that passed back will be evaluated as ClojureScript. Perhaps that's not what you want.

that just passed back the function call, i.e. (macros/square-limit (...) ...) except with the function qualified

Hmm. That sounds like it doesn't realize square-limit is a macro

right

ugh, i'm aware how much space i'm taking up in this channel (maybe fine since it's Sunday) but this is really baffling me