I've got a cljs.test failing. It's a simple (is (= a b)) test, and the Actual output shows no difference. Like I literally copy-paste the Actual into a repl and it returns true.

Fail (thrower) (com/grzm/dev/cljs_stacktrace/util_test.js:1269:51)
	Expected (= first-file-expected first-file-actual)
	Actual: (not (= "resources/public/js/compiled/out/test/com/grzm/dev/cljs_stacktrace/helpers.cljs" "resources/public/js/compiled/out/test/com/grzm/dev/cljs_stacktrace/helpers.cljs"))

Any ideas as to what I should look at?

I'm tempted to say types (but I don't think that makes sense)? have you tried == instead of =?

Yup. Same result. (btw, I think == is just for numerical comparisons, as I understand it. http://clojuredocs.org/clojure.core/==)

I've also checked (= (meta a) (meta b)), and that succeeds (just trying to cover all my bases)

maybe (= (str first-file-expected) (str first-file-actual))??

(clutching at straws...)

I appreciate you're efforts 🙂 Makes me feel less alone. That doesn't work either.

yeah, I can get lonely on this channel... I think I must be in the wrong timezone of something...

yeah, it's after 8PM local for me. It looks like it's midday for you?

@noisesmith I appreciated your help earlier with the stateful component question above. I thought through this for a while and came up with another alternative based on your idea I'm going to start using. This lets me create reusable stateful components without always needing to know the state-shape the component expects. here's a gist: https://gist.github.com/zakpatterson/4df511752b4f53d23deef4de309644c3

yeah, 1PM

so @grzm, what are you comparing exactly? are they file objects? can you get the absolute paths?

I'm parsing stacktrace frames.

The stacktrace is just a vector of maps.

and you're pulling out the first map from the vector... :thinking_face:

Yeah. The higher level (which is also failing, of course), is (vec (take 2 stacktrace)). Trying to dig down to see what's different, I'm comparing the values of the maps. That filename value is what's failing the equality test in each frame.

so that file value can't possibly be a simple string, right? otherwise the test would pass? odd... what could it be?

That's a good idea…

Expected (not= (type first-file-expected) (type first-file-actual))
	Actual: (not (not= #object[String] #object[String]))

I tested using not= so we can see the output. So, that's not it, either 😕

well, that's frustrating... how about using clojure.string/includes?

i.e. (and (includes? first-file-actual first-file-expected) (includes? first-file-expected first-file-actual))

maybe there's a \ff or \cr in there...

I'm thinking you need to use hexdump

Yeah, that's a good idea. There's definitely some extra characters there. The counts are different.

Now we're getting somewhere. 🙂

maybe just use trim and be done with it

Somehow there's a "base/" prefixed to the actual string! Why isn't that showing up?

Is the string some kind of array buffer that has an offset?

really, that would seem to be a greater mystery

what, like a golang slice? I don't think so!

I was thinking more like a Java bytebuffer, but yeah, neither should be the case in cljs

Cleary I'm missing something.

that would seem very un-clojure... but you know: javascript!

that's both the problem and one of cljs best selling points: you can't get away from the host system

so the diff of the counts is 5?

or is there like 5 backspaces so that the string prints that way?

Yup. There's a goog.string.StringBuffer class used in str: https://github.com/clojure/clojurescript/blob/master/src/main/cljs/cljs/core.cljs#L2923-L2935

how about

(is (= (class a) (class b)))
(is (= (count a) (count b)))
(is (= (str a) (str b)))

perhaps (seq a) (seq b) would be revealing, even if they print the same as one another

count are different, str is different, type is different. and yeah, I used seq to reveal the "base/" at the beginning of one of actual.

hmm, so you can use str in your test?

or just put "base/" in your expected string? You have this solved, right?

(Does class exist in cljs?)

(I didn't think so)

type is better regardless

Here's a summary of what I'm looking at:

Fail (thrower) (com/grzm/dev/cljs_stacktrace/util_test.js:2367:51)
	Expected (= first-file-expected first-file-actual)
	Actual: (not (= "resources/public/js/compiled/out/test/com/grzm/dev/cljs_stacktrace/helpers.cljs" "resources/public/js/compiled/out/test/com/grzm/dev/cljs_stacktrace/helpers.cljs"))
	
	Fail (thrower) (com/grzm/dev/cljs_stacktrace/util_test.js:2367:51)
	Expected (= (count first-file-actual) (count first-file-expected))
	Actual: (not (= 84 79))
	
	Fail (thrower) (com/grzm/dev/cljs_stacktrace/util_test.js:2367:51)
	Expected (= "" first-file-actual)
	Actual: (not (= "" "resources/public/js/compiled/out/test/com/grzm/dev/cljs_stacktrace/helpers.cljs"))
	
	Fail (thrower) (com/grzm/dev/cljs_stacktrace/util_test.js:2367:51)
	Expected (= [] (take 10 (seq first-file-actual)))
	Actual: (not (= [] ("b" "a" "s" "e" "/" "r" "e" "s" "o" "u")))

that's so strange, but if you use (= first-file-expected (str first-file-actual)) then it passes?

because goog.string.StringBuffer is doing something strange?

Nope. That fails, too.

I mean, yes, that's good

Expected (= (str first-file-actual) (str first-file-expected))
	Actual: (not (= "resources/public/js/compiled/out/test/com/grzm/dev/cljs_stacktrace/helpers.cljs" "resources/public/js/compiled/out/test/com/grzm/dev/cljs_stacktrace/helpers.cljs"))

try (= (str "base/" first-file-expected) (str first-file-actual))

or change your def of first-file-expected

or just (= (str "base/" first-file-expected) first-file-actual)

Appending "base/" works. What's so curious is why it prints the way it does.

yes, I agree

but you might have to take that up with the Google Closure folks 😕

It looks like "base/" means something special

you could have a race condition... goog.string.StringBuffer is mutable 😞

My test assertion looks like: (is (not= (str first-file-actual) (str "base/" first-file-expected))) (using not= so I can see the failure results)

The failure results look like:

Expected (not= (str first-file-actual) (str "" first-file-expected))
	Actual: (not (not= "resources/public/js/compiled/out/test/com/grzm/dev/cljs_stacktrace/helpers.cljs" "resources/public/js/compiled/out/test/com/grzm/dev/cljs_stacktrace/helpers.cljs"))

wait, why does "base/" just disappear?

I know, right?

oh... it's disappearing on the expected value? I thought it was the other way around... 😕

wait it even disappears in the first line of output...

No, I'm appending it to the expected like you suggested. But look at how it's printing.

This doesn't happen in my browser repl, so I don't know what's going on 😕

it's like the output is piped through sed /base\//s///

I think my brain is full.

😆 I don't blame you

I'm now thinking there's something else in your toolchain that's wreaking havoc

Yeah. It could be doo or karma or chrome

the toolchain is the worst part of web dev 😞

same thing happens with safari as the browser

Yeah. One of the things I'm trying to do here is ensure portability. Toolchain improvement is actually my goal.

I'm working on using source maps to improve stacktrace readability.

hey, maybe if you look at something else for while you'll come back to it with a fresh perspective, here is the problem I'm facing 😉

I'm attempting to use https://github.com/jasondavies/d3-parsets in a project... but I get this error from boot cljs:

java.lang.IllegalArgumentException: Duplicate module path after resolving: /home/corin/Projects/Demos/boot-parsets/node_modules/d3/d3.js

lol

nice segue

when I google the error, I get the source code the emits the error... I hate it when that happens

What else do you have on your filesystem besides the repo? From a fresh checkout, when I run boot cljs I get a different error: SyntaxError: Block-scoped declarations (let, const, function, class) not yet supported outside strict mode

oh... um, let me try a clean

I don't get a SyntaxError from a clean repo... I've built clojurescript from master in the past... maybe it's that...

I tried blowing away ~/.m2/repository/org/clojure/clojurescript but still no SyntaxError

is it complaining about es6/d3/parsets.js?

It looks like it's higher up than that:

Writing main.cljs.edn...
Compiling ClojureScript...
• main.js
[eval]:1
let fs = require('fs');
^^^

SyntaxError: Block-scoped declarations (let, const, function, class) not yet supported outside strict mode

what? eval? I don't know where that comes from

me, neither

what's your boot version?

I got 2.7.2

same

 boot -V
#
#Sun Nov 19 14:45:07 AEDT 2017
BOOT_CLOJURE_NAME=org.clojure/clojure
BOOT_CLOJURE_VERSION=1.8.0
BOOT_VERSION=2.7.2

I haven't compiled cljs with boot, though, so there might be something else wrong with my setup.

I can't even find "let fs" in the repo.

λ boot -V
#
#Sat Nov 18 21:51:17 CST 2017
BOOT_CLOJURE_NAME=org.clojure/clojure
BOOT_CLOJURE_VERSION=1.9.0-RC1
BOOT_VERSION=2.7.2

it looks like nodejs...

do you get

Classpath conflict: org.clojure/clojure version 1.9.0-RC1 already loaded, NOT loading version 1.8.0

yup.

I get the same thing when using 1.8.0

I tried BOOT_CLOJURE_VERSION=1.9.0-RC1 boot cljs but I don't get SyntaxError...

λ node --version
v4.8.4

v6.9.4

that's probably it

updating...

ok, so cljs is trying to npm install and can't

wow... how do I specify the node version in my project config... it's the toolchain again!

I think that's handled by $PATH

so, if you can update PATH via boot, then you're good

so, you get the same error as me now?

I'm still updating, actually. 😕

oh, fair enough

yay, toolchains. I'm dealing with dependency hell

well, now I get the same error you do.

one thought I had was that there were multiple inclusions of d3 in node_modeules, but I see only one.

yeah, often you get a node_modules dir inside other modules, but that thought didn't get me anywhere

The comment in the closure-compiler near that error is // Having root paths "a" and "b" and source files "a/f.js" and "b/f.js" is ambiguous. https://github.com/google/closure-compiler/blob/master/src/com/google/javascript/jscomp/deps/ModuleLoader.java#L275

Looking at the project:

λ find . -name "d3*"
./es6/d3
./node_modules/d3
./node_modules/d3/d3.js
./node_modules/d3/d3.min.js
./node_modules/d3/src/d3.js

Perhaps the conflict is between the two d3.js files. I'm spitballing here.

I'd think :npm-deps is smart enough to handle that, but I'm out of my depth here.

Looks like the same issue using the cljsjs package.

hmm, fair comment, changing the name of the namespace has no effect 😞

rename es6/{d3 => }/parsets.js

I feel like I need to throw in a lot of println into ModuleLoader.java....

ok, I'm doing it, I'm building com.google.javascript/closure-compiler-unshaded off master

good luck 🙂 An alternative would be to try lein. I think there are examples using d3 out there.

I know! I'm converting a project from lein to boot... I had it working with lein but for some reason it broke... I'm deep in the rabbit hole now!

ah. gotcha

Hi. Are there any well-known pointers available on the Internet on how to create a library targeting both Clojure and ClojureScript?

I have found that most examples and tutorials about cljs that have a server side component use Clojure (on the server side)

have you seen https://github.com/magomimmo/modern-cljs?

chapter 11 discusses the use of Clojure on the server

@au-phiware I guess the Modern-CLJS approach is too web heavy for me, and I am looking for ways to test the library under both CLJ and CLJS.

oh, I see, I've been focused on web, so I haven't come across any

so, your looking for cljc centric resources?

yes, CLJC with occasional CLJ and CLJS dependency for platform-specific bits

Is there a way to require macros in CLJS so that I can refer them namespaced? E.g. (foo/bar :baz) — bar is a macro here.

If I do (:require-macros [myapp.foo :as foo]), then I can no more refer to functions in foo?

@kumarshantanu See the Implicit macro loading: section of (doc ns). Also see more exposition at http://blog.fikesfarm.com/posts/2016-03-01-clojurescript-macro-sugar.html

I have a small spec-serialize library I only use myself, where almost everything is in one core.cljc file, using (ns spec-serialize.core (:require [#?(:clj clojure.spec.alpha :cljs cljs.spec.alpha :default clojure.spec.alpha) :as s])) then a few small .clj and .cljs files witch use specific code, and call a function on the core.cljc.

@gklijs You can simplify your ns form to

(ns spec-serialize.core
  (:require [clojure.spec.alpha :as s]))

Inside a defmacro, is the reader conditional supposed to always favor :clj? E.g. (catch #?(:clj Exception :cljs js/Error) e# (println e#))

"always" in your question is pretty tricky, think about self-hosted clojurescript where defmacro is actually cljs code

I don't remember the answer, but you should think about these corner cases if you want your library to be 100% portable including self-hosted

@darwin I’m using ClojureScript, not self-hosted one for now. (Digging into a puzzling error that complains: ReferenceError: java is not defined in CLJS)

I could imagine some clojure macro to be unaware of clojurescript and generating some lower-level java interop code. When you use this macro in clojurescript, it will happily spit such java-interopy code in the middle of your cljs code.

OK, so I was hit by the issue described here: https://groups.google.com/forum/#!topic/clojure/DvIxYnO1QLQ and here: https://dev.clojure.org/jira/browse/CLJ-1750 — fixed it with the (:ns &env) hack, but not sure if that’s supported by self-hosted CLJS. @darwin

Hello there! Question, is with-redefs somehow incompatible with :advanced compilation?

my codebase's tests use with-redefs extensively, suite breaks with :advanced wherever w-r is used

I get errors in the form $my$namespace$blah$$.$cljs$core$IFn$_invoke$arity$1$ is not a function

@kumarshantanu I'm sorry at some point I knew all the details of this, but I unloaded it from my brain... cannot help ATM

@vemv I'd suggest trying :static-fns false with :advanced to see if it might work for your case

seems a good call! trying, thanks

@vemv with-redefs is involved in a couple of compiler unit tests that run under :advanced, but none of them attempt to redefine a function (especially a multi-arity or variadic function)

@mfikes :static-fns: seems to have fixed it, thanks! 🍻 (can't be 100% sure, as suite now remains broken for new unrelated reasons)

Cool. That would make some sense, given the way with-redefs is implemented.

hi 😄 I am having trouble getting figwheel to work 😞 Figwheel compiles my js and starts but it reports Resource not Found when I try and connect

Aha, I see, I did not actually start the clojure-backend that generates the HTML page.

😄 nvm

om, reagent or re-frame, with which one should i start considering i'm just starting with ClojureScript?

from just taking a look over the README.md in these projects they seem to be the same thing, interfaces to React from ClojureScript

from my understanding om and reagent are independent projects that both attempt to provide a idiomatic clojure interface to react, and re-frame builds upon reagent

fwiw I chose reagent when faced with the same choice as I felt that api was simpler

@zignd I assume you have a javascript background, so will explain in those terms: reagent wraps React rum wraps React re-frame adds a redux-ish capabilities to reagent fulcro is om.next made easier. It wraps React, plus provides a client<->server story.

@zignd I would suggest you start with reagent , kick some easy early goals. Feel good about progress, and then consider next steps from that platform.

(but I am a bit biased, so listen also to others :-))

I was going to mention om.next but I have found it very hard to find resources for it... good to know that fulcro exists 🙂

thanks @au-phiware and @mikethompson, i think i will start with reagent as both of you recommend, and because its API seems more Clojure like and less JavaScript interop than Om, at least from what i could grasp by just checking their docs quickly, and i didn't know fulcro existed

If you do choose the reagent path, then look carefully through these "Domino 5" resources: https://github.com/Day8/re-frame/tree/master/docs#domino-5

(ignore the rest of that page, it relates to re-frame)

thanks for linking this resource, i'll check it out!

@au-phiware figured out the mystery of the missing "base/": it's stripped from output by the doo reporter: https://github.com/karma-runner/karma/issues/1706

unreal... I ended up finding a bug in cljs compiler with my issue, I even filed a patch... it was late at night so I won't be surprised if it gets rejected at triage!