[FIXED] just disable parallel build I'm using the "webpack method" and it's working ok for dev, but when I try "advanced" build, it warns about some goog's namespaces.... I can safely ignore it?

Compiling ["resources/public/javascript/main.js"] from ["src"]...
Jul 08, 2018 10:58:29 PM com.google.javascript.jscomp.LoggerErrorManager println
WARNING: target/cljsbuild-compiler-0/inferred_externs.js:39: WARNING - name goog is not defined in the externs.
goog.DEBUG;
^^^^

Jul 08, 2018 10:58:29 PM com.google.javascript.jscomp.LoggerErrorManager println
WARNING: target/cljsbuild-compiler-0/inferred_externs.js:40: WARNING - name goog is not defined in the externs.
goog.string;
^^^^

has anyone else encountered this error during closure compilation?

Jul 09, 2018 8:21:42 AM com.google.javascript.jscomp.LoggerErrorManager println
SEVERE: /out/cljs/core/async.js:7197: ERROR - Parse error. '}' expected
cljs.core.async.t_cljs$core$async61681.getBas
                                             ^
 
Jul 09, 2018 8:21:42 AM com.google.javascript.jscomp.LoggerErrorManager println
SEVERE: /out/js/radiant/test-runner/out/cljs/spec/alpha.js:1120: ERROR - Parse error. ',' expected
cljs.core._STAR_print_fn_STAR_ = ((function (_STAR_print_newline_STAR_61674_61684,_STAR_print_fn_STAR_61675_61685,sb__4432__auto__,ed,arg_spec,temp__5459__auto__,spec
                                                                                                                                                                      ^
 
Jul 09, 2018 8:21:42 AM com.google.javascript.jscomp.LoggerErrorManager printSummary
WARNING: 2 error(s), 0 warning(s)
ERROR: JSC_PARSE_ERROR. Parse error. '}' expected at /out/js/radiant/test-runner/out/cljs/core/async.js line 7197 : 45
ERROR: JSC_PARSE_ERROR. Parse error. ',' expected at /out/js/radiant/test-runner/out/cljs/spec/alpha.js line 1120 : 166
java.lang.Exception: Closure compilation failed

using: [org.clojure/clojurescript “1.10.126”]

and lein-doo for test running https://github.com/bensu/doo

@biscuitpants looks like that CLJS output files were truncated or did not finish writing. closure probably has nothing to do with this. I have seen similar strange results like this when running 2 instances of the CLJS compile in parallel that write to the same output-dir. or maybe the disk is just full? should get an exception during CLJS compilation then though so I'm just guessing

interesting thanks @thheller - i don’t think more than one instance of the compiler is running. is it possible that a logger could be messing with the writer to the file?

i don’t think the disk is full

no logger no but if you can check the /var/lib/buildkite-agent/builds/BuildkiteSlave-i-08751208534127368-1/cognician/monolith-develop/out/js/radiant/test-runner/out/cljs/core/async.js if it actually ends in .getBas

ok i’ll do that, trhying another build now

@thheller so it looks like if i turn off :language-in :ecmascript5 :language-out :ecmascript 5 compilation works again 🙂 we had that turned on for a bug in one of our dependencies. luckily, that bug has since been fixed upstream

I don't see how that would produce the above situation but when it works it works 😉

neither, but with an entirely clean build with the option ON, it breaks. with a clean build, with it OFF, it works

did you check if the files are actually truncated on disk?

not yet. trying to reproduce the broken situation again. i cleared .cljs/ and now i can’t hit that closure compilation issue again

What’s the state-of-the-art for re-frame table views, with sortable columns, filtering badges, etc?

Is there a component or group thereof that would be a good start?

I use https://react-table.js.org which works well. I do not use re-frame though so can't comment on that

@thheller I’ll check it out.

I'm looking to get a node based cljs.main repl with deps.edn up and running with an nrepl server, so I can connect from cider or proto-repl or anything. Seems like a common need, but I'm flailing 😞 I'm trying to move to deps.edn for new projects, after using lein project.clj until now. Does someone have a pointer to a gist or such maybe?

@vinai I'm going to recommend figwheel.main and CIDER here

There isn't a write up on how to do it yet, but it's pretty straight forward

I'll be happy to help if you have hiccups

@vinai you can also try shadow-cljs, it has a good integration with deps.edn and its strait forward to use a node repl with it: https://shadow-cljs.github.io/docs/UsersGuide.html#deps-edn / https://shadow-cljs.github.io/docs/UsersGuide.html#_node_repl

@bhauman Even with node? This is for a non-browser project. I thought figwheel was intended for browser use (at least that's how I've been using it).

@wilkerlucio Thanks! I will check it out. Would have liked to use plain cljs.main to be honest, just for the sake of understanding the setup better. Still, will try it out.

This seems like good advice for a cljs + figwheel browser repl, but like I said, no luck with node yet. https://gist.github.com/reedho/302fabea63ab3416200462e192c39827

@vinai figwheel.main works great with node

and with figwheel.sidecar you will need to launch node manually

after you complie

So it would be responsible for recompiling changes and loading them into cljs?

you can use it with or without auto reloading

but it will watch and compile files and notify you when you get things wrong

Brilliant!

and it will autoload them as well

I have this set up in one of my deps.edn-based projects. The one bit I haven’t explored is the nREPL aspect.

:aliases {:repl       {:main-opts ["-m" "figwheel.main" "-co" "compile-opts.edn" "-co" "{:main,my-project.core}" "-ro" "repl-opts.edn" "-d" "target/dev" "-t" "node" "-r"]}
           :build-dev  {:main-opts ["-m" "cljs.main" "-co" "compile-opts.edn" "-d" "target/dev" "-t" "node" "-c" "my-project.core"]}
           :test       {:main-opts ["-m" "cljs.main" "-co" "compile-opts.edn" "-d" "target/test" "-re" "node" "-m" "my-project.runner"]}
           :build-prod {:main-opts ["-m" "cljs.main" "-co" "compile-opts.edn" "-d" "target/prod" "-O" "simple" "-t" "node" "-c" "my-project.core"]}}}

https://github.com/bhauman/figwheel-main#working-with-nodejs

This gets you a figwheel.main running reloading in node with clojure -Arepl

@mfikes I really reccomend the build file route 🙂

Yeah, I’m willing to try that @bhauman. Just exploring different ways to do this myself. 🙂

its just that having a stable unique name makes repl connections and automatic output-dirs easier to isolate

and maintain

@vinai I have figwheel.main working quite nicely in cider

with a pretty simple set up

@vinai there is also a #figwheel-main channel

Thank you very much, this is great. I am unfortunately on a meeting call atm, but am eager to try it out!

Noted down all the information.

@bharathmuraleedharan There is also a #cljsrn channel if you don’t get help here

okay thanks @mfikes, will try there too

Hi. Am I right in assuming that the :pseudo-names flag should not make any difference to the dead code elimination?

I have a situation where my app runs fine with :pseudo-names true, but fails with a "is not a function" error when its false. Stepping through the code in the browser debugger seems to suggest that the error happens in a React function, but there should be no UI stuff involved at this point. Stepping through with pseudo-names on, React is nowhere to be seen

The code in question is pretty simple:

(merge validated-word
            (if (contains? (:next validated-word) :end)
               (parse accumulated-words)
               {:valid-cmd false}))

:pseudo-names is a debugging feature

it’s not really meant to be used

yes, I enabled it in an attempt to debug 🙂

your issues sounds like an externs problem to me

yes, but there is no javascript dependenies involved here

the parse function is a javascript thing, but I can tell it finishes its job by littering the code with console.log statements. The failure seems to happen in merge itself

does parse have an extern?

yes

just three lines:

var parser;
parser.parse;
parser.parseError;

and they have been working fine for a long time

I can try moving the call to parse out into a let binding to see if it fails earlier

did you try :pseudo-names false :pretty-print true?

Yes, it ran passed the parser, and I printed the result of the parsing which is just a map. It still crashes in the call to merge

to confirm it’s not renamed?

“and crashes in the call to merge” is lacking in detail 🙂

yes, I've tried all combinations of pseudo-names and pretty print

also how can merge work if parse is a JS thing?

I realize that. sorry. I am not able to make a small example

well, its a js thing wrapped in a cljs function

but what is the actual error message?

TypeError: v.j is not a function

what does this mean “its a js thing wrapped in a cljs function”

is parse a js thing or not?

v.j seems to indicate exactly what I guessed - externs aren’t working

I would ignore that “it worked before”

this is the actual function called

(defn validate-with-parser
  "transform result of parsing s into a predictable hash"
  [s]
  (let [parser-js-response (build-cmd (string/trim s))

        parser-response    (js->clj parser-js-response :keywordize-keys true)]
    (when-not (helpers/d "parser-response type " (:type parser-response))
      (helpers/error-log! "WARNING! Parser disagrees with Rose"))
    {:command    parser-response
     :type       (:type parser-response)
     :valid-cmd  (not (nil? (:type parser-response)))}))

still not enough information

what is parse?

what you’re showing is very disconnected

is parse parser.parse??

sorry, I renamed validate-with-parser to parse in the first snippet. "validate-with-parser" is the actual function called, and build-cmd is an even thinner wrapper:

(defn build-cmd
  "Wrap the javascript RAIRD syntax parser. Returns cmd objects that can be
  sent to server. Makes exceptions look like regular responses"
  [s]
  (helpers/d "Calling parser with " s)
  (try
    (.parse js/parser s)
    (catch :default e e)))

so build-cmd is what’s failing, yes or no?

no, build-cmd is returning a map, like expected

then did you get nitty gritty and try to log right before the failure?

in validate-with-parser

to find the offending line

yes, I log the response from validate-with-parser (which is now run before the merge in a let), and it prints out the expected map

trying to avoid talking in circles here 🙂

if it prints, then what’s wrong?

after it prints, I get the v.j is not a function error

so it can’t be in validate-with-parser

that invalidates the path you’re leading us down

exectly

so it has to be somewhere else

I would stop looking here

or get ready to at least

do you believe it’s validate-with-parser because source maps points there?

But if this is an externs thing, should it not fail even with :pseudo-names enabled?

I would stop thinking about :pseudo-names

ignore whatever other variables that might be introducing

just fine the actual location of the error

no, because I've littered the code with console.logs, and then searched for those strings in the compiled code, placed breakpoints and stepped through

:pseudo-names really can’t tell you anything - in general i find it works only if it fails with :pseudo-names too

if it doesn’t - don’t worry about it - you can’t use it

ok. good to know

one trick I’ve used that could help find the source of the error, more so than printing

you want to find the last log before the error

the error should prevent logging after

then you can search the advanced compiled code :pretty-print true for your debug string

Closure won’t move side-effects

I have logging before and after the call to merge. The one before prints, the one after does not. both maps that are fed to merge prints fine

I think this is probably the clue

you’re not using JS values in the map keys are you?

map keys should be immutable JS values (booleans, strings, numbers) or ClojureScript values only

this is the map returned from the parser:

{:command {:type "statOp", :command "tabulate", :variables ["kjon"], :subset nil, :options []}, :type "statOp", :valid-cmd true}

that’s merged?

no, that is before the merge. Its the second parameter to merge

and the other map is the same? Also what is v in the debugger? null? something else?

bbiab

the other map is just s cljs map. not touched by any external js libraries. v looks like a map also

its not null

Hi all, I'm trying to run a cljs function using the cli like: clj -m cljs.main -re node -m seeds.core however I'm getting that it can't find a node dependency (even though it is inside of the node_modules folder). Anyone know what I might be doing wrong?

@dehli you can't use npm dependencies without declaring them first to the compiler

Would that be using npm-deps?

@dehli actually it depends on how you are requiring it

I'm using node/require

hmmmm try (js/require and make sure that you are launching from the root dir of your project

which I'm sure you are doing

@dnolen I replaced the call to build-cmd with a hardcoded map, and that made the problem disapear. I am somewhat confused, but at least I now know that the problem is indeed the js parser 🙂 Thank you for pointing me in the right direction, and sorry for wasting your time.

Thanks! Using js/require gives the same error. Note that I don't get this problem when I use cider-jack-in-cljs and I can build the project fine. It just throws this error when I'm trying to run that individual main

@dehli what is the specific error you are getting

Error: Cannot find module 'aws-sdk'
    at Function.Module._resolveFilename (internal/modules/cjs/loader.js:594:15)
    at Function.Module._load (internal/modules/cjs/loader.js:520:25)

yeah so this is getting out of my area of expertise

it does look like main is doing something differrent

No worries 🙂 Thanks a lot for your help! I'll keep poking around

guys, does somebody know why i get false here

(extend-type object
	IPrintWithWriter
	(-pr-writer [obj writer _]
		(write-all writer "blah-blah")))
(print (implements? IPrintWithWriter (js-obj))) ;; => false

@akond implements? doesn’t check native objects like satisfies? does. In short, since the printing code doesn’t check satisfies? you can’t override printing of object

is it made for a purpose?

i mean the fact that i cannot override printing of object?

https://github.com/clojure/clojurescript/commit/7d3ce70a1c2f4d54b9044d2e44b62c8252e5dea6

I don’t know the rationale for why printing doesn’t fall back to satisfies?

Perhaps it was an oversight, with the last revision in the commit referenced above. Evidently prior to that change it was possible.

you think i should ask david?

@akond I’ve posed the question in #cljs-dev

oh, thank you, pal.

@akond @mfikes this seems like a misunderstanding about the original question

no your example will not work, intentionally

but you can still customize object printing and that should work

consider implements? a performance footgun - satisfies? should be preferred

@dnolen you mean i am not allowed to extend js base types?

@akond no, I’m just talking about implements? don’t use it

@akond there is a bug here, @mfikes filing a minor issue etc.

ok. thank you.

What’s the best way to elide some code in Clojurescript? I’ve been using goog-define, but am wondering if I could do so with a macro

@colindresj For library code, goog-define is better because it is compatible with the AOT cache feature. (See the note at https://clojurescript.org/news/2018-03-28-shared-aot-cache)

I’m running into a funny problem: I am building a small Tic Tac Toe page to get familiar with basic web design, react, ant design, and clojurescript. Here is the snippert of my code: [antizer.reagent/button {:value i :on-click #(do (println “Button: ” (.-value (.-target %))) (re-frame/dispatch [::events/square-clicked (-> % .-target .-value)]))} (or @sq “foo”)] Here is the translated HTML <button value=“7” type=“button” class=“ant-btn”> <span>foo</span> </button> So the text foo is a “span” tag in the button. What i see: Now if i click on the “foo”, the “on-click” handler on the button get triggered, but the “value” is nil. If i click outside the “foo” text (on the actual button) then the value is what i am expecting, in this case “7” I think this is becasue when i click on “foo” text, i am picking up the attributes of the “span” tag and it doesn’t have a “value” attribute attached to it. In this case, why does the “:on-click” handler still get activated? What i want: click anywhere on the button, and extract the correct “value” attribute.

reason for adding a “value” attribute to each button: i have 9 squares for “Tic Tac Toe”, and I am associating a value to each square from 1 to 9. when a square is clicked, I know which square from reading the value attribute is clicked and I can update my DB accordingly and calculate a winner after each click.

Got it, thanks @mfikes

@dhruv1 First a tip: if you put triple backticks around your code, it will format nicely in Slack. Second, I’d recommend passing in value as a parameter to your component code, like this (or similar)

(defn my-tic-tac-toe-button [value]
  [antizer.reagent/button {:on-click (fn [e] 
                                       (println "Button:" value)
                                       (rf/dispatch [::events/square-clicked value]))}
   (or @sq "foo")])

ahh that’s right, i forgot about tripple backticks. thank you!

@manutter51 thanks. the value is associated with the button so I know which button was clicked if you’re asking me to pass in the value, then i am recording the value somewhere else. then how do i know which button was clicked?

What you do is make a button component (which you’ll probably call something nicer than my-tic-tac-toe-button like I did in my example). Then you’ll build your tic-tac-toe matrix using [my-tic-tac-toe-button 1], [my-tic-tac-toe-button 2], and so on

Each button will have its own value, which you passed in as the argument to my-tic-tac-toe-button

hmm. that’s actually what I’m doing here is the whole functino:

(defn square [i]
  (let [sq (re-frame/subscribe [::subs/square (str i)])]
    
    [ant/button {:value i
                 :on-click #(do
                              (println "Button: " (.-value (.-target %)))
                              (re-frame/dispatch [::events/square-clicked (-> % .-target .-value)]))}
     [:i {:class-name "far fa-circle"}]])) 

sorry ignore the [:i …] in the end. it was me trying something else

Ok, good, but when you dispatch your event, don’t get value from .-target .-value, get it from your argument i

oh okay

neat

is there a way to tell the closure compiler not to eliminate a namespace?

atm I'm having to add in very explicit side effects like a println in order for it to do what I want <_<