Thanks!

No prob

I think I can make due with that. It seems like a nice pattern emerges with data structured as :company/name instead of {:company {:name …}}, --- things get flatter. If I have a lot of keys that I’m destructuring though and one is off then it seems like we aren’t able to let that be the exception and state it twice and everything else once. We have to state everything twice. 😕

You’re actually not committed to just that method of destructuring, if it helps - only the stuff that needs a different name

cljs.user=> (let [user {:user/keys {:id 56 :name "Jim Berlage" :cl "js"}}
       #_=>       company {:company/keys {:id 28 :name "Guaranteed Rate" :business "financial"}}
       #_=>       {{user-id :id :keys [name cl]} :user/keys} user
       #_=>       {{id :id company-name :name :keys [business]} :company/keys} company]
       #_=>   [user-id name cl id company-name business])
[56 "Jim Berlage" "js" 28 "Guaranteed Rate" "financial"]

You can still use :keys, :as, etc., and only specify that some variables have different names

(fn [obj] (let [v1 (goog.object/get obj "k1") v2 (goog.object/get v1 "k2") v3 (goog.object/get v1 "k3")] ... ))

is there a better way to do this? when obj is #js {} it crashes because goog.object/get will return undefined but won't take undefined

right now my solution is to always use nil as default value, but something tells me I would forget and crash.. Not to mention very verbose

anyone thought about destructuring macro for js objects yet?

@poernahi try getValueByKeys https://google.github.io/closure-library/api/goog.object.html

I've thought about a destructuring macro, yes, and I think it would be a good idea in principle (perhaps a js-obj aware version of let)

well you could always use (.-k1 obj) and so on. all it takes is one ^js obj hint.

is that an extension in shadow-cljs or available in the cljs compiler by default?

should work in CLJS but you may need to turn on some :infer-externs option first

it might complain if you only use ^js and not ^js/SomeClass. not sure actually.

Another alternative is to extend object to ILookup and then you can do

(get-in obj [:k1 :k2 :k3])

See https://stackoverflow.com/a/49848352/4284484

@thheller I can't get it to work with pure cljs: https://github.com/pesterhazy/see-el-jay-ess/blob/master/src/my/core.cljs https://github.com/pesterhazy/see-el-jay-ess/blob/master/README.md

@mfikes interesing, wonder if there's any side effect?

Personally extending to ILookup feels risky

actually, it may not work when the js object is deliberately created without inheriting js/Object

@pesterhazy I agree. I've been trying it for a while to see if anything breaks.

@poernahi The make effect is to make get work in the cond branch that checks ILookup. But I'd be curious if anything else goes wrong.

If we could get ^js obj type hints to work generally, like @thheller suggested, that would be a big step forward

When I try :infer-externs true with cljs.main, that leads to a NPE: https://github.com/pesterhazy/see-el-jay-ess/blob/master/README.md#infering-externs

@thheller I've updated my test repo to include a non-nested case in addition to the nested map

With infer-externs false property access returns nil; with infer-externs true it crashes.

@pesterhazy Even though you don't have a minimal repro, if your NPE stack is safe to put in a gist I'd be happy to take a quick look at it.

Oh the NPE is in your repo?

@mfikes it's a pretty minimal repro: https://github.com/pesterhazy/see-el-jay-ess/blob/master/scripts/run-test-deps

Cool!

just clone and run that script

Thanks!

scripts/run-test-deps --infer

Cool. I can repro.

Crap looks like it is Closure

Looks like it might be filed https://github.com/google/closure-compiler/issues/2629

@mfikes, I wonder if it works for others, and if so what's special about my repro-repo

just the fact that I'm using cljs.main?

@pesterhazy If you revise your deps.edn to point at ClojureScript master, it no longer crashes, and the printed output seems to imply it it is working (I haven't read what that repo's code is doing). Specifically this coordinate seems to be OK

{:git/url "" :sha "132d3aa232921a3cea66f830d61c89be78c581cb"}

@mfikes I can confirm that it doesn't crash anymore, thanks for that!

I'm trying to bisect to see what ClojureScript change caused it to work.

two observations: - with infer-externs the type hints don't seem to matter, given that demo, demo2 and demo3 print the same output - nested key access using (some-> obj (.-very) (.-lengthy)) doesn't seem to work

@pesterhazy This commit fixed it https://github.com/clojure/clojurescript/commit/30bc7c2ef251d74c2ab398c2a1461984f9c80469

that looks very plausible

By the way, as an aside, setting a :local/root to ClojureScript lets you do a git bisect directly on that repo to find what fixed or broke a project like Paulus's which uses deps.edn. TL;DR deps.edn made bisecting this trivial. 🙂

Isn't it great when a simplification (like using clj instead of lein) suddenly opens up unexpected possibilities?

(I didn't know, and if anyone else is wondering, :local/root is explained here: https://clojure.org/guides/deps_and_cli#_using_local_libraries)

@mfikes out of curiosity, do you manually accept or reject a step when using git bisect, or do you rely on the exit status of a script?

@pesterhazy Manually. Especially since I was running your repro to see if it worked or failed each time. It was only about 5 steps IIRC.

cool

you can destructure with something like {cname :company/name, uname :user/name, :company/keys [address], :user/keys [birthday id]} etc

so using a mixture of keys destructuring and renaming

I’m trying to add an npm dependency (`react-ace`) to my project (via :npm-deps), but I’m getting an error:

index.js:20 Uncaught ReferenceError: module is not defined
    at index.js:20

var freeModule$$module$project_path$node_modules$lodash_isequal$index=freeExports$$module$project_path$node_modules$lodash_isequal$index&&"object"=="object"&&module&&!module.nodeType&&module

you’re missing a ] in calcs

and a }

ah yeah I just typed that here, it’s right in the code 🙂

:thumbsup:

I’m in macro hell lol

@alex-dixon huh, this works?

(let [{:course/keys [header-image tags]} {:course/header-image "foo" :course/tags "bar"}] [header-image tags])

Is it still the case that CommonJS modules won’t work properly with cljs?

@borkdude this is new in 1.9 and is doc’ed at https://clojure.org/reference/special_forms#_map_binding_destructuring

but there is a ticket to improve that and extend the destructuring guide

Whoah! Super cool. TIL. That needs to be promulgated 🙂

I’m trying to google for the module is not defined error I pasted above, and this is all I can find — trying to figure out if it’s out of date or not: https://stackoverflow.com/questions/35489797/using-react-components-in-reagent

@wombawomba not sure about npm-deps, but you can make it work pretty easily using the "double bundle" method: https://github.com/pesterhazy/presumably/blob/master/posts/double-bundle.md and https://github.com/pesterhazy/double-bundle

that's still my personal go-to method until all the cljs-node-module-require kinks are ironed out

until now I didn’t realize you can mix keys with the other notation. I guess it’s just the unqualified version of the new feature:

(let [{a :a :keys [b]} {:a 1 :b 2}] [a b])

I’ve been using mostly :keys so far

multiple also supported:

(let [{:myns/keys [a] :keys [b] :myotherns/keys [c]} {:myns/a 1 :b 2 :myotherns/c 3}] [a b c])

yeah, none of that is new other than :myns/keys

yes, I thought they were exclusive and I managed to hold up that believe for years 😉

well, lots of people combine :keys and :as at the same time

but other than that, most people don’t do multiple kinds of associative destructuring in a single case

yes, I did that, but I mean the combination of {a :a} and :keys

ah, yeah, nothing new there

that’s 1.0 stuff :)

I’m just catching up with 1.0…

increasing use of namespaced keys has put a lot of pressure on associative destructuring over the last few years

@pesterhazy cool, I’ll try that approach! thanks

Is there a function that creates a namespaced map or is it just the reader macro?

@alex-dixon The concept of “namespaced map” doesn’t really exist. The keys in a map can be namespaced.

in other words, this is a syntax - the objects are identical