Can someone explain me how and if this should work in Clojure?

(let [{a :a
       {b :b} (name a)}
      {:a :a
       "a" {:b :b}}]
  (println "value of b" b))

(it was reported by a user of clj-kondo)

It seems the bound value of a is used in the "second" binding, but since it's a map destructuring, the bindings aren't ordered maybe, so I'm puzzled if this would work at all

whoa, that definitely smells like undefined behaviour, I had no idea that was valid syntax

Here's a more minimal example:

(let [{a :a, b a} {:a :x, :x :y}]
  b)
;; => :y

But reordering the keys in the destructuring form breaks it:

(let [{b a, a :a} {:a :x, :x :y}]
  b)
;; Syntax error compiling at (src/repro.clj:4:1).
;; Unable to resolve symbol: a in this context

(let [{_x  :a
       b1 :a
       c :a
       d :a
       e :a, f :a, g :a, h :a, i :a
       {b :b} (name _x)}
      {:a :a
       "a" {:b :b}}]
  (println "value of b" b))

Syntax error compiling at (form-init191682565753033842.clj:6:15).
Unable to resolve symbol: _x in this context

When you create a map using {}, it creates an array map if the number of items is <= 8. Otherwise, it creates a hash map. With array map, the order and hence the behavior is well defined. With a hash map - no so much.

This also doesn't seem to work:

(let [{a :a, b a, c b} {:a :x, :x :y}] c)
nil

I'm confused:

(let [{a :a, b a, c b, d c, e d} {:a :x, :x :y}]
  [a b c d e])
=> [:x :y nil nil nil]

How did it get to :y?

Oh wait, nevermind.

(this goes to demonstrate that such chaining can be confusing)

and undefined probably

Yeah, given your example, I retract my previous words about an array map resulting in a well defined behavior.

it makes more sense looking at the macroexpanded form of let , which is just a "dumb" syntactical transform that doesn't know about symbol resolution

I think clj-kondo is right to flag such things with at least a warning

yeah, might be nice

(as in, the current behavior already seems right)

But the (let [{a :a, b a, c b} {:a :x, :x :y}] c) is actually correct to return nil. Because there's no :y key in the map.

Wow, these examples are fun. My brain keeps turning back to them thinking that I still didn't get something.

Maybe a good reason to avoid it altogether

undefined behavior

it's not supposed to work

Thanks for confirming

agreed, you should not rely on this working

there is actually a ticket about this I think

not it, but similar: https://clojure.atlassian.net/browse/CLJ-1613

also not it, but similar: https://clojure.atlassian.net/browse/CLJ-1881

I did not find a ticket specifically about this (not involving :or)

WDYT of the following pattern for authoring macros that work in clj + cljs + self-hosted cljs?

(defmacro foo [compilation-target]
  (case compilation-target
    :clj  :clj
    :cljs :cljs))

(foo #?(:clj  :clj
        :cljs :cljs))

It's a bit noisy for consumers, but clean for the macro itself

(haven't tried it yet in self-hosted cljs btw - elsewhere it appears to work correctly)

hello, has anyone used GitHub Actions to run tests on a clojure project with leiningen, where a couple of the project dependencies are in a private S3 bucket? locally i've been using s3-wagon-private and aws access/secret keys in a ~/.aws/credentials file. i was hoping something like this would do the trick:

name: Clojure CI
on: [push]
jobs:
  build:
    runs-on: ubuntu-latest

    steps:
      - name: Checkout
        uses: actions/checkout@v1

      - name: Configure AWS Credentials
        uses: aws-actions/configure-aws-credentials@v1
        with:
          aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
          aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
          aws-region: eu-west-1

      - name: Install dependencies
        run: lein deps

      - name: Run tests
        run: lein test

have you tested the exact same policy locally?

if you can't use the same credentials, you can assume a role with the same policy and give it a shot locally

or add a call to aws sts get-caller-identity after you configure credentials to make sure everything is 👌:skin-tone-4:

(you may have to delete a private jar from ~/.m2/repository to make sure that lein re-pulls it down locally)

good call yeah, i've wiped the local repo in ~/.m2/repository, used the same access/secret key pair locally and have gotten a successful download

compare aws sts get-caller-identity locally and in Github to eliminate doubt

github has masked the Account number, ***, but the UserId and role part of the Arn match up

weird -- which wagon plugin specifically are you using?

[s3-wagon-private "1.3.1"] i could bump that to the latest, 1.3.3 and see what happens

do you have :no-auth true set in the :repositories section?

:repositories configured with :no-auth true as well in project.clj

jinx

😂 :thumbsup:

I'm fresh out of ideas, and we've debugged scientifically

I think bumping to 1.3.3 is in order

(I use the same exact setup but on 1.3.1 🙂 )

sigh

the s3 region is of critical importance when accessing s3 buckets - you had "eu-west-1" above - is that where the bucket is?

aws s3api get-bucket-location  --bucket YOURBUCKET

{
    "LocationConstraint": "eu-west-1"
}

😬

haven't bumped the version yet, i'll run that

not gonna work

I'm putting money on it

nah, it didn't 😂 i was clutching at straws

can you paste the config and redact the bucket?

:plugins           [[s3-wagon-private "1.3.1"]]
  :repositories [["releases" {:url           ""
                              :no-auth       true
                              :sign-releases false}]])

mine looks like that ^

yep, mine's the mirror image

s3p proto?

s3p too

another thing to debug is assume the same exact role then:

aws s3 cp --recursive  /tmp

^ that shouldn't fail inside github actions

Hi, I'm working on adding support for schema.core/defn to clojure-lsp. An example call is

(s/defn stamped-names :- StampedNames
  [names :- [s/Str]])

https://github.com/danielcompton/defn-spec#alternatives

I was thinking of ghostwheel. Not sure how widely used it is though

Reading the clojure.walk api docs, and I came across this:

Note: "walk" supports all Clojure data structures EXCEPT maps
created with sorted-map-by.  There is no (obvious) way to retrieve
the sorting function.

The Java implementation has a field storing the sorting function, so via appropriate kind of field access it is straightforward. It isn't published as a Clojure-level access function that I know of.

Do you have a link to that text you found?

https://clojure.github.io/clojure/clojure.walk-api.html

Overview section at the top

@flyboarder Not sure what that means, but in a quick REPL test, it seems that walk supports sorted-map-by

user=> (clojure.walk/postwalk-demo (sorted-map-by > 1 10 2 20))
Walked: 2
Walked: 20
Walked: [2 20]
Walked: 1
Walked: 10
Walked: [1 10]
Walked: {2 20, 1 10}
{2 20, 1 10}
user=> (clojure.walk/postwalk-demo (sorted-map-by < 1 10 2 20))
Walked: 1
Walked: 10
Walked: [1 10]
Walked: 2
Walked: 20
Walked: [2 20]
Walked: {1 10, 2 20}
{1 10, 2 20}

it might be that it doesn’t preserve the order when you transform nodes?

at least nominally postwalk-demo is transforming nodes, it just happens to return the same node it received

ostensibly you could change the value of something in place and it would not re-sort it?

when I check the type returned by the postwalk, it's a clojure.lang.PeristentTreeMap

I'll try a version that inverts sign of the keys

(ins)user=> (clojure.walk/postwalk (fn [x] (if (number? x) (- x) x)) (sorted-map-by > 1 10 2 20))
{-1 -10, -2 -20}
(ins)user=> (type *1)
clojure.lang.PersistentTreeMap

it makes sense, if this works, then postwalk should work

)user=> (let [m (sorted-map-by >)] (into m {1 :a 2 :b 3 :c}))
{3 :c, 2 :b, 1 :a}

relevant clause of clojure.walk/walk cond form:

(coll? form) (outer (into (empty form) (map inner form)))

then I have no idea what the docstring means

it might be the doc is stale on that point?

it calls into, into preserves your sort function