does clojure.test have a convenient way to do a global fixture? as in creating an account once per test suite, then cleaning up afterword? the (use-fixtures) is on a namespace level, not global, so can I use the namespace dependencies (or something similar) to ensure a catch-all file that’s loaded first?

If you’re using leiningen you can do global test setup code in a :injections vector within the project.clj and a jvm shutdown hook for cleanup

I’m sure there are other ways but that’s been one that has worked nicely for me.. I use that to start up the server exposed by my application and shut it down after the tests have run. It’s only needed when running on CI agents since if I’m working on the tests myself I’ll just do the setup in my repl

something like…

what do people use for linting clojure code? So far I’ve tried kibit, eastwood and joker, but all of these tools have lots of false postives.

joker works very well for me

integrated into emacs it's a huge productivity boost

what kind of false positives are you seeing?

just Cursive

does GraalVM work with clojure?

@jgh some things work, it has some limitations

I’m conjing onto a list and calling peek on it. This seems to work for a while in my app until it doesn’t and I get java.lang.ClassCastException: clojure.lang.LazySeq cannot be cast to clojure.lang.IPersistentStack. It is possible that conj returns a different kind of list depending on the size or should I look elsewhere for a bug?

conj does return different concrete implementations depending on size (e.g. array-map to hash-map), but not different collection types

but more importantly conj never returns a LazySeq

so you’re likely maping over the list before calling peek on it

peek works on stacks, vectors and lists are stacks but seqs aren’t

and it looks like you’re using a seq as a stack somewhere

@bronsa alright, thanks a lot for the clarification 🙏

I am using the clojure command line tool, it locks up, strace shows chasing the POMs, on a cyclic dependency (hickey-rage-face) in [org.apache.xmlgraphics/batik-transcoder "1.7"] similar to https://github.com/technomancy/leiningen/issues/1127 - does anyone know a workaround ? @alexmiller

I was burnt by this in gradle, 7 years ago, to the point of submitting a pull-request https://github.com/gradle/gradle/pull/13/commits/cebaae69dd21049a11240b320c00e79e9ae20a2b I think that time the library was apache fop ...

What's the JVM equivalent of the cljs.test/async macro? https://github.com/clojure/clojurescript-site/blob/master/content/tools/testing.adoc#async-testing

I am trying to bring up a service using compojure-api and the swagger-ui seems to have auto-sorted.... is there some way to control the sorting of the JSON responses?

can you explain what you are trying to accomplish @sidiv14.it. It sounds like you are trying to sort a hashmap which is asking more from a data structure than it is willing to give

I am trying to see if we can sort the fields in the JSON..

json objects are unordered. so there's no guaranteeing that your sorting would be honored

you could wrap whatever objects you want sorted in an array, the order would be honored then.

but you obviously would have to iterate through it to access the objects.

I suppose you could see if there's a JSON serializer that sorts keys, but why would you want/need that?

but there's no guarantee that a client would treat the serialized order as important

thanks for the responses guys... wanted to see if thats a possibility.

it's probably better to just have the client sort it however the client wants to sort it

yep I agree... sorting doesn't serve any real purpose in parsing

I am writing a simple parser/compiler for a hiccup-like DSL. atm I’m using a protocol that dispatches on the type of the first element in the vector, e.g. if it’s a keyword, symbol.

I’d like to allow a consumer to extend the parser in a way that isn’t just based on type, though. such as for a special :keyword, transform it in some way

protocols don’t really seem to work well for this. but also, multimethods kind of fall short for this too it feels like.

is there a common pattern to allow consumers of a library to extend functionality this way?

what if your multimethod first tried to match on the value, and if nothing found, then attempted to dispatch on the type?

(defmulti dsl-type-based type) (defmulti dsl-value-based identity) (defmethod dsl-value-based :default [x] (dsl-value-based x))

:thinking_face: that could work!

I was thinking of allowing a consumer to pass in a custom predicate. But maybe I don't need to be that flexible

@lilactown expand on your reason for not using multi-methods though, they're extremely flexible

@lilactown If one were to use ‘spec’ how would you do without multi-methods?

I suggest you take a look at spec, it fits perfectly for your use case and you can write a hiccup like parser using spec in 10-15 minutes.

hi everyone. where am I going wrong? 😄

(defn- resolve-video-provider
  [url]
  (condp (partial s/includes? url)
      "youtube" "YouTube"
      "vimeo" "Vimeo"))

@bravilogy condp takes a binary predicate and an expression, followed by a sequence of alternating value / result clauses.

yeah, I followed the docs and to my understanding, that was supposed to work but I guess I don’t quite get it then

I thought I would refactor this

(defn- resolve-video-provider
  [url]
  (cond
    (s/includes? url "youtube") "YouTube"
    (s/includes? url "vimeo")   "Vimeo"))

and make it more DRY

It's difficult with s/includes? because condp tests (pred test-expr expr) where expr is the initial expression and test-expr is each of the values in the clauses that follow.

you’d need to do something like,

(condp #(s/includes? %2 %1) url "youtube" "YouTube" ...)

you'd need

(condp (fn [t e] (s/includes? e t)) url
  "youtube" "YouTube"
  "vimeo" "Vimeo")

jinx 🙂

ah I see! ok now I understand

thank you so much

The docstring/reference docs are not really very clear (IMO), but clojuredocs has good examples: http://clojuredocs.org/clojure.core/condp

I too am sad that this situation doesn’t let you use partial 😉

(fn [{:keys [:keywords]}] keywords) destructure with keywords is allowed. (fn [{:prefix/keys [:keywords]}] keywords) destructure with namespaced keys keyword not allowed. intentional? bug?

@mg yeah, that’d be very clojury 😄

We have a utility function at work, called flip (taken from Haskell), so we could say

(condp (flip s/includes?) url
  "youtube" "YouTube"
  "vimeo" "Vimeo")

yeah that would be great

(defn flip
  "Like partial except you supply everything but the first argument.
  Also like Haskell's flip for single arity call."
  ([f] (fn [b a] (f a b)))
  ([f b] (fn [a] (f a b)))
  ([f b c] (fn [a] (f a b c)))
  ([f b c d & more]
   (fn [a] (apply f a b c d more))))

@dpsutton it’s (fn [{:keys [prefix/sym]} m] ...)

the binding will be to sym

@mg you can use qualified-`keys`:

user=> ( (fn [{:a/keys [b c]}] (println b c)) {:a/b 1 :a/c 2} )
1 2
nil

a new version is {:prefix/keys [without-prefix]}. however, when using a keyword there (as opposed to just a symbol) it barfs with :prefix/keys but not on :keys

oh, haven’t seen that form yet

yeah its a 1.9 goodie i believe

@seancorfield I use flip all the time in js from ramda.js library 😄 that’s a neat implementation, thank you

@dpsutton I would say it's intentional because :a/k and :k are different -- and I think it's weird that :k is allowed inside [ .. ] in the unqualified case.

understood. i can see a rationale either way. just was a bit surprising but makes sense

@seancorfield the 1-arity is kind of subtly different from the others and doesn’t quite match the docstring… I guess it’s semantically meaningless for fewer than 2 args

@michael.gaare Well, we don't have currying in Clojure -- and if the 1-arity was curried, it would be harder to use.

You couldn't use it in condp for a start...

(but your point is well-taken)

it would be interesting to see it more fully support partial, I think

hm, but how would that work

yeah, doesn’t make sense

@michael.gaare Yeah, it's a bit of an odd compromise due to the fundamental difference between Haskell and Clojure in terms of currying. Haskell doesn't need partial at all, and Haskell's flip can be fully defined in one arity.

I never quite understood why condp was written that way, actually. It seems like it’s forcing you to give up some expressivity in terms of how you construct your predicate, and I don’t see what the benefit is

I guess it’s to support the :>> syntax which I have never seen used

what would be the alternative in your view @michael.gaare?

basically the way Bravi was trying to use it

So, something like:

(condq pred
  test-1 result-1
  test-2 :>> result-2)

(since it could still pass (pred test-2) as an argument to result-2 (as a function))

Yes, something like that

Probably exactly like that

Basically I just removed all the references to expr from condp...

btw, I always wondered why we don’t have currying in clojure

currying or variable arg count, pick one

I guess there's maybe some hacky way to do both but I think it would be too clever?

not quite sure what is meant by variable arg count

Bravi: one function can take any number of arguments

ah I see, yeah

(+) => 0 (+ 1) => 1 (+ 1 1) => 2 (+ 1 1 1) => 3 etc.

you can't really curry that

yeah true. I guess languages like Elm have currying by default because the functions know exactly how many args they’re going to take

I am new to using spec. Am trying to define spec for vector of keywords. Could someone please suggest a way to do this? Sample structure

[:val1 :val2]

have you looked at the spec guide?

https://clojure.org/guides/spec#_a_game_of_cards

(s/coll-of keyword?)