https://docs.oracle.com/javase/9/docs/api/java/time/LocalDate.html#datesUntil-java.time.LocalDate- this is an instance method. So you need (.datesUntil <instance> <end-exclusive>)

(.datesUntil (LocalDate/now) (LocalDate/parse "2022-12-25")) -> LocalDate.now().datesUntil(then)

(iterator-seq (.iterator (.datesUntil (LocalDate/now) (LocalDate/parse "2022-12-25"))))

(map val vec-of-mapentries)

val works on map-entries, vals works on maps. (map val ... is not very common, and it'll break when you aren't using map-entries

(map val [{:x 1} {:x 2}])
=> ExceptionInfo
Error printing return value (ClassCastException) at null (REPL:1).

i was incorrect in my assumption of how map would handle single entry maps

(mapcat vals ...) is better

Yes, I thought map would convert each map to a set of map entries as well.

it does if you do (map val {})

which is very similar to update-vals or update-keys

Thanks @U0LAJQLQ1.

Ah, of course. It doesn't matter how many are in each one. Thanks!

if the maps all have a single key (not necessarily specified as such in the original question) you could (map :that-key coll)

moving to thread...

mvn-install is for installing a jar you have into the Maven repository, which you usually don't need to do (usually by referring to it as a dep it is automatically downloaded)

can you clarify the full command you ran and what you're trying to do?

I have a deps.edn with these contents:

{:deps
{org.clojure/core.async {:mvn/version "1.5.648"}}
{org.clojure/data.json {:mvn/version "2.4.0"}}}

which I did but I get the above error

I could've sworn that previously I used clj -X:deps mvn-install to add core.async to my classpath and make it available in the REPL but I wouldn't stake my life on it

You should be able to just run clj and it will download those deps before starting the repl

correct, you don't need to call mvn-install here

I don't see any downloads happening when I start the REPL and I get ClassNotFoundException when I try to require the deps

you may already have those downloads in your .m2 so they don't need to be downloaded

they would be at ~/.m2/repository/org/clojure/core.async/1.5.648/core.async-1.5.648.jar for example

you can check your classpath with clj -Spath

what's the actual repl invocation and response you're seeing?

just plain

clj
Clojure 1.11.1
user=>

and it does appear that data.json is on my classpath as you said

where's the error?

and the thing that causes the error

user=> (require [org.clojure/data.json :as json])
Syntax error (ClassNotFoundException) compiling at (REPL:1:1).
org.clojure

just in general, a good recipe for asking for help to make it easy for others to help is: I want to do X. I did Y. I got Z, which is wrong/confusing/etc

the namespaces in an artifact are different than the name of artifact itself (which is confusing!)

and they don't even necessarily have any relationship

here you'd want:

(require '[clojure.data.json :as json])

how would you know that, you might ask... and the answer is, only by reading the documentation

duly noted, on all counts. I will do that more studiously in the future. Thanks a lot for your help

for example, the git readme has an example - https://github.com/clojure/data.json

also note that I quoted the vector passed to require - this is data passed to a function so to avoid evaluating the symbols inside, it needs to be quoted (which means "read, but don't evaluate")

is there a reason why there is a divergence between artifact naming conventions? Like something to do with the reader syntax? or is it just that there is no hard rule and so things just went that way

jars / artifacts are just containers to hold code

in the early days of Java, there was never any established linkage between how to name / reference libraries and the code inside

there was no package/dependency manager, and it was the wild west

eventually Maven established some conventions, but no one ever created any hard relationship or self describability for jars (beyond just looking at what's in them)

Clojure is really a source-based language with a convention for lookup, but again, basically relied on Java's ecosystem for publishing and consuming artifacts containing code

so, missed opportunities at multiple points in the timeline

interesting history :thinking_face: maybe not the choices I would make but I know very little indeed

I think figuring out library deployment and consumption should probably be on the list of things a language does very early in it's life, otherwise you end up stuck with ad hoc decisions

I think there is more Clojure can do here though and it's something we've been kicking around for many years

the only other language I use with any degree of regularity is Rust since I'm basically a hobbyist and I am very much accustomed to cargo allowing me to not really think about anything re: dependency management, it's very straightforward

https://clojurians.slack.com/archives/C03S1KBA2/p1657731629991189

Vestigial, eh? 🙂

Interesting that there's a linter in kondo for it. I first spotted the with-colon variety in the kondo code. I suppose if there's a linter for it and it's somewhat of an artifact, the without-colon variety should be preferred. But I think using the colon would offer some benefit when searching the code base for all instances of a keyword.

this is the first time i've seen this form of :keys

the latter is preferred

the keyword form was added to support use of autoreferred keywords (`::k`, ::n/k) but those are now better supported via ::keys or ::n/keys

Thanks for the confirmation.

This is a good one from the illustrious @U04V70XH6: https://www.youtube.com/watch?v=gIoadGfm5T8

just some weird stuff in clojure. sometimes names are well designed, sometimes not.

some-fn returns the first result of applying each function in turn -- so it will return "logical true"; every-pred returns true if everything returns "logical true".

But, yeah, some of this is historical quirks. Like we have every? and some -- but any? means something very different.

@U0NCTKEV8 (moving to thread)--is there a good way to narrow down which library is at fault, and/or figure out how to reset it? I suppose if all else fails, I can try analyzing the dependencies one by one, too.

the stack trace will tell you what code is calling loadLibrary

Sounds good, thanks for the help! I will dig a little deeper and see if I can find the problem.

@U03DKTAF52P I'm definitely curious to hear what you find and which libraries seem to be doing this. @U0NCTKEV8 Any idea how to clean things up once they get into a bad state like that, or is it just a case of restarting the REPL and letting it finish loading?

I am not sure, it will depend on what is actually doing the loading and what it is trying to load

This has happened a couple times across a couple projects for me, now--once on a Shadow-CLJS React Native/Expo project, and again as I'm currently working through LambdaIsland's Buddy tutorial (so it's possible that it's a my-machine-problem). Both have had a large stack of their own dependencies. Unfortunately, the issue continues across multiple restarted REPL sessions, so I'm wondering if it's something that's being saved into some hidden configuration. </Speculation>

I have this really wild guess that in this case it might be due the library having native code for osx x86, and trying to load that into the process of a apple silicon native jvm, but that is way out over my skis (not a lot of evidence, and some of the evidence may contradict that)

@U03DKTAF52P Is yours an M1 Mac or an x86 Mac?

M1 on my end. However, I would note that the same REPL load and functional call worked before I interrupted the CIDER REPL load, so it would be weird to me if it was an architecture issue, but that is also far out of my wheelhouse.

if can get the error to happen when using a repl start with clj, that is likely the easiest path to getting a good stacktrace

(FWIW, I always start my REPL manually and just connect my editor to it -- because my REPLs tend to be longer-lived than my editor sessions)

Wow, bizarre--the first library I can see in here appears to be Figwheel?

(if there's a prettier way to print this, feel free to let me know 😅 )

It's better to share actual text -- you can format

code with triple backticks

hawk is, if I recall, a file watching library, and I bet that uses native code for monitoring the fileystem for changes

Based on that stacktrace, it looks like Hawk is the probelm

so that makes sense

Yeah, I bet it either doesn't have an M1 version or you need a more up to date version that the project currently has (via Figwheel Sidecar)

Ah. I'll see if bumping the Figwheel dependency up a few versions helps, but that pretty much answers my short-term question. Thanks so much for your help!!

well the barbarysoftware thing that hawk uses on osx seems unlikely to get an update https://github.com/gjoseph/BarbaryWatchService

Related https://github.com/bhauman/figwheel-main/issues/253

For future reference--from a fresh install on another M1 Mac, everything seems to work out-of-the-box; both Hawk and Figwheel are behaving correctly. Not sure what I could've done to my existing setup to have borked it (or exactly what I can do to fix it).

Did updating the version(s) of Figwheel and Sidecar help at all?

Unfortunately not 😕...

have you compared the jvm version on the two macs?

18.0.2 on the broken one, and 18.0.1 on the working one. I'll upgrade the working one and see if it breaks.

Confirmed that it works OOTB on 18.0.2 as well. Something must be really messed up on my other machine.

besides the version number, are they the same? like maybe the working one is using an x86 jvm above the emulation layer and the broken one is using a native apple silicon jvm

java -version outputs the following for both:

openjdk version "18.0.2" 2022-07-19
OpenJDK Runtime Environment Homebrew (build 18.0.2+0)
OpenJDK 64-Bit Server VM Homebrew (build 18.0.2+0, mixed mode, sharing)

not super useful I guess, but if you are getting it from homebrew it seems like they should be the same

(no native architecture stuff in the -version output of java)

Hmm. I'll keep digging--at the end of the day, maybe I just need to factory reset my whole machine. Really not sure how this happened.

iterate and nth sounds like what you're looking for?

((comp f f f f f f f) foo) if you want to rewrite your first example

user=> (take 4 (iterate inc 0))
(0 1 2 3)
user=> (nth (iterate inc 0) 2)
2
user=> (nth (iterate inc 42) 4)
46
user=>

Thanks sean, that's better!

BTW, you can do (repeat n rotate...) to get n of that function.

(so you don't need take in what you're doing now)

interesting use of repeat

user=> (doc repeat)
-------------------------
clojure.core/repeat
([x] [n x])
  Returns a lazy (infinite!, or length n if supplied) sequence of xs.
nil
user=>

It'd be helpful to comment in this thread. 🙏

me! is there some kind of registration for ds4clj?

Wonderful 🙏 There is no registration yet.