#{a b} set initialization doesn't throw if there are duplicates (in clj it does) about babashka HOT 6 CLOSED

Since CLJS accepts it and I haven't heard anyone complain about that, I think it's better to just leave it since then the SCI code can remain the same for both implementations.

from babashka.

ClojureScript also accepts (let [a 1, b 1] #{a b}). I would consider this undefined behavior or an incidental implementation detail.

from babashka.

Ah wait:

user=> (defn foo [a b] #{a b})
#'user/foo
user=> (foo 1 1)
Execution error (IllegalArgumentException) at user/foo (REPL:1).
Duplicate key: 1

from babashka.

It seems JVM Clojure has a createWithCheck mechanism which is used for literals:

https://github.com/search?q=repo%3Aclojure%2Fclojure%20createWithCheck&type=code

from babashka.

Still I think the question should be asked: why is this important, exactly? I never had a situation myself where leaving the check off would have made a big difference.

from babashka.

I agree it's probably "incidental implementation". I simply just realized that there's this strange difference between clj and babashka and though it might be nice to at be aware know about it, that's all.

Personally I never even realized that set initialization using #{} form actually throws in clojure. Instinctively, I'd expect it to behave similar to into #{}, i.e. duplicities get silently ignored. Because, well.. why not.

When I think about it, the fact that the form using values #{1 2 3 1} throws might make sense. If the expression is such a simple static literal, then the compile error might make come handy (in a way it can be thought as a kind of late compile-time "linter")

I agree I wouldn't "fix" this discrepancy in behavior as there's probably no "agreed contract" regarding how all this should "officially" behave. Let's close this issue.

from babashka.