Clojurians Log v2
Clojure programming
Channels
Apps
clj-kondo 2026-07-28
Not sure if this is the newly enhanced type linter, but I'm now getting a Redundant int coercion: expression already has type int on
(defn foo [bar]
(when (integer? bar)
(case (int bar) ;; <<<< here
0 "Low"
1 "Medium"
2 "High"
nil)))case perspective, but I find the lint message somewhat confusing, as bar could be Short, Byte, BigInteger etc. and still pass the integer? check
borkdude 2026-07-28T15:58:06.629019Z
If that's the case, then our integer? specs are wrong. Issue welcome
borkdude 2026-07-28T15:58:30.716659Z
Preferably with a repro of runtime code going wrong when changing it
by that you mean something like "if I remove the coerce flagged redundant, the behavior of my system changes in a bad way"?
borkdude 2026-07-28T16:00:35.322489Z
yes
Kinda like:
(let [x 5]
(when (integer? x)
(= Integer (type (int x)))))
; true
(let [x 5]
(when (integer? x)
(= Integer (type x))))
; falseborkdude 2026-07-28T16:05:02.859599Z
yeah but more realistic? :)
I might have a hard time coming up with something non-convoluted. There are only a few places in code I know that has a call to (int) and the reasons are always something to do with how data is modeled or strict type checks, so the cast is required.
However, I still do not agree with the linter message in this case, because (integer? x) doesn't mean x is an int
borkdude 2026-07-28T16:12:14.310939Z
I agree. I guess we could ask a robot to come up with a realistic case but we have to change the spec anyway, I agree
clj-kondo --lint - <<< "(fn [x] (when (integer? x) (int x)))"
<stdin>:1:28: info: Redundant int coercion: expression already has type int
linting took 122ms, errors: 0, warnings: 0
clj-kondo --lint - <<< "(fn [x] (when (integer? x) (long x)))"
linting took 60ms, errors: 0, warnings: 0longI'll put these into a ticket
thank you
What would be the best way to get warnings about redefinition of clojure spec keywords? Edge cases seem to be ::kw in two different namespaces being different keywords, and :some-alias/kw in two different namespaces being the same keyword.
Can a hook properly resolve a keyword's namespace?