Clojure Q&A - Recent activity in Metadata

Commented: Is it valid for metadata to have metadata?

Wed, 18 Feb 2026 22:01:59 +0000

Thanks for logging this – I've left a comment on that ticket with a patch and detailed the approach/patch summary there too. I should have a signed CA on file since I've contributed to ClojureScript and tools.reader in the past.

Metadata on quoted forms

Sun, 06 Jul 2025 15:29:14 +0000

Reader metadata is attached to IObj forms at read time, but if the form is quoted, then the quote form gets the metadata and the quoted form doesn't get the metadata. This means that to attach metadata to forms, you need to write the slightly awkward quote -> metadata -> form instead of the (imo) more natural metadata-> quote -> form.

user=> (meta ^:foo '(1 2 3))
{:line 1 :column 1}
user=> (meta '^:foo (1 2 3))
{:lime 1 :column 1 :foo true}

I know that this is because ' is implemented as a "WrappingReader" and so the former is read as ^:foo (quote (1 2 3)) but that feels like an implementation detail leaking through. I think it would be helpful/more consistent if quote passed any metadata to IObjs.

Answered: Bind ns and require before evaluating namespace metadata

Tue, 20 May 2025 19:51:37 +0000

Regardless of whether this is desirable or not, I don't believe this is currently feasible. The reader reads in each form one at a time, converting auto-namespaced keywords and symbols to fully-qualified keywords and symbols, and then evaluates the whole form. The target namespace alias (tape.mvc) has to exist before the form ::mvc/interceptors can even be read, which can't happen when the entire (ns) form will be read together and the aliases don't exist yet.

Commented: File and location data on namespaces

Tue, 20 May 2025 19:16:47 +0000

Patch submitted.

Answered: A function with metadata is slower than it could be

Wed, 05 Mar 2025 18:51:01 +0000

Wouldn't it make more sense to focus on making applyTo faster?

Closed: strange behavior with `range` and `with-meta`

Wed, 05 Jun 2024 18:57:51 +0000

from slack: https://clojurians.slack.com/archives/C03S1KBA2/p1662437976861689

(range 0 5 1.0) ;=> (0 1.0 2.0 3.0 4.0)
(with-meta (range 0 5 1.0) {}) ;=> (5)

user=> (def x (range 0 5 1.0))
#'user/x
user=> x
(0 1.0 2.0 3.0 4.0)
user=> (with-meta x {})
(5 1.0 2.0 3.0 4.0)

I noticed that it looks like there might be some out-of-order argument called

the with-meta call in Range.java:

public Obj withMeta(IPersistentMap meta){
	if(meta == _meta)
		return this;
	return new Range(meta, end, start, step, boundsCheck, _chunk, _chunkNext);
}

and that constructor has the end and start swapped:

private Range(IPersistentMap meta, Object start, Object end, Object step, BoundsCheck boundsCheck, IChunk chunk, ISeq chunkNext){

Answer selected: Add type hints to `ex-info`

Fri, 01 Dec 2023 20:17:15 +0000

Jira ticket https://clojure.atlassian.net/browse/CLJ-2815 has been created to track the issue.

Answer selected: what does ^:once meta applied for anon function mean?

Thu, 06 Jul 2023 12:05:14 +0000

it promises the compielr that the lambda will only be invoked once, and thus any variables that are captured in the lexical context can be garbage collected.

This is potentially unsafe if used incorrectly, but in some contexts it's necessary to avoid unnecessarily holding onto unused memory

This example shows how it works:

user=> (let [x (Object.)] (def x (^:once fn* [] x)))
#'user/x
user=> (x)
#object[java.lang.Object 0x2189e7a7 "java.lang.Object@2189e7a7"]
user=> (x)
nil

Commented: which two objects share the same sequence?

Tue, 13 Jul 2021 13:47:10 +0000

Clojure web site issues/prs should be filed at https://github.com/clojure/clojure-site/issues

Answered: Enhance multimethods metadata

Wed, 26 Jun 2019 12:00:00 +0000

Reference: https://clojure.atlassian.net/browse/CLJ-1522 (reported by bozhidar)