Sweet! Just while I was looking for it. I was about to try to write it out, but I'm still pretty new to clojure. So i was scratching my head a little on this one (the rum hook does look like a good place to start though)

Let me know if you need any help or have any questions

The docs for hooks are here: https://github.com/borkdude/clj-kondo/blob/master/doc/hooks.md

@borkdude Just spotted this in the analysis of deps.edn:

#:worldsingles{worldsingles-test
                  #:local{:root "../worldsingles-test"}}

crap. I just pushed the button for release ;)

And it actually recommends changing the above to this, which is definitely wrong:

{worldsingles-test {:local/root "../worldsingles-test"}}/{worldsingles-test {:local/root "../worldsingles-test"}}

@lee ^ interesting issue with namespaced maps in rewrite-clj wrt analysis..

yes, that's because it doesn't look at a namespaced map like it should.

issue welcome

I'll fix it in the next release, I just hit the release button

Interesting, thanks for the ping! I should have a look at how clj-kondo is using rewrite-clj here and make sure we are properly covered for rewrite-cljc.

It could in fact be a bug in my own fork of rewrite-clj now I think about it. I'll hold off on releasing clj-kondo for today and take a look.

In any case, I’m interested to learn more. I’ll be lurking!

I'm calling sexpr on those nodes... and probably my version of sexpr for this thing doesn't work correctly... https://github.com/borkdude/clj-kondo/blob/master/src/clj_kondo/impl/linters/deps_edn.clj

hmm, it does appear to work:

user=> (n/sexpr (p/parse-string "#:foo{:a 1}"))
#:foo{:a 1}

Then maybe the bug is in the deps_edn.clj code itself.

yeah, namespaced element support is weak in rewrite-clj but my notes say that particular case is covered.

Hmm, this looks wrong:

user=> (n/sexpr (p/parse-string "#:worldsingles{worldsingles-test #:local{:root \"../worldsingles-test\"}}"))
{worldsingles-test #:local{:root "../worldsingles-test"}}

It appears to go wrong when the map has symbols:

user=> (n/sexpr (p/parse-string "#:foo{x 1}"))
{x 1}

I have never seen this case before. TIL.

But symbols shouldn’t be namespaced, only keywords, right?

user=> (keys (edn/read-string "#:foo{x 1}"))
(foo/x)

hmmm

otoh:

user=> (keys (edn/read-string "#:foo{\"x\" 1}"))
("x")

@lee does rewrite-cljc do this correctly?

So I’m looking at my handy reference https://clojure.atlassian.net/browse/CLJ-1910

> A keyword or symbol key without a namespace is read with the default namespace as its namespace.

hmm… I think I missed that “symbol key” detail…

so no, rewrite-cljc namespaced element support WIP does not handle this correctly yet!

In clj-kondo this is a small fix:

user=> (keys (n/sexpr (p/parse-string "#:foo{x 1}")))
(foo/x)

modified   parser/clj_kondo/impl/rewrite_clj/node/seq.clj
@@ -56,9 +56,11 @@
           m (first (node/sexprs children))
           nspace (name nspace-k)]
       (->> (for [[k v] m
-                 :let [k' (cond (not (keyword? k)) k
+                 :let [k' (cond (not (ident? k)) k
                                 (namespace k)      k
-                                :else (keyword nspace (name k)))]]
+                                :else (if (keyword? k)
+                                        (keyword nspace (name k))
+                                        (symbol nspace (name k))))]]

oh foo! …but am happy to realize this fact!

CLJ-1910 does deliver the gritty details, it just needs to be absorbed.

@seancorfield Pushed a fix for https://github.com/borkdude/clj-kondo/issues/1093 to master. Re-scheduled the release for coming weekend.

Nice! Thank you, sir!