I think it would be possible to infer the var just by looking at the namespace / alias and name of the symbol, just for navigation, perhaps.

And perhaps only allow navigation (or even completion?) for qualified symbols

yes, as a final user that sounds good. Not that easy I think to support that, as we don't have any analysis of symbols 'clojure.test/is for example, so we would need to do hacky things like if no analysis is present try to infer that with the logic you mentioned

Why is that hacky? Btw, Cursive supports this

@U04V15CAJ thanks for starting this conversation! @ericdallo some context if you’re interested: in Clerk a render-fn is a quoted form defined in Clojure that’s then transmitted to the browser. In cursive you can use jump to definition for a symbol like https://github.com/nextjournal/clerk/blob/edd0c2fc22bd0207cc8d3762f00f239ce63ee87f/src/nextjournal/clerk/viewer.cljc#L733 and I’d love it if lsp users could also do that

one level up would be to resolve ns-aliases for a quoted form so when you see a quoted symbol like v/html to first try to resolve it as-is, and if there’s an ns-alias defined for this, also try with that resolved

I agreed that for user this makes sense, I'm just saying that it'd be something hacky to support on clojure-lsp as ATM we completely rely on analysis and we are saying we would for cases where we don't find a analysis we would introduce a new logic to find a definition

https://github.com/nextjournal/clerk/pull/276 is even more context if you’re interested

Yeah, that makes sense, we would need to have some kind of resolver for symbols with fallbacking to use rewrite-clj on that code and parse because ATM we receive on server: URI line col

so we would need to parse the code on that uri line and col with rewrite-clj, check that is a symbol and try to find analysis for that ns or alias

ah so currently this is fully handled by clj-kondo and you don’t even see the symbol?

yes, ATM we find any element from kondo analysis that are inside the range checking name locations

https://github.com/clojure-lsp/clojure-lsp/blob/master/lib/src/clojure_lsp/queries.clj#L501-L505

alright, I have to run now but happy to take a look and see if I can implement this next week if you agree having this makes sense in general

Sure, feel free to create a issue where we can discuss the implementation better, the feature support makes sense, but is not clear to me if we can support it without lots of hacky code

This would be a really nice feature 👍 I remember thinking about opening a ticket about this when I was playing around with lazy loading functions with requiring-resolve

that's a good example!

I think just looking at the aliases / used namespaces and using the full namespace name as fallback should be sufficient to navigate to the var (or to do completions)

The dev jar is a one time only mostly, you build and the connect to the running nrepl port, there is even a lsp-clojure-nrepl-connect command that does automatically for you :)

Then you can change how clojure-lsp works and build your feature without restarting or re-building anything

@ericdallo That works, but why even go through the jar? You can just start a -main function in the script directly?

Yeah, works too, there is run deps alias which runs the main, it's just that was never a issue, as bb debug-cli takes a few seconds and usually you don't need to re-build that anymore

Merged the first @U04V15CAJ's PR 🎉 it should be available a nightly build soon on #C032YH7P0R2 @mkvlr in case you want to give it a try

It's not the first though ;) https://github.com/clojure-lsp/clojure-lsp/commits?author=borkdude

Oh, I meant related to this thread 😅 but yeah, you already helped a lot in the past!

oh you meant that 😅

@ericdallo @U04V15CAJ awesome, thanks a ton to both of you!