@borkdude what paths are actually needed for clj-kondo? 🙂 see this: https://github.com/thheller/shadow-cljs/issues/558#event-2643788138

thheller asked: > Does clj-kondo really require the full classpath? Wouldn't the actual :source-paths suffice?

I thought clj-kondo wanted all the things, but I don't know

It depends. If you lint your libraries then clj-kondo knows about arities and other information about library functions. If it doesn't know about them, it won't emit warnings when you call them incorrectly.

okay, thanks. so then it is useful to cache your libraries as well 🙂

yes

thanks!

@borkdude i'm investigating the idea of generating TAGS / tags files using clj-kondo analysis data, but iiuc, there are few pieces of info that don't get picked up (or stored at any rate). the motivation is that, iiuc, clj-kondo understands clojure source better than the various ctags implementations so it ought to be able to index better. the two thoughts that come up are: 1) using the analysis data, revisit all of the relevant files to collect the info -OR- 2) is there some way to hook into / customize clj-kondo's process of visiting files so that the data might be collected in a customized fashion? the missing pieces for the etags version of TAGS are, iiuc, for each identifier: * some text preceding the identifier (as well as the identifier) also get stored separately from the identifier. e.g. for a namespace clojure.core, that text might be:

(ns ^{:doc "The core Clojure language."
       :author "Rich Hickey"}
  clojure.core

(defn start

I'll take a look later, I'm on a plane right now and can't visit many links. Please document what is required to the analysis output in a GH issue

tnx, will do