Clojurians Log v2

maybe cljdoc can leverage this data somehow: https://github.com/borkdude/clj-kondo/tree/dump-analysis/analysis

I expect that cljdoc will eventually switch to using the most awesome clj-kondo. cljdoc currently uses a fork of codox to get data about clojure code. Codox uses clojure and cljs analyzers which means the lies told by such tools as import-vars are incorporated into the data story. I think we’d need https://github.com/borkdude/clj-kondo/issues/237 before switching over.

yeah, I'll add that to the next release

that is exciting news!

I'm sure there will be more data missing, so as a transition period you can combine data from clj-kondo and codox maybe

One thing to keep in mind with this stuff is that Codox expects the runtime and clj-kondo inspects the source code (as far as I understand). The latter approach will always have some problems with extracting information from libraries like amazonica

that's also what I thought. so maybe you could do something like a left join on codox data and enrich is with what clj-kondo found

how does codox do this for clojurescript, it evals stuff and then collects stuff?

to make sure I understand: are you guys talking about using clojure/cljs analyzers (codox) vs parsing source code (clj-kondo)?

yes

thanks, I made the assumption that clj-kondo would eventually use analyzers as well - to support import-vars magic

clj-kondo only statically analyzes the source code

I'm not sure what you mean by "analyzer"

import-vars magic is possible to analyze statically

so codox uses on cljs side: https://github.com/cljdoc/codox/blob/cljs-proper/codox/src/codox/reader/clojurescript.clj#L5 and on clj side: https://github.com/cljdoc/codox/blob/cljs-proper/codox/src/codox/reader/clojure.clj#L7

(ya, “analyzer” was wrong term for clj)

codox does not make use of rewrite-clj

I didn't assume that 🙂 it's clever to use clojurescript/analyzer, that had not even occurred to me 😛

It is an alternate approach to consider. It will tell you the truth (including lies).

but I guess that's also doing some kind of static analysis

it's an approach, but I don't know if that will work with GraalVM and how it will perform. also, it will only work for CLJS

right… all variables in the equation.

and the Clojure part does need the dependency loaded in the JVM I guess: https://github.com/cljdoc/codox/blob/cljs-proper/codox/src/codox/reader/clojure.clj#L29

i.e. it evals code

yes

I guess it is a bit of a different beast than clj-kondo

codox only uses tools.namespace to find all the namespaces and then loads them in a JVM: https://github.com/cljdoc/codox/blob/cljs-proper/codox/src/codox/reader/clojure.clj#L95

so if the code doesn't run, you won't find any data, which defeats the purpose of a linter

excellent point sir

so although clj-kondo might play a role in cljdoc, it won’t likely replace the features we are using from codox

you were bringing up the idea of breaking changes if you remove a var or an arity. this is the thing you could use clj-kondo for

yes

so you could do a "left join" on the vars you already have and then enrich that with the data from clj-kondo and keep that in a database per version

but there’s the question of import-vars on cljs … we might use data returned from codox for that.

right

and then compare a new version to an old version and report any breakage

but maybe you can already use codox for this, I don't know. it's just that clj-kondo already has this information, I'd be happy to share. but if you don't want to use clj-kondo for anything other than linting, that's ok too

I’ve fallen in love with clj-kondo, so I am biased

but we’ll try to use it where it makes sense

you can also display a namespace graph, or anything other fancy you come up with based on data 😛 https://github.com/borkdude/clj-kondo/tree/dump-analysis/analysis#namespace-graph

I gotta go now

later, thanks for chats @borkdude!