want to help test a new feature? warnings about missing requires: linux https://7518-176829714-gh.circle-artifacts.com/0/release/clj-kondo-2019.11.24-SNAPSHOT-linux-amd64.zip mac: https://7517-176829714-gh.circle-artifacts.com/0/release/clj-kondo-2019.11.24-SNAPSHOT-macos-amd64.zip branch: missing-require example:

The name of the linter and the message are subject to change, feel free to post suggestions at https://github.com/borkdude/clj-kondo/issues/339

hello, is there a way to exclude certain namespaces (with a regex) from being checked for :unresolved-symbol?

https://github.com/borkdude/clj-kondo/blob/master/doc/config.md I don't think so (yet). Do you have an example of how this would be useful?

@borkdude I use hugsql in namespaces that all end in .data and I’d rather not declare all my hugsql-generated functions

so the vars are foo.data/bar?

correct

I guess the regex could be "\\.data/.*" in that linter. PR welcome I think

ok, thanks!

by the way, does this look ok to you? :unresolved-symbol {:exclude [(secretary.core [defroute])]}

could you first make an issue about it with a proposed solution and then a corresponding PR?

I think that looks ok

isn't it working?

it seems to crash clj-kondo

Unexpected error. Please report an issue.
java.lang.IllegalArgumentException: no conversion to symbol
	at clojure.core$symbol.invokeStatic(core.clj:596)
	at clojure.core$symbol.invoke(core.clj:589)
	at clj_kondo.impl.config$fn__4141$delayed_cfg__4142$fn__4144.invoke(config.clj:187)
	at clojure.lang.ArrayChunk.reduce(ArrayChunk.java:63)
	at clojure.core.protocols$fn__8139.invokeStatic(protocols.clj:136)
	at clojure.core.protocols$fn__8139.invoke(protocols.clj:124)
	at clojure.core.protocols$fn__8099$G__8094__8108.invoke(protocols.clj:19)
	at clojure.core.protocols$seq_reduce.invokeStatic(protocols.clj:31)
	at clojure.core.protocols$fn__8131.invokeStatic(protocols.clj:75)
	at clojure.core.protocols$fn__8131.invoke(protocols.clj:75)
	at clojure.core.protocols$fn__8073$G__8068__8086.invoke(protocols.clj:13)
	at clojure.core$reduce.invokeStatic(core.clj:6828)
	at clojure.core$reduce.invoke(core.clj:6810)
	at clj_kondo.impl.config$fn__4141$delayed_cfg__4142.invoke(config.clj:186)
	at clojure.lang.AFn.applyToHelper(AFn.java:154)
	at clojure.lang.AFn.applyTo(AFn.java:144)
	at clojure.core$apply.invokeStatic(core.clj:665)
	at clojure.core$memoize$fn__6862.doInvoke(core.clj:6353)
	at clojure.lang.RestFn.invoke(RestFn.java:408)
	at clj_kondo.impl.config$fn__4141$fn__4150.invoke(config.clj:195)
	at clj_kondo.impl.namespace$reg_unresolved_symbol_BANG_.invokeStatic(namespace.clj:162)
	at clj_kondo.impl.namespace$reg_unresolved_symbol_BANG_.invoke(namespace.clj:157)
	at clj_kondo.impl.linters$lint_var_usage$iter__5082__5090$fn__5091$iter__5084__5092$fn__5093.invoke(linters.clj:239)
	at clojure.lang.LazySeq.sval(LazySeq.java:42)
	at clojure.lang.LazySeq.seq(LazySeq.java:51)
	at clojure.lang.RT.seq(RT.java:531)
	at clojure.core$seq__5387.invokeStatic(core.clj:137)
	at clojure.core$seq__5387.invoke(core.clj:137)
	at clj_kondo.impl.linters$lint_var_usage$iter__5082__5090$fn__5091.invoke(linters.clj:201)

@stathissideris ah gotcha:

$ clj -A:clj-kondo --config '{:linters {:unresolved-symbol {:exclude [(secretary.core/defroute)]}}}' --lint - <<< "(require '[secretary.core]) (secretary.core/defroute foo)"
linting took 61ms, errors: 0, warnings: 0

the vector is for excluding specific symbols within a call

I'm thinking about deprecating this syntax maybe, since it's a bit odd maybe

oooh, sorry, I misread the manual

no problem. there is an issue for a config validator. once that's there, we should no longer have to bother with this kind of problem

so any pointers on where I could start with implementing the new exclusion for the :unresolved-symbol linter?

@stathissideris yes: here is the related config code: https://github.com/borkdude/clj-kondo/blob/de8add6672d5c795a0012e6ba9043f3bfe4b9c4b/src/clj_kondo/impl/config.clj#L179 and here is a config that already uses regexes, as an example: https://github.com/borkdude/clj-kondo/blob/de8add6672d5c795a0012e6ba9043f3bfe4b9c4b/src/clj_kondo/impl/config.clj#L161

thanks, I’ll have a look!

if you have a suggestion for a map-based configuration format instead of the vector with lists or symbols, feel free to come up with one as well

because one problem now is that this config isn't merged with built-in ones (e.g. clojure.test/is)

one thing that was a bit unexpected in the config is that each linter seems to expect different (but kinda similar) configs

The related issue for that one is: https://github.com/borkdude/clj-kondo/issues/608

Maybe two separate PRs would be nice. Optional of course.

I think it's good to discuss some alternative formats first before implementing it. But this is another feature really, so let's just focus on the regex first 🙂

It might be slightly related though, since the regex feature may affect the "ignore in call" exclusion or the symbol itself

@borkdude thank you so much by the way, I’m finding bugs with clj-kondo

I had a semi-working branch for that exact feature, could try and polish it up and submit a PR over the weekend!

My use case was wanting to support Meander macros and ignore all unresolved-symbol warnings for symbols ending with ! and ?

one odd quirk I found with the regex configurations was how they don't accept actual #"" regex literals and require string literals with double-escaped backslashes

@qythium that is because EDN doesn't support regexes

but I do actually support it with edamame (which is a configurable EDN parser), but clj-kondo predates it

Since US is on holiday, I finally have space to think about this clj-kondo current directory thing.

I think I'm concluding that ideally, I want to lint stdin, but supply the filename somehow so that the clj-kondo can find the config directory.

eslint (the first one I checked) does this with --stdin and --stdin-filename options.

@eraserhd - is already there for linting stdin, --stdin-filename would that only be used for finding the config dir? then we could also have --cwd which overrides the current working directory?

--stdin-filename would also provide a default for --lang.

true

alright, seems good then

Oh, to disclose: I'm still pondering. This proposal seems the most "correct", although it requires a bit of rework on the editor part.

Some other linters don't support linting stdin. cppcheck being one of them.

@marc-omorain’s plugin also doesn't use stdin, it uses the raw file + filename, so there it still make sense what we're currently doing

flycheck (emacs) sets the cwd to the dir of the file being linted

huh. That's actually an easy fix. That suggests all linters can find their config in a parent directory?

yes

ehm, I don't know for all linters

but I'm happy it works that way

But anything that works with flycheck, which is most.

because I anticipated the same problem for a multi-project git repo, then I checked, and I was happy that it worked in my case (having two .clj-kondo dirs in the same repo)

@eraserhd aren't you talking from the perspective of kakoune?

Well, I'm trying to have a wider perspective and do the "most natural" thing.

adopting the flycheck behavior requires no changes to clj-kondo

Hmm, well changing to the linted file's directory is easy, but it still writes the buffer to a temporary directory passed by absolute path.

in the intellij setup it's also working like this: the working dir is the file's dir: https://github.com/borkdude/clj-kondo/blob/master/doc/editor-integration.md#intellij-idea

Which means the PR I made would not be useful in this case.

@eraserhd you would of course set the cwd to the dir of the linted file, not the copied tmp file

Yeah.

but yeah, using the path of the copied tmp file for resolving the config dir is a bit pointless

@eraserhd https://github.com/flycheck/flycheck/blob/269237f6529a4ad7db5bbbc5a40c1f038accf3cd/flycheck.el#L4990

Well, adding a cd to the editor script definitely works.

I'm looking at the language-specific checkers in the flycheck org. If very few of them override that working-directory thing, then I'll just make that PR to Kakoune.

OK, there are a few, by global GitHub search - elsa, elixir, scons, android, credo.

Actually, that seems like a small list, except elixir is important.

and why does elixir do that?

I don't know, but it appears to try to find the "mix directory", which is the project root.