[feature request] add feature to find cross references for a function or var

[stardiviner]

I hope CIDER can add a feature to find all cross references for a function or var.
For examples, list out all places which used a function or var.
Common Lisp's SLIME has this function called slime-who-references. And Emacs Lisp has this function called xref-find-references [M-?].

[grammati]

I think clj-refactor's find-usages is what you are looking for: https://github.com/clojure-emacs/clj-refactor.el/wiki/cljr-find-usages

[stardiviner]

@grammati Thanks.

[timvisher]

Actually the SLIME feature finds actual calls to the function at point rather than just textual occurrences. If I'm reading the clj-refactor feature correctly, it's the later not the former.

Waaay back when cider was still hacked-SLIME, it did have this feature. I wish we had it still too. :)

[stardiviner]

@timvisher glad to hear this. I will re-open this feature request.

[bbatsov]

I want us to have something similar at CIDER itself eventually.

Waaay back when cider was still hacked-SLIME, it did have this feature. I wish we had it still too. :)

This was just a textual search, which is rather easy to add back. Doing something different that's not relying on an AST is a bit harder, but is definitely possible.

[timvisher]

@bbatsov Are you sure it was just a textual search? Did it at least try to filter for things that looked like calls? My memory (which is foggy) was that it really did search for references to the var or something.

[bbatsov]

This is the old implementation - https://github.com/technomancy/swank-clojure/blob/master/src/swank/commands/xref.clj

Everyone's welcome to port it to an nREPL middleware.

[bbatsov]

@sanjayl @cskksc Maybe one of you would be interested in porting this from SLIME? :-)

[sanjayl]

Will take a look at it

[ckoparkar]

@sanjayl Awesome! Let me know if I can help with anything.

[johngit22]

Thanks !

On Sun, Sep 4, 2016 at 6:52 AM, Chaitanya Koparkar <[email protected]

wrote:

@sanjayl Awesome! Let me know if I can help with anything.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#1840 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/ADFz2XEZMczlC33SSuV1vmrP7rqf2Vi0ks5qmqL0gaJpZM4Jzi0S
.

John Boyd

[bbatsov]

@sanjayl Thanks for tackling this!

[bbatsov]

@sanjayl any news on this?

[sanjayl]

Sorry, hit a wall and ran out of steam. I have the next few days off so I'll see what I can do on the clojure side of things, doubt I'll have time to work on the UI half.

[bbatsov]

Sorry, hit a wall and ran out of steam. I have the next few days off so I'll see what I can do on the clojure side of things, doubt I'll have time to work on the UI half.

One step at a time. I believe you'll be able to pretty much copy/paste the UI from SLIME or just leverage the new xref functionality from recent Emacsen.

[bkruczyk]

@bbatsov Is this still an issue? From what I can see nrepl has message handlar for finding a symbol.

[bbatsov]

@bkruczyk That's certainly not the case. Such functionality doesn't really belong in nREPL itself and will never be part of nREPL I guess. The only nREPL middleware that provides something like this is refactor-nrepl, but I don't like the approach taken there and I'd rather go with something that just relies on the state of the running application (similar to what swank-clojure did in the past).

[bkruczyk]

@bbatsov Just like you said, I was thinking of refactor-nrepl middleware (i.e. this).

What's wrong with it? Isn't it better than textual search from swank-clojure?

[expez]

@bkruczyk it has to evaluate all the code to understand it.

[ailisp]

@bkruczyk swank also need to load all common lisp code in slime repl before cross reference. But that's not a problem for cl, because there's a different workflow between cl and clj. CL programmers have habit to load all project files from slime-repl using (ql:quickload ...) or (asdf:load-system ...) before doing any development). In clojure, nrepl was launched from from shell and didn't load the whole project. Unless you load the whole system in repl, the cross reference is like doing so in python or ruby, which is quite inaccurate.

[stardiviner]

If so, then is it possible to provide two solution. use textual search by default. (As @ailisp upper said)
And provide this load a whole project solution as an optional solution for users who want it like me.
CIDER can provide two different keybinding for them.

[bbatsov]

@bkruczyk swank also need to load all common lisp code in slime repl before cross reference. But that's not a problem for cl, because there's a different workflow between cl and clj. CL programmers have habit to load all project files from slime-repl using (ql:quickload ...) or (asdf:load-system ...) before doing any development). In clojure, nrepl was launched from from shell and didn't load the whole project. Unless you load the whole system in repl, the cross reference is like doing so in python or ruby, which is quite inaccurate.

@ailisp We already have functionality to load the entire project if necessary, so that's not a big deal. I don't think swank-clojure was doing this for Clojure, and people were kind of OK with the feature, though, so I don't think that's a big deal.

@stardiviner The textual search is pretty much the same approach. It's not really related to loading everything or not. It's just a way to try to find call site for some functions.

[ailisp]

@stardiviner That's really a good idea. I prefer to use a same key binding to do semantic search first and if nothing found fall over to text search though. But this is trivial to do in init file after implement both options.

[gerdint]

1+ for having this!

Geiser implements this but I do believe it relies on the underlying Scheme for doing it accurately.

[harold]

Just discovered this today, and it's great! Thanks @bbatsov 🙇