@ericdallo I've raised: https://github.com/clojure-lsp/clojure-lsp/issues/396 Weird if I'm the only one experiencing this lol Thanks

FYI @UR37CBF8D https://github.com/clojure-lsp/clojure-lsp/issues/399#issuecomment-814922383

Oh, interesting

If you could test if that branch fix your issue, it'd help

you can just checkout it and run make debug-bin to generate a local clojure-lsp

Sweet, will do. Thanks

@ericdallo After using LSP for a few mins with that branch, haven't seen the issue so far 😄 I'll keep an eye out throughout the day tomorrow too

nice 🙂

it's not perfect as it uses a fuzzy search, the idea is that it scores all symbols with your query and order by the score

Fuzzy meaning the search term has to be a substring of the searched term, no?

If so, should create-topic be matched by the search term try-pop?

no, is just a "smart" search: https://whatis.techtarget.com/definition/fuzzy-search#:~:text=A%20fuzzy%20search%20is%20a,correspond%20to%20the%20desired%20information.

we could probably tune the config I just used the defaults from https://github.com/Yomguithereal/clj-fuzzy

Ah, I see. I had the wrong expectations. Reading the doc for clj-fuzzy, I couldnt tell whether it uses all the algorithms together or the client (of the lib) has to pick which ones.

the client has to pick one

https://github.com/clojure-lsp/clojure-lsp/blob/master/src/clojure_lsp/feature/workspace_symbols.clj#L15

Feel free to improve that, I think it can be improved somehow

Maybe the score filter on the line below should check if it's equal to 0?

I see. Is there any way to make the algo configurable by the user? Actually, generally-speaking, is there a way to configure other stuff for LSP (by the user)?

ATM no, and I don't see what options we could have for that, I think improving it to looks like a good filtering with the given user input would be ok for most of the cases. > is there a way to configure other stuff for LSP (by the user) what do you mean with other stuff?

Personally, I've found that substring matches are most intuitive. I don't know the name of this algo (if it even has one), but basically, given something like: • search term: abcde • matches: 1a23bc4d5e6 (because abcde's chars are consecutively within) (Vscode and Emacs's ivy-fuzzy uses this afaik)

> what do you mean with other stuff? I actually didn't have any concrete examples in mind when I said this. I suppose stuff like: Max matches to fetch for auto-completion?

> Personally, I've found that substring matches are most intuitive. I don't know the name of this algo (if it even has one), but basically, given something like: > > yeah, looks valid indeed

we could have a setting for that on .lsp/config.edn like:

{:workspace-symbol-filter :substring}

{:workspace-symbol-filter :fuzzy-dice}

AH, that sounds great. I'll try to find out what the exact algorithm is (for the one I'm describing)

I think the algo I'm talking about is used in regex-fuzzy. https://oremacs.com/2016/01/06/ivy-flx/ I can port the elisp code to a clojure lib?

maybe clj-fuzzy could have another algo with that ? 🙂

Ah sure, I'll make a PR to them

@ericdallo I opened https://github.com/Yomguithereal/clj-fuzzy/issues/49for clj-fuzzy. It doesn't feel appropriate to put this algorithm into clj-fuzzy. So, I'm thinking I just publish a separate lib (like https://github.com/lewang/flx, so maybe call it clj-flx or something). Then I'll make a PR to clojure-lsp so that {:workspace-symbol-filter :flx} uses the lib?

Oh, seems good 🙂

If the algo is better than the current one, I think we could just replace the clj-fuzzy with that new one

Right, ok. I'll write extensive unit tests to show how intuitive (or otherwise) it is

@ericdallo I've published https://github.com/anonimitoraf/clj-flx 🙂 I'll inject it into clojure-lsp later/tomorrow. Are there other parts of clojure-lsp that uses "fuzzy" candidate matching?

Nice! No, this is the only one :)