Closed WeetHet closed 8 months ago
same issue with golang:
A fix for this will go out with the next release. Thanks so much for the feedback! 🙏
Strangely, this seems to still not be fixed.
From the discord, I verified this behavior as well:
I think it would be great if it suggest typescript primitives types first too (for me at least, I use primitives more often)
Similar issue in vue template files while autocompleting imported components
Similar issue with PHP (in this case using the Laravel framework).
I start typing whereBetween
however the first result on the autocomplete is orWhereBetween
and the actual one I want is further down.
This seems to affect the project file search (cmd+p) as well. Here I search for the file mi.php
and it's the third option down when it should really be the first one. The first suggestion here doesn't make sense given it's only a partial match:
@DannyJJK that last one, that's a different issue: we rank recent items higher. There's a very thin line after the first 2 results. Admittedly, it's really hard to spot and we should probably improve that UI.
In https://github.com/zed-industries/zed/pull/7727 we just made some improvements that should help with the bug that Theo reported and some of the others reported in here (string
/String
and UIAlert
). The vec
vs. Vec
thing in Rust is still unchanged (and matches VS Code btw.!)
We just merged #7727 and it attempts to fix the order.
If problems like this pop up again, I think it would be best to have separate issues with clear descriptions on how to reproduce the problem and what the expected/actual behavior are, since there's a lot of nuance in how these completions are sorted and we don't want to make something worse by loosely making something else better :)
Same problem happens with C++ too.
Same problem happens with C++ too.
Can you provide more information? i.e. what did you type, what did you expect to see, what did you see instead, etc.?
Same problem happens with C++ too.
Can you provide more information? i.e. what did you type, what did you expect to see, what did you see instead, etc.?
Sure, when I write "for" it does not show for loop instead it shows for_each function. I assume for loop should be most used. And I never used for_each in my project or file so it is recommending it out of nowhere. Zed version 0.125.3 ![Uploading Screenshot 2024-03-09 at 4.59.28 PM.png…]()
I can't see that screenshot, but that does sound like an issue. Like I wrote above:
I think it would be best to have separate issues with clear descriptions on how to reproduce the problem and what the expected/actual behavior are
Can you create a new ticket for this?
Same here, not fixed yet...
@bernardinorafael can you create a ticket with a minimal reproducible example, what you expect to see, what you get instead?
I've created a new ticket https://github.com/zed-industries/zed/issues/12590 with steps to reproduce in Elm
Tailwind completion order is really not great as well
It should be suggesting just "px-2,..." when I scroll down, they are there, but this should be the first suggestion
Can somebody else confirm that this is still an issue? This was addressed in #2169, but it breaks completions for me. If I enable "smart case" in fuzzy::match_strings
, meaning it's always true
, then completions seem to work as expected. I've tested it for Rust and TypeScript.
@tinrab
running ok for me (zed 0.147.2)
Running version 0.151.2 I realize that the autocomplete order is really inconvenient...
I noticed this by trying zed in tailwind and typescript
Check for existing issues
Describe the feature
In VSCode rust-ananlyzer puts
Vec
type beforevec!
macro when autocompleting something likeIn Zed autocomplete order also definitely should respect at least the case of item being autocompleted, and if possible also try rearanging completions arrording to some heuristic or maybe even some pretrained neural network, which would assign weight to completions
Also, using macros for types is quite rare, so they can be moved down
If applicable, add mockups / screenshots to help present your vision of the feature