Global Search yields incorrect results sometimes

[nik-rev]

Summary

global search yields incorrect results sometimes

Reproduction Steps

1. open helix in helix's codebase
2. global search with space + /
3. input hk
4. you also get results for 繁

expected:

• do not get results for 繁, but do get results for hk

actual:

• do get results for 繁, and also get results for hk

Helix log

~/.cache/helix/helix.log

not relevant

2025-03-26T22:27:09.981 helix_lsp::transport [ERROR] Tried sending response into a closed channel (id=Num(9)), original request likely timed out
2025-03-26T22:32:43.674 helix_lsp::transport [ERROR] rust-analyzer <- InvalidParams: code action without data
2025-03-26T22:33:45.119 helix_lsp::transport [ERROR] Tried sending response into a closed channel (id=Num(60)), original request likely timed out
2025-03-26T22:45:03.058 helix_lsp::transport [ERROR] rust-analyzer <- InvalidParams: code action without data
2025-03-26T22:46:30.048 helix_lsp::transport [ERROR] Tried sending response into a closed channel (id=Num(178)), original request likely timed out
2025-03-26T22:46:55.775 helix_lsp::transport [ERROR] Tried sending response into a closed channel (id=Num(180)), original request likely timed out
2025-03-26T22:47:05.881 helix_lsp::transport [ERROR] Tried sending response into a closed channel (id=Num(185)), original request likely timed out
2025-03-26T22:47:07.267 helix_lsp::transport [ERROR] Tried sending response into a closed channel (id=Num(184)), original request likely timed out

Platform

Linux

Terminal Emulator

wezterm 8e9cf91

Installation Method

source

Helix Version

helix 25.01.1 (d238d51)

[pppKin]

the search results you see are from one of the files in helix-view/tests/encoding/, which is not encoded in utf8. And by default our searcher(which is essentially ripgrep) treats everything as utf8.
Since utf8 is ubiquitous these days, even if we want to detect file encoding and specify it in the searcher, we might not want to enable it by default.

[pascalkuthe]

Yeah I am pretty sure we dintwant to start doing encoding detection in the searcher. That adds a lot of overhead for something very niche. I am not even sure we can Implement that efficiently. Probably would have to be implemented upstream in ripgrep/grep-searcher