# PR #1467: Fix Unicode handling in chat_composer "@" token detection
Status: closed | Created: July 6, 2025 | Author: ryozi-tn
Base: main ← Head: fix-1450

## Description
 ## Issues Fixed

- **Primary Issue (#1450)**: Unicode cursor positioning was incorrect due to mixing  character positions with byte positions 
- **Additional Issue**: Full-width spaces (CJK whitespace like "　") weren't properly handled as token boundaries
  - ref: https://doc.rust-lang.org/std/primitive.char.html#method.is_whitespace


## Files Changed (1)
Additions: +175 | Deletions: -18 | Files: 1

## Comments (3)
### github-actions[bot] - July 6, 2025
All contributors have signed the CLA  ✍️ ✅<br/><sub>Posted by the ****CLA Assistant Lite bot****.</sub>

### ryozi-tn - July 6, 2025
I have read the CLA Document and I hereby sign the CLA

### bolinfest - July 7, 2025
Admittedly, when I introduced this code and asked Codex to code review it on https://github.com/openai/codex/pull/1401, it noted:

> `current_at_token`/`insert_selected_path` index by byte; non-ASCII filenames might misbehave—worth documenting or switching to `char_indices`.

## Reviews (1)
### bolinfest - COMMENTED - July 7, 2025
Really great work and thank you for the test!

Can you please apply the suggested changes and then run the formatter:

https://github.com/openai/codex/blob/c221eab0b5cad59ce3dafebf7ca630f217263cc6/codex-rs/justfile#L25

## Labels
No labels