Unicode Width and Multi-byte Character Support

This document lists all the places in the codebase where visual display width must be accounted for, particularly for:

• CJK (Chinese, Japanese, Korean) characters that occupy 2 terminal columns
• Emoji that occupy 2 terminal columns
• Zero-width characters (combining marks, zero-width space)
• Multi-byte UTF-8 characters (which may not correspond to visual width)

Core Utility Module

src/primitives/display_width.rs

The central utility module for all visual width calculations. Uses the unicode-width crate.

Key functions:

• char_width(c: char) -> usize - Returns visual width of a character (0, 1, or 2)
• str_width(s: &str) -> usize - Returns total visual width of a string
• DisplayWidth trait - Extension trait for convenient .display_width() method

Usage: Import from crate::primitives::display_width::{char_width, str_width}

Critical Rendering Code

src/primitives/line_wrapping.rs

Line wrapping must wrap based on visual width, not character count.

Key locations:

• wrap_line() function (lines 111-127) - Tracks segment_visual_width using char_width(c)
• Uses visual width to determine when to break lines
• Property tests verify visual width correctness

src/view/ui/split_rendering.rs

The main buffer rendering code that positions characters and cursors on screen.

Key locations:

• Line ~2150: visible_char_count += char_width(ch) - Tracks visual column during character iteration
• Line ~2927: str_width(&span.content) - Counts span width for cursor position detection (in test code)

Important: The visible_char_count variable tracks visual columns, not character indices.

src/view/viewport.rs

Viewport scrolling calculations for horizontal scroll.

Key locations:

• Lines ~381-394: Calculates cursor visual column by walking through characters and summing widths
• Line ~394: str_width(line_text) for line visual width
• ensure_column_visible_simple() uses visual widths for scroll calculations

src/primitives/ansi.rs

ANSI escape code handling must exclude invisible escape sequences from width.

Key locations:

• visible_char_count() function (lines ~302-319) - Returns visual width excluding ANSI codes
• Uses str_width() for fast path (no ANSI) and char_width() for character-by-character parsing

UI Components

src/view/ui/status_bar.rs

Status bar text truncation must account for visual width.

Key locations:

• Lines ~504-529: Truncates left status by visual width using str_width() and char_width()
• Lines ~538: Calculates padding using str_width(&displayed_left)
• Lines ~602-639: Similar truncation logic for narrow terminal case

Truncation pattern:

let visual_width = str_width(&text);
if visual_width > max_width {
    let mut width = 0;
    let truncated: String = text
        .chars()
        .take_while(|ch| {
            let w = char_width(*ch);
            if width + w <= truncate_at {
                width += w;
                true
            } else {
                false
            }
        })
        .collect();
    format!("{}...", truncated)
}

src/view/ui/tabs.rs

Tab bar width calculations.

Key locations:

• Lines ~200-205: str_width(&tab_name_text) and str_width(close_text) for tab sizing

src/view/ui/file_explorer.rs

File explorer tree view layout.

Key locations:

• Line ~176: str_width(&node.entry.name) for name column width
• Line ~249: str_width(&size_text) for file size alignment

src/view/ui/file_browser.rs

File browser popup for Open File dialog.

Key locations:

• Line ~567: str_width(label) + 2 for shortcut label width in navigation bar

src/view/ui/suggestions.rs

Autocomplete/command palette suggestion rendering.

Key locations (all use visual-width-aware truncation):

• Lines ~139-162: Command name column
• Lines ~188-213: Keybinding column
• Lines ~235-271: Description column
• Lines ~310-331: Source column

Each column uses the truncation pattern shown above.

Locations That DON'T Need Visual Width

Some .chars().count() usages are intentionally character-based:

Character Indexing into Data Structures

• char_mappings vectors in split_rendering.rs (lines ~1003, ~2162, ~2271)
  • These index into character-indexed data structures, not visual positions
  • The vectors have one element per character, not per visual column

Test Assertions

• Test code that verifies character count vs byte count (e.g., fancy_quote.chars().count() == 1)

Common Pitfalls

1. Byte length vs character count vs visual width:

   • s.len() - byte count (wrong for multi-byte UTF-8)
   • s.chars().count() - character count (wrong for double-width)
   • str_width(s) - visual width (correct for display)

2. Truncation must use visual width:

   rustCopy

   // WRONG: truncates by character count
   text.chars().take(max_width).collect()
   
   // RIGHT: truncates by visual width
   let mut width = 0;
   text.chars().take_while(|ch| {
       let w = char_width(*ch);
       if width + w <= max_width { width += w; true } else { false }
   }).collect()
   

3. Cursor positioning:

   • Screen X coordinate = sum of visual widths of characters before cursor
   • NOT the character index

4. Line wrapping:

   • Wrap when visual width exceeds terminal width
   • NOT when character count exceeds

Testing

Unit Tests

• src/primitives/display_width.rs - Tests for width functions
• src/primitives/line_wrapping.rs - Tests including:
  • test_visual_width_calculation
  • test_wrap_line_double_width_characters
  • test_wrap_line_emoji_visual_width
  • test_wrap_line_mixed_ascii_and_cjk
  • test_chars_count_vs_visual_width_bug (regression test)

Property Tests

• src/primitives/line_wrapping.rs::proptests module
• Tests wrap_line with various Unicode character combinations

E2E Tests

• tests/e2e/multibyte_characters.rs - End-to-end tests for multi-byte character handling

Dependencies

• unicode-width = "0.2" in Cargo.toml
• Uses UAX#11 (Unicode Standard Annex #11) East Asian Width property