Thai combining marks cause width miscalculation in modern terminals (Kitty, Ghostty) #3957
Description
Problem
Thai combining marks (vowels and tone marks) are marked as zero-width in CELL_WIDTHS, following Unicode's Mn (Nonspacing Mark) classification. However, modern terminal emulators (tested: Kitty, Ghostty) render Thai text with these characters taking visible width, causing table columns and other layouts to misalign.
Reproduction
from rich.console import Console
from rich.table import Table
from rich.cells import cell_len
console = Console()
table = Table(title='Thai Text Width Test')
table.add_column('Text', width=10)
table.add_column('cell_len')
for text in ['สวัสดี', 'น้ำ', 'กรุงเทพ']:
table.add_row(text, str(cell_len(text)))
console.print(table)Width Analysis
| Text | Codepoints | Combining marks | Rich cell_len | Terminal width |
|---|---|---|---|---|
| สวัสดี | 6 | 2 | 4 | 6 |
| น้ำ | 3 | 1 | 2 | 3 |
| กรุงเทพ | 7 | 1 | 6 | 7 |
The discrepancy equals the number of combining marks — terminals render each Thai codepoint as width 1.
Affected Ranges in CELL_WIDTHS
| Range | Characters | Unicode Category |
|---|---|---|
| U+0E31 | ั | Mn (sara am vowel) |
| U+0E34-U+0E3A | ิ ี ึ ื ุ ู ฺ | Mn (vowel signs) |
| U+0E47-U+0E4E | ็ ่ ้ ๊ ๋ ์ ํ ๎ | Mn (tone marks) |
Environment
- Rich: 13.7.1
- Python: 3.14.2
- Terminals tested:
- Kitty 0.45.0 — affected
- Ghostty 1.2.3 — affected
- Platform: macOS
Proposed Fix
Change Thai combining mark ranges from width 0 to width 1 in _cell_widths.py. This matches actual terminal rendering behavior for Thai script.