| 1 | //! Shared text helpers for TUI selection and clipboard workflows. |
| 2 | |
| 3 | use ratatui::text::{Line, Span}; |
| 4 | use unicode_segmentation::UnicodeSegmentation; |
| 5 | use unicode_width::{UnicodeWidthChar, UnicodeWidthStr}; |
| 6 | |
| 7 | use crate::tui::history::HistoryCell; |
| 8 | use crate::tui::osc8; |
| 9 | |
| 10 | #[derive(Debug, Clone, Copy, PartialEq, Eq)] |
| 11 | pub(crate) enum CopyLineSeparator { |
| 12 | None, |
| 13 | Space, |
| 14 | Newline, |
| 15 | } |
| 16 | |
| 17 | impl CopyLineSeparator { |
| 18 | #[must_use] |
| 19 | pub(crate) const fn as_str(self) -> &'static str { |
| 20 | match self { |
| 21 | Self::None => "", |
| 22 | Self::Space => " ", |
| 23 | Self::Newline => "\n", |
| 24 | } |
| 25 | } |
| 26 | } |
| 27 | |
| 28 | pub(crate) fn truncate_line_to_width(text: &str, max_width: usize) -> String { |
| 29 | if max_width == 0 { |
| 30 | return String::new(); |
| 31 | } |
| 32 | if text_display_width(text) <= max_width { |
| 33 | return text.to_string(); |
| 34 | } |
| 35 | // For very small budgets, take whole graphemes until the next one would |
| 36 | // exceed the display width. Never split an emoji or combining sequence. |
| 37 | if max_width <= 3 { |
| 38 | let mut out = String::new(); |
| 39 | let mut width = 0usize; |
| 40 | for grapheme in text.graphemes(true) { |
| 41 | let grapheme_width = grapheme_display_width(grapheme); |
| 42 | if width + grapheme_width > max_width { |
| 43 | break; |
| 44 | } |
| 45 | out.push_str(grapheme); |
| 46 | width += grapheme_width; |
| 47 | } |
| 48 | return out; |
| 49 | } |
| 50 | |
| 51 | let mut out = String::new(); |
| 52 | let mut width = 0usize; |
| 53 | let limit = max_width.saturating_sub(3); |
| 54 | for grapheme in text.graphemes(true) { |
| 55 | let grapheme_width = grapheme_display_width(grapheme); |
| 56 | if width + grapheme_width > limit { |
| 57 | break; |
| 58 | } |
| 59 | out.push_str(grapheme); |
| 60 | width += grapheme_width; |
| 61 | } |
| 62 | out.push_str("..."); |
| 63 | out |
| 64 | } |
| 65 | |
| 66 | /// Truncate `text` to `max_width` display columns, preferring whole words. |
| 67 | pub(crate) fn semantic_truncate(text: &str, max_width: usize) -> String { |
| 68 | if max_width == 0 { |
| 69 | return String::new(); |
| 70 | } |
| 71 | if text_display_width(text) <= max_width { |
| 72 | return text.to_string(); |
| 73 | } |
| 74 | |
| 75 | const ELLIPSIS: char = '…'; |
| 76 | let ellipsis_width = char_display_width(ELLIPSIS); |
| 77 | let limit = max_width.saturating_sub(ellipsis_width); |
| 78 | if limit == 0 { |
| 79 | return ELLIPSIS.to_string(); |
| 80 | } |
| 81 | |
| 82 | let mut width = 0usize; |
| 83 | let mut cut_byte = 0usize; |
| 84 | let mut last_word_end = None; |
| 85 | let mut in_word = false; |
| 86 | for (byte_idx, grapheme) in text.grapheme_indices(true) { |
| 87 | let grapheme_width = grapheme_display_width(grapheme); |
| 88 | if width + grapheme_width > limit { |
| 89 | break; |
| 90 | } |
| 91 | width += grapheme_width; |
| 92 | cut_byte = byte_idx + grapheme.len(); |
| 93 | if grapheme.chars().all(char::is_whitespace) { |
| 94 | if in_word { |
| 95 | last_word_end = Some(byte_idx); |
| 96 | in_word = false; |
| 97 | } |
| 98 | } else { |
| 99 | in_word = true; |
| 100 | } |
| 101 | } |
| 102 | if cut_byte == 0 { |
| 103 | return ELLIPSIS.to_string(); |
| 104 | } |
| 105 | |
| 106 | let mut body = if let Some(word_end) = last_word_end { |
| 107 | text[..word_end].trim_end() |
| 108 | } else { |
| 109 | text[..cut_byte].trim_end() |
| 110 | }; |
| 111 | if body.is_empty() { |
| 112 | body = text[..cut_byte].trim_end(); |
| 113 | } |
| 114 | let mut out = body.to_string(); |
| 115 | out.push(ELLIPSIS); |
| 116 | out |
| 117 | } |
| 118 | |
| 119 | pub(crate) fn semantic_truncate_with_affixes( |
| 120 | prefix: &str, |
| 121 | text: &str, |
| 122 | suffix: &str, |
| 123 | max_width: usize, |
| 124 | ) -> String { |
| 125 | let fixed_width = text_display_width(prefix) + text_display_width(suffix); |
| 126 | if fixed_width > max_width { |
| 127 | return semantic_truncate(&format!("{prefix}{text}{suffix}"), max_width); |
| 128 | } |
| 129 | format!( |
| 130 | "{prefix}{}{suffix}", |
| 131 | semantic_truncate_between_affixes(prefix, text, suffix, max_width) |
| 132 | ) |
| 133 | } |
| 134 | |
| 135 | pub(crate) fn semantic_truncate_between_affixes( |
| 136 | prefix: &str, |
| 137 | text: &str, |
| 138 | suffix: &str, |
| 139 | max_width: usize, |
| 140 | ) -> String { |
| 141 | let fixed_width = text_display_width(prefix) + text_display_width(suffix); |
| 142 | if fixed_width > max_width { |
| 143 | return String::new(); |
| 144 | } |
| 145 | semantic_truncate(text, max_width - fixed_width) |
| 146 | } |
| 147 | |
| 148 | pub(super) fn history_cell_to_text(cell: &HistoryCell, width: u16) -> String { |
| 149 | cell.transcript_lines(width) |
| 150 | .into_iter() |
| 151 | .map(line_to_string) |
| 152 | .collect::<Vec<_>>() |
| 153 | .join("\n") |
| 154 | } |
| 155 | |
| 156 | fn line_to_string(line: Line<'static>) -> String { |
| 157 | let mut out = String::new(); |
| 158 | append_spans_plain(line.spans.iter(), &mut out); |
| 159 | out |
| 160 | } |
| 161 | |
| 162 | /// Convert a rendered transcript line to plain text, stripping OSC-8 link |
| 163 | /// escape sequences. The caller is responsible for shifting selection columns |
| 164 | /// to account for any visual-only rail prefix (see |
| 165 | /// `TranscriptViewCache::rail_prefix_width`). |
| 166 | pub(super) fn line_to_plain(line: &Line<'static>) -> String { |
| 167 | let mut out = String::new(); |
| 168 | append_spans_plain(line.spans.iter(), &mut out); |
| 169 | out |
| 170 | } |
| 171 | |
| 172 | fn append_spans_plain<'a, I>(spans: I, out: &mut String) |
| 173 | where |
| 174 | I: Iterator<Item = &'a Span<'a>>, |
| 175 | { |
| 176 | for span in spans { |
| 177 | if span.content.contains('\x1b') { |
| 178 | osc8::strip_into(&span.content, out); |
| 179 | } else { |
| 180 | out.push_str(span.content.as_ref()); |
| 181 | } |
| 182 | } |
| 183 | } |
| 184 | |
| 185 | pub(crate) fn text_display_width(text: &str) -> usize { |
| 186 | text.graphemes(true).map(grapheme_display_width).sum() |
| 187 | } |
| 188 | |
| 189 | pub(super) fn slice_text(text: &str, start: usize, end: usize) -> String { |
| 190 | if end <= start { |
| 191 | return String::new(); |
| 192 | } |
| 193 | |
| 194 | let mut out = String::new(); |
| 195 | let mut col = 0usize; |
| 196 | for grapheme in text.graphemes(true) { |
| 197 | let grapheme_width = grapheme_display_width(grapheme); |
| 198 | let grapheme_start = col; |
| 199 | let grapheme_end = col.saturating_add(grapheme_width); |
| 200 | if grapheme_end > start && grapheme_start < end { |
| 201 | out.push_str(grapheme); |
| 202 | } |
| 203 | col = grapheme_end; |
| 204 | if col >= end { |
| 205 | break; |
| 206 | } |
| 207 | } |
| 208 | out |
| 209 | } |
| 210 | |
| 211 | pub(super) fn char_display_width(ch: char) -> usize { |
| 212 | match ch { |
| 213 | '\t' => 4, |
| 214 | // Enclosed Alphanumerics (U+2460-U+24FF), Dingbat Circled Digits |
| 215 | // (U+2776-U+2793), and Circled Numbers on Black Square (U+3248-U+324F) |
| 216 | // have East Asian Width "Ambiguous" but are rendered as 2 columns in |
| 217 | // CJK terminals. unicode-width's width() conservatively reports 1; we |
| 218 | // match the terminal. (#4479) |
| 219 | '\u{2460}'..='\u{24FF}' | '\u{2776}'..='\u{2793}' | '\u{3248}'..='\u{324F}' => 2, |
| 220 | _ => { |
| 221 | // `width()` returns `None` for control/unassigned chars (default them to |
| 222 | // one column so layout doesn't collapse) and `Some(0)` for genuinely |
| 223 | // zero-width chars — combining marks, ZWJ, zero-width spaces — which must |
| 224 | // stay 0 so display-width math (truncation, slicing, overflow, copy) |
| 225 | // matches what the terminal actually renders. |
| 226 | UnicodeWidthChar::width(ch).unwrap_or(1) |
| 227 | } |
| 228 | } |
| 229 | } |
| 230 | |
| 231 | /// Measure one extended grapheme using the same string-level Unicode rules as |
| 232 | /// Ratatui. String width intentionally differs from the sum of codepoint widths |
| 233 | /// for keycaps, ZWJ emoji, modifiers, and other terminal ligatures. |
| 234 | /// |
| 235 | /// Keycap sequences (such as 1\u{fe0f}\u{20e3}) that lack an FE0F variation |
| 236 | /// selector still render as 2 columns in terminals, but unicode-width's |
| 237 | /// `grapheme.width()` only reports 1. We force 2 when U+20E3 is present in a |
| 238 | /// multi-codepoint grapheme. (#4479) |
| 239 | pub(super) fn grapheme_display_width(grapheme: &str) -> usize { |
| 240 | if grapheme == "\t" { |
| 241 | return 4; |
| 242 | } |
| 243 | if let Some(ch) = grapheme.chars().next() |
| 244 | && ch.len_utf8() == grapheme.len() |
| 245 | { |
| 246 | return char_display_width(ch); |
| 247 | } |
| 248 | // Keycap sequences always render as 2 columns. unicode-width's |
| 249 | // `width()` undercounts the non-FE0F variant to 1. |
| 250 | if grapheme.contains('\u{20e3}') { |
| 251 | return 2; |
| 252 | } |
| 253 | UnicodeWidthStr::width(grapheme) |
| 254 | } |
| 255 | |
| 256 | #[cfg(test)] |
| 257 | mod tests { |
| 258 | use super::*; |
| 259 | use ratatui::text::Span; |
| 260 | |
| 261 | #[test] |
| 262 | fn line_to_plain_strips_osc_8_wrapper() { |
| 263 | let wrapped = format!( |
| 264 | "\x1b]8;;{}\x1b\\{}\x1b]8;;\x1b\\", |
| 265 | "https://example.com", "https://example.com" |
| 266 | ); |
| 267 | let line = Line::from(vec![ |
| 268 | Span::raw("see "), |
| 269 | Span::raw(wrapped), |
| 270 | Span::raw(" for details"), |
| 271 | ]); |
| 272 | let text = line_to_plain(&line); |
| 273 | assert_eq!(text, "see https://example.com for details"); |
| 274 | } |
| 275 | |
| 276 | #[test] |
| 277 | fn line_to_plain_passes_through_plain_spans() { |
| 278 | let line = Line::from(vec![Span::raw("plain "), Span::raw("text")]); |
| 279 | let text = line_to_plain(&line); |
| 280 | assert_eq!(text, "plain text"); |
| 281 | } |
| 282 | |
| 283 | #[test] |
| 284 | fn line_to_plain_includes_all_spans() { |
| 285 | // Visual-only rail spans are stripped by the caller using |
| 286 | // TranscriptViewCache::rail_prefix_width — line_to_plain itself |
| 287 | // is a faithful span-to-string pass-through. |
| 288 | let line = Line::from(vec![Span::raw("\u{2502} "), Span::raw("tool output")]); |
| 289 | let text = line_to_plain(&line); |
| 290 | assert_eq!(text, "\u{2502} tool output"); |
| 291 | } |
| 292 | |
| 293 | #[test] |
| 294 | fn slice_text_respects_column_bounds() { |
| 295 | let text = "hello world"; |
| 296 | assert_eq!(slice_text(text, 0, 5), "hello"); |
| 297 | assert_eq!(slice_text(text, 6, 11), "world"); |
| 298 | assert_eq!(slice_text(text, 0, 0), ""); |
| 299 | assert_eq!(slice_text(text, 0, 100), text); |
| 300 | } |
| 301 | |
| 302 | #[test] |
| 303 | fn slice_text_handles_multibyte_characters() { |
| 304 | let text = "a─b"; // U+2500 is 1 display column on supported terminals |
| 305 | assert_eq!(slice_text(text, 1, 2), "─"); |
| 306 | assert_eq!(slice_text(text, 0, 3), text); |
| 307 | } |
| 308 | |
| 309 | #[test] |
| 310 | fn slice_text_truncates_at_end() { |
| 311 | let text = "ab"; |
| 312 | assert_eq!(slice_text(text, 1, 5), "b"); |
| 313 | } |
| 314 | |
| 315 | // --- Unicode / CJK / terminal-width QA (issue #3488) ------------------- |
| 316 | // These exercise the production width helpers directly so the assertions |
| 317 | // track the same code path the renderer uses. |
| 318 | |
| 319 | #[test] |
| 320 | fn text_display_width_counts_cjk_as_two_columns() { |
| 321 | assert_eq!(text_display_width("中文"), 4); // two wide glyphs |
| 322 | assert_eq!(text_display_width("Hello世界"), 9); // 5 ASCII + 2×2 |
| 323 | // Full-width (ambiguous→wide) punctuation is two columns each. |
| 324 | assert_eq!(text_display_width("，。！？"), 8); |
| 325 | } |
| 326 | |
| 327 | #[test] |
| 328 | fn text_display_width_treats_zero_width_marks_as_zero() { |
| 329 | // A combining mark adds no column: "e" + U+0301 renders as one cell. |
| 330 | // (Regression guard: the old `.max(1)` counted it as 1, over-reporting |
| 331 | // width and causing premature truncation / border drift on text with |
| 332 | // combining marks or ZWJ emoji sequences.) |
| 333 | assert_eq!(text_display_width("e\u{0301}"), 1); |
| 334 | assert_eq!(text_display_width("cafe\u{0301}"), 4); |
| 335 | // The complete ZWJ emoji is one two-column grapheme, matching Ratatui. |
| 336 | assert_eq!(text_display_width("\u{1F469}\u{200D}\u{1F4BB}"), 2); |
| 337 | } |
| 338 | |
| 339 | #[test] |
| 340 | fn text_display_width_keeps_control_and_tab_widths() { |
| 341 | // Control chars still occupy a column (avoid layout collapse); tab = 4. |
| 342 | assert_eq!(text_display_width("a\u{0007}b"), 3); |
| 343 | assert_eq!(text_display_width("\t"), 4); |
| 344 | assert_eq!(text_display_width("\ta"), 5); |
| 345 | } |
| 346 | |
| 347 | #[test] |
| 348 | fn truncate_line_to_width_respects_display_width_not_byte_len() { |
| 349 | // No truncation when the string already fits by display width. |
| 350 | assert_eq!(truncate_line_to_width("中文", 10), "中文"); |
| 351 | // Oversized: reserve 3 cols for the ellipsis, fill the rest by width. |
| 352 | let out = truncate_line_to_width("中文测试", 7); |
| 353 | assert_eq!(out, "中文..."); |
| 354 | assert_eq!(text_display_width(&out), 7); |
| 355 | // Never split a wide glyph across the boundary, and never emit U+FFFD. |
| 356 | let clipped = truncate_line_to_width("界界界界界", 5); |
| 357 | assert!(text_display_width(&clipped) <= 5); |
| 358 | assert!(!clipped.contains('\u{FFFD}')); |
| 359 | } |
| 360 | |
| 361 | #[test] |
| 362 | fn semantic_truncate_prefers_word_boundaries() { |
| 363 | let out = semantic_truncate("hello world foo bar", 14); |
| 364 | assert_eq!(out, "hello world…"); |
| 365 | assert!(text_display_width(&out) <= 14); |
| 366 | } |
| 367 | |
| 368 | #[test] |
| 369 | fn semantic_truncate_falls_back_with_long_words_and_wide_glyphs() { |
| 370 | let long_word = semantic_truncate("supercalifragilistic", 8); |
| 371 | assert_eq!(long_word, "superca…"); |
| 372 | assert!(text_display_width(&long_word) <= 8); |
| 373 | |
| 374 | let cjk = semantic_truncate("中文测试文本", 7); |
| 375 | assert_eq!(cjk, "中文测…"); |
| 376 | assert!(text_display_width(&cjk) <= 7); |
| 377 | } |
| 378 | |
| 379 | #[test] |
| 380 | fn semantic_truncate_handles_empty_and_tiny_budgets() { |
| 381 | assert_eq!(semantic_truncate("", 10), ""); |
| 382 | assert_eq!(semantic_truncate("hello", 0), ""); |
| 383 | assert_eq!(semantic_truncate("hello", 1), "…"); |
| 384 | } |
| 385 | |
| 386 | #[test] |
| 387 | fn semantic_truncate_between_affixes_reserves_fixed_columns() { |
| 388 | let hint = semantic_truncate_between_affixes( |
| 389 | " > [ ] Prefix stability (", |
| 390 | "whether system/tools stayed cacheable", |
| 391 | ")", |
| 392 | 49, |
| 393 | ); |
| 394 | let row = format!(" > [ ] Prefix stability ({hint})"); |
| 395 | assert_eq!(hint, "whether system/tools…"); |
| 396 | assert!(text_display_width(&row) <= 49); |
| 397 | } |
| 398 | |
| 399 | #[test] |
| 400 | fn slice_text_slices_cjk_by_display_column() { |
| 401 | // Columns: 中=[0,2) 文=[2,4) a=[4,5) b=[5,6) |
| 402 | let text = "中文ab"; |
| 403 | assert_eq!(slice_text(text, 0, 2), "中"); |
| 404 | assert_eq!(slice_text(text, 2, 4), "文"); |
| 405 | assert_eq!(slice_text(text, 4, 6), "ab"); |
| 406 | } |
| 407 | |
| 408 | // --- New #3488 fixtures: CJK/wide-glyph truncation on selector-style rows. |
| 409 | // truncate_line_to_width is the production helper behind sidebar (file_tree), |
| 410 | // statusline (footer_ui), hotbar, and picker (mouse_ui) row rendering, so |
| 411 | // these exercise the same truncation path those surfaces use. |
| 412 | |
| 413 | #[test] |
| 414 | fn truncate_line_to_width_full_width_cjk_lands_on_glyph_boundary() { |
| 415 | // Each Han glyph is two columns. With an odd budget the truncation must |
| 416 | // land on a whole-glyph boundary (reserving three columns for the |
| 417 | // ellipsis), never leaving a half-rendered wide cell or emitting U+FFFD. |
| 418 | let title = "项目报告结果"; // 6 glyphs, 12 columns |
| 419 | let out = truncate_line_to_width(title, 7); |
| 420 | // Budget 7 -> limit 4 columns -> two glyphs fit, then the ellipsis. |
| 421 | assert_eq!(out, "项目..."); |
| 422 | assert_eq!(text_display_width(&out), 7); |
| 423 | // The kept prefix is composed only of whole wide glyphs (each 2 cols), |
| 424 | // proving the boundary glyph was dropped whole, not split. |
| 425 | let prefix = out.strip_suffix("...").expect("ellipsis present"); |
| 426 | assert!(prefix.chars().all(|c| char_display_width(c) == 2)); |
| 427 | assert!(!out.contains('\u{FFFD}')); |
| 428 | } |
| 429 | |
| 430 | #[test] |
| 431 | fn truncate_line_to_width_mixed_ascii_cjk_row_keeps_ellipsis_within_budget() { |
| 432 | // A sidebar/selector row mixing an ASCII label with a CJK title, wider |
| 433 | // than the column budget, must truncate with a trailing ellipsis that |
| 434 | // still fits by display width and must not split a wide glyph. |
| 435 | let row = "Task: 数据库迁移任务 done"; // ASCII label + 7 Han glyphs |
| 436 | let budget = 12; |
| 437 | let out = truncate_line_to_width(row, budget); |
| 438 | assert!(out.ends_with("..."), "expected ellipsis, got {out:?}"); |
| 439 | // Ellipsis-and-content fit within the budget by *display* width. |
| 440 | assert!(text_display_width(&out) <= budget); |
| 441 | // The non-ellipsis prefix stays within budget-minus-ellipsis, so the |
| 442 | // wide glyph on the boundary was dropped whole rather than half-drawn. |
| 443 | let prefix = out.strip_suffix("...").expect("ellipsis present"); |
| 444 | assert!(text_display_width(prefix) <= budget - 3); |
| 445 | assert!(!out.contains('\u{FFFD}')); |
| 446 | // The semantic ASCII prefix survives truncation. |
| 447 | assert!(out.starts_with("Task:")); |
| 448 | } |
| 449 | |
| 450 | #[test] |
| 451 | fn truncate_line_to_width_dense_cjk_selector_row_survives_narrow_widths() { |
| 452 | // Picker/selector rows degrade through truncate_line_to_width when the |
| 453 | // terminal is narrow. A dense row with a leading marker glyph and CJK |
| 454 | // content must stay within budget at tiny widths, without panicking or |
| 455 | // emitting a replacement char from a mid-glyph byte split. |
| 456 | let row = "▸ 中文项目 · main"; // marker + CJK + separator + branch |
| 457 | for width in [1usize, 2, 3, 4, 6, 8] { |
| 458 | let out = truncate_line_to_width(row, width); |
| 459 | assert!( |
| 460 | text_display_width(&out) <= width, |
| 461 | "width={width}: {out:?} exceeds budget" |
| 462 | ); |
| 463 | assert!( |
| 464 | !out.contains('\u{FFFD}'), |
| 465 | "width={width}: truncation split a wide glyph" |
| 466 | ); |
| 467 | } |
| 468 | } |
| 469 | |
| 470 | // --- keycap / grapheme regression guard (#4479) --------------------------- |
| 471 | // Fully qualified keycap sequences render as two columns. Codepoint sums |
| 472 | // report one; the canonical string/grapheme contract reports two. |
| 473 | |
| 474 | #[test] |
| 475 | fn text_display_width_treats_keycap_sequence_as_two_columns() { |
| 476 | for keycap in [ |
| 477 | "1\u{fe0f}\u{20e3}", |
| 478 | "9\u{fe0f}\u{20e3}", |
| 479 | "#\u{fe0f}\u{20e3}", |
| 480 | ] { |
| 481 | assert_eq!(text_display_width(keycap), 2); |
| 482 | assert_eq!(text_display_width(keycap), UnicodeWidthStr::width(keycap)); |
| 483 | } |
| 484 | // A digit directly followed by U+20E3 (without FE0F variation selector) |
| 485 | // still renders as a 2-column keycap in terminals. We force this in |
| 486 | // grapheme_display_width when the grapheme contains U+20E3. |
| 487 | // A standalone U+20E3 is a zero-width combining mark. |
| 488 | assert_eq!(text_display_width("1\u{20e3}"), 2); |
| 489 | assert_eq!(text_display_width("\u{20e3}"), 0); |
| 490 | } |
| 491 | |
| 492 | #[test] |
| 493 | fn circled_digit_display_width() { |
| 494 | assert_eq!(char_display_width('\u{2460}'), 2); |
| 495 | assert_eq!(char_display_width('\u{2461}'), 2); |
| 496 | assert_eq!(char_display_width('\u{24ea}'), 2); |
| 497 | assert_eq!(char_display_width('\u{2776}'), 2); |
| 498 | assert_eq!(text_display_width("\u{2460}\u{2461}\u{2462}"), 6); |
| 499 | assert_eq!(text_display_width("Step \u{2460}: init"), 13); |
| 500 | assert_eq!(text_display_width("A\u{24d0}B"), 4); |
| 501 | } |
| 502 | |
| 503 | #[test] |
| 504 | fn unicode_width_reports_circled_digits_as_two_columns() { |
| 505 | // Regression guard for the unicode-width patch (#4479): Ratatui |
| 506 | // renders text through UnicodeWidthChar::width(), so the patch must |
| 507 | // make even the raw crate API report 2 columns for ambiguous-width |
| 508 | // characters — otherwise Ratatui places them in 1 cell while the |
| 509 | // terminal paints 2, shifting every downstream column. |
| 510 | assert_eq!(UnicodeWidthChar::width('\u{2460}'), Some(2)); |
| 511 | assert_eq!(UnicodeWidthChar::width('\u{24ea}'), Some(2)); |
| 512 | assert_eq!(UnicodeWidthStr::width("\u{2460}\u{2461}\u{2462}"), 6); |
| 513 | } |
| 514 | |
| 515 | #[test] |
| 516 | fn slice_text_does_not_split_keycap_sequence() { |
| 517 | let row = "step 1\u{fe0f}\u{20e3} done"; |
| 518 | // The keycap occupies columns [5, 7). Any overlapping selection keeps |
| 519 | // the complete grapheme; no isolated FE0F/U+20E3 mark may escape. |
| 520 | for (start, end) in [(0, 7), (5, 6), (6, 7)] { |
| 521 | let sliced = slice_text(row, start, end); |
| 522 | assert!( |
| 523 | sliced.contains("1\u{fe0f}\u{20e3}"), |
| 524 | "range=({start}, {end}) split keycap: {sliced:?}" |
| 525 | ); |
| 526 | } |
| 527 | } |
| 528 | |
| 529 | #[test] |
| 530 | fn truncate_line_to_width_always_stays_within_budget_with_keycap() { |
| 531 | // Budgets from zero through wide, with and without surrounding text. |
| 532 | let cases = [ |
| 533 | "1\u{fe0f}\u{20e3}", |
| 534 | "A 1\u{fe0f}\u{20e3} B", |
| 535 | "step 2\u{fe0f}\u{20e3} and 3\u{fe0f}\u{20e3} continue", |
| 536 | ]; |
| 537 | for text in &cases { |
| 538 | for budget in 0..=text_display_width(text) + 4 { |
| 539 | let out = truncate_line_to_width(text, budget); |
| 540 | let width = text_display_width(&out); |
| 541 | assert!( |
| 542 | width <= budget, |
| 543 | "budget={budget} text={text:?} -> {out:?} (width={width})" |
| 544 | ); |
| 545 | assert!(!out.ends_with('\u{fe0f}')); |
| 546 | assert!(!out.starts_with('\u{20e3}')); |
| 547 | } |
| 548 | } |
| 549 | } |
| 550 | } |
| 551 |