1use super::{
2 Highlights,
3 fold_map::{self, Chunk, FoldChunks, FoldEdit, FoldPoint, FoldSnapshot},
4};
5
6use language::Point;
7use multi_buffer::MultiBufferSnapshot;
8use std::{cmp, mem, num::NonZeroU32, ops::Range};
9use sum_tree::Bias;
10
11const MAX_EXPANSION_COLUMN: u32 = 256;
12
13// Handles a tab width <= 128
14const SPACES: &[u8; u128::BITS as usize] = &[b' '; _];
15const MAX_TABS: NonZeroU32 = NonZeroU32::new(SPACES.len() as u32).unwrap();
16
17/// Keeps track of hard tabs in a text buffer.
18///
19/// See the [`display_map` module documentation](crate::display_map) for more information.
20pub struct TabMap(TabSnapshot);
21
22impl TabMap {
23 pub fn new(fold_snapshot: FoldSnapshot, tab_size: NonZeroU32) -> (Self, TabSnapshot) {
24 let snapshot = TabSnapshot {
25 fold_snapshot,
26 tab_size: tab_size.min(MAX_TABS),
27 max_expansion_column: MAX_EXPANSION_COLUMN,
28 version: 0,
29 };
30 (Self(snapshot.clone()), snapshot)
31 }
32
33 #[cfg(test)]
34 pub fn set_max_expansion_column(&mut self, column: u32) -> TabSnapshot {
35 self.0.max_expansion_column = column;
36 self.0.clone()
37 }
38
39 pub fn sync(
40 &mut self,
41 fold_snapshot: FoldSnapshot,
42 mut fold_edits: Vec<FoldEdit>,
43 tab_size: NonZeroU32,
44 ) -> (TabSnapshot, Vec<TabEdit>) {
45 let old_snapshot = &mut self.0;
46 let mut new_snapshot = TabSnapshot {
47 fold_snapshot,
48 tab_size: tab_size.min(MAX_TABS),
49 max_expansion_column: old_snapshot.max_expansion_column,
50 version: old_snapshot.version,
51 };
52
53 if old_snapshot.fold_snapshot.version != new_snapshot.fold_snapshot.version {
54 new_snapshot.version += 1;
55 }
56
57 let tab_edits = if old_snapshot.tab_size == new_snapshot.tab_size {
58 // Expand each edit to include the next tab on the same line as the edit,
59 // and any subsequent tabs on that line that moved across the tab expansion
60 // boundary.
61 for fold_edit in &mut fold_edits {
62 let old_end = fold_edit.old.end.to_point(&old_snapshot.fold_snapshot);
63 let old_end_row_successor_offset = cmp::min(
64 FoldPoint::new(old_end.row() + 1, 0),
65 old_snapshot.fold_snapshot.max_point(),
66 )
67 .to_offset(&old_snapshot.fold_snapshot);
68 let new_end = fold_edit.new.end.to_point(&new_snapshot.fold_snapshot);
69
70 let mut offset_from_edit = 0;
71 let mut first_tab_offset = None;
72 let mut last_tab_with_changed_expansion_offset = None;
73 'outer: for chunk in old_snapshot.fold_snapshot.chunks(
74 fold_edit.old.end..old_end_row_successor_offset,
75 false,
76 Highlights::default(),
77 ) {
78 // todo(performance use tabs bitmask)
79 for (ix, _) in chunk.text.match_indices('\t') {
80 let offset_from_edit = offset_from_edit + (ix as u32);
81 if first_tab_offset.is_none() {
82 first_tab_offset = Some(offset_from_edit);
83 }
84
85 let old_column = old_end.column() + offset_from_edit;
86 let new_column = new_end.column() + offset_from_edit;
87 let was_expanded = old_column < old_snapshot.max_expansion_column;
88 let is_expanded = new_column < new_snapshot.max_expansion_column;
89 if was_expanded != is_expanded {
90 last_tab_with_changed_expansion_offset = Some(offset_from_edit);
91 } else if !was_expanded && !is_expanded {
92 break 'outer;
93 }
94 }
95
96 offset_from_edit += chunk.text.len() as u32;
97 if old_end.column() + offset_from_edit >= old_snapshot.max_expansion_column
98 && new_end.column() + offset_from_edit >= new_snapshot.max_expansion_column
99 {
100 break;
101 }
102 }
103
104 if let Some(offset) = last_tab_with_changed_expansion_offset.or(first_tab_offset) {
105 fold_edit.old.end.0 += offset as usize + 1;
106 fold_edit.new.end.0 += offset as usize + 1;
107 }
108 }
109
110 let _old_alloc_ptr = fold_edits.as_ptr();
111 // Combine any edits that overlap due to the expansion.
112 let mut fold_edits = fold_edits.into_iter();
113 if let Some(mut first_edit) = fold_edits.next() {
114 // This code relies on reusing allocations from the Vec<_> - at the time of writing .flatten() prevents them.
115 #[allow(clippy::filter_map_identity)]
116 let mut v: Vec<_> = fold_edits
117 .scan(&mut first_edit, |state, edit| {
118 if state.old.end >= edit.old.start {
119 state.old.end = edit.old.end;
120 state.new.end = edit.new.end;
121 Some(None) // Skip this edit, it's merged
122 } else {
123 let new_state = edit;
124 let result = Some(Some(state.clone())); // Yield the previous edit
125 **state = new_state;
126 result
127 }
128 })
129 .filter_map(|x| x)
130 .collect();
131 v.push(first_edit);
132 debug_assert_eq!(v.as_ptr(), _old_alloc_ptr, "Fold edits were reallocated");
133 v.into_iter()
134 .map(|fold_edit| {
135 let old_start = fold_edit.old.start.to_point(&old_snapshot.fold_snapshot);
136 let old_end = fold_edit.old.end.to_point(&old_snapshot.fold_snapshot);
137 let new_start = fold_edit.new.start.to_point(&new_snapshot.fold_snapshot);
138 let new_end = fold_edit.new.end.to_point(&new_snapshot.fold_snapshot);
139 TabEdit {
140 old: old_snapshot.to_tab_point(old_start)
141 ..old_snapshot.to_tab_point(old_end),
142 new: new_snapshot.to_tab_point(new_start)
143 ..new_snapshot.to_tab_point(new_end),
144 }
145 })
146 .collect()
147 } else {
148 vec![]
149 }
150 } else {
151 new_snapshot.version += 1;
152 vec![TabEdit {
153 old: TabPoint::zero()..old_snapshot.max_point(),
154 new: TabPoint::zero()..new_snapshot.max_point(),
155 }]
156 };
157 *old_snapshot = new_snapshot;
158 (old_snapshot.clone(), tab_edits)
159 }
160}
161
162#[derive(Clone)]
163pub struct TabSnapshot {
164 pub fold_snapshot: FoldSnapshot,
165 pub tab_size: NonZeroU32,
166 pub max_expansion_column: u32,
167 pub version: usize,
168}
169
170impl TabSnapshot {
171 pub fn buffer_snapshot(&self) -> &MultiBufferSnapshot {
172 &self.fold_snapshot.inlay_snapshot.buffer
173 }
174
175 pub fn line_len(&self, row: u32) -> u32 {
176 let max_point = self.max_point();
177 if row < max_point.row() {
178 self.to_tab_point(FoldPoint::new(row, self.fold_snapshot.line_len(row)))
179 .0
180 .column
181 } else {
182 max_point.column()
183 }
184 }
185
186 pub fn text_summary(&self) -> TextSummary {
187 self.text_summary_for_range(TabPoint::zero()..self.max_point())
188 }
189
190 pub fn text_summary_for_range(&self, range: Range<TabPoint>) -> TextSummary {
191 let input_start = self.to_fold_point(range.start, Bias::Left).0;
192 let input_end = self.to_fold_point(range.end, Bias::Right).0;
193 let input_summary = self
194 .fold_snapshot
195 .text_summary_for_range(input_start..input_end);
196
197 let line_end = if range.start.row() == range.end.row() {
198 range.end
199 } else {
200 self.max_point()
201 };
202 let first_line_chars = self
203 .chunks(range.start..line_end, false, Highlights::default())
204 .flat_map(|chunk| chunk.text.chars())
205 .take_while(|&c| c != '\n')
206 .count() as u32;
207
208 let last_line_chars = if range.start.row() == range.end.row() {
209 first_line_chars
210 } else {
211 self.chunks(
212 TabPoint::new(range.end.row(), 0)..range.end,
213 false,
214 Highlights::default(),
215 )
216 .flat_map(|chunk| chunk.text.chars())
217 .count() as u32
218 };
219
220 TextSummary {
221 lines: range.end.0 - range.start.0,
222 first_line_chars,
223 last_line_chars,
224 longest_row: input_summary.longest_row,
225 longest_row_chars: input_summary.longest_row_chars,
226 }
227 }
228
229 pub(crate) fn chunks<'a>(
230 &'a self,
231 range: Range<TabPoint>,
232 language_aware: bool,
233 highlights: Highlights<'a>,
234 ) -> TabChunks<'a> {
235 let (input_start, expanded_char_column, to_next_stop) =
236 self.to_fold_point(range.start, Bias::Left);
237 let input_column = input_start.column();
238 let input_start = input_start.to_offset(&self.fold_snapshot);
239 let input_end = self
240 .to_fold_point(range.end, Bias::Right)
241 .0
242 .to_offset(&self.fold_snapshot);
243 let to_next_stop = if range.start.0 + Point::new(0, to_next_stop) > range.end.0 {
244 range.end.column() - range.start.column()
245 } else {
246 to_next_stop
247 };
248
249 TabChunks {
250 snapshot: self,
251 fold_chunks: self.fold_snapshot.chunks(
252 input_start..input_end,
253 language_aware,
254 highlights,
255 ),
256 input_column,
257 column: expanded_char_column,
258 max_expansion_column: self.max_expansion_column,
259 output_position: range.start.0,
260 max_output_position: range.end.0,
261 tab_size: self.tab_size,
262 chunk: Chunk {
263 text: unsafe { std::str::from_utf8_unchecked(&SPACES[..to_next_stop as usize]) },
264 is_tab: true,
265 ..Default::default()
266 },
267 inside_leading_tab: to_next_stop > 0,
268 }
269 }
270
271 pub fn rows(&self, row: u32) -> fold_map::FoldRows<'_> {
272 self.fold_snapshot.row_infos(row)
273 }
274
275 #[cfg(test)]
276 pub fn text(&self) -> String {
277 self.chunks(
278 TabPoint::zero()..self.max_point(),
279 false,
280 Highlights::default(),
281 )
282 .map(|chunk| chunk.text)
283 .collect()
284 }
285
286 pub fn max_point(&self) -> TabPoint {
287 self.to_tab_point(self.fold_snapshot.max_point())
288 }
289
290 pub fn clip_point(&self, point: TabPoint, bias: Bias) -> TabPoint {
291 self.to_tab_point(
292 self.fold_snapshot
293 .clip_point(self.to_fold_point(point, bias).0, bias),
294 )
295 }
296
297 pub fn to_tab_point(&self, input: FoldPoint) -> TabPoint {
298 let chunks = self.fold_snapshot.chunks_at(FoldPoint::new(input.row(), 0));
299 let tab_cursor = TabStopCursor::new(chunks);
300 let expanded = self.expand_tabs(tab_cursor, input.column());
301 TabPoint::new(input.row(), expanded)
302 }
303
304 pub fn to_fold_point(&self, output: TabPoint, bias: Bias) -> (FoldPoint, u32, u32) {
305 let chunks = self
306 .fold_snapshot
307 .chunks_at(FoldPoint::new(output.row(), 0));
308
309 let tab_cursor = TabStopCursor::new(chunks);
310 let expanded = output.column();
311 let (collapsed, expanded_char_column, to_next_stop) =
312 self.collapse_tabs(tab_cursor, expanded, bias);
313
314 (
315 FoldPoint::new(output.row(), collapsed),
316 expanded_char_column,
317 to_next_stop,
318 )
319 }
320
321 pub fn make_tab_point(&self, point: Point, bias: Bias) -> TabPoint {
322 let inlay_point = self.fold_snapshot.inlay_snapshot.to_inlay_point(point);
323 let fold_point = self.fold_snapshot.to_fold_point(inlay_point, bias);
324 self.to_tab_point(fold_point)
325 }
326
327 pub fn to_point(&self, point: TabPoint, bias: Bias) -> Point {
328 let fold_point = self.to_fold_point(point, bias).0;
329 let inlay_point = fold_point.to_inlay_point(&self.fold_snapshot);
330 self.fold_snapshot
331 .inlay_snapshot
332 .to_buffer_point(inlay_point)
333 }
334
335 fn expand_tabs<'a, I>(&self, mut cursor: TabStopCursor<'a, I>, column: u32) -> u32
336 where
337 I: Iterator<Item = Chunk<'a>>,
338 {
339 let tab_size = self.tab_size.get();
340
341 let end_column = column.min(self.max_expansion_column);
342 let mut seek_target = end_column;
343 let mut tab_count = 0;
344 let mut expanded_tab_len = 0;
345
346 while let Some(tab_stop) = cursor.seek(seek_target) {
347 let expanded_chars_old = tab_stop.char_offset + expanded_tab_len - tab_count;
348 let tab_len = tab_size - ((expanded_chars_old - 1) % tab_size);
349 tab_count += 1;
350 expanded_tab_len += tab_len;
351
352 seek_target = end_column - cursor.byte_offset;
353 }
354
355 let left_over_char_bytes = if !cursor.is_char_boundary() {
356 cursor.bytes_until_next_char().unwrap_or(0) as u32
357 } else {
358 0
359 };
360
361 let collapsed_bytes = cursor.byte_offset() + left_over_char_bytes;
362 let expanded_bytes =
363 cursor.byte_offset() + expanded_tab_len - tab_count + left_over_char_bytes;
364
365 expanded_bytes + column.saturating_sub(collapsed_bytes)
366 }
367
368 fn collapse_tabs<'a, I>(
369 &self,
370 mut cursor: TabStopCursor<'a, I>,
371 column: u32,
372 bias: Bias,
373 ) -> (u32, u32, u32)
374 where
375 I: Iterator<Item = Chunk<'a>>,
376 {
377 let tab_size = self.tab_size.get();
378 let mut collapsed_column = column;
379 let mut seek_target = column.min(self.max_expansion_column);
380 let mut tab_count = 0;
381 let mut expanded_tab_len = 0;
382
383 while let Some(tab_stop) = cursor.seek(seek_target) {
384 // Calculate how much we want to expand this tab stop (into spaces)
385 let expanded_chars_old = tab_stop.char_offset + expanded_tab_len - tab_count;
386 let tab_len = tab_size - ((expanded_chars_old - 1) % tab_size);
387 // Increment tab count
388 tab_count += 1;
389 // The count of how many spaces we've added to this line in place of tab bytes
390 expanded_tab_len += tab_len;
391
392 // The count of bytes at this point in the iteration while considering tab_count and previous expansions
393 let expanded_bytes = tab_stop.byte_offset + expanded_tab_len - tab_count;
394
395 // Did we expand past the search target?
396 if expanded_bytes > column {
397 let mut expanded_chars = tab_stop.char_offset + expanded_tab_len - tab_count;
398 // We expanded past the search target, so need to account for the offshoot
399 expanded_chars -= expanded_bytes - column;
400 return match bias {
401 Bias::Left => (
402 cursor.byte_offset() - 1,
403 expanded_chars,
404 expanded_bytes - column,
405 ),
406 Bias::Right => (cursor.byte_offset(), expanded_chars, 0),
407 };
408 } else {
409 // otherwise we only want to move the cursor collapse column forward
410 collapsed_column = collapsed_column - tab_len + 1;
411 seek_target = (collapsed_column - cursor.byte_offset)
412 .min(self.max_expansion_column - cursor.byte_offset);
413 }
414 }
415
416 let collapsed_bytes = cursor.byte_offset();
417 let expanded_bytes = cursor.byte_offset() + expanded_tab_len - tab_count;
418 let expanded_chars = cursor.char_offset() + expanded_tab_len - tab_count;
419 (
420 collapsed_bytes + column.saturating_sub(expanded_bytes),
421 expanded_chars,
422 0,
423 )
424 }
425}
426
427#[derive(Copy, Clone, Debug, Default, Eq, Ord, PartialOrd, PartialEq)]
428pub struct TabPoint(pub Point);
429
430impl TabPoint {
431 pub fn new(row: u32, column: u32) -> Self {
432 Self(Point::new(row, column))
433 }
434
435 pub fn zero() -> Self {
436 Self::new(0, 0)
437 }
438
439 pub fn row(self) -> u32 {
440 self.0.row
441 }
442
443 pub fn column(self) -> u32 {
444 self.0.column
445 }
446}
447
448impl From<Point> for TabPoint {
449 fn from(point: Point) -> Self {
450 Self(point)
451 }
452}
453
454pub type TabEdit = text::Edit<TabPoint>;
455
456#[derive(Clone, Debug, Default, Eq, PartialEq)]
457pub struct TextSummary {
458 pub lines: Point,
459 pub first_line_chars: u32,
460 pub last_line_chars: u32,
461 pub longest_row: u32,
462 pub longest_row_chars: u32,
463}
464
465impl<'a> From<&'a str> for TextSummary {
466 fn from(text: &'a str) -> Self {
467 let sum = text::TextSummary::from(text);
468
469 TextSummary {
470 lines: sum.lines,
471 first_line_chars: sum.first_line_chars,
472 last_line_chars: sum.last_line_chars,
473 longest_row: sum.longest_row,
474 longest_row_chars: sum.longest_row_chars,
475 }
476 }
477}
478
479impl<'a> std::ops::AddAssign<&'a Self> for TextSummary {
480 fn add_assign(&mut self, other: &'a Self) {
481 let joined_chars = self.last_line_chars + other.first_line_chars;
482 if joined_chars > self.longest_row_chars {
483 self.longest_row = self.lines.row;
484 self.longest_row_chars = joined_chars;
485 }
486 if other.longest_row_chars > self.longest_row_chars {
487 self.longest_row = self.lines.row + other.longest_row;
488 self.longest_row_chars = other.longest_row_chars;
489 }
490
491 if self.lines.row == 0 {
492 self.first_line_chars += other.first_line_chars;
493 }
494
495 if other.lines.row == 0 {
496 self.last_line_chars += other.first_line_chars;
497 } else {
498 self.last_line_chars = other.last_line_chars;
499 }
500
501 self.lines += &other.lines;
502 }
503}
504
505pub struct TabChunks<'a> {
506 snapshot: &'a TabSnapshot,
507 max_expansion_column: u32,
508 max_output_position: Point,
509 tab_size: NonZeroU32,
510 // region: iteration state
511 fold_chunks: FoldChunks<'a>,
512 chunk: Chunk<'a>,
513 column: u32,
514 output_position: Point,
515 input_column: u32,
516 inside_leading_tab: bool,
517 // endregion: iteration state
518}
519
520impl TabChunks<'_> {
521 pub(crate) fn seek(&mut self, range: Range<TabPoint>) {
522 let (input_start, expanded_char_column, to_next_stop) =
523 self.snapshot.to_fold_point(range.start, Bias::Left);
524 let input_column = input_start.column();
525 let input_start = input_start.to_offset(&self.snapshot.fold_snapshot);
526 let input_end = self
527 .snapshot
528 .to_fold_point(range.end, Bias::Right)
529 .0
530 .to_offset(&self.snapshot.fold_snapshot);
531 let to_next_stop = if range.start.0 + Point::new(0, to_next_stop) > range.end.0 {
532 range.end.column() - range.start.column()
533 } else {
534 to_next_stop
535 };
536
537 self.fold_chunks.seek(input_start..input_end);
538 self.input_column = input_column;
539 self.column = expanded_char_column;
540 self.output_position = range.start.0;
541 self.max_output_position = range.end.0;
542 self.chunk = Chunk {
543 text: unsafe { std::str::from_utf8_unchecked(&SPACES[..to_next_stop as usize]) },
544 is_tab: true,
545 chars: 1u128.unbounded_shl(to_next_stop) - 1,
546 ..Default::default()
547 };
548 self.inside_leading_tab = to_next_stop > 0;
549 }
550}
551
552impl<'a> Iterator for TabChunks<'a> {
553 type Item = Chunk<'a>;
554
555 fn next(&mut self) -> Option<Self::Item> {
556 if self.chunk.text.is_empty() {
557 if let Some(chunk) = self.fold_chunks.next() {
558 self.chunk = chunk;
559 if self.inside_leading_tab {
560 self.chunk.text = &self.chunk.text[1..];
561 self.inside_leading_tab = false;
562 self.input_column += 1;
563 }
564 } else {
565 return None;
566 }
567 }
568
569 //todo(improve performance by using tab cursor)
570 for (ix, c) in self.chunk.text.char_indices() {
571 match c {
572 '\t' => {
573 if ix > 0 {
574 let (prefix, suffix) = self.chunk.text.split_at(ix);
575
576 let (chars, tabs) = if ix == 128 {
577 let output = (self.chunk.chars, self.chunk.tabs);
578 self.chunk.chars = 0;
579 self.chunk.tabs = 0;
580 output
581 } else {
582 let mask = (1 << ix) - 1;
583 let output = (self.chunk.chars & mask, self.chunk.tabs & mask);
584 self.chunk.chars = self.chunk.chars >> ix;
585 self.chunk.tabs = self.chunk.tabs >> ix;
586 output
587 };
588
589 self.chunk.text = suffix;
590 return Some(Chunk {
591 text: prefix,
592 chars,
593 tabs,
594 ..self.chunk.clone()
595 });
596 } else {
597 self.chunk.text = &self.chunk.text[1..];
598 self.chunk.tabs >>= 1;
599 self.chunk.chars >>= 1;
600 let tab_size = if self.input_column < self.max_expansion_column {
601 self.tab_size.get()
602 } else {
603 1
604 };
605 let mut len = tab_size - self.column % tab_size;
606 let next_output_position = cmp::min(
607 self.output_position + Point::new(0, len),
608 self.max_output_position,
609 );
610 len = next_output_position.column - self.output_position.column;
611 self.column += len;
612 self.input_column += 1;
613 self.output_position = next_output_position;
614 return Some(Chunk {
615 text: unsafe { std::str::from_utf8_unchecked(&SPACES[..len as usize]) },
616 is_tab: true,
617 chars: 1u128.unbounded_shl(len) - 1,
618 tabs: 0,
619 ..self.chunk.clone()
620 });
621 }
622 }
623 '\n' => {
624 self.column = 0;
625 self.input_column = 0;
626 self.output_position += Point::new(1, 0);
627 }
628 _ => {
629 self.column += 1;
630 if !self.inside_leading_tab {
631 self.input_column += c.len_utf8() as u32;
632 }
633 self.output_position.column += c.len_utf8() as u32;
634 }
635 }
636 }
637
638 Some(mem::take(&mut self.chunk))
639 }
640}
641
642#[cfg(test)]
643mod tests {
644 use super::*;
645 use crate::{
646 MultiBuffer,
647 display_map::{
648 fold_map::{FoldMap, FoldOffset},
649 inlay_map::InlayMap,
650 },
651 };
652 use rand::{Rng, prelude::StdRng};
653 use util;
654
655 impl TabSnapshot {
656 fn expected_collapse_tabs(
657 &self,
658 chars: impl Iterator<Item = char>,
659 column: u32,
660 bias: Bias,
661 ) -> (u32, u32, u32) {
662 let tab_size = self.tab_size.get();
663
664 let mut expanded_bytes = 0;
665 let mut expanded_chars = 0;
666 let mut collapsed_bytes = 0;
667 for c in chars {
668 if expanded_bytes >= column {
669 break;
670 }
671 if collapsed_bytes >= self.max_expansion_column {
672 break;
673 }
674
675 if c == '\t' {
676 let tab_len = tab_size - (expanded_chars % tab_size);
677 expanded_chars += tab_len;
678 expanded_bytes += tab_len;
679 if expanded_bytes > column {
680 expanded_chars -= expanded_bytes - column;
681 return match bias {
682 Bias::Left => {
683 (collapsed_bytes, expanded_chars, expanded_bytes - column)
684 }
685 Bias::Right => (collapsed_bytes + 1, expanded_chars, 0),
686 };
687 }
688 } else {
689 expanded_chars += 1;
690 expanded_bytes += c.len_utf8() as u32;
691 }
692
693 if expanded_bytes > column && matches!(bias, Bias::Left) {
694 expanded_chars -= 1;
695 break;
696 }
697
698 collapsed_bytes += c.len_utf8() as u32;
699 }
700
701 (
702 collapsed_bytes + column.saturating_sub(expanded_bytes),
703 expanded_chars,
704 0,
705 )
706 }
707
708 pub fn expected_to_tab_point(&self, input: FoldPoint) -> TabPoint {
709 let chars = self.fold_snapshot.chars_at(FoldPoint::new(input.row(), 0));
710 let expanded = self.expected_expand_tabs(chars, input.column());
711 TabPoint::new(input.row(), expanded)
712 }
713
714 fn expected_expand_tabs(&self, chars: impl Iterator<Item = char>, column: u32) -> u32 {
715 let tab_size = self.tab_size.get();
716
717 let mut expanded_chars = 0;
718 let mut expanded_bytes = 0;
719 let mut collapsed_bytes = 0;
720 let end_column = column.min(self.max_expansion_column);
721 for c in chars {
722 if collapsed_bytes >= end_column {
723 break;
724 }
725 if c == '\t' {
726 let tab_len = tab_size - expanded_chars % tab_size;
727 expanded_bytes += tab_len;
728 expanded_chars += tab_len;
729 } else {
730 expanded_bytes += c.len_utf8() as u32;
731 expanded_chars += 1;
732 }
733 collapsed_bytes += c.len_utf8() as u32;
734 }
735
736 expanded_bytes + column.saturating_sub(collapsed_bytes)
737 }
738
739 fn expected_to_fold_point(&self, output: TabPoint, bias: Bias) -> (FoldPoint, u32, u32) {
740 let chars = self.fold_snapshot.chars_at(FoldPoint::new(output.row(), 0));
741 let expanded = output.column();
742 let (collapsed, expanded_char_column, to_next_stop) =
743 self.expected_collapse_tabs(chars, expanded, bias);
744 (
745 FoldPoint::new(output.row(), collapsed),
746 expanded_char_column,
747 to_next_stop,
748 )
749 }
750 }
751
752 #[gpui::test]
753 fn test_expand_tabs(cx: &mut gpui::App) {
754 let test_values = [
755 ("κg🏀 f\nwo🏀❌by🍐❎β🍗c\tβ❎ \ncλ🎉", 17),
756 (" \twςe", 4),
757 ("fε", 1),
758 ("i❎\t", 3),
759 ];
760 let buffer = MultiBuffer::build_simple("", cx);
761 let buffer_snapshot = buffer.read(cx).snapshot(cx);
762 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot);
763 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
764 let (_, tab_snapshot) = TabMap::new(fold_snapshot, 4.try_into().unwrap());
765
766 for (text, column) in test_values {
767 let mut tabs = 0u128;
768 let mut chars = 0u128;
769 for (idx, c) in text.char_indices() {
770 if c == '\t' {
771 tabs |= 1 << idx;
772 }
773 chars |= 1 << idx;
774 }
775
776 let chunks = [Chunk {
777 text,
778 tabs,
779 chars,
780 ..Default::default()
781 }];
782
783 let cursor = TabStopCursor::new(chunks);
784
785 assert_eq!(
786 tab_snapshot.expected_expand_tabs(text.chars(), column),
787 tab_snapshot.expand_tabs(cursor, column)
788 );
789 }
790 }
791
792 #[gpui::test]
793 fn test_collapse_tabs(cx: &mut gpui::App) {
794 let input = "A\tBC\tDEF\tG\tHI\tJ\tK\tL\tM";
795
796 let buffer = MultiBuffer::build_simple(input, cx);
797 let buffer_snapshot = buffer.read(cx).snapshot(cx);
798 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot);
799 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
800 let (_, tab_snapshot) = TabMap::new(fold_snapshot, 4.try_into().unwrap());
801
802 for (ix, _) in input.char_indices() {
803 let range = TabPoint::new(0, ix as u32)..tab_snapshot.max_point();
804
805 assert_eq!(
806 tab_snapshot.expected_to_fold_point(range.start, Bias::Left),
807 tab_snapshot.to_fold_point(range.start, Bias::Left),
808 "Failed with tab_point at column {ix}"
809 );
810 assert_eq!(
811 tab_snapshot.expected_to_fold_point(range.start, Bias::Right),
812 tab_snapshot.to_fold_point(range.start, Bias::Right),
813 "Failed with tab_point at column {ix}"
814 );
815
816 assert_eq!(
817 tab_snapshot.expected_to_fold_point(range.end, Bias::Left),
818 tab_snapshot.to_fold_point(range.end, Bias::Left),
819 "Failed with tab_point at column {ix}"
820 );
821 assert_eq!(
822 tab_snapshot.expected_to_fold_point(range.end, Bias::Right),
823 tab_snapshot.to_fold_point(range.end, Bias::Right),
824 "Failed with tab_point at column {ix}"
825 );
826 }
827 }
828
829 #[gpui::test]
830 fn test_to_fold_point_panic_reproduction(cx: &mut gpui::App) {
831 // This test reproduces a specific panic where to_fold_point returns incorrect results
832 let _text = "use macro_rules_attribute::apply;\nuse serde_json::Value;\nuse smol::{\n io::AsyncReadExt,\n process::{Command, Stdio},\n};\nuse smol_macros::main;\nuse std::io;\n\nfn test_random() {\n // Generate a random value\n let random_value = std::time::SystemTime::now()\n .duration_since(std::time::UNIX_EPOCH)\n .unwrap()\n .as_secs()\n % 100;\n\n // Create some complex nested data structures\n let mut vector = Vec::new();\n for i in 0..random_value {\n vector.push(i);\n }\n ";
833
834 let text = "γ\tw⭐\n🍐🍗 \t";
835 let buffer = MultiBuffer::build_simple(text, cx);
836 let buffer_snapshot = buffer.read(cx).snapshot(cx);
837 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot);
838 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
839 let (_, tab_snapshot) = TabMap::new(fold_snapshot, 4.try_into().unwrap());
840
841 // This should panic with the expected vs actual mismatch
842 let tab_point = TabPoint::new(0, 9);
843 let result = tab_snapshot.to_fold_point(tab_point, Bias::Left);
844 let expected = tab_snapshot.expected_to_fold_point(tab_point, Bias::Left);
845
846 assert_eq!(result, expected);
847 }
848
849 #[gpui::test(iterations = 100)]
850 fn test_collapse_tabs_random(cx: &mut gpui::App, mut rng: StdRng) {
851 // Generate random input string with up to 200 characters including tabs
852 // to stay within the MAX_EXPANSION_COLUMN limit of 256
853 let len = rng.random_range(0..=2048);
854 let tab_size = NonZeroU32::new(rng.random_range(1..=4)).unwrap();
855 let mut input = String::with_capacity(len);
856
857 for _ in 0..len {
858 if rng.random_bool(0.1) {
859 // 10% chance of inserting a tab
860 input.push('\t');
861 } else {
862 // 90% chance of inserting a random ASCII character (excluding tab, newline, carriage return)
863 let ch = loop {
864 let ascii_code = rng.random_range(32..=126); // printable ASCII range
865 let ch = ascii_code as u8 as char;
866 if ch != '\t' {
867 break ch;
868 }
869 };
870 input.push(ch);
871 }
872 }
873
874 let buffer = MultiBuffer::build_simple(&input, cx);
875 let buffer_snapshot = buffer.read(cx).snapshot(cx);
876 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot);
877 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
878 let (_, mut tab_snapshot) = TabMap::new(fold_snapshot, 4.try_into().unwrap());
879 tab_snapshot.max_expansion_column = rng.random_range(0..323);
880 tab_snapshot.tab_size = tab_size;
881
882 for (ix, _) in input.char_indices() {
883 let range = TabPoint::new(0, ix as u32)..tab_snapshot.max_point();
884
885 assert_eq!(
886 tab_snapshot.expected_to_fold_point(range.start, Bias::Left),
887 tab_snapshot.to_fold_point(range.start, Bias::Left),
888 "Failed with input: {}, with idx: {ix}",
889 input
890 );
891 assert_eq!(
892 tab_snapshot.expected_to_fold_point(range.start, Bias::Right),
893 tab_snapshot.to_fold_point(range.start, Bias::Right),
894 "Failed with input: {}, with idx: {ix}",
895 input
896 );
897
898 assert_eq!(
899 tab_snapshot.expected_to_fold_point(range.end, Bias::Left),
900 tab_snapshot.to_fold_point(range.end, Bias::Left),
901 "Failed with input: {}, with idx: {ix}",
902 input
903 );
904 assert_eq!(
905 tab_snapshot.expected_to_fold_point(range.end, Bias::Right),
906 tab_snapshot.to_fold_point(range.end, Bias::Right),
907 "Failed with input: {}, with idx: {ix}",
908 input
909 );
910 }
911 }
912
913 #[gpui::test]
914 fn test_long_lines(cx: &mut gpui::App) {
915 let max_expansion_column = 12;
916 let input = "A\tBC\tDEF\tG\tHI\tJ\tK\tL\tM";
917 let output = "A BC DEF G HI J K L M";
918
919 let buffer = MultiBuffer::build_simple(input, cx);
920 let buffer_snapshot = buffer.read(cx).snapshot(cx);
921 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot);
922 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
923 let (_, mut tab_snapshot) = TabMap::new(fold_snapshot, 4.try_into().unwrap());
924
925 tab_snapshot.max_expansion_column = max_expansion_column;
926 assert_eq!(tab_snapshot.text(), output);
927
928 for (ix, c) in input.char_indices() {
929 assert_eq!(
930 tab_snapshot
931 .chunks(
932 TabPoint::new(0, ix as u32)..tab_snapshot.max_point(),
933 false,
934 Highlights::default(),
935 )
936 .map(|c| c.text)
937 .collect::<String>(),
938 &output[ix..],
939 "text from index {ix}"
940 );
941
942 if c != '\t' {
943 let input_point = Point::new(0, ix as u32);
944 let output_point = Point::new(0, output.find(c).unwrap() as u32);
945 assert_eq!(
946 tab_snapshot.to_tab_point(FoldPoint(input_point)),
947 TabPoint(output_point),
948 "to_tab_point({input_point:?})"
949 );
950 assert_eq!(
951 tab_snapshot
952 .to_fold_point(TabPoint(output_point), Bias::Left)
953 .0,
954 FoldPoint(input_point),
955 "to_fold_point({output_point:?})"
956 );
957 }
958 }
959 }
960
961 #[gpui::test]
962 fn test_long_lines_with_character_spanning_max_expansion_column(cx: &mut gpui::App) {
963 let max_expansion_column = 8;
964 let input = "abcdefg⋯hij";
965
966 let buffer = MultiBuffer::build_simple(input, cx);
967 let buffer_snapshot = buffer.read(cx).snapshot(cx);
968 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot);
969 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
970 let (_, mut tab_snapshot) = TabMap::new(fold_snapshot, 4.try_into().unwrap());
971
972 tab_snapshot.max_expansion_column = max_expansion_column;
973 assert_eq!(tab_snapshot.text(), input);
974 }
975
976 #[gpui::test]
977 fn test_marking_tabs(cx: &mut gpui::App) {
978 let input = "\t \thello";
979
980 let buffer = MultiBuffer::build_simple(input, cx);
981 let buffer_snapshot = buffer.read(cx).snapshot(cx);
982 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot);
983 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
984 let (_, tab_snapshot) = TabMap::new(fold_snapshot, 4.try_into().unwrap());
985
986 assert_eq!(
987 chunks(&tab_snapshot, TabPoint::zero()),
988 vec![
989 (" ".to_string(), true),
990 (" ".to_string(), false),
991 (" ".to_string(), true),
992 ("hello".to_string(), false),
993 ]
994 );
995 assert_eq!(
996 chunks(&tab_snapshot, TabPoint::new(0, 2)),
997 vec![
998 (" ".to_string(), true),
999 (" ".to_string(), false),
1000 (" ".to_string(), true),
1001 ("hello".to_string(), false),
1002 ]
1003 );
1004
1005 fn chunks(snapshot: &TabSnapshot, start: TabPoint) -> Vec<(String, bool)> {
1006 let mut chunks = Vec::new();
1007 let mut was_tab = false;
1008 let mut text = String::new();
1009 for chunk in snapshot.chunks(start..snapshot.max_point(), false, Highlights::default())
1010 {
1011 if chunk.is_tab != was_tab {
1012 if !text.is_empty() {
1013 chunks.push((mem::take(&mut text), was_tab));
1014 }
1015 was_tab = chunk.is_tab;
1016 }
1017 text.push_str(chunk.text);
1018 }
1019
1020 if !text.is_empty() {
1021 chunks.push((text, was_tab));
1022 }
1023 chunks
1024 }
1025 }
1026
1027 #[gpui::test(iterations = 100)]
1028 fn test_random_tabs(cx: &mut gpui::App, mut rng: StdRng) {
1029 let tab_size = NonZeroU32::new(rng.random_range(1..=4)).unwrap();
1030 let len = rng.random_range(0..30);
1031 let buffer = if rng.random() {
1032 let text = util::RandomCharIter::new(&mut rng)
1033 .take(len)
1034 .collect::<String>();
1035 MultiBuffer::build_simple(&text, cx)
1036 } else {
1037 MultiBuffer::build_random(&mut rng, cx)
1038 };
1039 let buffer_snapshot = buffer.read(cx).snapshot(cx);
1040 log::info!("Buffer text: {:?}", buffer_snapshot.text());
1041
1042 let (mut inlay_map, inlay_snapshot) = InlayMap::new(buffer_snapshot);
1043 log::info!("InlayMap text: {:?}", inlay_snapshot.text());
1044 let (mut fold_map, _) = FoldMap::new(inlay_snapshot.clone());
1045 fold_map.randomly_mutate(&mut rng);
1046 let (fold_snapshot, _) = fold_map.read(inlay_snapshot, vec![]);
1047 log::info!("FoldMap text: {:?}", fold_snapshot.text());
1048 let (inlay_snapshot, _) = inlay_map.randomly_mutate(&mut 0, &mut rng);
1049 log::info!("InlayMap text: {:?}", inlay_snapshot.text());
1050
1051 let (mut tab_map, _) = TabMap::new(fold_snapshot, tab_size);
1052 let tabs_snapshot = tab_map.set_max_expansion_column(32);
1053
1054 let text = text::Rope::from(tabs_snapshot.text().as_str());
1055 log::info!(
1056 "TabMap text (tab size: {}): {:?}",
1057 tab_size,
1058 tabs_snapshot.text(),
1059 );
1060
1061 for _ in 0..5 {
1062 let end_row = rng.random_range(0..=text.max_point().row);
1063 let end_column = rng.random_range(0..=text.line_len(end_row));
1064 let mut end = TabPoint(text.clip_point(Point::new(end_row, end_column), Bias::Right));
1065 let start_row = rng.random_range(0..=text.max_point().row);
1066 let start_column = rng.random_range(0..=text.line_len(start_row));
1067 let mut start =
1068 TabPoint(text.clip_point(Point::new(start_row, start_column), Bias::Left));
1069 if start > end {
1070 mem::swap(&mut start, &mut end);
1071 }
1072
1073 let expected_text = text
1074 .chunks_in_range(text.point_to_offset(start.0)..text.point_to_offset(end.0))
1075 .collect::<String>();
1076 let expected_summary = TextSummary::from(expected_text.as_str());
1077 assert_eq!(
1078 tabs_snapshot
1079 .chunks(start..end, false, Highlights::default())
1080 .map(|c| c.text)
1081 .collect::<String>(),
1082 expected_text,
1083 "chunks({:?}..{:?})",
1084 start,
1085 end
1086 );
1087
1088 let mut actual_summary = tabs_snapshot.text_summary_for_range(start..end);
1089 if tab_size.get() > 1 && inlay_snapshot.text().contains('\t') {
1090 actual_summary.longest_row = expected_summary.longest_row;
1091 actual_summary.longest_row_chars = expected_summary.longest_row_chars;
1092 }
1093 assert_eq!(actual_summary, expected_summary);
1094 }
1095
1096 for row in 0..=text.max_point().row {
1097 assert_eq!(
1098 tabs_snapshot.line_len(row),
1099 text.line_len(row),
1100 "line_len({row})"
1101 );
1102 }
1103 }
1104
1105 #[gpui::test(iterations = 100)]
1106 fn test_to_tab_point_random(cx: &mut gpui::App, mut rng: StdRng) {
1107 let tab_size = NonZeroU32::new(rng.random_range(1..=16)).unwrap();
1108 let len = rng.random_range(0..=2000);
1109
1110 // Generate random text using RandomCharIter
1111 let text = util::RandomCharIter::new(&mut rng)
1112 .take(len)
1113 .collect::<String>();
1114
1115 // Create buffer and tab map
1116 let buffer = MultiBuffer::build_simple(&text, cx);
1117 let buffer_snapshot = buffer.read(cx).snapshot(cx);
1118 let (mut inlay_map, inlay_snapshot) = InlayMap::new(buffer_snapshot);
1119 let (mut fold_map, fold_snapshot) = FoldMap::new(inlay_snapshot);
1120 let (mut tab_map, _) = TabMap::new(fold_snapshot, tab_size);
1121
1122 let mut next_inlay_id = 0;
1123 let (inlay_snapshot, inlay_edits) = inlay_map.randomly_mutate(&mut next_inlay_id, &mut rng);
1124 let (fold_snapshot, fold_edits) = fold_map.read(inlay_snapshot, inlay_edits);
1125 let max_fold_point = fold_snapshot.max_point();
1126 let (mut tab_snapshot, _) = tab_map.sync(fold_snapshot.clone(), fold_edits, tab_size);
1127
1128 // Test random fold points
1129 for _ in 0..50 {
1130 tab_snapshot.max_expansion_column = rng.random_range(0..=256);
1131 // Generate random fold point
1132 let row = rng.random_range(0..=max_fold_point.row());
1133 let max_column = if row < max_fold_point.row() {
1134 fold_snapshot.line_len(row)
1135 } else {
1136 max_fold_point.column()
1137 };
1138 let column = rng.random_range(0..=max_column + 10);
1139 let fold_point = FoldPoint::new(row, column);
1140
1141 let actual = tab_snapshot.to_tab_point(fold_point);
1142 let expected = tab_snapshot.expected_to_tab_point(fold_point);
1143
1144 assert_eq!(
1145 actual, expected,
1146 "to_tab_point mismatch for fold_point {:?} in text {:?}",
1147 fold_point, text
1148 );
1149 }
1150 }
1151
1152 #[gpui::test]
1153 fn test_tab_stop_cursor_utf8(cx: &mut gpui::App) {
1154 let text = "\tfoo\tbarbarbar\t\tbaz\n";
1155 let buffer = MultiBuffer::build_simple(text, cx);
1156 let buffer_snapshot = buffer.read(cx).snapshot(cx);
1157 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot);
1158 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
1159 let chunks = fold_snapshot.chunks(
1160 FoldOffset(0)..fold_snapshot.len(),
1161 false,
1162 Default::default(),
1163 );
1164 let mut cursor = TabStopCursor::new(chunks);
1165 assert!(cursor.seek(0).is_none());
1166 let mut tab_stops = Vec::new();
1167
1168 let mut all_tab_stops = Vec::new();
1169 let mut byte_offset = 0;
1170 for (offset, ch) in buffer.read(cx).snapshot(cx).text().char_indices() {
1171 byte_offset += ch.len_utf8() as u32;
1172
1173 if ch == '\t' {
1174 all_tab_stops.push(TabStop {
1175 byte_offset,
1176 char_offset: offset as u32 + 1,
1177 });
1178 }
1179 }
1180
1181 while let Some(tab_stop) = cursor.seek(u32::MAX) {
1182 tab_stops.push(tab_stop);
1183 }
1184 pretty_assertions::assert_eq!(tab_stops.as_slice(), all_tab_stops.as_slice(),);
1185
1186 assert_eq!(cursor.byte_offset(), byte_offset);
1187 }
1188
1189 #[gpui::test]
1190 fn test_tab_stop_with_end_range_utf8(cx: &mut gpui::App) {
1191 let input = "A\tBC\t"; // DEF\tG\tHI\tJ\tK\tL\tM
1192
1193 let buffer = MultiBuffer::build_simple(input, cx);
1194 let buffer_snapshot = buffer.read(cx).snapshot(cx);
1195 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot);
1196 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
1197
1198 let chunks = fold_snapshot.chunks_at(FoldPoint::new(0, 0));
1199 let mut cursor = TabStopCursor::new(chunks);
1200
1201 let mut actual_tab_stops = Vec::new();
1202
1203 let mut expected_tab_stops = Vec::new();
1204 let mut byte_offset = 0;
1205 for (offset, ch) in buffer.read(cx).snapshot(cx).text().char_indices() {
1206 byte_offset += ch.len_utf8() as u32;
1207
1208 if ch == '\t' {
1209 expected_tab_stops.push(TabStop {
1210 byte_offset,
1211 char_offset: offset as u32 + 1,
1212 });
1213 }
1214 }
1215
1216 while let Some(tab_stop) = cursor.seek(u32::MAX) {
1217 actual_tab_stops.push(tab_stop);
1218 }
1219 pretty_assertions::assert_eq!(actual_tab_stops.as_slice(), expected_tab_stops.as_slice(),);
1220
1221 assert_eq!(cursor.byte_offset(), byte_offset);
1222 }
1223
1224 #[gpui::test(iterations = 100)]
1225 fn test_tab_stop_cursor_random_utf8(cx: &mut gpui::App, mut rng: StdRng) {
1226 // Generate random input string with up to 512 characters including tabs
1227 let len = rng.random_range(0..=2048);
1228 let mut input = String::with_capacity(len);
1229
1230 let mut skip_tabs = rng.random_bool(0.10);
1231 for idx in 0..len {
1232 if idx % 128 == 0 {
1233 skip_tabs = rng.random_bool(0.10);
1234 }
1235
1236 if rng.random_bool(0.15) && !skip_tabs {
1237 input.push('\t');
1238 } else {
1239 let ch = loop {
1240 let ascii_code = rng.random_range(32..=126); // printable ASCII range
1241 let ch = ascii_code as u8 as char;
1242 if ch != '\t' {
1243 break ch;
1244 }
1245 };
1246 input.push(ch);
1247 }
1248 }
1249
1250 // Build the buffer and create cursor
1251 let buffer = MultiBuffer::build_simple(&input, cx);
1252 let buffer_snapshot = buffer.read(cx).snapshot(cx);
1253 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot.clone());
1254 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
1255
1256 // First, collect all expected tab positions
1257 let mut all_tab_stops = Vec::new();
1258 let mut byte_offset = 1;
1259 let mut char_offset = 1;
1260 for ch in buffer_snapshot.text().chars() {
1261 if ch == '\t' {
1262 all_tab_stops.push(TabStop {
1263 byte_offset,
1264 char_offset,
1265 });
1266 }
1267 byte_offset += ch.len_utf8() as u32;
1268 char_offset += 1;
1269 }
1270
1271 // Test with various distances
1272 let distances = vec![1, 5, 10, 50, 100, u32::MAX];
1273 // let distances = vec![150];
1274
1275 for distance in distances {
1276 let chunks = fold_snapshot.chunks_at(FoldPoint::new(0, 0));
1277 let mut cursor = TabStopCursor::new(chunks);
1278
1279 let mut found_tab_stops = Vec::new();
1280 let mut position = distance;
1281 while let Some(tab_stop) = cursor.seek(position) {
1282 found_tab_stops.push(tab_stop);
1283 position = distance - tab_stop.byte_offset;
1284 }
1285
1286 let expected_found_tab_stops: Vec<_> = all_tab_stops
1287 .iter()
1288 .take_while(|tab_stop| tab_stop.byte_offset <= distance)
1289 .cloned()
1290 .collect();
1291
1292 pretty_assertions::assert_eq!(
1293 found_tab_stops,
1294 expected_found_tab_stops,
1295 "TabStopCursor output mismatch for distance {}. Input: {:?}",
1296 distance,
1297 input
1298 );
1299
1300 let final_position = cursor.byte_offset();
1301 if !found_tab_stops.is_empty() {
1302 let last_tab_stop = found_tab_stops.last().unwrap();
1303 assert!(
1304 final_position >= last_tab_stop.byte_offset,
1305 "Cursor final position {} is before last tab stop {}. Input: {:?}",
1306 final_position,
1307 last_tab_stop.byte_offset,
1308 input
1309 );
1310 }
1311 }
1312 }
1313
1314 #[gpui::test]
1315 fn test_tab_stop_cursor_utf16(cx: &mut gpui::App) {
1316 let text = "\r\t😁foo\tb😀arbar🤯bar\t\tbaz\n";
1317 let buffer = MultiBuffer::build_simple(text, cx);
1318 let buffer_snapshot = buffer.read(cx).snapshot(cx);
1319 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot);
1320 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
1321 let chunks = fold_snapshot.chunks(
1322 FoldOffset(0)..fold_snapshot.len(),
1323 false,
1324 Default::default(),
1325 );
1326 let mut cursor = TabStopCursor::new(chunks);
1327 assert!(cursor.seek(0).is_none());
1328
1329 let mut expected_tab_stops = Vec::new();
1330 let mut byte_offset = 0;
1331 for (i, ch) in fold_snapshot.chars_at(FoldPoint::new(0, 0)).enumerate() {
1332 byte_offset += ch.len_utf8() as u32;
1333
1334 if ch == '\t' {
1335 expected_tab_stops.push(TabStop {
1336 byte_offset,
1337 char_offset: i as u32 + 1,
1338 });
1339 }
1340 }
1341
1342 let mut actual_tab_stops = Vec::new();
1343 while let Some(tab_stop) = cursor.seek(u32::MAX) {
1344 actual_tab_stops.push(tab_stop);
1345 }
1346
1347 pretty_assertions::assert_eq!(actual_tab_stops.as_slice(), expected_tab_stops.as_slice(),);
1348
1349 assert_eq!(cursor.byte_offset(), byte_offset);
1350 }
1351
1352 #[gpui::test(iterations = 100)]
1353 fn test_tab_stop_cursor_random_utf16(cx: &mut gpui::App, mut rng: StdRng) {
1354 // Generate random input string with up to 512 characters including tabs
1355 let len = rng.random_range(0..=2048);
1356 let input = util::RandomCharIter::new(&mut rng)
1357 .take(len)
1358 .collect::<String>();
1359
1360 // Build the buffer and create cursor
1361 let buffer = MultiBuffer::build_simple(&input, cx);
1362 let buffer_snapshot = buffer.read(cx).snapshot(cx);
1363 let (_, inlay_snapshot) = InlayMap::new(buffer_snapshot.clone());
1364 let (_, fold_snapshot) = FoldMap::new(inlay_snapshot);
1365
1366 // First, collect all expected tab positions
1367 let mut all_tab_stops = Vec::new();
1368 let mut byte_offset = 0;
1369 for (i, ch) in buffer_snapshot.text().chars().enumerate() {
1370 byte_offset += ch.len_utf8() as u32;
1371 if ch == '\t' {
1372 all_tab_stops.push(TabStop {
1373 byte_offset,
1374 char_offset: i as u32 + 1,
1375 });
1376 }
1377 }
1378
1379 // Test with various distances
1380 // let distances = vec![1, 5, 10, 50, 100, u32::MAX];
1381 let distances = vec![150];
1382
1383 for distance in distances {
1384 let chunks = fold_snapshot.chunks_at(FoldPoint::new(0, 0));
1385 let mut cursor = TabStopCursor::new(chunks);
1386
1387 let mut found_tab_stops = Vec::new();
1388 let mut position = distance;
1389 while let Some(tab_stop) = cursor.seek(position) {
1390 found_tab_stops.push(tab_stop);
1391 position = distance - tab_stop.byte_offset;
1392 }
1393
1394 let expected_found_tab_stops: Vec<_> = all_tab_stops
1395 .iter()
1396 .take_while(|tab_stop| tab_stop.byte_offset <= distance)
1397 .cloned()
1398 .collect();
1399
1400 pretty_assertions::assert_eq!(
1401 found_tab_stops,
1402 expected_found_tab_stops,
1403 "TabStopCursor output mismatch for distance {}. Input: {:?}",
1404 distance,
1405 input
1406 );
1407
1408 let final_position = cursor.byte_offset();
1409 if !found_tab_stops.is_empty() {
1410 let last_tab_stop = found_tab_stops.last().unwrap();
1411 assert!(
1412 final_position >= last_tab_stop.byte_offset,
1413 "Cursor final position {} is before last tab stop {}. Input: {:?}",
1414 final_position,
1415 last_tab_stop.byte_offset,
1416 input
1417 );
1418 }
1419 }
1420 }
1421}
1422
1423struct TabStopCursor<'a, I>
1424where
1425 I: Iterator<Item = Chunk<'a>>,
1426{
1427 chunks: I,
1428 byte_offset: u32,
1429 char_offset: u32,
1430 /// Chunk
1431 /// last tab position iterated through
1432 current_chunk: Option<(Chunk<'a>, u32)>,
1433}
1434
1435impl<'a, I> TabStopCursor<'a, I>
1436where
1437 I: Iterator<Item = Chunk<'a>>,
1438{
1439 fn new(chunks: impl IntoIterator<Item = Chunk<'a>, IntoIter = I>) -> Self {
1440 Self {
1441 chunks: chunks.into_iter(),
1442 byte_offset: 0,
1443 char_offset: 0,
1444 current_chunk: None,
1445 }
1446 }
1447
1448 fn bytes_until_next_char(&self) -> Option<usize> {
1449 self.current_chunk.as_ref().and_then(|(chunk, idx)| {
1450 let mut idx = *idx;
1451 let mut diff = 0;
1452 while idx > 0 && chunk.chars & (1 << idx) == 0 {
1453 idx -= 1;
1454 diff += 1;
1455 }
1456
1457 if chunk.chars & (1 << idx) != 0 {
1458 Some(
1459 (chunk.text[idx as usize..].chars().next()?)
1460 .len_utf8()
1461 .saturating_sub(diff),
1462 )
1463 } else {
1464 None
1465 }
1466 })
1467 }
1468
1469 fn is_char_boundary(&self) -> bool {
1470 self.current_chunk
1471 .as_ref()
1472 .is_some_and(|(chunk, idx)| (chunk.chars & (1 << *idx.min(&127))) != 0)
1473 }
1474
1475 /// distance: length to move forward while searching for the next tab stop
1476 fn seek(&mut self, distance: u32) -> Option<TabStop> {
1477 if distance == 0 {
1478 return None;
1479 }
1480
1481 let mut distance_traversed = 0;
1482
1483 while let Some((mut chunk, chunk_position)) = self
1484 .current_chunk
1485 .take()
1486 .or_else(|| self.chunks.next().zip(Some(0)))
1487 {
1488 if chunk.tabs == 0 {
1489 let chunk_distance = chunk.text.len() as u32 - chunk_position;
1490 if chunk_distance + distance_traversed >= distance {
1491 let overshoot = distance_traversed.abs_diff(distance);
1492
1493 self.byte_offset += overshoot;
1494 self.char_offset += get_char_offset(
1495 chunk_position..(chunk_position + overshoot).saturating_sub(1).min(127),
1496 chunk.chars,
1497 );
1498
1499 self.current_chunk = Some((chunk, chunk_position + overshoot));
1500
1501 return None;
1502 }
1503
1504 self.byte_offset += chunk_distance;
1505 self.char_offset += get_char_offset(
1506 chunk_position..(chunk_position + chunk_distance).saturating_sub(1).min(127),
1507 chunk.chars,
1508 );
1509 distance_traversed += chunk_distance;
1510 continue;
1511 }
1512 let tab_position = chunk.tabs.trailing_zeros() + 1;
1513
1514 if distance_traversed + tab_position - chunk_position > distance {
1515 let cursor_position = distance_traversed.abs_diff(distance);
1516
1517 self.char_offset += get_char_offset(
1518 chunk_position..(chunk_position + cursor_position - 1),
1519 chunk.chars,
1520 );
1521 self.current_chunk = Some((chunk, cursor_position + chunk_position));
1522 self.byte_offset += cursor_position;
1523
1524 return None;
1525 }
1526
1527 self.byte_offset += tab_position - chunk_position;
1528 self.char_offset += get_char_offset(chunk_position..(tab_position - 1), chunk.chars);
1529
1530 let tabstop = TabStop {
1531 char_offset: self.char_offset,
1532 byte_offset: self.byte_offset,
1533 };
1534
1535 chunk.tabs = (chunk.tabs - 1) & chunk.tabs;
1536
1537 if tab_position as usize != chunk.text.len() {
1538 self.current_chunk = Some((chunk, tab_position));
1539 }
1540
1541 return Some(tabstop);
1542 }
1543
1544 None
1545 }
1546
1547 fn byte_offset(&self) -> u32 {
1548 self.byte_offset
1549 }
1550
1551 fn char_offset(&self) -> u32 {
1552 self.char_offset
1553 }
1554}
1555
1556#[inline(always)]
1557fn get_char_offset(range: Range<u32>, bit_map: u128) -> u32 {
1558 // This edge case can happen when we're at chunk position 128
1559
1560 if range.start == range.end {
1561 return if (1u128 << range.start) & bit_map == 0 {
1562 0
1563 } else {
1564 1
1565 };
1566 }
1567 let end_shift: u128 = 127u128 - range.end.min(127) as u128;
1568 let mut bit_mask = (u128::MAX >> range.start) << range.start;
1569 bit_mask = (bit_mask << end_shift) >> end_shift;
1570 let bit_map = bit_map & bit_mask;
1571
1572 bit_map.count_ones()
1573}
1574
1575#[derive(Clone, Copy, Debug, PartialEq, Eq)]
1576struct TabStop {
1577 char_offset: u32,
1578 byte_offset: u32,
1579}