1use rope::Rope;
2use std::{iter, ops::Range};
3use sum_tree::SumTree;
4use text::{Anchor, BufferId, BufferSnapshot, OffsetRangeExt, Point};
5
6pub use git2 as libgit;
7use libgit::{DiffLineType as GitDiffLineType, DiffOptions as GitOptions, Patch as GitPatch};
8
9#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
10pub enum DiffHunkStatus {
11 Added,
12 Modified,
13 Removed,
14}
15
16/// A diff hunk, representing a range of consequent lines in a singleton buffer, associated with a generic range.
17#[derive(Debug, Clone, PartialEq, Eq)]
18pub struct DiffHunk<T> {
19 /// E.g. a range in multibuffer, that has an excerpt added, singleton buffer for which has this diff hunk.
20 /// Consider a singleton buffer with 10 lines, all of them are modified — so a corresponding diff hunk would have a range 0..10.
21 /// And a multibuffer with the excerpt of lines 2-6 from the singleton buffer.
22 /// If the multibuffer is searched for diff hunks, the associated range would be multibuffer rows, corresponding to rows 2..6 from the singleton buffer.
23 /// But the hunk range would be 0..10, same for any other excerpts from the same singleton buffer.
24 pub associated_range: Range<T>,
25 /// Singleton buffer ID this hunk belongs to.
26 pub buffer_id: BufferId,
27 /// A consequent range of lines in the singleton buffer, that were changed and produced this diff hunk.
28 pub buffer_range: Range<Anchor>,
29 /// Original singleton buffer text before the change, that was instead of the `buffer_range`.
30 pub diff_base_byte_range: Range<usize>,
31}
32
33impl sum_tree::Item for DiffHunk<Anchor> {
34 type Summary = DiffHunkSummary;
35
36 fn summary(&self) -> Self::Summary {
37 DiffHunkSummary {
38 buffer_range: self.associated_range.clone(),
39 }
40 }
41}
42
43#[derive(Debug, Default, Clone)]
44pub struct DiffHunkSummary {
45 buffer_range: Range<Anchor>,
46}
47
48impl sum_tree::Summary for DiffHunkSummary {
49 type Context = text::BufferSnapshot;
50
51 fn add_summary(&mut self, other: &Self, buffer: &Self::Context) {
52 self.buffer_range.start = self
53 .buffer_range
54 .start
55 .min(&other.buffer_range.start, buffer);
56 self.buffer_range.end = self.buffer_range.end.max(&other.buffer_range.end, buffer);
57 }
58}
59
60#[derive(Debug, Clone)]
61pub struct BufferDiff {
62 last_buffer_version: Option<clock::Global>,
63 tree: SumTree<DiffHunk<Anchor>>,
64}
65
66impl Default for BufferDiff {
67 fn default() -> Self {
68 Self::new()
69 }
70}
71
72impl BufferDiff {
73 pub fn new() -> BufferDiff {
74 BufferDiff {
75 last_buffer_version: None,
76 tree: SumTree::new(),
77 }
78 }
79
80 pub fn is_empty(&self) -> bool {
81 self.tree.is_empty()
82 }
83
84 pub fn hunks_in_row_range<'a>(
85 &'a self,
86 range: Range<u32>,
87 buffer: &'a BufferSnapshot,
88 ) -> impl 'a + Iterator<Item = DiffHunk<u32>> {
89 let start = buffer.anchor_before(Point::new(range.start, 0));
90 let end = buffer.anchor_after(Point::new(range.end, 0));
91
92 self.hunks_intersecting_range(start..end, buffer)
93 }
94
95 pub fn hunks_intersecting_range<'a>(
96 &'a self,
97 range: Range<Anchor>,
98 buffer: &'a BufferSnapshot,
99 ) -> impl 'a + Iterator<Item = DiffHunk<u32>> {
100 let mut cursor = self.tree.filter::<_, DiffHunkSummary>(move |summary| {
101 let before_start = summary.buffer_range.end.cmp(&range.start, buffer).is_lt();
102 let after_end = summary.buffer_range.start.cmp(&range.end, buffer).is_gt();
103 !before_start && !after_end
104 });
105
106 let anchor_iter = std::iter::from_fn(move || {
107 cursor.next(buffer);
108 cursor.item()
109 })
110 .flat_map(move |hunk| {
111 [
112 (
113 &hunk.associated_range.start,
114 hunk.diff_base_byte_range.start,
115 ),
116 (&hunk.associated_range.end, hunk.diff_base_byte_range.end),
117 ]
118 .into_iter()
119 });
120
121 let mut summaries = buffer.summaries_for_anchors_with_payload::<Point, _, _>(anchor_iter);
122 iter::from_fn(move || {
123 let (start_point, start_base) = summaries.next()?;
124 let (mut end_point, end_base) = summaries.next()?;
125
126 if end_point.column > 0 {
127 end_point.row += 1;
128 end_point.column = 0;
129 }
130
131 Some(DiffHunk {
132 associated_range: start_point.row..end_point.row,
133 diff_base_byte_range: start_base..end_base,
134 buffer_range: buffer.anchor_before(start_point)..buffer.anchor_after(end_point),
135 buffer_id: buffer.remote_id(),
136 })
137 })
138 }
139
140 pub fn hunks_intersecting_range_rev<'a>(
141 &'a self,
142 range: Range<Anchor>,
143 buffer: &'a BufferSnapshot,
144 ) -> impl 'a + Iterator<Item = DiffHunk<u32>> {
145 let mut cursor = self.tree.filter::<_, DiffHunkSummary>(move |summary| {
146 let before_start = summary.buffer_range.end.cmp(&range.start, buffer).is_lt();
147 let after_end = summary.buffer_range.start.cmp(&range.end, buffer).is_gt();
148 !before_start && !after_end
149 });
150
151 std::iter::from_fn(move || {
152 cursor.prev(buffer);
153
154 let hunk = cursor.item()?;
155 let range = hunk.associated_range.to_point(buffer);
156 let end_row = if range.end.column > 0 {
157 range.end.row + 1
158 } else {
159 range.end.row
160 };
161
162 Some(DiffHunk {
163 associated_range: range.start.row..end_row,
164 diff_base_byte_range: hunk.diff_base_byte_range.clone(),
165 buffer_range: hunk.buffer_range.clone(),
166 buffer_id: hunk.buffer_id,
167 })
168 })
169 }
170
171 #[cfg(test)]
172 fn clear(&mut self, buffer: &text::BufferSnapshot) {
173 self.last_buffer_version = Some(buffer.version().clone());
174 self.tree = SumTree::new();
175 }
176
177 pub async fn update(&mut self, diff_base: &Rope, buffer: &text::BufferSnapshot) {
178 let mut tree = SumTree::new();
179
180 let diff_base_text = diff_base.to_string();
181 let buffer_text = buffer.as_rope().to_string();
182 let patch = Self::diff(&diff_base_text, &buffer_text);
183
184 if let Some(patch) = patch {
185 let mut divergence = 0;
186 for hunk_index in 0..patch.num_hunks() {
187 let hunk = Self::process_patch_hunk(&patch, hunk_index, buffer, &mut divergence);
188 tree.push(hunk, buffer);
189 }
190 }
191
192 self.tree = tree;
193 self.last_buffer_version = Some(buffer.version().clone());
194 }
195
196 #[cfg(test)]
197 fn hunks<'a>(&'a self, text: &'a BufferSnapshot) -> impl 'a + Iterator<Item = DiffHunk<u32>> {
198 let start = text.anchor_before(Point::new(0, 0));
199 let end = text.anchor_after(Point::new(u32::MAX, u32::MAX));
200 self.hunks_intersecting_range(start..end, text)
201 }
202
203 fn diff<'a>(head: &'a str, current: &'a str) -> Option<GitPatch<'a>> {
204 let mut options = GitOptions::default();
205 options.context_lines(0);
206
207 let patch = GitPatch::from_buffers(
208 head.as_bytes(),
209 None,
210 current.as_bytes(),
211 None,
212 Some(&mut options),
213 );
214
215 match patch {
216 Ok(patch) => Some(patch),
217
218 Err(err) => {
219 log::error!("`GitPatch::from_buffers` failed: {}", err);
220 None
221 }
222 }
223 }
224
225 fn process_patch_hunk(
226 patch: &GitPatch<'_>,
227 hunk_index: usize,
228 buffer: &text::BufferSnapshot,
229 buffer_row_divergence: &mut i64,
230 ) -> DiffHunk<Anchor> {
231 let line_item_count = patch.num_lines_in_hunk(hunk_index).unwrap();
232 assert!(line_item_count > 0);
233
234 let mut first_deletion_buffer_row: Option<u32> = None;
235 let mut buffer_row_range: Option<Range<u32>> = None;
236 let mut diff_base_byte_range: Option<Range<usize>> = None;
237
238 for line_index in 0..line_item_count {
239 let line = patch.line_in_hunk(hunk_index, line_index).unwrap();
240 let kind = line.origin_value();
241 let content_offset = line.content_offset() as isize;
242 let content_len = line.content().len() as isize;
243
244 if kind == GitDiffLineType::Addition {
245 *buffer_row_divergence += 1;
246 let row = line.new_lineno().unwrap().saturating_sub(1);
247
248 match &mut buffer_row_range {
249 Some(buffer_row_range) => buffer_row_range.end = row + 1,
250 None => buffer_row_range = Some(row..row + 1),
251 }
252 }
253
254 if kind == GitDiffLineType::Deletion {
255 let end = content_offset + content_len;
256
257 match &mut diff_base_byte_range {
258 Some(head_byte_range) => head_byte_range.end = end as usize,
259 None => diff_base_byte_range = Some(content_offset as usize..end as usize),
260 }
261
262 if first_deletion_buffer_row.is_none() {
263 let old_row = line.old_lineno().unwrap().saturating_sub(1);
264 let row = old_row as i64 + *buffer_row_divergence;
265 first_deletion_buffer_row = Some(row as u32);
266 }
267
268 *buffer_row_divergence -= 1;
269 }
270 }
271
272 //unwrap_or deletion without addition
273 let buffer_row_range = buffer_row_range.unwrap_or_else(|| {
274 //we cannot have an addition-less hunk without deletion(s) or else there would be no hunk
275 let row = first_deletion_buffer_row.unwrap();
276 row..row
277 });
278
279 //unwrap_or addition without deletion
280 let diff_base_byte_range = diff_base_byte_range.unwrap_or(0..0);
281
282 let start = Point::new(buffer_row_range.start, 0);
283 let end = Point::new(buffer_row_range.end, 0);
284 let buffer_range = buffer.anchor_before(start)..buffer.anchor_before(end);
285 DiffHunk {
286 associated_range: buffer_range.clone(),
287 buffer_range,
288 diff_base_byte_range,
289 buffer_id: buffer.remote_id(),
290 }
291 }
292}
293
294/// Range (crossing new lines), old, new
295#[cfg(any(test, feature = "test-support"))]
296#[track_caller]
297pub fn assert_hunks<Iter>(
298 diff_hunks: Iter,
299 buffer: &BufferSnapshot,
300 diff_base: &str,
301 expected_hunks: &[(Range<u32>, &str, &str)],
302) where
303 Iter: Iterator<Item = DiffHunk<u32>>,
304{
305 let actual_hunks = diff_hunks
306 .map(|hunk| {
307 (
308 hunk.associated_range.clone(),
309 &diff_base[hunk.diff_base_byte_range],
310 buffer
311 .text_for_range(
312 Point::new(hunk.associated_range.start, 0)
313 ..Point::new(hunk.associated_range.end, 0),
314 )
315 .collect::<String>(),
316 )
317 })
318 .collect::<Vec<_>>();
319
320 let expected_hunks: Vec<_> = expected_hunks
321 .iter()
322 .map(|(r, s, h)| (r.clone(), *s, h.to_string()))
323 .collect();
324
325 assert_eq!(actual_hunks, expected_hunks);
326}
327
328#[cfg(test)]
329mod tests {
330 use std::assert_eq;
331
332 use super::*;
333 use text::{Buffer, BufferId};
334 use unindent::Unindent as _;
335
336 #[test]
337 fn test_buffer_diff_simple() {
338 let diff_base = "
339 one
340 two
341 three
342 "
343 .unindent();
344 let diff_base_rope = Rope::from(diff_base.clone());
345
346 let buffer_text = "
347 one
348 HELLO
349 three
350 "
351 .unindent();
352
353 let mut buffer = Buffer::new(0, BufferId::new(1).unwrap(), buffer_text);
354 let mut diff = BufferDiff::new();
355 smol::block_on(diff.update(&diff_base_rope, &buffer));
356 assert_hunks(
357 diff.hunks(&buffer),
358 &buffer,
359 &diff_base,
360 &[(1..2, "two\n", "HELLO\n")],
361 );
362
363 buffer.edit([(0..0, "point five\n")]);
364 smol::block_on(diff.update(&diff_base_rope, &buffer));
365 assert_hunks(
366 diff.hunks(&buffer),
367 &buffer,
368 &diff_base,
369 &[(0..1, "", "point five\n"), (2..3, "two\n", "HELLO\n")],
370 );
371
372 diff.clear(&buffer);
373 assert_hunks(diff.hunks(&buffer), &buffer, &diff_base, &[]);
374 }
375
376 #[test]
377 fn test_buffer_diff_range() {
378 let diff_base = "
379 one
380 two
381 three
382 four
383 five
384 six
385 seven
386 eight
387 nine
388 ten
389 "
390 .unindent();
391 let diff_base_rope = Rope::from(diff_base.clone());
392
393 let buffer_text = "
394 A
395 one
396 B
397 two
398 C
399 three
400 HELLO
401 four
402 five
403 SIXTEEN
404 seven
405 eight
406 WORLD
407 nine
408
409 ten
410
411 "
412 .unindent();
413
414 let buffer = Buffer::new(0, BufferId::new(1).unwrap(), buffer_text);
415 let mut diff = BufferDiff::new();
416 smol::block_on(diff.update(&diff_base_rope, &buffer));
417 assert_eq!(diff.hunks(&buffer).count(), 8);
418
419 assert_hunks(
420 diff.hunks_in_row_range(7..12, &buffer),
421 &buffer,
422 &diff_base,
423 &[
424 (6..7, "", "HELLO\n"),
425 (9..10, "six\n", "SIXTEEN\n"),
426 (12..13, "", "WORLD\n"),
427 ],
428 );
429 }
430}