1use std::ops::Range;
2
3use sum_tree::{Bias, SumTree};
4use text::{Anchor, BufferSnapshot, OffsetRangeExt, Point, Rope, ToOffset, ToPoint};
5
6pub use git2 as libgit;
7use libgit::{
8 DiffLine as GitDiffLine, DiffLineType as GitDiffLineType, DiffOptions as GitOptions,
9 Patch as GitPatch,
10};
11
12#[derive(Debug, Clone, Copy)]
13pub enum DiffHunkStatus {
14 Added,
15 Modified,
16 Removed,
17}
18
19#[derive(Debug, Clone, PartialEq, Eq)]
20pub struct DiffHunk<T> {
21 pub buffer_range: Range<T>,
22 pub head_byte_range: Range<usize>,
23}
24
25impl DiffHunk<u32> {
26 pub fn status(&self) -> DiffHunkStatus {
27 if self.head_byte_range.is_empty() {
28 DiffHunkStatus::Added
29 } else if self.buffer_range.is_empty() {
30 DiffHunkStatus::Removed
31 } else {
32 DiffHunkStatus::Modified
33 }
34 }
35}
36
37impl sum_tree::Item for DiffHunk<Anchor> {
38 type Summary = DiffHunkSummary;
39
40 fn summary(&self) -> Self::Summary {
41 DiffHunkSummary {
42 buffer_range: self.buffer_range.clone(),
43 head_range: self.head_byte_range.clone(),
44 }
45 }
46}
47
48#[derive(Debug, Default, Clone)]
49pub struct DiffHunkSummary {
50 buffer_range: Range<Anchor>,
51 head_range: Range<usize>,
52}
53
54impl sum_tree::Summary for DiffHunkSummary {
55 type Context = text::BufferSnapshot;
56
57 fn add_summary(&mut self, other: &Self, _: &Self::Context) {
58 self.head_range.start = self.head_range.start.min(other.head_range.start);
59 self.head_range.end = self.head_range.end.max(other.head_range.end);
60 }
61}
62
63#[derive(Debug, Default, Clone, PartialEq, Eq, PartialOrd, Ord)]
64struct HunkHeadEnd(usize);
65
66impl<'a> sum_tree::Dimension<'a, DiffHunkSummary> for HunkHeadEnd {
67 fn add_summary(&mut self, summary: &'a DiffHunkSummary, _: &text::BufferSnapshot) {
68 self.0 = summary.head_range.end;
69 }
70
71 fn from_summary(summary: &'a DiffHunkSummary, _: &text::BufferSnapshot) -> Self {
72 HunkHeadEnd(summary.head_range.end)
73 }
74}
75
76#[derive(Debug, Default, Clone, PartialEq, Eq, PartialOrd, Ord)]
77struct HunkBufferStart(u32);
78
79impl<'a> sum_tree::Dimension<'a, DiffHunkSummary> for HunkBufferStart {
80 fn add_summary(&mut self, summary: &'a DiffHunkSummary, buffer: &text::BufferSnapshot) {
81 self.0 = summary.buffer_range.start.to_point(buffer).row;
82 }
83
84 fn from_summary(summary: &'a DiffHunkSummary, buffer: &text::BufferSnapshot) -> Self {
85 HunkBufferStart(summary.buffer_range.start.to_point(buffer).row)
86 }
87}
88
89#[derive(Debug, Default, Clone, PartialEq, Eq, PartialOrd, Ord)]
90struct HunkBufferEnd(u32);
91
92impl<'a> sum_tree::Dimension<'a, DiffHunkSummary> for HunkBufferEnd {
93 fn add_summary(&mut self, summary: &'a DiffHunkSummary, buffer: &text::BufferSnapshot) {
94 self.0 = summary.buffer_range.end.to_point(buffer).row;
95 }
96
97 fn from_summary(summary: &'a DiffHunkSummary, buffer: &text::BufferSnapshot) -> Self {
98 HunkBufferEnd(summary.buffer_range.end.to_point(buffer).row)
99 }
100}
101
102struct HunkLineIter<'a, 'b> {
103 patch: &'a GitPatch<'b>,
104 hunk_index: usize,
105 line_index: usize,
106}
107
108impl<'a, 'b> HunkLineIter<'a, 'b> {
109 fn new(patch: &'a GitPatch<'b>, hunk_index: usize) -> Self {
110 HunkLineIter {
111 patch,
112 hunk_index,
113 line_index: 0,
114 }
115 }
116}
117
118impl<'a, 'b> std::iter::Iterator for HunkLineIter<'a, 'b> {
119 type Item = GitDiffLine<'b>;
120
121 fn next(&mut self) -> Option<Self::Item> {
122 if self.line_index >= self.patch.num_lines_in_hunk(self.hunk_index).unwrap() {
123 return None;
124 }
125
126 let line_index = self.line_index;
127 self.line_index += 1;
128 Some(
129 self.patch
130 .line_in_hunk(self.hunk_index, line_index)
131 .unwrap(),
132 )
133 }
134}
135
136#[derive(Clone)]
137pub struct BufferDiffSnapshot {
138 tree: SumTree<DiffHunk<Anchor>>,
139}
140
141impl BufferDiffSnapshot {
142 pub fn hunks_in_range<'a>(
143 &'a self,
144 query_row_range: Range<u32>,
145 buffer: &'a BufferSnapshot,
146 ) -> impl 'a + Iterator<Item = DiffHunk<u32>> {
147 self.tree.iter().filter_map(move |hunk| {
148 let range = hunk.buffer_range.to_point(&buffer);
149
150 if range.start.row <= query_row_range.end && query_row_range.start <= range.end.row {
151 let end_row = if range.end.column > 0 {
152 range.end.row + 1
153 } else {
154 range.end.row
155 };
156
157 Some(DiffHunk {
158 buffer_range: range.start.row..end_row,
159 head_byte_range: hunk.head_byte_range.clone(),
160 })
161 } else {
162 None
163 }
164 })
165 }
166
167 #[cfg(test)]
168 fn hunks<'a>(&'a self, text: &'a BufferSnapshot) -> impl 'a + Iterator<Item = DiffHunk<u32>> {
169 self.hunks_in_range(0..u32::MAX, text)
170 }
171}
172
173pub struct BufferDiff {
174 last_update_version: clock::Global,
175 snapshot: BufferDiffSnapshot,
176}
177
178impl BufferDiff {
179 pub fn new(head_text: &Option<String>, buffer: &text::BufferSnapshot) -> BufferDiff {
180 let mut tree = SumTree::new();
181
182 if let Some(head_text) = head_text {
183 let buffer_text = buffer.as_rope().to_string();
184 let patch = Self::diff(&head_text, &buffer_text);
185
186 if let Some(patch) = patch {
187 for hunk_index in 0..patch.num_hunks() {
188 let hunk = Self::process_patch_hunk(&patch, hunk_index, buffer);
189 tree.push(hunk, buffer);
190 }
191 }
192 }
193
194 BufferDiff {
195 last_update_version: buffer.version().clone(),
196 snapshot: BufferDiffSnapshot { tree },
197 }
198 }
199
200 pub fn snapshot(&self) -> BufferDiffSnapshot {
201 self.snapshot.clone()
202 }
203
204 pub fn update(&mut self, head_text: &str, buffer: &text::BufferSnapshot) {
205 let mut tree = SumTree::new();
206
207 let buffer_text = buffer.as_rope().to_string();
208 let patch = Self::diff(&head_text, &buffer_text);
209
210 if let Some(patch) = patch {
211 for hunk_index in 0..patch.num_hunks() {
212 let hunk = Self::process_patch_hunk(&patch, hunk_index, buffer);
213 tree.push(hunk, buffer);
214 }
215 }
216
217 self.last_update_version = buffer.version().clone();
218 self.snapshot.tree = tree;
219 }
220
221 fn diff<'a>(head: &'a str, current: &'a str) -> Option<GitPatch<'a>> {
222 let mut options = GitOptions::default();
223 options.context_lines(0);
224
225 let patch = GitPatch::from_buffers(
226 head.as_bytes(),
227 None,
228 current.as_bytes(),
229 None,
230 Some(&mut options),
231 );
232
233 match patch {
234 Ok(patch) => Some(patch),
235
236 Err(err) => {
237 log::error!("`GitPatch::from_buffers` failed: {}", err);
238 None
239 }
240 }
241 }
242
243 fn group_edit_ranges(&self, buffer: &text::BufferSnapshot) -> Vec<Range<u32>> {
244 const EXPAND_BY: u32 = 20;
245 const COMBINE_DISTANCE: u32 = 5;
246
247 // let mut cursor = self.snapshot.tree.cursor::<HunkBufferStart>();
248
249 let mut ranges = Vec::<Range<u32>>::new();
250
251 for edit in buffer.edits_since::<Point>(&self.last_update_version) {
252 let buffer_start = edit.new.start.row.saturating_sub(EXPAND_BY);
253 let buffer_end = (edit.new.end.row + EXPAND_BY).min(buffer.row_count());
254
255 match ranges.last_mut() {
256 Some(last_range) if last_range.end.abs_diff(buffer_end) <= COMBINE_DISTANCE => {
257 last_range.start = last_range.start.min(buffer_start);
258 last_range.end = last_range.end.max(buffer_end);
259 }
260
261 _ => ranges.push(buffer_start..buffer_end),
262 }
263 }
264
265 ranges
266 }
267
268 fn process_patch_hunk<'a>(
269 patch: &GitPatch<'a>,
270 hunk_index: usize,
271 buffer: &text::BufferSnapshot,
272 ) -> DiffHunk<Anchor> {
273 let line_item_count = patch.num_lines_in_hunk(hunk_index).unwrap();
274 assert!(line_item_count > 0);
275
276 let mut first_deletion_buffer_row: Option<u32> = None;
277 let mut buffer_byte_range: Option<Range<usize>> = None;
278 let mut head_byte_range: Option<Range<usize>> = None;
279
280 for line_index in 0..line_item_count {
281 let line = patch.line_in_hunk(hunk_index, line_index).unwrap();
282 let kind = line.origin_value();
283 let content_offset = line.content_offset() as isize;
284 let content_len = line.content().len() as isize;
285
286 match (kind, &mut buffer_byte_range, &mut head_byte_range) {
287 (GitDiffLineType::Addition, None, _) => {
288 let end = content_offset + content_len;
289 buffer_byte_range = Some(content_offset as usize..end as usize);
290 }
291
292 (GitDiffLineType::Addition, Some(buffer_byte_range), _) => {
293 let end = content_offset + content_len;
294 buffer_byte_range.end = end as usize;
295 }
296
297 (GitDiffLineType::Deletion, _, None) => {
298 let end = content_offset + content_len;
299 head_byte_range = Some(content_offset as usize..end as usize);
300 }
301
302 (GitDiffLineType::Deletion, _, Some(head_byte_range)) => {
303 let end = content_offset + content_len;
304 head_byte_range.end = end as usize;
305 }
306
307 _ => {}
308 }
309
310 if kind == GitDiffLineType::Deletion && first_deletion_buffer_row.is_none() {
311 //old_lineno is guarenteed to be Some for deletions
312 //libgit gives us line numbers that are 1-indexed but also returns a 0 for some states
313 let row = line.old_lineno().unwrap().saturating_sub(1);
314 first_deletion_buffer_row = Some(row);
315 }
316 }
317
318 //unwrap_or deletion without addition
319 let buffer_byte_range = buffer_byte_range.unwrap_or_else(|| {
320 //we cannot have an addition-less hunk without deletion(s) or else there would be no hunk
321 let row = first_deletion_buffer_row.unwrap();
322 let anchor = buffer.anchor_before(Point::new(row, 0));
323 let offset = anchor.to_offset(buffer);
324 offset..offset
325 });
326
327 //unwrap_or addition without deletion
328 let head_byte_range = head_byte_range.unwrap_or(0..0);
329
330 DiffHunk {
331 buffer_range: buffer.anchor_before(buffer_byte_range.start)
332 ..buffer.anchor_before(buffer_byte_range.end),
333 head_byte_range,
334 }
335 }
336}
337
338#[cfg(test)]
339mod tests {
340 use super::*;
341 use text::Buffer;
342 use unindent::Unindent as _;
343
344 #[gpui::test]
345 fn test_buffer_diff_simple() {
346 let head_text = "
347 one
348 two
349 three
350 "
351 .unindent();
352
353 let buffer_text = "
354 one
355 hello
356 three
357 "
358 .unindent();
359
360 let mut buffer = Buffer::new(0, 0, buffer_text);
361 let diff = BufferDiff::new(&Some(head_text.clone()), &buffer);
362 assert_hunks(&diff, &buffer, &head_text, &[(1..2, "two\n")]);
363
364 buffer.edit([(0..0, "point five\n")]);
365 assert_hunks(&diff, &buffer, &head_text, &[(2..3, "two\n")]);
366 }
367
368 #[track_caller]
369 fn assert_hunks(
370 diff: &BufferDiff,
371 buffer: &BufferSnapshot,
372 head_text: &str,
373 expected_hunks: &[(Range<u32>, &str)],
374 ) {
375 let hunks = diff.snapshot.hunks(buffer).collect::<Vec<_>>();
376 assert_eq!(
377 hunks.len(),
378 expected_hunks.len(),
379 "actual hunks are {hunks:#?}"
380 );
381
382 let diff_iter = hunks.iter().enumerate();
383 for ((index, hunk), (expected_range, expected_str)) in diff_iter.zip(expected_hunks) {
384 assert_eq!(&hunk.buffer_range, expected_range, "for hunk {index}");
385 assert_eq!(
386 &head_text[hunk.head_byte_range.clone()],
387 *expected_str,
388 "for hunk {index}"
389 );
390 }
391 }
392
393 // use rand::rngs::StdRng;
394 // #[gpui::test(iterations = 100)]
395 // fn test_buffer_diff_random(mut rng: StdRng) {}
396}