1use std::{
2 cell::LazyCell,
3 collections::BTreeSet,
4 ops::Range,
5 path::{Path, PathBuf},
6 pin::pin,
7 sync::Arc,
8};
9
10use anyhow::Context;
11use collections::HashSet;
12use fs::Fs;
13use futures::{AsyncBufReadExt, SinkExt, StreamExt, select_biased, stream::FuturesOrdered};
14use gpui::{App, AppContext, AsyncApp, Entity, Task};
15use language::{Buffer, BufferSnapshot};
16use parking_lot::Mutex;
17use postage::oneshot;
18use rpc::{AnyProtoClient, proto};
19use smol::{
20 channel::{Receiver, Sender, bounded, unbounded},
21 future::FutureExt,
22};
23
24use text::BufferId;
25use util::{ResultExt, maybe, paths::compare_rel_paths, rel_path::RelPath};
26use worktree::{Entry, ProjectEntryId, Snapshot, Worktree, WorktreeSettings};
27
28use crate::{
29 Project, ProjectItem, ProjectPath, RemotelyCreatedModels,
30 buffer_store::BufferStore,
31 search::{SearchQuery, SearchResult},
32 worktree_store::WorktreeStore,
33};
34
35pub struct Search {
36 buffer_store: Entity<BufferStore>,
37 worktree_store: Entity<WorktreeStore>,
38 limit: usize,
39 kind: SearchKind,
40}
41
42/// Represents search setup, before it is actually kicked off with Search::into_results
43enum SearchKind {
44 /// Search for candidates by inspecting file contents on file system, avoiding loading the buffer unless we know that a given file contains a match.
45 Local {
46 fs: Arc<dyn Fs>,
47 worktrees: Vec<Entity<Worktree>>,
48 },
49 /// Query remote host for candidates. As of writing, the host runs a local search in "buffers with matches only" mode.
50 Remote {
51 client: AnyProtoClient,
52 remote_id: u64,
53 models: Arc<Mutex<RemotelyCreatedModels>>,
54 },
55 /// Run search against a known set of candidates. Even when working with a remote host, this won't round-trip to host.
56 OpenBuffersOnly,
57}
58
59/// Represents results of project search and allows one to either obtain match positions OR
60/// just the handles to buffers that may match the search. Grabbing the handles is cheaper than obtaining full match positions, because in that case we'll look for
61/// at most one match in each file.
62#[must_use]
63pub struct SearchResultsHandle {
64 results: Receiver<SearchResult>,
65 matching_buffers: Receiver<Entity<Buffer>>,
66 trigger_search: Box<dyn FnOnce(&mut App) -> Task<()> + Send + Sync>,
67}
68
69impl SearchResultsHandle {
70 pub fn results(self, cx: &mut App) -> (Receiver<SearchResult>, Task<()>) {
71 let task = (self.trigger_search)(cx);
72 (self.results, task)
73 }
74
75 pub fn matching_buffers(self, cx: &mut App) -> (Receiver<Entity<Buffer>>, Task<()>) {
76 let task = (self.trigger_search)(cx);
77 (self.matching_buffers, task)
78 }
79}
80
81#[derive(Clone)]
82enum FindSearchCandidates {
83 Local {
84 fs: Arc<dyn Fs>,
85 /// Start off with all paths in project and filter them based on:
86 /// - Include filters
87 /// - Exclude filters
88 /// - Only open buffers
89 /// - Scan ignored files
90 /// Put another way: filter out files that can't match (without looking at file contents)
91 input_paths_rx: Receiver<InputPath>,
92 /// After that, if the buffer is not yet loaded, we'll figure out if it contains at least one match
93 /// based on disk contents of a buffer. This step is not performed for buffers we already have in memory.
94 confirm_contents_will_match_tx: Sender<MatchingEntry>,
95 confirm_contents_will_match_rx: Receiver<MatchingEntry>,
96 },
97 Remote,
98 OpenBuffersOnly,
99}
100
101impl Search {
102 pub fn local(
103 fs: Arc<dyn Fs>,
104 buffer_store: Entity<BufferStore>,
105 worktree_store: Entity<WorktreeStore>,
106 limit: usize,
107 cx: &mut App,
108 ) -> Self {
109 let worktrees = worktree_store.read(cx).visible_worktrees(cx).collect();
110 Self {
111 kind: SearchKind::Local { fs, worktrees },
112 buffer_store,
113 worktree_store,
114 limit,
115 }
116 }
117
118 pub(crate) fn remote(
119 buffer_store: Entity<BufferStore>,
120 worktree_store: Entity<WorktreeStore>,
121 limit: usize,
122 client_state: (AnyProtoClient, u64, Arc<Mutex<RemotelyCreatedModels>>),
123 ) -> Self {
124 Self {
125 kind: SearchKind::Remote {
126 client: client_state.0,
127 remote_id: client_state.1,
128 models: client_state.2,
129 },
130 buffer_store,
131 worktree_store,
132 limit,
133 }
134 }
135 pub(crate) fn open_buffers_only(
136 buffer_store: Entity<BufferStore>,
137 worktree_store: Entity<WorktreeStore>,
138 limit: usize,
139 ) -> Self {
140 Self {
141 kind: SearchKind::OpenBuffersOnly,
142 buffer_store,
143 worktree_store,
144 limit,
145 }
146 }
147
148 pub(crate) const MAX_SEARCH_RESULT_FILES: usize = 5_000;
149 pub(crate) const MAX_SEARCH_RESULT_RANGES: usize = 10_000;
150 /// Prepares a project search run. The resulting [`SearchResultsHandle`] has to be used to specify whether you're interested in matching buffers
151 /// or full search results.
152 pub fn into_handle(mut self, query: SearchQuery, cx: &mut App) -> SearchResultsHandle {
153 let mut open_buffers = HashSet::default();
154 let mut unnamed_buffers = Vec::new();
155 const MAX_CONCURRENT_BUFFER_OPENS: usize = 64;
156 let buffers = self.buffer_store.read(cx);
157 for handle in buffers.buffers() {
158 let buffer = handle.read(cx);
159 if !buffers.is_searchable(&buffer.remote_id()) {
160 continue;
161 } else if let Some(entry_id) = buffer.entry_id(cx) {
162 open_buffers.insert(entry_id);
163 } else {
164 self.limit = self.limit.saturating_sub(1);
165 unnamed_buffers.push(handle)
166 };
167 }
168 let executor = cx.background_executor().clone();
169 let (tx, rx) = unbounded();
170 let (grab_buffer_snapshot_tx, grab_buffer_snapshot_rx) = unbounded();
171 let matching_buffers = grab_buffer_snapshot_rx.clone();
172 let trigger_search = Box::new(move |cx: &mut App| {
173 cx.spawn(async move |cx| {
174 for buffer in unnamed_buffers {
175 _ = grab_buffer_snapshot_tx.send(buffer).await;
176 }
177
178 let (find_all_matches_tx, find_all_matches_rx) =
179 bounded(MAX_CONCURRENT_BUFFER_OPENS);
180 let query = Arc::new(query);
181 let (candidate_searcher, tasks) = match self.kind {
182 SearchKind::OpenBuffersOnly => {
183 let Ok(open_buffers) = cx.update(|cx| self.all_loaded_buffers(&query, cx))
184 else {
185 return;
186 };
187 let fill_requests = cx
188 .background_spawn(async move {
189 for buffer in open_buffers {
190 if let Err(_) = grab_buffer_snapshot_tx.send(buffer).await {
191 return;
192 }
193 }
194 })
195 .boxed_local();
196 (FindSearchCandidates::OpenBuffersOnly, vec![fill_requests])
197 }
198 SearchKind::Local {
199 fs,
200 ref mut worktrees,
201 } => {
202 let (get_buffer_for_full_scan_tx, get_buffer_for_full_scan_rx) =
203 unbounded();
204 let (confirm_contents_will_match_tx, confirm_contents_will_match_rx) =
205 bounded(64);
206 let (sorted_search_results_tx, sorted_search_results_rx) = unbounded();
207
208 let (input_paths_tx, input_paths_rx) = unbounded();
209 let tasks = vec![
210 cx.spawn(Self::provide_search_paths(
211 std::mem::take(worktrees),
212 query.clone(),
213 input_paths_tx,
214 sorted_search_results_tx,
215 ))
216 .boxed_local(),
217 Self::open_buffers(
218 &self.buffer_store,
219 get_buffer_for_full_scan_rx,
220 grab_buffer_snapshot_tx,
221 cx.clone(),
222 )
223 .boxed_local(),
224 cx.background_spawn(Self::maintain_sorted_search_results(
225 sorted_search_results_rx,
226 get_buffer_for_full_scan_tx,
227 self.limit,
228 ))
229 .boxed_local(),
230 ];
231 (
232 FindSearchCandidates::Local {
233 fs,
234 confirm_contents_will_match_tx,
235 confirm_contents_will_match_rx,
236 input_paths_rx,
237 },
238 tasks,
239 )
240 }
241 SearchKind::Remote {
242 client,
243 remote_id,
244 models,
245 } => {
246 let request = client.request(proto::FindSearchCandidates {
247 project_id: remote_id,
248 query: Some(query.to_proto()),
249 limit: self.limit as _,
250 });
251 let Ok(guard) = cx.update(|cx| {
252 Project::retain_remotely_created_models_impl(
253 &models,
254 &self.buffer_store,
255 &self.worktree_store,
256 cx,
257 )
258 }) else {
259 return;
260 };
261 let buffer_store = self.buffer_store.downgrade();
262 let issue_remote_buffers_request = cx
263 .spawn(async move |cx| {
264 let _ = maybe!(async move {
265 let response = request.await?;
266 for buffer_id in response.buffer_ids {
267 let buffer_id = BufferId::new(buffer_id)?;
268 let buffer = buffer_store
269 .update(cx, |buffer_store, cx| {
270 buffer_store.wait_for_remote_buffer(buffer_id, cx)
271 })?
272 .await?;
273 let _ = grab_buffer_snapshot_tx.send(buffer).await;
274 }
275
276 drop(guard);
277 anyhow::Ok(())
278 })
279 .await
280 .log_err();
281 })
282 .boxed_local();
283 (
284 FindSearchCandidates::Remote,
285 vec![issue_remote_buffers_request],
286 )
287 }
288 };
289
290 let should_find_all_matches = !tx.is_closed();
291
292 let worker_pool = executor.scoped(|scope| {
293 let num_cpus = executor.num_cpus();
294
295 assert!(num_cpus > 0);
296 for _ in 0..executor.num_cpus() - 1 {
297 let worker = Worker {
298 query: &query,
299 open_buffers: &open_buffers,
300 candidates: candidate_searcher.clone(),
301 find_all_matches_rx: find_all_matches_rx.clone(),
302 };
303 scope.spawn(worker.run());
304 }
305
306 drop(find_all_matches_rx);
307 drop(candidate_searcher);
308 });
309
310 let (sorted_matches_tx, sorted_matches_rx) = unbounded();
311 // The caller of `into_handle` decides whether they're interested in all matches (files that matched + all matching ranges) or
312 // just the files. *They are using the same stream as the guts of the project search do*.
313 // This means that we cannot grab values off of that stream unless it's strictly needed for making a progress in project search.
314 //
315 // Grabbing buffer snapshots is only necessary when we're looking for all matches. If the caller decided that they're not interested
316 // in all matches, running that task unconditionally would hinder caller's ability to observe all matching file paths.
317 let buffer_snapshots = if should_find_all_matches {
318 Some(
319 Self::grab_buffer_snapshots(
320 grab_buffer_snapshot_rx,
321 find_all_matches_tx,
322 sorted_matches_tx,
323 cx.clone(),
324 )
325 .boxed_local(),
326 )
327 } else {
328 drop(find_all_matches_tx);
329
330 None
331 };
332 let ensure_matches_are_reported_in_order = if should_find_all_matches {
333 Some(
334 Self::ensure_matched_ranges_are_reported_in_order(sorted_matches_rx, tx)
335 .boxed_local(),
336 )
337 } else {
338 drop(tx);
339 None
340 };
341
342 futures::future::join_all(
343 [worker_pool.boxed_local()]
344 .into_iter()
345 .chain(buffer_snapshots)
346 .chain(ensure_matches_are_reported_in_order)
347 .chain(tasks),
348 )
349 .await;
350 })
351 });
352
353 SearchResultsHandle {
354 results: rx,
355 matching_buffers,
356 trigger_search,
357 }
358 }
359
360 fn provide_search_paths(
361 worktrees: Vec<Entity<Worktree>>,
362 query: Arc<SearchQuery>,
363 tx: Sender<InputPath>,
364 results: Sender<oneshot::Receiver<ProjectPath>>,
365 ) -> impl AsyncFnOnce(&mut AsyncApp) {
366 async move |cx| {
367 _ = maybe!(async move {
368 let gitignored_tracker = PathInclusionMatcher::new(query.clone());
369 for worktree in worktrees {
370 let (mut snapshot, worktree_settings) = worktree
371 .read_with(cx, |this, _| {
372 Some((this.snapshot(), this.as_local()?.settings()))
373 })?
374 .context("The worktree is not local")?;
375 if query.include_ignored() {
376 // Pre-fetch all of the ignored directories as they're going to be searched.
377 let mut entries_to_refresh = vec![];
378
379 for entry in snapshot.entries(query.include_ignored(), 0) {
380 if gitignored_tracker.should_scan_gitignored_dir(
381 entry,
382 &snapshot,
383 &worktree_settings,
384 ) {
385 entries_to_refresh.push(entry.path.clone());
386 }
387 }
388 let barrier = worktree.update(cx, |this, _| {
389 let local = this.as_local_mut()?;
390 let barrier = entries_to_refresh
391 .into_iter()
392 .map(|path| local.add_path_prefix_to_scan(path).into_future())
393 .collect::<Vec<_>>();
394 Some(barrier)
395 })?;
396 if let Some(barriers) = barrier {
397 futures::future::join_all(barriers).await;
398 }
399 snapshot = worktree.read_with(cx, |this, _| this.snapshot())?;
400 }
401 cx.background_executor()
402 .scoped(|scope| {
403 scope.spawn(async {
404 for entry in snapshot.files(query.include_ignored(), 0) {
405 let (should_scan_tx, should_scan_rx) = oneshot::channel();
406
407 let Ok(_) = tx
408 .send(InputPath {
409 entry: entry.clone(),
410 snapshot: snapshot.clone(),
411 should_scan_tx,
412 })
413 .await
414 else {
415 return;
416 };
417 if results.send(should_scan_rx).await.is_err() {
418 return;
419 };
420 }
421 })
422 })
423 .await;
424 }
425 anyhow::Ok(())
426 })
427 .await;
428 }
429 }
430
431 async fn maintain_sorted_search_results(
432 rx: Receiver<oneshot::Receiver<ProjectPath>>,
433 paths_for_full_scan: Sender<ProjectPath>,
434 limit: usize,
435 ) {
436 let mut rx = pin!(rx);
437 let mut matched = 0;
438 while let Some(mut next_path_result) = rx.next().await {
439 let Some(successful_path) = next_path_result.next().await else {
440 // This file did not produce a match, hence skip it.
441 continue;
442 };
443 if paths_for_full_scan.send(successful_path).await.is_err() {
444 return;
445 };
446 matched += 1;
447 if matched >= limit {
448 break;
449 }
450 }
451 }
452
453 /// Background workers cannot open buffers by themselves, hence main thread will do it on their behalf.
454 async fn open_buffers(
455 buffer_store: &Entity<BufferStore>,
456 rx: Receiver<ProjectPath>,
457 find_all_matches_tx: Sender<Entity<Buffer>>,
458 mut cx: AsyncApp,
459 ) {
460 let mut rx = pin!(rx.ready_chunks(64));
461 _ = maybe!(async move {
462 while let Some(requested_paths) = rx.next().await {
463 let mut buffers = buffer_store.update(&mut cx, |this, cx| {
464 requested_paths
465 .into_iter()
466 .map(|path| this.open_buffer(path, cx))
467 .collect::<FuturesOrdered<_>>()
468 })?;
469
470 while let Some(buffer) = buffers.next().await {
471 if let Some(buffer) = buffer.log_err() {
472 find_all_matches_tx.send(buffer).await?;
473 }
474 }
475 }
476 Result::<_, anyhow::Error>::Ok(())
477 })
478 .await;
479 }
480
481 async fn grab_buffer_snapshots(
482 rx: Receiver<Entity<Buffer>>,
483 find_all_matches_tx: Sender<(
484 Entity<Buffer>,
485 BufferSnapshot,
486 oneshot::Sender<(Entity<Buffer>, Vec<Range<language::Anchor>>)>,
487 )>,
488 results: Sender<oneshot::Receiver<(Entity<Buffer>, Vec<Range<language::Anchor>>)>>,
489 mut cx: AsyncApp,
490 ) {
491 _ = maybe!(async move {
492 while let Ok(buffer) = rx.recv().await {
493 let snapshot = buffer.read_with(&mut cx, |this, _| this.snapshot())?;
494 let (tx, rx) = oneshot::channel();
495 find_all_matches_tx.send((buffer, snapshot, tx)).await?;
496 results.send(rx).await?;
497 }
498 debug_assert!(rx.is_empty());
499 Result::<_, anyhow::Error>::Ok(())
500 })
501 .await;
502 }
503
504 async fn ensure_matched_ranges_are_reported_in_order(
505 rx: Receiver<oneshot::Receiver<(Entity<Buffer>, Vec<Range<language::Anchor>>)>>,
506 tx: Sender<SearchResult>,
507 ) {
508 use postage::stream::Stream;
509 _ = maybe!(async move {
510 let mut matched_buffers = 0;
511 let mut matches = 0;
512 while let Ok(mut next_buffer_matches) = rx.recv().await {
513 let Some((buffer, ranges)) = next_buffer_matches.recv().await else {
514 continue;
515 };
516
517 if matched_buffers > Search::MAX_SEARCH_RESULT_FILES
518 || matches > Search::MAX_SEARCH_RESULT_RANGES
519 {
520 _ = tx.send(SearchResult::LimitReached).await;
521 break;
522 }
523 matched_buffers += 1;
524 matches += ranges.len();
525
526 _ = tx.send(SearchResult::Buffer { buffer, ranges }).await?;
527 }
528 anyhow::Ok(())
529 })
530 .await;
531 }
532
533 fn all_loaded_buffers(&self, search_query: &SearchQuery, cx: &App) -> Vec<Entity<Buffer>> {
534 let worktree_store = self.worktree_store.read(cx);
535 let mut buffers = search_query
536 .buffers()
537 .into_iter()
538 .flatten()
539 .filter(|buffer| {
540 let b = buffer.read(cx);
541 if let Some(file) = b.file() {
542 if !search_query.match_path(file.path()) {
543 return false;
544 }
545 if !search_query.include_ignored()
546 && let Some(entry) = b
547 .entry_id(cx)
548 .and_then(|entry_id| worktree_store.entry_for_id(entry_id, cx))
549 && entry.is_ignored
550 {
551 return false;
552 }
553 }
554 true
555 })
556 .cloned()
557 .collect::<Vec<_>>();
558 buffers.sort_by(|a, b| {
559 let a = a.read(cx);
560 let b = b.read(cx);
561 match (a.file(), b.file()) {
562 (None, None) => a.remote_id().cmp(&b.remote_id()),
563 (None, Some(_)) => std::cmp::Ordering::Less,
564 (Some(_), None) => std::cmp::Ordering::Greater,
565 (Some(a), Some(b)) => compare_rel_paths((a.path(), true), (b.path(), true)),
566 }
567 });
568
569 buffers
570 }
571}
572
573struct Worker<'search> {
574 query: &'search SearchQuery,
575 open_buffers: &'search HashSet<ProjectEntryId>,
576 candidates: FindSearchCandidates,
577 /// Ok, we're back in background: run full scan & find all matches in a given buffer snapshot.
578 /// Then, when you're done, share them via the channel you were given.
579 find_all_matches_rx: Receiver<(
580 Entity<Buffer>,
581 BufferSnapshot,
582 oneshot::Sender<(Entity<Buffer>, Vec<Range<language::Anchor>>)>,
583 )>,
584}
585
586impl Worker<'_> {
587 async fn run(self) {
588 let (
589 input_paths_rx,
590 confirm_contents_will_match_rx,
591 mut confirm_contents_will_match_tx,
592 fs,
593 ) = match self.candidates {
594 FindSearchCandidates::Local {
595 fs,
596 input_paths_rx,
597 confirm_contents_will_match_rx,
598 confirm_contents_will_match_tx,
599 } => (
600 input_paths_rx,
601 confirm_contents_will_match_rx,
602 confirm_contents_will_match_tx,
603 Some(fs),
604 ),
605 FindSearchCandidates::Remote | FindSearchCandidates::OpenBuffersOnly => {
606 (unbounded().1, unbounded().1, unbounded().0, None)
607 }
608 };
609 // WorkerA: grabs a request for "find all matches in file/a" <- takes 5 minutes
610 // right after: WorkerB: grabs a request for "find all matches in file/b" <- takes 5 seconds
611 let mut find_all_matches = pin!(self.find_all_matches_rx.fuse());
612 let mut find_first_match = pin!(confirm_contents_will_match_rx.fuse());
613 let mut scan_path = pin!(input_paths_rx.fuse());
614
615 loop {
616 let handler = RequestHandler {
617 query: self.query,
618 open_entries: &self.open_buffers,
619 fs: fs.as_deref(),
620 confirm_contents_will_match_tx: &confirm_contents_will_match_tx,
621 };
622 // Whenever we notice that some step of a pipeline is closed, we don't want to close subsequent
623 // steps straight away. Another worker might be about to produce a value that will
624 // be pushed there, thus we'll replace current worker's pipe with a dummy one.
625 // That way, we'll only ever close a next-stage channel when ALL workers do so.
626 select_biased! {
627 find_all_matches = find_all_matches.next() => {
628 let Some(matches) = find_all_matches else {
629 continue;
630 };
631 handler.handle_find_all_matches(matches).await;
632 },
633 find_first_match = find_first_match.next() => {
634 if let Some(buffer_with_at_least_one_match) = find_first_match {
635 handler.handle_find_first_match(buffer_with_at_least_one_match).await;
636 }
637 },
638 scan_path = scan_path.next() => {
639 if let Some(path_to_scan) = scan_path {
640 handler.handle_scan_path(path_to_scan).await;
641 } else {
642 // If we're the last worker to notice that this is not producing values, close the upstream.
643 confirm_contents_will_match_tx = bounded(1).0;
644 }
645
646 }
647 complete => {
648 break
649 },
650
651 }
652 }
653 }
654}
655
656struct RequestHandler<'worker> {
657 query: &'worker SearchQuery,
658 fs: Option<&'worker dyn Fs>,
659 open_entries: &'worker HashSet<ProjectEntryId>,
660 confirm_contents_will_match_tx: &'worker Sender<MatchingEntry>,
661}
662
663impl RequestHandler<'_> {
664 async fn handle_find_all_matches(
665 &self,
666 (buffer, snapshot, mut report_matches): (
667 Entity<Buffer>,
668 BufferSnapshot,
669 oneshot::Sender<(Entity<Buffer>, Vec<Range<language::Anchor>>)>,
670 ),
671 ) {
672 let ranges = self
673 .query
674 .search(&snapshot, None)
675 .await
676 .iter()
677 .map(|range| snapshot.anchor_before(range.start)..snapshot.anchor_after(range.end))
678 .collect::<Vec<_>>();
679
680 _ = report_matches.send((buffer, ranges)).await;
681 }
682
683 async fn handle_find_first_match(&self, mut entry: MatchingEntry) {
684 _=maybe!(async move {
685 let abs_path = entry.worktree_root.join(entry.path.path.as_std_path());
686
687 let file = self
688 .fs
689 .context("Trying to query filesystem in remote project search")?
690 .open_read(&abs_path)
691 .await?;
692
693 let mut file = futures::io::BufReader::new(file);
694
695 // Before attempting to match the file content, throw away files that have invalid UTF-8 sequences early on;
696 // That way we can still match files in a streaming fashion without having look at "obviously binary" files.
697 //
698 // Use `fill_buf` so we can "peek" without consuming bytes, and reuse the same stream for detection.
699 let buffer = file.fill_buf().await?;
700 let prefix_len = buffer.len().min(8192);
701
702 if let Err(error) = std::str::from_utf8(&buffer[..prefix_len]) {
703 if let Some(starting_position) = error.error_len() {
704 log::debug!(
705 "Invalid UTF-8 sequence in file {abs_path:?} at byte position {starting_position}"
706 );
707 return Ok(());
708 }
709 }
710
711 if self.query.detect(file).await.unwrap_or(false) {
712 // Yes, we should scan the whole file.
713 entry.should_scan_tx.send(entry.path).await?;
714 }
715
716 anyhow::Ok(())
717 }).await;
718 }
719
720 async fn handle_scan_path(&self, req: InputPath) {
721 _ = maybe!(async move {
722 let InputPath {
723 entry,
724 snapshot,
725 mut should_scan_tx,
726 } = req;
727
728 if entry.is_fifo || !entry.is_file() {
729 return Ok(());
730 }
731
732 if self.query.filters_path() {
733 let matched_path = if self.query.match_full_paths() {
734 let mut full_path = snapshot.root_name().to_owned();
735 full_path.push(&entry.path);
736 self.query.match_path(&full_path)
737 } else {
738 self.query.match_path(&entry.path)
739 };
740 if !matched_path {
741 return Ok(());
742 }
743 }
744
745 if self.open_entries.contains(&entry.id) {
746 // The buffer is already in memory and that's the version we want to scan;
747 // hence skip the dilly-dally and look for all matches straight away.
748 should_scan_tx
749 .send(ProjectPath {
750 worktree_id: snapshot.id(),
751 path: entry.path.clone(),
752 })
753 .await?;
754 } else {
755 self.confirm_contents_will_match_tx
756 .send(MatchingEntry {
757 should_scan_tx: should_scan_tx,
758 worktree_root: snapshot.abs_path().clone(),
759 path: ProjectPath {
760 worktree_id: snapshot.id(),
761 path: entry.path.clone(),
762 },
763 })
764 .await?;
765 }
766
767 anyhow::Ok(())
768 })
769 .await;
770 }
771}
772
773struct InputPath {
774 entry: Entry,
775 snapshot: Snapshot,
776 should_scan_tx: oneshot::Sender<ProjectPath>,
777}
778
779struct MatchingEntry {
780 worktree_root: Arc<Path>,
781 path: ProjectPath,
782 should_scan_tx: oneshot::Sender<ProjectPath>,
783}
784
785/// This struct encapsulates the logic to decide whether a given gitignored directory should be
786/// scanned based on include/exclude patterns of a search query (as include/exclude parameters may match paths inside it).
787/// It is kind-of doing an inverse of glob. Given a glob pattern like `src/**/` and a parent path like `src`, we need to decide whether the parent
788/// may contain glob hits.
789struct PathInclusionMatcher {
790 included: BTreeSet<PathBuf>,
791 query: Arc<SearchQuery>,
792}
793
794impl PathInclusionMatcher {
795 fn new(query: Arc<SearchQuery>) -> Self {
796 let mut included = BTreeSet::new();
797 // To do an inverse glob match, we split each glob into it's prefix and the glob part.
798 // For example, `src/**/*.rs` becomes `src/` and `**/*.rs`. The glob part gets dropped.
799 // Then, when checking whether a given directory should be scanned, we check whether it is a non-empty substring of any glob prefix.
800 if query.filters_path() {
801 included.extend(
802 query
803 .files_to_include()
804 .sources()
805 .flat_map(|glob| Some(wax::Glob::new(glob).ok()?.partition().0)),
806 );
807 }
808 Self { included, query }
809 }
810
811 fn should_scan_gitignored_dir(
812 &self,
813 entry: &Entry,
814 snapshot: &Snapshot,
815 worktree_settings: &WorktreeSettings,
816 ) -> bool {
817 if !entry.is_ignored || !entry.kind.is_unloaded() {
818 return false;
819 }
820 if !self.query.include_ignored() {
821 return false;
822 }
823 if worktree_settings.is_path_excluded(&entry.path) {
824 return false;
825 }
826 if !self.query.filters_path() {
827 return true;
828 }
829
830 let as_abs_path = LazyCell::new(move || snapshot.absolutize(&entry.path));
831 let entry_path = &entry.path;
832 // 3. Check Exclusions (Pruning)
833 // If the current path is a child of an excluded path, we stop.
834 let is_excluded = self.path_is_definitely_excluded(&entry_path, snapshot);
835
836 if is_excluded {
837 return false;
838 }
839
840 // 4. Check Inclusions (Traversal)
841 if self.included.is_empty() {
842 return true;
843 }
844
845 // We scan if the current path is a descendant of an include prefix
846 // OR if the current path is an ancestor of an include prefix (we need to go deeper to find it).
847 let is_included = self.included.iter().any(|prefix| {
848 let (prefix_matches_entry, entry_matches_prefix) = if prefix.is_absolute() {
849 (
850 prefix.starts_with(&**as_abs_path),
851 as_abs_path.starts_with(prefix),
852 )
853 } else {
854 RelPath::new(prefix, snapshot.path_style()).map_or((false, false), |prefix| {
855 (
856 prefix.starts_with(entry_path),
857 entry_path.starts_with(&prefix),
858 )
859 })
860 };
861
862 // Logic:
863 // 1. entry_matches_prefix: We are inside the target zone (e.g. glob: src/, current: src/lib/). Keep scanning.
864 // 2. prefix_matches_entry: We are above the target zone (e.g. glob: src/foo/, current: src/). Keep scanning to reach foo.
865 prefix_matches_entry || entry_matches_prefix
866 });
867
868 is_included
869 }
870 fn path_is_definitely_excluded(&self, path: &RelPath, snapshot: &Snapshot) -> bool {
871 if !self.query.files_to_exclude().sources().next().is_none() {
872 let mut path = if self.query.match_full_paths() {
873 let mut full_path = snapshot.root_name().to_owned();
874 full_path.push(path);
875 full_path
876 } else {
877 path.to_owned()
878 };
879 loop {
880 if self.query.files_to_exclude().is_match(&path) {
881 return true;
882 } else if !path.pop() {
883 return false;
884 }
885 }
886 } else {
887 false
888 }
889 }
890}
891
892#[cfg(test)]
893mod tests {
894 use super::*;
895 use fs::FakeFs;
896 use serde_json::json;
897 use settings::Settings;
898 use util::{
899 path,
900 paths::{PathMatcher, PathStyle},
901 rel_path::RelPath,
902 };
903 use worktree::{Entry, EntryKind, WorktreeSettings};
904
905 use crate::{
906 Project, project_search::PathInclusionMatcher, project_tests::init_test,
907 search::SearchQuery,
908 };
909
910 #[gpui::test]
911 async fn test_path_inclusion_matcher(cx: &mut gpui::TestAppContext) {
912 init_test(cx);
913
914 let fs = FakeFs::new(cx.background_executor.clone());
915 fs.insert_tree(
916 "/root",
917 json!({
918 ".gitignore": "src/data/\n",
919 "src": {
920 "data": {
921 "main.csv": "field_1,field_2,field_3",
922 },
923 "lib": {
924 "main.txt": "Are you familiar with fields?",
925 },
926 },
927 }),
928 )
929 .await;
930
931 let project = Project::test(fs.clone(), [path!("/root").as_ref()], cx).await;
932 let worktree = project.update(cx, |project, cx| project.worktrees(cx).next().unwrap());
933 let (worktree_settings, worktree_snapshot) = worktree.update(cx, |worktree, cx| {
934 let settings_location = worktree.settings_location(cx);
935 return (
936 WorktreeSettings::get(Some(settings_location), cx).clone(),
937 worktree.snapshot(),
938 );
939 });
940
941 // Manually create a test entry for the gitignored directory since it won't
942 // be loaded by the worktree
943 let entry = Entry {
944 id: ProjectEntryId::from_proto(1),
945 kind: EntryKind::UnloadedDir,
946 path: Arc::from(RelPath::unix(Path::new("src/data")).unwrap()),
947 inode: 0,
948 mtime: None,
949 canonical_path: None,
950 is_ignored: true,
951 is_hidden: false,
952 is_always_included: false,
953 is_external: false,
954 is_private: false,
955 size: 0,
956 char_bag: Default::default(),
957 is_fifo: false,
958 };
959
960 // 1. Test searching for `field`, including ignored files without any
961 // inclusion and exclusion filters.
962 let include_ignored = true;
963 let files_to_include = PathMatcher::default();
964 let files_to_exclude = PathMatcher::default();
965 let match_full_paths = false;
966 let search_query = SearchQuery::text(
967 "field",
968 false,
969 false,
970 include_ignored,
971 files_to_include,
972 files_to_exclude,
973 match_full_paths,
974 None,
975 )
976 .unwrap();
977
978 let path_matcher = PathInclusionMatcher::new(Arc::new(search_query));
979 assert!(path_matcher.should_scan_gitignored_dir(
980 &entry,
981 &worktree_snapshot,
982 &worktree_settings
983 ));
984
985 // 2. Test searching for `field`, including ignored files but updating
986 // `files_to_include` to only include files under `src/lib`.
987 let include_ignored = true;
988 let files_to_include = PathMatcher::new(vec!["src/lib"], PathStyle::Posix).unwrap();
989 let files_to_exclude = PathMatcher::default();
990 let match_full_paths = false;
991 let search_query = SearchQuery::text(
992 "field",
993 false,
994 false,
995 include_ignored,
996 files_to_include,
997 files_to_exclude,
998 match_full_paths,
999 None,
1000 )
1001 .unwrap();
1002
1003 let path_matcher = PathInclusionMatcher::new(Arc::new(search_query));
1004 assert!(!path_matcher.should_scan_gitignored_dir(
1005 &entry,
1006 &worktree_snapshot,
1007 &worktree_settings
1008 ));
1009 }
1010}