1mod ignore;
2mod worktree_settings;
3
4use ::ignore::gitignore::{Gitignore, GitignoreBuilder};
5use anyhow::{Context as _, Result, anyhow};
6use chardetng::EncodingDetector;
7use clock::ReplicaId;
8use collections::{HashMap, HashSet, VecDeque};
9use encoding_rs::Encoding;
10use fs::{Fs, MTime, PathEvent, RemoveOptions, Watcher, copy_recursive, read_dir_items};
11use futures::{
12 FutureExt as _, Stream, StreamExt,
13 channel::{
14 mpsc::{self, UnboundedSender},
15 oneshot,
16 },
17 select_biased, stream,
18 task::Poll,
19};
20use fuzzy::CharBag;
21use git::{
22 COMMIT_MESSAGE, DOT_GIT, FSMONITOR_DAEMON, GITIGNORE, INDEX_LOCK, LFS_DIR, REPO_EXCLUDE,
23 status::GitSummary,
24};
25use gpui::{
26 App, AppContext as _, AsyncApp, BackgroundExecutor, Context, Entity, EventEmitter, Priority,
27 Task,
28};
29use ignore::IgnoreStack;
30use language::DiskState;
31
32use parking_lot::Mutex;
33use paths::{local_settings_folder_name, local_vscode_folder_name};
34use postage::{
35 barrier,
36 prelude::{Sink as _, Stream as _},
37 watch,
38};
39use rpc::{
40 AnyProtoClient,
41 proto::{self, split_worktree_update},
42};
43pub use settings::WorktreeId;
44use settings::{Settings, SettingsLocation, SettingsStore};
45use smallvec::{SmallVec, smallvec};
46use smol::channel::{self, Sender};
47use std::{
48 any::Any,
49 borrow::Borrow as _,
50 cmp::Ordering,
51 collections::hash_map,
52 convert::TryFrom,
53 ffi::OsStr,
54 fmt,
55 future::Future,
56 mem::{self},
57 ops::{Deref, DerefMut, Range},
58 path::{Path, PathBuf},
59 pin::Pin,
60 sync::{
61 Arc,
62 atomic::{AtomicUsize, Ordering::SeqCst},
63 },
64 time::{Duration, Instant},
65};
66use sum_tree::{Bias, Dimensions, Edit, KeyedItem, SeekTarget, SumTree, Summary, TreeMap, TreeSet};
67use text::{LineEnding, Rope};
68use util::{
69 ResultExt, debug_panic, maybe,
70 paths::{PathMatcher, PathStyle, SanitizedPath, home_dir},
71 rel_path::RelPath,
72};
73pub use worktree_settings::WorktreeSettings;
74
75use crate::ignore::IgnoreKind;
76
77pub const FS_WATCH_LATENCY: Duration = Duration::from_millis(100);
78
79/// A set of local or remote files that are being opened as part of a project.
80/// Responsible for tracking related FS (for local)/collab (for remote) events and corresponding updates.
81/// Stores git repositories data and the diagnostics for the file(s).
82///
83/// Has an absolute path, and may be set to be visible in Zed UI or not.
84/// May correspond to a directory or a single file.
85/// Possible examples:
86/// * a drag and dropped file — may be added as an invisible, "ephemeral" entry to the current worktree
87/// * a directory opened in Zed — may be added as a visible entry to the current worktree
88///
89/// Uses [`Entry`] to track the state of each file/directory, can look up absolute paths for entries.
90pub enum Worktree {
91 Local(LocalWorktree),
92 Remote(RemoteWorktree),
93}
94
95/// An entry, created in the worktree.
96#[derive(Debug)]
97pub enum CreatedEntry {
98 /// Got created and indexed by the worktree, receiving a corresponding entry.
99 Included(Entry),
100 /// Got created, but not indexed due to falling under exclusion filters.
101 Excluded { abs_path: PathBuf },
102}
103
104#[derive(Debug)]
105pub struct LoadedFile {
106 pub file: Arc<File>,
107 pub text: String,
108 pub encoding: &'static Encoding,
109 pub has_bom: bool,
110}
111
112pub struct LoadedBinaryFile {
113 pub file: Arc<File>,
114 pub content: Vec<u8>,
115}
116
117impl fmt::Debug for LoadedBinaryFile {
118 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
119 f.debug_struct("LoadedBinaryFile")
120 .field("file", &self.file)
121 .field("content_bytes", &self.content.len())
122 .finish()
123 }
124}
125
126pub struct LocalWorktree {
127 snapshot: LocalSnapshot,
128 scan_requests_tx: channel::Sender<ScanRequest>,
129 path_prefixes_to_scan_tx: channel::Sender<PathPrefixScanRequest>,
130 is_scanning: (watch::Sender<bool>, watch::Receiver<bool>),
131 _background_scanner_tasks: Vec<Task<()>>,
132 update_observer: Option<UpdateObservationState>,
133 fs: Arc<dyn Fs>,
134 fs_case_sensitive: bool,
135 visible: bool,
136 next_entry_id: Arc<AtomicUsize>,
137 settings: WorktreeSettings,
138 share_private_files: bool,
139 scanning_enabled: bool,
140}
141
142pub struct PathPrefixScanRequest {
143 path: Arc<RelPath>,
144 done: SmallVec<[barrier::Sender; 1]>,
145}
146
147struct ScanRequest {
148 relative_paths: Vec<Arc<RelPath>>,
149 done: SmallVec<[barrier::Sender; 1]>,
150}
151
152pub struct RemoteWorktree {
153 snapshot: Snapshot,
154 background_snapshot: Arc<Mutex<(Snapshot, Vec<proto::UpdateWorktree>)>>,
155 project_id: u64,
156 client: AnyProtoClient,
157 file_scan_inclusions: PathMatcher,
158 updates_tx: Option<UnboundedSender<proto::UpdateWorktree>>,
159 update_observer: Option<mpsc::UnboundedSender<proto::UpdateWorktree>>,
160 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
161 replica_id: ReplicaId,
162 visible: bool,
163 disconnected: bool,
164}
165
166#[derive(Clone)]
167pub struct Snapshot {
168 id: WorktreeId,
169 /// The absolute path of the worktree root.
170 abs_path: Arc<SanitizedPath>,
171 path_style: PathStyle,
172 root_name: Arc<RelPath>,
173 root_char_bag: CharBag,
174 entries_by_path: SumTree<Entry>,
175 entries_by_id: SumTree<PathEntry>,
176 always_included_entries: Vec<Arc<RelPath>>,
177
178 /// A number that increases every time the worktree begins scanning
179 /// a set of paths from the filesystem. This scanning could be caused
180 /// by some operation performed on the worktree, such as reading or
181 /// writing a file, or by an event reported by the filesystem.
182 scan_id: usize,
183
184 /// The latest scan id that has completed, and whose preceding scans
185 /// have all completed. The current `scan_id` could be more than one
186 /// greater than the `completed_scan_id` if operations are performed
187 /// on the worktree while it is processing a file-system event.
188 completed_scan_id: usize,
189}
190
191/// This path corresponds to the 'content path' of a repository in relation
192/// to Zed's project root.
193/// In the majority of the cases, this is the folder that contains the .git folder.
194/// But if a sub-folder of a git repository is opened, this corresponds to the
195/// project root and the .git folder is located in a parent directory.
196#[derive(Clone, Debug, Ord, PartialOrd, Eq, PartialEq, Hash)]
197pub enum WorkDirectory {
198 InProject {
199 relative_path: Arc<RelPath>,
200 },
201 AboveProject {
202 absolute_path: Arc<Path>,
203 location_in_repo: Arc<Path>,
204 },
205}
206
207impl WorkDirectory {
208 fn path_key(&self) -> PathKey {
209 match self {
210 WorkDirectory::InProject { relative_path } => PathKey(relative_path.clone()),
211 WorkDirectory::AboveProject { .. } => PathKey(RelPath::empty().into()),
212 }
213 }
214
215 /// Returns true if the given path is a child of the work directory.
216 ///
217 /// Note that the path may not be a member of this repository, if there
218 /// is a repository in a directory between these two paths
219 /// external .git folder in a parent folder of the project root.
220 #[track_caller]
221 pub fn directory_contains(&self, path: &RelPath) -> bool {
222 match self {
223 WorkDirectory::InProject { relative_path } => path.starts_with(relative_path),
224 WorkDirectory::AboveProject { .. } => true,
225 }
226 }
227}
228
229impl Default for WorkDirectory {
230 fn default() -> Self {
231 Self::InProject {
232 relative_path: Arc::from(RelPath::empty()),
233 }
234 }
235}
236
237#[derive(Clone)]
238pub struct LocalSnapshot {
239 snapshot: Snapshot,
240 global_gitignore: Option<Arc<Gitignore>>,
241 /// Exclude files for all git repositories in the worktree, indexed by their absolute path.
242 /// The boolean indicates whether the gitignore needs to be updated.
243 repo_exclude_by_work_dir_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
244 /// All of the gitignore files in the worktree, indexed by their absolute path.
245 /// The boolean indicates whether the gitignore needs to be updated.
246 ignores_by_parent_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
247 /// All of the git repositories in the worktree, indexed by the project entry
248 /// id of their parent directory.
249 git_repositories: TreeMap<ProjectEntryId, LocalRepositoryEntry>,
250 /// The file handle of the worktree root
251 /// (so we can find it after it's been moved)
252 root_file_handle: Option<Arc<dyn fs::FileHandle>>,
253}
254
255struct BackgroundScannerState {
256 snapshot: LocalSnapshot,
257 scanned_dirs: HashSet<ProjectEntryId>,
258 path_prefixes_to_scan: HashSet<Arc<RelPath>>,
259 paths_to_scan: HashSet<Arc<RelPath>>,
260 /// The ids of all of the entries that were removed from the snapshot
261 /// as part of the current update. These entry ids may be re-used
262 /// if the same inode is discovered at a new path, or if the given
263 /// path is re-created after being deleted.
264 removed_entries: HashMap<u64, Entry>,
265 changed_paths: Vec<Arc<RelPath>>,
266 prev_snapshot: Snapshot,
267 scanning_enabled: bool,
268}
269
270#[derive(Debug, Clone)]
271struct LocalRepositoryEntry {
272 work_directory_id: ProjectEntryId,
273 work_directory: WorkDirectory,
274 work_directory_abs_path: Arc<Path>,
275 git_dir_scan_id: usize,
276 /// Absolute path to the original .git entry that caused us to create this repository.
277 ///
278 /// This is normally a directory, but may be a "gitfile" that points to a directory elsewhere
279 /// (whose path we then store in `repository_dir_abs_path`).
280 dot_git_abs_path: Arc<Path>,
281 /// Absolute path to the "commondir" for this repository.
282 ///
283 /// This is always a directory. For a normal repository, this is the same as dot_git_abs_path,
284 /// but in the case of a submodule or a worktree it is the path to the "parent" .git directory
285 /// from which the submodule/worktree was derived.
286 common_dir_abs_path: Arc<Path>,
287 /// Absolute path to the directory holding the repository's state.
288 ///
289 /// For a normal repository, this is a directory and coincides with `dot_git_abs_path` and
290 /// `common_dir_abs_path`. For a submodule or worktree, this is some subdirectory of the
291 /// commondir like `/project/.git/modules/foo`.
292 repository_dir_abs_path: Arc<Path>,
293}
294
295impl sum_tree::Item for LocalRepositoryEntry {
296 type Summary = PathSummary<sum_tree::NoSummary>;
297
298 fn summary(&self, _: <Self::Summary as Summary>::Context<'_>) -> Self::Summary {
299 PathSummary {
300 max_path: self.work_directory.path_key().0,
301 item_summary: sum_tree::NoSummary,
302 }
303 }
304}
305
306impl KeyedItem for LocalRepositoryEntry {
307 type Key = PathKey;
308
309 fn key(&self) -> Self::Key {
310 self.work_directory.path_key()
311 }
312}
313
314impl Deref for LocalRepositoryEntry {
315 type Target = WorkDirectory;
316
317 fn deref(&self) -> &Self::Target {
318 &self.work_directory
319 }
320}
321
322impl Deref for LocalSnapshot {
323 type Target = Snapshot;
324
325 fn deref(&self) -> &Self::Target {
326 &self.snapshot
327 }
328}
329
330impl DerefMut for LocalSnapshot {
331 fn deref_mut(&mut self) -> &mut Self::Target {
332 &mut self.snapshot
333 }
334}
335
336enum ScanState {
337 Started,
338 Updated {
339 snapshot: LocalSnapshot,
340 changes: UpdatedEntriesSet,
341 barrier: SmallVec<[barrier::Sender; 1]>,
342 scanning: bool,
343 },
344 RootUpdated {
345 new_path: Arc<SanitizedPath>,
346 },
347}
348
349struct UpdateObservationState {
350 snapshots_tx: mpsc::UnboundedSender<(LocalSnapshot, UpdatedEntriesSet)>,
351 resume_updates: watch::Sender<()>,
352 _maintain_remote_snapshot: Task<Option<()>>,
353}
354
355#[derive(Debug, Clone)]
356pub enum Event {
357 UpdatedEntries(UpdatedEntriesSet),
358 UpdatedGitRepositories(UpdatedGitRepositoriesSet),
359 DeletedEntry(ProjectEntryId),
360}
361
362impl EventEmitter<Event> for Worktree {}
363
364impl Worktree {
365 pub async fn local(
366 path: impl Into<Arc<Path>>,
367 visible: bool,
368 fs: Arc<dyn Fs>,
369 next_entry_id: Arc<AtomicUsize>,
370 scanning_enabled: bool,
371 cx: &mut AsyncApp,
372 ) -> Result<Entity<Self>> {
373 let abs_path = path.into();
374 let metadata = fs
375 .metadata(&abs_path)
376 .await
377 .context("failed to stat worktree path")?;
378
379 let fs_case_sensitive = fs.is_case_sensitive().await.unwrap_or_else(|e| {
380 log::error!(
381 "Failed to determine whether filesystem is case sensitive (falling back to true) due to error: {e:#}"
382 );
383 true
384 });
385
386 let root_file_handle = if metadata.as_ref().is_some() {
387 fs.open_handle(&abs_path)
388 .await
389 .with_context(|| {
390 format!(
391 "failed to open local worktree root at {}",
392 abs_path.display()
393 )
394 })
395 .log_err()
396 } else {
397 None
398 };
399
400 Ok(cx.new(move |cx: &mut Context<Worktree>| {
401 let mut snapshot = LocalSnapshot {
402 ignores_by_parent_abs_path: Default::default(),
403 global_gitignore: Default::default(),
404 repo_exclude_by_work_dir_abs_path: Default::default(),
405 git_repositories: Default::default(),
406 snapshot: Snapshot::new(
407 cx.entity_id().as_u64(),
408 abs_path
409 .file_name()
410 .and_then(|f| f.to_str())
411 .map_or(RelPath::empty().into(), |f| {
412 RelPath::unix(f).unwrap().into()
413 }),
414 abs_path.clone(),
415 PathStyle::local(),
416 ),
417 root_file_handle,
418 };
419
420 let worktree_id = snapshot.id();
421 let settings_location = Some(SettingsLocation {
422 worktree_id,
423 path: RelPath::empty(),
424 });
425
426 let settings = WorktreeSettings::get(settings_location, cx).clone();
427 cx.observe_global::<SettingsStore>(move |this, cx| {
428 if let Self::Local(this) = this {
429 let settings = WorktreeSettings::get(settings_location, cx).clone();
430 if this.settings != settings {
431 this.settings = settings;
432 this.restart_background_scanners(cx);
433 }
434 }
435 })
436 .detach();
437
438 let share_private_files = false;
439 if let Some(metadata) = metadata {
440 let mut entry = Entry::new(
441 RelPath::empty().into(),
442 &metadata,
443 ProjectEntryId::new(&next_entry_id),
444 snapshot.root_char_bag,
445 None,
446 );
447 if metadata.is_dir {
448 if !scanning_enabled {
449 entry.kind = EntryKind::UnloadedDir;
450 }
451 } else {
452 if let Some(file_name) = abs_path.file_name()
453 && let Some(file_name) = file_name.to_str()
454 && let Ok(path) = RelPath::unix(file_name)
455 {
456 entry.is_private = !share_private_files && settings.is_path_private(path);
457 entry.is_hidden = settings.is_path_hidden(path);
458 }
459 }
460 cx.foreground_executor()
461 .block_on(snapshot.insert_entry(entry, fs.as_ref()));
462 }
463
464 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
465 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
466 let mut worktree = LocalWorktree {
467 share_private_files,
468 next_entry_id,
469 snapshot,
470 is_scanning: watch::channel_with(true),
471 update_observer: None,
472 scan_requests_tx,
473 path_prefixes_to_scan_tx,
474 _background_scanner_tasks: Vec::new(),
475 fs,
476 fs_case_sensitive,
477 visible,
478 settings,
479 scanning_enabled,
480 };
481 worktree.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
482 Worktree::Local(worktree)
483 }))
484 }
485
486 pub fn remote(
487 project_id: u64,
488 replica_id: ReplicaId,
489 worktree: proto::WorktreeMetadata,
490 client: AnyProtoClient,
491 path_style: PathStyle,
492 cx: &mut App,
493 ) -> Entity<Self> {
494 cx.new(|cx: &mut Context<Self>| {
495 let snapshot = Snapshot::new(
496 worktree.id,
497 RelPath::from_proto(&worktree.root_name)
498 .unwrap_or_else(|_| RelPath::empty().into()),
499 Path::new(&worktree.abs_path).into(),
500 path_style,
501 );
502
503 let background_snapshot = Arc::new(Mutex::new((
504 snapshot.clone(),
505 Vec::<proto::UpdateWorktree>::new(),
506 )));
507 let (background_updates_tx, mut background_updates_rx) =
508 mpsc::unbounded::<proto::UpdateWorktree>();
509 let (mut snapshot_updated_tx, mut snapshot_updated_rx) = watch::channel();
510
511 let worktree_id = snapshot.id();
512 let settings_location = Some(SettingsLocation {
513 worktree_id,
514 path: RelPath::empty(),
515 });
516
517 let settings = WorktreeSettings::get(settings_location, cx).clone();
518 let worktree = RemoteWorktree {
519 client,
520 project_id,
521 replica_id,
522 snapshot,
523 file_scan_inclusions: settings.parent_dir_scan_inclusions.clone(),
524 background_snapshot: background_snapshot.clone(),
525 updates_tx: Some(background_updates_tx),
526 update_observer: None,
527 snapshot_subscriptions: Default::default(),
528 visible: worktree.visible,
529 disconnected: false,
530 };
531
532 // Apply updates to a separate snapshot in a background task, then
533 // send them to a foreground task which updates the model.
534 cx.background_spawn(async move {
535 while let Some(update) = background_updates_rx.next().await {
536 {
537 let mut lock = background_snapshot.lock();
538 lock.0.apply_remote_update(
539 update.clone(),
540 &settings.parent_dir_scan_inclusions,
541 );
542 lock.1.push(update);
543 }
544 snapshot_updated_tx.send(()).await.ok();
545 }
546 })
547 .detach();
548
549 // On the foreground task, update to the latest snapshot and notify
550 // any update observer of all updates that led to that snapshot.
551 cx.spawn(async move |this, cx| {
552 while (snapshot_updated_rx.recv().await).is_some() {
553 this.update(cx, |this, cx| {
554 let mut entries_changed = false;
555 let this = this.as_remote_mut().unwrap();
556 {
557 let mut lock = this.background_snapshot.lock();
558 this.snapshot = lock.0.clone();
559 for update in lock.1.drain(..) {
560 entries_changed |= !update.updated_entries.is_empty()
561 || !update.removed_entries.is_empty();
562 if let Some(tx) = &this.update_observer {
563 tx.unbounded_send(update).ok();
564 }
565 }
566 };
567
568 if entries_changed {
569 cx.emit(Event::UpdatedEntries(Arc::default()));
570 }
571 cx.notify();
572 while let Some((scan_id, _)) = this.snapshot_subscriptions.front() {
573 if this.observed_snapshot(*scan_id) {
574 let (_, tx) = this.snapshot_subscriptions.pop_front().unwrap();
575 let _ = tx.send(());
576 } else {
577 break;
578 }
579 }
580 })?;
581 }
582 anyhow::Ok(())
583 })
584 .detach();
585
586 Worktree::Remote(worktree)
587 })
588 }
589
590 pub fn as_local(&self) -> Option<&LocalWorktree> {
591 if let Worktree::Local(worktree) = self {
592 Some(worktree)
593 } else {
594 None
595 }
596 }
597
598 pub fn as_remote(&self) -> Option<&RemoteWorktree> {
599 if let Worktree::Remote(worktree) = self {
600 Some(worktree)
601 } else {
602 None
603 }
604 }
605
606 pub fn as_local_mut(&mut self) -> Option<&mut LocalWorktree> {
607 if let Worktree::Local(worktree) = self {
608 Some(worktree)
609 } else {
610 None
611 }
612 }
613
614 pub fn as_remote_mut(&mut self) -> Option<&mut RemoteWorktree> {
615 if let Worktree::Remote(worktree) = self {
616 Some(worktree)
617 } else {
618 None
619 }
620 }
621
622 pub fn is_local(&self) -> bool {
623 matches!(self, Worktree::Local(_))
624 }
625
626 pub fn is_remote(&self) -> bool {
627 !self.is_local()
628 }
629
630 pub fn settings_location(&self, _: &Context<Self>) -> SettingsLocation<'static> {
631 SettingsLocation {
632 worktree_id: self.id(),
633 path: RelPath::empty(),
634 }
635 }
636
637 pub fn snapshot(&self) -> Snapshot {
638 match self {
639 Worktree::Local(worktree) => worktree.snapshot.snapshot.clone(),
640 Worktree::Remote(worktree) => worktree.snapshot.clone(),
641 }
642 }
643
644 pub fn scan_id(&self) -> usize {
645 match self {
646 Worktree::Local(worktree) => worktree.snapshot.scan_id,
647 Worktree::Remote(worktree) => worktree.snapshot.scan_id,
648 }
649 }
650
651 pub fn metadata_proto(&self) -> proto::WorktreeMetadata {
652 proto::WorktreeMetadata {
653 id: self.id().to_proto(),
654 root_name: self.root_name().to_proto(),
655 visible: self.is_visible(),
656 abs_path: self.abs_path().to_string_lossy().into_owned(),
657 }
658 }
659
660 pub fn completed_scan_id(&self) -> usize {
661 match self {
662 Worktree::Local(worktree) => worktree.snapshot.completed_scan_id,
663 Worktree::Remote(worktree) => worktree.snapshot.completed_scan_id,
664 }
665 }
666
667 pub fn is_visible(&self) -> bool {
668 match self {
669 Worktree::Local(worktree) => worktree.visible,
670 Worktree::Remote(worktree) => worktree.visible,
671 }
672 }
673
674 pub fn replica_id(&self) -> ReplicaId {
675 match self {
676 Worktree::Local(_) => ReplicaId::LOCAL,
677 Worktree::Remote(worktree) => worktree.replica_id,
678 }
679 }
680
681 pub fn abs_path(&self) -> Arc<Path> {
682 match self {
683 Worktree::Local(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
684 Worktree::Remote(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
685 }
686 }
687
688 pub fn root_file(&self, cx: &Context<Self>) -> Option<Arc<File>> {
689 let entry = self.root_entry()?;
690 Some(File::for_entry(entry.clone(), cx.entity()))
691 }
692
693 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
694 where
695 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
696 Fut: 'static + Send + Future<Output = bool>,
697 {
698 match self {
699 Worktree::Local(this) => this.observe_updates(project_id, cx, callback),
700 Worktree::Remote(this) => this.observe_updates(project_id, cx, callback),
701 }
702 }
703
704 pub fn stop_observing_updates(&mut self) {
705 match self {
706 Worktree::Local(this) => {
707 this.update_observer.take();
708 }
709 Worktree::Remote(this) => {
710 this.update_observer.take();
711 }
712 }
713 }
714
715 #[cfg(feature = "test-support")]
716 pub fn has_update_observer(&self) -> bool {
717 match self {
718 Worktree::Local(this) => this.update_observer.is_some(),
719 Worktree::Remote(this) => this.update_observer.is_some(),
720 }
721 }
722
723 pub fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
724 match self {
725 Worktree::Local(this) => this.load_file(path, cx),
726 Worktree::Remote(_) => {
727 Task::ready(Err(anyhow!("remote worktrees can't yet load files")))
728 }
729 }
730 }
731
732 pub fn load_binary_file(
733 &self,
734 path: &RelPath,
735 cx: &Context<Worktree>,
736 ) -> Task<Result<LoadedBinaryFile>> {
737 match self {
738 Worktree::Local(this) => this.load_binary_file(path, cx),
739 Worktree::Remote(_) => {
740 Task::ready(Err(anyhow!("remote worktrees can't yet load binary files")))
741 }
742 }
743 }
744
745 pub fn write_file(
746 &self,
747 path: Arc<RelPath>,
748 text: Rope,
749 line_ending: LineEnding,
750 encoding: &'static Encoding,
751 has_bom: bool,
752 cx: &Context<Worktree>,
753 ) -> Task<Result<Arc<File>>> {
754 match self {
755 Worktree::Local(this) => {
756 this.write_file(path, text, line_ending, encoding, has_bom, cx)
757 }
758 Worktree::Remote(_) => {
759 Task::ready(Err(anyhow!("remote worktree can't yet write files")))
760 }
761 }
762 }
763
764 pub fn create_entry(
765 &mut self,
766 path: Arc<RelPath>,
767 is_directory: bool,
768 content: Option<Vec<u8>>,
769 cx: &Context<Worktree>,
770 ) -> Task<Result<CreatedEntry>> {
771 let worktree_id = self.id();
772 match self {
773 Worktree::Local(this) => this.create_entry(path, is_directory, content, cx),
774 Worktree::Remote(this) => {
775 let project_id = this.project_id;
776 let request = this.client.request(proto::CreateProjectEntry {
777 worktree_id: worktree_id.to_proto(),
778 project_id,
779 path: path.as_ref().to_proto(),
780 content,
781 is_directory,
782 });
783 cx.spawn(async move |this, cx| {
784 let response = request.await?;
785 match response.entry {
786 Some(entry) => this
787 .update(cx, |worktree, cx| {
788 worktree.as_remote_mut().unwrap().insert_entry(
789 entry,
790 response.worktree_scan_id as usize,
791 cx,
792 )
793 })?
794 .await
795 .map(CreatedEntry::Included),
796 None => {
797 let abs_path =
798 this.read_with(cx, |worktree, _| worktree.absolutize(&path))?;
799 Ok(CreatedEntry::Excluded { abs_path })
800 }
801 }
802 })
803 }
804 }
805 }
806
807 pub fn delete_entry(
808 &mut self,
809 entry_id: ProjectEntryId,
810 trash: bool,
811 cx: &mut Context<Worktree>,
812 ) -> Option<Task<Result<()>>> {
813 let task = match self {
814 Worktree::Local(this) => this.delete_entry(entry_id, trash, cx),
815 Worktree::Remote(this) => this.delete_entry(entry_id, trash, cx),
816 }?;
817
818 let entry = match &*self {
819 Worktree::Local(this) => this.entry_for_id(entry_id),
820 Worktree::Remote(this) => this.entry_for_id(entry_id),
821 }?;
822
823 let mut ids = vec![entry_id];
824 let path = &*entry.path;
825
826 self.get_children_ids_recursive(path, &mut ids);
827
828 for id in ids {
829 cx.emit(Event::DeletedEntry(id));
830 }
831 Some(task)
832 }
833
834 fn get_children_ids_recursive(&self, path: &RelPath, ids: &mut Vec<ProjectEntryId>) {
835 let children_iter = self.child_entries(path);
836 for child in children_iter {
837 ids.push(child.id);
838 self.get_children_ids_recursive(&child.path, ids);
839 }
840 }
841
842 // pub fn rename_entry(
843 // &mut self,
844 // entry_id: ProjectEntryId,
845 // new_path: Arc<RelPath>,
846 // cx: &Context<Self>,
847 // ) -> Task<Result<CreatedEntry>> {
848 // match self {
849 // Worktree::Local(this) => this.rename_entry(entry_id, new_path, cx),
850 // Worktree::Remote(this) => this.rename_entry(entry_id, new_path, cx),
851 // }
852 // }
853
854 pub fn copy_external_entries(
855 &mut self,
856 target_directory: Arc<RelPath>,
857 paths: Vec<Arc<Path>>,
858 fs: Arc<dyn Fs>,
859 cx: &Context<Worktree>,
860 ) -> Task<Result<Vec<ProjectEntryId>>> {
861 match self {
862 Worktree::Local(this) => this.copy_external_entries(target_directory, paths, cx),
863 Worktree::Remote(this) => this.copy_external_entries(target_directory, paths, fs, cx),
864 }
865 }
866
867 pub fn expand_entry(
868 &mut self,
869 entry_id: ProjectEntryId,
870 cx: &Context<Worktree>,
871 ) -> Option<Task<Result<()>>> {
872 match self {
873 Worktree::Local(this) => this.expand_entry(entry_id, cx),
874 Worktree::Remote(this) => {
875 let response = this.client.request(proto::ExpandProjectEntry {
876 project_id: this.project_id,
877 entry_id: entry_id.to_proto(),
878 });
879 Some(cx.spawn(async move |this, cx| {
880 let response = response.await?;
881 this.update(cx, |this, _| {
882 this.as_remote_mut()
883 .unwrap()
884 .wait_for_snapshot(response.worktree_scan_id as usize)
885 })?
886 .await?;
887 Ok(())
888 }))
889 }
890 }
891 }
892
893 pub fn expand_all_for_entry(
894 &mut self,
895 entry_id: ProjectEntryId,
896 cx: &Context<Worktree>,
897 ) -> Option<Task<Result<()>>> {
898 match self {
899 Worktree::Local(this) => this.expand_all_for_entry(entry_id, cx),
900 Worktree::Remote(this) => {
901 let response = this.client.request(proto::ExpandAllForProjectEntry {
902 project_id: this.project_id,
903 entry_id: entry_id.to_proto(),
904 });
905 Some(cx.spawn(async move |this, cx| {
906 let response = response.await?;
907 this.update(cx, |this, _| {
908 this.as_remote_mut()
909 .unwrap()
910 .wait_for_snapshot(response.worktree_scan_id as usize)
911 })?
912 .await?;
913 Ok(())
914 }))
915 }
916 }
917 }
918
919 pub async fn handle_create_entry(
920 this: Entity<Self>,
921 request: proto::CreateProjectEntry,
922 mut cx: AsyncApp,
923 ) -> Result<proto::ProjectEntryResponse> {
924 let (scan_id, entry) = this.update(&mut cx, |this, cx| {
925 anyhow::Ok((
926 this.scan_id(),
927 this.create_entry(
928 RelPath::from_proto(&request.path).with_context(|| {
929 format!("received invalid relative path {:?}", request.path)
930 })?,
931 request.is_directory,
932 request.content,
933 cx,
934 ),
935 ))
936 })?;
937 Ok(proto::ProjectEntryResponse {
938 entry: match &entry.await? {
939 CreatedEntry::Included(entry) => Some(entry.into()),
940 CreatedEntry::Excluded { .. } => None,
941 },
942 worktree_scan_id: scan_id as u64,
943 })
944 }
945
946 pub async fn handle_delete_entry(
947 this: Entity<Self>,
948 request: proto::DeleteProjectEntry,
949 mut cx: AsyncApp,
950 ) -> Result<proto::ProjectEntryResponse> {
951 let (scan_id, task) = this.update(&mut cx, |this, cx| {
952 (
953 this.scan_id(),
954 this.delete_entry(
955 ProjectEntryId::from_proto(request.entry_id),
956 request.use_trash,
957 cx,
958 ),
959 )
960 });
961 task.ok_or_else(|| anyhow::anyhow!("invalid entry"))?
962 .await?;
963 Ok(proto::ProjectEntryResponse {
964 entry: None,
965 worktree_scan_id: scan_id as u64,
966 })
967 }
968
969 pub async fn handle_expand_entry(
970 this: Entity<Self>,
971 request: proto::ExpandProjectEntry,
972 mut cx: AsyncApp,
973 ) -> Result<proto::ExpandProjectEntryResponse> {
974 let task = this.update(&mut cx, |this, cx| {
975 this.expand_entry(ProjectEntryId::from_proto(request.entry_id), cx)
976 });
977 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
978 .await?;
979 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
980 Ok(proto::ExpandProjectEntryResponse {
981 worktree_scan_id: scan_id as u64,
982 })
983 }
984
985 pub async fn handle_expand_all_for_entry(
986 this: Entity<Self>,
987 request: proto::ExpandAllForProjectEntry,
988 mut cx: AsyncApp,
989 ) -> Result<proto::ExpandAllForProjectEntryResponse> {
990 let task = this.update(&mut cx, |this, cx| {
991 this.expand_all_for_entry(ProjectEntryId::from_proto(request.entry_id), cx)
992 });
993 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
994 .await?;
995 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
996 Ok(proto::ExpandAllForProjectEntryResponse {
997 worktree_scan_id: scan_id as u64,
998 })
999 }
1000
1001 pub fn is_single_file(&self) -> bool {
1002 self.root_dir().is_none()
1003 }
1004
1005 /// For visible worktrees, returns the path with the worktree name as the first component.
1006 /// Otherwise, returns an absolute path.
1007 pub fn full_path(&self, worktree_relative_path: &RelPath) -> PathBuf {
1008 if self.is_visible() {
1009 self.root_name()
1010 .join(worktree_relative_path)
1011 .display(self.path_style)
1012 .to_string()
1013 .into()
1014 } else {
1015 let full_path = self.abs_path();
1016 let mut full_path_string = if self.is_local()
1017 && let Ok(stripped) = full_path.strip_prefix(home_dir())
1018 {
1019 self.path_style
1020 .join("~", &*stripped.to_string_lossy())
1021 .unwrap()
1022 } else {
1023 full_path.to_string_lossy().into_owned()
1024 };
1025
1026 if worktree_relative_path.components().next().is_some() {
1027 full_path_string.push_str(self.path_style.primary_separator());
1028 full_path_string.push_str(&worktree_relative_path.display(self.path_style));
1029 }
1030
1031 full_path_string.into()
1032 }
1033 }
1034}
1035
1036impl LocalWorktree {
1037 pub fn fs(&self) -> &Arc<dyn Fs> {
1038 &self.fs
1039 }
1040
1041 pub fn is_path_private(&self, path: &RelPath) -> bool {
1042 !self.share_private_files && self.settings.is_path_private(path)
1043 }
1044
1045 pub fn fs_is_case_sensitive(&self) -> bool {
1046 self.fs_case_sensitive
1047 }
1048
1049 fn restart_background_scanners(&mut self, cx: &Context<Worktree>) {
1050 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
1051 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
1052 self.scan_requests_tx = scan_requests_tx;
1053 self.path_prefixes_to_scan_tx = path_prefixes_to_scan_tx;
1054
1055 self.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
1056 let always_included_entries = mem::take(&mut self.snapshot.always_included_entries);
1057 log::debug!(
1058 "refreshing entries for the following always included paths: {:?}",
1059 always_included_entries
1060 );
1061
1062 // Cleans up old always included entries to ensure they get updated properly. Otherwise,
1063 // nested always included entries may not get updated and will result in out-of-date info.
1064 self.refresh_entries_for_paths(always_included_entries);
1065 }
1066
1067 fn start_background_scanner(
1068 &mut self,
1069 scan_requests_rx: channel::Receiver<ScanRequest>,
1070 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
1071 cx: &Context<Worktree>,
1072 ) {
1073 let snapshot = self.snapshot();
1074 let share_private_files = self.share_private_files;
1075 let next_entry_id = self.next_entry_id.clone();
1076 let fs = self.fs.clone();
1077 let scanning_enabled = self.scanning_enabled;
1078 let settings = self.settings.clone();
1079 let (scan_states_tx, mut scan_states_rx) = mpsc::unbounded();
1080 let background_scanner = cx.background_spawn({
1081 let abs_path = snapshot.abs_path.as_path().to_path_buf();
1082 let background = cx.background_executor().clone();
1083 async move {
1084 let (events, watcher) = if scanning_enabled {
1085 fs.watch(&abs_path, FS_WATCH_LATENCY).await
1086 } else {
1087 (Box::pin(stream::pending()) as _, Arc::new(NullWatcher) as _)
1088 };
1089 let fs_case_sensitive = fs.is_case_sensitive().await.unwrap_or_else(|e| {
1090 log::error!("Failed to determine whether filesystem is case sensitive: {e:#}");
1091 true
1092 });
1093
1094 let mut scanner = BackgroundScanner {
1095 fs,
1096 fs_case_sensitive,
1097 status_updates_tx: scan_states_tx,
1098 executor: background,
1099 scan_requests_rx,
1100 path_prefixes_to_scan_rx,
1101 next_entry_id,
1102 state: async_lock::Mutex::new(BackgroundScannerState {
1103 prev_snapshot: snapshot.snapshot.clone(),
1104 snapshot,
1105 scanned_dirs: Default::default(),
1106 scanning_enabled,
1107 path_prefixes_to_scan: Default::default(),
1108 paths_to_scan: Default::default(),
1109 removed_entries: Default::default(),
1110 changed_paths: Default::default(),
1111 }),
1112 phase: BackgroundScannerPhase::InitialScan,
1113 share_private_files,
1114 settings,
1115 watcher,
1116 };
1117
1118 scanner
1119 .run(Box::pin(events.map(|events| events.into_iter().collect())))
1120 .await;
1121 }
1122 });
1123 let scan_state_updater = cx.spawn(async move |this, cx| {
1124 while let Some((state, this)) = scan_states_rx.next().await.zip(this.upgrade()) {
1125 this.update(cx, |this, cx| {
1126 let this = this.as_local_mut().unwrap();
1127 match state {
1128 ScanState::Started => {
1129 *this.is_scanning.0.borrow_mut() = true;
1130 }
1131 ScanState::Updated {
1132 snapshot,
1133 changes,
1134 barrier,
1135 scanning,
1136 } => {
1137 *this.is_scanning.0.borrow_mut() = scanning;
1138 this.set_snapshot(snapshot, changes, cx);
1139 drop(barrier);
1140 }
1141 ScanState::RootUpdated { new_path } => {
1142 this.update_abs_path_and_refresh(new_path, cx);
1143 }
1144 }
1145 });
1146 }
1147 });
1148 self._background_scanner_tasks = vec![background_scanner, scan_state_updater];
1149 *self.is_scanning.0.borrow_mut() = true;
1150 }
1151
1152 fn set_snapshot(
1153 &mut self,
1154 mut new_snapshot: LocalSnapshot,
1155 entry_changes: UpdatedEntriesSet,
1156 cx: &mut Context<Worktree>,
1157 ) {
1158 let repo_changes = self.changed_repos(&self.snapshot, &mut new_snapshot);
1159 self.snapshot = new_snapshot;
1160
1161 if let Some(share) = self.update_observer.as_mut() {
1162 share
1163 .snapshots_tx
1164 .unbounded_send((self.snapshot.clone(), entry_changes.clone()))
1165 .ok();
1166 }
1167
1168 if !entry_changes.is_empty() {
1169 cx.emit(Event::UpdatedEntries(entry_changes));
1170 }
1171 if !repo_changes.is_empty() {
1172 cx.emit(Event::UpdatedGitRepositories(repo_changes));
1173 }
1174 }
1175
1176 fn changed_repos(
1177 &self,
1178 old_snapshot: &LocalSnapshot,
1179 new_snapshot: &mut LocalSnapshot,
1180 ) -> UpdatedGitRepositoriesSet {
1181 let mut changes = Vec::new();
1182 let mut old_repos = old_snapshot.git_repositories.iter().peekable();
1183 let new_repos = new_snapshot.git_repositories.clone();
1184 let mut new_repos = new_repos.iter().peekable();
1185
1186 loop {
1187 match (new_repos.peek().map(clone), old_repos.peek().map(clone)) {
1188 (Some((new_entry_id, new_repo)), Some((old_entry_id, old_repo))) => {
1189 match Ord::cmp(&new_entry_id, &old_entry_id) {
1190 Ordering::Less => {
1191 changes.push(UpdatedGitRepository {
1192 work_directory_id: new_entry_id,
1193 old_work_directory_abs_path: None,
1194 new_work_directory_abs_path: Some(
1195 new_repo.work_directory_abs_path.clone(),
1196 ),
1197 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1198 repository_dir_abs_path: Some(
1199 new_repo.repository_dir_abs_path.clone(),
1200 ),
1201 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1202 });
1203 new_repos.next();
1204 }
1205 Ordering::Equal => {
1206 if new_repo.git_dir_scan_id != old_repo.git_dir_scan_id
1207 || new_repo.work_directory_abs_path
1208 != old_repo.work_directory_abs_path
1209 {
1210 changes.push(UpdatedGitRepository {
1211 work_directory_id: new_entry_id,
1212 old_work_directory_abs_path: Some(
1213 old_repo.work_directory_abs_path.clone(),
1214 ),
1215 new_work_directory_abs_path: Some(
1216 new_repo.work_directory_abs_path.clone(),
1217 ),
1218 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1219 repository_dir_abs_path: Some(
1220 new_repo.repository_dir_abs_path.clone(),
1221 ),
1222 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1223 });
1224 }
1225 new_repos.next();
1226 old_repos.next();
1227 }
1228 Ordering::Greater => {
1229 changes.push(UpdatedGitRepository {
1230 work_directory_id: old_entry_id,
1231 old_work_directory_abs_path: Some(
1232 old_repo.work_directory_abs_path.clone(),
1233 ),
1234 new_work_directory_abs_path: None,
1235 dot_git_abs_path: None,
1236 repository_dir_abs_path: None,
1237 common_dir_abs_path: None,
1238 });
1239 old_repos.next();
1240 }
1241 }
1242 }
1243 (Some((entry_id, repo)), None) => {
1244 changes.push(UpdatedGitRepository {
1245 work_directory_id: entry_id,
1246 old_work_directory_abs_path: None,
1247 new_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1248 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1249 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1250 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1251 });
1252 new_repos.next();
1253 }
1254 (None, Some((entry_id, repo))) => {
1255 changes.push(UpdatedGitRepository {
1256 work_directory_id: entry_id,
1257 old_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1258 new_work_directory_abs_path: None,
1259 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1260 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1261 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1262 });
1263 old_repos.next();
1264 }
1265 (None, None) => break,
1266 }
1267 }
1268
1269 fn clone<T: Clone, U: Clone>(value: &(&T, &U)) -> (T, U) {
1270 (value.0.clone(), value.1.clone())
1271 }
1272
1273 changes.into()
1274 }
1275
1276 pub fn scan_complete(&self) -> impl Future<Output = ()> + use<> {
1277 let mut is_scanning_rx = self.is_scanning.1.clone();
1278 async move {
1279 let mut is_scanning = *is_scanning_rx.borrow();
1280 while is_scanning {
1281 if let Some(value) = is_scanning_rx.recv().await {
1282 is_scanning = value;
1283 } else {
1284 break;
1285 }
1286 }
1287 }
1288 }
1289
1290 pub fn snapshot(&self) -> LocalSnapshot {
1291 self.snapshot.clone()
1292 }
1293
1294 pub fn settings(&self) -> WorktreeSettings {
1295 self.settings.clone()
1296 }
1297
1298 fn load_binary_file(
1299 &self,
1300 path: &RelPath,
1301 cx: &Context<Worktree>,
1302 ) -> Task<Result<LoadedBinaryFile>> {
1303 let path = Arc::from(path);
1304 let abs_path = self.absolutize(&path);
1305 let fs = self.fs.clone();
1306 let entry = self.refresh_entry(path.clone(), None, cx);
1307 let is_private = self.is_path_private(&path);
1308
1309 let worktree = cx.weak_entity();
1310 cx.background_spawn(async move {
1311 let content = fs.load_bytes(&abs_path).await?;
1312
1313 let worktree = worktree.upgrade().context("worktree was dropped")?;
1314 let file = match entry.await? {
1315 Some(entry) => File::for_entry(entry, worktree),
1316 None => {
1317 let metadata = fs
1318 .metadata(&abs_path)
1319 .await
1320 .with_context(|| {
1321 format!("Loading metadata for excluded file {abs_path:?}")
1322 })?
1323 .with_context(|| {
1324 format!("Excluded file {abs_path:?} got removed during loading")
1325 })?;
1326 Arc::new(File {
1327 entry_id: None,
1328 worktree,
1329 path,
1330 disk_state: DiskState::Present {
1331 mtime: metadata.mtime,
1332 },
1333 is_local: true,
1334 is_private,
1335 })
1336 }
1337 };
1338
1339 Ok(LoadedBinaryFile { file, content })
1340 })
1341 }
1342
1343 #[ztracing::instrument(skip_all)]
1344 fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
1345 let path = Arc::from(path);
1346 let abs_path = self.absolutize(&path);
1347 let fs = self.fs.clone();
1348 let entry = self.refresh_entry(path.clone(), None, cx);
1349 let is_private = self.is_path_private(path.as_ref());
1350
1351 let this = cx.weak_entity();
1352 cx.background_spawn(async move {
1353 // WARN: Temporary workaround for #27283.
1354 // We are not efficient with our memory usage per file, and use in excess of 64GB for a 10GB file
1355 // Therefore, as a temporary workaround to prevent system freezes, we just bail before opening a file
1356 // if it is too large
1357 // 5GB seems to be more reasonable, peaking at ~16GB, while 6GB jumps up to >24GB which seems like a
1358 // reasonable limit
1359 {
1360 const FILE_SIZE_MAX: u64 = 6 * 1024 * 1024 * 1024; // 6GB
1361 if let Ok(Some(metadata)) = fs.metadata(&abs_path).await
1362 && metadata.len >= FILE_SIZE_MAX
1363 {
1364 anyhow::bail!("File is too large to load");
1365 }
1366 }
1367 let (text, encoding, has_bom) = decode_file_text(fs.as_ref(), &abs_path).await?;
1368
1369 let worktree = this.upgrade().context("worktree was dropped")?;
1370 let file = match entry.await? {
1371 Some(entry) => File::for_entry(entry, worktree),
1372 None => {
1373 let metadata = fs
1374 .metadata(&abs_path)
1375 .await
1376 .with_context(|| {
1377 format!("Loading metadata for excluded file {abs_path:?}")
1378 })?
1379 .with_context(|| {
1380 format!("Excluded file {abs_path:?} got removed during loading")
1381 })?;
1382 Arc::new(File {
1383 entry_id: None,
1384 worktree,
1385 path,
1386 disk_state: DiskState::Present {
1387 mtime: metadata.mtime,
1388 },
1389 is_local: true,
1390 is_private,
1391 })
1392 }
1393 };
1394
1395 Ok(LoadedFile {
1396 file,
1397 text,
1398 encoding,
1399 has_bom,
1400 })
1401 })
1402 }
1403
1404 /// Find the lowest path in the worktree's datastructures that is an ancestor
1405 fn lowest_ancestor(&self, path: &RelPath) -> Arc<RelPath> {
1406 let mut lowest_ancestor = None;
1407 for path in path.ancestors() {
1408 if self.entry_for_path(path).is_some() {
1409 lowest_ancestor = Some(path.into());
1410 break;
1411 }
1412 }
1413
1414 lowest_ancestor.unwrap_or_else(|| RelPath::empty().into())
1415 }
1416
1417 pub fn create_entry(
1418 &self,
1419 path: Arc<RelPath>,
1420 is_dir: bool,
1421 content: Option<Vec<u8>>,
1422 cx: &Context<Worktree>,
1423 ) -> Task<Result<CreatedEntry>> {
1424 let abs_path = self.absolutize(&path);
1425 let path_excluded = self.settings.is_path_excluded(&path);
1426 let fs = self.fs.clone();
1427 let task_abs_path = abs_path.clone();
1428 let write = cx.background_spawn(async move {
1429 if is_dir {
1430 fs.create_dir(&task_abs_path)
1431 .await
1432 .with_context(|| format!("creating directory {task_abs_path:?}"))
1433 } else {
1434 fs.write(&task_abs_path, content.as_deref().unwrap_or(&[]))
1435 .await
1436 .with_context(|| format!("creating file {task_abs_path:?}"))
1437 }
1438 });
1439
1440 let lowest_ancestor = self.lowest_ancestor(&path);
1441 cx.spawn(async move |this, cx| {
1442 write.await?;
1443 if path_excluded {
1444 return Ok(CreatedEntry::Excluded { abs_path });
1445 }
1446
1447 let (result, refreshes) = this.update(cx, |this, cx| {
1448 let mut refreshes = Vec::new();
1449 let refresh_paths = path.strip_prefix(&lowest_ancestor).unwrap();
1450 for refresh_path in refresh_paths.ancestors() {
1451 if refresh_path == RelPath::empty() {
1452 continue;
1453 }
1454 let refresh_full_path = lowest_ancestor.join(refresh_path);
1455
1456 refreshes.push(this.as_local_mut().unwrap().refresh_entry(
1457 refresh_full_path,
1458 None,
1459 cx,
1460 ));
1461 }
1462 (
1463 this.as_local_mut().unwrap().refresh_entry(path, None, cx),
1464 refreshes,
1465 )
1466 })?;
1467 for refresh in refreshes {
1468 refresh.await.log_err();
1469 }
1470
1471 Ok(result
1472 .await?
1473 .map(CreatedEntry::Included)
1474 .unwrap_or_else(|| CreatedEntry::Excluded { abs_path }))
1475 })
1476 }
1477
1478 pub fn write_file(
1479 &self,
1480 path: Arc<RelPath>,
1481 text: Rope,
1482 line_ending: LineEnding,
1483 encoding: &'static Encoding,
1484 has_bom: bool,
1485 cx: &Context<Worktree>,
1486 ) -> Task<Result<Arc<File>>> {
1487 let fs = self.fs.clone();
1488 let is_private = self.is_path_private(&path);
1489 let abs_path = self.absolutize(&path);
1490
1491 let write = cx.background_spawn({
1492 let fs = fs.clone();
1493 let abs_path = abs_path.clone();
1494 async move {
1495 // For UTF-8, use the optimized `fs.save` which writes Rope chunks directly to disk
1496 // without allocating a contiguous string.
1497 if encoding == encoding_rs::UTF_8 && !has_bom {
1498 return fs.save(&abs_path, &text, line_ending).await;
1499 }
1500
1501 // For legacy encodings (e.g. Shift-JIS), we fall back to converting the entire Rope
1502 // to a String/Bytes in memory before writing.
1503 //
1504 // Note: This is inefficient for very large files compared to the streaming approach above,
1505 // but supporting streaming writes for arbitrary encodings would require a significant
1506 // refactor of the `fs` crate to expose a Writer interface.
1507 let text_string = text.to_string();
1508 let normalized_text = match line_ending {
1509 LineEnding::Unix => text_string,
1510 LineEnding::Windows => text_string.replace('\n', "\r\n"),
1511 };
1512
1513 // Create the byte vector manually for UTF-16 encodings because encoding_rs encodes to UTF-8 by default (per WHATWG standards),
1514 // which is not what we want for saving files.
1515 let bytes = if encoding == encoding_rs::UTF_16BE {
1516 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1517 if has_bom {
1518 data.extend_from_slice(&[0xFE, 0xFF]); // BOM
1519 }
1520 let utf16be_bytes =
1521 normalized_text.encode_utf16().flat_map(|u| u.to_be_bytes());
1522 data.extend(utf16be_bytes);
1523 data.into()
1524 } else if encoding == encoding_rs::UTF_16LE {
1525 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1526 if has_bom {
1527 data.extend_from_slice(&[0xFF, 0xFE]); // BOM
1528 }
1529 let utf16le_bytes =
1530 normalized_text.encode_utf16().flat_map(|u| u.to_le_bytes());
1531 data.extend(utf16le_bytes);
1532 data.into()
1533 } else {
1534 // For other encodings (Shift-JIS, UTF-8 with BOM, etc.), delegate to encoding_rs.
1535 let bom_bytes = if has_bom {
1536 if encoding == encoding_rs::UTF_8 {
1537 vec![0xEF, 0xBB, 0xBF]
1538 } else {
1539 vec![]
1540 }
1541 } else {
1542 vec![]
1543 };
1544 let (cow, _, _) = encoding.encode(&normalized_text);
1545 if !bom_bytes.is_empty() {
1546 let mut bytes = bom_bytes;
1547 bytes.extend_from_slice(&cow);
1548 bytes.into()
1549 } else {
1550 cow
1551 }
1552 };
1553
1554 fs.write(&abs_path, &bytes).await
1555 }
1556 });
1557
1558 cx.spawn(async move |this, cx| {
1559 write.await?;
1560 let entry = this
1561 .update(cx, |this, cx| {
1562 this.as_local_mut()
1563 .unwrap()
1564 .refresh_entry(path.clone(), None, cx)
1565 })?
1566 .await?;
1567 let worktree = this.upgrade().context("worktree dropped")?;
1568 if let Some(entry) = entry {
1569 Ok(File::for_entry(entry, worktree))
1570 } else {
1571 let metadata = fs
1572 .metadata(&abs_path)
1573 .await
1574 .with_context(|| {
1575 format!("Fetching metadata after saving the excluded buffer {abs_path:?}")
1576 })?
1577 .with_context(|| {
1578 format!("Excluded buffer {path:?} got removed during saving")
1579 })?;
1580 Ok(Arc::new(File {
1581 worktree,
1582 path,
1583 disk_state: DiskState::Present {
1584 mtime: metadata.mtime,
1585 },
1586 entry_id: None,
1587 is_local: true,
1588 is_private,
1589 }))
1590 }
1591 })
1592 }
1593
1594 pub fn delete_entry(
1595 &self,
1596 entry_id: ProjectEntryId,
1597 trash: bool,
1598 cx: &Context<Worktree>,
1599 ) -> Option<Task<Result<()>>> {
1600 let entry = self.entry_for_id(entry_id)?.clone();
1601 let abs_path = self.absolutize(&entry.path);
1602 let fs = self.fs.clone();
1603
1604 let delete = cx.background_spawn(async move {
1605 if entry.is_file() {
1606 if trash {
1607 fs.trash_file(&abs_path, Default::default()).await?;
1608 } else {
1609 fs.remove_file(&abs_path, Default::default()).await?;
1610 }
1611 } else if trash {
1612 fs.trash_dir(
1613 &abs_path,
1614 RemoveOptions {
1615 recursive: true,
1616 ignore_if_not_exists: false,
1617 },
1618 )
1619 .await?;
1620 } else {
1621 fs.remove_dir(
1622 &abs_path,
1623 RemoveOptions {
1624 recursive: true,
1625 ignore_if_not_exists: false,
1626 },
1627 )
1628 .await?;
1629 }
1630 anyhow::Ok(entry.path)
1631 });
1632
1633 Some(cx.spawn(async move |this, cx| {
1634 let path = delete.await?;
1635 this.update(cx, |this, _| {
1636 this.as_local_mut()
1637 .unwrap()
1638 .refresh_entries_for_paths(vec![path])
1639 })?
1640 .recv()
1641 .await;
1642 Ok(())
1643 }))
1644 }
1645
1646 pub fn copy_external_entries(
1647 &self,
1648 target_directory: Arc<RelPath>,
1649 paths: Vec<Arc<Path>>,
1650 cx: &Context<Worktree>,
1651 ) -> Task<Result<Vec<ProjectEntryId>>> {
1652 let target_directory = self.absolutize(&target_directory);
1653 let worktree_path = self.abs_path().clone();
1654 let fs = self.fs.clone();
1655 let paths = paths
1656 .into_iter()
1657 .filter_map(|source| {
1658 let file_name = source.file_name()?;
1659 let mut target = target_directory.clone();
1660 target.push(file_name);
1661
1662 // Do not allow copying the same file to itself.
1663 if source.as_ref() != target.as_path() {
1664 Some((source, target))
1665 } else {
1666 None
1667 }
1668 })
1669 .collect::<Vec<_>>();
1670
1671 let paths_to_refresh = paths
1672 .iter()
1673 .filter_map(|(_, target)| {
1674 RelPath::new(
1675 target.strip_prefix(&worktree_path).ok()?,
1676 PathStyle::local(),
1677 )
1678 .ok()
1679 .map(|path| path.into_arc())
1680 })
1681 .collect::<Vec<_>>();
1682
1683 cx.spawn(async move |this, cx| {
1684 cx.background_spawn(async move {
1685 for (source, target) in paths {
1686 copy_recursive(
1687 fs.as_ref(),
1688 &source,
1689 &target,
1690 fs::CopyOptions {
1691 overwrite: true,
1692 ..Default::default()
1693 },
1694 )
1695 .await
1696 .with_context(|| {
1697 format!("Failed to copy file from {source:?} to {target:?}")
1698 })?;
1699 }
1700 anyhow::Ok(())
1701 })
1702 .await
1703 .log_err();
1704 let mut refresh = cx.read_entity(
1705 &this.upgrade().with_context(|| "Dropped worktree")?,
1706 |this, _| {
1707 anyhow::Ok::<postage::barrier::Receiver>(
1708 this.as_local()
1709 .with_context(|| "Worktree is not local")?
1710 .refresh_entries_for_paths(paths_to_refresh.clone()),
1711 )
1712 },
1713 )?;
1714
1715 cx.background_spawn(async move {
1716 refresh.next().await;
1717 anyhow::Ok(())
1718 })
1719 .await
1720 .log_err();
1721
1722 let this = this.upgrade().with_context(|| "Dropped worktree")?;
1723 Ok(cx.read_entity(&this, |this, _| {
1724 paths_to_refresh
1725 .iter()
1726 .filter_map(|path| Some(this.entry_for_path(path)?.id))
1727 .collect()
1728 }))
1729 })
1730 }
1731
1732 fn expand_entry(
1733 &self,
1734 entry_id: ProjectEntryId,
1735 cx: &Context<Worktree>,
1736 ) -> Option<Task<Result<()>>> {
1737 let path = self.entry_for_id(entry_id)?.path.clone();
1738 let mut refresh = self.refresh_entries_for_paths(vec![path]);
1739 Some(cx.background_spawn(async move {
1740 refresh.next().await;
1741 Ok(())
1742 }))
1743 }
1744
1745 fn expand_all_for_entry(
1746 &self,
1747 entry_id: ProjectEntryId,
1748 cx: &Context<Worktree>,
1749 ) -> Option<Task<Result<()>>> {
1750 let path = self.entry_for_id(entry_id).unwrap().path.clone();
1751 let mut rx = self.add_path_prefix_to_scan(path);
1752 Some(cx.background_spawn(async move {
1753 rx.next().await;
1754 Ok(())
1755 }))
1756 }
1757
1758 pub fn refresh_entries_for_paths(&self, paths: Vec<Arc<RelPath>>) -> barrier::Receiver {
1759 let (tx, rx) = barrier::channel();
1760 self.scan_requests_tx
1761 .try_send(ScanRequest {
1762 relative_paths: paths,
1763 done: smallvec![tx],
1764 })
1765 .ok();
1766 rx
1767 }
1768
1769 #[cfg(feature = "test-support")]
1770 pub fn manually_refresh_entries_for_paths(
1771 &self,
1772 paths: Vec<Arc<RelPath>>,
1773 ) -> barrier::Receiver {
1774 self.refresh_entries_for_paths(paths)
1775 }
1776
1777 pub fn add_path_prefix_to_scan(&self, path_prefix: Arc<RelPath>) -> barrier::Receiver {
1778 let (tx, rx) = barrier::channel();
1779 self.path_prefixes_to_scan_tx
1780 .try_send(PathPrefixScanRequest {
1781 path: path_prefix,
1782 done: smallvec![tx],
1783 })
1784 .ok();
1785 rx
1786 }
1787
1788 pub fn refresh_entry(
1789 &self,
1790 path: Arc<RelPath>,
1791 old_path: Option<Arc<RelPath>>,
1792 cx: &Context<Worktree>,
1793 ) -> Task<Result<Option<Entry>>> {
1794 if self.settings.is_path_excluded(&path) {
1795 return Task::ready(Ok(None));
1796 }
1797 let paths = if let Some(old_path) = old_path.as_ref() {
1798 vec![old_path.clone(), path.clone()]
1799 } else {
1800 vec![path.clone()]
1801 };
1802 let t0 = Instant::now();
1803 let mut refresh = self.refresh_entries_for_paths(paths);
1804 // todo(lw): Hot foreground spawn
1805 cx.spawn(async move |this, cx| {
1806 refresh.recv().await;
1807 log::trace!("refreshed entry {path:?} in {:?}", t0.elapsed());
1808 let new_entry = this.read_with(cx, |this, _| {
1809 this.entry_for_path(&path).cloned().with_context(|| {
1810 format!("Could not find entry in worktree for {path:?} after refresh")
1811 })
1812 })??;
1813 Ok(Some(new_entry))
1814 })
1815 }
1816
1817 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1818 where
1819 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1820 Fut: 'static + Send + Future<Output = bool>,
1821 {
1822 if let Some(observer) = self.update_observer.as_mut() {
1823 *observer.resume_updates.borrow_mut() = ();
1824 return;
1825 }
1826
1827 let (resume_updates_tx, mut resume_updates_rx) = watch::channel::<()>();
1828 let (snapshots_tx, mut snapshots_rx) =
1829 mpsc::unbounded::<(LocalSnapshot, UpdatedEntriesSet)>();
1830 snapshots_tx
1831 .unbounded_send((self.snapshot(), Arc::default()))
1832 .ok();
1833
1834 let worktree_id = cx.entity_id().as_u64();
1835 let _maintain_remote_snapshot = cx.background_spawn(async move {
1836 let mut is_first = true;
1837 while let Some((snapshot, entry_changes)) = snapshots_rx.next().await {
1838 let update = if is_first {
1839 is_first = false;
1840 snapshot.build_initial_update(project_id, worktree_id)
1841 } else {
1842 snapshot.build_update(project_id, worktree_id, entry_changes)
1843 };
1844
1845 for update in proto::split_worktree_update(update) {
1846 let _ = resume_updates_rx.try_recv();
1847 loop {
1848 let result = callback(update.clone());
1849 if result.await {
1850 break;
1851 } else {
1852 log::info!("waiting to resume updates");
1853 if resume_updates_rx.next().await.is_none() {
1854 return Some(());
1855 }
1856 }
1857 }
1858 }
1859 }
1860 Some(())
1861 });
1862
1863 self.update_observer = Some(UpdateObservationState {
1864 snapshots_tx,
1865 resume_updates: resume_updates_tx,
1866 _maintain_remote_snapshot,
1867 });
1868 }
1869
1870 pub fn share_private_files(&mut self, cx: &Context<Worktree>) {
1871 self.share_private_files = true;
1872 self.restart_background_scanners(cx);
1873 }
1874
1875 pub fn update_abs_path_and_refresh(
1876 &mut self,
1877 new_path: Arc<SanitizedPath>,
1878 cx: &Context<Worktree>,
1879 ) {
1880 self.snapshot.git_repositories = Default::default();
1881 self.snapshot.ignores_by_parent_abs_path = Default::default();
1882 let root_name = new_path
1883 .as_path()
1884 .file_name()
1885 .and_then(|f| f.to_str())
1886 .map_or(RelPath::empty().into(), |f| {
1887 RelPath::unix(f).unwrap().into()
1888 });
1889 self.snapshot.update_abs_path(new_path, root_name);
1890 self.restart_background_scanners(cx);
1891 }
1892 #[cfg(feature = "test-support")]
1893 pub fn repositories(&self) -> Vec<Arc<Path>> {
1894 self.git_repositories
1895 .values()
1896 .map(|entry| entry.work_directory_abs_path.clone())
1897 .collect::<Vec<_>>()
1898 }
1899}
1900
1901impl RemoteWorktree {
1902 pub fn project_id(&self) -> u64 {
1903 self.project_id
1904 }
1905
1906 pub fn client(&self) -> AnyProtoClient {
1907 self.client.clone()
1908 }
1909
1910 pub fn disconnected_from_host(&mut self) {
1911 self.updates_tx.take();
1912 self.snapshot_subscriptions.clear();
1913 self.disconnected = true;
1914 }
1915
1916 pub fn update_from_remote(&self, update: proto::UpdateWorktree) {
1917 if let Some(updates_tx) = &self.updates_tx {
1918 updates_tx
1919 .unbounded_send(update)
1920 .expect("consumer runs to completion");
1921 }
1922 }
1923
1924 fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1925 where
1926 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1927 Fut: 'static + Send + Future<Output = bool>,
1928 {
1929 let (tx, mut rx) = mpsc::unbounded();
1930 let initial_update = self
1931 .snapshot
1932 .build_initial_update(project_id, self.id().to_proto());
1933 self.update_observer = Some(tx);
1934 cx.spawn(async move |this, cx| {
1935 let mut update = initial_update;
1936 'outer: loop {
1937 // SSH projects use a special project ID of 0, and we need to
1938 // remap it to the correct one here.
1939 update.project_id = project_id;
1940
1941 for chunk in split_worktree_update(update) {
1942 if !callback(chunk).await {
1943 break 'outer;
1944 }
1945 }
1946
1947 if let Some(next_update) = rx.next().await {
1948 update = next_update;
1949 } else {
1950 break;
1951 }
1952 }
1953 this.update(cx, |this, _| {
1954 let this = this.as_remote_mut().unwrap();
1955 this.update_observer.take();
1956 })
1957 })
1958 .detach();
1959 }
1960
1961 fn observed_snapshot(&self, scan_id: usize) -> bool {
1962 self.completed_scan_id >= scan_id
1963 }
1964
1965 pub fn wait_for_snapshot(
1966 &mut self,
1967 scan_id: usize,
1968 ) -> impl Future<Output = Result<()>> + use<> {
1969 let (tx, rx) = oneshot::channel();
1970 if self.observed_snapshot(scan_id) {
1971 let _ = tx.send(());
1972 } else if self.disconnected {
1973 drop(tx);
1974 } else {
1975 match self
1976 .snapshot_subscriptions
1977 .binary_search_by_key(&scan_id, |probe| probe.0)
1978 {
1979 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
1980 }
1981 }
1982
1983 async move {
1984 rx.await?;
1985 Ok(())
1986 }
1987 }
1988
1989 pub fn insert_entry(
1990 &mut self,
1991 entry: proto::Entry,
1992 scan_id: usize,
1993 cx: &Context<Worktree>,
1994 ) -> Task<Result<Entry>> {
1995 let wait_for_snapshot = self.wait_for_snapshot(scan_id);
1996 cx.spawn(async move |this, cx| {
1997 wait_for_snapshot.await?;
1998 this.update(cx, |worktree, _| {
1999 let worktree = worktree.as_remote_mut().unwrap();
2000 let snapshot = &mut worktree.background_snapshot.lock().0;
2001 let entry = snapshot.insert_entry(entry, &worktree.file_scan_inclusions);
2002 worktree.snapshot = snapshot.clone();
2003 entry
2004 })?
2005 })
2006 }
2007
2008 fn delete_entry(
2009 &self,
2010 entry_id: ProjectEntryId,
2011 trash: bool,
2012 cx: &Context<Worktree>,
2013 ) -> Option<Task<Result<()>>> {
2014 let response = self.client.request(proto::DeleteProjectEntry {
2015 project_id: self.project_id,
2016 entry_id: entry_id.to_proto(),
2017 use_trash: trash,
2018 });
2019 Some(cx.spawn(async move |this, cx| {
2020 let response = response.await?;
2021 let scan_id = response.worktree_scan_id as usize;
2022
2023 this.update(cx, move |this, _| {
2024 this.as_remote_mut().unwrap().wait_for_snapshot(scan_id)
2025 })?
2026 .await?;
2027
2028 this.update(cx, |this, _| {
2029 let this = this.as_remote_mut().unwrap();
2030 let snapshot = &mut this.background_snapshot.lock().0;
2031 snapshot.delete_entry(entry_id);
2032 this.snapshot = snapshot.clone();
2033 })
2034 }))
2035 }
2036
2037 // fn rename_entry(
2038 // &self,
2039 // entry_id: ProjectEntryId,
2040 // new_path: impl Into<Arc<RelPath>>,
2041 // cx: &Context<Worktree>,
2042 // ) -> Task<Result<CreatedEntry>> {
2043 // let new_path: Arc<RelPath> = new_path.into();
2044 // let response = self.client.request(proto::RenameProjectEntry {
2045 // project_id: self.project_id,
2046 // entry_id: entry_id.to_proto(),
2047 // new_worktree_id: new_path.worktree_id,
2048 // new_path: new_path.as_ref().to_proto(),
2049 // });
2050 // cx.spawn(async move |this, cx| {
2051 // let response = response.await?;
2052 // match response.entry {
2053 // Some(entry) => this
2054 // .update(cx, |this, cx| {
2055 // this.as_remote_mut().unwrap().insert_entry(
2056 // entry,
2057 // response.worktree_scan_id as usize,
2058 // cx,
2059 // )
2060 // })?
2061 // .await
2062 // .map(CreatedEntry::Included),
2063 // None => {
2064 // let abs_path =
2065 // this.read_with(cx, |worktree, _| worktree.absolutize(&new_path))?;
2066 // Ok(CreatedEntry::Excluded { abs_path })
2067 // }
2068 // }
2069 // })
2070 // }
2071
2072 fn copy_external_entries(
2073 &self,
2074 target_directory: Arc<RelPath>,
2075 paths_to_copy: Vec<Arc<Path>>,
2076 local_fs: Arc<dyn Fs>,
2077 cx: &Context<Worktree>,
2078 ) -> Task<anyhow::Result<Vec<ProjectEntryId>>> {
2079 let client = self.client.clone();
2080 let worktree_id = self.id().to_proto();
2081 let project_id = self.project_id;
2082
2083 cx.background_spawn(async move {
2084 let mut requests = Vec::new();
2085 for root_path_to_copy in paths_to_copy {
2086 let Some(filename) = root_path_to_copy
2087 .file_name()
2088 .and_then(|name| name.to_str())
2089 .and_then(|filename| RelPath::unix(filename).ok())
2090 else {
2091 continue;
2092 };
2093 for (abs_path, is_directory) in
2094 read_dir_items(local_fs.as_ref(), &root_path_to_copy).await?
2095 {
2096 let Some(relative_path) = abs_path
2097 .strip_prefix(&root_path_to_copy)
2098 .map_err(|e| anyhow::Error::from(e))
2099 .and_then(|relative_path| RelPath::new(relative_path, PathStyle::local()))
2100 .log_err()
2101 else {
2102 continue;
2103 };
2104 let content = if is_directory {
2105 None
2106 } else {
2107 Some(local_fs.load_bytes(&abs_path).await?)
2108 };
2109
2110 let mut target_path = target_directory.join(filename);
2111 if relative_path.file_name().is_some() {
2112 target_path = target_path.join(&relative_path);
2113 }
2114
2115 requests.push(proto::CreateProjectEntry {
2116 project_id,
2117 worktree_id,
2118 path: target_path.to_proto(),
2119 is_directory,
2120 content,
2121 });
2122 }
2123 }
2124 requests.sort_unstable_by(|a, b| a.path.cmp(&b.path));
2125 requests.dedup();
2126
2127 let mut copied_entry_ids = Vec::new();
2128 for request in requests {
2129 let response = client.request(request).await?;
2130 copied_entry_ids.extend(response.entry.map(|e| ProjectEntryId::from_proto(e.id)));
2131 }
2132
2133 Ok(copied_entry_ids)
2134 })
2135 }
2136}
2137
2138impl Snapshot {
2139 pub fn new(
2140 id: u64,
2141 root_name: Arc<RelPath>,
2142 abs_path: Arc<Path>,
2143 path_style: PathStyle,
2144 ) -> Self {
2145 Snapshot {
2146 id: WorktreeId::from_usize(id as usize),
2147 abs_path: SanitizedPath::from_arc(abs_path),
2148 path_style,
2149 root_char_bag: root_name
2150 .as_unix_str()
2151 .chars()
2152 .map(|c| c.to_ascii_lowercase())
2153 .collect(),
2154 root_name,
2155 always_included_entries: Default::default(),
2156 entries_by_path: Default::default(),
2157 entries_by_id: Default::default(),
2158 scan_id: 1,
2159 completed_scan_id: 0,
2160 }
2161 }
2162
2163 pub fn id(&self) -> WorktreeId {
2164 self.id
2165 }
2166
2167 // TODO:
2168 // Consider the following:
2169 //
2170 // ```rust
2171 // let abs_path: Arc<Path> = snapshot.abs_path(); // e.g. "C:\Users\user\Desktop\project"
2172 // let some_non_trimmed_path = Path::new("\\\\?\\C:\\Users\\user\\Desktop\\project\\main.rs");
2173 // // The caller perform some actions here:
2174 // some_non_trimmed_path.strip_prefix(abs_path); // This fails
2175 // some_non_trimmed_path.starts_with(abs_path); // This fails too
2176 // ```
2177 //
2178 // This is definitely a bug, but it's not clear if we should handle it here or not.
2179 pub fn abs_path(&self) -> &Arc<Path> {
2180 SanitizedPath::cast_arc_ref(&self.abs_path)
2181 }
2182
2183 fn build_initial_update(&self, project_id: u64, worktree_id: u64) -> proto::UpdateWorktree {
2184 let mut updated_entries = self
2185 .entries_by_path
2186 .iter()
2187 .map(proto::Entry::from)
2188 .collect::<Vec<_>>();
2189 updated_entries.sort_unstable_by_key(|e| e.id);
2190
2191 proto::UpdateWorktree {
2192 project_id,
2193 worktree_id,
2194 abs_path: self.abs_path().to_string_lossy().into_owned(),
2195 root_name: self.root_name().to_proto(),
2196 updated_entries,
2197 removed_entries: Vec::new(),
2198 scan_id: self.scan_id as u64,
2199 is_last_update: self.completed_scan_id == self.scan_id,
2200 // Sent in separate messages.
2201 updated_repositories: Vec::new(),
2202 removed_repositories: Vec::new(),
2203 }
2204 }
2205
2206 pub fn work_directory_abs_path(&self, work_directory: &WorkDirectory) -> PathBuf {
2207 match work_directory {
2208 WorkDirectory::InProject { relative_path } => self.absolutize(relative_path),
2209 WorkDirectory::AboveProject { absolute_path, .. } => absolute_path.as_ref().to_owned(),
2210 }
2211 }
2212
2213 pub fn absolutize(&self, path: &RelPath) -> PathBuf {
2214 if path.file_name().is_some() {
2215 let mut abs_path = self.abs_path.to_string();
2216 for component in path.components() {
2217 if !abs_path.ends_with(self.path_style.primary_separator()) {
2218 abs_path.push_str(self.path_style.primary_separator());
2219 }
2220 abs_path.push_str(component);
2221 }
2222 PathBuf::from(abs_path)
2223 } else {
2224 self.abs_path.as_path().to_path_buf()
2225 }
2226 }
2227
2228 pub fn contains_entry(&self, entry_id: ProjectEntryId) -> bool {
2229 self.entries_by_id.get(&entry_id, ()).is_some()
2230 }
2231
2232 fn insert_entry(
2233 &mut self,
2234 entry: proto::Entry,
2235 always_included_paths: &PathMatcher,
2236 ) -> Result<Entry> {
2237 let entry = Entry::try_from((&self.root_char_bag, always_included_paths, entry))?;
2238 let old_entry = self.entries_by_id.insert_or_replace(
2239 PathEntry {
2240 id: entry.id,
2241 path: entry.path.clone(),
2242 is_ignored: entry.is_ignored,
2243 scan_id: 0,
2244 },
2245 (),
2246 );
2247 if let Some(old_entry) = old_entry {
2248 self.entries_by_path.remove(&PathKey(old_entry.path), ());
2249 }
2250 self.entries_by_path.insert_or_replace(entry.clone(), ());
2251 Ok(entry)
2252 }
2253
2254 fn delete_entry(&mut self, entry_id: ProjectEntryId) -> Option<Arc<RelPath>> {
2255 let removed_entry = self.entries_by_id.remove(&entry_id, ())?;
2256 self.entries_by_path = {
2257 let mut cursor = self.entries_by_path.cursor::<TraversalProgress>(());
2258 let mut new_entries_by_path =
2259 cursor.slice(&TraversalTarget::path(&removed_entry.path), Bias::Left);
2260 while let Some(entry) = cursor.item() {
2261 if entry.path.starts_with(&removed_entry.path) {
2262 self.entries_by_id.remove(&entry.id, ());
2263 cursor.next();
2264 } else {
2265 break;
2266 }
2267 }
2268 new_entries_by_path.append(cursor.suffix(), ());
2269 new_entries_by_path
2270 };
2271
2272 Some(removed_entry.path)
2273 }
2274
2275 fn update_abs_path(&mut self, abs_path: Arc<SanitizedPath>, root_name: Arc<RelPath>) {
2276 self.abs_path = abs_path;
2277 if root_name != self.root_name {
2278 self.root_char_bag = root_name
2279 .as_unix_str()
2280 .chars()
2281 .map(|c| c.to_ascii_lowercase())
2282 .collect();
2283 self.root_name = root_name;
2284 }
2285 }
2286
2287 pub fn apply_remote_update(
2288 &mut self,
2289 update: proto::UpdateWorktree,
2290 always_included_paths: &PathMatcher,
2291 ) {
2292 log::debug!(
2293 "applying remote worktree update. {} entries updated, {} removed",
2294 update.updated_entries.len(),
2295 update.removed_entries.len()
2296 );
2297 if let Some(root_name) = RelPath::from_proto(&update.root_name).log_err() {
2298 self.update_abs_path(
2299 SanitizedPath::new_arc(&Path::new(&update.abs_path)),
2300 root_name,
2301 );
2302 }
2303
2304 let mut entries_by_path_edits = Vec::new();
2305 let mut entries_by_id_edits = Vec::new();
2306
2307 for entry_id in update.removed_entries {
2308 let entry_id = ProjectEntryId::from_proto(entry_id);
2309 entries_by_id_edits.push(Edit::Remove(entry_id));
2310 if let Some(entry) = self.entry_for_id(entry_id) {
2311 entries_by_path_edits.push(Edit::Remove(PathKey(entry.path.clone())));
2312 }
2313 }
2314
2315 for entry in update.updated_entries {
2316 let Some(entry) =
2317 Entry::try_from((&self.root_char_bag, always_included_paths, entry)).log_err()
2318 else {
2319 continue;
2320 };
2321 if let Some(PathEntry { path, .. }) = self.entries_by_id.get(&entry.id, ()) {
2322 entries_by_path_edits.push(Edit::Remove(PathKey(path.clone())));
2323 }
2324 if let Some(old_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2325 && old_entry.id != entry.id
2326 {
2327 entries_by_id_edits.push(Edit::Remove(old_entry.id));
2328 }
2329 entries_by_id_edits.push(Edit::Insert(PathEntry {
2330 id: entry.id,
2331 path: entry.path.clone(),
2332 is_ignored: entry.is_ignored,
2333 scan_id: 0,
2334 }));
2335 entries_by_path_edits.push(Edit::Insert(entry));
2336 }
2337
2338 self.entries_by_path.edit(entries_by_path_edits, ());
2339 self.entries_by_id.edit(entries_by_id_edits, ());
2340
2341 self.scan_id = update.scan_id as usize;
2342 if update.is_last_update {
2343 self.completed_scan_id = update.scan_id as usize;
2344 }
2345 }
2346
2347 pub fn entry_count(&self) -> usize {
2348 self.entries_by_path.summary().count
2349 }
2350
2351 pub fn visible_entry_count(&self) -> usize {
2352 self.entries_by_path.summary().non_ignored_count
2353 }
2354
2355 pub fn dir_count(&self) -> usize {
2356 let summary = self.entries_by_path.summary();
2357 summary.count - summary.file_count
2358 }
2359
2360 pub fn visible_dir_count(&self) -> usize {
2361 let summary = self.entries_by_path.summary();
2362 summary.non_ignored_count - summary.non_ignored_file_count
2363 }
2364
2365 pub fn file_count(&self) -> usize {
2366 self.entries_by_path.summary().file_count
2367 }
2368
2369 pub fn visible_file_count(&self) -> usize {
2370 self.entries_by_path.summary().non_ignored_file_count
2371 }
2372
2373 fn traverse_from_offset(
2374 &self,
2375 include_files: bool,
2376 include_dirs: bool,
2377 include_ignored: bool,
2378 start_offset: usize,
2379 ) -> Traversal<'_> {
2380 let mut cursor = self.entries_by_path.cursor(());
2381 cursor.seek(
2382 &TraversalTarget::Count {
2383 count: start_offset,
2384 include_files,
2385 include_dirs,
2386 include_ignored,
2387 },
2388 Bias::Right,
2389 );
2390 Traversal {
2391 snapshot: self,
2392 cursor,
2393 include_files,
2394 include_dirs,
2395 include_ignored,
2396 }
2397 }
2398
2399 pub fn traverse_from_path(
2400 &self,
2401 include_files: bool,
2402 include_dirs: bool,
2403 include_ignored: bool,
2404 path: &RelPath,
2405 ) -> Traversal<'_> {
2406 Traversal::new(self, include_files, include_dirs, include_ignored, path)
2407 }
2408
2409 pub fn files(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2410 self.traverse_from_offset(true, false, include_ignored, start)
2411 }
2412
2413 pub fn directories(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2414 self.traverse_from_offset(false, true, include_ignored, start)
2415 }
2416
2417 pub fn entries(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2418 self.traverse_from_offset(true, true, include_ignored, start)
2419 }
2420
2421 pub fn paths(&self) -> impl Iterator<Item = &RelPath> {
2422 self.entries_by_path
2423 .cursor::<()>(())
2424 .filter(move |entry| !entry.path.is_empty())
2425 .map(|entry| entry.path.as_ref())
2426 }
2427
2428 pub fn child_entries<'a>(&'a self, parent_path: &'a RelPath) -> ChildEntriesIter<'a> {
2429 let options = ChildEntriesOptions {
2430 include_files: true,
2431 include_dirs: true,
2432 include_ignored: true,
2433 };
2434 self.child_entries_with_options(parent_path, options)
2435 }
2436
2437 pub fn child_entries_with_options<'a>(
2438 &'a self,
2439 parent_path: &'a RelPath,
2440 options: ChildEntriesOptions,
2441 ) -> ChildEntriesIter<'a> {
2442 let mut cursor = self.entries_by_path.cursor(());
2443 cursor.seek(&TraversalTarget::path(parent_path), Bias::Right);
2444 let traversal = Traversal {
2445 snapshot: self,
2446 cursor,
2447 include_files: options.include_files,
2448 include_dirs: options.include_dirs,
2449 include_ignored: options.include_ignored,
2450 };
2451 ChildEntriesIter {
2452 traversal,
2453 parent_path,
2454 }
2455 }
2456
2457 pub fn root_entry(&self) -> Option<&Entry> {
2458 self.entries_by_path.first()
2459 }
2460
2461 /// Returns `None` for a single file worktree, or `Some(self.abs_path())` if
2462 /// it is a directory.
2463 pub fn root_dir(&self) -> Option<Arc<Path>> {
2464 self.root_entry()
2465 .filter(|entry| entry.is_dir())
2466 .map(|_| self.abs_path().clone())
2467 }
2468
2469 pub fn root_name(&self) -> &RelPath {
2470 &self.root_name
2471 }
2472
2473 pub fn root_name_str(&self) -> &str {
2474 self.root_name.as_unix_str()
2475 }
2476
2477 pub fn scan_id(&self) -> usize {
2478 self.scan_id
2479 }
2480
2481 pub fn entry_for_path(&self, path: &RelPath) -> Option<&Entry> {
2482 self.traverse_from_path(true, true, true, path)
2483 .entry()
2484 .and_then(|entry| {
2485 if entry.path.as_ref() == path {
2486 Some(entry)
2487 } else {
2488 None
2489 }
2490 })
2491 }
2492
2493 /// Resolves a path to an executable using the following heuristics:
2494 ///
2495 /// 1. If the path starts with `~`, it is expanded to the user's home directory.
2496 /// 2. If the path is relative and contains more than one component,
2497 /// it is joined to the worktree root path.
2498 /// 3. If the path is relative and exists in the worktree
2499 /// (even if falls under an exclusion filter),
2500 /// it is joined to the worktree root path.
2501 /// 4. Otherwise the path is returned unmodified.
2502 ///
2503 /// Relative paths that do not exist in the worktree may
2504 /// still be found using the `PATH` environment variable.
2505 pub fn resolve_executable_path(&self, path: PathBuf) -> PathBuf {
2506 if let Some(path_str) = path.to_str() {
2507 if let Some(remaining_path) = path_str.strip_prefix("~/") {
2508 return home_dir().join(remaining_path);
2509 } else if path_str == "~" {
2510 return home_dir().to_path_buf();
2511 }
2512 }
2513
2514 if let Ok(rel_path) = RelPath::new(&path, self.path_style)
2515 && (path.components().count() > 1 || self.entry_for_path(&rel_path).is_some())
2516 {
2517 self.abs_path().join(path)
2518 } else {
2519 path
2520 }
2521 }
2522
2523 pub fn entry_for_id(&self, id: ProjectEntryId) -> Option<&Entry> {
2524 let entry = self.entries_by_id.get(&id, ())?;
2525 self.entry_for_path(&entry.path)
2526 }
2527
2528 pub fn path_style(&self) -> PathStyle {
2529 self.path_style
2530 }
2531}
2532
2533impl LocalSnapshot {
2534 fn local_repo_for_work_directory_path(&self, path: &RelPath) -> Option<&LocalRepositoryEntry> {
2535 self.git_repositories
2536 .iter()
2537 .map(|(_, entry)| entry)
2538 .find(|entry| entry.work_directory.path_key() == PathKey(path.into()))
2539 }
2540
2541 fn build_update(
2542 &self,
2543 project_id: u64,
2544 worktree_id: u64,
2545 entry_changes: UpdatedEntriesSet,
2546 ) -> proto::UpdateWorktree {
2547 let mut updated_entries = Vec::new();
2548 let mut removed_entries = Vec::new();
2549
2550 for (_, entry_id, path_change) in entry_changes.iter() {
2551 if let PathChange::Removed = path_change {
2552 removed_entries.push(entry_id.0 as u64);
2553 } else if let Some(entry) = self.entry_for_id(*entry_id) {
2554 updated_entries.push(proto::Entry::from(entry));
2555 }
2556 }
2557
2558 removed_entries.sort_unstable();
2559 updated_entries.sort_unstable_by_key(|e| e.id);
2560
2561 // TODO - optimize, knowing that removed_entries are sorted.
2562 removed_entries.retain(|id| updated_entries.binary_search_by_key(id, |e| e.id).is_err());
2563
2564 proto::UpdateWorktree {
2565 project_id,
2566 worktree_id,
2567 abs_path: self.abs_path().to_string_lossy().into_owned(),
2568 root_name: self.root_name().to_proto(),
2569 updated_entries,
2570 removed_entries,
2571 scan_id: self.scan_id as u64,
2572 is_last_update: self.completed_scan_id == self.scan_id,
2573 // Sent in separate messages.
2574 updated_repositories: Vec::new(),
2575 removed_repositories: Vec::new(),
2576 }
2577 }
2578
2579 async fn insert_entry(&mut self, mut entry: Entry, fs: &dyn Fs) -> Entry {
2580 log::trace!("insert entry {:?}", entry.path);
2581 if entry.is_file() && entry.path.file_name() == Some(&GITIGNORE) {
2582 let abs_path = self.absolutize(&entry.path);
2583 match build_gitignore(&abs_path, fs).await {
2584 Ok(ignore) => {
2585 self.ignores_by_parent_abs_path
2586 .insert(abs_path.parent().unwrap().into(), (Arc::new(ignore), true));
2587 }
2588 Err(error) => {
2589 log::error!(
2590 "error loading .gitignore file {:?} - {:?}",
2591 &entry.path,
2592 error
2593 );
2594 }
2595 }
2596 }
2597
2598 if entry.kind == EntryKind::PendingDir
2599 && let Some(existing_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2600 {
2601 entry.kind = existing_entry.kind;
2602 }
2603
2604 let scan_id = self.scan_id;
2605 let removed = self.entries_by_path.insert_or_replace(entry.clone(), ());
2606 if let Some(removed) = removed
2607 && removed.id != entry.id
2608 {
2609 self.entries_by_id.remove(&removed.id, ());
2610 }
2611 self.entries_by_id.insert_or_replace(
2612 PathEntry {
2613 id: entry.id,
2614 path: entry.path.clone(),
2615 is_ignored: entry.is_ignored,
2616 scan_id,
2617 },
2618 (),
2619 );
2620
2621 entry
2622 }
2623
2624 fn ancestor_inodes_for_path(&self, path: &RelPath) -> TreeSet<u64> {
2625 let mut inodes = TreeSet::default();
2626 for ancestor in path.ancestors().skip(1) {
2627 if let Some(entry) = self.entry_for_path(ancestor) {
2628 inodes.insert(entry.inode);
2629 }
2630 }
2631 inodes
2632 }
2633
2634 async fn ignore_stack_for_abs_path(
2635 &self,
2636 abs_path: &Path,
2637 is_dir: bool,
2638 fs: &dyn Fs,
2639 ) -> IgnoreStack {
2640 let mut new_ignores = Vec::new();
2641 let mut repo_root = None;
2642 for (index, ancestor) in abs_path.ancestors().enumerate() {
2643 if index > 0 {
2644 if let Some((ignore, _)) = self.ignores_by_parent_abs_path.get(ancestor) {
2645 new_ignores.push((ancestor, Some(ignore.clone())));
2646 } else {
2647 new_ignores.push((ancestor, None));
2648 }
2649 }
2650
2651 let metadata = fs.metadata(&ancestor.join(DOT_GIT)).await.ok().flatten();
2652 if metadata.is_some() {
2653 repo_root = Some(Arc::from(ancestor));
2654 break;
2655 }
2656 }
2657
2658 let mut ignore_stack = if let Some(global_gitignore) = self.global_gitignore.clone() {
2659 IgnoreStack::global(global_gitignore)
2660 } else {
2661 IgnoreStack::none()
2662 };
2663
2664 if let Some((repo_exclude, _)) = repo_root
2665 .as_ref()
2666 .and_then(|abs_path| self.repo_exclude_by_work_dir_abs_path.get(abs_path))
2667 {
2668 ignore_stack = ignore_stack.append(IgnoreKind::RepoExclude, repo_exclude.clone());
2669 }
2670 ignore_stack.repo_root = repo_root;
2671 for (parent_abs_path, ignore) in new_ignores.into_iter().rev() {
2672 if ignore_stack.is_abs_path_ignored(parent_abs_path, true) {
2673 ignore_stack = IgnoreStack::all();
2674 break;
2675 } else if let Some(ignore) = ignore {
2676 ignore_stack =
2677 ignore_stack.append(IgnoreKind::Gitignore(parent_abs_path.into()), ignore);
2678 }
2679 }
2680
2681 if ignore_stack.is_abs_path_ignored(abs_path, is_dir) {
2682 ignore_stack = IgnoreStack::all();
2683 }
2684
2685 ignore_stack
2686 }
2687
2688 #[cfg(feature = "test-support")]
2689 pub fn expanded_entries(&self) -> impl Iterator<Item = &Entry> {
2690 self.entries_by_path
2691 .cursor::<()>(())
2692 .filter(|entry| entry.kind == EntryKind::Dir && (entry.is_external || entry.is_ignored))
2693 }
2694
2695 #[cfg(feature = "test-support")]
2696 pub fn check_invariants(&self, git_state: bool) {
2697 use pretty_assertions::assert_eq;
2698
2699 assert_eq!(
2700 self.entries_by_path
2701 .cursor::<()>(())
2702 .map(|e| (&e.path, e.id))
2703 .collect::<Vec<_>>(),
2704 self.entries_by_id
2705 .cursor::<()>(())
2706 .map(|e| (&e.path, e.id))
2707 .collect::<collections::BTreeSet<_>>()
2708 .into_iter()
2709 .collect::<Vec<_>>(),
2710 "entries_by_path and entries_by_id are inconsistent"
2711 );
2712
2713 let mut files = self.files(true, 0);
2714 let mut visible_files = self.files(false, 0);
2715 for entry in self.entries_by_path.cursor::<()>(()) {
2716 if entry.is_file() {
2717 assert_eq!(files.next().unwrap().inode, entry.inode);
2718 if (!entry.is_ignored && !entry.is_external) || entry.is_always_included {
2719 assert_eq!(visible_files.next().unwrap().inode, entry.inode);
2720 }
2721 }
2722 }
2723
2724 assert!(files.next().is_none());
2725 assert!(visible_files.next().is_none());
2726
2727 let mut bfs_paths = Vec::new();
2728 let mut stack = self
2729 .root_entry()
2730 .map(|e| e.path.as_ref())
2731 .into_iter()
2732 .collect::<Vec<_>>();
2733 while let Some(path) = stack.pop() {
2734 bfs_paths.push(path);
2735 let ix = stack.len();
2736 for child_entry in self.child_entries(path) {
2737 stack.insert(ix, &child_entry.path);
2738 }
2739 }
2740
2741 let dfs_paths_via_iter = self
2742 .entries_by_path
2743 .cursor::<()>(())
2744 .map(|e| e.path.as_ref())
2745 .collect::<Vec<_>>();
2746 assert_eq!(bfs_paths, dfs_paths_via_iter);
2747
2748 let dfs_paths_via_traversal = self
2749 .entries(true, 0)
2750 .map(|e| e.path.as_ref())
2751 .collect::<Vec<_>>();
2752
2753 assert_eq!(dfs_paths_via_traversal, dfs_paths_via_iter);
2754
2755 if git_state {
2756 for ignore_parent_abs_path in self.ignores_by_parent_abs_path.keys() {
2757 let ignore_parent_path = &RelPath::new(
2758 ignore_parent_abs_path
2759 .strip_prefix(self.abs_path.as_path())
2760 .unwrap(),
2761 PathStyle::local(),
2762 )
2763 .unwrap();
2764 assert!(self.entry_for_path(ignore_parent_path).is_some());
2765 assert!(
2766 self.entry_for_path(
2767 &ignore_parent_path.join(RelPath::unix(GITIGNORE).unwrap())
2768 )
2769 .is_some()
2770 );
2771 }
2772 }
2773 }
2774
2775 #[cfg(feature = "test-support")]
2776 pub fn entries_without_ids(&self, include_ignored: bool) -> Vec<(&RelPath, u64, bool)> {
2777 let mut paths = Vec::new();
2778 for entry in self.entries_by_path.cursor::<()>(()) {
2779 if include_ignored || !entry.is_ignored {
2780 paths.push((entry.path.as_ref(), entry.inode, entry.is_ignored));
2781 }
2782 }
2783 paths.sort_by(|a, b| a.0.cmp(b.0));
2784 paths
2785 }
2786}
2787
2788impl BackgroundScannerState {
2789 fn should_scan_directory(&self, entry: &Entry) -> bool {
2790 (self.scanning_enabled && !entry.is_external && (!entry.is_ignored || entry.is_always_included))
2791 || entry.path.file_name() == Some(DOT_GIT)
2792 || entry.path.file_name() == Some(local_settings_folder_name())
2793 || entry.path.file_name() == Some(local_vscode_folder_name())
2794 || self.scanned_dirs.contains(&entry.id) // If we've ever scanned it, keep scanning
2795 || self
2796 .paths_to_scan
2797 .iter()
2798 .any(|p| p.starts_with(&entry.path))
2799 || self
2800 .path_prefixes_to_scan
2801 .iter()
2802 .any(|p| entry.path.starts_with(p))
2803 }
2804
2805 async fn enqueue_scan_dir(
2806 &self,
2807 abs_path: Arc<Path>,
2808 entry: &Entry,
2809 scan_job_tx: &Sender<ScanJob>,
2810 fs: &dyn Fs,
2811 ) {
2812 let path = entry.path.clone();
2813 let ignore_stack = self
2814 .snapshot
2815 .ignore_stack_for_abs_path(&abs_path, true, fs)
2816 .await;
2817 let mut ancestor_inodes = self.snapshot.ancestor_inodes_for_path(&path);
2818
2819 if !ancestor_inodes.contains(&entry.inode) {
2820 ancestor_inodes.insert(entry.inode);
2821 scan_job_tx
2822 .try_send(ScanJob {
2823 abs_path,
2824 path,
2825 ignore_stack,
2826 scan_queue: scan_job_tx.clone(),
2827 ancestor_inodes,
2828 is_external: entry.is_external,
2829 })
2830 .unwrap();
2831 }
2832 }
2833
2834 fn reuse_entry_id(&mut self, entry: &mut Entry) {
2835 if let Some(mtime) = entry.mtime {
2836 // If an entry with the same inode was removed from the worktree during this scan,
2837 // then it *might* represent the same file or directory. But the OS might also have
2838 // re-used the inode for a completely different file or directory.
2839 //
2840 // Conditionally reuse the old entry's id:
2841 // * if the mtime is the same, the file was probably been renamed.
2842 // * if the path is the same, the file may just have been updated
2843 if let Some(removed_entry) = self.removed_entries.remove(&entry.inode) {
2844 if removed_entry.mtime == Some(mtime) || removed_entry.path == entry.path {
2845 entry.id = removed_entry.id;
2846 }
2847 } else if let Some(existing_entry) = self.snapshot.entry_for_path(&entry.path) {
2848 entry.id = existing_entry.id;
2849 }
2850 }
2851 }
2852
2853 fn entry_id_for(
2854 &mut self,
2855 next_entry_id: &AtomicUsize,
2856 path: &RelPath,
2857 metadata: &fs::Metadata,
2858 ) -> ProjectEntryId {
2859 // If an entry with the same inode was removed from the worktree during this scan,
2860 // then it *might* represent the same file or directory. But the OS might also have
2861 // re-used the inode for a completely different file or directory.
2862 //
2863 // Conditionally reuse the old entry's id:
2864 // * if the mtime is the same, the file was probably been renamed.
2865 // * if the path is the same, the file may just have been updated
2866 if let Some(removed_entry) = self.removed_entries.remove(&metadata.inode) {
2867 if removed_entry.mtime == Some(metadata.mtime) || *removed_entry.path == *path {
2868 return removed_entry.id;
2869 }
2870 } else if let Some(existing_entry) = self.snapshot.entry_for_path(path) {
2871 return existing_entry.id;
2872 }
2873 ProjectEntryId::new(next_entry_id)
2874 }
2875
2876 async fn insert_entry(&mut self, entry: Entry, fs: &dyn Fs, watcher: &dyn Watcher) -> Entry {
2877 let entry = self.snapshot.insert_entry(entry, fs).await;
2878 if entry.path.file_name() == Some(&DOT_GIT) {
2879 self.insert_git_repository(entry.path.clone(), fs, watcher)
2880 .await;
2881 }
2882
2883 #[cfg(feature = "test-support")]
2884 self.snapshot.check_invariants(false);
2885
2886 entry
2887 }
2888
2889 fn populate_dir(
2890 &mut self,
2891 parent_path: Arc<RelPath>,
2892 entries: impl IntoIterator<Item = Entry>,
2893 ignore: Option<Arc<Gitignore>>,
2894 ) {
2895 let mut parent_entry = if let Some(parent_entry) = self
2896 .snapshot
2897 .entries_by_path
2898 .get(&PathKey(parent_path.clone()), ())
2899 {
2900 parent_entry.clone()
2901 } else {
2902 log::warn!(
2903 "populating a directory {:?} that has been removed",
2904 parent_path
2905 );
2906 return;
2907 };
2908
2909 match parent_entry.kind {
2910 EntryKind::PendingDir | EntryKind::UnloadedDir => parent_entry.kind = EntryKind::Dir,
2911 EntryKind::Dir => {}
2912 _ => return,
2913 }
2914
2915 if let Some(ignore) = ignore {
2916 let abs_parent_path = self
2917 .snapshot
2918 .abs_path
2919 .as_path()
2920 .join(parent_path.as_std_path())
2921 .into();
2922 self.snapshot
2923 .ignores_by_parent_abs_path
2924 .insert(abs_parent_path, (ignore, false));
2925 }
2926
2927 let parent_entry_id = parent_entry.id;
2928 self.scanned_dirs.insert(parent_entry_id);
2929 let mut entries_by_path_edits = vec![Edit::Insert(parent_entry)];
2930 let mut entries_by_id_edits = Vec::new();
2931
2932 for entry in entries {
2933 entries_by_id_edits.push(Edit::Insert(PathEntry {
2934 id: entry.id,
2935 path: entry.path.clone(),
2936 is_ignored: entry.is_ignored,
2937 scan_id: self.snapshot.scan_id,
2938 }));
2939 entries_by_path_edits.push(Edit::Insert(entry));
2940 }
2941
2942 self.snapshot
2943 .entries_by_path
2944 .edit(entries_by_path_edits, ());
2945 self.snapshot.entries_by_id.edit(entries_by_id_edits, ());
2946
2947 if let Err(ix) = self.changed_paths.binary_search(&parent_path) {
2948 self.changed_paths.insert(ix, parent_path.clone());
2949 }
2950
2951 #[cfg(feature = "test-support")]
2952 self.snapshot.check_invariants(false);
2953 }
2954
2955 fn remove_path(&mut self, path: &RelPath) {
2956 log::trace!("background scanner removing path {path:?}");
2957 let mut new_entries;
2958 let removed_entries;
2959 {
2960 let mut cursor = self
2961 .snapshot
2962 .entries_by_path
2963 .cursor::<TraversalProgress>(());
2964 new_entries = cursor.slice(&TraversalTarget::path(path), Bias::Left);
2965 removed_entries = cursor.slice(&TraversalTarget::successor(path), Bias::Left);
2966 new_entries.append(cursor.suffix(), ());
2967 }
2968 self.snapshot.entries_by_path = new_entries;
2969
2970 let mut removed_ids = Vec::with_capacity(removed_entries.summary().count);
2971 for entry in removed_entries.cursor::<()>(()) {
2972 match self.removed_entries.entry(entry.inode) {
2973 hash_map::Entry::Occupied(mut e) => {
2974 let prev_removed_entry = e.get_mut();
2975 if entry.id > prev_removed_entry.id {
2976 *prev_removed_entry = entry.clone();
2977 }
2978 }
2979 hash_map::Entry::Vacant(e) => {
2980 e.insert(entry.clone());
2981 }
2982 }
2983
2984 if entry.path.file_name() == Some(GITIGNORE) {
2985 let abs_parent_path = self.snapshot.absolutize(&entry.path.parent().unwrap());
2986 if let Some((_, needs_update)) = self
2987 .snapshot
2988 .ignores_by_parent_abs_path
2989 .get_mut(abs_parent_path.as_path())
2990 {
2991 *needs_update = true;
2992 }
2993 }
2994
2995 if let Err(ix) = removed_ids.binary_search(&entry.id) {
2996 removed_ids.insert(ix, entry.id);
2997 }
2998 }
2999
3000 self.snapshot
3001 .entries_by_id
3002 .edit(removed_ids.iter().map(|&id| Edit::Remove(id)).collect(), ());
3003 self.snapshot
3004 .git_repositories
3005 .retain(|id, _| removed_ids.binary_search(id).is_err());
3006
3007 #[cfg(feature = "test-support")]
3008 self.snapshot.check_invariants(false);
3009 }
3010
3011 async fn insert_git_repository(
3012 &mut self,
3013 dot_git_path: Arc<RelPath>,
3014 fs: &dyn Fs,
3015 watcher: &dyn Watcher,
3016 ) {
3017 let work_dir_path: Arc<RelPath> = match dot_git_path.parent() {
3018 Some(parent_dir) => {
3019 // Guard against repositories inside the repository metadata
3020 if parent_dir
3021 .components()
3022 .any(|component| component == DOT_GIT)
3023 {
3024 log::debug!(
3025 "not building git repository for nested `.git` directory, `.git` path in the worktree: {dot_git_path:?}"
3026 );
3027 return;
3028 };
3029
3030 parent_dir.into()
3031 }
3032 None => {
3033 // `dot_git_path.parent().is_none()` means `.git` directory is the opened worktree itself,
3034 // no files inside that directory are tracked by git, so no need to build the repo around it
3035 log::debug!(
3036 "not building git repository for the worktree itself, `.git` path in the worktree: {dot_git_path:?}"
3037 );
3038 return;
3039 }
3040 };
3041
3042 let dot_git_abs_path = Arc::from(self.snapshot.absolutize(&dot_git_path).as_ref());
3043
3044 self.insert_git_repository_for_path(
3045 WorkDirectory::InProject {
3046 relative_path: work_dir_path,
3047 },
3048 dot_git_abs_path,
3049 fs,
3050 watcher,
3051 )
3052 .await
3053 .log_err();
3054 }
3055
3056 async fn insert_git_repository_for_path(
3057 &mut self,
3058 work_directory: WorkDirectory,
3059 dot_git_abs_path: Arc<Path>,
3060 fs: &dyn Fs,
3061 watcher: &dyn Watcher,
3062 ) -> Result<LocalRepositoryEntry> {
3063 let work_dir_entry = self
3064 .snapshot
3065 .entry_for_path(&work_directory.path_key().0)
3066 .with_context(|| {
3067 format!(
3068 "working directory `{}` not indexed",
3069 work_directory
3070 .path_key()
3071 .0
3072 .display(self.snapshot.path_style)
3073 )
3074 })?;
3075 let work_directory_abs_path = self.snapshot.work_directory_abs_path(&work_directory);
3076
3077 let (repository_dir_abs_path, common_dir_abs_path) =
3078 discover_git_paths(&dot_git_abs_path, fs).await;
3079 watcher
3080 .add(&common_dir_abs_path)
3081 .context("failed to add common directory to watcher")
3082 .log_err();
3083 if !repository_dir_abs_path.starts_with(&common_dir_abs_path) {
3084 watcher
3085 .add(&repository_dir_abs_path)
3086 .context("failed to add repository directory to watcher")
3087 .log_err();
3088 }
3089
3090 let work_directory_id = work_dir_entry.id;
3091
3092 let local_repository = LocalRepositoryEntry {
3093 work_directory_id,
3094 work_directory,
3095 work_directory_abs_path: work_directory_abs_path.as_path().into(),
3096 git_dir_scan_id: 0,
3097 dot_git_abs_path,
3098 common_dir_abs_path,
3099 repository_dir_abs_path,
3100 };
3101
3102 self.snapshot
3103 .git_repositories
3104 .insert(work_directory_id, local_repository.clone());
3105
3106 log::trace!("inserting new local git repository");
3107 Ok(local_repository)
3108 }
3109}
3110
3111async fn is_git_dir(path: &Path, fs: &dyn Fs) -> bool {
3112 if let Some(file_name) = path.file_name()
3113 && file_name == DOT_GIT
3114 {
3115 return true;
3116 }
3117
3118 // If we're in a bare repository, we are not inside a `.git` folder. In a
3119 // bare repository, the root folder contains what would normally be in the
3120 // `.git` folder.
3121 let head_metadata = fs.metadata(&path.join("HEAD")).await;
3122 if !matches!(head_metadata, Ok(Some(_))) {
3123 return false;
3124 }
3125 let config_metadata = fs.metadata(&path.join("config")).await;
3126 matches!(config_metadata, Ok(Some(_)))
3127}
3128
3129async fn build_gitignore(abs_path: &Path, fs: &dyn Fs) -> Result<Gitignore> {
3130 let contents = fs
3131 .load(abs_path)
3132 .await
3133 .with_context(|| format!("failed to load gitignore file at {}", abs_path.display()))?;
3134 let parent = abs_path.parent().unwrap_or_else(|| Path::new("/"));
3135 let mut builder = GitignoreBuilder::new(parent);
3136 for line in contents.lines() {
3137 builder.add_line(Some(abs_path.into()), line)?;
3138 }
3139 Ok(builder.build()?)
3140}
3141
3142impl Deref for Worktree {
3143 type Target = Snapshot;
3144
3145 fn deref(&self) -> &Self::Target {
3146 match self {
3147 Worktree::Local(worktree) => &worktree.snapshot,
3148 Worktree::Remote(worktree) => &worktree.snapshot,
3149 }
3150 }
3151}
3152
3153impl Deref for LocalWorktree {
3154 type Target = LocalSnapshot;
3155
3156 fn deref(&self) -> &Self::Target {
3157 &self.snapshot
3158 }
3159}
3160
3161impl Deref for RemoteWorktree {
3162 type Target = Snapshot;
3163
3164 fn deref(&self) -> &Self::Target {
3165 &self.snapshot
3166 }
3167}
3168
3169impl fmt::Debug for LocalWorktree {
3170 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3171 self.snapshot.fmt(f)
3172 }
3173}
3174
3175impl fmt::Debug for Snapshot {
3176 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3177 struct EntriesById<'a>(&'a SumTree<PathEntry>);
3178 struct EntriesByPath<'a>(&'a SumTree<Entry>);
3179
3180 impl fmt::Debug for EntriesByPath<'_> {
3181 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3182 f.debug_map()
3183 .entries(self.0.iter().map(|entry| (&entry.path, entry.id)))
3184 .finish()
3185 }
3186 }
3187
3188 impl fmt::Debug for EntriesById<'_> {
3189 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3190 f.debug_list().entries(self.0.iter()).finish()
3191 }
3192 }
3193
3194 f.debug_struct("Snapshot")
3195 .field("id", &self.id)
3196 .field("root_name", &self.root_name)
3197 .field("entries_by_path", &EntriesByPath(&self.entries_by_path))
3198 .field("entries_by_id", &EntriesById(&self.entries_by_id))
3199 .finish()
3200 }
3201}
3202
3203#[derive(Debug, Clone, PartialEq)]
3204pub struct File {
3205 pub worktree: Entity<Worktree>,
3206 pub path: Arc<RelPath>,
3207 pub disk_state: DiskState,
3208 pub entry_id: Option<ProjectEntryId>,
3209 pub is_local: bool,
3210 pub is_private: bool,
3211}
3212
3213impl language::File for File {
3214 fn as_local(&self) -> Option<&dyn language::LocalFile> {
3215 if self.is_local { Some(self) } else { None }
3216 }
3217
3218 fn disk_state(&self) -> DiskState {
3219 self.disk_state
3220 }
3221
3222 fn path(&self) -> &Arc<RelPath> {
3223 &self.path
3224 }
3225
3226 fn full_path(&self, cx: &App) -> PathBuf {
3227 self.worktree.read(cx).full_path(&self.path)
3228 }
3229
3230 /// Returns the last component of this handle's absolute path. If this handle refers to the root
3231 /// of its worktree, then this method will return the name of the worktree itself.
3232 fn file_name<'a>(&'a self, cx: &'a App) -> &'a str {
3233 self.path
3234 .file_name()
3235 .unwrap_or_else(|| self.worktree.read(cx).root_name_str())
3236 }
3237
3238 fn worktree_id(&self, cx: &App) -> WorktreeId {
3239 self.worktree.read(cx).id()
3240 }
3241
3242 fn to_proto(&self, cx: &App) -> rpc::proto::File {
3243 rpc::proto::File {
3244 worktree_id: self.worktree.read(cx).id().to_proto(),
3245 entry_id: self.entry_id.map(|id| id.to_proto()),
3246 path: self.path.as_ref().to_proto(),
3247 mtime: self.disk_state.mtime().map(|time| time.into()),
3248 is_deleted: self.disk_state.is_deleted(),
3249 is_historic: matches!(self.disk_state, DiskState::Historic { .. }),
3250 }
3251 }
3252
3253 fn is_private(&self) -> bool {
3254 self.is_private
3255 }
3256
3257 fn path_style(&self, cx: &App) -> PathStyle {
3258 self.worktree.read(cx).path_style()
3259 }
3260
3261 fn can_open(&self) -> bool {
3262 true
3263 }
3264}
3265
3266impl language::LocalFile for File {
3267 fn abs_path(&self, cx: &App) -> PathBuf {
3268 self.worktree.read(cx).absolutize(&self.path)
3269 }
3270
3271 fn load(&self, cx: &App) -> Task<Result<String>> {
3272 let worktree = self.worktree.read(cx).as_local().unwrap();
3273 let abs_path = worktree.absolutize(&self.path);
3274 let fs = worktree.fs.clone();
3275 cx.background_spawn(async move { fs.load(&abs_path).await })
3276 }
3277
3278 fn load_bytes(&self, cx: &App) -> Task<Result<Vec<u8>>> {
3279 let worktree = self.worktree.read(cx).as_local().unwrap();
3280 let abs_path = worktree.absolutize(&self.path);
3281 let fs = worktree.fs.clone();
3282 cx.background_spawn(async move { fs.load_bytes(&abs_path).await })
3283 }
3284}
3285
3286impl File {
3287 pub fn for_entry(entry: Entry, worktree: Entity<Worktree>) -> Arc<Self> {
3288 Arc::new(Self {
3289 worktree,
3290 path: entry.path.clone(),
3291 disk_state: if let Some(mtime) = entry.mtime {
3292 DiskState::Present { mtime }
3293 } else {
3294 DiskState::New
3295 },
3296 entry_id: Some(entry.id),
3297 is_local: true,
3298 is_private: entry.is_private,
3299 })
3300 }
3301
3302 pub fn from_proto(
3303 proto: rpc::proto::File,
3304 worktree: Entity<Worktree>,
3305 cx: &App,
3306 ) -> Result<Self> {
3307 let worktree_id = worktree.read(cx).as_remote().context("not remote")?.id();
3308
3309 anyhow::ensure!(
3310 worktree_id.to_proto() == proto.worktree_id,
3311 "worktree id does not match file"
3312 );
3313
3314 let disk_state = if proto.is_historic {
3315 DiskState::Historic {
3316 was_deleted: proto.is_deleted,
3317 }
3318 } else if proto.is_deleted {
3319 DiskState::Deleted
3320 } else if let Some(mtime) = proto.mtime.map(&Into::into) {
3321 DiskState::Present { mtime }
3322 } else {
3323 DiskState::New
3324 };
3325
3326 Ok(Self {
3327 worktree,
3328 path: RelPath::from_proto(&proto.path).context("invalid path in file protobuf")?,
3329 disk_state,
3330 entry_id: proto.entry_id.map(ProjectEntryId::from_proto),
3331 is_local: false,
3332 is_private: false,
3333 })
3334 }
3335
3336 pub fn from_dyn(file: Option<&Arc<dyn language::File>>) -> Option<&Self> {
3337 file.and_then(|f| {
3338 let f: &dyn language::File = f.borrow();
3339 let f: &dyn Any = f;
3340 f.downcast_ref()
3341 })
3342 }
3343
3344 pub fn worktree_id(&self, cx: &App) -> WorktreeId {
3345 self.worktree.read(cx).id()
3346 }
3347
3348 pub fn project_entry_id(&self) -> Option<ProjectEntryId> {
3349 match self.disk_state {
3350 DiskState::Deleted => None,
3351 _ => self.entry_id,
3352 }
3353 }
3354}
3355
3356#[derive(Clone, Debug, PartialEq, Eq)]
3357pub struct Entry {
3358 pub id: ProjectEntryId,
3359 pub kind: EntryKind,
3360 pub path: Arc<RelPath>,
3361 pub inode: u64,
3362 pub mtime: Option<MTime>,
3363
3364 pub canonical_path: Option<Arc<Path>>,
3365 /// Whether this entry is ignored by Git.
3366 ///
3367 /// We only scan ignored entries once the directory is expanded and
3368 /// exclude them from searches.
3369 pub is_ignored: bool,
3370
3371 /// Whether this entry is hidden or inside hidden directory.
3372 ///
3373 /// We only scan hidden entries once the directory is expanded.
3374 pub is_hidden: bool,
3375
3376 /// Whether this entry is always included in searches.
3377 ///
3378 /// This is used for entries that are always included in searches, even
3379 /// if they are ignored by git. Overridden by file_scan_exclusions.
3380 pub is_always_included: bool,
3381
3382 /// Whether this entry's canonical path is outside of the worktree.
3383 /// This means the entry is only accessible from the worktree root via a
3384 /// symlink.
3385 ///
3386 /// We only scan entries outside of the worktree once the symlinked
3387 /// directory is expanded. External entries are treated like gitignored
3388 /// entries in that they are not included in searches.
3389 pub is_external: bool,
3390
3391 /// Whether this entry is considered to be a `.env` file.
3392 pub is_private: bool,
3393 /// The entry's size on disk, in bytes.
3394 pub size: u64,
3395 pub char_bag: CharBag,
3396 pub is_fifo: bool,
3397}
3398
3399#[derive(Clone, Copy, Debug, PartialEq, Eq)]
3400pub enum EntryKind {
3401 UnloadedDir,
3402 PendingDir,
3403 Dir,
3404 File,
3405}
3406
3407#[derive(Clone, Copy, Debug, PartialEq)]
3408pub enum PathChange {
3409 /// A filesystem entry was was created.
3410 Added,
3411 /// A filesystem entry was removed.
3412 Removed,
3413 /// A filesystem entry was updated.
3414 Updated,
3415 /// A filesystem entry was either updated or added. We don't know
3416 /// whether or not it already existed, because the path had not
3417 /// been loaded before the event.
3418 AddedOrUpdated,
3419 /// A filesystem entry was found during the initial scan of the worktree.
3420 Loaded,
3421}
3422
3423#[derive(Clone, Debug, PartialEq, Eq)]
3424pub struct UpdatedGitRepository {
3425 /// ID of the repository's working directory.
3426 ///
3427 /// For a repo that's above the worktree root, this is the ID of the worktree root, and hence not unique.
3428 /// It's included here to aid the GitStore in detecting when a repository's working directory is renamed.
3429 pub work_directory_id: ProjectEntryId,
3430 pub old_work_directory_abs_path: Option<Arc<Path>>,
3431 pub new_work_directory_abs_path: Option<Arc<Path>>,
3432 /// For a normal git repository checkout, the absolute path to the .git directory.
3433 /// For a worktree, the absolute path to the worktree's subdirectory inside the .git directory.
3434 pub dot_git_abs_path: Option<Arc<Path>>,
3435 pub repository_dir_abs_path: Option<Arc<Path>>,
3436 pub common_dir_abs_path: Option<Arc<Path>>,
3437}
3438
3439pub type UpdatedEntriesSet = Arc<[(Arc<RelPath>, ProjectEntryId, PathChange)]>;
3440pub type UpdatedGitRepositoriesSet = Arc<[UpdatedGitRepository]>;
3441
3442#[derive(Clone, Debug)]
3443pub struct PathProgress<'a> {
3444 pub max_path: &'a RelPath,
3445}
3446
3447#[derive(Clone, Debug)]
3448pub struct PathSummary<S> {
3449 pub max_path: Arc<RelPath>,
3450 pub item_summary: S,
3451}
3452
3453impl<S: Summary> Summary for PathSummary<S> {
3454 type Context<'a> = S::Context<'a>;
3455
3456 fn zero(cx: Self::Context<'_>) -> Self {
3457 Self {
3458 max_path: RelPath::empty().into(),
3459 item_summary: S::zero(cx),
3460 }
3461 }
3462
3463 fn add_summary(&mut self, rhs: &Self, cx: Self::Context<'_>) {
3464 self.max_path = rhs.max_path.clone();
3465 self.item_summary.add_summary(&rhs.item_summary, cx);
3466 }
3467}
3468
3469impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathProgress<'a> {
3470 fn zero(_: <PathSummary<S> as Summary>::Context<'_>) -> Self {
3471 Self {
3472 max_path: RelPath::empty(),
3473 }
3474 }
3475
3476 fn add_summary(
3477 &mut self,
3478 summary: &'a PathSummary<S>,
3479 _: <PathSummary<S> as Summary>::Context<'_>,
3480 ) {
3481 self.max_path = summary.max_path.as_ref()
3482 }
3483}
3484
3485impl<'a> sum_tree::Dimension<'a, PathSummary<GitSummary>> for GitSummary {
3486 fn zero(_cx: ()) -> Self {
3487 Default::default()
3488 }
3489
3490 fn add_summary(&mut self, summary: &'a PathSummary<GitSummary>, _: ()) {
3491 *self += summary.item_summary
3492 }
3493}
3494
3495impl<'a>
3496 sum_tree::SeekTarget<'a, PathSummary<GitSummary>, Dimensions<TraversalProgress<'a>, GitSummary>>
3497 for PathTarget<'_>
3498{
3499 fn cmp(
3500 &self,
3501 cursor_location: &Dimensions<TraversalProgress<'a>, GitSummary>,
3502 _: (),
3503 ) -> Ordering {
3504 self.cmp_path(cursor_location.0.max_path)
3505 }
3506}
3507
3508impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathKey {
3509 fn zero(_: S::Context<'_>) -> Self {
3510 Default::default()
3511 }
3512
3513 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3514 self.0 = summary.max_path.clone();
3515 }
3516}
3517
3518impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for TraversalProgress<'a> {
3519 fn zero(_cx: S::Context<'_>) -> Self {
3520 Default::default()
3521 }
3522
3523 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3524 self.max_path = summary.max_path.as_ref();
3525 }
3526}
3527
3528impl Entry {
3529 fn new(
3530 path: Arc<RelPath>,
3531 metadata: &fs::Metadata,
3532 id: ProjectEntryId,
3533 root_char_bag: CharBag,
3534 canonical_path: Option<Arc<Path>>,
3535 ) -> Self {
3536 let char_bag = char_bag_for_path(root_char_bag, &path);
3537 Self {
3538 id,
3539 kind: if metadata.is_dir {
3540 EntryKind::PendingDir
3541 } else {
3542 EntryKind::File
3543 },
3544 path,
3545 inode: metadata.inode,
3546 mtime: Some(metadata.mtime),
3547 size: metadata.len,
3548 canonical_path,
3549 is_ignored: false,
3550 is_hidden: false,
3551 is_always_included: false,
3552 is_external: false,
3553 is_private: false,
3554 char_bag,
3555 is_fifo: metadata.is_fifo,
3556 }
3557 }
3558
3559 pub fn is_created(&self) -> bool {
3560 self.mtime.is_some()
3561 }
3562
3563 pub fn is_dir(&self) -> bool {
3564 self.kind.is_dir()
3565 }
3566
3567 pub fn is_file(&self) -> bool {
3568 self.kind.is_file()
3569 }
3570}
3571
3572impl EntryKind {
3573 pub fn is_dir(&self) -> bool {
3574 matches!(
3575 self,
3576 EntryKind::Dir | EntryKind::PendingDir | EntryKind::UnloadedDir
3577 )
3578 }
3579
3580 pub fn is_unloaded(&self) -> bool {
3581 matches!(self, EntryKind::UnloadedDir)
3582 }
3583
3584 pub fn is_file(&self) -> bool {
3585 matches!(self, EntryKind::File)
3586 }
3587}
3588
3589impl sum_tree::Item for Entry {
3590 type Summary = EntrySummary;
3591
3592 fn summary(&self, _cx: ()) -> Self::Summary {
3593 let non_ignored_count = if (self.is_ignored || self.is_external) && !self.is_always_included
3594 {
3595 0
3596 } else {
3597 1
3598 };
3599 let file_count;
3600 let non_ignored_file_count;
3601 if self.is_file() {
3602 file_count = 1;
3603 non_ignored_file_count = non_ignored_count;
3604 } else {
3605 file_count = 0;
3606 non_ignored_file_count = 0;
3607 }
3608
3609 EntrySummary {
3610 max_path: self.path.clone(),
3611 count: 1,
3612 non_ignored_count,
3613 file_count,
3614 non_ignored_file_count,
3615 }
3616 }
3617}
3618
3619impl sum_tree::KeyedItem for Entry {
3620 type Key = PathKey;
3621
3622 fn key(&self) -> Self::Key {
3623 PathKey(self.path.clone())
3624 }
3625}
3626
3627#[derive(Clone, Debug)]
3628pub struct EntrySummary {
3629 max_path: Arc<RelPath>,
3630 count: usize,
3631 non_ignored_count: usize,
3632 file_count: usize,
3633 non_ignored_file_count: usize,
3634}
3635
3636impl Default for EntrySummary {
3637 fn default() -> Self {
3638 Self {
3639 max_path: Arc::from(RelPath::empty()),
3640 count: 0,
3641 non_ignored_count: 0,
3642 file_count: 0,
3643 non_ignored_file_count: 0,
3644 }
3645 }
3646}
3647
3648impl sum_tree::ContextLessSummary for EntrySummary {
3649 fn zero() -> Self {
3650 Default::default()
3651 }
3652
3653 fn add_summary(&mut self, rhs: &Self) {
3654 self.max_path = rhs.max_path.clone();
3655 self.count += rhs.count;
3656 self.non_ignored_count += rhs.non_ignored_count;
3657 self.file_count += rhs.file_count;
3658 self.non_ignored_file_count += rhs.non_ignored_file_count;
3659 }
3660}
3661
3662#[derive(Clone, Debug)]
3663struct PathEntry {
3664 id: ProjectEntryId,
3665 path: Arc<RelPath>,
3666 is_ignored: bool,
3667 scan_id: usize,
3668}
3669
3670impl sum_tree::Item for PathEntry {
3671 type Summary = PathEntrySummary;
3672
3673 fn summary(&self, _cx: ()) -> Self::Summary {
3674 PathEntrySummary { max_id: self.id }
3675 }
3676}
3677
3678impl sum_tree::KeyedItem for PathEntry {
3679 type Key = ProjectEntryId;
3680
3681 fn key(&self) -> Self::Key {
3682 self.id
3683 }
3684}
3685
3686#[derive(Clone, Debug, Default)]
3687struct PathEntrySummary {
3688 max_id: ProjectEntryId,
3689}
3690
3691impl sum_tree::ContextLessSummary for PathEntrySummary {
3692 fn zero() -> Self {
3693 Default::default()
3694 }
3695
3696 fn add_summary(&mut self, summary: &Self) {
3697 self.max_id = summary.max_id;
3698 }
3699}
3700
3701impl<'a> sum_tree::Dimension<'a, PathEntrySummary> for ProjectEntryId {
3702 fn zero(_cx: ()) -> Self {
3703 Default::default()
3704 }
3705
3706 fn add_summary(&mut self, summary: &'a PathEntrySummary, _: ()) {
3707 *self = summary.max_id;
3708 }
3709}
3710
3711#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
3712pub struct PathKey(pub Arc<RelPath>);
3713
3714impl Default for PathKey {
3715 fn default() -> Self {
3716 Self(RelPath::empty().into())
3717 }
3718}
3719
3720impl<'a> sum_tree::Dimension<'a, EntrySummary> for PathKey {
3721 fn zero(_cx: ()) -> Self {
3722 Default::default()
3723 }
3724
3725 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
3726 self.0 = summary.max_path.clone();
3727 }
3728}
3729
3730struct BackgroundScanner {
3731 state: async_lock::Mutex<BackgroundScannerState>,
3732 fs: Arc<dyn Fs>,
3733 fs_case_sensitive: bool,
3734 status_updates_tx: UnboundedSender<ScanState>,
3735 executor: BackgroundExecutor,
3736 scan_requests_rx: channel::Receiver<ScanRequest>,
3737 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
3738 next_entry_id: Arc<AtomicUsize>,
3739 phase: BackgroundScannerPhase,
3740 watcher: Arc<dyn Watcher>,
3741 settings: WorktreeSettings,
3742 share_private_files: bool,
3743}
3744
3745#[derive(Copy, Clone, PartialEq)]
3746enum BackgroundScannerPhase {
3747 InitialScan,
3748 EventsReceivedDuringInitialScan,
3749 Events,
3750}
3751
3752impl BackgroundScanner {
3753 async fn run(&mut self, mut fs_events_rx: Pin<Box<dyn Send + Stream<Item = Vec<PathEvent>>>>) {
3754 let root_abs_path;
3755 let scanning_enabled;
3756 {
3757 let state = self.state.lock().await;
3758 root_abs_path = state.snapshot.abs_path.clone();
3759 scanning_enabled = state.scanning_enabled;
3760 }
3761
3762 // If the worktree root does not contain a git repository, then find
3763 // the git repository in an ancestor directory. Find any gitignore files
3764 // in ancestor directories.
3765 let repo = if scanning_enabled {
3766 let (ignores, exclude, repo) =
3767 discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await;
3768 self.state
3769 .lock()
3770 .await
3771 .snapshot
3772 .ignores_by_parent_abs_path
3773 .extend(ignores);
3774 if let Some(exclude) = exclude {
3775 self.state
3776 .lock()
3777 .await
3778 .snapshot
3779 .repo_exclude_by_work_dir_abs_path
3780 .insert(root_abs_path.as_path().into(), (exclude, false));
3781 }
3782
3783 repo
3784 } else {
3785 None
3786 };
3787
3788 let containing_git_repository = if let Some((ancestor_dot_git, work_directory)) = repo
3789 && scanning_enabled
3790 {
3791 maybe!(async {
3792 self.state
3793 .lock()
3794 .await
3795 .insert_git_repository_for_path(
3796 work_directory,
3797 ancestor_dot_git.clone().into(),
3798 self.fs.as_ref(),
3799 self.watcher.as_ref(),
3800 )
3801 .await
3802 .log_err()?;
3803 Some(ancestor_dot_git)
3804 })
3805 .await
3806 } else {
3807 None
3808 };
3809
3810 log::trace!("containing git repository: {containing_git_repository:?}");
3811
3812 let mut global_gitignore_events = if let Some(global_gitignore_path) =
3813 &paths::global_gitignore_path()
3814 && scanning_enabled
3815 {
3816 let is_file = self.fs.is_file(&global_gitignore_path).await;
3817 self.state.lock().await.snapshot.global_gitignore = if is_file {
3818 build_gitignore(global_gitignore_path, self.fs.as_ref())
3819 .await
3820 .ok()
3821 .map(Arc::new)
3822 } else {
3823 None
3824 };
3825 if is_file
3826 || matches!(global_gitignore_path.parent(), Some(path) if self.fs.is_dir(path).await)
3827 {
3828 self.fs
3829 .watch(global_gitignore_path, FS_WATCH_LATENCY)
3830 .await
3831 .0
3832 } else {
3833 Box::pin(futures::stream::pending())
3834 }
3835 } else {
3836 self.state.lock().await.snapshot.global_gitignore = None;
3837 Box::pin(futures::stream::pending())
3838 };
3839
3840 let (scan_job_tx, scan_job_rx) = channel::unbounded();
3841 {
3842 let mut state = self.state.lock().await;
3843 state.snapshot.scan_id += 1;
3844 if let Some(mut root_entry) = state.snapshot.root_entry().cloned() {
3845 let ignore_stack = state
3846 .snapshot
3847 .ignore_stack_for_abs_path(root_abs_path.as_path(), true, self.fs.as_ref())
3848 .await;
3849 if ignore_stack.is_abs_path_ignored(root_abs_path.as_path(), true) {
3850 root_entry.is_ignored = true;
3851 let mut root_entry = root_entry.clone();
3852 state.reuse_entry_id(&mut root_entry);
3853 state
3854 .insert_entry(root_entry, self.fs.as_ref(), self.watcher.as_ref())
3855 .await;
3856 }
3857 if root_entry.is_dir() && state.scanning_enabled {
3858 state
3859 .enqueue_scan_dir(
3860 root_abs_path.as_path().into(),
3861 &root_entry,
3862 &scan_job_tx,
3863 self.fs.as_ref(),
3864 )
3865 .await;
3866 }
3867 }
3868 };
3869
3870 // Perform an initial scan of the directory.
3871 drop(scan_job_tx);
3872 self.scan_dirs(true, scan_job_rx).await;
3873 {
3874 let mut state = self.state.lock().await;
3875 state.snapshot.completed_scan_id = state.snapshot.scan_id;
3876 }
3877
3878 self.send_status_update(false, SmallVec::new()).await;
3879
3880 // Process any any FS events that occurred while performing the initial scan.
3881 // For these events, update events cannot be as precise, because we didn't
3882 // have the previous state loaded yet.
3883 self.phase = BackgroundScannerPhase::EventsReceivedDuringInitialScan;
3884 if let Poll::Ready(Some(mut paths)) = futures::poll!(fs_events_rx.next()) {
3885 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
3886 paths.extend(more_paths);
3887 }
3888 self.process_events(
3889 paths
3890 .into_iter()
3891 .filter(|e| e.kind.is_some())
3892 .map(Into::into)
3893 .collect(),
3894 )
3895 .await;
3896 }
3897 if let Some(abs_path) = containing_git_repository {
3898 self.process_events(vec![abs_path]).await;
3899 }
3900
3901 // Continue processing events until the worktree is dropped.
3902 self.phase = BackgroundScannerPhase::Events;
3903
3904 loop {
3905 select_biased! {
3906 // Process any path refresh requests from the worktree. Prioritize
3907 // these before handling changes reported by the filesystem.
3908 request = self.next_scan_request().fuse() => {
3909 let Ok(request) = request else { break };
3910 if !self.process_scan_request(request, false).await {
3911 return;
3912 }
3913 }
3914
3915 path_prefix_request = self.path_prefixes_to_scan_rx.recv().fuse() => {
3916 let Ok(request) = path_prefix_request else { break };
3917 log::trace!("adding path prefix {:?}", request.path);
3918
3919 let did_scan = self.forcibly_load_paths(std::slice::from_ref(&request.path)).await;
3920 if did_scan {
3921 let abs_path =
3922 {
3923 let mut state = self.state.lock().await;
3924 state.path_prefixes_to_scan.insert(request.path.clone());
3925 state.snapshot.absolutize(&request.path)
3926 };
3927
3928 if let Some(abs_path) = self.fs.canonicalize(&abs_path).await.log_err() {
3929 self.process_events(vec![abs_path]).await;
3930 }
3931 }
3932 self.send_status_update(false, request.done).await;
3933 }
3934
3935 paths = fs_events_rx.next().fuse() => {
3936 let Some(mut paths) = paths else { break };
3937 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
3938 paths.extend(more_paths);
3939 }
3940 self.process_events(paths.into_iter().filter(|e| e.kind.is_some()).map(Into::into).collect()).await;
3941 }
3942
3943 paths = global_gitignore_events.next().fuse() => {
3944 match paths.as_deref() {
3945 Some([event, ..]) => {
3946 self.update_global_gitignore(&event.path).await;
3947 }
3948 _ => (),
3949 }
3950 }
3951 }
3952 }
3953 }
3954
3955 async fn process_scan_request(&self, mut request: ScanRequest, scanning: bool) -> bool {
3956 log::debug!("rescanning paths {:?}", request.relative_paths);
3957
3958 request.relative_paths.sort_unstable();
3959 self.forcibly_load_paths(&request.relative_paths).await;
3960
3961 let root_path = self.state.lock().await.snapshot.abs_path.clone();
3962 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
3963 let root_canonical_path = match &root_canonical_path {
3964 Ok(path) => SanitizedPath::new(path),
3965 Err(err) => {
3966 log::error!("failed to canonicalize root path {root_path:?}: {err:#}");
3967 return true;
3968 }
3969 };
3970 let abs_paths = request
3971 .relative_paths
3972 .iter()
3973 .map(|path| {
3974 if path.file_name().is_some() {
3975 root_canonical_path.as_path().join(path.as_std_path())
3976 } else {
3977 root_canonical_path.as_path().to_path_buf()
3978 }
3979 })
3980 .collect::<Vec<_>>();
3981
3982 {
3983 let mut state = self.state.lock().await;
3984 let is_idle = state.snapshot.completed_scan_id == state.snapshot.scan_id;
3985 state.snapshot.scan_id += 1;
3986 if is_idle {
3987 state.snapshot.completed_scan_id = state.snapshot.scan_id;
3988 }
3989 }
3990
3991 self.reload_entries_for_paths(
3992 &root_path,
3993 &root_canonical_path,
3994 &request.relative_paths,
3995 abs_paths,
3996 None,
3997 )
3998 .await;
3999
4000 self.send_status_update(scanning, request.done).await
4001 }
4002
4003 async fn process_events(&self, mut abs_paths: Vec<PathBuf>) {
4004 log::trace!("process events: {abs_paths:?}");
4005 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4006 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4007 let root_canonical_path = match &root_canonical_path {
4008 Ok(path) => SanitizedPath::new(path),
4009 Err(err) => {
4010 let new_path = self
4011 .state
4012 .lock()
4013 .await
4014 .snapshot
4015 .root_file_handle
4016 .clone()
4017 .and_then(|handle| match handle.current_path(&self.fs) {
4018 Ok(new_path) => Some(new_path),
4019 Err(e) => {
4020 log::error!("Failed to refresh worktree root path: {e:#}");
4021 None
4022 }
4023 })
4024 .map(|path| SanitizedPath::new_arc(&path))
4025 .filter(|new_path| *new_path != root_path);
4026
4027 if let Some(new_path) = new_path {
4028 log::info!(
4029 "root renamed from {:?} to {:?}",
4030 root_path.as_path(),
4031 new_path.as_path(),
4032 );
4033 self.status_updates_tx
4034 .unbounded_send(ScanState::RootUpdated { new_path })
4035 .ok();
4036 } else {
4037 log::error!("root path could not be canonicalized: {err:#}");
4038 }
4039 return;
4040 }
4041 };
4042
4043 // Certain directories may have FS changes, but do not lead to git data changes that Zed cares about.
4044 // Ignore these, to avoid Zed unnecessarily rescanning git metadata.
4045 let skipped_files_in_dot_git = [COMMIT_MESSAGE, INDEX_LOCK];
4046 let skipped_dirs_in_dot_git = [FSMONITOR_DAEMON, LFS_DIR];
4047
4048 let mut relative_paths = Vec::with_capacity(abs_paths.len());
4049 let mut dot_git_abs_paths = Vec::new();
4050 let mut work_dirs_needing_exclude_update = Vec::new();
4051 abs_paths.sort_unstable();
4052 abs_paths.dedup_by(|a, b| a.starts_with(b));
4053 {
4054 let snapshot = &self.state.lock().await.snapshot;
4055
4056 let mut ranges_to_drop = SmallVec::<[Range<usize>; 4]>::new();
4057
4058 fn skip_ix(ranges: &mut SmallVec<[Range<usize>; 4]>, ix: usize) {
4059 if let Some(last_range) = ranges.last_mut()
4060 && last_range.end == ix
4061 {
4062 last_range.end += 1;
4063 } else {
4064 ranges.push(ix..ix + 1);
4065 }
4066 }
4067
4068 for (ix, abs_path) in abs_paths.iter().enumerate() {
4069 let abs_path = &SanitizedPath::new(&abs_path);
4070
4071 let mut is_git_related = false;
4072 let mut dot_git_paths = None;
4073
4074 for ancestor in abs_path.as_path().ancestors() {
4075 if is_git_dir(ancestor, self.fs.as_ref()).await {
4076 let path_in_git_dir = abs_path
4077 .as_path()
4078 .strip_prefix(ancestor)
4079 .expect("stripping off the ancestor");
4080 dot_git_paths = Some((ancestor.to_owned(), path_in_git_dir.to_owned()));
4081 break;
4082 }
4083 }
4084
4085 if let Some((dot_git_abs_path, path_in_git_dir)) = dot_git_paths {
4086 if skipped_files_in_dot_git
4087 .iter()
4088 .any(|skipped| OsStr::new(skipped) == path_in_git_dir.as_path().as_os_str())
4089 || skipped_dirs_in_dot_git.iter().any(|skipped_git_subdir| {
4090 path_in_git_dir.starts_with(skipped_git_subdir)
4091 })
4092 {
4093 log::debug!(
4094 "ignoring event {abs_path:?} as it's in the .git directory among skipped files or directories"
4095 );
4096 skip_ix(&mut ranges_to_drop, ix);
4097 continue;
4098 }
4099
4100 is_git_related = true;
4101 if !dot_git_abs_paths.contains(&dot_git_abs_path) {
4102 dot_git_abs_paths.push(dot_git_abs_path);
4103 }
4104 }
4105
4106 let relative_path = if let Ok(path) = abs_path.strip_prefix(&root_canonical_path)
4107 && let Ok(path) = RelPath::new(path, PathStyle::local())
4108 {
4109 path
4110 } else {
4111 if is_git_related {
4112 log::debug!(
4113 "ignoring event {abs_path:?}, since it's in git dir outside of root path {root_canonical_path:?}",
4114 );
4115 } else {
4116 log::error!(
4117 "ignoring event {abs_path:?} outside of root path {root_canonical_path:?}",
4118 );
4119 }
4120 skip_ix(&mut ranges_to_drop, ix);
4121 continue;
4122 };
4123
4124 let absolute_path = abs_path.to_path_buf();
4125 if absolute_path.ends_with(Path::new(DOT_GIT).join(REPO_EXCLUDE)) {
4126 if let Some(repository) = snapshot
4127 .git_repositories
4128 .values()
4129 .find(|repo| repo.common_dir_abs_path.join(REPO_EXCLUDE) == absolute_path)
4130 {
4131 work_dirs_needing_exclude_update
4132 .push(repository.work_directory_abs_path.clone());
4133 }
4134 }
4135
4136 if abs_path.file_name() == Some(OsStr::new(GITIGNORE)) {
4137 for (_, repo) in snapshot
4138 .git_repositories
4139 .iter()
4140 .filter(|(_, repo)| repo.directory_contains(&relative_path))
4141 {
4142 if !dot_git_abs_paths.iter().any(|dot_git_abs_path| {
4143 dot_git_abs_path == repo.common_dir_abs_path.as_ref()
4144 }) {
4145 dot_git_abs_paths.push(repo.common_dir_abs_path.to_path_buf());
4146 }
4147 }
4148 }
4149
4150 let parent_dir_is_loaded = relative_path.parent().is_none_or(|parent| {
4151 snapshot
4152 .entry_for_path(parent)
4153 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4154 });
4155 if !parent_dir_is_loaded {
4156 log::debug!("ignoring event {relative_path:?} within unloaded directory");
4157 skip_ix(&mut ranges_to_drop, ix);
4158 continue;
4159 }
4160
4161 if self.settings.is_path_excluded(&relative_path) {
4162 if !is_git_related {
4163 log::debug!("ignoring FS event for excluded path {relative_path:?}");
4164 }
4165 skip_ix(&mut ranges_to_drop, ix);
4166 continue;
4167 }
4168
4169 relative_paths.push(relative_path.into_arc());
4170 }
4171
4172 for range_to_drop in ranges_to_drop.into_iter().rev() {
4173 abs_paths.drain(range_to_drop);
4174 }
4175 }
4176
4177 if relative_paths.is_empty() && dot_git_abs_paths.is_empty() {
4178 return;
4179 }
4180
4181 if !work_dirs_needing_exclude_update.is_empty() {
4182 let mut state = self.state.lock().await;
4183 for work_dir_abs_path in work_dirs_needing_exclude_update {
4184 if let Some((_, needs_update)) = state
4185 .snapshot
4186 .repo_exclude_by_work_dir_abs_path
4187 .get_mut(&work_dir_abs_path)
4188 {
4189 *needs_update = true;
4190 }
4191 }
4192 }
4193
4194 self.state.lock().await.snapshot.scan_id += 1;
4195
4196 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4197 log::debug!("received fs events {:?}", relative_paths);
4198 self.reload_entries_for_paths(
4199 &root_path,
4200 &root_canonical_path,
4201 &relative_paths,
4202 abs_paths,
4203 Some(scan_job_tx.clone()),
4204 )
4205 .await;
4206
4207 let affected_repo_roots = if !dot_git_abs_paths.is_empty() {
4208 self.update_git_repositories(dot_git_abs_paths).await
4209 } else {
4210 Vec::new()
4211 };
4212
4213 {
4214 let mut ignores_to_update = self.ignores_needing_update().await;
4215 ignores_to_update.extend(affected_repo_roots);
4216 let ignores_to_update = self.order_ignores(ignores_to_update).await;
4217 let snapshot = self.state.lock().await.snapshot.clone();
4218 self.update_ignore_statuses_for_paths(scan_job_tx, snapshot, ignores_to_update)
4219 .await;
4220 self.scan_dirs(false, scan_job_rx).await;
4221 }
4222
4223 {
4224 let mut state = self.state.lock().await;
4225 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4226 for (_, entry) in mem::take(&mut state.removed_entries) {
4227 state.scanned_dirs.remove(&entry.id);
4228 }
4229 }
4230 self.send_status_update(false, SmallVec::new()).await;
4231 }
4232
4233 async fn update_global_gitignore(&self, abs_path: &Path) {
4234 let ignore = build_gitignore(abs_path, self.fs.as_ref())
4235 .await
4236 .log_err()
4237 .map(Arc::new);
4238 let (prev_snapshot, ignore_stack, abs_path) = {
4239 let mut state = self.state.lock().await;
4240 state.snapshot.global_gitignore = ignore;
4241 let abs_path = state.snapshot.abs_path().clone();
4242 let ignore_stack = state
4243 .snapshot
4244 .ignore_stack_for_abs_path(&abs_path, true, self.fs.as_ref())
4245 .await;
4246 (state.snapshot.clone(), ignore_stack, abs_path)
4247 };
4248 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4249 self.update_ignore_statuses_for_paths(
4250 scan_job_tx,
4251 prev_snapshot,
4252 vec![(abs_path, ignore_stack)],
4253 )
4254 .await;
4255 self.scan_dirs(false, scan_job_rx).await;
4256 self.send_status_update(false, SmallVec::new()).await;
4257 }
4258
4259 async fn forcibly_load_paths(&self, paths: &[Arc<RelPath>]) -> bool {
4260 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4261 {
4262 let mut state = self.state.lock().await;
4263 let root_path = state.snapshot.abs_path.clone();
4264 for path in paths {
4265 for ancestor in path.ancestors() {
4266 if let Some(entry) = state.snapshot.entry_for_path(ancestor)
4267 && entry.kind == EntryKind::UnloadedDir
4268 {
4269 let abs_path = root_path.join(ancestor.as_std_path());
4270 state
4271 .enqueue_scan_dir(
4272 abs_path.into(),
4273 entry,
4274 &scan_job_tx,
4275 self.fs.as_ref(),
4276 )
4277 .await;
4278 state.paths_to_scan.insert(path.clone());
4279 break;
4280 }
4281 }
4282 }
4283 drop(scan_job_tx);
4284 }
4285 while let Ok(job) = scan_job_rx.recv().await {
4286 self.scan_dir(&job).await.log_err();
4287 }
4288
4289 !mem::take(&mut self.state.lock().await.paths_to_scan).is_empty()
4290 }
4291
4292 async fn scan_dirs(
4293 &self,
4294 enable_progress_updates: bool,
4295 scan_jobs_rx: channel::Receiver<ScanJob>,
4296 ) {
4297 if self
4298 .status_updates_tx
4299 .unbounded_send(ScanState::Started)
4300 .is_err()
4301 {
4302 return;
4303 }
4304
4305 let progress_update_count = AtomicUsize::new(0);
4306 self.executor
4307 .scoped_priority(Priority::Low, |scope| {
4308 for _ in 0..self.executor.num_cpus() {
4309 scope.spawn(async {
4310 let mut last_progress_update_count = 0;
4311 let progress_update_timer = self.progress_timer(enable_progress_updates).fuse();
4312 futures::pin_mut!(progress_update_timer);
4313
4314 loop {
4315 select_biased! {
4316 // Process any path refresh requests before moving on to process
4317 // the scan queue, so that user operations are prioritized.
4318 request = self.next_scan_request().fuse() => {
4319 let Ok(request) = request else { break };
4320 if !self.process_scan_request(request, true).await {
4321 return;
4322 }
4323 }
4324
4325 // Send periodic progress updates to the worktree. Use an atomic counter
4326 // to ensure that only one of the workers sends a progress update after
4327 // the update interval elapses.
4328 _ = progress_update_timer => {
4329 match progress_update_count.compare_exchange(
4330 last_progress_update_count,
4331 last_progress_update_count + 1,
4332 SeqCst,
4333 SeqCst
4334 ) {
4335 Ok(_) => {
4336 last_progress_update_count += 1;
4337 self.send_status_update(true, SmallVec::new()).await;
4338 }
4339 Err(count) => {
4340 last_progress_update_count = count;
4341 }
4342 }
4343 progress_update_timer.set(self.progress_timer(enable_progress_updates).fuse());
4344 }
4345
4346 // Recursively load directories from the file system.
4347 job = scan_jobs_rx.recv().fuse() => {
4348 let Ok(job) = job else { break };
4349 if let Err(err) = self.scan_dir(&job).await
4350 && job.path.is_empty() {
4351 log::error!("error scanning directory {:?}: {}", job.abs_path, err);
4352 }
4353 }
4354 }
4355 }
4356 });
4357 }
4358 })
4359 .await;
4360 }
4361
4362 async fn send_status_update(
4363 &self,
4364 scanning: bool,
4365 barrier: SmallVec<[barrier::Sender; 1]>,
4366 ) -> bool {
4367 let mut state = self.state.lock().await;
4368 if state.changed_paths.is_empty() && scanning {
4369 return true;
4370 }
4371
4372 let new_snapshot = state.snapshot.clone();
4373 let old_snapshot = mem::replace(&mut state.prev_snapshot, new_snapshot.snapshot.clone());
4374 let changes = build_diff(
4375 self.phase,
4376 &old_snapshot,
4377 &new_snapshot,
4378 &state.changed_paths,
4379 );
4380 state.changed_paths.clear();
4381
4382 self.status_updates_tx
4383 .unbounded_send(ScanState::Updated {
4384 snapshot: new_snapshot,
4385 changes,
4386 scanning,
4387 barrier,
4388 })
4389 .is_ok()
4390 }
4391
4392 async fn scan_dir(&self, job: &ScanJob) -> Result<()> {
4393 let root_abs_path;
4394 let root_char_bag;
4395 {
4396 let snapshot = &self.state.lock().await.snapshot;
4397 if self.settings.is_path_excluded(&job.path) {
4398 log::error!("skipping excluded directory {:?}", job.path);
4399 return Ok(());
4400 }
4401 log::trace!("scanning directory {:?}", job.path);
4402 root_abs_path = snapshot.abs_path().clone();
4403 root_char_bag = snapshot.root_char_bag;
4404 }
4405
4406 let next_entry_id = self.next_entry_id.clone();
4407 let mut ignore_stack = job.ignore_stack.clone();
4408 let mut new_ignore = None;
4409 let mut root_canonical_path = None;
4410 let mut new_entries: Vec<Entry> = Vec::new();
4411 let mut new_jobs: Vec<Option<ScanJob>> = Vec::new();
4412 let mut child_paths = self
4413 .fs
4414 .read_dir(&job.abs_path)
4415 .await?
4416 .filter_map(|entry| async {
4417 match entry {
4418 Ok(entry) => Some(entry),
4419 Err(error) => {
4420 log::error!("error processing entry {:?}", error);
4421 None
4422 }
4423 }
4424 })
4425 .collect::<Vec<_>>()
4426 .await;
4427
4428 // Ensure that .git and .gitignore are processed first.
4429 swap_to_front(&mut child_paths, GITIGNORE);
4430 swap_to_front(&mut child_paths, DOT_GIT);
4431
4432 if let Some(path) = child_paths.first()
4433 && path.ends_with(DOT_GIT)
4434 {
4435 ignore_stack.repo_root = Some(job.abs_path.clone());
4436 }
4437
4438 for child_abs_path in child_paths {
4439 let child_abs_path: Arc<Path> = child_abs_path.into();
4440 let child_name = child_abs_path.file_name().unwrap();
4441 let Some(child_path) = child_name
4442 .to_str()
4443 .and_then(|name| Some(job.path.join(RelPath::unix(name).ok()?)))
4444 else {
4445 continue;
4446 };
4447
4448 if child_name == DOT_GIT {
4449 let mut state = self.state.lock().await;
4450 state
4451 .insert_git_repository(
4452 child_path.clone(),
4453 self.fs.as_ref(),
4454 self.watcher.as_ref(),
4455 )
4456 .await;
4457 } else if child_name == GITIGNORE {
4458 match build_gitignore(&child_abs_path, self.fs.as_ref()).await {
4459 Ok(ignore) => {
4460 let ignore = Arc::new(ignore);
4461 ignore_stack = ignore_stack
4462 .append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4463 new_ignore = Some(ignore);
4464 }
4465 Err(error) => {
4466 log::error!(
4467 "error loading .gitignore file {:?} - {:?}",
4468 child_name,
4469 error
4470 );
4471 }
4472 }
4473 }
4474
4475 if self.settings.is_path_excluded(&child_path) {
4476 log::debug!("skipping excluded child entry {child_path:?}");
4477 self.state.lock().await.remove_path(&child_path);
4478 continue;
4479 }
4480
4481 let child_metadata = match self.fs.metadata(&child_abs_path).await {
4482 Ok(Some(metadata)) => metadata,
4483 Ok(None) => continue,
4484 Err(err) => {
4485 log::error!("error processing {child_abs_path:?}: {err:#}");
4486 continue;
4487 }
4488 };
4489
4490 let mut child_entry = Entry::new(
4491 child_path.clone(),
4492 &child_metadata,
4493 ProjectEntryId::new(&next_entry_id),
4494 root_char_bag,
4495 None,
4496 );
4497
4498 if job.is_external {
4499 child_entry.is_external = true;
4500 } else if child_metadata.is_symlink {
4501 let canonical_path = match self.fs.canonicalize(&child_abs_path).await {
4502 Ok(path) => path,
4503 Err(err) => {
4504 log::error!("error reading target of symlink {child_abs_path:?}: {err:#}",);
4505 continue;
4506 }
4507 };
4508
4509 // lazily canonicalize the root path in order to determine if
4510 // symlinks point outside of the worktree.
4511 let root_canonical_path = match &root_canonical_path {
4512 Some(path) => path,
4513 None => match self.fs.canonicalize(&root_abs_path).await {
4514 Ok(path) => root_canonical_path.insert(path),
4515 Err(err) => {
4516 log::error!("error canonicalizing root {:?}: {:?}", root_abs_path, err);
4517 continue;
4518 }
4519 },
4520 };
4521
4522 if !canonical_path.starts_with(root_canonical_path) {
4523 child_entry.is_external = true;
4524 }
4525
4526 child_entry.canonical_path = Some(canonical_path.into());
4527 }
4528
4529 if child_entry.is_dir() {
4530 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, true);
4531 child_entry.is_always_included =
4532 self.settings.is_path_always_included(&child_path, true);
4533
4534 // Avoid recursing until crash in the case of a recursive symlink
4535 if job.ancestor_inodes.contains(&child_entry.inode) {
4536 new_jobs.push(None);
4537 } else {
4538 let mut ancestor_inodes = job.ancestor_inodes.clone();
4539 ancestor_inodes.insert(child_entry.inode);
4540
4541 new_jobs.push(Some(ScanJob {
4542 abs_path: child_abs_path.clone(),
4543 path: child_path,
4544 is_external: child_entry.is_external,
4545 ignore_stack: if child_entry.is_ignored {
4546 IgnoreStack::all()
4547 } else {
4548 ignore_stack.clone()
4549 },
4550 ancestor_inodes,
4551 scan_queue: job.scan_queue.clone(),
4552 }));
4553 }
4554 } else {
4555 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, false);
4556 child_entry.is_always_included =
4557 self.settings.is_path_always_included(&child_path, false);
4558 }
4559
4560 {
4561 let relative_path = job
4562 .path
4563 .join(RelPath::unix(child_name.to_str().unwrap()).unwrap());
4564 if self.is_path_private(&relative_path) {
4565 log::debug!("detected private file: {relative_path:?}");
4566 child_entry.is_private = true;
4567 }
4568 if self.settings.is_path_hidden(&relative_path) {
4569 log::debug!("detected hidden file: {relative_path:?}");
4570 child_entry.is_hidden = true;
4571 }
4572 }
4573
4574 new_entries.push(child_entry);
4575 }
4576
4577 let mut state = self.state.lock().await;
4578
4579 // Identify any subdirectories that should not be scanned.
4580 let mut job_ix = 0;
4581 for entry in &mut new_entries {
4582 state.reuse_entry_id(entry);
4583 if entry.is_dir() {
4584 if state.should_scan_directory(entry) {
4585 job_ix += 1;
4586 } else {
4587 log::debug!("defer scanning directory {:?}", entry.path);
4588 entry.kind = EntryKind::UnloadedDir;
4589 new_jobs.remove(job_ix);
4590 }
4591 }
4592 if entry.is_always_included {
4593 state
4594 .snapshot
4595 .always_included_entries
4596 .push(entry.path.clone());
4597 }
4598 }
4599
4600 state.populate_dir(job.path.clone(), new_entries, new_ignore);
4601 self.watcher.add(job.abs_path.as_ref()).log_err();
4602
4603 for new_job in new_jobs.into_iter().flatten() {
4604 job.scan_queue
4605 .try_send(new_job)
4606 .expect("channel is unbounded");
4607 }
4608
4609 Ok(())
4610 }
4611
4612 /// All list arguments should be sorted before calling this function
4613 async fn reload_entries_for_paths(
4614 &self,
4615 root_abs_path: &SanitizedPath,
4616 root_canonical_path: &SanitizedPath,
4617 relative_paths: &[Arc<RelPath>],
4618 abs_paths: Vec<PathBuf>,
4619 scan_queue_tx: Option<Sender<ScanJob>>,
4620 ) {
4621 // grab metadata for all requested paths
4622 let metadata = futures::future::join_all(
4623 abs_paths
4624 .iter()
4625 .map(|abs_path| async move {
4626 let metadata = self.fs.metadata(abs_path).await?;
4627 if let Some(metadata) = metadata {
4628 let canonical_path = self.fs.canonicalize(abs_path).await?;
4629
4630 // If we're on a case-insensitive filesystem (default on macOS), we want
4631 // to only ignore metadata for non-symlink files if their absolute-path matches
4632 // the canonical-path.
4633 // Because if not, this might be a case-only-renaming (`mv test.txt TEST.TXT`)
4634 // and we want to ignore the metadata for the old path (`test.txt`) so it's
4635 // treated as removed.
4636 if !self.fs_case_sensitive && !metadata.is_symlink {
4637 let canonical_file_name = canonical_path.file_name();
4638 let file_name = abs_path.file_name();
4639 if canonical_file_name != file_name {
4640 return Ok(None);
4641 }
4642 }
4643
4644 anyhow::Ok(Some((metadata, SanitizedPath::new_arc(&canonical_path))))
4645 } else {
4646 Ok(None)
4647 }
4648 })
4649 .collect::<Vec<_>>(),
4650 )
4651 .await;
4652
4653 let mut new_ancestor_repo = if relative_paths.iter().any(|path| path.is_empty()) {
4654 Some(discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await)
4655 } else {
4656 None
4657 };
4658
4659 let mut state = self.state.lock().await;
4660 let doing_recursive_update = scan_queue_tx.is_some();
4661
4662 // Remove any entries for paths that no longer exist or are being recursively
4663 // refreshed. Do this before adding any new entries, so that renames can be
4664 // detected regardless of the order of the paths.
4665 for (path, metadata) in relative_paths.iter().zip(metadata.iter()) {
4666 if matches!(metadata, Ok(None)) || doing_recursive_update {
4667 state.remove_path(path);
4668 }
4669 }
4670
4671 for (path, metadata) in relative_paths.iter().zip(metadata.into_iter()) {
4672 let abs_path: Arc<Path> = root_abs_path.join(path.as_std_path()).into();
4673 match metadata {
4674 Ok(Some((metadata, canonical_path))) => {
4675 let ignore_stack = state
4676 .snapshot
4677 .ignore_stack_for_abs_path(&abs_path, metadata.is_dir, self.fs.as_ref())
4678 .await;
4679 let is_external = !canonical_path.starts_with(&root_canonical_path);
4680 let entry_id = state.entry_id_for(self.next_entry_id.as_ref(), path, &metadata);
4681 let mut fs_entry = Entry::new(
4682 path.clone(),
4683 &metadata,
4684 entry_id,
4685 state.snapshot.root_char_bag,
4686 if metadata.is_symlink {
4687 Some(canonical_path.as_path().to_path_buf().into())
4688 } else {
4689 None
4690 },
4691 );
4692
4693 let is_dir = fs_entry.is_dir();
4694 fs_entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, is_dir);
4695 fs_entry.is_external = is_external;
4696 fs_entry.is_private = self.is_path_private(path);
4697 fs_entry.is_always_included =
4698 self.settings.is_path_always_included(path, is_dir);
4699 fs_entry.is_hidden = self.settings.is_path_hidden(path);
4700
4701 if let (Some(scan_queue_tx), true) = (&scan_queue_tx, is_dir) {
4702 if state.should_scan_directory(&fs_entry)
4703 || (fs_entry.path.is_empty()
4704 && abs_path.file_name() == Some(OsStr::new(DOT_GIT)))
4705 {
4706 state
4707 .enqueue_scan_dir(
4708 abs_path,
4709 &fs_entry,
4710 scan_queue_tx,
4711 self.fs.as_ref(),
4712 )
4713 .await;
4714 } else {
4715 fs_entry.kind = EntryKind::UnloadedDir;
4716 }
4717 }
4718
4719 state
4720 .insert_entry(fs_entry.clone(), self.fs.as_ref(), self.watcher.as_ref())
4721 .await;
4722
4723 if path.is_empty()
4724 && let Some((ignores, exclude, repo)) = new_ancestor_repo.take()
4725 {
4726 log::trace!("updating ancestor git repository");
4727 state.snapshot.ignores_by_parent_abs_path.extend(ignores);
4728 if let Some((ancestor_dot_git, work_directory)) = repo {
4729 if let Some(exclude) = exclude {
4730 let work_directory_abs_path = self
4731 .state
4732 .lock()
4733 .await
4734 .snapshot
4735 .work_directory_abs_path(&work_directory);
4736
4737 state
4738 .snapshot
4739 .repo_exclude_by_work_dir_abs_path
4740 .insert(work_directory_abs_path.into(), (exclude, false));
4741 }
4742 state
4743 .insert_git_repository_for_path(
4744 work_directory,
4745 ancestor_dot_git.into(),
4746 self.fs.as_ref(),
4747 self.watcher.as_ref(),
4748 )
4749 .await
4750 .log_err();
4751 }
4752 }
4753 }
4754 Ok(None) => {
4755 self.remove_repo_path(path.clone(), &mut state.snapshot);
4756 }
4757 Err(err) => {
4758 log::error!("error reading file {abs_path:?} on event: {err:#}");
4759 }
4760 }
4761 }
4762
4763 util::extend_sorted(
4764 &mut state.changed_paths,
4765 relative_paths.iter().cloned(),
4766 usize::MAX,
4767 Ord::cmp,
4768 );
4769 }
4770
4771 fn remove_repo_path(&self, path: Arc<RelPath>, snapshot: &mut LocalSnapshot) -> Option<()> {
4772 if !path.components().any(|component| component == DOT_GIT)
4773 && let Some(local_repo) = snapshot.local_repo_for_work_directory_path(&path)
4774 {
4775 let id = local_repo.work_directory_id;
4776 log::debug!("remove repo path: {:?}", path);
4777 snapshot.git_repositories.remove(&id);
4778 return Some(());
4779 }
4780
4781 Some(())
4782 }
4783
4784 async fn update_ignore_statuses_for_paths(
4785 &self,
4786 scan_job_tx: Sender<ScanJob>,
4787 prev_snapshot: LocalSnapshot,
4788 ignores_to_update: Vec<(Arc<Path>, IgnoreStack)>,
4789 ) {
4790 let (ignore_queue_tx, ignore_queue_rx) = channel::unbounded();
4791 {
4792 for (parent_abs_path, ignore_stack) in ignores_to_update {
4793 ignore_queue_tx
4794 .send_blocking(UpdateIgnoreStatusJob {
4795 abs_path: parent_abs_path,
4796 ignore_stack,
4797 ignore_queue: ignore_queue_tx.clone(),
4798 scan_queue: scan_job_tx.clone(),
4799 })
4800 .unwrap();
4801 }
4802 }
4803 drop(ignore_queue_tx);
4804
4805 self.executor
4806 .scoped(|scope| {
4807 for _ in 0..self.executor.num_cpus() {
4808 scope.spawn(async {
4809 loop {
4810 select_biased! {
4811 // Process any path refresh requests before moving on to process
4812 // the queue of ignore statuses.
4813 request = self.next_scan_request().fuse() => {
4814 let Ok(request) = request else { break };
4815 if !self.process_scan_request(request, true).await {
4816 return;
4817 }
4818 }
4819
4820 // Recursively process directories whose ignores have changed.
4821 job = ignore_queue_rx.recv().fuse() => {
4822 let Ok(job) = job else { break };
4823 self.update_ignore_status(job, &prev_snapshot).await;
4824 }
4825 }
4826 }
4827 });
4828 }
4829 })
4830 .await;
4831 }
4832
4833 async fn ignores_needing_update(&self) -> Vec<Arc<Path>> {
4834 let mut ignores_to_update = Vec::new();
4835 let mut excludes_to_load: Vec<(Arc<Path>, PathBuf)> = Vec::new();
4836
4837 // First pass: collect updates and drop stale entries without awaiting.
4838 {
4839 let snapshot = &mut self.state.lock().await.snapshot;
4840 let abs_path = snapshot.abs_path.clone();
4841 let mut repo_exclude_keys_to_remove: Vec<Arc<Path>> = Vec::new();
4842
4843 for (work_dir_abs_path, (_, needs_update)) in
4844 snapshot.repo_exclude_by_work_dir_abs_path.iter_mut()
4845 {
4846 let repository = snapshot
4847 .git_repositories
4848 .iter()
4849 .find(|(_, repo)| &repo.work_directory_abs_path == work_dir_abs_path);
4850
4851 if *needs_update {
4852 *needs_update = false;
4853 ignores_to_update.push(work_dir_abs_path.clone());
4854
4855 if let Some((_, repository)) = repository {
4856 let exclude_abs_path = repository.common_dir_abs_path.join(REPO_EXCLUDE);
4857 excludes_to_load.push((work_dir_abs_path.clone(), exclude_abs_path));
4858 }
4859 }
4860
4861 if repository.is_none() {
4862 repo_exclude_keys_to_remove.push(work_dir_abs_path.clone());
4863 }
4864 }
4865
4866 for key in repo_exclude_keys_to_remove {
4867 snapshot.repo_exclude_by_work_dir_abs_path.remove(&key);
4868 }
4869
4870 snapshot
4871 .ignores_by_parent_abs_path
4872 .retain(|parent_abs_path, (_, needs_update)| {
4873 if let Ok(parent_path) = parent_abs_path.strip_prefix(abs_path.as_path())
4874 && let Some(parent_path) =
4875 RelPath::new(&parent_path, PathStyle::local()).log_err()
4876 {
4877 if *needs_update {
4878 *needs_update = false;
4879 if snapshot.snapshot.entry_for_path(&parent_path).is_some() {
4880 ignores_to_update.push(parent_abs_path.clone());
4881 }
4882 }
4883
4884 let ignore_path = parent_path.join(RelPath::unix(GITIGNORE).unwrap());
4885 if snapshot.snapshot.entry_for_path(&ignore_path).is_none() {
4886 return false;
4887 }
4888 }
4889 true
4890 });
4891 }
4892
4893 // Load gitignores asynchronously (outside the lock)
4894 let mut loaded_excludes: Vec<(Arc<Path>, Arc<Gitignore>)> = Vec::new();
4895 for (work_dir_abs_path, exclude_abs_path) in excludes_to_load {
4896 if let Ok(current_exclude) = build_gitignore(&exclude_abs_path, self.fs.as_ref()).await
4897 {
4898 loaded_excludes.push((work_dir_abs_path, Arc::new(current_exclude)));
4899 }
4900 }
4901
4902 // Second pass: apply updates.
4903 if !loaded_excludes.is_empty() {
4904 let snapshot = &mut self.state.lock().await.snapshot;
4905
4906 for (work_dir_abs_path, exclude) in loaded_excludes {
4907 if let Some((existing_exclude, _)) = snapshot
4908 .repo_exclude_by_work_dir_abs_path
4909 .get_mut(&work_dir_abs_path)
4910 {
4911 *existing_exclude = exclude;
4912 }
4913 }
4914 }
4915
4916 ignores_to_update
4917 }
4918
4919 async fn order_ignores(&self, mut ignores: Vec<Arc<Path>>) -> Vec<(Arc<Path>, IgnoreStack)> {
4920 let fs = self.fs.clone();
4921 let snapshot = self.state.lock().await.snapshot.clone();
4922 ignores.sort_unstable();
4923 let mut ignores_to_update = ignores.into_iter().peekable();
4924
4925 let mut result = vec![];
4926 while let Some(parent_abs_path) = ignores_to_update.next() {
4927 while ignores_to_update
4928 .peek()
4929 .map_or(false, |p| p.starts_with(&parent_abs_path))
4930 {
4931 ignores_to_update.next().unwrap();
4932 }
4933 let ignore_stack = snapshot
4934 .ignore_stack_for_abs_path(&parent_abs_path, true, fs.as_ref())
4935 .await;
4936 result.push((parent_abs_path, ignore_stack));
4937 }
4938
4939 result
4940 }
4941
4942 async fn update_ignore_status(&self, job: UpdateIgnoreStatusJob, snapshot: &LocalSnapshot) {
4943 log::trace!("update ignore status {:?}", job.abs_path);
4944
4945 let mut ignore_stack = job.ignore_stack;
4946 if let Some((ignore, _)) = snapshot.ignores_by_parent_abs_path.get(&job.abs_path) {
4947 ignore_stack =
4948 ignore_stack.append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4949 }
4950
4951 let mut entries_by_id_edits = Vec::new();
4952 let mut entries_by_path_edits = Vec::new();
4953 let Some(path) = job
4954 .abs_path
4955 .strip_prefix(snapshot.abs_path.as_path())
4956 .map_err(|_| {
4957 anyhow::anyhow!(
4958 "Failed to strip prefix '{}' from path '{}'",
4959 snapshot.abs_path.as_path().display(),
4960 job.abs_path.display()
4961 )
4962 })
4963 .log_err()
4964 else {
4965 return;
4966 };
4967
4968 let Some(path) = RelPath::new(&path, PathStyle::local()).log_err() else {
4969 return;
4970 };
4971
4972 if let Ok(Some(metadata)) = self.fs.metadata(&job.abs_path.join(DOT_GIT)).await
4973 && metadata.is_dir
4974 {
4975 ignore_stack.repo_root = Some(job.abs_path.clone());
4976 }
4977
4978 for mut entry in snapshot.child_entries(&path).cloned() {
4979 let was_ignored = entry.is_ignored;
4980 let abs_path: Arc<Path> = snapshot.absolutize(&entry.path).into();
4981 entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, entry.is_dir());
4982
4983 if entry.is_dir() {
4984 let child_ignore_stack = if entry.is_ignored {
4985 IgnoreStack::all()
4986 } else {
4987 ignore_stack.clone()
4988 };
4989
4990 // Scan any directories that were previously ignored and weren't previously scanned.
4991 if was_ignored && !entry.is_ignored && entry.kind.is_unloaded() {
4992 let state = self.state.lock().await;
4993 if state.should_scan_directory(&entry) {
4994 state
4995 .enqueue_scan_dir(
4996 abs_path.clone(),
4997 &entry,
4998 &job.scan_queue,
4999 self.fs.as_ref(),
5000 )
5001 .await;
5002 }
5003 }
5004
5005 job.ignore_queue
5006 .send(UpdateIgnoreStatusJob {
5007 abs_path: abs_path.clone(),
5008 ignore_stack: child_ignore_stack,
5009 ignore_queue: job.ignore_queue.clone(),
5010 scan_queue: job.scan_queue.clone(),
5011 })
5012 .await
5013 .unwrap();
5014 }
5015
5016 if entry.is_ignored != was_ignored {
5017 let mut path_entry = snapshot.entries_by_id.get(&entry.id, ()).unwrap().clone();
5018 path_entry.scan_id = snapshot.scan_id;
5019 path_entry.is_ignored = entry.is_ignored;
5020 entries_by_id_edits.push(Edit::Insert(path_entry));
5021 entries_by_path_edits.push(Edit::Insert(entry));
5022 }
5023 }
5024
5025 let state = &mut self.state.lock().await;
5026 for edit in &entries_by_path_edits {
5027 if let Edit::Insert(entry) = edit
5028 && let Err(ix) = state.changed_paths.binary_search(&entry.path)
5029 {
5030 state.changed_paths.insert(ix, entry.path.clone());
5031 }
5032 }
5033
5034 state
5035 .snapshot
5036 .entries_by_path
5037 .edit(entries_by_path_edits, ());
5038 state.snapshot.entries_by_id.edit(entries_by_id_edits, ());
5039 }
5040
5041 async fn update_git_repositories(&self, dot_git_paths: Vec<PathBuf>) -> Vec<Arc<Path>> {
5042 log::trace!("reloading repositories: {dot_git_paths:?}");
5043 let mut state = self.state.lock().await;
5044 let scan_id = state.snapshot.scan_id;
5045 let mut affected_repo_roots = Vec::new();
5046 for dot_git_dir in dot_git_paths {
5047 let existing_repository_entry =
5048 state
5049 .snapshot
5050 .git_repositories
5051 .iter()
5052 .find_map(|(_, repo)| {
5053 let dot_git_dir = SanitizedPath::new(&dot_git_dir);
5054 if SanitizedPath::new(repo.common_dir_abs_path.as_ref()) == dot_git_dir
5055 || SanitizedPath::new(repo.repository_dir_abs_path.as_ref())
5056 == dot_git_dir
5057 {
5058 Some(repo.clone())
5059 } else {
5060 None
5061 }
5062 });
5063
5064 match existing_repository_entry {
5065 None => {
5066 let Ok(relative) = dot_git_dir.strip_prefix(state.snapshot.abs_path()) else {
5067 debug_panic!(
5068 "update_git_repositories called with .git directory outside the worktree root"
5069 );
5070 return Vec::new();
5071 };
5072 affected_repo_roots.push(dot_git_dir.parent().unwrap().into());
5073 state
5074 .insert_git_repository(
5075 RelPath::new(relative, PathStyle::local())
5076 .unwrap()
5077 .into_arc(),
5078 self.fs.as_ref(),
5079 self.watcher.as_ref(),
5080 )
5081 .await;
5082 }
5083 Some(local_repository) => {
5084 state.snapshot.git_repositories.update(
5085 &local_repository.work_directory_id,
5086 |entry| {
5087 entry.git_dir_scan_id = scan_id;
5088 },
5089 );
5090 }
5091 };
5092 }
5093
5094 // Remove any git repositories whose .git entry no longer exists.
5095 let snapshot = &mut state.snapshot;
5096 let mut ids_to_preserve = HashSet::default();
5097 for (&work_directory_id, entry) in snapshot.git_repositories.iter() {
5098 let exists_in_snapshot =
5099 snapshot
5100 .entry_for_id(work_directory_id)
5101 .is_some_and(|entry| {
5102 snapshot
5103 .entry_for_path(&entry.path.join(RelPath::unix(DOT_GIT).unwrap()))
5104 .is_some()
5105 });
5106
5107 if exists_in_snapshot
5108 || matches!(
5109 self.fs.metadata(&entry.common_dir_abs_path).await,
5110 Ok(Some(_))
5111 )
5112 {
5113 ids_to_preserve.insert(work_directory_id);
5114 }
5115 }
5116
5117 snapshot
5118 .git_repositories
5119 .retain(|work_directory_id, entry| {
5120 let preserve = ids_to_preserve.contains(work_directory_id);
5121 if !preserve {
5122 affected_repo_roots.push(entry.dot_git_abs_path.parent().unwrap().into());
5123 snapshot
5124 .repo_exclude_by_work_dir_abs_path
5125 .remove(&entry.work_directory_abs_path);
5126 }
5127 preserve
5128 });
5129
5130 affected_repo_roots
5131 }
5132
5133 async fn progress_timer(&self, running: bool) {
5134 if !running {
5135 return futures::future::pending().await;
5136 }
5137
5138 #[cfg(feature = "test-support")]
5139 if self.fs.is_fake() {
5140 return self.executor.simulate_random_delay().await;
5141 }
5142
5143 self.executor.timer(FS_WATCH_LATENCY).await
5144 }
5145
5146 fn is_path_private(&self, path: &RelPath) -> bool {
5147 !self.share_private_files && self.settings.is_path_private(path)
5148 }
5149
5150 async fn next_scan_request(&self) -> Result<ScanRequest> {
5151 let mut request = self.scan_requests_rx.recv().await?;
5152 while let Ok(next_request) = self.scan_requests_rx.try_recv() {
5153 request.relative_paths.extend(next_request.relative_paths);
5154 request.done.extend(next_request.done);
5155 }
5156 Ok(request)
5157 }
5158}
5159
5160async fn discover_ancestor_git_repo(
5161 fs: Arc<dyn Fs>,
5162 root_abs_path: &SanitizedPath,
5163) -> (
5164 HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
5165 Option<Arc<Gitignore>>,
5166 Option<(PathBuf, WorkDirectory)>,
5167) {
5168 let mut exclude = None;
5169 let mut ignores = HashMap::default();
5170 for (index, ancestor) in root_abs_path.as_path().ancestors().enumerate() {
5171 if index != 0 {
5172 if ancestor == paths::home_dir() {
5173 // Unless $HOME is itself the worktree root, don't consider it as a
5174 // containing git repository---expensive and likely unwanted.
5175 break;
5176 } else if let Ok(ignore) = build_gitignore(&ancestor.join(GITIGNORE), fs.as_ref()).await
5177 {
5178 ignores.insert(ancestor.into(), (ignore.into(), false));
5179 }
5180 }
5181
5182 let ancestor_dot_git = ancestor.join(DOT_GIT);
5183 log::trace!("considering ancestor: {ancestor_dot_git:?}");
5184 // Check whether the directory or file called `.git` exists (in the
5185 // case of worktrees it's a file.)
5186 if fs
5187 .metadata(&ancestor_dot_git)
5188 .await
5189 .is_ok_and(|metadata| metadata.is_some())
5190 {
5191 if index != 0 {
5192 // We canonicalize, since the FS events use the canonicalized path.
5193 if let Some(ancestor_dot_git) = fs.canonicalize(&ancestor_dot_git).await.log_err() {
5194 let location_in_repo = root_abs_path
5195 .as_path()
5196 .strip_prefix(ancestor)
5197 .unwrap()
5198 .into();
5199 log::info!("inserting parent git repo for this worktree: {location_in_repo:?}");
5200 // We associate the external git repo with our root folder and
5201 // also mark where in the git repo the root folder is located.
5202 return (
5203 ignores,
5204 exclude,
5205 Some((
5206 ancestor_dot_git,
5207 WorkDirectory::AboveProject {
5208 absolute_path: ancestor.into(),
5209 location_in_repo,
5210 },
5211 )),
5212 );
5213 };
5214 }
5215
5216 let repo_exclude_abs_path = ancestor_dot_git.join(REPO_EXCLUDE);
5217 if let Ok(repo_exclude) = build_gitignore(&repo_exclude_abs_path, fs.as_ref()).await {
5218 exclude = Some(Arc::new(repo_exclude));
5219 }
5220
5221 // Reached root of git repository.
5222 break;
5223 }
5224 }
5225
5226 (ignores, exclude, None)
5227}
5228
5229fn build_diff(
5230 phase: BackgroundScannerPhase,
5231 old_snapshot: &Snapshot,
5232 new_snapshot: &Snapshot,
5233 event_paths: &[Arc<RelPath>],
5234) -> UpdatedEntriesSet {
5235 use BackgroundScannerPhase::*;
5236 use PathChange::{Added, AddedOrUpdated, Loaded, Removed, Updated};
5237
5238 // Identify which paths have changed. Use the known set of changed
5239 // parent paths to optimize the search.
5240 let mut changes = Vec::new();
5241 let mut old_paths = old_snapshot.entries_by_path.cursor::<PathKey>(());
5242 let mut new_paths = new_snapshot.entries_by_path.cursor::<PathKey>(());
5243 let mut last_newly_loaded_dir_path = None;
5244 old_paths.next();
5245 new_paths.next();
5246 for path in event_paths {
5247 let path = PathKey(path.clone());
5248 if old_paths.item().is_some_and(|e| e.path < path.0) {
5249 old_paths.seek_forward(&path, Bias::Left);
5250 }
5251 if new_paths.item().is_some_and(|e| e.path < path.0) {
5252 new_paths.seek_forward(&path, Bias::Left);
5253 }
5254 loop {
5255 match (old_paths.item(), new_paths.item()) {
5256 (Some(old_entry), Some(new_entry)) => {
5257 if old_entry.path > path.0
5258 && new_entry.path > path.0
5259 && !old_entry.path.starts_with(&path.0)
5260 && !new_entry.path.starts_with(&path.0)
5261 {
5262 break;
5263 }
5264
5265 match Ord::cmp(&old_entry.path, &new_entry.path) {
5266 Ordering::Less => {
5267 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5268 old_paths.next();
5269 }
5270 Ordering::Equal => {
5271 if phase == EventsReceivedDuringInitialScan {
5272 if old_entry.id != new_entry.id {
5273 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5274 }
5275 // If the worktree was not fully initialized when this event was generated,
5276 // we can't know whether this entry was added during the scan or whether
5277 // it was merely updated.
5278 changes.push((
5279 new_entry.path.clone(),
5280 new_entry.id,
5281 AddedOrUpdated,
5282 ));
5283 } else if old_entry.id != new_entry.id {
5284 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5285 changes.push((new_entry.path.clone(), new_entry.id, Added));
5286 } else if old_entry != new_entry {
5287 if old_entry.kind.is_unloaded() {
5288 last_newly_loaded_dir_path = Some(&new_entry.path);
5289 changes.push((new_entry.path.clone(), new_entry.id, Loaded));
5290 } else {
5291 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5292 }
5293 }
5294 old_paths.next();
5295 new_paths.next();
5296 }
5297 Ordering::Greater => {
5298 let is_newly_loaded = phase == InitialScan
5299 || last_newly_loaded_dir_path
5300 .as_ref()
5301 .is_some_and(|dir| new_entry.path.starts_with(dir));
5302 changes.push((
5303 new_entry.path.clone(),
5304 new_entry.id,
5305 if is_newly_loaded { Loaded } else { Added },
5306 ));
5307 new_paths.next();
5308 }
5309 }
5310 }
5311 (Some(old_entry), None) => {
5312 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5313 old_paths.next();
5314 }
5315 (None, Some(new_entry)) => {
5316 let is_newly_loaded = phase == InitialScan
5317 || last_newly_loaded_dir_path
5318 .as_ref()
5319 .is_some_and(|dir| new_entry.path.starts_with(dir));
5320 changes.push((
5321 new_entry.path.clone(),
5322 new_entry.id,
5323 if is_newly_loaded { Loaded } else { Added },
5324 ));
5325 new_paths.next();
5326 }
5327 (None, None) => break,
5328 }
5329 }
5330 }
5331
5332 changes.into()
5333}
5334
5335fn swap_to_front(child_paths: &mut Vec<PathBuf>, file: &str) {
5336 let position = child_paths
5337 .iter()
5338 .position(|path| path.file_name().unwrap() == file);
5339 if let Some(position) = position {
5340 let temp = child_paths.remove(position);
5341 child_paths.insert(0, temp);
5342 }
5343}
5344
5345fn char_bag_for_path(root_char_bag: CharBag, path: &RelPath) -> CharBag {
5346 let mut result = root_char_bag;
5347 result.extend(path.as_unix_str().chars().map(|c| c.to_ascii_lowercase()));
5348 result
5349}
5350
5351#[derive(Debug)]
5352struct ScanJob {
5353 abs_path: Arc<Path>,
5354 path: Arc<RelPath>,
5355 ignore_stack: IgnoreStack,
5356 scan_queue: Sender<ScanJob>,
5357 ancestor_inodes: TreeSet<u64>,
5358 is_external: bool,
5359}
5360
5361struct UpdateIgnoreStatusJob {
5362 abs_path: Arc<Path>,
5363 ignore_stack: IgnoreStack,
5364 ignore_queue: Sender<UpdateIgnoreStatusJob>,
5365 scan_queue: Sender<ScanJob>,
5366}
5367
5368pub trait WorktreeModelHandle {
5369 #[cfg(feature = "test-support")]
5370 fn flush_fs_events<'a>(
5371 &self,
5372 cx: &'a mut gpui::TestAppContext,
5373 ) -> futures::future::LocalBoxFuture<'a, ()>;
5374
5375 #[cfg(feature = "test-support")]
5376 fn flush_fs_events_in_root_git_repository<'a>(
5377 &self,
5378 cx: &'a mut gpui::TestAppContext,
5379 ) -> futures::future::LocalBoxFuture<'a, ()>;
5380}
5381
5382impl WorktreeModelHandle for Entity<Worktree> {
5383 // When the worktree's FS event stream sometimes delivers "redundant" events for FS changes that
5384 // occurred before the worktree was constructed. These events can cause the worktree to perform
5385 // extra directory scans, and emit extra scan-state notifications.
5386 //
5387 // This function mutates the worktree's directory and waits for those mutations to be picked up,
5388 // to ensure that all redundant FS events have already been processed.
5389 #[cfg(feature = "test-support")]
5390 fn flush_fs_events<'a>(
5391 &self,
5392 cx: &'a mut gpui::TestAppContext,
5393 ) -> futures::future::LocalBoxFuture<'a, ()> {
5394 let file_name = "fs-event-sentinel";
5395
5396 let tree = self.clone();
5397 let (fs, root_path) = self.read_with(cx, |tree, _| {
5398 let tree = tree.as_local().unwrap();
5399 (tree.fs.clone(), tree.abs_path.clone())
5400 });
5401
5402 async move {
5403 // Subscribe to events BEFORE creating the file to avoid race condition
5404 // where events fire before subscription is set up
5405 let mut events = cx.events(&tree);
5406
5407 fs.create_file(&root_path.join(file_name), Default::default())
5408 .await
5409 .unwrap();
5410
5411 // Check if condition is already met before waiting for events
5412 let file_exists = || {
5413 tree.read_with(cx, |tree, _| {
5414 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5415 .is_some()
5416 })
5417 };
5418
5419 // Use select to avoid blocking indefinitely if events are delayed
5420 while !file_exists() {
5421 futures::select_biased! {
5422 _ = events.next() => {}
5423 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5424 }
5425 }
5426
5427 fs.remove_file(&root_path.join(file_name), Default::default())
5428 .await
5429 .unwrap();
5430
5431 // Check if condition is already met before waiting for events
5432 let file_gone = || {
5433 tree.read_with(cx, |tree, _| {
5434 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5435 .is_none()
5436 })
5437 };
5438
5439 // Use select to avoid blocking indefinitely if events are delayed
5440 while !file_gone() {
5441 futures::select_biased! {
5442 _ = events.next() => {}
5443 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5444 }
5445 }
5446
5447 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5448 .await;
5449 }
5450 .boxed_local()
5451 }
5452
5453 // This function is similar to flush_fs_events, except that it waits for events to be flushed in
5454 // the .git folder of the root repository.
5455 // The reason for its existence is that a repository's .git folder might live *outside* of the
5456 // worktree and thus its FS events might go through a different path.
5457 // In order to flush those, we need to create artificial events in the .git folder and wait
5458 // for the repository to be reloaded.
5459 #[cfg(feature = "test-support")]
5460 fn flush_fs_events_in_root_git_repository<'a>(
5461 &self,
5462 cx: &'a mut gpui::TestAppContext,
5463 ) -> futures::future::LocalBoxFuture<'a, ()> {
5464 let file_name = "fs-event-sentinel";
5465
5466 let tree = self.clone();
5467 let (fs, root_path, mut git_dir_scan_id) = self.read_with(cx, |tree, _| {
5468 let tree = tree.as_local().unwrap();
5469 let local_repo_entry = tree
5470 .git_repositories
5471 .values()
5472 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5473 .unwrap();
5474 (
5475 tree.fs.clone(),
5476 local_repo_entry.common_dir_abs_path.clone(),
5477 local_repo_entry.git_dir_scan_id,
5478 )
5479 });
5480
5481 let scan_id_increased = |tree: &mut Worktree, git_dir_scan_id: &mut usize| {
5482 let tree = tree.as_local().unwrap();
5483 // let repository = tree.repositories.first().unwrap();
5484 let local_repo_entry = tree
5485 .git_repositories
5486 .values()
5487 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5488 .unwrap();
5489
5490 if local_repo_entry.git_dir_scan_id > *git_dir_scan_id {
5491 *git_dir_scan_id = local_repo_entry.git_dir_scan_id;
5492 true
5493 } else {
5494 false
5495 }
5496 };
5497
5498 async move {
5499 // Subscribe to events BEFORE creating the file to avoid race condition
5500 // where events fire before subscription is set up
5501 let mut events = cx.events(&tree);
5502
5503 fs.create_file(&root_path.join(file_name), Default::default())
5504 .await
5505 .unwrap();
5506
5507 // Use select to avoid blocking indefinitely if events are delayed
5508 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5509 futures::select_biased! {
5510 _ = events.next() => {}
5511 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5512 }
5513 }
5514
5515 fs.remove_file(&root_path.join(file_name), Default::default())
5516 .await
5517 .unwrap();
5518
5519 // Use select to avoid blocking indefinitely if events are delayed
5520 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5521 futures::select_biased! {
5522 _ = events.next() => {}
5523 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5524 }
5525 }
5526
5527 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5528 .await;
5529 }
5530 .boxed_local()
5531 }
5532}
5533
5534#[derive(Clone, Debug)]
5535struct TraversalProgress<'a> {
5536 max_path: &'a RelPath,
5537 count: usize,
5538 non_ignored_count: usize,
5539 file_count: usize,
5540 non_ignored_file_count: usize,
5541}
5542
5543impl TraversalProgress<'_> {
5544 fn count(&self, include_files: bool, include_dirs: bool, include_ignored: bool) -> usize {
5545 match (include_files, include_dirs, include_ignored) {
5546 (true, true, true) => self.count,
5547 (true, true, false) => self.non_ignored_count,
5548 (true, false, true) => self.file_count,
5549 (true, false, false) => self.non_ignored_file_count,
5550 (false, true, true) => self.count - self.file_count,
5551 (false, true, false) => self.non_ignored_count - self.non_ignored_file_count,
5552 (false, false, _) => 0,
5553 }
5554 }
5555}
5556
5557impl<'a> sum_tree::Dimension<'a, EntrySummary> for TraversalProgress<'a> {
5558 fn zero(_cx: ()) -> Self {
5559 Default::default()
5560 }
5561
5562 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
5563 self.max_path = summary.max_path.as_ref();
5564 self.count += summary.count;
5565 self.non_ignored_count += summary.non_ignored_count;
5566 self.file_count += summary.file_count;
5567 self.non_ignored_file_count += summary.non_ignored_file_count;
5568 }
5569}
5570
5571impl Default for TraversalProgress<'_> {
5572 fn default() -> Self {
5573 Self {
5574 max_path: RelPath::empty(),
5575 count: 0,
5576 non_ignored_count: 0,
5577 file_count: 0,
5578 non_ignored_file_count: 0,
5579 }
5580 }
5581}
5582
5583#[derive(Debug)]
5584pub struct Traversal<'a> {
5585 snapshot: &'a Snapshot,
5586 cursor: sum_tree::Cursor<'a, 'static, Entry, TraversalProgress<'a>>,
5587 include_ignored: bool,
5588 include_files: bool,
5589 include_dirs: bool,
5590}
5591
5592impl<'a> Traversal<'a> {
5593 fn new(
5594 snapshot: &'a Snapshot,
5595 include_files: bool,
5596 include_dirs: bool,
5597 include_ignored: bool,
5598 start_path: &RelPath,
5599 ) -> Self {
5600 let mut cursor = snapshot.entries_by_path.cursor(());
5601 cursor.seek(&TraversalTarget::path(start_path), Bias::Left);
5602 let mut traversal = Self {
5603 snapshot,
5604 cursor,
5605 include_files,
5606 include_dirs,
5607 include_ignored,
5608 };
5609 if traversal.end_offset() == traversal.start_offset() {
5610 traversal.next();
5611 }
5612 traversal
5613 }
5614
5615 pub fn advance(&mut self) -> bool {
5616 self.advance_by(1)
5617 }
5618
5619 pub fn advance_by(&mut self, count: usize) -> bool {
5620 self.cursor.seek_forward(
5621 &TraversalTarget::Count {
5622 count: self.end_offset() + count,
5623 include_dirs: self.include_dirs,
5624 include_files: self.include_files,
5625 include_ignored: self.include_ignored,
5626 },
5627 Bias::Left,
5628 )
5629 }
5630
5631 pub fn advance_to_sibling(&mut self) -> bool {
5632 while let Some(entry) = self.cursor.item() {
5633 self.cursor
5634 .seek_forward(&TraversalTarget::successor(&entry.path), Bias::Left);
5635 if let Some(entry) = self.cursor.item()
5636 && (self.include_files || !entry.is_file())
5637 && (self.include_dirs || !entry.is_dir())
5638 && (self.include_ignored || !entry.is_ignored || entry.is_always_included)
5639 {
5640 return true;
5641 }
5642 }
5643 false
5644 }
5645
5646 pub fn back_to_parent(&mut self) -> bool {
5647 let Some(parent_path) = self.cursor.item().and_then(|entry| entry.path.parent()) else {
5648 return false;
5649 };
5650 self.cursor
5651 .seek(&TraversalTarget::path(parent_path), Bias::Left)
5652 }
5653
5654 pub fn entry(&self) -> Option<&'a Entry> {
5655 self.cursor.item()
5656 }
5657
5658 pub fn snapshot(&self) -> &'a Snapshot {
5659 self.snapshot
5660 }
5661
5662 pub fn start_offset(&self) -> usize {
5663 self.cursor
5664 .start()
5665 .count(self.include_files, self.include_dirs, self.include_ignored)
5666 }
5667
5668 pub fn end_offset(&self) -> usize {
5669 self.cursor
5670 .end()
5671 .count(self.include_files, self.include_dirs, self.include_ignored)
5672 }
5673}
5674
5675impl<'a> Iterator for Traversal<'a> {
5676 type Item = &'a Entry;
5677
5678 fn next(&mut self) -> Option<Self::Item> {
5679 if let Some(item) = self.entry() {
5680 self.advance();
5681 Some(item)
5682 } else {
5683 None
5684 }
5685 }
5686}
5687
5688#[derive(Debug, Clone, Copy)]
5689pub enum PathTarget<'a> {
5690 Path(&'a RelPath),
5691 Successor(&'a RelPath),
5692}
5693
5694impl PathTarget<'_> {
5695 fn cmp_path(&self, other: &RelPath) -> Ordering {
5696 match self {
5697 PathTarget::Path(path) => path.cmp(&other),
5698 PathTarget::Successor(path) => {
5699 if other.starts_with(path) {
5700 Ordering::Greater
5701 } else {
5702 Ordering::Equal
5703 }
5704 }
5705 }
5706 }
5707}
5708
5709impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, PathProgress<'a>> for PathTarget<'_> {
5710 fn cmp(&self, cursor_location: &PathProgress<'a>, _: S::Context<'_>) -> Ordering {
5711 self.cmp_path(cursor_location.max_path)
5712 }
5713}
5714
5715impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, TraversalProgress<'a>> for PathTarget<'_> {
5716 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: S::Context<'_>) -> Ordering {
5717 self.cmp_path(cursor_location.max_path)
5718 }
5719}
5720
5721#[derive(Debug)]
5722enum TraversalTarget<'a> {
5723 Path(PathTarget<'a>),
5724 Count {
5725 count: usize,
5726 include_files: bool,
5727 include_ignored: bool,
5728 include_dirs: bool,
5729 },
5730}
5731
5732impl<'a> TraversalTarget<'a> {
5733 fn path(path: &'a RelPath) -> Self {
5734 Self::Path(PathTarget::Path(path))
5735 }
5736
5737 fn successor(path: &'a RelPath) -> Self {
5738 Self::Path(PathTarget::Successor(path))
5739 }
5740
5741 fn cmp_progress(&self, progress: &TraversalProgress) -> Ordering {
5742 match self {
5743 TraversalTarget::Path(path) => path.cmp_path(progress.max_path),
5744 TraversalTarget::Count {
5745 count,
5746 include_files,
5747 include_dirs,
5748 include_ignored,
5749 } => Ord::cmp(
5750 count,
5751 &progress.count(*include_files, *include_dirs, *include_ignored),
5752 ),
5753 }
5754 }
5755}
5756
5757impl<'a> SeekTarget<'a, EntrySummary, TraversalProgress<'a>> for TraversalTarget<'_> {
5758 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5759 self.cmp_progress(cursor_location)
5760 }
5761}
5762
5763impl<'a> SeekTarget<'a, PathSummary<sum_tree::NoSummary>, TraversalProgress<'a>>
5764 for TraversalTarget<'_>
5765{
5766 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5767 self.cmp_progress(cursor_location)
5768 }
5769}
5770
5771pub struct ChildEntriesOptions {
5772 pub include_files: bool,
5773 pub include_dirs: bool,
5774 pub include_ignored: bool,
5775}
5776
5777pub struct ChildEntriesIter<'a> {
5778 parent_path: &'a RelPath,
5779 traversal: Traversal<'a>,
5780}
5781
5782impl<'a> Iterator for ChildEntriesIter<'a> {
5783 type Item = &'a Entry;
5784
5785 fn next(&mut self) -> Option<Self::Item> {
5786 if let Some(item) = self.traversal.entry()
5787 && item.path.starts_with(self.parent_path)
5788 {
5789 self.traversal.advance_to_sibling();
5790 return Some(item);
5791 }
5792 None
5793 }
5794}
5795
5796impl<'a> From<&'a Entry> for proto::Entry {
5797 fn from(entry: &'a Entry) -> Self {
5798 Self {
5799 id: entry.id.to_proto(),
5800 is_dir: entry.is_dir(),
5801 path: entry.path.as_ref().to_proto(),
5802 inode: entry.inode,
5803 mtime: entry.mtime.map(|time| time.into()),
5804 is_ignored: entry.is_ignored,
5805 is_hidden: entry.is_hidden,
5806 is_external: entry.is_external,
5807 is_fifo: entry.is_fifo,
5808 size: Some(entry.size),
5809 canonical_path: entry
5810 .canonical_path
5811 .as_ref()
5812 .map(|path| path.to_string_lossy().into_owned()),
5813 }
5814 }
5815}
5816
5817impl TryFrom<(&CharBag, &PathMatcher, proto::Entry)> for Entry {
5818 type Error = anyhow::Error;
5819
5820 fn try_from(
5821 (root_char_bag, always_included, entry): (&CharBag, &PathMatcher, proto::Entry),
5822 ) -> Result<Self> {
5823 let kind = if entry.is_dir {
5824 EntryKind::Dir
5825 } else {
5826 EntryKind::File
5827 };
5828
5829 let path =
5830 RelPath::from_proto(&entry.path).context("invalid relative path in proto message")?;
5831 let char_bag = char_bag_for_path(*root_char_bag, &path);
5832 let is_always_included = always_included.is_match(&path);
5833 Ok(Entry {
5834 id: ProjectEntryId::from_proto(entry.id),
5835 kind,
5836 path,
5837 inode: entry.inode,
5838 mtime: entry.mtime.map(|time| time.into()),
5839 size: entry.size.unwrap_or(0),
5840 canonical_path: entry
5841 .canonical_path
5842 .map(|path_string| Arc::from(PathBuf::from(path_string))),
5843 is_ignored: entry.is_ignored,
5844 is_hidden: entry.is_hidden,
5845 is_always_included,
5846 is_external: entry.is_external,
5847 is_private: false,
5848 char_bag,
5849 is_fifo: entry.is_fifo,
5850 })
5851 }
5852}
5853
5854#[derive(Clone, Copy, Debug, Default, Hash, PartialEq, Eq, PartialOrd, Ord)]
5855pub struct ProjectEntryId(usize);
5856
5857impl ProjectEntryId {
5858 pub const MAX: Self = Self(usize::MAX);
5859 pub const MIN: Self = Self(usize::MIN);
5860
5861 pub fn new(counter: &AtomicUsize) -> Self {
5862 Self(counter.fetch_add(1, SeqCst))
5863 }
5864
5865 pub fn from_proto(id: u64) -> Self {
5866 Self(id as usize)
5867 }
5868
5869 pub fn to_proto(self) -> u64 {
5870 self.0 as u64
5871 }
5872
5873 pub fn from_usize(id: usize) -> Self {
5874 ProjectEntryId(id)
5875 }
5876
5877 pub fn to_usize(self) -> usize {
5878 self.0
5879 }
5880}
5881
5882#[cfg(feature = "test-support")]
5883impl CreatedEntry {
5884 pub fn into_included(self) -> Option<Entry> {
5885 match self {
5886 CreatedEntry::Included(entry) => Some(entry),
5887 CreatedEntry::Excluded { .. } => None,
5888 }
5889 }
5890}
5891
5892fn parse_gitfile(content: &str) -> anyhow::Result<&Path> {
5893 let path = content
5894 .strip_prefix("gitdir:")
5895 .with_context(|| format!("parsing gitfile content {content:?}"))?;
5896 Ok(Path::new(path.trim()))
5897}
5898
5899async fn discover_git_paths(dot_git_abs_path: &Arc<Path>, fs: &dyn Fs) -> (Arc<Path>, Arc<Path>) {
5900 let mut repository_dir_abs_path = dot_git_abs_path.clone();
5901 let mut common_dir_abs_path = dot_git_abs_path.clone();
5902
5903 if let Some(path) = fs
5904 .load(dot_git_abs_path)
5905 .await
5906 .ok()
5907 .as_ref()
5908 .and_then(|contents| parse_gitfile(contents).log_err())
5909 {
5910 let path = dot_git_abs_path
5911 .parent()
5912 .unwrap_or(Path::new(""))
5913 .join(path);
5914 if let Some(path) = fs.canonicalize(&path).await.log_err() {
5915 repository_dir_abs_path = Path::new(&path).into();
5916 common_dir_abs_path = repository_dir_abs_path.clone();
5917
5918 if let Some(commondir_contents) = fs.load(&path.join("commondir")).await.ok()
5919 && let Some(commondir_path) = fs
5920 .canonicalize(&path.join(commondir_contents.trim()))
5921 .await
5922 .log_err()
5923 {
5924 common_dir_abs_path = commondir_path.as_path().into();
5925 }
5926 }
5927 };
5928 (repository_dir_abs_path, common_dir_abs_path)
5929}
5930
5931struct NullWatcher;
5932
5933impl fs::Watcher for NullWatcher {
5934 fn add(&self, _path: &Path) -> Result<()> {
5935 Ok(())
5936 }
5937
5938 fn remove(&self, _path: &Path) -> Result<()> {
5939 Ok(())
5940 }
5941}
5942
5943const FILE_ANALYSIS_BYTES: usize = 1024;
5944
5945async fn decode_file_text(
5946 fs: &dyn Fs,
5947 abs_path: &Path,
5948) -> Result<(String, &'static Encoding, bool)> {
5949 let mut file = fs
5950 .open_sync(&abs_path)
5951 .await
5952 .with_context(|| format!("opening file {abs_path:?}"))?;
5953
5954 // First, read the beginning of the file to determine its kind and encoding.
5955 // We do not want to load an entire large blob into memory only to discard it.
5956 let mut file_first_bytes = Vec::with_capacity(FILE_ANALYSIS_BYTES);
5957 let mut buf = [0u8; FILE_ANALYSIS_BYTES];
5958 let mut reached_eof = false;
5959 loop {
5960 if file_first_bytes.len() >= FILE_ANALYSIS_BYTES {
5961 break;
5962 }
5963 let n = file
5964 .read(&mut buf)
5965 .with_context(|| format!("reading bytes of the file {abs_path:?}"))?;
5966 if n == 0 {
5967 reached_eof = true;
5968 break;
5969 }
5970 file_first_bytes.extend_from_slice(&buf[..n]);
5971 }
5972 let (bom_encoding, byte_content) = decode_byte_header(&file_first_bytes);
5973 anyhow::ensure!(
5974 byte_content != ByteContent::Binary,
5975 "Binary files are not supported"
5976 );
5977
5978 // If the file is eligible for opening, read the rest of the file.
5979 let mut content = file_first_bytes;
5980 if !reached_eof {
5981 let mut buf = [0u8; 8 * 1024];
5982 loop {
5983 let n = file
5984 .read(&mut buf)
5985 .with_context(|| format!("reading remaining bytes of the file {abs_path:?}"))?;
5986 if n == 0 {
5987 break;
5988 }
5989 content.extend_from_slice(&buf[..n]);
5990 }
5991 }
5992 decode_byte_full(content, bom_encoding, byte_content)
5993}
5994
5995fn decode_byte_header(prefix: &[u8]) -> (Option<&'static Encoding>, ByteContent) {
5996 if let Some((encoding, _bom_len)) = Encoding::for_bom(prefix) {
5997 return (Some(encoding), ByteContent::Unknown);
5998 }
5999 (None, analyze_byte_content(prefix))
6000}
6001
6002fn decode_byte_full(
6003 bytes: Vec<u8>,
6004 bom_encoding: Option<&'static Encoding>,
6005 byte_content: ByteContent,
6006) -> Result<(String, &'static Encoding, bool)> {
6007 if let Some(encoding) = bom_encoding {
6008 let (cow, _) = encoding.decode_with_bom_removal(&bytes);
6009 return Ok((cow.into_owned(), encoding, true));
6010 }
6011
6012 match byte_content {
6013 ByteContent::Utf16Le => {
6014 let encoding = encoding_rs::UTF_16LE;
6015 let (cow, _, _) = encoding.decode(&bytes);
6016 return Ok((cow.into_owned(), encoding, false));
6017 }
6018 ByteContent::Utf16Be => {
6019 let encoding = encoding_rs::UTF_16BE;
6020 let (cow, _, _) = encoding.decode(&bytes);
6021 return Ok((cow.into_owned(), encoding, false));
6022 }
6023 ByteContent::Binary => {
6024 anyhow::bail!("Binary files are not supported");
6025 }
6026 ByteContent::Unknown => {}
6027 }
6028
6029 fn detect_encoding(bytes: Vec<u8>) -> (String, &'static Encoding) {
6030 let mut detector = EncodingDetector::new();
6031 detector.feed(&bytes, true);
6032
6033 let encoding = detector.guess(None, true); // Use None for TLD hint to ensure neutral detection logic.
6034
6035 let (cow, _, _) = encoding.decode(&bytes);
6036 (cow.into_owned(), encoding)
6037 }
6038
6039 match String::from_utf8(bytes) {
6040 Ok(text) => {
6041 // ISO-2022-JP (and other ISO-2022 variants) consists entirely of 7-bit ASCII bytes,
6042 // so it is valid UTF-8. However, it contains escape sequences starting with '\x1b'.
6043 // If we find an escape character, we double-check the encoding to prevent
6044 // displaying raw escape sequences instead of the correct characters.
6045 if text.contains('\x1b') {
6046 let (s, enc) = detect_encoding(text.into_bytes());
6047 Ok((s, enc, false))
6048 } else {
6049 Ok((text, encoding_rs::UTF_8, false))
6050 }
6051 }
6052 Err(e) => {
6053 let (s, enc) = detect_encoding(e.into_bytes());
6054 Ok((s, enc, false))
6055 }
6056 }
6057}
6058
6059#[derive(PartialEq)]
6060enum ByteContent {
6061 Utf16Le,
6062 Utf16Be,
6063 Binary,
6064 Unknown,
6065}
6066
6067// Heuristic check using null byte distribution plus a generic text-likeness
6068// heuristic. This prefers UTF-16 when many bytes are NUL and otherwise
6069// distinguishes between text-like and binary-like content.
6070fn analyze_byte_content(bytes: &[u8]) -> ByteContent {
6071 if bytes.len() < 2 {
6072 return ByteContent::Unknown;
6073 }
6074
6075 if is_known_binary_header(bytes) {
6076 return ByteContent::Binary;
6077 }
6078
6079 let limit = bytes.len().min(FILE_ANALYSIS_BYTES);
6080 let mut even_null_count = 0usize;
6081 let mut odd_null_count = 0usize;
6082 let mut non_text_like_count = 0usize;
6083
6084 for (i, &byte) in bytes[..limit].iter().enumerate() {
6085 if byte == 0 {
6086 if i % 2 == 0 {
6087 even_null_count += 1;
6088 } else {
6089 odd_null_count += 1;
6090 }
6091 non_text_like_count += 1;
6092 continue;
6093 }
6094
6095 let is_text_like = match byte {
6096 b'\t' | b'\n' | b'\r' | 0x0C => true,
6097 0x20..=0x7E => true,
6098 // Treat bytes that are likely part of UTF-8 or single-byte encodings as text-like.
6099 0x80..=0xBF | 0xC2..=0xF4 => true,
6100 _ => false,
6101 };
6102
6103 if !is_text_like {
6104 non_text_like_count += 1;
6105 }
6106 }
6107
6108 let total_null_count = even_null_count + odd_null_count;
6109
6110 // If there are no NUL bytes at all, this is overwhelmingly likely to be text.
6111 if total_null_count == 0 {
6112 return ByteContent::Unknown;
6113 }
6114
6115 if total_null_count >= limit / 16 {
6116 if even_null_count > odd_null_count * 4 {
6117 return ByteContent::Utf16Be;
6118 }
6119 if odd_null_count > even_null_count * 4 {
6120 return ByteContent::Utf16Le;
6121 }
6122 return ByteContent::Binary;
6123 }
6124
6125 if non_text_like_count * 100 < limit * 8 {
6126 ByteContent::Unknown
6127 } else {
6128 ByteContent::Binary
6129 }
6130}
6131
6132fn is_known_binary_header(bytes: &[u8]) -> bool {
6133 bytes.starts_with(b"%PDF-") // PDF
6134 || bytes.starts_with(b"PK\x03\x04") // ZIP local header
6135 || bytes.starts_with(b"PK\x05\x06") // ZIP end of central directory
6136 || bytes.starts_with(b"PK\x07\x08") // ZIP spanning/splitting
6137 || bytes.starts_with(b"\x89PNG\r\n\x1a\n") // PNG
6138 || bytes.starts_with(b"\xFF\xD8\xFF") // JPEG
6139 || bytes.starts_with(b"GIF87a") // GIF87a
6140 || bytes.starts_with(b"GIF89a") // GIF89a
6141}