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::{
11 Fs, MTime, PathEvent, PathEventKind, RemoveOptions, TrashedEntry, Watcher, copy_recursive,
12 read_dir_items,
13};
14use futures::{
15 FutureExt as _, Stream, StreamExt,
16 channel::{
17 mpsc::{self, UnboundedSender},
18 oneshot,
19 },
20 select_biased, stream,
21 task::Poll,
22};
23use fuzzy::CharBag;
24use git::{
25 COMMIT_MESSAGE, DOT_GIT, FSMONITOR_DAEMON, GITIGNORE, INDEX_LOCK, LFS_DIR, REPO_EXCLUDE,
26 status::GitSummary,
27};
28use gpui::{
29 App, AppContext as _, AsyncApp, BackgroundExecutor, Context, Entity, EventEmitter, Priority,
30 Task,
31};
32use ignore::IgnoreStack;
33use language::DiskState;
34
35use parking_lot::Mutex;
36use paths::{local_settings_folder_name, local_vscode_folder_name};
37use postage::{
38 barrier,
39 prelude::{Sink as _, Stream as _},
40 watch,
41};
42use rpc::{
43 AnyProtoClient,
44 proto::{self, split_worktree_update},
45};
46pub use settings::WorktreeId;
47use settings::{Settings, SettingsLocation, SettingsStore};
48use smallvec::{SmallVec, smallvec};
49use smol::channel::{self, Sender};
50use std::{
51 any::Any,
52 borrow::Borrow as _,
53 cmp::Ordering,
54 collections::hash_map,
55 convert::TryFrom,
56 ffi::OsStr,
57 fmt,
58 future::Future,
59 mem::{self},
60 ops::{Deref, DerefMut, Range},
61 path::{Path, PathBuf},
62 pin::Pin,
63 sync::{
64 Arc,
65 atomic::{AtomicUsize, Ordering::SeqCst},
66 },
67 time::{Duration, Instant},
68};
69use sum_tree::{Bias, Dimensions, Edit, KeyedItem, SeekTarget, SumTree, Summary, TreeMap, TreeSet};
70use text::{LineEnding, Rope};
71use util::{
72 ResultExt, maybe,
73 paths::{PathMatcher, PathStyle, SanitizedPath, home_dir},
74 rel_path::{RelPath, RelPathBuf},
75};
76pub use worktree_settings::WorktreeSettings;
77
78use crate::ignore::IgnoreKind;
79
80pub const FS_WATCH_LATENCY: Duration = Duration::from_millis(100);
81
82/// A set of local or remote files that are being opened as part of a project.
83/// Responsible for tracking related FS (for local)/collab (for remote) events and corresponding updates.
84/// Stores git repositories data and the diagnostics for the file(s).
85///
86/// Has an absolute path, and may be set to be visible in Zed UI or not.
87/// May correspond to a directory or a single file.
88/// Possible examples:
89/// * a drag and dropped file — may be added as an invisible, "ephemeral" entry to the current worktree
90/// * a directory opened in Zed — may be added as a visible entry to the current worktree
91///
92/// Uses [`Entry`] to track the state of each file/directory, can look up absolute paths for entries.
93pub enum Worktree {
94 Local(LocalWorktree),
95 Remote(RemoteWorktree),
96}
97
98/// An entry, created in the worktree.
99#[derive(Debug)]
100pub enum CreatedEntry {
101 /// Got created and indexed by the worktree, receiving a corresponding entry.
102 Included(Entry),
103 /// Got created, but not indexed due to falling under exclusion filters.
104 Excluded { abs_path: PathBuf },
105}
106
107#[derive(Debug)]
108pub struct LoadedFile {
109 pub file: Arc<File>,
110 pub text: String,
111 pub encoding: &'static Encoding,
112 pub has_bom: bool,
113}
114
115pub struct LoadedBinaryFile {
116 pub file: Arc<File>,
117 pub content: Vec<u8>,
118}
119
120impl fmt::Debug for LoadedBinaryFile {
121 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
122 f.debug_struct("LoadedBinaryFile")
123 .field("file", &self.file)
124 .field("content_bytes", &self.content.len())
125 .finish()
126 }
127}
128
129pub struct LocalWorktree {
130 snapshot: LocalSnapshot,
131 scan_requests_tx: channel::Sender<ScanRequest>,
132 path_prefixes_to_scan_tx: channel::Sender<PathPrefixScanRequest>,
133 is_scanning: (watch::Sender<bool>, watch::Receiver<bool>),
134 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
135 _background_scanner_tasks: Vec<Task<()>>,
136 update_observer: Option<UpdateObservationState>,
137 fs: Arc<dyn Fs>,
138 fs_case_sensitive: bool,
139 visible: bool,
140 next_entry_id: Arc<AtomicUsize>,
141 settings: WorktreeSettings,
142 share_private_files: bool,
143 scanning_enabled: bool,
144}
145
146pub struct PathPrefixScanRequest {
147 path: Arc<RelPath>,
148 done: SmallVec<[barrier::Sender; 1]>,
149}
150
151struct ScanRequest {
152 relative_paths: Vec<Arc<RelPath>>,
153 done: SmallVec<[barrier::Sender; 1]>,
154}
155
156pub struct RemoteWorktree {
157 snapshot: Snapshot,
158 background_snapshot: Arc<Mutex<(Snapshot, Vec<proto::UpdateWorktree>)>>,
159 project_id: u64,
160 client: AnyProtoClient,
161 file_scan_inclusions: PathMatcher,
162 updates_tx: Option<UnboundedSender<proto::UpdateWorktree>>,
163 update_observer: Option<mpsc::UnboundedSender<proto::UpdateWorktree>>,
164 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
165 replica_id: ReplicaId,
166 visible: bool,
167 disconnected: bool,
168 received_initial_update: bool,
169}
170
171#[derive(Clone)]
172pub struct Snapshot {
173 id: WorktreeId,
174 /// The absolute path of the worktree root.
175 abs_path: Arc<SanitizedPath>,
176 path_style: PathStyle,
177 root_name: Arc<RelPath>,
178 root_char_bag: CharBag,
179 entries_by_path: SumTree<Entry>,
180 entries_by_id: SumTree<PathEntry>,
181 root_repo_common_dir: Option<Arc<SanitizedPath>>,
182 always_included_entries: Vec<Arc<RelPath>>,
183
184 /// A number that increases every time the worktree begins scanning
185 /// a set of paths from the filesystem. This scanning could be caused
186 /// by some operation performed on the worktree, such as reading or
187 /// writing a file, or by an event reported by the filesystem.
188 scan_id: usize,
189
190 /// The latest scan id that has completed, and whose preceding scans
191 /// have all completed. The current `scan_id` could be more than one
192 /// greater than the `completed_scan_id` if operations are performed
193 /// on the worktree while it is processing a file-system event.
194 completed_scan_id: usize,
195}
196
197/// This path corresponds to the 'content path' of a repository in relation
198/// to Zed's project root.
199/// In the majority of the cases, this is the folder that contains the .git folder.
200/// But if a sub-folder of a git repository is opened, this corresponds to the
201/// project root and the .git folder is located in a parent directory.
202#[derive(Clone, Debug, Ord, PartialOrd, Eq, PartialEq, Hash)]
203pub enum WorkDirectory {
204 InProject {
205 relative_path: Arc<RelPath>,
206 },
207 AboveProject {
208 absolute_path: Arc<Path>,
209 location_in_repo: Arc<Path>,
210 },
211}
212
213impl WorkDirectory {
214 fn path_key(&self) -> PathKey {
215 match self {
216 WorkDirectory::InProject { relative_path } => PathKey(relative_path.clone()),
217 WorkDirectory::AboveProject { .. } => PathKey(RelPath::empty().into()),
218 }
219 }
220
221 /// Returns true if the given path is a child of the work directory.
222 ///
223 /// Note that the path may not be a member of this repository, if there
224 /// is a repository in a directory between these two paths
225 /// external .git folder in a parent folder of the project root.
226 #[track_caller]
227 pub fn directory_contains(&self, path: &RelPath) -> bool {
228 match self {
229 WorkDirectory::InProject { relative_path } => path.starts_with(relative_path),
230 WorkDirectory::AboveProject { .. } => true,
231 }
232 }
233}
234
235impl Default for WorkDirectory {
236 fn default() -> Self {
237 Self::InProject {
238 relative_path: Arc::from(RelPath::empty()),
239 }
240 }
241}
242
243#[derive(Clone)]
244pub struct LocalSnapshot {
245 snapshot: Snapshot,
246 global_gitignore: Option<Arc<Gitignore>>,
247 /// Exclude files for all git repositories in the worktree, indexed by their absolute path.
248 /// The boolean indicates whether the gitignore needs to be updated.
249 repo_exclude_by_work_dir_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
250 /// All of the gitignore files in the worktree, indexed by their absolute path.
251 /// The boolean indicates whether the gitignore needs to be updated.
252 ignores_by_parent_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
253 /// All of the git repositories in the worktree, indexed by the project entry
254 /// id of their parent directory.
255 git_repositories: TreeMap<ProjectEntryId, LocalRepositoryEntry>,
256 /// The file handle of the worktree root
257 /// (so we can find it after it's been moved)
258 root_file_handle: Option<Arc<dyn fs::FileHandle>>,
259}
260
261struct BackgroundScannerState {
262 snapshot: LocalSnapshot,
263 symlink_paths_by_target: HashMap<Arc<Path>, SmallVec<[Arc<RelPath>; 1]>>,
264 scanned_dirs: HashSet<ProjectEntryId>,
265 watched_dir_abs_paths_by_entry_id: HashMap<ProjectEntryId, Arc<Path>>,
266 path_prefixes_to_scan: HashSet<Arc<RelPath>>,
267 paths_to_scan: HashSet<Arc<RelPath>>,
268 /// The ids of all of the entries that were removed from the snapshot
269 /// as part of the current update. These entry ids may be re-used
270 /// if the same inode is discovered at a new path, or if the given
271 /// path is re-created after being deleted.
272 removed_entries: HashMap<u64, Entry>,
273 changed_paths: Vec<Arc<RelPath>>,
274 prev_snapshot: Snapshot,
275 scanning_enabled: bool,
276}
277
278#[derive(Clone, Debug, Eq, PartialEq)]
279struct EventRoot {
280 path: Arc<RelPath>,
281 was_rescanned: bool,
282}
283
284#[derive(Debug, Clone)]
285struct LocalRepositoryEntry {
286 work_directory_id: ProjectEntryId,
287 work_directory: WorkDirectory,
288 work_directory_abs_path: Arc<Path>,
289 git_dir_scan_id: usize,
290 /// Absolute path to the original .git entry that caused us to create this repository.
291 ///
292 /// This is normally a directory, but may be a "gitfile" that points to a directory elsewhere
293 /// (whose path we then store in `repository_dir_abs_path`).
294 dot_git_abs_path: Arc<Path>,
295 /// Absolute path to the "commondir" for this repository.
296 ///
297 /// This is always a directory. For a normal repository, this is the same as
298 /// `dot_git_abs_path`. For a linked worktree, this is the main repo's `.git`
299 /// directory (resolved from the worktree's `commondir` file). For a submodule,
300 /// this equals `repository_dir_abs_path` (submodules don't have a `commondir`
301 /// file).
302 common_dir_abs_path: Arc<Path>,
303 /// Absolute path to the directory holding the repository's state.
304 ///
305 /// For a normal repository, this is a directory and coincides with `dot_git_abs_path` and
306 /// `common_dir_abs_path`. For a submodule or worktree, this is some subdirectory of the
307 /// commondir like `/project/.git/modules/foo`.
308 repository_dir_abs_path: Arc<Path>,
309}
310
311impl sum_tree::Item for LocalRepositoryEntry {
312 type Summary = PathSummary<sum_tree::NoSummary>;
313
314 fn summary(&self, _: <Self::Summary as Summary>::Context<'_>) -> Self::Summary {
315 PathSummary {
316 max_path: self.work_directory.path_key().0,
317 item_summary: sum_tree::NoSummary,
318 }
319 }
320}
321
322impl KeyedItem for LocalRepositoryEntry {
323 type Key = PathKey;
324
325 fn key(&self) -> Self::Key {
326 self.work_directory.path_key()
327 }
328}
329
330impl Deref for LocalRepositoryEntry {
331 type Target = WorkDirectory;
332
333 fn deref(&self) -> &Self::Target {
334 &self.work_directory
335 }
336}
337
338impl Deref for LocalSnapshot {
339 type Target = Snapshot;
340
341 fn deref(&self) -> &Self::Target {
342 &self.snapshot
343 }
344}
345
346impl DerefMut for LocalSnapshot {
347 fn deref_mut(&mut self) -> &mut Self::Target {
348 &mut self.snapshot
349 }
350}
351
352enum ScanState {
353 Started,
354 Updated {
355 snapshot: LocalSnapshot,
356 changes: UpdatedEntriesSet,
357 barrier: SmallVec<[barrier::Sender; 1]>,
358 scanning: bool,
359 },
360 RootUpdated {
361 new_path: Arc<SanitizedPath>,
362 },
363 RootDeleted,
364}
365
366struct UpdateObservationState {
367 snapshots_tx: mpsc::UnboundedSender<(LocalSnapshot, UpdatedEntriesSet)>,
368 resume_updates: watch::Sender<()>,
369 _maintain_remote_snapshot: Task<Option<()>>,
370}
371
372#[derive(Debug, Clone)]
373pub enum Event {
374 UpdatedEntries(UpdatedEntriesSet),
375 UpdatedGitRepositories(UpdatedGitRepositoriesSet),
376 UpdatedRootRepoCommonDir {
377 old: Option<Arc<SanitizedPath>>,
378 },
379 DeletedEntry(ProjectEntryId),
380 /// The worktree root itself has been deleted (for single-file worktrees)
381 Deleted,
382}
383
384impl EventEmitter<Event> for Worktree {}
385
386impl Worktree {
387 pub async fn local(
388 path: impl Into<Arc<Path>>,
389 visible: bool,
390 fs: Arc<dyn Fs>,
391 next_entry_id: Arc<AtomicUsize>,
392 scanning_enabled: bool,
393 worktree_id: WorktreeId,
394 cx: &mut AsyncApp,
395 ) -> Result<Entity<Self>> {
396 let abs_path = path.into();
397 let metadata = fs
398 .metadata(&abs_path)
399 .await
400 .context("failed to stat worktree path")?;
401
402 let fs_case_sensitive = fs.is_case_sensitive().await;
403
404 let root_file_handle = if metadata.as_ref().is_some() {
405 fs.open_handle(&abs_path)
406 .await
407 .with_context(|| {
408 format!(
409 "failed to open local worktree root at {}",
410 abs_path.display()
411 )
412 })
413 .log_err()
414 } else {
415 None
416 };
417
418 let root_repo_common_dir = discover_root_repo_common_dir(&abs_path, fs.as_ref())
419 .await
420 .map(SanitizedPath::from_arc);
421
422 Ok(cx.new(move |cx: &mut Context<Worktree>| {
423 let mut snapshot = LocalSnapshot {
424 ignores_by_parent_abs_path: Default::default(),
425 global_gitignore: Default::default(),
426 repo_exclude_by_work_dir_abs_path: Default::default(),
427 git_repositories: Default::default(),
428 snapshot: Snapshot::new(
429 worktree_id,
430 abs_path
431 .file_name()
432 .and_then(|f| f.to_str())
433 .map_or(RelPath::empty().into(), |f| {
434 RelPath::unix(f).unwrap().into()
435 }),
436 abs_path.clone(),
437 PathStyle::local(),
438 ),
439 root_file_handle,
440 };
441 snapshot.root_repo_common_dir = root_repo_common_dir;
442
443 let worktree_id = snapshot.id();
444 let settings_location = Some(SettingsLocation {
445 worktree_id,
446 path: RelPath::empty(),
447 });
448
449 let settings = WorktreeSettings::get(settings_location, cx).clone();
450 cx.observe_global::<SettingsStore>(move |this, cx| {
451 if let Self::Local(this) = this {
452 let settings = WorktreeSettings::get(settings_location, cx).clone();
453 if this.settings != settings {
454 this.settings = settings;
455 this.restart_background_scanners(cx);
456 }
457 }
458 })
459 .detach();
460
461 let share_private_files = false;
462 if let Some(metadata) = metadata {
463 let mut entry = Entry::new(
464 RelPath::empty().into(),
465 &metadata,
466 ProjectEntryId::new(&next_entry_id),
467 snapshot.root_char_bag,
468 None,
469 );
470 if metadata.is_dir {
471 if !scanning_enabled {
472 entry.kind = EntryKind::UnloadedDir;
473 }
474 } else {
475 if let Some(file_name) = abs_path.file_name()
476 && let Some(file_name) = file_name.to_str()
477 && let Ok(path) = RelPath::unix(file_name)
478 {
479 entry.is_private = !share_private_files && settings.is_path_private(path);
480 entry.is_hidden = settings.is_path_hidden(path);
481 }
482 }
483 cx.foreground_executor()
484 .block_on(snapshot.insert_entry(entry, fs.as_ref()));
485 }
486
487 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
488 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
489 let mut worktree = LocalWorktree {
490 share_private_files,
491 next_entry_id,
492 snapshot,
493 is_scanning: watch::channel_with(true),
494 snapshot_subscriptions: Default::default(),
495 update_observer: None,
496 scan_requests_tx,
497 path_prefixes_to_scan_tx,
498 _background_scanner_tasks: Vec::new(),
499 fs,
500 fs_case_sensitive,
501 visible,
502 settings,
503 scanning_enabled,
504 };
505 worktree.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
506 Worktree::Local(worktree)
507 }))
508 }
509
510 pub fn remote(
511 project_id: u64,
512 replica_id: ReplicaId,
513 worktree: proto::WorktreeMetadata,
514 client: AnyProtoClient,
515 path_style: PathStyle,
516 cx: &mut App,
517 ) -> Entity<Self> {
518 cx.new(|cx: &mut Context<Self>| {
519 let mut snapshot = Snapshot::new(
520 WorktreeId::from_proto(worktree.id),
521 RelPath::from_proto(&worktree.root_name)
522 .unwrap_or_else(|_| RelPath::empty().into()),
523 Path::new(&worktree.abs_path).into(),
524 path_style,
525 );
526
527 snapshot.root_repo_common_dir = worktree
528 .root_repo_common_dir
529 .map(|p| SanitizedPath::new_arc(Path::new(&p)));
530
531 let background_snapshot = Arc::new(Mutex::new((
532 snapshot.clone(),
533 Vec::<proto::UpdateWorktree>::new(),
534 )));
535 let (background_updates_tx, mut background_updates_rx) =
536 mpsc::unbounded::<proto::UpdateWorktree>();
537 let (mut snapshot_updated_tx, mut snapshot_updated_rx) = watch::channel();
538
539 let worktree_id = snapshot.id();
540 let settings_location = Some(SettingsLocation {
541 worktree_id,
542 path: RelPath::empty(),
543 });
544
545 let settings = WorktreeSettings::get(settings_location, cx).clone();
546 let worktree = RemoteWorktree {
547 client,
548 project_id,
549 replica_id,
550 snapshot,
551 file_scan_inclusions: settings.parent_dir_scan_inclusions.clone(),
552 background_snapshot: background_snapshot.clone(),
553 updates_tx: Some(background_updates_tx),
554 update_observer: None,
555 snapshot_subscriptions: Default::default(),
556 visible: worktree.visible,
557 disconnected: false,
558 received_initial_update: false,
559 };
560
561 // Apply updates to a separate snapshot in a background task, then
562 // send them to a foreground task which updates the model.
563 cx.background_spawn(async move {
564 while let Some(update) = background_updates_rx.next().await {
565 {
566 let mut lock = background_snapshot.lock();
567 lock.0.apply_remote_update(
568 update.clone(),
569 &settings.parent_dir_scan_inclusions,
570 );
571 lock.1.push(update);
572 }
573 snapshot_updated_tx.send(()).await.ok();
574 }
575 })
576 .detach();
577
578 // On the foreground task, update to the latest snapshot and notify
579 // any update observer of all updates that led to that snapshot.
580 cx.spawn(async move |this, cx| {
581 while (snapshot_updated_rx.recv().await).is_some() {
582 this.update(cx, |this, cx| {
583 let this = this.as_remote_mut().unwrap();
584
585 // The watch channel delivers an initial signal before
586 // any real updates arrive. Skip these spurious wakeups.
587 if this.background_snapshot.lock().1.is_empty() {
588 return;
589 }
590
591 let old_root_repo_common_dir = this.snapshot.root_repo_common_dir.clone();
592 let mut entries_changed = false;
593 {
594 let mut lock = this.background_snapshot.lock();
595 this.snapshot = lock.0.clone();
596 for update in lock.1.drain(..) {
597 entries_changed |= !update.updated_entries.is_empty()
598 || !update.removed_entries.is_empty();
599 if let Some(tx) = &this.update_observer {
600 tx.unbounded_send(update).ok();
601 }
602 }
603 };
604
605 if entries_changed {
606 cx.emit(Event::UpdatedEntries(Arc::default()));
607 }
608 let is_first_update = !this.received_initial_update;
609 this.received_initial_update = true;
610 if this.snapshot.root_repo_common_dir != old_root_repo_common_dir
611 || (is_first_update && this.snapshot.root_repo_common_dir.is_none())
612 {
613 cx.emit(Event::UpdatedRootRepoCommonDir {
614 old: old_root_repo_common_dir,
615 });
616 }
617 cx.notify();
618 while let Some((scan_id, _)) = this.snapshot_subscriptions.front() {
619 if this.observed_snapshot(*scan_id) {
620 let (_, tx) = this.snapshot_subscriptions.pop_front().unwrap();
621 let _ = tx.send(());
622 } else {
623 break;
624 }
625 }
626 })?;
627 }
628 anyhow::Ok(())
629 })
630 .detach();
631
632 Worktree::Remote(worktree)
633 })
634 }
635
636 pub fn as_local(&self) -> Option<&LocalWorktree> {
637 if let Worktree::Local(worktree) = self {
638 Some(worktree)
639 } else {
640 None
641 }
642 }
643
644 pub fn as_remote(&self) -> Option<&RemoteWorktree> {
645 if let Worktree::Remote(worktree) = self {
646 Some(worktree)
647 } else {
648 None
649 }
650 }
651
652 pub fn as_local_mut(&mut self) -> Option<&mut LocalWorktree> {
653 if let Worktree::Local(worktree) = self {
654 Some(worktree)
655 } else {
656 None
657 }
658 }
659
660 pub fn as_remote_mut(&mut self) -> Option<&mut RemoteWorktree> {
661 if let Worktree::Remote(worktree) = self {
662 Some(worktree)
663 } else {
664 None
665 }
666 }
667
668 pub fn is_local(&self) -> bool {
669 matches!(self, Worktree::Local(_))
670 }
671
672 pub fn is_remote(&self) -> bool {
673 !self.is_local()
674 }
675
676 pub fn settings_location(&self, _: &Context<Self>) -> SettingsLocation<'static> {
677 SettingsLocation {
678 worktree_id: self.id(),
679 path: RelPath::empty(),
680 }
681 }
682
683 pub fn snapshot(&self) -> Snapshot {
684 match self {
685 Worktree::Local(worktree) => worktree.snapshot.snapshot.clone(),
686 Worktree::Remote(worktree) => worktree.snapshot.clone(),
687 }
688 }
689
690 pub fn scan_id(&self) -> usize {
691 match self {
692 Worktree::Local(worktree) => worktree.snapshot.scan_id,
693 Worktree::Remote(worktree) => worktree.snapshot.scan_id,
694 }
695 }
696
697 pub fn metadata_proto(&self) -> proto::WorktreeMetadata {
698 proto::WorktreeMetadata {
699 id: self.id().to_proto(),
700 root_name: self.root_name().to_proto(),
701 visible: self.is_visible(),
702 abs_path: self.abs_path().to_string_lossy().into_owned(),
703 root_repo_common_dir: self
704 .root_repo_common_dir()
705 .map(|p| p.to_string_lossy().into_owned()),
706 }
707 }
708
709 pub fn completed_scan_id(&self) -> usize {
710 match self {
711 Worktree::Local(worktree) => worktree.snapshot.completed_scan_id,
712 Worktree::Remote(worktree) => worktree.snapshot.completed_scan_id,
713 }
714 }
715
716 pub fn is_visible(&self) -> bool {
717 match self {
718 Worktree::Local(worktree) => worktree.visible,
719 Worktree::Remote(worktree) => worktree.visible,
720 }
721 }
722
723 pub fn replica_id(&self) -> ReplicaId {
724 match self {
725 Worktree::Local(_) => ReplicaId::LOCAL,
726 Worktree::Remote(worktree) => worktree.replica_id,
727 }
728 }
729
730 pub fn abs_path(&self) -> Arc<Path> {
731 match self {
732 Worktree::Local(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
733 Worktree::Remote(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
734 }
735 }
736
737 pub fn root_file(&self, cx: &Context<Self>) -> Option<Arc<File>> {
738 let entry = self.root_entry()?;
739 Some(File::for_entry(entry.clone(), cx.entity()))
740 }
741
742 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
743 where
744 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
745 Fut: 'static + Send + Future<Output = bool>,
746 {
747 match self {
748 Worktree::Local(this) => this.observe_updates(project_id, cx, callback),
749 Worktree::Remote(this) => this.observe_updates(project_id, cx, callback),
750 }
751 }
752
753 pub fn stop_observing_updates(&mut self) {
754 match self {
755 Worktree::Local(this) => {
756 this.update_observer.take();
757 }
758 Worktree::Remote(this) => {
759 this.update_observer.take();
760 }
761 }
762 }
763
764 pub fn wait_for_snapshot(
765 &mut self,
766 scan_id: usize,
767 ) -> impl Future<Output = Result<()>> + use<> {
768 match self {
769 Worktree::Local(this) => this.wait_for_snapshot(scan_id).boxed(),
770 Worktree::Remote(this) => this.wait_for_snapshot(scan_id).boxed(),
771 }
772 }
773
774 #[cfg(feature = "test-support")]
775 pub fn has_update_observer(&self) -> bool {
776 match self {
777 Worktree::Local(this) => this.update_observer.is_some(),
778 Worktree::Remote(this) => this.update_observer.is_some(),
779 }
780 }
781
782 pub fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
783 match self {
784 Worktree::Local(this) => this.load_file(path, cx),
785 Worktree::Remote(_) => {
786 Task::ready(Err(anyhow!("remote worktrees can't yet load files")))
787 }
788 }
789 }
790
791 pub fn load_binary_file(
792 &self,
793 path: &RelPath,
794 cx: &Context<Worktree>,
795 ) -> Task<Result<LoadedBinaryFile>> {
796 match self {
797 Worktree::Local(this) => this.load_binary_file(path, cx),
798 Worktree::Remote(_) => {
799 Task::ready(Err(anyhow!("remote worktrees can't yet load binary files")))
800 }
801 }
802 }
803
804 pub fn write_file(
805 &self,
806 path: Arc<RelPath>,
807 text: Rope,
808 line_ending: LineEnding,
809 encoding: &'static Encoding,
810 has_bom: bool,
811 cx: &Context<Worktree>,
812 ) -> Task<Result<Arc<File>>> {
813 match self {
814 Worktree::Local(this) => {
815 this.write_file(path, text, line_ending, encoding, has_bom, cx)
816 }
817 Worktree::Remote(_) => {
818 Task::ready(Err(anyhow!("remote worktree can't yet write files")))
819 }
820 }
821 }
822
823 pub fn create_entry(
824 &mut self,
825 path: Arc<RelPath>,
826 is_directory: bool,
827 content: Option<Vec<u8>>,
828 cx: &Context<Worktree>,
829 ) -> Task<Result<CreatedEntry>> {
830 let worktree_id = self.id();
831 match self {
832 Worktree::Local(this) => this.create_entry(path, is_directory, content, cx),
833 Worktree::Remote(this) => {
834 let project_id = this.project_id;
835 let request = this.client.request(proto::CreateProjectEntry {
836 worktree_id: worktree_id.to_proto(),
837 project_id,
838 path: path.as_ref().to_proto(),
839 content,
840 is_directory,
841 });
842 cx.spawn(async move |this, cx| {
843 let response = request.await?;
844 match response.entry {
845 Some(entry) => this
846 .update(cx, |worktree, cx| {
847 worktree.as_remote_mut().unwrap().insert_entry(
848 entry,
849 response.worktree_scan_id as usize,
850 cx,
851 )
852 })?
853 .await
854 .map(CreatedEntry::Included),
855 None => {
856 let abs_path =
857 this.read_with(cx, |worktree, _| worktree.absolutize(&path))?;
858 Ok(CreatedEntry::Excluded { abs_path })
859 }
860 }
861 })
862 }
863 }
864 }
865
866 pub fn delete_entry(
867 &mut self,
868 entry_id: ProjectEntryId,
869 trash: bool,
870 cx: &mut Context<Worktree>,
871 ) -> Option<Task<Result<Option<TrashedEntry>>>> {
872 let task = match self {
873 Worktree::Local(this) => this.delete_entry(entry_id, trash, cx),
874 Worktree::Remote(this) => this.delete_entry(entry_id, trash, cx),
875 }?;
876
877 let entry = match &*self {
878 Worktree::Local(this) => this.entry_for_id(entry_id),
879 Worktree::Remote(this) => this.entry_for_id(entry_id),
880 }?;
881
882 let mut ids = vec![entry_id];
883 let path = &*entry.path;
884
885 self.get_children_ids_recursive(path, &mut ids);
886
887 for id in ids {
888 cx.emit(Event::DeletedEntry(id));
889 }
890 Some(task)
891 }
892
893 pub async fn restore_entry(
894 trash_entry: TrashedEntry,
895 worktree: Entity<Self>,
896 cx: &mut AsyncApp,
897 ) -> Result<RelPathBuf> {
898 let is_local = worktree.read_with(cx, |this, _| this.is_local());
899 if is_local {
900 LocalWorktree::restore_entry(trash_entry, worktree, cx).await
901 } else {
902 // TODO(dino): Add support for restoring entries in remote worktrees.
903 Err(anyhow!("Unsupported"))
904 }
905 }
906
907 fn get_children_ids_recursive(&self, path: &RelPath, ids: &mut Vec<ProjectEntryId>) {
908 let children_iter = self.child_entries(path);
909 for child in children_iter {
910 ids.push(child.id);
911 self.get_children_ids_recursive(&child.path, ids);
912 }
913 }
914
915 // pub fn rename_entry(
916 // &mut self,
917 // entry_id: ProjectEntryId,
918 // new_path: Arc<RelPath>,
919 // cx: &Context<Self>,
920 // ) -> Task<Result<CreatedEntry>> {
921 // match self {
922 // Worktree::Local(this) => this.rename_entry(entry_id, new_path, cx),
923 // Worktree::Remote(this) => this.rename_entry(entry_id, new_path, cx),
924 // }
925 // }
926
927 pub fn copy_external_entries(
928 &mut self,
929 target_directory: Arc<RelPath>,
930 paths: Vec<Arc<Path>>,
931 fs: Arc<dyn Fs>,
932 cx: &Context<Worktree>,
933 ) -> Task<Result<Vec<ProjectEntryId>>> {
934 match self {
935 Worktree::Local(this) => this.copy_external_entries(target_directory, paths, cx),
936 Worktree::Remote(this) => this.copy_external_entries(target_directory, paths, fs, cx),
937 }
938 }
939
940 pub fn expand_entry(
941 &mut self,
942 entry_id: ProjectEntryId,
943 cx: &Context<Worktree>,
944 ) -> Option<Task<Result<()>>> {
945 match self {
946 Worktree::Local(this) => this.expand_entry(entry_id, cx),
947 Worktree::Remote(this) => {
948 let response = this.client.request(proto::ExpandProjectEntry {
949 project_id: this.project_id,
950 entry_id: entry_id.to_proto(),
951 });
952 Some(cx.spawn(async move |this, cx| {
953 let response = response.await?;
954 this.update(cx, |this, _| {
955 this.as_remote_mut()
956 .unwrap()
957 .wait_for_snapshot(response.worktree_scan_id as usize)
958 })?
959 .await?;
960 Ok(())
961 }))
962 }
963 }
964 }
965
966 pub fn expand_all_for_entry(
967 &mut self,
968 entry_id: ProjectEntryId,
969 cx: &Context<Worktree>,
970 ) -> Option<Task<Result<()>>> {
971 match self {
972 Worktree::Local(this) => this.expand_all_for_entry(entry_id, cx),
973 Worktree::Remote(this) => {
974 let response = this.client.request(proto::ExpandAllForProjectEntry {
975 project_id: this.project_id,
976 entry_id: entry_id.to_proto(),
977 });
978 Some(cx.spawn(async move |this, cx| {
979 let response = response.await?;
980 this.update(cx, |this, _| {
981 this.as_remote_mut()
982 .unwrap()
983 .wait_for_snapshot(response.worktree_scan_id as usize)
984 })?
985 .await?;
986 Ok(())
987 }))
988 }
989 }
990 }
991
992 pub async fn handle_create_entry(
993 this: Entity<Self>,
994 request: proto::CreateProjectEntry,
995 mut cx: AsyncApp,
996 ) -> Result<proto::ProjectEntryResponse> {
997 let (scan_id, entry) = this.update(&mut cx, |this, cx| {
998 anyhow::Ok((
999 this.scan_id(),
1000 this.create_entry(
1001 RelPath::from_proto(&request.path).with_context(|| {
1002 format!("received invalid relative path {:?}", request.path)
1003 })?,
1004 request.is_directory,
1005 request.content,
1006 cx,
1007 ),
1008 ))
1009 })?;
1010 Ok(proto::ProjectEntryResponse {
1011 entry: match &entry.await? {
1012 CreatedEntry::Included(entry) => Some(entry.into()),
1013 CreatedEntry::Excluded { .. } => None,
1014 },
1015 worktree_scan_id: scan_id as u64,
1016 })
1017 }
1018
1019 pub async fn handle_delete_entry(
1020 this: Entity<Self>,
1021 request: proto::DeleteProjectEntry,
1022 mut cx: AsyncApp,
1023 ) -> Result<proto::ProjectEntryResponse> {
1024 let (scan_id, task) = this.update(&mut cx, |this, cx| {
1025 (
1026 this.scan_id(),
1027 this.delete_entry(
1028 ProjectEntryId::from_proto(request.entry_id),
1029 request.use_trash,
1030 cx,
1031 ),
1032 )
1033 });
1034 task.ok_or_else(|| anyhow::anyhow!("invalid entry"))?
1035 .await?;
1036 Ok(proto::ProjectEntryResponse {
1037 entry: None,
1038 worktree_scan_id: scan_id as u64,
1039 })
1040 }
1041
1042 pub async fn handle_expand_entry(
1043 this: Entity<Self>,
1044 request: proto::ExpandProjectEntry,
1045 mut cx: AsyncApp,
1046 ) -> Result<proto::ExpandProjectEntryResponse> {
1047 let task = this.update(&mut cx, |this, cx| {
1048 this.expand_entry(ProjectEntryId::from_proto(request.entry_id), cx)
1049 });
1050 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
1051 .await?;
1052 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
1053 Ok(proto::ExpandProjectEntryResponse {
1054 worktree_scan_id: scan_id as u64,
1055 })
1056 }
1057
1058 pub async fn handle_expand_all_for_entry(
1059 this: Entity<Self>,
1060 request: proto::ExpandAllForProjectEntry,
1061 mut cx: AsyncApp,
1062 ) -> Result<proto::ExpandAllForProjectEntryResponse> {
1063 let task = this.update(&mut cx, |this, cx| {
1064 this.expand_all_for_entry(ProjectEntryId::from_proto(request.entry_id), cx)
1065 });
1066 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
1067 .await?;
1068 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
1069 Ok(proto::ExpandAllForProjectEntryResponse {
1070 worktree_scan_id: scan_id as u64,
1071 })
1072 }
1073
1074 pub fn is_single_file(&self) -> bool {
1075 self.root_dir().is_none()
1076 }
1077
1078 /// For visible worktrees, returns the path with the worktree name as the first component.
1079 /// Otherwise, returns an absolute path.
1080 pub fn full_path(&self, worktree_relative_path: &RelPath) -> PathBuf {
1081 if self.is_visible() {
1082 self.root_name()
1083 .join(worktree_relative_path)
1084 .display(self.path_style)
1085 .to_string()
1086 .into()
1087 } else {
1088 let full_path = self.abs_path();
1089 let mut full_path_string = if self.is_local()
1090 && let Ok(stripped) = full_path.strip_prefix(home_dir())
1091 {
1092 self.path_style
1093 .join("~", &*stripped.to_string_lossy())
1094 .unwrap()
1095 } else {
1096 full_path.to_string_lossy().into_owned()
1097 };
1098
1099 if worktree_relative_path.components().next().is_some() {
1100 full_path_string.push_str(self.path_style.primary_separator());
1101 full_path_string.push_str(&worktree_relative_path.display(self.path_style));
1102 }
1103
1104 full_path_string.into()
1105 }
1106 }
1107}
1108
1109impl LocalWorktree {
1110 pub fn fs(&self) -> &Arc<dyn Fs> {
1111 &self.fs
1112 }
1113
1114 pub fn is_path_private(&self, path: &RelPath) -> bool {
1115 !self.share_private_files && self.settings.is_path_private(path)
1116 }
1117
1118 pub fn fs_is_case_sensitive(&self) -> bool {
1119 self.fs_case_sensitive
1120 }
1121
1122 fn restart_background_scanners(&mut self, cx: &Context<Worktree>) {
1123 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
1124 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
1125 self.scan_requests_tx = scan_requests_tx;
1126 self.path_prefixes_to_scan_tx = path_prefixes_to_scan_tx;
1127
1128 self.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
1129 let always_included_entries = mem::take(&mut self.snapshot.always_included_entries);
1130 log::debug!(
1131 "refreshing entries for the following always included paths: {:?}",
1132 always_included_entries
1133 );
1134
1135 // Cleans up old always included entries to ensure they get updated properly. Otherwise,
1136 // nested always included entries may not get updated and will result in out-of-date info.
1137 self.refresh_entries_for_paths(always_included_entries);
1138 }
1139
1140 fn start_background_scanner(
1141 &mut self,
1142 scan_requests_rx: channel::Receiver<ScanRequest>,
1143 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
1144 cx: &Context<Worktree>,
1145 ) {
1146 let snapshot = self.snapshot();
1147 let share_private_files = self.share_private_files;
1148 let next_entry_id = self.next_entry_id.clone();
1149 let fs = self.fs.clone();
1150 let scanning_enabled = self.scanning_enabled;
1151 let settings = self.settings.clone();
1152 let (scan_states_tx, mut scan_states_rx) = mpsc::unbounded();
1153 let background_scanner = cx.background_spawn({
1154 let abs_path = snapshot.abs_path.as_path().to_path_buf();
1155 let background = cx.background_executor().clone();
1156 async move {
1157 let (events, watcher) = if scanning_enabled {
1158 fs.watch(&abs_path, FS_WATCH_LATENCY).await
1159 } else {
1160 (Box::pin(stream::pending()) as _, Arc::new(NullWatcher) as _)
1161 };
1162 let fs_case_sensitive = fs.is_case_sensitive().await;
1163
1164 let is_single_file = snapshot.snapshot.root_dir().is_none();
1165 let mut scanner = BackgroundScanner {
1166 fs,
1167 fs_case_sensitive,
1168 status_updates_tx: scan_states_tx,
1169 executor: background,
1170 scan_requests_rx,
1171 path_prefixes_to_scan_rx,
1172 next_entry_id,
1173 state: async_lock::Mutex::new(BackgroundScannerState {
1174 prev_snapshot: snapshot.snapshot.clone(),
1175 snapshot,
1176 symlink_paths_by_target: Default::default(),
1177 scanned_dirs: Default::default(),
1178 watched_dir_abs_paths_by_entry_id: Default::default(),
1179 scanning_enabled,
1180 path_prefixes_to_scan: Default::default(),
1181 paths_to_scan: Default::default(),
1182 removed_entries: Default::default(),
1183 changed_paths: Default::default(),
1184 }),
1185 phase: BackgroundScannerPhase::InitialScan,
1186 share_private_files,
1187 settings,
1188 watcher,
1189 is_single_file,
1190 };
1191
1192 scanner
1193 .run(Box::pin(events.map(|events| events.into_iter().collect())))
1194 .await;
1195 }
1196 });
1197 let scan_state_updater = cx.spawn(async move |this, cx| {
1198 while let Some((state, this)) = scan_states_rx.next().await.zip(this.upgrade()) {
1199 this.update(cx, |this, cx| {
1200 let this = this.as_local_mut().unwrap();
1201 match state {
1202 ScanState::Started => {
1203 *this.is_scanning.0.borrow_mut() = true;
1204 }
1205 ScanState::Updated {
1206 snapshot,
1207 changes,
1208 barrier,
1209 scanning,
1210 } => {
1211 *this.is_scanning.0.borrow_mut() = scanning;
1212 this.set_snapshot(snapshot, changes, cx);
1213 drop(barrier);
1214 }
1215 ScanState::RootUpdated { new_path } => {
1216 this.update_abs_path_and_refresh(new_path, cx);
1217 }
1218 ScanState::RootDeleted => {
1219 log::info!(
1220 "worktree root {} no longer exists, closing worktree",
1221 this.abs_path().display()
1222 );
1223 cx.emit(Event::Deleted);
1224 }
1225 }
1226 });
1227 }
1228 });
1229 self._background_scanner_tasks = vec![background_scanner, scan_state_updater];
1230 *self.is_scanning.0.borrow_mut() = true;
1231 }
1232
1233 fn set_snapshot(
1234 &mut self,
1235 mut new_snapshot: LocalSnapshot,
1236 entry_changes: UpdatedEntriesSet,
1237 cx: &mut Context<Worktree>,
1238 ) {
1239 let repo_changes = self.changed_repos(&self.snapshot, &mut new_snapshot);
1240
1241 new_snapshot.root_repo_common_dir = new_snapshot
1242 .local_repo_for_work_directory_path(RelPath::empty())
1243 .map(|repo| SanitizedPath::from_arc(repo.common_dir_abs_path.clone()));
1244
1245 let old_root_repo_common_dir = (self.snapshot.root_repo_common_dir
1246 != new_snapshot.root_repo_common_dir)
1247 .then(|| self.snapshot.root_repo_common_dir.clone());
1248 self.snapshot = new_snapshot;
1249
1250 if let Some(share) = self.update_observer.as_mut() {
1251 share
1252 .snapshots_tx
1253 .unbounded_send((self.snapshot.clone(), entry_changes.clone()))
1254 .ok();
1255 }
1256
1257 if !entry_changes.is_empty() {
1258 cx.emit(Event::UpdatedEntries(entry_changes));
1259 }
1260 if !repo_changes.is_empty() {
1261 cx.emit(Event::UpdatedGitRepositories(repo_changes));
1262 }
1263 if let Some(old) = old_root_repo_common_dir {
1264 cx.emit(Event::UpdatedRootRepoCommonDir { old });
1265 }
1266
1267 while let Some((scan_id, _)) = self.snapshot_subscriptions.front() {
1268 if self.snapshot.completed_scan_id >= *scan_id {
1269 let (_, tx) = self.snapshot_subscriptions.pop_front().unwrap();
1270 tx.send(()).ok();
1271 } else {
1272 break;
1273 }
1274 }
1275 }
1276
1277 fn changed_repos(
1278 &self,
1279 old_snapshot: &LocalSnapshot,
1280 new_snapshot: &mut LocalSnapshot,
1281 ) -> UpdatedGitRepositoriesSet {
1282 let mut changes = Vec::new();
1283 let mut old_repos = old_snapshot.git_repositories.iter().peekable();
1284 let new_repos = new_snapshot.git_repositories.clone();
1285 let mut new_repos = new_repos.iter().peekable();
1286
1287 loop {
1288 match (new_repos.peek().map(clone), old_repos.peek().map(clone)) {
1289 (Some((new_entry_id, new_repo)), Some((old_entry_id, old_repo))) => {
1290 match Ord::cmp(&new_entry_id, &old_entry_id) {
1291 Ordering::Less => {
1292 changes.push(UpdatedGitRepository {
1293 work_directory_id: new_entry_id,
1294 old_work_directory_abs_path: None,
1295 new_work_directory_abs_path: Some(
1296 new_repo.work_directory_abs_path.clone(),
1297 ),
1298 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1299 repository_dir_abs_path: Some(
1300 new_repo.repository_dir_abs_path.clone(),
1301 ),
1302 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1303 });
1304 new_repos.next();
1305 }
1306 Ordering::Equal => {
1307 if new_repo.git_dir_scan_id != old_repo.git_dir_scan_id
1308 || new_repo.work_directory_abs_path
1309 != old_repo.work_directory_abs_path
1310 {
1311 changes.push(UpdatedGitRepository {
1312 work_directory_id: new_entry_id,
1313 old_work_directory_abs_path: Some(
1314 old_repo.work_directory_abs_path.clone(),
1315 ),
1316 new_work_directory_abs_path: Some(
1317 new_repo.work_directory_abs_path.clone(),
1318 ),
1319 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1320 repository_dir_abs_path: Some(
1321 new_repo.repository_dir_abs_path.clone(),
1322 ),
1323 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1324 });
1325 }
1326 new_repos.next();
1327 old_repos.next();
1328 }
1329 Ordering::Greater => {
1330 changes.push(UpdatedGitRepository {
1331 work_directory_id: old_entry_id,
1332 old_work_directory_abs_path: Some(
1333 old_repo.work_directory_abs_path.clone(),
1334 ),
1335 new_work_directory_abs_path: None,
1336 dot_git_abs_path: None,
1337 repository_dir_abs_path: None,
1338 common_dir_abs_path: None,
1339 });
1340 old_repos.next();
1341 }
1342 }
1343 }
1344 (Some((entry_id, repo)), None) => {
1345 changes.push(UpdatedGitRepository {
1346 work_directory_id: entry_id,
1347 old_work_directory_abs_path: None,
1348 new_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1349 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1350 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1351 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1352 });
1353 new_repos.next();
1354 }
1355 (None, Some((entry_id, repo))) => {
1356 changes.push(UpdatedGitRepository {
1357 work_directory_id: entry_id,
1358 old_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1359 new_work_directory_abs_path: None,
1360 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1361 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1362 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1363 });
1364 old_repos.next();
1365 }
1366 (None, None) => break,
1367 }
1368 }
1369
1370 fn clone<T: Clone, U: Clone>(value: &(&T, &U)) -> (T, U) {
1371 (value.0.clone(), value.1.clone())
1372 }
1373
1374 changes.into()
1375 }
1376
1377 pub fn scan_complete(&self) -> impl Future<Output = ()> + use<> {
1378 let mut is_scanning_rx = self.is_scanning.1.clone();
1379 async move {
1380 let mut is_scanning = *is_scanning_rx.borrow();
1381 while is_scanning {
1382 if let Some(value) = is_scanning_rx.recv().await {
1383 is_scanning = value;
1384 } else {
1385 break;
1386 }
1387 }
1388 }
1389 }
1390
1391 pub fn wait_for_snapshot(
1392 &mut self,
1393 scan_id: usize,
1394 ) -> impl Future<Output = Result<()>> + use<> {
1395 let (tx, rx) = oneshot::channel();
1396 if self.snapshot.completed_scan_id >= scan_id {
1397 tx.send(()).ok();
1398 } else {
1399 match self
1400 .snapshot_subscriptions
1401 .binary_search_by_key(&scan_id, |probe| probe.0)
1402 {
1403 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
1404 }
1405 }
1406
1407 async move {
1408 rx.await?;
1409 Ok(())
1410 }
1411 }
1412
1413 pub fn snapshot(&self) -> LocalSnapshot {
1414 self.snapshot.clone()
1415 }
1416
1417 pub fn settings(&self) -> WorktreeSettings {
1418 self.settings.clone()
1419 }
1420
1421 fn load_binary_file(
1422 &self,
1423 path: &RelPath,
1424 cx: &Context<Worktree>,
1425 ) -> Task<Result<LoadedBinaryFile>> {
1426 let path = Arc::from(path);
1427 let abs_path = self.absolutize(&path);
1428 let fs = self.fs.clone();
1429 let entry = self.refresh_entry(path.clone(), None, cx);
1430 let is_private = self.is_path_private(&path);
1431
1432 let worktree = cx.weak_entity();
1433 cx.background_spawn(async move {
1434 let content = fs.load_bytes(&abs_path).await?;
1435
1436 let worktree = worktree.upgrade().context("worktree was dropped")?;
1437 let file = match entry.await? {
1438 Some(entry) => File::for_entry(entry, worktree),
1439 None => {
1440 let metadata = fs
1441 .metadata(&abs_path)
1442 .await
1443 .with_context(|| {
1444 format!("Loading metadata for excluded file {abs_path:?}")
1445 })?
1446 .with_context(|| {
1447 format!("Excluded file {abs_path:?} got removed during loading")
1448 })?;
1449 Arc::new(File {
1450 entry_id: None,
1451 worktree,
1452 path,
1453 disk_state: DiskState::Present {
1454 mtime: metadata.mtime,
1455 size: metadata.len,
1456 },
1457 is_local: true,
1458 is_private,
1459 })
1460 }
1461 };
1462
1463 Ok(LoadedBinaryFile { file, content })
1464 })
1465 }
1466
1467 #[ztracing::instrument(skip_all)]
1468 fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
1469 let path = Arc::from(path);
1470 let abs_path = self.absolutize(&path);
1471 let fs = self.fs.clone();
1472 let entry = self.refresh_entry(path.clone(), None, cx);
1473 let is_private = self.is_path_private(path.as_ref());
1474
1475 let this = cx.weak_entity();
1476 cx.background_spawn(async move {
1477 // WARN: Temporary workaround for #27283.
1478 // We are not efficient with our memory usage per file, and use in excess of 64GB for a 10GB file
1479 // Therefore, as a temporary workaround to prevent system freezes, we just bail before opening a file
1480 // if it is too large
1481 // 5GB seems to be more reasonable, peaking at ~16GB, while 6GB jumps up to >24GB which seems like a
1482 // reasonable limit
1483 {
1484 const FILE_SIZE_MAX: u64 = 6 * 1024 * 1024 * 1024; // 6GB
1485 if let Ok(Some(metadata)) = fs.metadata(&abs_path).await
1486 && metadata.len >= FILE_SIZE_MAX
1487 {
1488 anyhow::bail!("File is too large to load");
1489 }
1490 }
1491 let (text, encoding, has_bom) = decode_file_text(fs.as_ref(), &abs_path).await?;
1492
1493 let worktree = this.upgrade().context("worktree was dropped")?;
1494 let file = match entry.await? {
1495 Some(entry) => File::for_entry(entry, worktree),
1496 None => {
1497 let metadata = fs
1498 .metadata(&abs_path)
1499 .await
1500 .with_context(|| {
1501 format!("Loading metadata for excluded file {abs_path:?}")
1502 })?
1503 .with_context(|| {
1504 format!("Excluded file {abs_path:?} got removed during loading")
1505 })?;
1506 Arc::new(File {
1507 entry_id: None,
1508 worktree,
1509 path,
1510 disk_state: DiskState::Present {
1511 mtime: metadata.mtime,
1512 size: metadata.len,
1513 },
1514 is_local: true,
1515 is_private,
1516 })
1517 }
1518 };
1519
1520 Ok(LoadedFile {
1521 file,
1522 text,
1523 encoding,
1524 has_bom,
1525 })
1526 })
1527 }
1528
1529 /// Find the lowest path in the worktree's datastructures that is an ancestor
1530 fn lowest_ancestor(&self, path: &RelPath) -> Arc<RelPath> {
1531 let mut lowest_ancestor = None;
1532 for path in path.ancestors() {
1533 if self.entry_for_path(path).is_some() {
1534 lowest_ancestor = Some(path.into());
1535 break;
1536 }
1537 }
1538
1539 lowest_ancestor.unwrap_or_else(|| RelPath::empty().into())
1540 }
1541
1542 pub fn create_entry(
1543 &self,
1544 path: Arc<RelPath>,
1545 is_dir: bool,
1546 content: Option<Vec<u8>>,
1547 cx: &Context<Worktree>,
1548 ) -> Task<Result<CreatedEntry>> {
1549 let abs_path = self.absolutize(&path);
1550 let path_excluded = self.settings.is_path_excluded(&path);
1551 let fs = self.fs.clone();
1552 let task_abs_path = abs_path.clone();
1553 let write = cx.background_spawn(async move {
1554 if is_dir {
1555 fs.create_dir(&task_abs_path)
1556 .await
1557 .with_context(|| format!("creating directory {task_abs_path:?}"))
1558 } else {
1559 fs.write(&task_abs_path, content.as_deref().unwrap_or(&[]))
1560 .await
1561 .with_context(|| format!("creating file {task_abs_path:?}"))
1562 }
1563 });
1564
1565 let lowest_ancestor = self.lowest_ancestor(&path);
1566 cx.spawn(async move |this, cx| {
1567 write.await?;
1568 if path_excluded {
1569 return Ok(CreatedEntry::Excluded { abs_path });
1570 }
1571
1572 let (result, refreshes) = this.update(cx, |this, cx| {
1573 let mut refreshes = Vec::new();
1574 let refresh_paths = path.strip_prefix(&lowest_ancestor).unwrap();
1575 for refresh_path in refresh_paths.ancestors() {
1576 if refresh_path == RelPath::empty() {
1577 continue;
1578 }
1579 let refresh_full_path = lowest_ancestor.join(refresh_path);
1580
1581 refreshes.push(this.as_local_mut().unwrap().refresh_entry(
1582 refresh_full_path,
1583 None,
1584 cx,
1585 ));
1586 }
1587 (
1588 this.as_local_mut().unwrap().refresh_entry(path, None, cx),
1589 refreshes,
1590 )
1591 })?;
1592 for refresh in refreshes {
1593 refresh.await.log_err();
1594 }
1595
1596 Ok(result
1597 .await?
1598 .map(CreatedEntry::Included)
1599 .unwrap_or_else(|| CreatedEntry::Excluded { abs_path }))
1600 })
1601 }
1602
1603 pub fn write_file(
1604 &self,
1605 path: Arc<RelPath>,
1606 text: Rope,
1607 line_ending: LineEnding,
1608 encoding: &'static Encoding,
1609 has_bom: bool,
1610 cx: &Context<Worktree>,
1611 ) -> Task<Result<Arc<File>>> {
1612 let fs = self.fs.clone();
1613 let is_private = self.is_path_private(&path);
1614 let abs_path = self.absolutize(&path);
1615
1616 let write = cx.background_spawn({
1617 let fs = fs.clone();
1618 let abs_path = abs_path.clone();
1619 async move {
1620 // For UTF-8, use the optimized `fs.save` which writes Rope chunks directly to disk
1621 // without allocating a contiguous string.
1622 if encoding == encoding_rs::UTF_8 && !has_bom {
1623 return fs.save(&abs_path, &text, line_ending).await;
1624 }
1625
1626 // For legacy encodings (e.g. Shift-JIS), we fall back to converting the entire Rope
1627 // to a String/Bytes in memory before writing.
1628 //
1629 // Note: This is inefficient for very large files compared to the streaming approach above,
1630 // but supporting streaming writes for arbitrary encodings would require a significant
1631 // refactor of the `fs` crate to expose a Writer interface.
1632 let text_string = text.to_string();
1633 let normalized_text = match line_ending {
1634 LineEnding::Unix => text_string,
1635 LineEnding::Windows => text_string.replace('\n', "\r\n"),
1636 };
1637
1638 // Create the byte vector manually for UTF-16 encodings because encoding_rs encodes to UTF-8 by default (per WHATWG standards),
1639 // which is not what we want for saving files.
1640 let bytes = if encoding == encoding_rs::UTF_16BE {
1641 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1642 if has_bom {
1643 data.extend_from_slice(&[0xFE, 0xFF]); // BOM
1644 }
1645 let utf16be_bytes =
1646 normalized_text.encode_utf16().flat_map(|u| u.to_be_bytes());
1647 data.extend(utf16be_bytes);
1648 data.into()
1649 } else if encoding == encoding_rs::UTF_16LE {
1650 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1651 if has_bom {
1652 data.extend_from_slice(&[0xFF, 0xFE]); // BOM
1653 }
1654 let utf16le_bytes =
1655 normalized_text.encode_utf16().flat_map(|u| u.to_le_bytes());
1656 data.extend(utf16le_bytes);
1657 data.into()
1658 } else {
1659 // For other encodings (Shift-JIS, UTF-8 with BOM, etc.), delegate to encoding_rs.
1660 let bom_bytes = if has_bom {
1661 if encoding == encoding_rs::UTF_8 {
1662 vec![0xEF, 0xBB, 0xBF]
1663 } else {
1664 vec![]
1665 }
1666 } else {
1667 vec![]
1668 };
1669 let (cow, _, _) = encoding.encode(&normalized_text);
1670 if !bom_bytes.is_empty() {
1671 let mut bytes = bom_bytes;
1672 bytes.extend_from_slice(&cow);
1673 bytes.into()
1674 } else {
1675 cow
1676 }
1677 };
1678
1679 fs.write(&abs_path, &bytes).await
1680 }
1681 });
1682
1683 cx.spawn(async move |this, cx| {
1684 write.await?;
1685 let entry = this
1686 .update(cx, |this, cx| {
1687 this.as_local_mut()
1688 .unwrap()
1689 .refresh_entry(path.clone(), None, cx)
1690 })?
1691 .await?;
1692 let worktree = this.upgrade().context("worktree dropped")?;
1693 if let Some(entry) = entry {
1694 Ok(File::for_entry(entry, worktree))
1695 } else {
1696 let metadata = fs
1697 .metadata(&abs_path)
1698 .await
1699 .with_context(|| {
1700 format!("Fetching metadata after saving the excluded buffer {abs_path:?}")
1701 })?
1702 .with_context(|| {
1703 format!("Excluded buffer {path:?} got removed during saving")
1704 })?;
1705 Ok(Arc::new(File {
1706 worktree,
1707 path,
1708 disk_state: DiskState::Present {
1709 mtime: metadata.mtime,
1710 size: metadata.len,
1711 },
1712 entry_id: None,
1713 is_local: true,
1714 is_private,
1715 }))
1716 }
1717 })
1718 }
1719
1720 pub fn delete_entry(
1721 &self,
1722 entry_id: ProjectEntryId,
1723 trash: bool,
1724 cx: &Context<Worktree>,
1725 ) -> Option<Task<Result<Option<TrashedEntry>>>> {
1726 let entry = self.entry_for_id(entry_id)?.clone();
1727 let abs_path = self.absolutize(&entry.path);
1728 let fs = self.fs.clone();
1729
1730 let delete = cx.background_spawn(async move {
1731 let trashed_entry = match (entry.is_file(), trash) {
1732 (true, true) => Some(fs.trash(&abs_path, Default::default()).await?),
1733 (false, true) => Some(
1734 fs.trash(
1735 &abs_path,
1736 RemoveOptions {
1737 recursive: true,
1738 ignore_if_not_exists: false,
1739 },
1740 )
1741 .await?,
1742 ),
1743 (true, false) => {
1744 fs.remove_file(&abs_path, Default::default()).await?;
1745 None
1746 }
1747 (false, false) => {
1748 fs.remove_dir(
1749 &abs_path,
1750 RemoveOptions {
1751 recursive: true,
1752 ignore_if_not_exists: false,
1753 },
1754 )
1755 .await?;
1756 None
1757 }
1758 };
1759
1760 anyhow::Ok((trashed_entry, entry.path))
1761 });
1762
1763 Some(cx.spawn(async move |this, cx| {
1764 let (trashed_entry, path) = delete.await?;
1765 this.update(cx, |this, _| {
1766 this.as_local_mut()
1767 .unwrap()
1768 .refresh_entries_for_paths(vec![path])
1769 })?
1770 .recv()
1771 .await;
1772
1773 Ok(trashed_entry)
1774 }))
1775 }
1776
1777 pub async fn restore_entry(
1778 trash_entry: TrashedEntry,
1779 this: Entity<Worktree>,
1780 cx: &mut AsyncApp,
1781 ) -> Result<RelPathBuf> {
1782 let Some((fs, worktree_abs_path, path_style)) = this.read_with(cx, |this, _cx| {
1783 let local_worktree = match this {
1784 Worktree::Local(local_worktree) => local_worktree,
1785 Worktree::Remote(_) => return None,
1786 };
1787
1788 let fs = local_worktree.fs.clone();
1789 let path_style = local_worktree.path_style();
1790 Some((fs, Arc::clone(local_worktree.abs_path()), path_style))
1791 }) else {
1792 return Err(anyhow!("Localworktree should not change into a remote one"));
1793 };
1794
1795 let path_buf = fs.restore(trash_entry).await?;
1796 let path = path_buf
1797 .strip_prefix(worktree_abs_path)
1798 .context("Could not strip prefix")?;
1799 let path = RelPath::new(&path, path_style)?;
1800 let path = path.into_owned();
1801
1802 Ok(path)
1803 }
1804
1805 pub fn copy_external_entries(
1806 &self,
1807 target_directory: Arc<RelPath>,
1808 paths: Vec<Arc<Path>>,
1809 cx: &Context<Worktree>,
1810 ) -> Task<Result<Vec<ProjectEntryId>>> {
1811 let target_directory = self.absolutize(&target_directory);
1812 let worktree_path = self.abs_path().clone();
1813 let fs = self.fs.clone();
1814 let paths = paths
1815 .into_iter()
1816 .filter_map(|source| {
1817 let file_name = source.file_name()?;
1818 let mut target = target_directory.clone();
1819 target.push(file_name);
1820
1821 // Do not allow copying the same file to itself.
1822 if source.as_ref() != target.as_path() {
1823 Some((source, target))
1824 } else {
1825 None
1826 }
1827 })
1828 .collect::<Vec<_>>();
1829
1830 let paths_to_refresh = paths
1831 .iter()
1832 .filter_map(|(_, target)| {
1833 RelPath::new(
1834 target.strip_prefix(&worktree_path).ok()?,
1835 PathStyle::local(),
1836 )
1837 .ok()
1838 .map(|path| path.into_arc())
1839 })
1840 .collect::<Vec<_>>();
1841
1842 cx.spawn(async move |this, cx| {
1843 cx.background_spawn(async move {
1844 for (source, target) in paths {
1845 copy_recursive(
1846 fs.as_ref(),
1847 &source,
1848 &target,
1849 fs::CopyOptions {
1850 overwrite: true,
1851 ..Default::default()
1852 },
1853 )
1854 .await
1855 .with_context(|| {
1856 format!("Failed to copy file from {source:?} to {target:?}")
1857 })?;
1858 }
1859 anyhow::Ok(())
1860 })
1861 .await
1862 .log_err();
1863 let mut refresh = cx.read_entity(
1864 &this.upgrade().with_context(|| "Dropped worktree")?,
1865 |this, _| {
1866 anyhow::Ok::<postage::barrier::Receiver>(
1867 this.as_local()
1868 .with_context(|| "Worktree is not local")?
1869 .refresh_entries_for_paths(paths_to_refresh.clone()),
1870 )
1871 },
1872 )?;
1873
1874 cx.background_spawn(async move {
1875 refresh.next().await;
1876 anyhow::Ok(())
1877 })
1878 .await
1879 .log_err();
1880
1881 let this = this.upgrade().with_context(|| "Dropped worktree")?;
1882 Ok(cx.read_entity(&this, |this, _| {
1883 paths_to_refresh
1884 .iter()
1885 .filter_map(|path| Some(this.entry_for_path(path)?.id))
1886 .collect()
1887 }))
1888 })
1889 }
1890
1891 fn expand_entry(
1892 &self,
1893 entry_id: ProjectEntryId,
1894 cx: &Context<Worktree>,
1895 ) -> Option<Task<Result<()>>> {
1896 let path = self.entry_for_id(entry_id)?.path.clone();
1897 let mut refresh = self.refresh_entries_for_paths(vec![path]);
1898 Some(cx.background_spawn(async move {
1899 refresh.next().await;
1900 Ok(())
1901 }))
1902 }
1903
1904 fn expand_all_for_entry(
1905 &self,
1906 entry_id: ProjectEntryId,
1907 cx: &Context<Worktree>,
1908 ) -> Option<Task<Result<()>>> {
1909 let path = self.entry_for_id(entry_id).unwrap().path.clone();
1910 let mut rx = self.add_path_prefix_to_scan(path);
1911 Some(cx.background_spawn(async move {
1912 rx.next().await;
1913 Ok(())
1914 }))
1915 }
1916
1917 pub fn refresh_entries_for_paths(&self, paths: Vec<Arc<RelPath>>) -> barrier::Receiver {
1918 let (tx, rx) = barrier::channel();
1919 self.scan_requests_tx
1920 .try_send(ScanRequest {
1921 relative_paths: paths,
1922 done: smallvec![tx],
1923 })
1924 .ok();
1925 rx
1926 }
1927
1928 #[cfg(feature = "test-support")]
1929 pub fn manually_refresh_entries_for_paths(
1930 &self,
1931 paths: Vec<Arc<RelPath>>,
1932 ) -> barrier::Receiver {
1933 self.refresh_entries_for_paths(paths)
1934 }
1935
1936 pub fn add_path_prefix_to_scan(&self, path_prefix: Arc<RelPath>) -> barrier::Receiver {
1937 let (tx, rx) = barrier::channel();
1938 self.path_prefixes_to_scan_tx
1939 .try_send(PathPrefixScanRequest {
1940 path: path_prefix,
1941 done: smallvec![tx],
1942 })
1943 .ok();
1944 rx
1945 }
1946
1947 pub fn refresh_entry(
1948 &self,
1949 path: Arc<RelPath>,
1950 old_path: Option<Arc<RelPath>>,
1951 cx: &Context<Worktree>,
1952 ) -> Task<Result<Option<Entry>>> {
1953 if self.settings.is_path_excluded(&path) {
1954 return Task::ready(Ok(None));
1955 }
1956 let paths = if let Some(old_path) = old_path.as_ref() {
1957 vec![old_path.clone(), path.clone()]
1958 } else {
1959 vec![path.clone()]
1960 };
1961 let t0 = Instant::now();
1962 let mut refresh = self.refresh_entries_for_paths(paths);
1963 // todo(lw): Hot foreground spawn
1964 cx.spawn(async move |this, cx| {
1965 refresh.recv().await;
1966 log::trace!("refreshed entry {path:?} in {:?}", t0.elapsed());
1967 let new_entry = this.read_with(cx, |this, _| {
1968 this.entry_for_path(&path).cloned().with_context(|| {
1969 format!("Could not find entry in worktree for {path:?} after refresh")
1970 })
1971 })??;
1972 Ok(Some(new_entry))
1973 })
1974 }
1975
1976 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1977 where
1978 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1979 Fut: 'static + Send + Future<Output = bool>,
1980 {
1981 if let Some(observer) = self.update_observer.as_mut() {
1982 *observer.resume_updates.borrow_mut() = ();
1983 return;
1984 }
1985
1986 let (resume_updates_tx, mut resume_updates_rx) = watch::channel::<()>();
1987 let (snapshots_tx, mut snapshots_rx) =
1988 mpsc::unbounded::<(LocalSnapshot, UpdatedEntriesSet)>();
1989 snapshots_tx
1990 .unbounded_send((self.snapshot(), Arc::default()))
1991 .ok();
1992
1993 let worktree_id = self.id.to_proto();
1994 let _maintain_remote_snapshot = cx.background_spawn(async move {
1995 let mut is_first = true;
1996 while let Some((snapshot, entry_changes)) = snapshots_rx.next().await {
1997 let update = if is_first {
1998 is_first = false;
1999 snapshot.build_initial_update(project_id, worktree_id)
2000 } else {
2001 snapshot.build_update(project_id, worktree_id, entry_changes)
2002 };
2003
2004 for update in proto::split_worktree_update(update) {
2005 let _ = resume_updates_rx.try_recv();
2006 loop {
2007 let result = callback(update.clone());
2008 if result.await {
2009 break;
2010 } else {
2011 log::info!("waiting to resume updates");
2012 if resume_updates_rx.next().await.is_none() {
2013 return Some(());
2014 }
2015 }
2016 }
2017 }
2018 }
2019 Some(())
2020 });
2021
2022 self.update_observer = Some(UpdateObservationState {
2023 snapshots_tx,
2024 resume_updates: resume_updates_tx,
2025 _maintain_remote_snapshot,
2026 });
2027 }
2028
2029 pub fn share_private_files(&mut self, cx: &Context<Worktree>) {
2030 self.share_private_files = true;
2031 self.restart_background_scanners(cx);
2032 }
2033
2034 pub fn update_abs_path_and_refresh(
2035 &mut self,
2036 new_path: Arc<SanitizedPath>,
2037 cx: &Context<Worktree>,
2038 ) {
2039 self.snapshot.git_repositories = Default::default();
2040 self.snapshot.ignores_by_parent_abs_path = Default::default();
2041 let root_name = new_path
2042 .as_path()
2043 .file_name()
2044 .and_then(|f| f.to_str())
2045 .map_or(RelPath::empty().into(), |f| {
2046 RelPath::unix(f).unwrap().into()
2047 });
2048 self.snapshot.update_abs_path(new_path, root_name);
2049 self.restart_background_scanners(cx);
2050 }
2051 #[cfg(feature = "test-support")]
2052 pub fn repositories(&self) -> Vec<Arc<Path>> {
2053 self.git_repositories
2054 .values()
2055 .map(|entry| entry.work_directory_abs_path.clone())
2056 .collect::<Vec<_>>()
2057 }
2058}
2059
2060impl RemoteWorktree {
2061 pub fn project_id(&self) -> u64 {
2062 self.project_id
2063 }
2064
2065 pub fn client(&self) -> AnyProtoClient {
2066 self.client.clone()
2067 }
2068
2069 pub fn disconnected_from_host(&mut self) {
2070 self.updates_tx.take();
2071 self.snapshot_subscriptions.clear();
2072 self.disconnected = true;
2073 }
2074
2075 pub fn update_from_remote(&self, update: proto::UpdateWorktree) {
2076 if let Some(updates_tx) = &self.updates_tx {
2077 updates_tx
2078 .unbounded_send(update)
2079 .expect("consumer runs to completion");
2080 }
2081 }
2082
2083 fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
2084 where
2085 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
2086 Fut: 'static + Send + Future<Output = bool>,
2087 {
2088 let (tx, mut rx) = mpsc::unbounded();
2089 let initial_update = self
2090 .snapshot
2091 .build_initial_update(project_id, self.id().to_proto());
2092 self.update_observer = Some(tx);
2093 cx.spawn(async move |this, cx| {
2094 let mut update = initial_update;
2095 'outer: loop {
2096 // SSH projects use a special project ID of 0, and we need to
2097 // remap it to the correct one here.
2098 update.project_id = project_id;
2099
2100 for chunk in split_worktree_update(update) {
2101 if !callback(chunk).await {
2102 break 'outer;
2103 }
2104 }
2105
2106 if let Some(next_update) = rx.next().await {
2107 update = next_update;
2108 } else {
2109 break;
2110 }
2111 }
2112 this.update(cx, |this, _| {
2113 let this = this.as_remote_mut().unwrap();
2114 this.update_observer.take();
2115 })
2116 })
2117 .detach();
2118 }
2119
2120 fn observed_snapshot(&self, scan_id: usize) -> bool {
2121 self.completed_scan_id >= scan_id
2122 }
2123
2124 pub fn wait_for_snapshot(
2125 &mut self,
2126 scan_id: usize,
2127 ) -> impl Future<Output = Result<()>> + use<> {
2128 let (tx, rx) = oneshot::channel();
2129 if self.observed_snapshot(scan_id) {
2130 let _ = tx.send(());
2131 } else if self.disconnected {
2132 drop(tx);
2133 } else {
2134 match self
2135 .snapshot_subscriptions
2136 .binary_search_by_key(&scan_id, |probe| probe.0)
2137 {
2138 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
2139 }
2140 }
2141
2142 async move {
2143 rx.await?;
2144 Ok(())
2145 }
2146 }
2147
2148 pub fn insert_entry(
2149 &mut self,
2150 entry: proto::Entry,
2151 scan_id: usize,
2152 cx: &Context<Worktree>,
2153 ) -> Task<Result<Entry>> {
2154 let wait_for_snapshot = self.wait_for_snapshot(scan_id);
2155 cx.spawn(async move |this, cx| {
2156 wait_for_snapshot.await?;
2157 this.update(cx, |worktree, _| {
2158 let worktree = worktree.as_remote_mut().unwrap();
2159 let snapshot = &mut worktree.background_snapshot.lock().0;
2160 let entry = snapshot.insert_entry(entry, &worktree.file_scan_inclusions);
2161 worktree.snapshot = snapshot.clone();
2162 entry
2163 })?
2164 })
2165 }
2166
2167 fn delete_entry(
2168 &self,
2169 entry_id: ProjectEntryId,
2170 trash: bool,
2171 cx: &Context<Worktree>,
2172 ) -> Option<Task<Result<Option<TrashedEntry>>>> {
2173 let response = self.client.request(proto::DeleteProjectEntry {
2174 project_id: self.project_id,
2175 entry_id: entry_id.to_proto(),
2176 use_trash: trash,
2177 });
2178 Some(cx.spawn(async move |this, cx| {
2179 let response = response.await?;
2180 let scan_id = response.worktree_scan_id as usize;
2181
2182 this.update(cx, move |this, _| {
2183 this.as_remote_mut().unwrap().wait_for_snapshot(scan_id)
2184 })?
2185 .await?;
2186
2187 this.update(cx, |this, _| {
2188 let this = this.as_remote_mut().unwrap();
2189 let snapshot = &mut this.background_snapshot.lock().0;
2190 snapshot.delete_entry(entry_id);
2191 this.snapshot = snapshot.clone();
2192
2193 // TODO: How can we actually track the deleted entry when
2194 // working in remote? We likely only need to keep this
2195 // information on the remote side in order to support restoring
2196 // the trashed file.
2197 None
2198 })
2199 }))
2200 }
2201
2202 // fn rename_entry(
2203 // &self,
2204 // entry_id: ProjectEntryId,
2205 // new_path: impl Into<Arc<RelPath>>,
2206 // cx: &Context<Worktree>,
2207 // ) -> Task<Result<CreatedEntry>> {
2208 // let new_path: Arc<RelPath> = new_path.into();
2209 // let response = self.client.request(proto::RenameProjectEntry {
2210 // project_id: self.project_id,
2211 // entry_id: entry_id.to_proto(),
2212 // new_worktree_id: new_path.worktree_id,
2213 // new_path: new_path.as_ref().to_proto(),
2214 // });
2215 // cx.spawn(async move |this, cx| {
2216 // let response = response.await?;
2217 // match response.entry {
2218 // Some(entry) => this
2219 // .update(cx, |this, cx| {
2220 // this.as_remote_mut().unwrap().insert_entry(
2221 // entry,
2222 // response.worktree_scan_id as usize,
2223 // cx,
2224 // )
2225 // })?
2226 // .await
2227 // .map(CreatedEntry::Included),
2228 // None => {
2229 // let abs_path =
2230 // this.read_with(cx, |worktree, _| worktree.absolutize(&new_path))?;
2231 // Ok(CreatedEntry::Excluded { abs_path })
2232 // }
2233 // }
2234 // })
2235 // }
2236
2237 fn copy_external_entries(
2238 &self,
2239 target_directory: Arc<RelPath>,
2240 paths_to_copy: Vec<Arc<Path>>,
2241 local_fs: Arc<dyn Fs>,
2242 cx: &Context<Worktree>,
2243 ) -> Task<anyhow::Result<Vec<ProjectEntryId>>> {
2244 let client = self.client.clone();
2245 let worktree_id = self.id().to_proto();
2246 let project_id = self.project_id;
2247
2248 cx.background_spawn(async move {
2249 let mut requests = Vec::new();
2250 for root_path_to_copy in paths_to_copy {
2251 let Some(filename) = root_path_to_copy
2252 .file_name()
2253 .and_then(|name| name.to_str())
2254 .and_then(|filename| RelPath::unix(filename).ok())
2255 else {
2256 continue;
2257 };
2258 for (abs_path, is_directory) in
2259 read_dir_items(local_fs.as_ref(), &root_path_to_copy).await?
2260 {
2261 let Some(relative_path) = abs_path
2262 .strip_prefix(&root_path_to_copy)
2263 .map_err(|e| anyhow::Error::from(e))
2264 .and_then(|relative_path| RelPath::new(relative_path, PathStyle::local()))
2265 .log_err()
2266 else {
2267 continue;
2268 };
2269 let content = if is_directory {
2270 None
2271 } else {
2272 Some(local_fs.load_bytes(&abs_path).await?)
2273 };
2274
2275 let mut target_path = target_directory.join(filename);
2276 if relative_path.file_name().is_some() {
2277 target_path = target_path.join(&relative_path);
2278 }
2279
2280 requests.push(proto::CreateProjectEntry {
2281 project_id,
2282 worktree_id,
2283 path: target_path.to_proto(),
2284 is_directory,
2285 content,
2286 });
2287 }
2288 }
2289 requests.sort_unstable_by(|a, b| a.path.cmp(&b.path));
2290 requests.dedup();
2291
2292 let mut copied_entry_ids = Vec::new();
2293 for request in requests {
2294 let response = client.request(request).await?;
2295 copied_entry_ids.extend(response.entry.map(|e| ProjectEntryId::from_proto(e.id)));
2296 }
2297
2298 Ok(copied_entry_ids)
2299 })
2300 }
2301}
2302
2303impl Snapshot {
2304 pub fn new(
2305 id: WorktreeId,
2306 root_name: Arc<RelPath>,
2307 abs_path: Arc<Path>,
2308 path_style: PathStyle,
2309 ) -> Self {
2310 Snapshot {
2311 id,
2312 abs_path: SanitizedPath::from_arc(abs_path),
2313 path_style,
2314 root_char_bag: root_name
2315 .as_unix_str()
2316 .chars()
2317 .map(|c| c.to_ascii_lowercase())
2318 .collect(),
2319 root_name,
2320 always_included_entries: Default::default(),
2321 entries_by_path: Default::default(),
2322 entries_by_id: Default::default(),
2323 root_repo_common_dir: None,
2324 scan_id: 1,
2325 completed_scan_id: 0,
2326 }
2327 }
2328
2329 pub fn id(&self) -> WorktreeId {
2330 self.id
2331 }
2332
2333 // TODO:
2334 // Consider the following:
2335 //
2336 // ```rust
2337 // let abs_path: Arc<Path> = snapshot.abs_path(); // e.g. "C:\Users\user\Desktop\project"
2338 // let some_non_trimmed_path = Path::new("\\\\?\\C:\\Users\\user\\Desktop\\project\\main.rs");
2339 // // The caller perform some actions here:
2340 // some_non_trimmed_path.strip_prefix(abs_path); // This fails
2341 // some_non_trimmed_path.starts_with(abs_path); // This fails too
2342 // ```
2343 //
2344 // This is definitely a bug, but it's not clear if we should handle it here or not.
2345 pub fn abs_path(&self) -> &Arc<Path> {
2346 SanitizedPath::cast_arc_ref(&self.abs_path)
2347 }
2348
2349 pub fn root_repo_common_dir(&self) -> Option<&Arc<Path>> {
2350 self.root_repo_common_dir
2351 .as_ref()
2352 .map(SanitizedPath::cast_arc_ref)
2353 }
2354
2355 fn build_initial_update(&self, project_id: u64, worktree_id: u64) -> proto::UpdateWorktree {
2356 let mut updated_entries = self
2357 .entries_by_path
2358 .iter()
2359 .map(proto::Entry::from)
2360 .collect::<Vec<_>>();
2361 updated_entries.sort_unstable_by_key(|e| e.id);
2362
2363 proto::UpdateWorktree {
2364 project_id,
2365 worktree_id,
2366 abs_path: self.abs_path().to_string_lossy().into_owned(),
2367 root_name: self.root_name().to_proto(),
2368 root_repo_common_dir: self
2369 .root_repo_common_dir()
2370 .map(|p| p.to_string_lossy().into_owned()),
2371 updated_entries,
2372 removed_entries: Vec::new(),
2373 scan_id: self.scan_id as u64,
2374 is_last_update: self.completed_scan_id == self.scan_id,
2375 // Sent in separate messages.
2376 updated_repositories: Vec::new(),
2377 removed_repositories: Vec::new(),
2378 }
2379 }
2380
2381 pub fn work_directory_abs_path(&self, work_directory: &WorkDirectory) -> PathBuf {
2382 match work_directory {
2383 WorkDirectory::InProject { relative_path } => self.absolutize(relative_path),
2384 WorkDirectory::AboveProject { absolute_path, .. } => absolute_path.as_ref().to_owned(),
2385 }
2386 }
2387
2388 pub fn absolutize(&self, path: &RelPath) -> PathBuf {
2389 if path.file_name().is_some() {
2390 let mut abs_path = self.abs_path.to_string();
2391 for component in path.components() {
2392 if !abs_path.ends_with(self.path_style.primary_separator()) {
2393 abs_path.push_str(self.path_style.primary_separator());
2394 }
2395 abs_path.push_str(component);
2396 }
2397 PathBuf::from(abs_path)
2398 } else {
2399 self.abs_path.as_path().to_path_buf()
2400 }
2401 }
2402
2403 pub fn contains_entry(&self, entry_id: ProjectEntryId) -> bool {
2404 self.entries_by_id.get(&entry_id, ()).is_some()
2405 }
2406
2407 fn insert_entry(
2408 &mut self,
2409 entry: proto::Entry,
2410 always_included_paths: &PathMatcher,
2411 ) -> Result<Entry> {
2412 let entry = Entry::try_from((&self.root_char_bag, always_included_paths, entry))?;
2413 let old_entry = self.entries_by_id.insert_or_replace(
2414 PathEntry {
2415 id: entry.id,
2416 path: entry.path.clone(),
2417 is_ignored: entry.is_ignored,
2418 scan_id: 0,
2419 },
2420 (),
2421 );
2422 if let Some(old_entry) = old_entry {
2423 self.entries_by_path.remove(&PathKey(old_entry.path), ());
2424 }
2425 self.entries_by_path.insert_or_replace(entry.clone(), ());
2426 Ok(entry)
2427 }
2428
2429 fn delete_entry(&mut self, entry_id: ProjectEntryId) -> Option<Arc<RelPath>> {
2430 let removed_entry = self.entries_by_id.remove(&entry_id, ())?;
2431 self.entries_by_path = {
2432 let mut cursor = self.entries_by_path.cursor::<TraversalProgress>(());
2433 let mut new_entries_by_path =
2434 cursor.slice(&TraversalTarget::path(&removed_entry.path), Bias::Left);
2435 while let Some(entry) = cursor.item() {
2436 if entry.path.starts_with(&removed_entry.path) {
2437 self.entries_by_id.remove(&entry.id, ());
2438 cursor.next();
2439 } else {
2440 break;
2441 }
2442 }
2443 new_entries_by_path.append(cursor.suffix(), ());
2444 new_entries_by_path
2445 };
2446
2447 Some(removed_entry.path)
2448 }
2449
2450 fn update_abs_path(&mut self, abs_path: Arc<SanitizedPath>, root_name: Arc<RelPath>) {
2451 self.abs_path = abs_path;
2452 if root_name != self.root_name {
2453 self.root_char_bag = root_name
2454 .as_unix_str()
2455 .chars()
2456 .map(|c| c.to_ascii_lowercase())
2457 .collect();
2458 self.root_name = root_name;
2459 }
2460 }
2461
2462 pub fn apply_remote_update(
2463 &mut self,
2464 update: proto::UpdateWorktree,
2465 always_included_paths: &PathMatcher,
2466 ) {
2467 log::debug!(
2468 "applying remote worktree update. {} entries updated, {} removed",
2469 update.updated_entries.len(),
2470 update.removed_entries.len()
2471 );
2472 if let Some(root_name) = RelPath::from_proto(&update.root_name).log_err() {
2473 self.update_abs_path(
2474 SanitizedPath::new_arc(&Path::new(&update.abs_path)),
2475 root_name,
2476 );
2477 }
2478
2479 let mut entries_by_path_edits = Vec::new();
2480 let mut entries_by_id_edits = Vec::new();
2481
2482 for entry_id in update.removed_entries {
2483 let entry_id = ProjectEntryId::from_proto(entry_id);
2484 entries_by_id_edits.push(Edit::Remove(entry_id));
2485 if let Some(entry) = self.entry_for_id(entry_id) {
2486 entries_by_path_edits.push(Edit::Remove(PathKey(entry.path.clone())));
2487 }
2488 }
2489
2490 for entry in update.updated_entries {
2491 let Some(entry) =
2492 Entry::try_from((&self.root_char_bag, always_included_paths, entry)).log_err()
2493 else {
2494 continue;
2495 };
2496 if let Some(PathEntry { path, .. }) = self.entries_by_id.get(&entry.id, ()) {
2497 entries_by_path_edits.push(Edit::Remove(PathKey(path.clone())));
2498 }
2499 if let Some(old_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2500 && old_entry.id != entry.id
2501 {
2502 entries_by_id_edits.push(Edit::Remove(old_entry.id));
2503 }
2504 entries_by_id_edits.push(Edit::Insert(PathEntry {
2505 id: entry.id,
2506 path: entry.path.clone(),
2507 is_ignored: entry.is_ignored,
2508 scan_id: 0,
2509 }));
2510 entries_by_path_edits.push(Edit::Insert(entry));
2511 }
2512
2513 self.entries_by_path.edit(entries_by_path_edits, ());
2514 self.entries_by_id.edit(entries_by_id_edits, ());
2515
2516 if let Some(dir) = update
2517 .root_repo_common_dir
2518 .map(|p| SanitizedPath::new_arc(Path::new(&p)))
2519 {
2520 self.root_repo_common_dir = Some(dir);
2521 }
2522
2523 self.scan_id = update.scan_id as usize;
2524 if update.is_last_update {
2525 self.completed_scan_id = update.scan_id as usize;
2526 }
2527 }
2528
2529 pub fn entry_count(&self) -> usize {
2530 self.entries_by_path.summary().count
2531 }
2532
2533 pub fn visible_entry_count(&self) -> usize {
2534 self.entries_by_path.summary().non_ignored_count
2535 }
2536
2537 pub fn dir_count(&self) -> usize {
2538 let summary = self.entries_by_path.summary();
2539 summary.count - summary.file_count
2540 }
2541
2542 pub fn visible_dir_count(&self) -> usize {
2543 let summary = self.entries_by_path.summary();
2544 summary.non_ignored_count - summary.non_ignored_file_count
2545 }
2546
2547 pub fn file_count(&self) -> usize {
2548 self.entries_by_path.summary().file_count
2549 }
2550
2551 pub fn visible_file_count(&self) -> usize {
2552 self.entries_by_path.summary().non_ignored_file_count
2553 }
2554
2555 fn traverse_from_offset(
2556 &self,
2557 include_files: bool,
2558 include_dirs: bool,
2559 include_ignored: bool,
2560 start_offset: usize,
2561 ) -> Traversal<'_> {
2562 let mut cursor = self.entries_by_path.cursor(());
2563 cursor.seek(
2564 &TraversalTarget::Count {
2565 count: start_offset,
2566 include_files,
2567 include_dirs,
2568 include_ignored,
2569 },
2570 Bias::Right,
2571 );
2572 Traversal {
2573 snapshot: self,
2574 cursor,
2575 include_files,
2576 include_dirs,
2577 include_ignored,
2578 }
2579 }
2580
2581 pub fn traverse_from_path(
2582 &self,
2583 include_files: bool,
2584 include_dirs: bool,
2585 include_ignored: bool,
2586 path: &RelPath,
2587 ) -> Traversal<'_> {
2588 Traversal::new(self, include_files, include_dirs, include_ignored, path)
2589 }
2590
2591 pub fn files(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2592 self.traverse_from_offset(true, false, include_ignored, start)
2593 }
2594
2595 pub fn directories(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2596 self.traverse_from_offset(false, true, include_ignored, start)
2597 }
2598
2599 pub fn entries(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2600 self.traverse_from_offset(true, true, include_ignored, start)
2601 }
2602
2603 pub fn paths(&self) -> impl Iterator<Item = &RelPath> {
2604 self.entries_by_path
2605 .cursor::<()>(())
2606 .filter(move |entry| !entry.path.is_empty())
2607 .map(|entry| entry.path.as_ref())
2608 }
2609
2610 pub fn child_entries<'a>(&'a self, parent_path: &'a RelPath) -> ChildEntriesIter<'a> {
2611 let options = ChildEntriesOptions {
2612 include_files: true,
2613 include_dirs: true,
2614 include_ignored: true,
2615 };
2616 self.child_entries_with_options(parent_path, options)
2617 }
2618
2619 pub fn child_entries_with_options<'a>(
2620 &'a self,
2621 parent_path: &'a RelPath,
2622 options: ChildEntriesOptions,
2623 ) -> ChildEntriesIter<'a> {
2624 let mut cursor = self.entries_by_path.cursor(());
2625 cursor.seek(&TraversalTarget::path(parent_path), Bias::Right);
2626 let traversal = Traversal {
2627 snapshot: self,
2628 cursor,
2629 include_files: options.include_files,
2630 include_dirs: options.include_dirs,
2631 include_ignored: options.include_ignored,
2632 };
2633 ChildEntriesIter {
2634 traversal,
2635 parent_path,
2636 }
2637 }
2638
2639 pub fn root_entry(&self) -> Option<&Entry> {
2640 self.entries_by_path.first()
2641 }
2642
2643 /// Returns `None` for a single file worktree, or `Some(self.abs_path())` if
2644 /// it is a directory.
2645 pub fn root_dir(&self) -> Option<Arc<Path>> {
2646 self.root_entry()
2647 .filter(|entry| entry.is_dir())
2648 .map(|_| self.abs_path().clone())
2649 }
2650
2651 pub fn root_name(&self) -> &RelPath {
2652 &self.root_name
2653 }
2654
2655 pub fn root_name_str(&self) -> &str {
2656 self.root_name.as_unix_str()
2657 }
2658
2659 pub fn scan_id(&self) -> usize {
2660 self.scan_id
2661 }
2662
2663 pub fn entry_for_path(&self, path: &RelPath) -> Option<&Entry> {
2664 let entry = self.traverse_from_path(true, true, true, path).entry();
2665 entry.and_then(|entry| {
2666 if entry.path.as_ref() == path {
2667 Some(entry)
2668 } else {
2669 None
2670 }
2671 })
2672 }
2673
2674 /// Resolves a path to an executable using the following heuristics:
2675 ///
2676 /// 1. If the path starts with `~`, it is expanded to the user's home directory.
2677 /// 2. If the path is relative and contains more than one component,
2678 /// it is joined to the worktree root path.
2679 /// 3. If the path is relative and exists in the worktree
2680 /// (even if falls under an exclusion filter),
2681 /// it is joined to the worktree root path.
2682 /// 4. Otherwise the path is returned unmodified.
2683 ///
2684 /// Relative paths that do not exist in the worktree may
2685 /// still be found using the `PATH` environment variable.
2686 pub fn resolve_relative_path(&self, path: PathBuf) -> PathBuf {
2687 if let Some(path_str) = path.to_str() {
2688 if let Some(remaining_path) = path_str.strip_prefix("~/") {
2689 return home_dir().join(remaining_path);
2690 } else if path_str == "~" {
2691 return home_dir().to_path_buf();
2692 }
2693 }
2694
2695 if let Ok(rel_path) = RelPath::new(&path, self.path_style)
2696 && (path.components().count() > 1 || self.entry_for_path(&rel_path).is_some())
2697 {
2698 self.abs_path().join(path)
2699 } else {
2700 path
2701 }
2702 }
2703
2704 pub fn entry_for_id(&self, id: ProjectEntryId) -> Option<&Entry> {
2705 let entry = self.entries_by_id.get(&id, ())?;
2706 self.entry_for_path(&entry.path)
2707 }
2708
2709 pub fn path_style(&self) -> PathStyle {
2710 self.path_style
2711 }
2712}
2713
2714impl LocalSnapshot {
2715 fn local_repo_for_work_directory_path(&self, path: &RelPath) -> Option<&LocalRepositoryEntry> {
2716 self.git_repositories
2717 .iter()
2718 .map(|(_, entry)| entry)
2719 .find(|entry| entry.work_directory.path_key() == PathKey(path.into()))
2720 }
2721
2722 fn build_update(
2723 &self,
2724 project_id: u64,
2725 worktree_id: u64,
2726 entry_changes: UpdatedEntriesSet,
2727 ) -> proto::UpdateWorktree {
2728 let mut updated_entries = Vec::new();
2729 let mut removed_entries = Vec::new();
2730
2731 for (_, entry_id, path_change) in entry_changes.iter() {
2732 if let PathChange::Removed = path_change {
2733 removed_entries.push(entry_id.0 as u64);
2734 } else if let Some(entry) = self.entry_for_id(*entry_id) {
2735 updated_entries.push(proto::Entry::from(entry));
2736 }
2737 }
2738
2739 removed_entries.sort_unstable();
2740 updated_entries.sort_unstable_by_key(|e| e.id);
2741
2742 // TODO - optimize, knowing that removed_entries are sorted.
2743 removed_entries.retain(|id| updated_entries.binary_search_by_key(id, |e| e.id).is_err());
2744
2745 proto::UpdateWorktree {
2746 project_id,
2747 worktree_id,
2748 abs_path: self.abs_path().to_string_lossy().into_owned(),
2749 root_name: self.root_name().to_proto(),
2750 root_repo_common_dir: self
2751 .root_repo_common_dir()
2752 .map(|p| p.to_string_lossy().into_owned()),
2753 updated_entries,
2754 removed_entries,
2755 scan_id: self.scan_id as u64,
2756 is_last_update: self.completed_scan_id == self.scan_id,
2757 // Sent in separate messages.
2758 updated_repositories: Vec::new(),
2759 removed_repositories: Vec::new(),
2760 }
2761 }
2762
2763 async fn insert_entry(&mut self, mut entry: Entry, fs: &dyn Fs) -> Entry {
2764 log::trace!("insert entry {:?}", entry.path);
2765 if entry.is_file() && entry.path.file_name() == Some(&GITIGNORE) {
2766 let abs_path = self.absolutize(&entry.path);
2767 match build_gitignore(&abs_path, fs).await {
2768 Ok(ignore) => {
2769 self.ignores_by_parent_abs_path
2770 .insert(abs_path.parent().unwrap().into(), (Arc::new(ignore), true));
2771 }
2772 Err(error) => {
2773 log::error!(
2774 "error loading .gitignore file {:?} - {:?}",
2775 &entry.path,
2776 error
2777 );
2778 }
2779 }
2780 }
2781
2782 if entry.kind == EntryKind::PendingDir
2783 && let Some(existing_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2784 {
2785 entry.kind = existing_entry.kind;
2786 }
2787
2788 let scan_id = self.scan_id;
2789 let removed = self.entries_by_path.insert_or_replace(entry.clone(), ());
2790 if let Some(removed) = removed
2791 && removed.id != entry.id
2792 {
2793 self.entries_by_id.remove(&removed.id, ());
2794 }
2795 self.entries_by_id.insert_or_replace(
2796 PathEntry {
2797 id: entry.id,
2798 path: entry.path.clone(),
2799 is_ignored: entry.is_ignored,
2800 scan_id,
2801 },
2802 (),
2803 );
2804
2805 entry
2806 }
2807
2808 fn ancestor_inodes_for_path(&self, path: &RelPath) -> TreeSet<u64> {
2809 let mut inodes = TreeSet::default();
2810 for ancestor in path.ancestors().skip(1) {
2811 if let Some(entry) = self.entry_for_path(ancestor) {
2812 inodes.insert(entry.inode);
2813 }
2814 }
2815 inodes
2816 }
2817
2818 async fn ignore_stack_for_abs_path(
2819 &self,
2820 abs_path: &Path,
2821 is_dir: bool,
2822 fs: &dyn Fs,
2823 ) -> IgnoreStack {
2824 let mut new_ignores = Vec::new();
2825 let mut repo_root = None;
2826 for (index, ancestor) in abs_path.ancestors().enumerate() {
2827 if index > 0 {
2828 if let Some((ignore, _)) = self.ignores_by_parent_abs_path.get(ancestor) {
2829 new_ignores.push((ancestor, Some(ignore.clone())));
2830 } else {
2831 new_ignores.push((ancestor, None));
2832 }
2833 }
2834
2835 let metadata = fs.metadata(&ancestor.join(DOT_GIT)).await.ok().flatten();
2836 if metadata.is_some() {
2837 repo_root = Some(Arc::from(ancestor));
2838 break;
2839 }
2840 }
2841
2842 let mut ignore_stack = if let Some(global_gitignore) = self.global_gitignore.clone() {
2843 IgnoreStack::global(global_gitignore)
2844 } else {
2845 IgnoreStack::none()
2846 };
2847
2848 if let Some((repo_exclude, _)) = repo_root
2849 .as_ref()
2850 .and_then(|abs_path| self.repo_exclude_by_work_dir_abs_path.get(abs_path))
2851 {
2852 ignore_stack = ignore_stack.append(IgnoreKind::RepoExclude, repo_exclude.clone());
2853 }
2854 ignore_stack.repo_root = repo_root;
2855 for (parent_abs_path, ignore) in new_ignores.into_iter().rev() {
2856 if ignore_stack.is_abs_path_ignored(parent_abs_path, true) {
2857 ignore_stack = IgnoreStack::all();
2858 break;
2859 } else if let Some(ignore) = ignore {
2860 ignore_stack =
2861 ignore_stack.append(IgnoreKind::Gitignore(parent_abs_path.into()), ignore);
2862 }
2863 }
2864
2865 if ignore_stack.is_abs_path_ignored(abs_path, is_dir) {
2866 ignore_stack = IgnoreStack::all();
2867 }
2868
2869 ignore_stack
2870 }
2871
2872 #[cfg(feature = "test-support")]
2873 pub fn expanded_entries(&self) -> impl Iterator<Item = &Entry> {
2874 self.entries_by_path
2875 .cursor::<()>(())
2876 .filter(|entry| entry.kind == EntryKind::Dir && (entry.is_external || entry.is_ignored))
2877 }
2878
2879 #[cfg(feature = "test-support")]
2880 pub fn check_invariants(&self, git_state: bool) {
2881 use pretty_assertions::assert_eq;
2882
2883 assert_eq!(
2884 self.entries_by_path
2885 .cursor::<()>(())
2886 .map(|e| (&e.path, e.id))
2887 .collect::<Vec<_>>(),
2888 self.entries_by_id
2889 .cursor::<()>(())
2890 .map(|e| (&e.path, e.id))
2891 .collect::<collections::BTreeSet<_>>()
2892 .into_iter()
2893 .collect::<Vec<_>>(),
2894 "entries_by_path and entries_by_id are inconsistent"
2895 );
2896
2897 let mut files = self.files(true, 0);
2898 let mut visible_files = self.files(false, 0);
2899 for entry in self.entries_by_path.cursor::<()>(()) {
2900 if entry.is_file() {
2901 assert_eq!(files.next().unwrap().inode, entry.inode);
2902 if !entry.is_ignored || entry.is_always_included {
2903 assert_eq!(visible_files.next().unwrap().inode, entry.inode);
2904 }
2905 }
2906 }
2907
2908 assert!(files.next().is_none());
2909 assert!(visible_files.next().is_none());
2910
2911 let mut bfs_paths = Vec::new();
2912 let mut stack = self
2913 .root_entry()
2914 .map(|e| e.path.as_ref())
2915 .into_iter()
2916 .collect::<Vec<_>>();
2917 while let Some(path) = stack.pop() {
2918 bfs_paths.push(path);
2919 let ix = stack.len();
2920 for child_entry in self.child_entries(path) {
2921 stack.insert(ix, &child_entry.path);
2922 }
2923 }
2924
2925 let dfs_paths_via_iter = self
2926 .entries_by_path
2927 .cursor::<()>(())
2928 .map(|e| e.path.as_ref())
2929 .collect::<Vec<_>>();
2930 assert_eq!(bfs_paths, dfs_paths_via_iter);
2931
2932 let dfs_paths_via_traversal = self
2933 .entries(true, 0)
2934 .map(|e| e.path.as_ref())
2935 .collect::<Vec<_>>();
2936
2937 assert_eq!(dfs_paths_via_traversal, dfs_paths_via_iter);
2938
2939 if git_state {
2940 for ignore_parent_abs_path in self.ignores_by_parent_abs_path.keys() {
2941 let ignore_parent_path = &RelPath::new(
2942 ignore_parent_abs_path
2943 .strip_prefix(self.abs_path.as_path())
2944 .unwrap(),
2945 PathStyle::local(),
2946 )
2947 .unwrap();
2948 assert!(self.entry_for_path(ignore_parent_path).is_some());
2949 assert!(
2950 self.entry_for_path(
2951 &ignore_parent_path.join(RelPath::unix(GITIGNORE).unwrap())
2952 )
2953 .is_some()
2954 );
2955 }
2956 }
2957 }
2958
2959 #[cfg(feature = "test-support")]
2960 pub fn entries_without_ids(&self, include_ignored: bool) -> Vec<(&RelPath, u64, bool)> {
2961 let mut paths = Vec::new();
2962 for entry in self.entries_by_path.cursor::<()>(()) {
2963 if include_ignored || !entry.is_ignored {
2964 paths.push((entry.path.as_ref(), entry.inode, entry.is_ignored));
2965 }
2966 }
2967 paths.sort_by(|a, b| a.0.cmp(b.0));
2968 paths
2969 }
2970}
2971
2972impl BackgroundScannerState {
2973 fn should_scan_directory(&self, entry: &Entry) -> bool {
2974 (self.scanning_enabled && !entry.is_external && (!entry.is_ignored || entry.is_always_included))
2975 || entry.path.file_name() == Some(DOT_GIT)
2976 || entry.path.file_name() == Some(local_settings_folder_name())
2977 || entry.path.file_name() == Some(local_vscode_folder_name())
2978 || self.scanned_dirs.contains(&entry.id) // If we've ever scanned it, keep scanning
2979 || self
2980 .paths_to_scan
2981 .iter()
2982 .any(|p| p.starts_with(&entry.path))
2983 || self
2984 .path_prefixes_to_scan
2985 .iter()
2986 .any(|p| entry.path.starts_with(p))
2987 }
2988
2989 async fn enqueue_scan_dir(
2990 &self,
2991 abs_path: Arc<Path>,
2992 entry: &Entry,
2993 scan_job_tx: &Sender<ScanJob>,
2994 fs: &dyn Fs,
2995 ) {
2996 let path = entry.path.clone();
2997 let ignore_stack = self
2998 .snapshot
2999 .ignore_stack_for_abs_path(&abs_path, true, fs)
3000 .await;
3001 let mut ancestor_inodes = self.snapshot.ancestor_inodes_for_path(&path);
3002
3003 if !ancestor_inodes.contains(&entry.inode) {
3004 ancestor_inodes.insert(entry.inode);
3005 scan_job_tx
3006 .try_send(ScanJob {
3007 abs_path,
3008 path,
3009 ignore_stack,
3010 scan_queue: scan_job_tx.clone(),
3011 ancestor_inodes,
3012 is_external: entry.is_external,
3013 })
3014 .unwrap();
3015 }
3016 }
3017
3018 fn reuse_entry_id(&mut self, entry: &mut Entry) {
3019 if let Some(mtime) = entry.mtime {
3020 // If an entry with the same inode was removed from the worktree during this scan,
3021 // then it *might* represent the same file or directory. But the OS might also have
3022 // re-used the inode for a completely different file or directory.
3023 //
3024 // Conditionally reuse the old entry's id:
3025 // * if the mtime is the same, the file was probably been renamed.
3026 // * if the path is the same, the file may just have been updated
3027 if let Some(removed_entry) = self.removed_entries.remove(&entry.inode) {
3028 if removed_entry.mtime == Some(mtime) || removed_entry.path == entry.path {
3029 entry.id = removed_entry.id;
3030 }
3031 } else if let Some(existing_entry) = self.snapshot.entry_for_path(&entry.path) {
3032 entry.id = existing_entry.id;
3033 }
3034 }
3035 }
3036
3037 fn entry_id_for(
3038 &mut self,
3039 next_entry_id: &AtomicUsize,
3040 path: &RelPath,
3041 metadata: &fs::Metadata,
3042 ) -> ProjectEntryId {
3043 // If an entry with the same inode was removed from the worktree during this scan,
3044 // then it *might* represent the same file or directory. But the OS might also have
3045 // re-used the inode for a completely different file or directory.
3046 //
3047 // Conditionally reuse the old entry's id:
3048 // * if the mtime is the same, the file was probably been renamed.
3049 // * if the path is the same, the file may just have been updated
3050 if let Some(removed_entry) = self.removed_entries.remove(&metadata.inode) {
3051 if removed_entry.mtime == Some(metadata.mtime) || *removed_entry.path == *path {
3052 return removed_entry.id;
3053 }
3054 } else if let Some(existing_entry) = self.snapshot.entry_for_path(path) {
3055 return existing_entry.id;
3056 }
3057 ProjectEntryId::new(next_entry_id)
3058 }
3059
3060 async fn insert_entry(&mut self, entry: Entry, fs: &dyn Fs, watcher: &dyn Watcher) -> Entry {
3061 let entry = self.snapshot.insert_entry(entry, fs).await;
3062 if entry.path.file_name() == Some(&DOT_GIT) {
3063 self.insert_git_repository(entry.path.clone(), fs, watcher)
3064 .await;
3065 }
3066
3067 #[cfg(feature = "test-support")]
3068 self.snapshot.check_invariants(false);
3069
3070 entry
3071 }
3072
3073 fn populate_dir(
3074 &mut self,
3075 parent_path: Arc<RelPath>,
3076 entries: impl IntoIterator<Item = Entry>,
3077 ignore: Option<Arc<Gitignore>>,
3078 ) {
3079 let mut parent_entry = if let Some(parent_entry) = self
3080 .snapshot
3081 .entries_by_path
3082 .get(&PathKey(parent_path.clone()), ())
3083 {
3084 parent_entry.clone()
3085 } else {
3086 log::warn!(
3087 "populating a directory {:?} that has been removed",
3088 parent_path
3089 );
3090 return;
3091 };
3092
3093 match parent_entry.kind {
3094 EntryKind::PendingDir | EntryKind::UnloadedDir => parent_entry.kind = EntryKind::Dir,
3095 EntryKind::Dir => {}
3096 _ => return,
3097 }
3098
3099 if let Some(ignore) = ignore {
3100 let abs_parent_path = self
3101 .snapshot
3102 .abs_path
3103 .as_path()
3104 .join(parent_path.as_std_path())
3105 .into();
3106 self.snapshot
3107 .ignores_by_parent_abs_path
3108 .insert(abs_parent_path, (ignore, false));
3109 }
3110
3111 let parent_entry_id = parent_entry.id;
3112 self.scanned_dirs.insert(parent_entry_id);
3113 let mut entries_by_path_edits = vec![Edit::Insert(parent_entry)];
3114 let mut entries_by_id_edits = Vec::new();
3115
3116 for entry in entries {
3117 entries_by_id_edits.push(Edit::Insert(PathEntry {
3118 id: entry.id,
3119 path: entry.path.clone(),
3120 is_ignored: entry.is_ignored,
3121 scan_id: self.snapshot.scan_id,
3122 }));
3123 entries_by_path_edits.push(Edit::Insert(entry));
3124 }
3125
3126 self.snapshot
3127 .entries_by_path
3128 .edit(entries_by_path_edits, ());
3129 self.snapshot.entries_by_id.edit(entries_by_id_edits, ());
3130
3131 if let Err(ix) = self.changed_paths.binary_search(&parent_path) {
3132 self.changed_paths.insert(ix, parent_path.clone());
3133 }
3134
3135 #[cfg(feature = "test-support")]
3136 self.snapshot.check_invariants(false);
3137 }
3138
3139 fn remove_path(&mut self, path: &RelPath, watcher: &dyn Watcher) {
3140 log::trace!("background scanner removing path {path:?}");
3141 let mut new_entries;
3142 let removed_entries;
3143 {
3144 let mut cursor = self
3145 .snapshot
3146 .entries_by_path
3147 .cursor::<TraversalProgress>(());
3148 new_entries = cursor.slice(&TraversalTarget::path(path), Bias::Left);
3149 removed_entries = cursor.slice(&TraversalTarget::successor(path), Bias::Left);
3150 new_entries.append(cursor.suffix(), ());
3151 }
3152 self.snapshot.entries_by_path = new_entries;
3153
3154 let mut removed_ids = Vec::with_capacity(removed_entries.summary().count);
3155 let mut removed_dir_abs_paths = Vec::new();
3156 for entry in removed_entries.cursor::<()>(()) {
3157 if entry.is_dir() {
3158 let watch_path = self
3159 .watched_dir_abs_paths_by_entry_id
3160 .remove(&entry.id)
3161 .map(|path| path.as_ref().to_path_buf())
3162 .unwrap_or_else(|| self.snapshot.absolutize(&entry.path));
3163 removed_dir_abs_paths.push(watch_path);
3164 }
3165
3166 match self.removed_entries.entry(entry.inode) {
3167 hash_map::Entry::Occupied(mut e) => {
3168 let prev_removed_entry = e.get_mut();
3169 if entry.id > prev_removed_entry.id {
3170 *prev_removed_entry = entry.clone();
3171 }
3172 }
3173 hash_map::Entry::Vacant(e) => {
3174 e.insert(entry.clone());
3175 }
3176 }
3177
3178 if entry.path.file_name() == Some(GITIGNORE) {
3179 let abs_parent_path = self.snapshot.absolutize(&entry.path.parent().unwrap());
3180 if let Some((_, needs_update)) = self
3181 .snapshot
3182 .ignores_by_parent_abs_path
3183 .get_mut(abs_parent_path.as_path())
3184 {
3185 *needs_update = true;
3186 }
3187 }
3188
3189 if let Err(ix) = removed_ids.binary_search(&entry.id) {
3190 removed_ids.insert(ix, entry.id);
3191 }
3192 }
3193
3194 self.snapshot
3195 .entries_by_id
3196 .edit(removed_ids.iter().map(|&id| Edit::Remove(id)).collect(), ());
3197 self.snapshot
3198 .git_repositories
3199 .retain(|id, _| removed_ids.binary_search(id).is_err());
3200
3201 for removed_dir_abs_path in removed_dir_abs_paths {
3202 watcher.remove(&removed_dir_abs_path).log_err();
3203 }
3204
3205 #[cfg(feature = "test-support")]
3206 self.snapshot.check_invariants(false);
3207 }
3208
3209 async fn insert_git_repository(
3210 &mut self,
3211 dot_git_path: Arc<RelPath>,
3212 fs: &dyn Fs,
3213 watcher: &dyn Watcher,
3214 ) {
3215 let work_dir_path: Arc<RelPath> = match dot_git_path.parent() {
3216 Some(parent_dir) => {
3217 // Guard against repositories inside the repository metadata
3218 if parent_dir
3219 .components()
3220 .any(|component| component == DOT_GIT)
3221 {
3222 log::debug!(
3223 "not building git repository for nested `.git` directory, `.git` path in the worktree: {dot_git_path:?}"
3224 );
3225 return;
3226 };
3227
3228 parent_dir.into()
3229 }
3230 None => {
3231 // `dot_git_path.parent().is_none()` means `.git` directory is the opened worktree itself,
3232 // no files inside that directory are tracked by git, so no need to build the repo around it
3233 log::debug!(
3234 "not building git repository for the worktree itself, `.git` path in the worktree: {dot_git_path:?}"
3235 );
3236 return;
3237 }
3238 };
3239
3240 let dot_git_abs_path = Arc::from(self.snapshot.absolutize(&dot_git_path).as_ref());
3241
3242 self.insert_git_repository_for_path(
3243 WorkDirectory::InProject {
3244 relative_path: work_dir_path,
3245 },
3246 dot_git_abs_path,
3247 fs,
3248 watcher,
3249 )
3250 .await
3251 .log_err();
3252 }
3253
3254 async fn insert_git_repository_for_path(
3255 &mut self,
3256 work_directory: WorkDirectory,
3257 dot_git_abs_path: Arc<Path>,
3258 fs: &dyn Fs,
3259 watcher: &dyn Watcher,
3260 ) -> Result<LocalRepositoryEntry> {
3261 let work_dir_entry = self
3262 .snapshot
3263 .entry_for_path(&work_directory.path_key().0)
3264 .with_context(|| {
3265 format!(
3266 "working directory `{}` not indexed",
3267 work_directory
3268 .path_key()
3269 .0
3270 .display(self.snapshot.path_style)
3271 )
3272 })?;
3273 let work_directory_abs_path = self.snapshot.work_directory_abs_path(&work_directory);
3274
3275 let (repository_dir_abs_path, common_dir_abs_path) =
3276 discover_git_paths(&dot_git_abs_path, fs).await;
3277 watcher
3278 .add(&common_dir_abs_path)
3279 .context("failed to add common directory to watcher")
3280 .log_err();
3281 watcher
3282 .add(&repository_dir_abs_path)
3283 .context("failed to add repository directory to watcher")
3284 .log_err();
3285
3286 let work_directory_id = work_dir_entry.id;
3287
3288 let local_repository = LocalRepositoryEntry {
3289 work_directory_id,
3290 work_directory,
3291 work_directory_abs_path: work_directory_abs_path.as_path().into(),
3292 git_dir_scan_id: 0,
3293 dot_git_abs_path,
3294 common_dir_abs_path,
3295 repository_dir_abs_path,
3296 };
3297
3298 self.snapshot
3299 .git_repositories
3300 .insert(work_directory_id, local_repository.clone());
3301
3302 log::trace!("inserting new local git repository");
3303 Ok(local_repository)
3304 }
3305}
3306
3307async fn is_dot_git(path: &Path, fs: &dyn Fs) -> bool {
3308 if let Some(file_name) = path.file_name()
3309 && file_name == DOT_GIT
3310 {
3311 return true;
3312 }
3313
3314 // If we're in a bare repository, we are not inside a `.git` folder. In a
3315 // bare repository, the root folder contains what would normally be in the
3316 // `.git` folder.
3317 let head_metadata = fs.metadata(&path.join("HEAD")).await;
3318 if !matches!(head_metadata, Ok(Some(_))) {
3319 return false;
3320 }
3321 let config_metadata = fs.metadata(&path.join("config")).await;
3322 matches!(config_metadata, Ok(Some(_)))
3323}
3324
3325async fn build_gitignore(abs_path: &Path, fs: &dyn Fs) -> Result<Gitignore> {
3326 let contents = fs
3327 .load(abs_path)
3328 .await
3329 .with_context(|| format!("failed to load gitignore file at {}", abs_path.display()))?;
3330 let parent = abs_path.parent().unwrap_or_else(|| Path::new("/"));
3331 let mut builder = GitignoreBuilder::new(parent);
3332 for line in contents.lines() {
3333 builder.add_line(Some(abs_path.into()), line)?;
3334 }
3335 Ok(builder.build()?)
3336}
3337
3338impl Deref for Worktree {
3339 type Target = Snapshot;
3340
3341 fn deref(&self) -> &Self::Target {
3342 match self {
3343 Worktree::Local(worktree) => &worktree.snapshot,
3344 Worktree::Remote(worktree) => &worktree.snapshot,
3345 }
3346 }
3347}
3348
3349impl Deref for LocalWorktree {
3350 type Target = LocalSnapshot;
3351
3352 fn deref(&self) -> &Self::Target {
3353 &self.snapshot
3354 }
3355}
3356
3357impl Deref for RemoteWorktree {
3358 type Target = Snapshot;
3359
3360 fn deref(&self) -> &Self::Target {
3361 &self.snapshot
3362 }
3363}
3364
3365impl fmt::Debug for LocalWorktree {
3366 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3367 self.snapshot.fmt(f)
3368 }
3369}
3370
3371impl fmt::Debug for Snapshot {
3372 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3373 struct EntriesById<'a>(&'a SumTree<PathEntry>);
3374 struct EntriesByPath<'a>(&'a SumTree<Entry>);
3375
3376 impl fmt::Debug for EntriesByPath<'_> {
3377 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3378 f.debug_map()
3379 .entries(self.0.iter().map(|entry| (&entry.path, entry.id)))
3380 .finish()
3381 }
3382 }
3383
3384 impl fmt::Debug for EntriesById<'_> {
3385 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3386 f.debug_list().entries(self.0.iter()).finish()
3387 }
3388 }
3389
3390 f.debug_struct("Snapshot")
3391 .field("id", &self.id)
3392 .field("root_name", &self.root_name)
3393 .field("entries_by_path", &EntriesByPath(&self.entries_by_path))
3394 .field("entries_by_id", &EntriesById(&self.entries_by_id))
3395 .finish()
3396 }
3397}
3398
3399#[derive(Debug, Clone, PartialEq)]
3400pub struct File {
3401 pub worktree: Entity<Worktree>,
3402 pub path: Arc<RelPath>,
3403 pub disk_state: DiskState,
3404 pub entry_id: Option<ProjectEntryId>,
3405 pub is_local: bool,
3406 pub is_private: bool,
3407}
3408
3409impl language::File for File {
3410 fn as_local(&self) -> Option<&dyn language::LocalFile> {
3411 if self.is_local { Some(self) } else { None }
3412 }
3413
3414 fn disk_state(&self) -> DiskState {
3415 self.disk_state
3416 }
3417
3418 fn path(&self) -> &Arc<RelPath> {
3419 &self.path
3420 }
3421
3422 fn full_path(&self, cx: &App) -> PathBuf {
3423 self.worktree.read(cx).full_path(&self.path)
3424 }
3425
3426 /// Returns the last component of this handle's absolute path. If this handle refers to the root
3427 /// of its worktree, then this method will return the name of the worktree itself.
3428 fn file_name<'a>(&'a self, cx: &'a App) -> &'a str {
3429 self.path
3430 .file_name()
3431 .unwrap_or_else(|| self.worktree.read(cx).root_name_str())
3432 }
3433
3434 fn worktree_id(&self, cx: &App) -> WorktreeId {
3435 self.worktree.read(cx).id()
3436 }
3437
3438 fn to_proto(&self, cx: &App) -> rpc::proto::File {
3439 rpc::proto::File {
3440 worktree_id: self.worktree.read(cx).id().to_proto(),
3441 entry_id: self.entry_id.map(|id| id.to_proto()),
3442 path: self.path.as_ref().to_proto(),
3443 mtime: self.disk_state.mtime().map(|time| time.into()),
3444 is_deleted: self.disk_state.is_deleted(),
3445 is_historic: matches!(self.disk_state, DiskState::Historic { .. }),
3446 }
3447 }
3448
3449 fn is_private(&self) -> bool {
3450 self.is_private
3451 }
3452
3453 fn path_style(&self, cx: &App) -> PathStyle {
3454 self.worktree.read(cx).path_style()
3455 }
3456
3457 fn can_open(&self) -> bool {
3458 true
3459 }
3460}
3461
3462impl language::LocalFile for File {
3463 fn abs_path(&self, cx: &App) -> PathBuf {
3464 self.worktree.read(cx).absolutize(&self.path)
3465 }
3466
3467 fn load(&self, cx: &App) -> Task<Result<String>> {
3468 let worktree = self.worktree.read(cx).as_local().unwrap();
3469 let abs_path = worktree.absolutize(&self.path);
3470 let fs = worktree.fs.clone();
3471 cx.background_spawn(async move { fs.load(&abs_path).await })
3472 }
3473
3474 fn load_bytes(&self, cx: &App) -> Task<Result<Vec<u8>>> {
3475 let worktree = self.worktree.read(cx).as_local().unwrap();
3476 let abs_path = worktree.absolutize(&self.path);
3477 let fs = worktree.fs.clone();
3478 cx.background_spawn(async move { fs.load_bytes(&abs_path).await })
3479 }
3480}
3481
3482impl File {
3483 pub fn for_entry(entry: Entry, worktree: Entity<Worktree>) -> Arc<Self> {
3484 Arc::new(Self {
3485 worktree,
3486 path: entry.path.clone(),
3487 disk_state: if let Some(mtime) = entry.mtime {
3488 DiskState::Present {
3489 mtime,
3490 size: entry.size,
3491 }
3492 } else {
3493 DiskState::New
3494 },
3495 entry_id: Some(entry.id),
3496 is_local: true,
3497 is_private: entry.is_private,
3498 })
3499 }
3500
3501 pub fn from_proto(
3502 proto: rpc::proto::File,
3503 worktree: Entity<Worktree>,
3504 cx: &App,
3505 ) -> Result<Self> {
3506 let worktree_id = worktree.read(cx).as_remote().context("not remote")?.id();
3507
3508 anyhow::ensure!(
3509 worktree_id.to_proto() == proto.worktree_id,
3510 "worktree id does not match file"
3511 );
3512
3513 let disk_state = if proto.is_historic {
3514 DiskState::Historic {
3515 was_deleted: proto.is_deleted,
3516 }
3517 } else if proto.is_deleted {
3518 DiskState::Deleted
3519 } else if let Some(mtime) = proto.mtime.map(&Into::into) {
3520 DiskState::Present { mtime, size: 0 }
3521 } else {
3522 DiskState::New
3523 };
3524
3525 Ok(Self {
3526 worktree,
3527 path: RelPath::from_proto(&proto.path).context("invalid path in file protobuf")?,
3528 disk_state,
3529 entry_id: proto.entry_id.map(ProjectEntryId::from_proto),
3530 is_local: false,
3531 is_private: false,
3532 })
3533 }
3534
3535 pub fn from_dyn(file: Option<&Arc<dyn language::File>>) -> Option<&Self> {
3536 file.and_then(|f| {
3537 let f: &dyn language::File = f.borrow();
3538 let f: &dyn Any = f;
3539 f.downcast_ref()
3540 })
3541 }
3542
3543 pub fn worktree_id(&self, cx: &App) -> WorktreeId {
3544 self.worktree.read(cx).id()
3545 }
3546
3547 pub fn project_entry_id(&self) -> Option<ProjectEntryId> {
3548 match self.disk_state {
3549 DiskState::Deleted => None,
3550 _ => self.entry_id,
3551 }
3552 }
3553}
3554
3555#[derive(Clone, Debug, PartialEq, Eq)]
3556pub struct Entry {
3557 pub id: ProjectEntryId,
3558 pub kind: EntryKind,
3559 pub path: Arc<RelPath>,
3560 pub inode: u64,
3561 pub mtime: Option<MTime>,
3562
3563 pub canonical_path: Option<Arc<Path>>,
3564 /// Whether this entry is ignored by Git.
3565 ///
3566 /// We only scan ignored entries once the directory is expanded and
3567 /// exclude them from searches.
3568 pub is_ignored: bool,
3569
3570 /// Whether this entry is hidden or inside hidden directory.
3571 ///
3572 /// We only scan hidden entries once the directory is expanded.
3573 pub is_hidden: bool,
3574
3575 /// Whether this entry is always included in searches.
3576 ///
3577 /// This is used for entries that are always included in searches, even
3578 /// if they are ignored by git. Overridden by file_scan_exclusions.
3579 pub is_always_included: bool,
3580
3581 /// Whether this entry's canonical path is outside of the worktree.
3582 /// This means the entry is only accessible from the worktree root via a
3583 /// symlink.
3584 ///
3585 /// We only scan entries outside of the worktree once the symlinked
3586 /// directory is expanded.
3587 pub is_external: bool,
3588
3589 /// Whether this entry is considered to be a `.env` file.
3590 pub is_private: bool,
3591 /// The entry's size on disk, in bytes.
3592 pub size: u64,
3593 pub char_bag: CharBag,
3594 pub is_fifo: bool,
3595}
3596
3597#[derive(Clone, Copy, Debug, PartialEq, Eq)]
3598pub enum EntryKind {
3599 UnloadedDir,
3600 PendingDir,
3601 Dir,
3602 File,
3603}
3604
3605#[derive(Clone, Copy, Debug, PartialEq)]
3606pub enum PathChange {
3607 /// A filesystem entry was was created.
3608 Added,
3609 /// A filesystem entry was removed.
3610 Removed,
3611 /// A filesystem entry was updated.
3612 Updated,
3613 /// A filesystem entry was either updated or added. We don't know
3614 /// whether or not it already existed, because the path had not
3615 /// been loaded before the event.
3616 AddedOrUpdated,
3617 /// A filesystem entry was found during the initial scan of the worktree.
3618 Loaded,
3619}
3620
3621#[derive(Clone, Debug, PartialEq, Eq)]
3622pub struct UpdatedGitRepository {
3623 /// ID of the repository's working directory.
3624 ///
3625 /// For a repo that's above the worktree root, this is the ID of the worktree root, and hence not unique.
3626 /// It's included here to aid the GitStore in detecting when a repository's working directory is renamed.
3627 pub work_directory_id: ProjectEntryId,
3628 pub old_work_directory_abs_path: Option<Arc<Path>>,
3629 pub new_work_directory_abs_path: Option<Arc<Path>>,
3630 /// For a normal git repository checkout, the absolute path to the .git directory.
3631 /// For a worktree, the absolute path to the worktree's subdirectory inside the .git directory.
3632 pub dot_git_abs_path: Option<Arc<Path>>,
3633 pub repository_dir_abs_path: Option<Arc<Path>>,
3634 pub common_dir_abs_path: Option<Arc<Path>>,
3635}
3636
3637pub type UpdatedEntriesSet = Arc<[(Arc<RelPath>, ProjectEntryId, PathChange)]>;
3638pub type UpdatedGitRepositoriesSet = Arc<[UpdatedGitRepository]>;
3639
3640#[derive(Clone, Debug)]
3641pub struct PathProgress<'a> {
3642 pub max_path: &'a RelPath,
3643}
3644
3645#[derive(Clone, Debug)]
3646pub struct PathSummary<S> {
3647 pub max_path: Arc<RelPath>,
3648 pub item_summary: S,
3649}
3650
3651impl<S: Summary> Summary for PathSummary<S> {
3652 type Context<'a> = S::Context<'a>;
3653
3654 fn zero(cx: Self::Context<'_>) -> Self {
3655 Self {
3656 max_path: RelPath::empty().into(),
3657 item_summary: S::zero(cx),
3658 }
3659 }
3660
3661 fn add_summary(&mut self, rhs: &Self, cx: Self::Context<'_>) {
3662 self.max_path = rhs.max_path.clone();
3663 self.item_summary.add_summary(&rhs.item_summary, cx);
3664 }
3665}
3666
3667impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathProgress<'a> {
3668 fn zero(_: <PathSummary<S> as Summary>::Context<'_>) -> Self {
3669 Self {
3670 max_path: RelPath::empty(),
3671 }
3672 }
3673
3674 fn add_summary(
3675 &mut self,
3676 summary: &'a PathSummary<S>,
3677 _: <PathSummary<S> as Summary>::Context<'_>,
3678 ) {
3679 self.max_path = summary.max_path.as_ref()
3680 }
3681}
3682
3683impl<'a> sum_tree::Dimension<'a, PathSummary<GitSummary>> for GitSummary {
3684 fn zero(_cx: ()) -> Self {
3685 Default::default()
3686 }
3687
3688 fn add_summary(&mut self, summary: &'a PathSummary<GitSummary>, _: ()) {
3689 *self += summary.item_summary
3690 }
3691}
3692
3693impl<'a>
3694 sum_tree::SeekTarget<'a, PathSummary<GitSummary>, Dimensions<TraversalProgress<'a>, GitSummary>>
3695 for PathTarget<'_>
3696{
3697 fn cmp(
3698 &self,
3699 cursor_location: &Dimensions<TraversalProgress<'a>, GitSummary>,
3700 _: (),
3701 ) -> Ordering {
3702 self.cmp_path(cursor_location.0.max_path)
3703 }
3704}
3705
3706impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathKey {
3707 fn zero(_: S::Context<'_>) -> Self {
3708 Default::default()
3709 }
3710
3711 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3712 self.0 = summary.max_path.clone();
3713 }
3714}
3715
3716impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for TraversalProgress<'a> {
3717 fn zero(_cx: S::Context<'_>) -> Self {
3718 Default::default()
3719 }
3720
3721 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3722 self.max_path = summary.max_path.as_ref();
3723 }
3724}
3725
3726impl Entry {
3727 fn new(
3728 path: Arc<RelPath>,
3729 metadata: &fs::Metadata,
3730 id: ProjectEntryId,
3731 root_char_bag: CharBag,
3732 canonical_path: Option<Arc<Path>>,
3733 ) -> Self {
3734 let char_bag = char_bag_for_path(root_char_bag, &path);
3735 Self {
3736 id,
3737 kind: if metadata.is_dir {
3738 EntryKind::PendingDir
3739 } else {
3740 EntryKind::File
3741 },
3742 path,
3743 inode: metadata.inode,
3744 mtime: Some(metadata.mtime),
3745 size: metadata.len,
3746 canonical_path,
3747 is_ignored: false,
3748 is_hidden: false,
3749 is_always_included: false,
3750 is_external: false,
3751 is_private: false,
3752 char_bag,
3753 is_fifo: metadata.is_fifo,
3754 }
3755 }
3756
3757 pub fn is_created(&self) -> bool {
3758 self.mtime.is_some()
3759 }
3760
3761 pub fn is_dir(&self) -> bool {
3762 self.kind.is_dir()
3763 }
3764
3765 pub fn is_file(&self) -> bool {
3766 self.kind.is_file()
3767 }
3768}
3769
3770impl EntryKind {
3771 pub fn is_dir(&self) -> bool {
3772 matches!(
3773 self,
3774 EntryKind::Dir | EntryKind::PendingDir | EntryKind::UnloadedDir
3775 )
3776 }
3777
3778 pub fn is_unloaded(&self) -> bool {
3779 matches!(self, EntryKind::UnloadedDir)
3780 }
3781
3782 pub fn is_file(&self) -> bool {
3783 matches!(self, EntryKind::File)
3784 }
3785}
3786
3787impl sum_tree::Item for Entry {
3788 type Summary = EntrySummary;
3789
3790 fn summary(&self, _cx: ()) -> Self::Summary {
3791 let non_ignored_count = if self.is_ignored && !self.is_always_included {
3792 0
3793 } else {
3794 1
3795 };
3796 let file_count;
3797 let non_ignored_file_count;
3798 if self.is_file() {
3799 file_count = 1;
3800 non_ignored_file_count = non_ignored_count;
3801 } else {
3802 file_count = 0;
3803 non_ignored_file_count = 0;
3804 }
3805
3806 EntrySummary {
3807 max_path: self.path.clone(),
3808 count: 1,
3809 non_ignored_count,
3810 file_count,
3811 non_ignored_file_count,
3812 }
3813 }
3814}
3815
3816impl sum_tree::KeyedItem for Entry {
3817 type Key = PathKey;
3818
3819 fn key(&self) -> Self::Key {
3820 PathKey(self.path.clone())
3821 }
3822}
3823
3824#[derive(Clone, Debug)]
3825pub struct EntrySummary {
3826 max_path: Arc<RelPath>,
3827 count: usize,
3828 non_ignored_count: usize,
3829 file_count: usize,
3830 non_ignored_file_count: usize,
3831}
3832
3833impl Default for EntrySummary {
3834 fn default() -> Self {
3835 Self {
3836 max_path: Arc::from(RelPath::empty()),
3837 count: 0,
3838 non_ignored_count: 0,
3839 file_count: 0,
3840 non_ignored_file_count: 0,
3841 }
3842 }
3843}
3844
3845impl sum_tree::ContextLessSummary for EntrySummary {
3846 fn zero() -> Self {
3847 Default::default()
3848 }
3849
3850 fn add_summary(&mut self, rhs: &Self) {
3851 self.max_path = rhs.max_path.clone();
3852 self.count += rhs.count;
3853 self.non_ignored_count += rhs.non_ignored_count;
3854 self.file_count += rhs.file_count;
3855 self.non_ignored_file_count += rhs.non_ignored_file_count;
3856 }
3857}
3858
3859#[derive(Clone, Debug)]
3860struct PathEntry {
3861 id: ProjectEntryId,
3862 path: Arc<RelPath>,
3863 is_ignored: bool,
3864 scan_id: usize,
3865}
3866
3867impl sum_tree::Item for PathEntry {
3868 type Summary = PathEntrySummary;
3869
3870 fn summary(&self, _cx: ()) -> Self::Summary {
3871 PathEntrySummary { max_id: self.id }
3872 }
3873}
3874
3875impl sum_tree::KeyedItem for PathEntry {
3876 type Key = ProjectEntryId;
3877
3878 fn key(&self) -> Self::Key {
3879 self.id
3880 }
3881}
3882
3883#[derive(Clone, Debug, Default)]
3884struct PathEntrySummary {
3885 max_id: ProjectEntryId,
3886}
3887
3888impl sum_tree::ContextLessSummary for PathEntrySummary {
3889 fn zero() -> Self {
3890 Default::default()
3891 }
3892
3893 fn add_summary(&mut self, summary: &Self) {
3894 self.max_id = summary.max_id;
3895 }
3896}
3897
3898impl<'a> sum_tree::Dimension<'a, PathEntrySummary> for ProjectEntryId {
3899 fn zero(_cx: ()) -> Self {
3900 Default::default()
3901 }
3902
3903 fn add_summary(&mut self, summary: &'a PathEntrySummary, _: ()) {
3904 *self = summary.max_id;
3905 }
3906}
3907
3908#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
3909pub struct PathKey(pub Arc<RelPath>);
3910
3911impl Default for PathKey {
3912 fn default() -> Self {
3913 Self(RelPath::empty().into())
3914 }
3915}
3916
3917impl<'a> sum_tree::Dimension<'a, EntrySummary> for PathKey {
3918 fn zero(_cx: ()) -> Self {
3919 Default::default()
3920 }
3921
3922 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
3923 self.0 = summary.max_path.clone();
3924 }
3925}
3926
3927struct BackgroundScanner {
3928 state: async_lock::Mutex<BackgroundScannerState>,
3929 fs: Arc<dyn Fs>,
3930 fs_case_sensitive: bool,
3931 status_updates_tx: UnboundedSender<ScanState>,
3932 executor: BackgroundExecutor,
3933 scan_requests_rx: channel::Receiver<ScanRequest>,
3934 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
3935 next_entry_id: Arc<AtomicUsize>,
3936 phase: BackgroundScannerPhase,
3937 watcher: Arc<dyn Watcher>,
3938 settings: WorktreeSettings,
3939 share_private_files: bool,
3940 /// Whether this is a single-file worktree (root is a file, not a directory).
3941 /// Used to determine if we should give up after repeated canonicalization failures.
3942 is_single_file: bool,
3943}
3944
3945#[derive(Copy, Clone, PartialEq)]
3946enum BackgroundScannerPhase {
3947 InitialScan,
3948 EventsReceivedDuringInitialScan,
3949 Events,
3950}
3951
3952impl BackgroundScanner {
3953 async fn run(&mut self, mut fs_events_rx: Pin<Box<dyn Send + Stream<Item = Vec<PathEvent>>>>) {
3954 let root_abs_path;
3955 let scanning_enabled;
3956 {
3957 let state = self.state.lock().await;
3958 root_abs_path = state.snapshot.abs_path.clone();
3959 scanning_enabled = state.scanning_enabled;
3960 }
3961
3962 // If the worktree root does not contain a git repository, then find
3963 // the git repository in an ancestor directory. Find any gitignore files
3964 // in ancestor directories.
3965 let repo = if scanning_enabled {
3966 let (ignores, exclude, repo) =
3967 discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await;
3968 self.state
3969 .lock()
3970 .await
3971 .snapshot
3972 .ignores_by_parent_abs_path
3973 .extend(ignores);
3974 if let Some(exclude) = exclude {
3975 self.state
3976 .lock()
3977 .await
3978 .snapshot
3979 .repo_exclude_by_work_dir_abs_path
3980 .insert(root_abs_path.as_path().into(), (exclude, false));
3981 }
3982
3983 repo
3984 } else {
3985 None
3986 };
3987
3988 let containing_git_repository = if let Some((ancestor_dot_git, work_directory)) = repo
3989 && scanning_enabled
3990 {
3991 maybe!(async {
3992 self.state
3993 .lock()
3994 .await
3995 .insert_git_repository_for_path(
3996 work_directory,
3997 ancestor_dot_git.clone().into(),
3998 self.fs.as_ref(),
3999 self.watcher.as_ref(),
4000 )
4001 .await
4002 .log_err()?;
4003 Some(ancestor_dot_git)
4004 })
4005 .await
4006 } else {
4007 None
4008 };
4009
4010 log::trace!("containing git repository: {containing_git_repository:?}");
4011
4012 let global_gitignore_file = paths::global_gitignore_path();
4013 let mut global_gitignore_events = if let Some(global_gitignore_path) =
4014 &global_gitignore_file
4015 && scanning_enabled
4016 {
4017 let is_file = self.fs.is_file(&global_gitignore_path).await;
4018 self.state.lock().await.snapshot.global_gitignore = if is_file {
4019 build_gitignore(global_gitignore_path, self.fs.as_ref())
4020 .await
4021 .ok()
4022 .map(Arc::new)
4023 } else {
4024 None
4025 };
4026 if is_file {
4027 self.fs
4028 .watch(global_gitignore_path, FS_WATCH_LATENCY)
4029 .await
4030 .0
4031 } else {
4032 Box::pin(futures::stream::pending())
4033 }
4034 } else {
4035 self.state.lock().await.snapshot.global_gitignore = None;
4036 Box::pin(futures::stream::pending())
4037 };
4038
4039 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4040 {
4041 let mut state = self.state.lock().await;
4042 state.snapshot.scan_id += 1;
4043 if let Some(mut root_entry) = state.snapshot.root_entry().cloned() {
4044 let ignore_stack = state
4045 .snapshot
4046 .ignore_stack_for_abs_path(root_abs_path.as_path(), true, self.fs.as_ref())
4047 .await;
4048 if ignore_stack.is_abs_path_ignored(root_abs_path.as_path(), true) {
4049 root_entry.is_ignored = true;
4050 let mut root_entry = root_entry.clone();
4051 state.reuse_entry_id(&mut root_entry);
4052 state
4053 .insert_entry(root_entry, self.fs.as_ref(), self.watcher.as_ref())
4054 .await;
4055 }
4056 if root_entry.is_dir() && state.scanning_enabled {
4057 state
4058 .enqueue_scan_dir(
4059 root_abs_path.as_path().into(),
4060 &root_entry,
4061 &scan_job_tx,
4062 self.fs.as_ref(),
4063 )
4064 .await;
4065 }
4066 }
4067 };
4068
4069 // Perform an initial scan of the directory.
4070 drop(scan_job_tx);
4071 self.scan_dirs(true, scan_job_rx).await;
4072 {
4073 let mut state = self.state.lock().await;
4074 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4075 }
4076
4077 self.send_status_update(false, SmallVec::new(), &[]).await;
4078
4079 // Process any any FS events that occurred while performing the initial scan.
4080 // For these events, update events cannot be as precise, because we didn't
4081 // have the previous state loaded yet.
4082 self.phase = BackgroundScannerPhase::EventsReceivedDuringInitialScan;
4083 if let Poll::Ready(Some(mut paths)) = futures::poll!(fs_events_rx.next()) {
4084 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
4085 paths.extend(more_paths);
4086 }
4087 self.process_events(
4088 paths
4089 .into_iter()
4090 .filter(|event| event.kind.is_some())
4091 .collect(),
4092 )
4093 .await;
4094 }
4095 if let Some(abs_path) = containing_git_repository {
4096 self.process_events(vec![PathEvent {
4097 path: abs_path,
4098 kind: Some(fs::PathEventKind::Changed),
4099 }])
4100 .await;
4101 }
4102
4103 // Continue processing events until the worktree is dropped.
4104 self.phase = BackgroundScannerPhase::Events;
4105
4106 loop {
4107 select_biased! {
4108 // Process any path refresh requests from the worktree. Prioritize
4109 // these before handling changes reported by the filesystem.
4110 request = self.next_scan_request().fuse() => {
4111 let Ok(request) = request else { break };
4112 if !self.process_scan_request(request, false).await {
4113 return;
4114 }
4115 }
4116
4117 path_prefix_request = self.path_prefixes_to_scan_rx.recv().fuse() => {
4118 let Ok(request) = path_prefix_request else { break };
4119 log::trace!("adding path prefix {:?}", request.path);
4120
4121 let did_scan = self.forcibly_load_paths(std::slice::from_ref(&request.path)).await;
4122 if did_scan {
4123 let abs_path =
4124 {
4125 let mut state = self.state.lock().await;
4126 state.path_prefixes_to_scan.insert(request.path.clone());
4127 state.snapshot.absolutize(&request.path)
4128 };
4129
4130 if let Some(abs_path) = self.fs.canonicalize(&abs_path).await.log_err() {
4131 self.process_events(vec![PathEvent {
4132 path: abs_path,
4133 kind: Some(fs::PathEventKind::Changed),
4134 }])
4135 .await;
4136 }
4137 }
4138 self.send_status_update(false, request.done, &[]).await;
4139 }
4140
4141 paths = fs_events_rx.next().fuse() => {
4142 let Some(mut paths) = paths else { break };
4143 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
4144 paths.extend(more_paths);
4145 }
4146 self.process_events(paths.into_iter().filter(|event| event.kind.is_some()).collect()).await;
4147 }
4148
4149 _ = global_gitignore_events.next().fuse() => {
4150 if let Some(path) = &global_gitignore_file {
4151 self.update_global_gitignore(&path).await;
4152 }
4153 }
4154 }
4155 }
4156 }
4157
4158 async fn process_scan_request(&self, mut request: ScanRequest, scanning: bool) -> bool {
4159 log::debug!("rescanning paths {:?}", request.relative_paths);
4160
4161 request.relative_paths.sort_unstable();
4162 self.forcibly_load_paths(&request.relative_paths).await;
4163
4164 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4165 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4166 let root_canonical_path = match &root_canonical_path {
4167 Ok(path) => SanitizedPath::new(path),
4168 Err(err) => {
4169 log::error!("failed to canonicalize root path {root_path:?}: {err:#}");
4170 return true;
4171 }
4172 };
4173 let abs_paths = request
4174 .relative_paths
4175 .iter()
4176 .map(|path| {
4177 if path.file_name().is_some() {
4178 root_canonical_path.as_path().join(path.as_std_path())
4179 } else {
4180 root_canonical_path.as_path().to_path_buf()
4181 }
4182 })
4183 .collect::<Vec<_>>();
4184
4185 {
4186 let mut state = self.state.lock().await;
4187 let is_idle = state.snapshot.completed_scan_id == state.snapshot.scan_id;
4188 state.snapshot.scan_id += 1;
4189 if is_idle {
4190 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4191 }
4192 }
4193
4194 self.reload_entries_for_paths(
4195 &root_path,
4196 &root_canonical_path,
4197 &request.relative_paths,
4198 abs_paths,
4199 None,
4200 )
4201 .await;
4202
4203 self.send_status_update(scanning, request.done, &[]).await
4204 }
4205
4206 fn normalized_events_for_worktree(
4207 state: &BackgroundScannerState,
4208 root_canonical_path: &SanitizedPath,
4209 mut events: Vec<PathEvent>,
4210 ) -> Vec<PathEvent> {
4211 if state.symlink_paths_by_target.is_empty() {
4212 return events;
4213 }
4214 let mut mapped_events = Vec::new();
4215
4216 events.retain(|event| {
4217 let abs_path = SanitizedPath::new(&event.path);
4218
4219 let mut best_match: Option<(&Arc<Path>, &SmallVec<[Arc<RelPath>; 1]>)> = None;
4220 let mut best_depth = 0;
4221 for (target_root, symlink_paths) in &state.symlink_paths_by_target {
4222 if abs_path.as_path().starts_with(target_root.as_ref()) {
4223 let depth = target_root.as_ref().components().count();
4224 if depth > best_depth {
4225 best_depth = depth;
4226 best_match = Some((target_root, symlink_paths));
4227 }
4228 }
4229 }
4230
4231 let Some((target_root, symlink_paths)) = best_match else {
4232 return true;
4233 };
4234
4235 let Ok(suffix) = abs_path.as_path().strip_prefix(target_root.as_ref()) else {
4236 return true;
4237 };
4238
4239 // If the symlink's real target is outside this worktree, the original path
4240 // isn't visible to the worktree. Keep only the remapped symlink events.
4241 let keep_original = target_root.starts_with(root_canonical_path.as_path());
4242
4243 for symlink_path in symlink_paths {
4244 let mapped_path = if suffix.as_os_str().is_empty() {
4245 root_canonical_path
4246 .as_path()
4247 .join(symlink_path.as_std_path())
4248 } else {
4249 root_canonical_path
4250 .as_path()
4251 .join(symlink_path.as_std_path())
4252 .join(suffix)
4253 };
4254 if mapped_path != event.path {
4255 mapped_events.push(PathEvent {
4256 path: mapped_path,
4257 kind: event.kind,
4258 });
4259 }
4260 }
4261 keep_original
4262 });
4263 events.extend(mapped_events);
4264 events
4265 }
4266
4267 async fn process_events(&self, mut events: Vec<PathEvent>) {
4268 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4269 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4270 let root_canonical_path = match &root_canonical_path {
4271 Ok(path) => SanitizedPath::new(path),
4272 Err(err) => {
4273 let new_path = self
4274 .state
4275 .lock()
4276 .await
4277 .snapshot
4278 .root_file_handle
4279 .clone()
4280 .and_then(|handle| match handle.current_path(&self.fs) {
4281 Ok(new_path) => Some(new_path),
4282 Err(e) => {
4283 log::error!("Failed to refresh worktree root path: {e:#}");
4284 None
4285 }
4286 })
4287 .map(|path| SanitizedPath::new_arc(&path))
4288 .filter(|new_path| *new_path != root_path);
4289
4290 if let Some(new_path) = new_path {
4291 log::info!(
4292 "root renamed from {:?} to {:?}",
4293 root_path.as_path(),
4294 new_path.as_path(),
4295 );
4296 self.status_updates_tx
4297 .unbounded_send(ScanState::RootUpdated { new_path })
4298 .ok();
4299 } else {
4300 log::error!("root path could not be canonicalized: {err:#}");
4301
4302 // For single-file worktrees, if we can't canonicalize and the file handle
4303 // fallback also failed, the file is gone - close the worktree
4304 if self.is_single_file {
4305 log::info!(
4306 "single-file worktree root {:?} no longer exists, marking as deleted",
4307 root_path.as_path()
4308 );
4309 self.status_updates_tx
4310 .unbounded_send(ScanState::RootDeleted)
4311 .ok();
4312 }
4313 }
4314 return;
4315 }
4316 };
4317
4318 {
4319 let state = self.state.lock().await;
4320 events = Self::normalized_events_for_worktree(&state, &root_canonical_path, events);
4321 }
4322
4323 // Certain directories may have FS changes, but do not lead to git data changes that Zed cares about.
4324 // Ignore these, to avoid Zed unnecessarily rescanning git metadata.
4325 let skipped_files_in_dot_git = [COMMIT_MESSAGE, INDEX_LOCK];
4326 let skipped_dirs_in_dot_git = [FSMONITOR_DAEMON, LFS_DIR];
4327
4328 let mut relative_paths = Vec::with_capacity(events.len());
4329 let mut dot_git_abs_paths = Vec::new();
4330 let mut work_dirs_needing_exclude_update = Vec::new();
4331 events.sort_unstable_by(|left, right| left.path.cmp(&right.path));
4332 events.dedup_by(|left, right| {
4333 if left.path == right.path {
4334 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4335 right.kind = left.kind;
4336 }
4337 true
4338 } else if left.path.starts_with(&right.path) {
4339 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4340 right.kind = left.kind;
4341 }
4342 true
4343 } else {
4344 false
4345 }
4346 });
4347 {
4348 let snapshot = &self.state.lock().await.snapshot;
4349
4350 let mut ranges_to_drop = SmallVec::<[Range<usize>; 4]>::new();
4351
4352 fn skip_ix(ranges: &mut SmallVec<[Range<usize>; 4]>, ix: usize) {
4353 if let Some(last_range) = ranges.last_mut()
4354 && last_range.end == ix
4355 {
4356 last_range.end += 1;
4357 } else {
4358 ranges.push(ix..ix + 1);
4359 }
4360 }
4361
4362 for (ix, event) in events.iter().enumerate() {
4363 let abs_path = SanitizedPath::new(&event.path);
4364
4365 let mut is_git_related = false;
4366 let mut dot_git_paths = None;
4367
4368 for ancestor in abs_path.as_path().ancestors() {
4369 if is_dot_git(ancestor, self.fs.as_ref()).await {
4370 let path_in_git_dir = abs_path
4371 .as_path()
4372 .strip_prefix(ancestor)
4373 .expect("stripping off the ancestor");
4374 dot_git_paths = Some((ancestor.to_owned(), path_in_git_dir.to_owned()));
4375 break;
4376 }
4377 }
4378
4379 if let Some((dot_git_abs_path, path_in_git_dir)) = dot_git_paths {
4380 // We ignore `""` as well, as that is going to be the
4381 // `.git` folder itself. WE do not care about it, if
4382 // there are changes within we will see them, we need
4383 // this ignore to prevent us from accidentally observing
4384 // the ignored created file due to the events not being
4385 // empty after filtering.
4386
4387 let is_dot_git_changed = {
4388 path_in_git_dir == Path::new("")
4389 && event.kind == Some(PathEventKind::Changed)
4390 && self.fs.is_dir(abs_path.as_path()).await
4391 };
4392 let condition = skipped_files_in_dot_git.iter().any(|skipped| {
4393 OsStr::new(skipped) == path_in_git_dir.as_path().as_os_str()
4394 }) || skipped_dirs_in_dot_git
4395 .iter()
4396 .any(|skipped_git_subdir| path_in_git_dir.starts_with(skipped_git_subdir))
4397 || is_dot_git_changed;
4398 if condition {
4399 log::debug!(
4400 "ignoring event {abs_path:?} as it's in the .git directory among skipped files or directories"
4401 );
4402 skip_ix(&mut ranges_to_drop, ix);
4403 continue;
4404 }
4405
4406 is_git_related = true;
4407 if !dot_git_abs_paths.contains(&dot_git_abs_path) {
4408 dot_git_abs_paths.push(dot_git_abs_path);
4409 }
4410 }
4411
4412 let relative_path = if let Ok(path) = abs_path.strip_prefix(&root_canonical_path)
4413 && let Ok(path) = RelPath::new(path, PathStyle::local())
4414 {
4415 path
4416 } else {
4417 if is_git_related {
4418 log::debug!(
4419 "ignoring event {abs_path:?}, since it's in git dir outside of root path {root_canonical_path:?}",
4420 );
4421 } else {
4422 log::error!(
4423 "ignoring event {abs_path:?} outside of root path {root_canonical_path:?}",
4424 );
4425 }
4426 skip_ix(&mut ranges_to_drop, ix);
4427 continue;
4428 };
4429
4430 let absolute_path = abs_path.to_path_buf();
4431 if absolute_path.ends_with(Path::new(DOT_GIT).join(REPO_EXCLUDE)) {
4432 if let Some(repository) = snapshot
4433 .git_repositories
4434 .values()
4435 .find(|repo| repo.common_dir_abs_path.join(REPO_EXCLUDE) == absolute_path)
4436 {
4437 work_dirs_needing_exclude_update
4438 .push(repository.work_directory_abs_path.clone());
4439 }
4440 }
4441
4442 if abs_path.file_name() == Some(OsStr::new(GITIGNORE)) {
4443 for (_, repo) in snapshot
4444 .git_repositories
4445 .iter()
4446 .filter(|(_, repo)| repo.directory_contains(&relative_path))
4447 {
4448 if !dot_git_abs_paths.iter().any(|dot_git_abs_path| {
4449 dot_git_abs_path == repo.common_dir_abs_path.as_ref()
4450 }) {
4451 dot_git_abs_paths.push(repo.common_dir_abs_path.to_path_buf());
4452 }
4453 }
4454 }
4455
4456 let parent_dir_is_loaded = relative_path.parent().is_none_or(|parent| {
4457 snapshot
4458 .entry_for_path(parent)
4459 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4460 });
4461 if !parent_dir_is_loaded {
4462 log::debug!("ignoring event {relative_path:?} within unloaded directory");
4463 skip_ix(&mut ranges_to_drop, ix);
4464 continue;
4465 }
4466
4467 if self.settings.is_path_excluded(&relative_path) {
4468 if !is_git_related {
4469 log::debug!("ignoring FS event for excluded path {relative_path:?}");
4470 }
4471 skip_ix(&mut ranges_to_drop, ix);
4472 continue;
4473 }
4474
4475 relative_paths.push(EventRoot {
4476 path: relative_path.into_arc(),
4477 was_rescanned: matches!(event.kind, Some(fs::PathEventKind::Rescan)),
4478 });
4479 }
4480
4481 for range_to_drop in ranges_to_drop.into_iter().rev() {
4482 events.drain(range_to_drop);
4483 }
4484 }
4485
4486 if relative_paths.is_empty() && dot_git_abs_paths.is_empty() {
4487 return;
4488 }
4489
4490 if !work_dirs_needing_exclude_update.is_empty() {
4491 let mut state = self.state.lock().await;
4492 for work_dir_abs_path in work_dirs_needing_exclude_update {
4493 if let Some((_, needs_update)) = state
4494 .snapshot
4495 .repo_exclude_by_work_dir_abs_path
4496 .get_mut(&work_dir_abs_path)
4497 {
4498 *needs_update = true;
4499 }
4500 }
4501 }
4502
4503 self.state.lock().await.snapshot.scan_id += 1;
4504
4505 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4506 log::debug!(
4507 "received fs events {:?}",
4508 relative_paths
4509 .iter()
4510 .map(|event_root| &event_root.path)
4511 .collect::<Vec<_>>()
4512 );
4513 self.reload_entries_for_paths(
4514 &root_path,
4515 &root_canonical_path,
4516 &relative_paths
4517 .iter()
4518 .map(|event_root| event_root.path.clone())
4519 .collect::<Vec<_>>(),
4520 events
4521 .into_iter()
4522 .map(|event| event.path)
4523 .collect::<Vec<_>>(),
4524 Some(scan_job_tx.clone()),
4525 )
4526 .await;
4527
4528 let affected_repo_roots = if !dot_git_abs_paths.is_empty() {
4529 self.update_git_repositories(dot_git_abs_paths).await
4530 } else {
4531 Vec::new()
4532 };
4533
4534 {
4535 let mut ignores_to_update = self.ignores_needing_update().await;
4536 ignores_to_update.extend(affected_repo_roots);
4537 let ignores_to_update = self.order_ignores(ignores_to_update).await;
4538 let snapshot = self.state.lock().await.snapshot.clone();
4539 self.update_ignore_statuses_for_paths(scan_job_tx, snapshot, ignores_to_update)
4540 .await;
4541 self.scan_dirs(false, scan_job_rx).await;
4542 }
4543
4544 {
4545 let mut state = self.state.lock().await;
4546 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4547 for (_, entry) in mem::take(&mut state.removed_entries) {
4548 state.scanned_dirs.remove(&entry.id);
4549 }
4550 }
4551 self.send_status_update(false, SmallVec::new(), &relative_paths)
4552 .await;
4553 }
4554
4555 async fn update_global_gitignore(&self, abs_path: &Path) {
4556 let ignore = build_gitignore(abs_path, self.fs.as_ref())
4557 .await
4558 .log_err()
4559 .map(Arc::new);
4560 let (prev_snapshot, ignore_stack, abs_path) = {
4561 let mut state = self.state.lock().await;
4562 state.snapshot.global_gitignore = ignore;
4563 let abs_path = state.snapshot.abs_path().clone();
4564 let ignore_stack = state
4565 .snapshot
4566 .ignore_stack_for_abs_path(&abs_path, true, self.fs.as_ref())
4567 .await;
4568 (state.snapshot.clone(), ignore_stack, abs_path)
4569 };
4570 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4571 self.update_ignore_statuses_for_paths(
4572 scan_job_tx,
4573 prev_snapshot,
4574 vec![(abs_path, ignore_stack)],
4575 )
4576 .await;
4577 self.scan_dirs(false, scan_job_rx).await;
4578 self.send_status_update(false, SmallVec::new(), &[]).await;
4579 }
4580
4581 async fn forcibly_load_paths(&self, paths: &[Arc<RelPath>]) -> bool {
4582 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4583 {
4584 let mut state = self.state.lock().await;
4585 let root_path = state.snapshot.abs_path.clone();
4586 for path in paths {
4587 for ancestor in path.ancestors() {
4588 if let Some(entry) = state.snapshot.entry_for_path(ancestor)
4589 && entry.kind == EntryKind::UnloadedDir
4590 {
4591 let abs_path = if entry.is_external {
4592 entry
4593 .canonical_path
4594 .as_ref()
4595 .map(|path| path.as_ref().to_path_buf())
4596 .unwrap_or_else(|| root_path.join(ancestor.as_std_path()))
4597 } else {
4598 root_path.join(ancestor.as_std_path())
4599 };
4600 state
4601 .enqueue_scan_dir(
4602 abs_path.into(),
4603 entry,
4604 &scan_job_tx,
4605 self.fs.as_ref(),
4606 )
4607 .await;
4608 state.paths_to_scan.insert(path.clone());
4609 break;
4610 }
4611 }
4612 }
4613 drop(scan_job_tx);
4614 }
4615 while let Ok(job) = scan_job_rx.recv().await {
4616 self.scan_dir(&job).await.log_err();
4617 }
4618
4619 !mem::take(&mut self.state.lock().await.paths_to_scan).is_empty()
4620 }
4621
4622 async fn scan_dirs(
4623 &self,
4624 enable_progress_updates: bool,
4625 scan_jobs_rx: channel::Receiver<ScanJob>,
4626 ) {
4627 if self
4628 .status_updates_tx
4629 .unbounded_send(ScanState::Started)
4630 .is_err()
4631 {
4632 return;
4633 }
4634
4635 let progress_update_count = AtomicUsize::new(0);
4636 self.executor
4637 .scoped_priority(Priority::Low, |scope| {
4638 for _ in 0..self.executor.num_cpus() {
4639 scope.spawn(async {
4640 let mut last_progress_update_count = 0;
4641 let progress_update_timer = self.progress_timer(enable_progress_updates).fuse();
4642 futures::pin_mut!(progress_update_timer);
4643
4644 loop {
4645 select_biased! {
4646 // Process any path refresh requests before moving on to process
4647 // the scan queue, so that user operations are prioritized.
4648 request = self.next_scan_request().fuse() => {
4649 let Ok(request) = request else { break };
4650 if !self.process_scan_request(request, true).await {
4651 return;
4652 }
4653 }
4654
4655 // Send periodic progress updates to the worktree. Use an atomic counter
4656 // to ensure that only one of the workers sends a progress update after
4657 // the update interval elapses.
4658 _ = progress_update_timer => {
4659 match progress_update_count.compare_exchange(
4660 last_progress_update_count,
4661 last_progress_update_count + 1,
4662 SeqCst,
4663 SeqCst
4664 ) {
4665 Ok(_) => {
4666 last_progress_update_count += 1;
4667 self.send_status_update(true, SmallVec::new(), &[])
4668 .await;
4669 }
4670 Err(count) => {
4671 last_progress_update_count = count;
4672 }
4673 }
4674 progress_update_timer.set(self.progress_timer(enable_progress_updates).fuse());
4675 }
4676
4677 // Recursively load directories from the file system.
4678 job = scan_jobs_rx.recv().fuse() => {
4679 let Ok(job) = job else { break };
4680 if let Err(err) = self.scan_dir(&job).await
4681 && job.path.is_empty() {
4682 log::error!("error scanning directory {:?}: {}", job.abs_path, err);
4683 }
4684 }
4685 }
4686 }
4687 });
4688 }
4689 })
4690 .await;
4691 }
4692
4693 async fn send_status_update(
4694 &self,
4695 scanning: bool,
4696 barrier: SmallVec<[barrier::Sender; 1]>,
4697 event_roots: &[EventRoot],
4698 ) -> bool {
4699 let mut state = self.state.lock().await;
4700 if state.changed_paths.is_empty() && event_roots.is_empty() && scanning {
4701 return true;
4702 }
4703
4704 let merged_event_roots = merge_event_roots(&state.changed_paths, event_roots);
4705
4706 let new_snapshot = state.snapshot.clone();
4707 let old_snapshot = mem::replace(&mut state.prev_snapshot, new_snapshot.snapshot.clone());
4708 let changes = build_diff(
4709 self.phase,
4710 &old_snapshot,
4711 &new_snapshot,
4712 &merged_event_roots,
4713 );
4714 state.changed_paths.clear();
4715
4716 self.status_updates_tx
4717 .unbounded_send(ScanState::Updated {
4718 snapshot: new_snapshot,
4719 changes,
4720 scanning,
4721 barrier,
4722 })
4723 .is_ok()
4724 }
4725
4726 async fn scan_dir(&self, job: &ScanJob) -> Result<()> {
4727 let root_abs_path;
4728 let root_char_bag;
4729 {
4730 let snapshot = &self.state.lock().await.snapshot;
4731 if self.settings.is_path_excluded(&job.path) {
4732 log::error!("skipping excluded directory {:?}", job.path);
4733 return Ok(());
4734 }
4735 log::trace!("scanning directory {:?}", job.path);
4736 root_abs_path = snapshot.abs_path().clone();
4737 root_char_bag = snapshot.root_char_bag;
4738 }
4739
4740 let next_entry_id = self.next_entry_id.clone();
4741 let mut ignore_stack = job.ignore_stack.clone();
4742 let mut new_ignore = None;
4743 let mut root_canonical_path = None;
4744 let mut new_entries: Vec<Entry> = Vec::new();
4745 let mut new_jobs: Vec<Option<ScanJob>> = Vec::new();
4746 let mut child_paths = self
4747 .fs
4748 .read_dir(&job.abs_path)
4749 .await?
4750 .filter_map(|entry| async {
4751 match entry {
4752 Ok(entry) => Some(entry),
4753 Err(error) => {
4754 log::error!("error processing entry {:?}", error);
4755 None
4756 }
4757 }
4758 })
4759 .collect::<Vec<_>>()
4760 .await;
4761
4762 // Ensure that .git and .gitignore are processed first.
4763 swap_to_front(&mut child_paths, GITIGNORE);
4764 swap_to_front(&mut child_paths, DOT_GIT);
4765
4766 if let Some(path) = child_paths.first()
4767 && path.ends_with(DOT_GIT)
4768 {
4769 ignore_stack.repo_root = Some(job.abs_path.clone());
4770 }
4771
4772 for child_abs_path in child_paths {
4773 let child_abs_path: Arc<Path> = child_abs_path.into();
4774 let child_name = child_abs_path.file_name().unwrap();
4775 let Some(child_path) = child_name
4776 .to_str()
4777 .and_then(|name| Some(job.path.join(RelPath::unix(name).ok()?)))
4778 else {
4779 continue;
4780 };
4781
4782 if child_name == DOT_GIT {
4783 let mut state = self.state.lock().await;
4784 state
4785 .insert_git_repository(
4786 child_path.clone(),
4787 self.fs.as_ref(),
4788 self.watcher.as_ref(),
4789 )
4790 .await;
4791 } else if child_name == GITIGNORE {
4792 match build_gitignore(&child_abs_path, self.fs.as_ref()).await {
4793 Ok(ignore) => {
4794 let ignore = Arc::new(ignore);
4795 ignore_stack = ignore_stack
4796 .append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4797 new_ignore = Some(ignore);
4798 }
4799 Err(error) => {
4800 log::error!(
4801 "error loading .gitignore file {:?} - {:?}",
4802 child_name,
4803 error
4804 );
4805 }
4806 }
4807 }
4808
4809 if self.settings.is_path_excluded(&child_path) {
4810 log::debug!("skipping excluded child entry {child_path:?}");
4811 self.state
4812 .lock()
4813 .await
4814 .remove_path(&child_path, self.watcher.as_ref());
4815 continue;
4816 }
4817
4818 let child_metadata = match self.fs.metadata(&child_abs_path).await {
4819 Ok(Some(metadata)) => metadata,
4820 Ok(None) => continue,
4821 Err(err) => {
4822 log::error!("error processing {:?}: {err:#}", child_abs_path.display());
4823 continue;
4824 }
4825 };
4826
4827 let mut child_entry = Entry::new(
4828 child_path.clone(),
4829 &child_metadata,
4830 ProjectEntryId::new(&next_entry_id),
4831 root_char_bag,
4832 None,
4833 );
4834
4835 if job.is_external {
4836 child_entry.is_external = true;
4837 } else if child_metadata.is_symlink {
4838 let canonical_path = match self.fs.canonicalize(&child_abs_path).await {
4839 Ok(path) => path,
4840 Err(err) => {
4841 log::error!("error reading target of symlink {child_abs_path:?}: {err:#}",);
4842 continue;
4843 }
4844 };
4845
4846 // lazily canonicalize the root path in order to determine if
4847 // symlinks point outside of the worktree.
4848 let root_canonical_path = match &root_canonical_path {
4849 Some(path) => path,
4850 None => match self.fs.canonicalize(&root_abs_path).await {
4851 Ok(path) => root_canonical_path.insert(path),
4852 Err(err) => {
4853 log::error!("error canonicalizing root {:?}: {:?}", root_abs_path, err);
4854 continue;
4855 }
4856 },
4857 };
4858
4859 if !canonical_path.starts_with(root_canonical_path) {
4860 child_entry.is_external = true;
4861 }
4862
4863 if child_metadata.is_dir {
4864 let mut state = self.state.lock().await;
4865 let paths = state
4866 .symlink_paths_by_target
4867 .entry(Arc::from(canonical_path.clone()))
4868 .or_default();
4869 if !paths.iter().any(|path| path == &child_path) {
4870 paths.push(child_path.clone());
4871 }
4872 }
4873
4874 child_entry.canonical_path = Some(canonical_path.into());
4875 }
4876
4877 if child_entry.is_dir() {
4878 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, true);
4879 child_entry.is_always_included =
4880 self.settings.is_path_always_included(&child_path, true);
4881
4882 // Avoid recursing until crash in the case of a recursive symlink
4883 if job.ancestor_inodes.contains(&child_entry.inode) {
4884 new_jobs.push(None);
4885 } else {
4886 let mut ancestor_inodes = job.ancestor_inodes.clone();
4887 ancestor_inodes.insert(child_entry.inode);
4888
4889 new_jobs.push(Some(ScanJob {
4890 abs_path: child_abs_path.clone(),
4891 path: child_path,
4892 is_external: child_entry.is_external,
4893 ignore_stack: if child_entry.is_ignored {
4894 IgnoreStack::all()
4895 } else {
4896 ignore_stack.clone()
4897 },
4898 ancestor_inodes,
4899 scan_queue: job.scan_queue.clone(),
4900 }));
4901 }
4902 } else {
4903 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, false);
4904 child_entry.is_always_included =
4905 self.settings.is_path_always_included(&child_path, false);
4906 }
4907
4908 {
4909 let relative_path = job
4910 .path
4911 .join(RelPath::unix(child_name.to_str().unwrap()).unwrap());
4912 if self.is_path_private(&relative_path) {
4913 log::debug!("detected private file: {relative_path:?}");
4914 child_entry.is_private = true;
4915 }
4916 if self.settings.is_path_hidden(&relative_path) {
4917 log::debug!("detected hidden file: {relative_path:?}");
4918 child_entry.is_hidden = true;
4919 }
4920 }
4921
4922 new_entries.push(child_entry);
4923 }
4924
4925 let mut state = self.state.lock().await;
4926
4927 // Identify any subdirectories that should not be scanned.
4928 let mut job_ix = 0;
4929 for entry in &mut new_entries {
4930 state.reuse_entry_id(entry);
4931 if entry.is_dir() {
4932 if state.should_scan_directory(entry) {
4933 job_ix += 1;
4934 } else {
4935 log::debug!("defer scanning directory {:?}", entry.path);
4936 entry.kind = EntryKind::UnloadedDir;
4937 new_jobs.remove(job_ix);
4938 }
4939 }
4940 if entry.is_always_included {
4941 state
4942 .snapshot
4943 .always_included_entries
4944 .push(entry.path.clone());
4945 }
4946 }
4947
4948 state.populate_dir(job.path.clone(), new_entries, new_ignore);
4949
4950 self.watcher.add(job.abs_path.as_ref()).log_err();
4951
4952 let entry_id = state
4953 .snapshot
4954 .entry_for_path(&job.path)
4955 .map(|entry| entry.id);
4956 if let Some(entry_id) = entry_id {
4957 state
4958 .watched_dir_abs_paths_by_entry_id
4959 .insert(entry_id, job.abs_path.clone());
4960 }
4961
4962 for new_job in new_jobs.into_iter().flatten() {
4963 job.scan_queue
4964 .try_send(new_job)
4965 .expect("channel is unbounded");
4966 }
4967
4968 Ok(())
4969 }
4970
4971 /// All list arguments should be sorted before calling this function
4972 async fn reload_entries_for_paths(
4973 &self,
4974 root_abs_path: &SanitizedPath,
4975 root_canonical_path: &SanitizedPath,
4976 relative_paths: &[Arc<RelPath>],
4977 abs_paths: Vec<PathBuf>,
4978 scan_queue_tx: Option<Sender<ScanJob>>,
4979 ) {
4980 // grab metadata for all requested paths
4981 let metadata = futures::future::join_all(
4982 abs_paths
4983 .iter()
4984 .map(|abs_path| async move {
4985 let metadata = self.fs.metadata(abs_path).await?;
4986 if let Some(metadata) = metadata {
4987 let canonical_path = self.fs.canonicalize(abs_path).await?;
4988
4989 // If we're on a case-insensitive filesystem (default on macOS), we want
4990 // to only ignore metadata for non-symlink files if their absolute-path matches
4991 // the canonical-path.
4992 // Because if not, this might be a case-only-renaming (`mv test.txt TEST.TXT`)
4993 // and we want to ignore the metadata for the old path (`test.txt`) so it's
4994 // treated as removed.
4995 if !self.fs_case_sensitive && !metadata.is_symlink {
4996 let canonical_file_name = canonical_path.file_name();
4997 let file_name = abs_path.file_name();
4998 if canonical_file_name != file_name {
4999 return Ok(None);
5000 }
5001 }
5002
5003 anyhow::Ok(Some((metadata, SanitizedPath::new_arc(&canonical_path))))
5004 } else {
5005 Ok(None)
5006 }
5007 })
5008 .collect::<Vec<_>>(),
5009 )
5010 .await;
5011
5012 let mut new_ancestor_repo = if relative_paths.iter().any(|path| path.is_empty()) {
5013 Some(discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await)
5014 } else {
5015 None
5016 };
5017
5018 let mut state = self.state.lock().await;
5019 let doing_recursive_update = scan_queue_tx.is_some();
5020
5021 // Remove any entries for paths that no longer exist or are being recursively
5022 // refreshed. Do this before adding any new entries, so that renames can be
5023 // detected regardless of the order of the paths.
5024 for (path, metadata) in relative_paths.iter().zip(metadata.iter()) {
5025 if matches!(metadata, Ok(None)) || doing_recursive_update {
5026 state.remove_path(path, self.watcher.as_ref());
5027 }
5028 }
5029
5030 for (path, metadata) in relative_paths.iter().zip(metadata.into_iter()) {
5031 let abs_path: Arc<Path> = root_abs_path.join(path.as_std_path()).into();
5032 match metadata {
5033 Ok(Some((metadata, canonical_path))) => {
5034 let ignore_stack = state
5035 .snapshot
5036 .ignore_stack_for_abs_path(&abs_path, metadata.is_dir, self.fs.as_ref())
5037 .await;
5038 let is_external = !canonical_path.starts_with(&root_canonical_path);
5039 let entry_id = state.entry_id_for(self.next_entry_id.as_ref(), path, &metadata);
5040 let mut fs_entry = Entry::new(
5041 path.clone(),
5042 &metadata,
5043 entry_id,
5044 state.snapshot.root_char_bag,
5045 if metadata.is_symlink {
5046 Some(canonical_path.as_path().to_path_buf().into())
5047 } else {
5048 None
5049 },
5050 );
5051
5052 let is_dir = fs_entry.is_dir();
5053 fs_entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, is_dir);
5054 fs_entry.is_external = is_external;
5055 fs_entry.is_private = self.is_path_private(path);
5056 fs_entry.is_always_included =
5057 self.settings.is_path_always_included(path, is_dir);
5058 fs_entry.is_hidden = self.settings.is_path_hidden(path);
5059
5060 if let (Some(scan_queue_tx), true) = (&scan_queue_tx, is_dir) {
5061 if state.should_scan_directory(&fs_entry)
5062 || (fs_entry.path.is_empty()
5063 && abs_path.file_name() == Some(OsStr::new(DOT_GIT)))
5064 {
5065 state
5066 .enqueue_scan_dir(
5067 abs_path,
5068 &fs_entry,
5069 scan_queue_tx,
5070 self.fs.as_ref(),
5071 )
5072 .await;
5073 } else {
5074 fs_entry.kind = EntryKind::UnloadedDir;
5075 }
5076 }
5077
5078 state
5079 .insert_entry(fs_entry.clone(), self.fs.as_ref(), self.watcher.as_ref())
5080 .await;
5081
5082 if path.is_empty()
5083 && let Some((ignores, exclude, repo)) = new_ancestor_repo.take()
5084 {
5085 log::trace!("updating ancestor git repository");
5086 state.snapshot.ignores_by_parent_abs_path.extend(ignores);
5087 if let Some((ancestor_dot_git, work_directory)) = repo {
5088 if let Some(exclude) = exclude {
5089 let work_directory_abs_path = self
5090 .state
5091 .lock()
5092 .await
5093 .snapshot
5094 .work_directory_abs_path(&work_directory);
5095
5096 state
5097 .snapshot
5098 .repo_exclude_by_work_dir_abs_path
5099 .insert(work_directory_abs_path.into(), (exclude, false));
5100 }
5101 state
5102 .insert_git_repository_for_path(
5103 work_directory,
5104 ancestor_dot_git.into(),
5105 self.fs.as_ref(),
5106 self.watcher.as_ref(),
5107 )
5108 .await
5109 .log_err();
5110 }
5111 }
5112 }
5113 Ok(None) => {
5114 self.remove_repo_path(path.clone(), &mut state.snapshot);
5115 }
5116 Err(err) => {
5117 log::error!("error reading file {abs_path:?} on event: {err:#}");
5118 }
5119 }
5120 }
5121
5122 util::extend_sorted(
5123 &mut state.changed_paths,
5124 relative_paths.iter().cloned(),
5125 usize::MAX,
5126 Ord::cmp,
5127 );
5128 }
5129
5130 fn remove_repo_path(&self, path: Arc<RelPath>, snapshot: &mut LocalSnapshot) -> Option<()> {
5131 if !path.components().any(|component| component == DOT_GIT)
5132 && let Some(local_repo) = snapshot.local_repo_for_work_directory_path(&path)
5133 {
5134 let id = local_repo.work_directory_id;
5135 log::debug!("remove repo path: {:?}", path);
5136 snapshot.git_repositories.remove(&id);
5137 return Some(());
5138 }
5139
5140 Some(())
5141 }
5142
5143 async fn update_ignore_statuses_for_paths(
5144 &self,
5145 scan_job_tx: Sender<ScanJob>,
5146 prev_snapshot: LocalSnapshot,
5147 ignores_to_update: Vec<(Arc<Path>, IgnoreStack)>,
5148 ) {
5149 let (ignore_queue_tx, ignore_queue_rx) = channel::unbounded();
5150 {
5151 for (parent_abs_path, ignore_stack) in ignores_to_update {
5152 ignore_queue_tx
5153 .send_blocking(UpdateIgnoreStatusJob {
5154 abs_path: parent_abs_path,
5155 ignore_stack,
5156 ignore_queue: ignore_queue_tx.clone(),
5157 scan_queue: scan_job_tx.clone(),
5158 })
5159 .unwrap();
5160 }
5161 }
5162 drop(ignore_queue_tx);
5163
5164 self.executor
5165 .scoped(|scope| {
5166 for _ in 0..self.executor.num_cpus() {
5167 scope.spawn(async {
5168 loop {
5169 select_biased! {
5170 // Process any path refresh requests before moving on to process
5171 // the queue of ignore statuses.
5172 request = self.next_scan_request().fuse() => {
5173 let Ok(request) = request else { break };
5174 if !self.process_scan_request(request, true).await {
5175 return;
5176 }
5177 }
5178
5179 // Recursively process directories whose ignores have changed.
5180 job = ignore_queue_rx.recv().fuse() => {
5181 let Ok(job) = job else { break };
5182 self.update_ignore_status(job, &prev_snapshot).await;
5183 }
5184 }
5185 }
5186 });
5187 }
5188 })
5189 .await;
5190 }
5191
5192 async fn ignores_needing_update(&self) -> Vec<Arc<Path>> {
5193 let mut ignores_to_update = Vec::new();
5194 let mut excludes_to_load: Vec<(Arc<Path>, PathBuf)> = Vec::new();
5195
5196 // First pass: collect updates and drop stale entries without awaiting.
5197 {
5198 let snapshot = &mut self.state.lock().await.snapshot;
5199 let abs_path = snapshot.abs_path.clone();
5200 let mut repo_exclude_keys_to_remove: Vec<Arc<Path>> = Vec::new();
5201
5202 for (work_dir_abs_path, (_, needs_update)) in
5203 snapshot.repo_exclude_by_work_dir_abs_path.iter_mut()
5204 {
5205 let repository = snapshot
5206 .git_repositories
5207 .iter()
5208 .find(|(_, repo)| &repo.work_directory_abs_path == work_dir_abs_path);
5209
5210 if *needs_update {
5211 *needs_update = false;
5212 ignores_to_update.push(work_dir_abs_path.clone());
5213
5214 if let Some((_, repository)) = repository {
5215 let exclude_abs_path = repository.common_dir_abs_path.join(REPO_EXCLUDE);
5216 excludes_to_load.push((work_dir_abs_path.clone(), exclude_abs_path));
5217 }
5218 }
5219
5220 if repository.is_none() {
5221 repo_exclude_keys_to_remove.push(work_dir_abs_path.clone());
5222 }
5223 }
5224
5225 for key in repo_exclude_keys_to_remove {
5226 snapshot.repo_exclude_by_work_dir_abs_path.remove(&key);
5227 }
5228
5229 snapshot
5230 .ignores_by_parent_abs_path
5231 .retain(|parent_abs_path, (_, needs_update)| {
5232 if let Ok(parent_path) = parent_abs_path.strip_prefix(abs_path.as_path())
5233 && let Some(parent_path) =
5234 RelPath::new(&parent_path, PathStyle::local()).log_err()
5235 {
5236 if *needs_update {
5237 *needs_update = false;
5238 if snapshot.snapshot.entry_for_path(&parent_path).is_some() {
5239 ignores_to_update.push(parent_abs_path.clone());
5240 }
5241 }
5242
5243 let ignore_path = parent_path.join(RelPath::unix(GITIGNORE).unwrap());
5244 if snapshot.snapshot.entry_for_path(&ignore_path).is_none() {
5245 return false;
5246 }
5247 }
5248 true
5249 });
5250 }
5251
5252 // Load gitignores asynchronously (outside the lock)
5253 let mut loaded_excludes: Vec<(Arc<Path>, Arc<Gitignore>)> = Vec::new();
5254 for (work_dir_abs_path, exclude_abs_path) in excludes_to_load {
5255 if let Ok(current_exclude) = build_gitignore(&exclude_abs_path, self.fs.as_ref()).await
5256 {
5257 loaded_excludes.push((work_dir_abs_path, Arc::new(current_exclude)));
5258 }
5259 }
5260
5261 // Second pass: apply updates.
5262 if !loaded_excludes.is_empty() {
5263 let snapshot = &mut self.state.lock().await.snapshot;
5264
5265 for (work_dir_abs_path, exclude) in loaded_excludes {
5266 if let Some((existing_exclude, _)) = snapshot
5267 .repo_exclude_by_work_dir_abs_path
5268 .get_mut(&work_dir_abs_path)
5269 {
5270 *existing_exclude = exclude;
5271 }
5272 }
5273 }
5274
5275 ignores_to_update
5276 }
5277
5278 async fn order_ignores(&self, mut ignores: Vec<Arc<Path>>) -> Vec<(Arc<Path>, IgnoreStack)> {
5279 let fs = self.fs.clone();
5280 let snapshot = self.state.lock().await.snapshot.clone();
5281 ignores.sort_unstable();
5282 let mut ignores_to_update = ignores.into_iter().peekable();
5283
5284 let mut result = vec![];
5285 while let Some(parent_abs_path) = ignores_to_update.next() {
5286 while ignores_to_update
5287 .peek()
5288 .map_or(false, |p| p.starts_with(&parent_abs_path))
5289 {
5290 ignores_to_update.next().unwrap();
5291 }
5292 let ignore_stack = snapshot
5293 .ignore_stack_for_abs_path(&parent_abs_path, true, fs.as_ref())
5294 .await;
5295 result.push((parent_abs_path, ignore_stack));
5296 }
5297
5298 result
5299 }
5300
5301 async fn update_ignore_status(&self, job: UpdateIgnoreStatusJob, snapshot: &LocalSnapshot) {
5302 log::trace!("update ignore status {:?}", job.abs_path);
5303
5304 let mut ignore_stack = job.ignore_stack;
5305 if let Some((ignore, _)) = snapshot.ignores_by_parent_abs_path.get(&job.abs_path) {
5306 ignore_stack =
5307 ignore_stack.append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
5308 }
5309
5310 let mut entries_by_id_edits = Vec::new();
5311 let mut entries_by_path_edits = Vec::new();
5312 let Some(path) = job
5313 .abs_path
5314 .strip_prefix(snapshot.abs_path.as_path())
5315 .map_err(|_| {
5316 anyhow::anyhow!(
5317 "Failed to strip prefix '{}' from path '{}'",
5318 snapshot.abs_path.as_path().display(),
5319 job.abs_path.display()
5320 )
5321 })
5322 .log_err()
5323 else {
5324 return;
5325 };
5326
5327 let Some(path) = RelPath::new(&path, PathStyle::local()).log_err() else {
5328 return;
5329 };
5330
5331 if let Ok(Some(metadata)) = self.fs.metadata(&job.abs_path.join(DOT_GIT)).await
5332 && metadata.is_dir
5333 {
5334 ignore_stack.repo_root = Some(job.abs_path.clone());
5335 }
5336
5337 for mut entry in snapshot.child_entries(&path).cloned() {
5338 let was_ignored = entry.is_ignored;
5339 let abs_path: Arc<Path> = snapshot.absolutize(&entry.path).into();
5340 entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, entry.is_dir());
5341
5342 if entry.is_dir() {
5343 let child_ignore_stack = if entry.is_ignored {
5344 IgnoreStack::all()
5345 } else {
5346 ignore_stack.clone()
5347 };
5348
5349 // Scan any directories that were previously ignored and weren't previously scanned.
5350 if was_ignored && !entry.is_ignored && entry.kind.is_unloaded() {
5351 let state = self.state.lock().await;
5352 if state.should_scan_directory(&entry) {
5353 state
5354 .enqueue_scan_dir(
5355 abs_path.clone(),
5356 &entry,
5357 &job.scan_queue,
5358 self.fs.as_ref(),
5359 )
5360 .await;
5361 }
5362 }
5363
5364 job.ignore_queue
5365 .send(UpdateIgnoreStatusJob {
5366 abs_path: abs_path.clone(),
5367 ignore_stack: child_ignore_stack,
5368 ignore_queue: job.ignore_queue.clone(),
5369 scan_queue: job.scan_queue.clone(),
5370 })
5371 .await
5372 .unwrap();
5373 }
5374
5375 if entry.is_ignored != was_ignored {
5376 let mut path_entry = snapshot.entries_by_id.get(&entry.id, ()).unwrap().clone();
5377 path_entry.scan_id = snapshot.scan_id;
5378 path_entry.is_ignored = entry.is_ignored;
5379 entries_by_id_edits.push(Edit::Insert(path_entry));
5380 entries_by_path_edits.push(Edit::Insert(entry));
5381 }
5382 }
5383
5384 let state = &mut self.state.lock().await;
5385 for edit in &entries_by_path_edits {
5386 if let Edit::Insert(entry) = edit
5387 && let Err(ix) = state.changed_paths.binary_search(&entry.path)
5388 {
5389 state.changed_paths.insert(ix, entry.path.clone());
5390 }
5391 }
5392
5393 state
5394 .snapshot
5395 .entries_by_path
5396 .edit(entries_by_path_edits, ());
5397 state.snapshot.entries_by_id.edit(entries_by_id_edits, ());
5398 }
5399
5400 async fn update_git_repositories(&self, dot_git_paths: Vec<PathBuf>) -> Vec<Arc<Path>> {
5401 log::trace!("reloading repositories: {dot_git_paths:?}");
5402 let mut state = self.state.lock().await;
5403 let scan_id = state.snapshot.scan_id;
5404 let mut affected_repo_roots = Vec::new();
5405 for dot_git_dir in dot_git_paths {
5406 let existing_repository_entry =
5407 state
5408 .snapshot
5409 .git_repositories
5410 .iter()
5411 .find_map(|(_, repo)| {
5412 let dot_git_dir = SanitizedPath::new(&dot_git_dir);
5413 if SanitizedPath::new(repo.common_dir_abs_path.as_ref()) == dot_git_dir
5414 || SanitizedPath::new(repo.repository_dir_abs_path.as_ref())
5415 == dot_git_dir
5416 {
5417 Some(repo.clone())
5418 } else {
5419 None
5420 }
5421 });
5422
5423 match existing_repository_entry {
5424 None => {
5425 let Ok(relative) = dot_git_dir.strip_prefix(state.snapshot.abs_path()) else {
5426 // A `.git` path outside the worktree root is not
5427 // ours to register. This happens legitimately when
5428 // `.git` is a gitfile pointing outside the worktree
5429 // (linked worktrees and submodules), and also when
5430 // a rescan of a linked worktree's commondir arrives
5431 // after the worktree's repository has already been
5432 // unregistered.
5433 continue;
5434 };
5435 affected_repo_roots.push(dot_git_dir.parent().unwrap().into());
5436 state
5437 .insert_git_repository(
5438 RelPath::new(relative, PathStyle::local())
5439 .unwrap()
5440 .into_arc(),
5441 self.fs.as_ref(),
5442 self.watcher.as_ref(),
5443 )
5444 .await;
5445 }
5446 Some(local_repository) => {
5447 state.snapshot.git_repositories.update(
5448 &local_repository.work_directory_id,
5449 |entry| {
5450 entry.git_dir_scan_id = scan_id;
5451 },
5452 );
5453 }
5454 };
5455 }
5456
5457 // Remove any git repositories whose .git entry no longer exists.
5458 let snapshot = &mut state.snapshot;
5459 let mut ids_to_preserve = HashSet::default();
5460 for (&work_directory_id, entry) in snapshot.git_repositories.iter() {
5461 let exists_in_snapshot =
5462 snapshot
5463 .entry_for_id(work_directory_id)
5464 .is_some_and(|entry| {
5465 snapshot
5466 .entry_for_path(&entry.path.join(RelPath::unix(DOT_GIT).unwrap()))
5467 .is_some()
5468 });
5469
5470 if exists_in_snapshot
5471 || matches!(
5472 self.fs.metadata(&entry.common_dir_abs_path).await,
5473 Ok(Some(_))
5474 )
5475 {
5476 ids_to_preserve.insert(work_directory_id);
5477 }
5478 }
5479
5480 snapshot
5481 .git_repositories
5482 .retain(|work_directory_id, entry| {
5483 let preserve = ids_to_preserve.contains(work_directory_id);
5484 if !preserve {
5485 affected_repo_roots.push(entry.dot_git_abs_path.parent().unwrap().into());
5486 snapshot
5487 .repo_exclude_by_work_dir_abs_path
5488 .remove(&entry.work_directory_abs_path);
5489 }
5490 preserve
5491 });
5492
5493 affected_repo_roots
5494 }
5495
5496 async fn progress_timer(&self, running: bool) {
5497 if !running {
5498 return futures::future::pending().await;
5499 }
5500
5501 #[cfg(feature = "test-support")]
5502 if self.fs.is_fake() {
5503 return self.executor.simulate_random_delay().await;
5504 }
5505
5506 self.executor.timer(FS_WATCH_LATENCY).await
5507 }
5508
5509 fn is_path_private(&self, path: &RelPath) -> bool {
5510 !self.share_private_files && self.settings.is_path_private(path)
5511 }
5512
5513 async fn next_scan_request(&self) -> Result<ScanRequest> {
5514 let mut request = self.scan_requests_rx.recv().await?;
5515 while let Ok(next_request) = self.scan_requests_rx.try_recv() {
5516 request.relative_paths.extend(next_request.relative_paths);
5517 request.done.extend(next_request.done);
5518 }
5519 Ok(request)
5520 }
5521}
5522
5523async fn discover_ancestor_git_repo(
5524 fs: Arc<dyn Fs>,
5525 root_abs_path: &SanitizedPath,
5526) -> (
5527 HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
5528 Option<Arc<Gitignore>>,
5529 Option<(PathBuf, WorkDirectory)>,
5530) {
5531 let mut exclude = None;
5532 let mut ignores = HashMap::default();
5533 for (index, ancestor) in root_abs_path.as_path().ancestors().enumerate() {
5534 if index != 0 {
5535 if ancestor == paths::home_dir() {
5536 // Unless $HOME is itself the worktree root, don't consider it as a
5537 // containing git repository---expensive and likely unwanted.
5538 break;
5539 } else if let Ok(ignore) = build_gitignore(&ancestor.join(GITIGNORE), fs.as_ref()).await
5540 {
5541 ignores.insert(ancestor.into(), (ignore.into(), false));
5542 }
5543 }
5544
5545 let ancestor_dot_git = ancestor.join(DOT_GIT);
5546 log::trace!("considering ancestor: {ancestor_dot_git:?}");
5547 // Check whether the directory or file called `.git` exists (in the
5548 // case of worktrees it's a file.)
5549 if fs
5550 .metadata(&ancestor_dot_git)
5551 .await
5552 .is_ok_and(|metadata| metadata.is_some())
5553 {
5554 if index != 0 {
5555 // We canonicalize, since the FS events use the canonicalized path.
5556 if let Some(ancestor_dot_git) = fs.canonicalize(&ancestor_dot_git).await.log_err() {
5557 let location_in_repo = root_abs_path
5558 .as_path()
5559 .strip_prefix(ancestor)
5560 .unwrap()
5561 .into();
5562 log::info!("inserting parent git repo for this worktree: {location_in_repo:?}");
5563 // We associate the external git repo with our root folder and
5564 // also mark where in the git repo the root folder is located.
5565 return (
5566 ignores,
5567 exclude,
5568 Some((
5569 ancestor_dot_git,
5570 WorkDirectory::AboveProject {
5571 absolute_path: ancestor.into(),
5572 location_in_repo,
5573 },
5574 )),
5575 );
5576 };
5577 }
5578
5579 let repo_exclude_abs_path = ancestor_dot_git.join(REPO_EXCLUDE);
5580 if let Ok(repo_exclude) = build_gitignore(&repo_exclude_abs_path, fs.as_ref()).await {
5581 exclude = Some(Arc::new(repo_exclude));
5582 }
5583
5584 // Reached root of git repository.
5585 break;
5586 }
5587 }
5588
5589 (ignores, exclude, None)
5590}
5591
5592fn merge_event_roots(changed_paths: &[Arc<RelPath>], event_roots: &[EventRoot]) -> Vec<EventRoot> {
5593 let mut merged_event_roots = Vec::with_capacity(changed_paths.len() + event_roots.len());
5594 let mut changed_paths = changed_paths.iter().peekable();
5595 let mut event_roots = event_roots.iter().peekable();
5596 while let (Some(path), Some(event_root)) = (changed_paths.peek(), event_roots.peek()) {
5597 match path.cmp(&&event_root.path) {
5598 Ordering::Less => {
5599 merged_event_roots.push(EventRoot {
5600 path: (*changed_paths.next().expect("peeked changed path")).clone(),
5601 was_rescanned: false,
5602 });
5603 }
5604 Ordering::Equal => {
5605 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5606 changed_paths.next();
5607 }
5608 Ordering::Greater => {
5609 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5610 }
5611 }
5612 }
5613 merged_event_roots.extend(changed_paths.map(|path| EventRoot {
5614 path: path.clone(),
5615 was_rescanned: false,
5616 }));
5617 merged_event_roots.extend(event_roots.cloned());
5618 merged_event_roots
5619}
5620
5621fn build_diff(
5622 phase: BackgroundScannerPhase,
5623 old_snapshot: &Snapshot,
5624 new_snapshot: &Snapshot,
5625 event_roots: &[EventRoot],
5626) -> UpdatedEntriesSet {
5627 use BackgroundScannerPhase::*;
5628 use PathChange::{Added, AddedOrUpdated, Loaded, Removed, Updated};
5629
5630 // Identify which paths have changed. Use the known set of changed
5631 // parent paths to optimize the search.
5632 let mut changes = Vec::new();
5633
5634 let mut old_paths = old_snapshot.entries_by_path.cursor::<PathKey>(());
5635 let mut new_paths = new_snapshot.entries_by_path.cursor::<PathKey>(());
5636 let mut last_newly_loaded_dir_path = None;
5637 old_paths.next();
5638 new_paths.next();
5639 for event_root in event_roots {
5640 let path = PathKey(event_root.path.clone());
5641 if old_paths.item().is_some_and(|e| e.path < path.0) {
5642 old_paths.seek_forward(&path, Bias::Left);
5643 }
5644 if new_paths.item().is_some_and(|e| e.path < path.0) {
5645 new_paths.seek_forward(&path, Bias::Left);
5646 }
5647 loop {
5648 match (old_paths.item(), new_paths.item()) {
5649 (Some(old_entry), Some(new_entry)) => {
5650 if old_entry.path > path.0
5651 && new_entry.path > path.0
5652 && !old_entry.path.starts_with(&path.0)
5653 && !new_entry.path.starts_with(&path.0)
5654 {
5655 break;
5656 }
5657
5658 match Ord::cmp(&old_entry.path, &new_entry.path) {
5659 Ordering::Less => {
5660 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5661 old_paths.next();
5662 }
5663 Ordering::Equal => {
5664 if phase == EventsReceivedDuringInitialScan {
5665 if old_entry.id != new_entry.id {
5666 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5667 }
5668 // If the worktree was not fully initialized when this event was generated,
5669 // we can't know whether this entry was added during the scan or whether
5670 // it was merely updated.
5671 changes.push((
5672 new_entry.path.clone(),
5673 new_entry.id,
5674 AddedOrUpdated,
5675 ));
5676 } else if old_entry.id != new_entry.id {
5677 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5678 changes.push((new_entry.path.clone(), new_entry.id, Added));
5679 } else if old_entry != new_entry {
5680 if old_entry.kind.is_unloaded() {
5681 last_newly_loaded_dir_path = Some(&new_entry.path);
5682 changes.push((new_entry.path.clone(), new_entry.id, Loaded));
5683 } else {
5684 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5685 }
5686 } else if event_root.was_rescanned {
5687 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5688 }
5689 old_paths.next();
5690 new_paths.next();
5691 }
5692 Ordering::Greater => {
5693 let is_newly_loaded = phase == InitialScan
5694 || last_newly_loaded_dir_path
5695 .as_ref()
5696 .is_some_and(|dir| new_entry.path.starts_with(dir));
5697 changes.push((
5698 new_entry.path.clone(),
5699 new_entry.id,
5700 if is_newly_loaded { Loaded } else { Added },
5701 ));
5702 new_paths.next();
5703 }
5704 }
5705 }
5706 (Some(old_entry), None) => {
5707 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5708 old_paths.next();
5709 }
5710 (None, Some(new_entry)) => {
5711 let is_newly_loaded = phase == InitialScan
5712 || last_newly_loaded_dir_path
5713 .as_ref()
5714 .is_some_and(|dir| new_entry.path.starts_with(dir));
5715 changes.push((
5716 new_entry.path.clone(),
5717 new_entry.id,
5718 if is_newly_loaded { Loaded } else { Added },
5719 ));
5720 new_paths.next();
5721 }
5722 (None, None) => break,
5723 }
5724 }
5725 }
5726
5727 changes.into()
5728}
5729
5730fn swap_to_front(child_paths: &mut Vec<PathBuf>, file: &str) {
5731 let position = child_paths
5732 .iter()
5733 .position(|path| path.file_name().unwrap() == file);
5734 if let Some(position) = position {
5735 let temp = child_paths.remove(position);
5736 child_paths.insert(0, temp);
5737 }
5738}
5739
5740fn char_bag_for_path(root_char_bag: CharBag, path: &RelPath) -> CharBag {
5741 let mut result = root_char_bag;
5742 result.extend(path.as_unix_str().chars().map(|c| c.to_ascii_lowercase()));
5743 result
5744}
5745
5746#[derive(Debug)]
5747struct ScanJob {
5748 abs_path: Arc<Path>,
5749 path: Arc<RelPath>,
5750 ignore_stack: IgnoreStack,
5751 scan_queue: Sender<ScanJob>,
5752 ancestor_inodes: TreeSet<u64>,
5753 is_external: bool,
5754}
5755
5756struct UpdateIgnoreStatusJob {
5757 abs_path: Arc<Path>,
5758 ignore_stack: IgnoreStack,
5759 ignore_queue: Sender<UpdateIgnoreStatusJob>,
5760 scan_queue: Sender<ScanJob>,
5761}
5762
5763pub trait WorktreeModelHandle {
5764 #[cfg(feature = "test-support")]
5765 fn flush_fs_events<'a>(
5766 &self,
5767 cx: &'a mut gpui::TestAppContext,
5768 ) -> futures::future::LocalBoxFuture<'a, ()>;
5769
5770 #[cfg(feature = "test-support")]
5771 fn flush_fs_events_in_root_git_repository<'a>(
5772 &self,
5773 cx: &'a mut gpui::TestAppContext,
5774 ) -> futures::future::LocalBoxFuture<'a, ()>;
5775}
5776
5777impl WorktreeModelHandle for Entity<Worktree> {
5778 // When the worktree's FS event stream sometimes delivers "redundant" events for FS changes that
5779 // occurred before the worktree was constructed. These events can cause the worktree to perform
5780 // extra directory scans, and emit extra scan-state notifications.
5781 //
5782 // This function mutates the worktree's directory and waits for those mutations to be picked up,
5783 // to ensure that all redundant FS events have already been processed.
5784 #[cfg(feature = "test-support")]
5785 fn flush_fs_events<'a>(
5786 &self,
5787 cx: &'a mut gpui::TestAppContext,
5788 ) -> futures::future::LocalBoxFuture<'a, ()> {
5789 let file_name = "fs-event-sentinel";
5790
5791 let tree = self.clone();
5792 let (fs, root_path) = self.read_with(cx, |tree, _| {
5793 let tree = tree.as_local().unwrap();
5794 (tree.fs.clone(), tree.abs_path.clone())
5795 });
5796
5797 async move {
5798 // Subscribe to events BEFORE creating the file to avoid race condition
5799 // where events fire before subscription is set up
5800 let mut events = cx.events(&tree);
5801
5802 fs.create_file(&root_path.join(file_name), Default::default())
5803 .await
5804 .unwrap();
5805
5806 // Check if condition is already met before waiting for events
5807 let file_exists = || {
5808 tree.read_with(cx, |tree, _| {
5809 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5810 .is_some()
5811 })
5812 };
5813
5814 // Use select to avoid blocking indefinitely if events are delayed
5815 while !file_exists() {
5816 futures::select_biased! {
5817 _ = events.next() => {}
5818 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5819 }
5820 }
5821
5822 fs.remove_file(&root_path.join(file_name), Default::default())
5823 .await
5824 .unwrap();
5825
5826 // Check if condition is already met before waiting for events
5827 let file_gone = || {
5828 tree.read_with(cx, |tree, _| {
5829 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5830 .is_none()
5831 })
5832 };
5833
5834 // Use select to avoid blocking indefinitely if events are delayed
5835 while !file_gone() {
5836 futures::select_biased! {
5837 _ = events.next() => {}
5838 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5839 }
5840 }
5841
5842 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5843 .await;
5844 }
5845 .boxed_local()
5846 }
5847
5848 // This function is similar to flush_fs_events, except that it waits for events to be flushed in
5849 // the .git folder of the root repository.
5850 // The reason for its existence is that a repository's .git folder might live *outside* of the
5851 // worktree and thus its FS events might go through a different path.
5852 // In order to flush those, we need to create artificial events in the .git folder and wait
5853 // for the repository to be reloaded.
5854 #[cfg(feature = "test-support")]
5855 fn flush_fs_events_in_root_git_repository<'a>(
5856 &self,
5857 cx: &'a mut gpui::TestAppContext,
5858 ) -> futures::future::LocalBoxFuture<'a, ()> {
5859 let file_name = "fs-event-sentinel";
5860
5861 let tree = self.clone();
5862 let (fs, root_path, mut git_dir_scan_id) = self.read_with(cx, |tree, _| {
5863 let tree = tree.as_local().unwrap();
5864 let local_repo_entry = tree
5865 .git_repositories
5866 .values()
5867 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5868 .unwrap();
5869 (
5870 tree.fs.clone(),
5871 local_repo_entry.common_dir_abs_path.clone(),
5872 local_repo_entry.git_dir_scan_id,
5873 )
5874 });
5875
5876 let scan_id_increased = |tree: &mut Worktree, git_dir_scan_id: &mut usize| {
5877 let tree = tree.as_local().unwrap();
5878 // let repository = tree.repositories.first().unwrap();
5879 let local_repo_entry = tree
5880 .git_repositories
5881 .values()
5882 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5883 .unwrap();
5884
5885 if local_repo_entry.git_dir_scan_id > *git_dir_scan_id {
5886 *git_dir_scan_id = local_repo_entry.git_dir_scan_id;
5887 true
5888 } else {
5889 false
5890 }
5891 };
5892
5893 async move {
5894 // Subscribe to events BEFORE creating the file to avoid race condition
5895 // where events fire before subscription is set up
5896 let mut events = cx.events(&tree);
5897
5898 fs.create_file(&root_path.join(file_name), Default::default())
5899 .await
5900 .unwrap();
5901
5902 // Use select to avoid blocking indefinitely if events are delayed
5903 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5904 futures::select_biased! {
5905 _ = events.next() => {}
5906 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5907 }
5908 }
5909
5910 fs.remove_file(&root_path.join(file_name), Default::default())
5911 .await
5912 .unwrap();
5913
5914 // Use select to avoid blocking indefinitely if events are delayed
5915 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5916 futures::select_biased! {
5917 _ = events.next() => {}
5918 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5919 }
5920 }
5921
5922 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5923 .await;
5924 }
5925 .boxed_local()
5926 }
5927}
5928
5929#[derive(Clone, Debug)]
5930struct TraversalProgress<'a> {
5931 max_path: &'a RelPath,
5932 count: usize,
5933 non_ignored_count: usize,
5934 file_count: usize,
5935 non_ignored_file_count: usize,
5936}
5937
5938impl TraversalProgress<'_> {
5939 fn count(&self, include_files: bool, include_dirs: bool, include_ignored: bool) -> usize {
5940 match (include_files, include_dirs, include_ignored) {
5941 (true, true, true) => self.count,
5942 (true, true, false) => self.non_ignored_count,
5943 (true, false, true) => self.file_count,
5944 (true, false, false) => self.non_ignored_file_count,
5945 (false, true, true) => self.count - self.file_count,
5946 (false, true, false) => self.non_ignored_count - self.non_ignored_file_count,
5947 (false, false, _) => 0,
5948 }
5949 }
5950}
5951
5952impl<'a> sum_tree::Dimension<'a, EntrySummary> for TraversalProgress<'a> {
5953 fn zero(_cx: ()) -> Self {
5954 Default::default()
5955 }
5956
5957 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
5958 self.max_path = summary.max_path.as_ref();
5959 self.count += summary.count;
5960 self.non_ignored_count += summary.non_ignored_count;
5961 self.file_count += summary.file_count;
5962 self.non_ignored_file_count += summary.non_ignored_file_count;
5963 }
5964}
5965
5966impl Default for TraversalProgress<'_> {
5967 fn default() -> Self {
5968 Self {
5969 max_path: RelPath::empty(),
5970 count: 0,
5971 non_ignored_count: 0,
5972 file_count: 0,
5973 non_ignored_file_count: 0,
5974 }
5975 }
5976}
5977
5978#[derive(Debug)]
5979pub struct Traversal<'a> {
5980 snapshot: &'a Snapshot,
5981 cursor: sum_tree::Cursor<'a, 'static, Entry, TraversalProgress<'a>>,
5982 include_ignored: bool,
5983 include_files: bool,
5984 include_dirs: bool,
5985}
5986
5987impl<'a> Traversal<'a> {
5988 fn new(
5989 snapshot: &'a Snapshot,
5990 include_files: bool,
5991 include_dirs: bool,
5992 include_ignored: bool,
5993 start_path: &RelPath,
5994 ) -> Self {
5995 let mut cursor = snapshot.entries_by_path.cursor(());
5996 cursor.seek(&TraversalTarget::path(start_path), Bias::Left);
5997 let mut traversal = Self {
5998 snapshot,
5999 cursor,
6000 include_files,
6001 include_dirs,
6002 include_ignored,
6003 };
6004 if traversal.end_offset() == traversal.start_offset() {
6005 traversal.next();
6006 }
6007 traversal
6008 }
6009
6010 pub fn advance(&mut self) -> bool {
6011 self.advance_by(1)
6012 }
6013
6014 pub fn advance_by(&mut self, count: usize) -> bool {
6015 self.cursor.seek_forward(
6016 &TraversalTarget::Count {
6017 count: self.end_offset() + count,
6018 include_dirs: self.include_dirs,
6019 include_files: self.include_files,
6020 include_ignored: self.include_ignored,
6021 },
6022 Bias::Left,
6023 )
6024 }
6025
6026 pub fn advance_to_sibling(&mut self) -> bool {
6027 while let Some(entry) = self.cursor.item() {
6028 self.cursor
6029 .seek_forward(&TraversalTarget::successor(&entry.path), Bias::Left);
6030 if let Some(entry) = self.cursor.item()
6031 && (self.include_files || !entry.is_file())
6032 && (self.include_dirs || !entry.is_dir())
6033 && (self.include_ignored || !entry.is_ignored || entry.is_always_included)
6034 {
6035 return true;
6036 }
6037 }
6038 false
6039 }
6040
6041 pub fn back_to_parent(&mut self) -> bool {
6042 let Some(parent_path) = self.cursor.item().and_then(|entry| entry.path.parent()) else {
6043 return false;
6044 };
6045 self.cursor
6046 .seek(&TraversalTarget::path(parent_path), Bias::Left)
6047 }
6048
6049 pub fn entry(&self) -> Option<&'a Entry> {
6050 self.cursor.item()
6051 }
6052
6053 pub fn snapshot(&self) -> &'a Snapshot {
6054 self.snapshot
6055 }
6056
6057 pub fn start_offset(&self) -> usize {
6058 self.cursor
6059 .start()
6060 .count(self.include_files, self.include_dirs, self.include_ignored)
6061 }
6062
6063 pub fn end_offset(&self) -> usize {
6064 self.cursor
6065 .end()
6066 .count(self.include_files, self.include_dirs, self.include_ignored)
6067 }
6068}
6069
6070impl<'a> Iterator for Traversal<'a> {
6071 type Item = &'a Entry;
6072
6073 fn next(&mut self) -> Option<Self::Item> {
6074 if let Some(item) = self.entry() {
6075 self.advance();
6076 Some(item)
6077 } else {
6078 None
6079 }
6080 }
6081}
6082
6083#[derive(Debug, Clone, Copy)]
6084pub enum PathTarget<'a> {
6085 Path(&'a RelPath),
6086 Successor(&'a RelPath),
6087}
6088
6089impl PathTarget<'_> {
6090 fn cmp_path(&self, other: &RelPath) -> Ordering {
6091 match self {
6092 PathTarget::Path(path) => path.cmp(&other),
6093 PathTarget::Successor(path) => {
6094 if other.starts_with(path) {
6095 Ordering::Greater
6096 } else {
6097 Ordering::Equal
6098 }
6099 }
6100 }
6101 }
6102}
6103
6104impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, PathProgress<'a>> for PathTarget<'_> {
6105 fn cmp(&self, cursor_location: &PathProgress<'a>, _: S::Context<'_>) -> Ordering {
6106 self.cmp_path(cursor_location.max_path)
6107 }
6108}
6109
6110impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, TraversalProgress<'a>> for PathTarget<'_> {
6111 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: S::Context<'_>) -> Ordering {
6112 self.cmp_path(cursor_location.max_path)
6113 }
6114}
6115
6116#[derive(Debug)]
6117enum TraversalTarget<'a> {
6118 Path(PathTarget<'a>),
6119 Count {
6120 count: usize,
6121 include_files: bool,
6122 include_ignored: bool,
6123 include_dirs: bool,
6124 },
6125}
6126
6127impl<'a> TraversalTarget<'a> {
6128 fn path(path: &'a RelPath) -> Self {
6129 Self::Path(PathTarget::Path(path))
6130 }
6131
6132 fn successor(path: &'a RelPath) -> Self {
6133 Self::Path(PathTarget::Successor(path))
6134 }
6135
6136 fn cmp_progress(&self, progress: &TraversalProgress) -> Ordering {
6137 match self {
6138 TraversalTarget::Path(path) => path.cmp_path(progress.max_path),
6139 TraversalTarget::Count {
6140 count,
6141 include_files,
6142 include_dirs,
6143 include_ignored,
6144 } => Ord::cmp(
6145 count,
6146 &progress.count(*include_files, *include_dirs, *include_ignored),
6147 ),
6148 }
6149 }
6150}
6151
6152impl<'a> SeekTarget<'a, EntrySummary, TraversalProgress<'a>> for TraversalTarget<'_> {
6153 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
6154 self.cmp_progress(cursor_location)
6155 }
6156}
6157
6158impl<'a> SeekTarget<'a, PathSummary<sum_tree::NoSummary>, TraversalProgress<'a>>
6159 for TraversalTarget<'_>
6160{
6161 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
6162 self.cmp_progress(cursor_location)
6163 }
6164}
6165
6166pub struct ChildEntriesOptions {
6167 pub include_files: bool,
6168 pub include_dirs: bool,
6169 pub include_ignored: bool,
6170}
6171
6172pub struct ChildEntriesIter<'a> {
6173 parent_path: &'a RelPath,
6174 traversal: Traversal<'a>,
6175}
6176
6177impl<'a> Iterator for ChildEntriesIter<'a> {
6178 type Item = &'a Entry;
6179
6180 fn next(&mut self) -> Option<Self::Item> {
6181 if let Some(item) = self.traversal.entry()
6182 && item.path.starts_with(self.parent_path)
6183 {
6184 self.traversal.advance_to_sibling();
6185 return Some(item);
6186 }
6187 None
6188 }
6189}
6190
6191impl<'a> From<&'a Entry> for proto::Entry {
6192 fn from(entry: &'a Entry) -> Self {
6193 Self {
6194 id: entry.id.to_proto(),
6195 is_dir: entry.is_dir(),
6196 path: entry.path.as_ref().to_proto(),
6197 inode: entry.inode,
6198 mtime: entry.mtime.map(|time| time.into()),
6199 is_ignored: entry.is_ignored,
6200 is_hidden: entry.is_hidden,
6201 is_external: entry.is_external,
6202 is_fifo: entry.is_fifo,
6203 size: Some(entry.size),
6204 canonical_path: entry
6205 .canonical_path
6206 .as_ref()
6207 .map(|path| path.to_string_lossy().into_owned()),
6208 }
6209 }
6210}
6211
6212impl TryFrom<(&CharBag, &PathMatcher, proto::Entry)> for Entry {
6213 type Error = anyhow::Error;
6214
6215 fn try_from(
6216 (root_char_bag, always_included, entry): (&CharBag, &PathMatcher, proto::Entry),
6217 ) -> Result<Self> {
6218 let kind = if entry.is_dir {
6219 EntryKind::Dir
6220 } else {
6221 EntryKind::File
6222 };
6223
6224 let path =
6225 RelPath::from_proto(&entry.path).context("invalid relative path in proto message")?;
6226 let char_bag = char_bag_for_path(*root_char_bag, &path);
6227 let is_always_included = always_included.is_match(&path);
6228 Ok(Entry {
6229 id: ProjectEntryId::from_proto(entry.id),
6230 kind,
6231 path,
6232 inode: entry.inode,
6233 mtime: entry.mtime.map(|time| time.into()),
6234 size: entry.size.unwrap_or(0),
6235 canonical_path: entry
6236 .canonical_path
6237 .map(|path_string| Arc::from(PathBuf::from(path_string))),
6238 is_ignored: entry.is_ignored,
6239 is_hidden: entry.is_hidden,
6240 is_always_included,
6241 is_external: entry.is_external,
6242 is_private: false,
6243 char_bag,
6244 is_fifo: entry.is_fifo,
6245 })
6246 }
6247}
6248
6249#[derive(Clone, Copy, Debug, Default, Hash, PartialEq, Eq, PartialOrd, Ord)]
6250pub struct ProjectEntryId(usize);
6251
6252impl ProjectEntryId {
6253 pub const MAX: Self = Self(usize::MAX);
6254 pub const MIN: Self = Self(usize::MIN);
6255
6256 pub fn new(counter: &AtomicUsize) -> Self {
6257 Self(counter.fetch_add(1, SeqCst))
6258 }
6259
6260 pub fn from_proto(id: u64) -> Self {
6261 Self(id as usize)
6262 }
6263
6264 pub fn to_proto(self) -> u64 {
6265 self.0 as u64
6266 }
6267
6268 pub fn from_usize(id: usize) -> Self {
6269 ProjectEntryId(id)
6270 }
6271
6272 pub fn to_usize(self) -> usize {
6273 self.0
6274 }
6275}
6276
6277#[cfg(feature = "test-support")]
6278impl CreatedEntry {
6279 pub fn into_included(self) -> Option<Entry> {
6280 match self {
6281 CreatedEntry::Included(entry) => Some(entry),
6282 CreatedEntry::Excluded { .. } => None,
6283 }
6284 }
6285}
6286
6287fn parse_gitfile(content: &str) -> anyhow::Result<&Path> {
6288 let path = content
6289 .strip_prefix("gitdir:")
6290 .with_context(|| format!("parsing gitfile content {content:?}"))?;
6291 Ok(Path::new(path.trim()))
6292}
6293
6294pub async fn discover_root_repo_common_dir(root_abs_path: &Path, fs: &dyn Fs) -> Option<Arc<Path>> {
6295 let root_dot_git = root_abs_path.join(DOT_GIT);
6296 if !fs.metadata(&root_dot_git).await.is_ok_and(|m| m.is_some()) {
6297 return None;
6298 }
6299 let dot_git_path: Arc<Path> = root_dot_git.into();
6300 let (_, common_dir) = discover_git_paths(&dot_git_path, fs).await;
6301 Some(common_dir)
6302}
6303
6304async fn discover_git_paths(dot_git_abs_path: &Arc<Path>, fs: &dyn Fs) -> (Arc<Path>, Arc<Path>) {
6305 let mut repository_dir_abs_path = dot_git_abs_path.clone();
6306 let mut common_dir_abs_path = dot_git_abs_path.clone();
6307
6308 if let Some(path) = fs
6309 .load(dot_git_abs_path)
6310 .await
6311 .ok()
6312 .as_ref()
6313 .and_then(|contents| parse_gitfile(contents).log_err())
6314 {
6315 let path = dot_git_abs_path
6316 .parent()
6317 .unwrap_or(Path::new(""))
6318 .join(path);
6319 if let Some(path) = fs.canonicalize(&path).await.log_err() {
6320 repository_dir_abs_path = Path::new(&path).into();
6321 common_dir_abs_path = repository_dir_abs_path.clone();
6322
6323 if let Some(commondir_contents) = fs.load(&path.join("commondir")).await.ok()
6324 && let Some(commondir_path) = fs
6325 .canonicalize(&path.join(commondir_contents.trim()))
6326 .await
6327 .log_err()
6328 {
6329 common_dir_abs_path = commondir_path.as_path().into();
6330 }
6331 }
6332 };
6333 (repository_dir_abs_path, common_dir_abs_path)
6334}
6335
6336struct NullWatcher;
6337
6338impl fs::Watcher for NullWatcher {
6339 fn add(&self, _path: &Path) -> Result<()> {
6340 Ok(())
6341 }
6342
6343 fn remove(&self, _path: &Path) -> Result<()> {
6344 Ok(())
6345 }
6346}
6347
6348const FILE_ANALYSIS_BYTES: usize = 1024;
6349
6350async fn decode_file_text(
6351 fs: &dyn Fs,
6352 abs_path: &Path,
6353) -> Result<(String, &'static Encoding, bool)> {
6354 let mut file = fs
6355 .open_sync(&abs_path)
6356 .await
6357 .with_context(|| format!("opening file {abs_path:?}"))?;
6358
6359 // First, read the beginning of the file to determine its kind and encoding.
6360 // We do not want to load an entire large blob into memory only to discard it.
6361 let mut file_first_bytes = Vec::with_capacity(FILE_ANALYSIS_BYTES);
6362 let mut buf = [0u8; FILE_ANALYSIS_BYTES];
6363 let mut reached_eof = false;
6364 loop {
6365 if file_first_bytes.len() >= FILE_ANALYSIS_BYTES {
6366 break;
6367 }
6368 let n = file
6369 .read(&mut buf)
6370 .with_context(|| format!("reading bytes of the file {abs_path:?}"))?;
6371 if n == 0 {
6372 reached_eof = true;
6373 break;
6374 }
6375 file_first_bytes.extend_from_slice(&buf[..n]);
6376 }
6377 let (bom_encoding, byte_content) = decode_byte_header(&file_first_bytes);
6378 anyhow::ensure!(
6379 byte_content != ByteContent::Binary,
6380 "Binary files are not supported"
6381 );
6382
6383 // If the file is eligible for opening, read the rest of the file.
6384 let mut content = file_first_bytes;
6385 if !reached_eof {
6386 let mut buf = [0u8; 8 * 1024];
6387 loop {
6388 let n = file
6389 .read(&mut buf)
6390 .with_context(|| format!("reading remaining bytes of the file {abs_path:?}"))?;
6391 if n == 0 {
6392 break;
6393 }
6394 content.extend_from_slice(&buf[..n]);
6395 }
6396 }
6397 decode_byte_full(content, bom_encoding, byte_content)
6398}
6399
6400fn decode_byte_header(prefix: &[u8]) -> (Option<&'static Encoding>, ByteContent) {
6401 if let Some((encoding, _bom_len)) = Encoding::for_bom(prefix) {
6402 return (Some(encoding), ByteContent::Unknown);
6403 }
6404 (None, analyze_byte_content(prefix))
6405}
6406
6407fn decode_byte_full(
6408 bytes: Vec<u8>,
6409 bom_encoding: Option<&'static Encoding>,
6410 byte_content: ByteContent,
6411) -> Result<(String, &'static Encoding, bool)> {
6412 if let Some(encoding) = bom_encoding {
6413 let (cow, _) = encoding.decode_with_bom_removal(&bytes);
6414 return Ok((cow.into_owned(), encoding, true));
6415 }
6416
6417 match byte_content {
6418 ByteContent::Utf16Le => {
6419 let encoding = encoding_rs::UTF_16LE;
6420 let (cow, _, _) = encoding.decode(&bytes);
6421 return Ok((cow.into_owned(), encoding, false));
6422 }
6423 ByteContent::Utf16Be => {
6424 let encoding = encoding_rs::UTF_16BE;
6425 let (cow, _, _) = encoding.decode(&bytes);
6426 return Ok((cow.into_owned(), encoding, false));
6427 }
6428 ByteContent::Binary => {
6429 anyhow::bail!("Binary files are not supported");
6430 }
6431 ByteContent::Unknown => {}
6432 }
6433
6434 fn detect_encoding(bytes: Vec<u8>) -> (String, &'static Encoding) {
6435 let mut detector = EncodingDetector::new();
6436 detector.feed(&bytes, true);
6437
6438 let encoding = detector.guess(None, true); // Use None for TLD hint to ensure neutral detection logic.
6439
6440 let (cow, _, _) = encoding.decode(&bytes);
6441 (cow.into_owned(), encoding)
6442 }
6443
6444 match String::from_utf8(bytes) {
6445 Ok(text) => {
6446 // ISO-2022-JP (and other ISO-2022 variants) consists entirely of 7-bit ASCII bytes,
6447 // so it is valid UTF-8. However, it contains escape sequences starting with '\x1b'.
6448 // If we find an escape character, we double-check the encoding to prevent
6449 // displaying raw escape sequences instead of the correct characters.
6450 if text.contains('\x1b') {
6451 let (s, enc) = detect_encoding(text.into_bytes());
6452 Ok((s, enc, false))
6453 } else {
6454 Ok((text, encoding_rs::UTF_8, false))
6455 }
6456 }
6457 Err(e) => {
6458 let (s, enc) = detect_encoding(e.into_bytes());
6459 Ok((s, enc, false))
6460 }
6461 }
6462}
6463
6464#[derive(Debug, PartialEq)]
6465enum ByteContent {
6466 Utf16Le,
6467 Utf16Be,
6468 Binary,
6469 Unknown,
6470}
6471
6472// Heuristic check using null byte distribution plus a generic text-likeness
6473// heuristic. This prefers UTF-16 when many bytes are NUL and otherwise
6474// distinguishes between text-like and binary-like content.
6475fn analyze_byte_content(bytes: &[u8]) -> ByteContent {
6476 if bytes.len() < 2 {
6477 return ByteContent::Unknown;
6478 }
6479
6480 if is_known_binary_header(bytes) {
6481 return ByteContent::Binary;
6482 }
6483
6484 let limit = bytes.len().min(FILE_ANALYSIS_BYTES);
6485 let mut even_null_count = 0usize;
6486 let mut odd_null_count = 0usize;
6487 let mut non_text_like_count = 0usize;
6488
6489 for (i, &byte) in bytes[..limit].iter().enumerate() {
6490 if byte == 0 {
6491 if i % 2 == 0 {
6492 even_null_count += 1;
6493 } else {
6494 odd_null_count += 1;
6495 }
6496 non_text_like_count += 1;
6497 continue;
6498 }
6499
6500 let is_text_like = match byte {
6501 b'\t' | b'\n' | b'\r' | 0x0C => true,
6502 0x20..=0x7E => true,
6503 // Treat bytes that are likely part of UTF-8 or single-byte encodings as text-like.
6504 0x80..=0xBF | 0xC2..=0xF4 => true,
6505 _ => false,
6506 };
6507
6508 if !is_text_like {
6509 non_text_like_count += 1;
6510 }
6511 }
6512
6513 let total_null_count = even_null_count + odd_null_count;
6514
6515 // If there are no NUL bytes at all, this is overwhelmingly likely to be text.
6516 if total_null_count == 0 {
6517 return ByteContent::Unknown;
6518 }
6519
6520 let has_significant_nulls = total_null_count >= limit / 16;
6521 let nulls_skew_to_even = even_null_count > odd_null_count * 4;
6522 let nulls_skew_to_odd = odd_null_count > even_null_count * 4;
6523
6524 if has_significant_nulls {
6525 let sample = &bytes[..limit];
6526
6527 // UTF-16BE ASCII: [0x00, char] — nulls at even positions (high byte first)
6528 // UTF-16LE ASCII: [char, 0x00] — nulls at odd positions (low byte first)
6529
6530 if nulls_skew_to_even && is_plausible_utf16_text(sample, false) {
6531 return ByteContent::Utf16Be;
6532 }
6533
6534 if nulls_skew_to_odd && is_plausible_utf16_text(sample, true) {
6535 return ByteContent::Utf16Le;
6536 }
6537
6538 return ByteContent::Binary;
6539 }
6540
6541 if non_text_like_count * 100 < limit * 8 {
6542 ByteContent::Unknown
6543 } else {
6544 ByteContent::Binary
6545 }
6546}
6547
6548fn is_known_binary_header(bytes: &[u8]) -> bool {
6549 bytes.starts_with(b"%PDF-") // PDF
6550 || bytes.starts_with(b"PK\x03\x04") // ZIP local header
6551 || bytes.starts_with(b"PK\x05\x06") // ZIP end of central directory
6552 || bytes.starts_with(b"PK\x07\x08") // ZIP spanning/splitting
6553 || bytes.starts_with(b"\x89PNG\r\n\x1a\n") // PNG
6554 || bytes.starts_with(b"\xFF\xD8\xFF") // JPEG
6555 || bytes.starts_with(b"GIF87a") // GIF87a
6556 || bytes.starts_with(b"GIF89a") // GIF89a
6557 || bytes.starts_with(b"IWAD") // Doom IWAD archive
6558 || bytes.starts_with(b"PWAD") // Doom PWAD archive
6559 || bytes.starts_with(b"RIFF") // WAV, AVI, WebP
6560 || bytes.starts_with(b"OggS") // OGG (Vorbis, Opus, FLAC)
6561 || bytes.starts_with(b"fLaC") // FLAC
6562 || bytes.starts_with(b"ID3") // MP3 with ID3v2 tag
6563 || bytes.starts_with(b"\xFF\xFB") // MP3 frame sync (MPEG1 Layer3)
6564 || bytes.starts_with(b"\xFF\xFA") // MP3 frame sync (MPEG1 Layer3)
6565 || bytes.starts_with(b"\xFF\xF3") // MP3 frame sync (MPEG2 Layer3)
6566 || bytes.starts_with(b"\xFF\xF2") // MP3 frame sync (MPEG2 Layer3)
6567}
6568
6569// Null byte skew alone is not enough to identify UTF-16 -- binary formats with
6570// small 16-bit values (like PCM audio) produce the same pattern. Decode the
6571// bytes as UTF-16 and reject if too many code units land in control character
6572// ranges or form unpaired surrogates, which real text almost never contains.
6573fn is_plausible_utf16_text(bytes: &[u8], little_endian: bool) -> bool {
6574 let mut suspicious_count = 0usize;
6575 let mut total = 0usize;
6576
6577 let mut i = 0;
6578 while let Some(code_unit) = read_u16(bytes, i, little_endian) {
6579 total += 1;
6580
6581 match code_unit {
6582 0x0009 | 0x000A | 0x000C | 0x000D => {}
6583 // C0/C1 control characters and non-characters
6584 0x0000..=0x001F | 0x007F..=0x009F | 0xFFFE | 0xFFFF => suspicious_count += 1,
6585 0xD800..=0xDBFF => {
6586 let next_offset = i + 2;
6587 let has_low_surrogate = read_u16(bytes, next_offset, little_endian)
6588 .is_some_and(|next| (0xDC00..=0xDFFF).contains(&next));
6589 if has_low_surrogate {
6590 total += 1;
6591 i += 2;
6592 } else {
6593 suspicious_count += 1;
6594 }
6595 }
6596 // Lone low surrogate without a preceding high surrogate
6597 0xDC00..=0xDFFF => suspicious_count += 1,
6598 _ => {}
6599 }
6600
6601 i += 2;
6602 }
6603
6604 if total == 0 {
6605 return false;
6606 }
6607
6608 // Real UTF-16 text has near-zero control characters; binary data with
6609 // small 16-bit values typically exceeds 5%. 2% provides a safe margin.
6610 suspicious_count * 100 < total * 2
6611}
6612
6613fn read_u16(bytes: &[u8], offset: usize, little_endian: bool) -> Option<u16> {
6614 let pair = [*bytes.get(offset)?, *bytes.get(offset + 1)?];
6615 if little_endian {
6616 return Some(u16::from_le_bytes(pair));
6617 }
6618 Some(u16::from_be_bytes(pair))
6619}
6620
6621#[cfg(test)]
6622mod tests {
6623 use super::*;
6624
6625 /// reproduction of issue #50785
6626 fn build_pcm16_wav_bytes() -> Vec<u8> {
6627 let header: Vec<u8> = vec![
6628 /* RIFF header */
6629 0x52, 0x49, 0x46, 0x46, // "RIFF"
6630 0xc6, 0xcf, 0x00, 0x00, // file size: 8
6631 0x57, 0x41, 0x56, 0x45, // "WAVE"
6632 /* fmt chunk */
6633 0x66, 0x6d, 0x74, 0x20, // "fmt "
6634 0x10, 0x00, 0x00, 0x00, // chunk size: 16
6635 0x01, 0x00, // format: PCM (1)
6636 0x01, 0x00, // channels: 1 (mono)
6637 0x80, 0x3e, 0x00, 0x00, // sample rate: 16000
6638 0x00, 0x7d, 0x00, 0x00, // byte rate: 32000
6639 0x02, 0x00, // block align: 2
6640 0x10, 0x00, // bits per sample: 16
6641 /* LIST chunk */
6642 0x4c, 0x49, 0x53, 0x54, // "LIST"
6643 0x1a, 0x00, 0x00, 0x00, // chunk size: 26
6644 0x49, 0x4e, 0x46, 0x4f, // "INFO"
6645 0x49, 0x53, 0x46, 0x54, // "ISFT"
6646 0x0d, 0x00, 0x00, 0x00, // sub-chunk size: 13
6647 0x4c, 0x61, 0x76, 0x66, 0x36, 0x32, 0x2e, 0x33, // "Lavf62.3"
6648 0x2e, 0x31, 0x30, 0x30, 0x00, // ".100\0"
6649 /* padding byte for word alignment */
6650 0x00, // data chunk header
6651 0x64, 0x61, 0x74, 0x61, // "data"
6652 0x80, 0xcf, 0x00, 0x00, // chunk size
6653 ];
6654
6655 let mut bytes = header;
6656
6657 // fill remaining space up to `FILE_ANALYSIS_BYTES` with synthetic PCM
6658 let audio_bytes_needed = FILE_ANALYSIS_BYTES - bytes.len();
6659 for i in 0..(audio_bytes_needed / 2) {
6660 let sample = (i & 0xFF) as u8;
6661 bytes.push(sample); // low byte: varies
6662 bytes.push(0x00); // high byte: zero for small values
6663 }
6664
6665 bytes
6666 }
6667
6668 #[test]
6669 fn test_pcm16_wav_detected_as_binary() {
6670 let wav_bytes = build_pcm16_wav_bytes();
6671 assert_eq!(wav_bytes.len(), FILE_ANALYSIS_BYTES);
6672
6673 let result = analyze_byte_content(&wav_bytes);
6674 assert_eq!(
6675 result,
6676 ByteContent::Binary,
6677 "PCM 16-bit WAV should be detected as Binary via RIFF header"
6678 );
6679 }
6680
6681 #[test]
6682 fn test_le16_binary_not_misdetected_as_utf16le() {
6683 let mut bytes = b"FAKE".to_vec();
6684 while bytes.len() < FILE_ANALYSIS_BYTES {
6685 let sample = (bytes.len() & 0xFF) as u8;
6686 bytes.push(sample);
6687 bytes.push(0x00);
6688 }
6689 bytes.truncate(FILE_ANALYSIS_BYTES);
6690
6691 let result = analyze_byte_content(&bytes);
6692 assert_eq!(
6693 result,
6694 ByteContent::Binary,
6695 "LE 16-bit binary with control characters should be detected as Binary"
6696 );
6697 }
6698
6699 #[test]
6700 fn test_be16_binary_not_misdetected_as_utf16be() {
6701 let mut bytes = b"FAKE".to_vec();
6702 while bytes.len() < FILE_ANALYSIS_BYTES {
6703 bytes.push(0x00);
6704 let sample = (bytes.len() & 0xFF) as u8;
6705 bytes.push(sample);
6706 }
6707 bytes.truncate(FILE_ANALYSIS_BYTES);
6708
6709 let result = analyze_byte_content(&bytes);
6710 assert_eq!(
6711 result,
6712 ByteContent::Binary,
6713 "BE 16-bit binary with control characters should be detected as Binary"
6714 );
6715 }
6716
6717 #[test]
6718 fn test_utf16le_text_detected_as_utf16le() {
6719 let text = "Hello, world! This is a UTF-16 test string. ";
6720 let mut bytes = Vec::new();
6721 while bytes.len() < FILE_ANALYSIS_BYTES {
6722 bytes.extend(text.encode_utf16().flat_map(|u| u.to_le_bytes()));
6723 }
6724 bytes.truncate(FILE_ANALYSIS_BYTES);
6725
6726 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Le);
6727 }
6728
6729 #[test]
6730 fn test_utf16be_text_detected_as_utf16be() {
6731 let text = "Hello, world! This is a UTF-16 test string. ";
6732 let mut bytes = Vec::new();
6733 while bytes.len() < FILE_ANALYSIS_BYTES {
6734 bytes.extend(text.encode_utf16().flat_map(|u| u.to_be_bytes()));
6735 }
6736 bytes.truncate(FILE_ANALYSIS_BYTES);
6737
6738 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Be);
6739 }
6740
6741 #[test]
6742 fn test_known_binary_headers() {
6743 let cases: &[(&[u8], &str)] = &[
6744 (b"RIFF\x00\x00\x00\x00WAVE", "WAV"),
6745 (b"RIFF\x00\x00\x00\x00AVI ", "AVI"),
6746 (b"OggS\x00\x02", "OGG"),
6747 (b"fLaC\x00\x00", "FLAC"),
6748 (b"ID3\x03\x00", "MP3 ID3v2"),
6749 (b"\xFF\xFB\x90\x00", "MP3 MPEG1 Layer3"),
6750 (b"\xFF\xF3\x90\x00", "MP3 MPEG2 Layer3"),
6751 ];
6752
6753 for (header, label) in cases {
6754 let mut bytes = header.to_vec();
6755 bytes.resize(FILE_ANALYSIS_BYTES, 0x41); // pad with 'A'
6756 assert_eq!(
6757 analyze_byte_content(&bytes),
6758 ByteContent::Binary,
6759 "{label} should be detected as Binary"
6760 );
6761 }
6762 }
6763}