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