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