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