1mod ignore;
2mod worktree_settings;
3
4use ::ignore::gitignore::{Gitignore, GitignoreBuilder};
5use anyhow::{Context as _, Result, anyhow};
6use chardetng::EncodingDetector;
7use clock::ReplicaId;
8use collections::{HashMap, HashSet, VecDeque};
9use encoding_rs::Encoding;
10use fs::{
11 Fs, MTime, PathEvent, PathEventKind, RemoveOptions, TrashedEntry, Watcher, copy_recursive,
12 read_dir_items,
13};
14use futures::{
15 FutureExt as _, Stream, StreamExt,
16 channel::{
17 mpsc::{self, UnboundedSender},
18 oneshot,
19 },
20 select_biased, stream,
21 task::Poll,
22};
23use fuzzy::CharBag;
24use git::{
25 COMMIT_MESSAGE, DOT_GIT, FSMONITOR_DAEMON, GITIGNORE, INDEX_LOCK, LFS_DIR, REPO_EXCLUDE,
26 status::GitSummary,
27};
28use gpui::{
29 App, AppContext as _, AsyncApp, BackgroundExecutor, Context, Entity, EventEmitter, Priority,
30 Task,
31};
32use ignore::IgnoreStack;
33use language::DiskState;
34
35use parking_lot::Mutex;
36use paths::{local_settings_folder_name, local_vscode_folder_name};
37use postage::{
38 barrier,
39 prelude::{Sink as _, Stream as _},
40 watch,
41};
42use rpc::{
43 AnyProtoClient,
44 proto::{self, split_worktree_update},
45};
46pub use settings::WorktreeId;
47use settings::{Settings, SettingsLocation, SettingsStore};
48use smallvec::{SmallVec, smallvec};
49use smol::channel::{self, Sender};
50use std::{
51 any::Any,
52 borrow::Borrow as _,
53 cmp::Ordering,
54 collections::hash_map,
55 convert::TryFrom,
56 ffi::OsStr,
57 fmt,
58 future::Future,
59 mem::{self},
60 ops::{Deref, DerefMut, Range},
61 path::{Path, PathBuf},
62 pin::Pin,
63 sync::{
64 Arc,
65 atomic::{AtomicUsize, Ordering::SeqCst},
66 },
67 time::{Duration, Instant},
68};
69use sum_tree::{Bias, Dimensions, Edit, KeyedItem, SeekTarget, SumTree, Summary, TreeMap, TreeSet};
70use text::{LineEnding, Rope};
71use util::{
72 ResultExt, debug_panic, maybe,
73 paths::{PathMatcher, PathStyle, SanitizedPath, home_dir},
74 rel_path::{RelPath, RelPathBuf},
75};
76pub use worktree_settings::WorktreeSettings;
77
78use crate::ignore::IgnoreKind;
79
80pub const FS_WATCH_LATENCY: Duration = Duration::from_millis(100);
81
82/// A set of local or remote files that are being opened as part of a project.
83/// Responsible for tracking related FS (for local)/collab (for remote) events and corresponding updates.
84/// Stores git repositories data and the diagnostics for the file(s).
85///
86/// Has an absolute path, and may be set to be visible in Zed UI or not.
87/// May correspond to a directory or a single file.
88/// Possible examples:
89/// * a drag and dropped file — may be added as an invisible, "ephemeral" entry to the current worktree
90/// * a directory opened in Zed — may be added as a visible entry to the current worktree
91///
92/// Uses [`Entry`] to track the state of each file/directory, can look up absolute paths for entries.
93pub enum Worktree {
94 Local(LocalWorktree),
95 Remote(RemoteWorktree),
96}
97
98/// An entry, created in the worktree.
99#[derive(Debug)]
100pub enum CreatedEntry {
101 /// Got created and indexed by the worktree, receiving a corresponding entry.
102 Included(Entry),
103 /// Got created, but not indexed due to falling under exclusion filters.
104 Excluded { abs_path: PathBuf },
105}
106
107#[derive(Debug)]
108pub struct LoadedFile {
109 pub file: Arc<File>,
110 pub text: String,
111 pub encoding: &'static Encoding,
112 pub has_bom: bool,
113}
114
115pub struct LoadedBinaryFile {
116 pub file: Arc<File>,
117 pub content: Vec<u8>,
118}
119
120impl fmt::Debug for LoadedBinaryFile {
121 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
122 f.debug_struct("LoadedBinaryFile")
123 .field("file", &self.file)
124 .field("content_bytes", &self.content.len())
125 .finish()
126 }
127}
128
129pub struct LocalWorktree {
130 snapshot: LocalSnapshot,
131 scan_requests_tx: channel::Sender<ScanRequest>,
132 path_prefixes_to_scan_tx: channel::Sender<PathPrefixScanRequest>,
133 is_scanning: (watch::Sender<bool>, watch::Receiver<bool>),
134 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
135 _background_scanner_tasks: Vec<Task<()>>,
136 update_observer: Option<UpdateObservationState>,
137 fs: Arc<dyn Fs>,
138 fs_case_sensitive: bool,
139 visible: bool,
140 next_entry_id: Arc<AtomicUsize>,
141 settings: WorktreeSettings,
142 share_private_files: bool,
143 scanning_enabled: bool,
144}
145
146pub struct PathPrefixScanRequest {
147 path: Arc<RelPath>,
148 done: SmallVec<[barrier::Sender; 1]>,
149}
150
151struct ScanRequest {
152 relative_paths: Vec<Arc<RelPath>>,
153 done: SmallVec<[barrier::Sender; 1]>,
154}
155
156pub struct RemoteWorktree {
157 snapshot: Snapshot,
158 background_snapshot: Arc<Mutex<(Snapshot, Vec<proto::UpdateWorktree>)>>,
159 project_id: u64,
160 client: AnyProtoClient,
161 file_scan_inclusions: PathMatcher,
162 updates_tx: Option<UnboundedSender<proto::UpdateWorktree>>,
163 update_observer: Option<mpsc::UnboundedSender<proto::UpdateWorktree>>,
164 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
165 replica_id: ReplicaId,
166 visible: bool,
167 disconnected: bool,
168}
169
170#[derive(Clone)]
171pub struct Snapshot {
172 id: WorktreeId,
173 /// The absolute path of the worktree root.
174 abs_path: Arc<SanitizedPath>,
175 path_style: PathStyle,
176 root_name: Arc<RelPath>,
177 root_char_bag: CharBag,
178 entries_by_path: SumTree<Entry>,
179 entries_by_id: SumTree<PathEntry>,
180 root_repo_common_dir: Option<Arc<SanitizedPath>>,
181 always_included_entries: Vec<Arc<RelPath>>,
182
183 /// A number that increases every time the worktree begins scanning
184 /// a set of paths from the filesystem. This scanning could be caused
185 /// by some operation performed on the worktree, such as reading or
186 /// writing a file, or by an event reported by the filesystem.
187 scan_id: usize,
188
189 /// The latest scan id that has completed, and whose preceding scans
190 /// have all completed. The current `scan_id` could be more than one
191 /// greater than the `completed_scan_id` if operations are performed
192 /// on the worktree while it is processing a file-system event.
193 completed_scan_id: usize,
194}
195
196/// This path corresponds to the 'content path' of a repository in relation
197/// to Zed's project root.
198/// In the majority of the cases, this is the folder that contains the .git folder.
199/// But if a sub-folder of a git repository is opened, this corresponds to the
200/// project root and the .git folder is located in a parent directory.
201#[derive(Clone, Debug, Ord, PartialOrd, Eq, PartialEq, Hash)]
202pub enum WorkDirectory {
203 InProject {
204 relative_path: Arc<RelPath>,
205 },
206 AboveProject {
207 absolute_path: Arc<Path>,
208 location_in_repo: Arc<Path>,
209 },
210}
211
212impl WorkDirectory {
213 fn path_key(&self) -> PathKey {
214 match self {
215 WorkDirectory::InProject { relative_path } => PathKey(relative_path.clone()),
216 WorkDirectory::AboveProject { .. } => PathKey(RelPath::empty().into()),
217 }
218 }
219
220 /// Returns true if the given path is a child of the work directory.
221 ///
222 /// Note that the path may not be a member of this repository, if there
223 /// is a repository in a directory between these two paths
224 /// external .git folder in a parent folder of the project root.
225 #[track_caller]
226 pub fn directory_contains(&self, path: &RelPath) -> bool {
227 match self {
228 WorkDirectory::InProject { relative_path } => path.starts_with(relative_path),
229 WorkDirectory::AboveProject { .. } => true,
230 }
231 }
232}
233
234impl Default for WorkDirectory {
235 fn default() -> Self {
236 Self::InProject {
237 relative_path: Arc::from(RelPath::empty()),
238 }
239 }
240}
241
242#[derive(Clone)]
243pub struct LocalSnapshot {
244 snapshot: Snapshot,
245 global_gitignore: Option<Arc<Gitignore>>,
246 /// Exclude files for all git repositories in the worktree, indexed by their absolute path.
247 /// The boolean indicates whether the gitignore needs to be updated.
248 repo_exclude_by_work_dir_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
249 /// All of the gitignore files in the worktree, indexed by their absolute path.
250 /// The boolean indicates whether the gitignore needs to be updated.
251 ignores_by_parent_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
252 /// All of the git repositories in the worktree, indexed by the project entry
253 /// id of their parent directory.
254 git_repositories: TreeMap<ProjectEntryId, LocalRepositoryEntry>,
255 /// The file handle of the worktree root
256 /// (so we can find it after it's been moved)
257 root_file_handle: Option<Arc<dyn fs::FileHandle>>,
258}
259
260struct BackgroundScannerState {
261 snapshot: LocalSnapshot,
262 scanned_dirs: HashSet<ProjectEntryId>,
263 path_prefixes_to_scan: HashSet<Arc<RelPath>>,
264 paths_to_scan: HashSet<Arc<RelPath>>,
265 /// The ids of all of the entries that were removed from the snapshot
266 /// as part of the current update. These entry ids may be re-used
267 /// if the same inode is discovered at a new path, or if the given
268 /// path is re-created after being deleted.
269 removed_entries: HashMap<u64, Entry>,
270 changed_paths: Vec<Arc<RelPath>>,
271 prev_snapshot: Snapshot,
272 scanning_enabled: bool,
273}
274
275#[derive(Clone, Debug, Eq, PartialEq)]
276struct EventRoot {
277 path: Arc<RelPath>,
278 was_rescanned: bool,
279}
280
281#[derive(Debug, Clone)]
282struct LocalRepositoryEntry {
283 work_directory_id: ProjectEntryId,
284 work_directory: WorkDirectory,
285 work_directory_abs_path: Arc<Path>,
286 git_dir_scan_id: usize,
287 /// Absolute path to the original .git entry that caused us to create this repository.
288 ///
289 /// This is normally a directory, but may be a "gitfile" that points to a directory elsewhere
290 /// (whose path we then store in `repository_dir_abs_path`).
291 dot_git_abs_path: Arc<Path>,
292 /// Absolute path to the "commondir" for this repository.
293 ///
294 /// This is always a directory. For a normal repository, this is the same as
295 /// `dot_git_abs_path`. For a linked worktree, this is the main repo's `.git`
296 /// directory (resolved from the worktree's `commondir` file). For a submodule,
297 /// this equals `repository_dir_abs_path` (submodules don't have a `commondir`
298 /// file).
299 common_dir_abs_path: Arc<Path>,
300 /// Absolute path to the directory holding the repository's state.
301 ///
302 /// For a normal repository, this is a directory and coincides with `dot_git_abs_path` and
303 /// `common_dir_abs_path`. For a submodule or worktree, this is some subdirectory of the
304 /// commondir like `/project/.git/modules/foo`.
305 repository_dir_abs_path: Arc<Path>,
306}
307
308impl sum_tree::Item for LocalRepositoryEntry {
309 type Summary = PathSummary<sum_tree::NoSummary>;
310
311 fn summary(&self, _: <Self::Summary as Summary>::Context<'_>) -> Self::Summary {
312 PathSummary {
313 max_path: self.work_directory.path_key().0,
314 item_summary: sum_tree::NoSummary,
315 }
316 }
317}
318
319impl KeyedItem for LocalRepositoryEntry {
320 type Key = PathKey;
321
322 fn key(&self) -> Self::Key {
323 self.work_directory.path_key()
324 }
325}
326
327impl Deref for LocalRepositoryEntry {
328 type Target = WorkDirectory;
329
330 fn deref(&self) -> &Self::Target {
331 &self.work_directory
332 }
333}
334
335impl Deref for LocalSnapshot {
336 type Target = Snapshot;
337
338 fn deref(&self) -> &Self::Target {
339 &self.snapshot
340 }
341}
342
343impl DerefMut for LocalSnapshot {
344 fn deref_mut(&mut self) -> &mut Self::Target {
345 &mut self.snapshot
346 }
347}
348
349enum ScanState {
350 Started,
351 Updated {
352 snapshot: LocalSnapshot,
353 changes: UpdatedEntriesSet,
354 barrier: SmallVec<[barrier::Sender; 1]>,
355 scanning: bool,
356 },
357 RootUpdated {
358 new_path: Arc<SanitizedPath>,
359 },
360 RootDeleted,
361}
362
363struct UpdateObservationState {
364 snapshots_tx: mpsc::UnboundedSender<(LocalSnapshot, UpdatedEntriesSet)>,
365 resume_updates: watch::Sender<()>,
366 _maintain_remote_snapshot: Task<Option<()>>,
367}
368
369#[derive(Debug, Clone)]
370pub enum Event {
371 UpdatedEntries(UpdatedEntriesSet),
372 UpdatedGitRepositories(UpdatedGitRepositoriesSet),
373 UpdatedRootRepoCommonDir,
374 DeletedEntry(ProjectEntryId),
375 /// The worktree root itself has been deleted (for single-file worktrees)
376 Deleted,
377}
378
379impl EventEmitter<Event> for Worktree {}
380
381impl Worktree {
382 pub async fn local(
383 path: impl Into<Arc<Path>>,
384 visible: bool,
385 fs: Arc<dyn Fs>,
386 next_entry_id: Arc<AtomicUsize>,
387 scanning_enabled: bool,
388 worktree_id: WorktreeId,
389 cx: &mut AsyncApp,
390 ) -> Result<Entity<Self>> {
391 let abs_path = path.into();
392 let metadata = fs
393 .metadata(&abs_path)
394 .await
395 .context("failed to stat worktree path")?;
396
397 let fs_case_sensitive = fs.is_case_sensitive().await;
398
399 let root_file_handle = if metadata.as_ref().is_some() {
400 fs.open_handle(&abs_path)
401 .await
402 .with_context(|| {
403 format!(
404 "failed to open local worktree root at {}",
405 abs_path.display()
406 )
407 })
408 .log_err()
409 } else {
410 None
411 };
412
413 let root_repo_common_dir = discover_root_repo_common_dir(&abs_path, fs.as_ref())
414 .await
415 .map(SanitizedPath::from_arc);
416
417 Ok(cx.new(move |cx: &mut Context<Worktree>| {
418 let mut snapshot = LocalSnapshot {
419 ignores_by_parent_abs_path: Default::default(),
420 global_gitignore: Default::default(),
421 repo_exclude_by_work_dir_abs_path: Default::default(),
422 git_repositories: Default::default(),
423 snapshot: Snapshot::new(
424 worktree_id,
425 abs_path
426 .file_name()
427 .and_then(|f| f.to_str())
428 .map_or(RelPath::empty().into(), |f| {
429 RelPath::unix(f).unwrap().into()
430 }),
431 abs_path.clone(),
432 PathStyle::local(),
433 ),
434 root_file_handle,
435 };
436 snapshot.root_repo_common_dir = root_repo_common_dir;
437
438 let worktree_id = snapshot.id();
439 let settings_location = Some(SettingsLocation {
440 worktree_id,
441 path: RelPath::empty(),
442 });
443
444 let settings = WorktreeSettings::get(settings_location, cx).clone();
445 cx.observe_global::<SettingsStore>(move |this, cx| {
446 if let Self::Local(this) = this {
447 let settings = WorktreeSettings::get(settings_location, cx).clone();
448 if this.settings != settings {
449 this.settings = settings;
450 this.restart_background_scanners(cx);
451 }
452 }
453 })
454 .detach();
455
456 let share_private_files = false;
457 if let Some(metadata) = metadata {
458 let mut entry = Entry::new(
459 RelPath::empty().into(),
460 &metadata,
461 ProjectEntryId::new(&next_entry_id),
462 snapshot.root_char_bag,
463 None,
464 );
465 if metadata.is_dir {
466 if !scanning_enabled {
467 entry.kind = EntryKind::UnloadedDir;
468 }
469 } else {
470 if let Some(file_name) = abs_path.file_name()
471 && let Some(file_name) = file_name.to_str()
472 && let Ok(path) = RelPath::unix(file_name)
473 {
474 entry.is_private = !share_private_files && settings.is_path_private(path);
475 entry.is_hidden = settings.is_path_hidden(path);
476 }
477 }
478 cx.foreground_executor()
479 .block_on(snapshot.insert_entry(entry, fs.as_ref()));
480 }
481
482 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
483 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
484 let mut worktree = LocalWorktree {
485 share_private_files,
486 next_entry_id,
487 snapshot,
488 is_scanning: watch::channel_with(true),
489 snapshot_subscriptions: Default::default(),
490 update_observer: None,
491 scan_requests_tx,
492 path_prefixes_to_scan_tx,
493 _background_scanner_tasks: Vec::new(),
494 fs,
495 fs_case_sensitive,
496 visible,
497 settings,
498 scanning_enabled,
499 };
500 worktree.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
501 Worktree::Local(worktree)
502 }))
503 }
504
505 pub fn remote(
506 project_id: u64,
507 replica_id: ReplicaId,
508 worktree: proto::WorktreeMetadata,
509 client: AnyProtoClient,
510 path_style: PathStyle,
511 cx: &mut App,
512 ) -> Entity<Self> {
513 cx.new(|cx: &mut Context<Self>| {
514 let snapshot = Snapshot::new(
515 WorktreeId::from_proto(worktree.id),
516 RelPath::from_proto(&worktree.root_name)
517 .unwrap_or_else(|_| RelPath::empty().into()),
518 Path::new(&worktree.abs_path).into(),
519 path_style,
520 );
521
522 let background_snapshot = Arc::new(Mutex::new((
523 snapshot.clone(),
524 Vec::<proto::UpdateWorktree>::new(),
525 )));
526 let (background_updates_tx, mut background_updates_rx) =
527 mpsc::unbounded::<proto::UpdateWorktree>();
528 let (mut snapshot_updated_tx, mut snapshot_updated_rx) = watch::channel();
529
530 let worktree_id = snapshot.id();
531 let settings_location = Some(SettingsLocation {
532 worktree_id,
533 path: RelPath::empty(),
534 });
535
536 let settings = WorktreeSettings::get(settings_location, cx).clone();
537 let worktree = RemoteWorktree {
538 client,
539 project_id,
540 replica_id,
541 snapshot,
542 file_scan_inclusions: settings.parent_dir_scan_inclusions.clone(),
543 background_snapshot: background_snapshot.clone(),
544 updates_tx: Some(background_updates_tx),
545 update_observer: None,
546 snapshot_subscriptions: Default::default(),
547 visible: worktree.visible,
548 disconnected: false,
549 };
550
551 // Apply updates to a separate snapshot in a background task, then
552 // send them to a foreground task which updates the model.
553 cx.background_spawn(async move {
554 while let Some(update) = background_updates_rx.next().await {
555 {
556 let mut lock = background_snapshot.lock();
557 lock.0.apply_remote_update(
558 update.clone(),
559 &settings.parent_dir_scan_inclusions,
560 );
561 lock.1.push(update);
562 }
563 snapshot_updated_tx.send(()).await.ok();
564 }
565 })
566 .detach();
567
568 // On the foreground task, update to the latest snapshot and notify
569 // any update observer of all updates that led to that snapshot.
570 cx.spawn(async move |this, cx| {
571 while (snapshot_updated_rx.recv().await).is_some() {
572 this.update(cx, |this, cx| {
573 let mut entries_changed = false;
574 let this = this.as_remote_mut().unwrap();
575 let old_root_repo_common_dir = this.snapshot.root_repo_common_dir.clone();
576 {
577 let mut lock = this.background_snapshot.lock();
578 this.snapshot = lock.0.clone();
579 for update in lock.1.drain(..) {
580 entries_changed |= !update.updated_entries.is_empty()
581 || !update.removed_entries.is_empty();
582 if let Some(tx) = &this.update_observer {
583 tx.unbounded_send(update).ok();
584 }
585 }
586 };
587
588 if entries_changed {
589 cx.emit(Event::UpdatedEntries(Arc::default()));
590 }
591 if this.snapshot.root_repo_common_dir != old_root_repo_common_dir {
592 cx.emit(Event::UpdatedRootRepoCommonDir);
593 }
594 cx.notify();
595 while let Some((scan_id, _)) = this.snapshot_subscriptions.front() {
596 if this.observed_snapshot(*scan_id) {
597 let (_, tx) = this.snapshot_subscriptions.pop_front().unwrap();
598 let _ = tx.send(());
599 } else {
600 break;
601 }
602 }
603 })?;
604 }
605 anyhow::Ok(())
606 })
607 .detach();
608
609 Worktree::Remote(worktree)
610 })
611 }
612
613 pub fn as_local(&self) -> Option<&LocalWorktree> {
614 if let Worktree::Local(worktree) = self {
615 Some(worktree)
616 } else {
617 None
618 }
619 }
620
621 pub fn as_remote(&self) -> Option<&RemoteWorktree> {
622 if let Worktree::Remote(worktree) = self {
623 Some(worktree)
624 } else {
625 None
626 }
627 }
628
629 pub fn as_local_mut(&mut self) -> Option<&mut LocalWorktree> {
630 if let Worktree::Local(worktree) = self {
631 Some(worktree)
632 } else {
633 None
634 }
635 }
636
637 pub fn as_remote_mut(&mut self) -> Option<&mut RemoteWorktree> {
638 if let Worktree::Remote(worktree) = self {
639 Some(worktree)
640 } else {
641 None
642 }
643 }
644
645 pub fn is_local(&self) -> bool {
646 matches!(self, Worktree::Local(_))
647 }
648
649 pub fn is_remote(&self) -> bool {
650 !self.is_local()
651 }
652
653 pub fn settings_location(&self, _: &Context<Self>) -> SettingsLocation<'static> {
654 SettingsLocation {
655 worktree_id: self.id(),
656 path: RelPath::empty(),
657 }
658 }
659
660 pub fn snapshot(&self) -> Snapshot {
661 match self {
662 Worktree::Local(worktree) => worktree.snapshot.snapshot.clone(),
663 Worktree::Remote(worktree) => worktree.snapshot.clone(),
664 }
665 }
666
667 pub fn scan_id(&self) -> usize {
668 match self {
669 Worktree::Local(worktree) => worktree.snapshot.scan_id,
670 Worktree::Remote(worktree) => worktree.snapshot.scan_id,
671 }
672 }
673
674 pub fn metadata_proto(&self) -> proto::WorktreeMetadata {
675 proto::WorktreeMetadata {
676 id: self.id().to_proto(),
677 root_name: self.root_name().to_proto(),
678 visible: self.is_visible(),
679 abs_path: self.abs_path().to_string_lossy().into_owned(),
680 }
681 }
682
683 pub fn completed_scan_id(&self) -> usize {
684 match self {
685 Worktree::Local(worktree) => worktree.snapshot.completed_scan_id,
686 Worktree::Remote(worktree) => worktree.snapshot.completed_scan_id,
687 }
688 }
689
690 pub fn is_visible(&self) -> bool {
691 match self {
692 Worktree::Local(worktree) => worktree.visible,
693 Worktree::Remote(worktree) => worktree.visible,
694 }
695 }
696
697 pub fn replica_id(&self) -> ReplicaId {
698 match self {
699 Worktree::Local(_) => ReplicaId::LOCAL,
700 Worktree::Remote(worktree) => worktree.replica_id,
701 }
702 }
703
704 pub fn abs_path(&self) -> Arc<Path> {
705 match self {
706 Worktree::Local(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
707 Worktree::Remote(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
708 }
709 }
710
711 pub fn root_file(&self, cx: &Context<Self>) -> Option<Arc<File>> {
712 let entry = self.root_entry()?;
713 Some(File::for_entry(entry.clone(), cx.entity()))
714 }
715
716 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
717 where
718 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
719 Fut: 'static + Send + Future<Output = bool>,
720 {
721 match self {
722 Worktree::Local(this) => this.observe_updates(project_id, cx, callback),
723 Worktree::Remote(this) => this.observe_updates(project_id, cx, callback),
724 }
725 }
726
727 pub fn stop_observing_updates(&mut self) {
728 match self {
729 Worktree::Local(this) => {
730 this.update_observer.take();
731 }
732 Worktree::Remote(this) => {
733 this.update_observer.take();
734 }
735 }
736 }
737
738 pub fn wait_for_snapshot(
739 &mut self,
740 scan_id: usize,
741 ) -> impl Future<Output = Result<()>> + use<> {
742 match self {
743 Worktree::Local(this) => this.wait_for_snapshot(scan_id).boxed(),
744 Worktree::Remote(this) => this.wait_for_snapshot(scan_id).boxed(),
745 }
746 }
747
748 #[cfg(feature = "test-support")]
749 pub fn has_update_observer(&self) -> bool {
750 match self {
751 Worktree::Local(this) => this.update_observer.is_some(),
752 Worktree::Remote(this) => this.update_observer.is_some(),
753 }
754 }
755
756 pub fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
757 match self {
758 Worktree::Local(this) => this.load_file(path, cx),
759 Worktree::Remote(_) => {
760 Task::ready(Err(anyhow!("remote worktrees can't yet load files")))
761 }
762 }
763 }
764
765 pub fn load_binary_file(
766 &self,
767 path: &RelPath,
768 cx: &Context<Worktree>,
769 ) -> Task<Result<LoadedBinaryFile>> {
770 match self {
771 Worktree::Local(this) => this.load_binary_file(path, cx),
772 Worktree::Remote(_) => {
773 Task::ready(Err(anyhow!("remote worktrees can't yet load binary files")))
774 }
775 }
776 }
777
778 pub fn write_file(
779 &self,
780 path: Arc<RelPath>,
781 text: Rope,
782 line_ending: LineEnding,
783 encoding: &'static Encoding,
784 has_bom: bool,
785 cx: &Context<Worktree>,
786 ) -> Task<Result<Arc<File>>> {
787 match self {
788 Worktree::Local(this) => {
789 this.write_file(path, text, line_ending, encoding, has_bom, cx)
790 }
791 Worktree::Remote(_) => {
792 Task::ready(Err(anyhow!("remote worktree can't yet write files")))
793 }
794 }
795 }
796
797 pub fn create_entry(
798 &mut self,
799 path: Arc<RelPath>,
800 is_directory: bool,
801 content: Option<Vec<u8>>,
802 cx: &Context<Worktree>,
803 ) -> Task<Result<CreatedEntry>> {
804 let worktree_id = self.id();
805 match self {
806 Worktree::Local(this) => this.create_entry(path, is_directory, content, cx),
807 Worktree::Remote(this) => {
808 let project_id = this.project_id;
809 let request = this.client.request(proto::CreateProjectEntry {
810 worktree_id: worktree_id.to_proto(),
811 project_id,
812 path: path.as_ref().to_proto(),
813 content,
814 is_directory,
815 });
816 cx.spawn(async move |this, cx| {
817 let response = request.await?;
818 match response.entry {
819 Some(entry) => this
820 .update(cx, |worktree, cx| {
821 worktree.as_remote_mut().unwrap().insert_entry(
822 entry,
823 response.worktree_scan_id as usize,
824 cx,
825 )
826 })?
827 .await
828 .map(CreatedEntry::Included),
829 None => {
830 let abs_path =
831 this.read_with(cx, |worktree, _| worktree.absolutize(&path))?;
832 Ok(CreatedEntry::Excluded { abs_path })
833 }
834 }
835 })
836 }
837 }
838 }
839
840 pub fn delete_entry(
841 &mut self,
842 entry_id: ProjectEntryId,
843 trash: bool,
844 cx: &mut Context<Worktree>,
845 ) -> Option<Task<Result<Option<TrashedEntry>>>> {
846 let task = match self {
847 Worktree::Local(this) => this.delete_entry(entry_id, trash, cx),
848 Worktree::Remote(this) => this.delete_entry(entry_id, trash, cx),
849 }?;
850
851 let entry = match &*self {
852 Worktree::Local(this) => this.entry_for_id(entry_id),
853 Worktree::Remote(this) => this.entry_for_id(entry_id),
854 }?;
855
856 let mut ids = vec![entry_id];
857 let path = &*entry.path;
858
859 self.get_children_ids_recursive(path, &mut ids);
860
861 for id in ids {
862 cx.emit(Event::DeletedEntry(id));
863 }
864 Some(task)
865 }
866
867 pub async fn restore_entry(
868 trash_entry: TrashedEntry,
869 worktree: Entity<Self>,
870 cx: &mut AsyncApp,
871 ) -> Result<RelPathBuf> {
872 let is_local = worktree.read_with(cx, |this, _| this.is_local());
873 if is_local {
874 LocalWorktree::restore_entry(trash_entry, worktree, cx).await
875 } else {
876 // TODO(dino): Add support for restoring entries in remote worktrees.
877 Err(anyhow!("Unsupported"))
878 }
879 }
880
881 fn get_children_ids_recursive(&self, path: &RelPath, ids: &mut Vec<ProjectEntryId>) {
882 let children_iter = self.child_entries(path);
883 for child in children_iter {
884 ids.push(child.id);
885 self.get_children_ids_recursive(&child.path, ids);
886 }
887 }
888
889 // pub fn rename_entry(
890 // &mut self,
891 // entry_id: ProjectEntryId,
892 // new_path: Arc<RelPath>,
893 // cx: &Context<Self>,
894 // ) -> Task<Result<CreatedEntry>> {
895 // match self {
896 // Worktree::Local(this) => this.rename_entry(entry_id, new_path, cx),
897 // Worktree::Remote(this) => this.rename_entry(entry_id, new_path, cx),
898 // }
899 // }
900
901 pub fn copy_external_entries(
902 &mut self,
903 target_directory: Arc<RelPath>,
904 paths: Vec<Arc<Path>>,
905 fs: Arc<dyn Fs>,
906 cx: &Context<Worktree>,
907 ) -> Task<Result<Vec<ProjectEntryId>>> {
908 match self {
909 Worktree::Local(this) => this.copy_external_entries(target_directory, paths, cx),
910 Worktree::Remote(this) => this.copy_external_entries(target_directory, paths, fs, cx),
911 }
912 }
913
914 pub fn expand_entry(
915 &mut self,
916 entry_id: ProjectEntryId,
917 cx: &Context<Worktree>,
918 ) -> Option<Task<Result<()>>> {
919 match self {
920 Worktree::Local(this) => this.expand_entry(entry_id, cx),
921 Worktree::Remote(this) => {
922 let response = this.client.request(proto::ExpandProjectEntry {
923 project_id: this.project_id,
924 entry_id: entry_id.to_proto(),
925 });
926 Some(cx.spawn(async move |this, cx| {
927 let response = response.await?;
928 this.update(cx, |this, _| {
929 this.as_remote_mut()
930 .unwrap()
931 .wait_for_snapshot(response.worktree_scan_id as usize)
932 })?
933 .await?;
934 Ok(())
935 }))
936 }
937 }
938 }
939
940 pub fn expand_all_for_entry(
941 &mut self,
942 entry_id: ProjectEntryId,
943 cx: &Context<Worktree>,
944 ) -> Option<Task<Result<()>>> {
945 match self {
946 Worktree::Local(this) => this.expand_all_for_entry(entry_id, cx),
947 Worktree::Remote(this) => {
948 let response = this.client.request(proto::ExpandAllForProjectEntry {
949 project_id: this.project_id,
950 entry_id: entry_id.to_proto(),
951 });
952 Some(cx.spawn(async move |this, cx| {
953 let response = response.await?;
954 this.update(cx, |this, _| {
955 this.as_remote_mut()
956 .unwrap()
957 .wait_for_snapshot(response.worktree_scan_id as usize)
958 })?
959 .await?;
960 Ok(())
961 }))
962 }
963 }
964 }
965
966 pub async fn handle_create_entry(
967 this: Entity<Self>,
968 request: proto::CreateProjectEntry,
969 mut cx: AsyncApp,
970 ) -> Result<proto::ProjectEntryResponse> {
971 let (scan_id, entry) = this.update(&mut cx, |this, cx| {
972 anyhow::Ok((
973 this.scan_id(),
974 this.create_entry(
975 RelPath::from_proto(&request.path).with_context(|| {
976 format!("received invalid relative path {:?}", request.path)
977 })?,
978 request.is_directory,
979 request.content,
980 cx,
981 ),
982 ))
983 })?;
984 Ok(proto::ProjectEntryResponse {
985 entry: match &entry.await? {
986 CreatedEntry::Included(entry) => Some(entry.into()),
987 CreatedEntry::Excluded { .. } => None,
988 },
989 worktree_scan_id: scan_id as u64,
990 })
991 }
992
993 pub async fn handle_delete_entry(
994 this: Entity<Self>,
995 request: proto::DeleteProjectEntry,
996 mut cx: AsyncApp,
997 ) -> Result<proto::ProjectEntryResponse> {
998 let (scan_id, task) = this.update(&mut cx, |this, cx| {
999 (
1000 this.scan_id(),
1001 this.delete_entry(
1002 ProjectEntryId::from_proto(request.entry_id),
1003 request.use_trash,
1004 cx,
1005 ),
1006 )
1007 });
1008 task.ok_or_else(|| anyhow::anyhow!("invalid entry"))?
1009 .await?;
1010 Ok(proto::ProjectEntryResponse {
1011 entry: None,
1012 worktree_scan_id: scan_id as u64,
1013 })
1014 }
1015
1016 pub async fn handle_expand_entry(
1017 this: Entity<Self>,
1018 request: proto::ExpandProjectEntry,
1019 mut cx: AsyncApp,
1020 ) -> Result<proto::ExpandProjectEntryResponse> {
1021 let task = this.update(&mut cx, |this, cx| {
1022 this.expand_entry(ProjectEntryId::from_proto(request.entry_id), cx)
1023 });
1024 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
1025 .await?;
1026 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
1027 Ok(proto::ExpandProjectEntryResponse {
1028 worktree_scan_id: scan_id as u64,
1029 })
1030 }
1031
1032 pub async fn handle_expand_all_for_entry(
1033 this: Entity<Self>,
1034 request: proto::ExpandAllForProjectEntry,
1035 mut cx: AsyncApp,
1036 ) -> Result<proto::ExpandAllForProjectEntryResponse> {
1037 let task = this.update(&mut cx, |this, cx| {
1038 this.expand_all_for_entry(ProjectEntryId::from_proto(request.entry_id), cx)
1039 });
1040 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
1041 .await?;
1042 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
1043 Ok(proto::ExpandAllForProjectEntryResponse {
1044 worktree_scan_id: scan_id as u64,
1045 })
1046 }
1047
1048 pub fn is_single_file(&self) -> bool {
1049 self.root_dir().is_none()
1050 }
1051
1052 /// For visible worktrees, returns the path with the worktree name as the first component.
1053 /// Otherwise, returns an absolute path.
1054 pub fn full_path(&self, worktree_relative_path: &RelPath) -> PathBuf {
1055 if self.is_visible() {
1056 self.root_name()
1057 .join(worktree_relative_path)
1058 .display(self.path_style)
1059 .to_string()
1060 .into()
1061 } else {
1062 let full_path = self.abs_path();
1063 let mut full_path_string = if self.is_local()
1064 && let Ok(stripped) = full_path.strip_prefix(home_dir())
1065 {
1066 self.path_style
1067 .join("~", &*stripped.to_string_lossy())
1068 .unwrap()
1069 } else {
1070 full_path.to_string_lossy().into_owned()
1071 };
1072
1073 if worktree_relative_path.components().next().is_some() {
1074 full_path_string.push_str(self.path_style.primary_separator());
1075 full_path_string.push_str(&worktree_relative_path.display(self.path_style));
1076 }
1077
1078 full_path_string.into()
1079 }
1080 }
1081}
1082
1083impl LocalWorktree {
1084 pub fn fs(&self) -> &Arc<dyn Fs> {
1085 &self.fs
1086 }
1087
1088 pub fn is_path_private(&self, path: &RelPath) -> bool {
1089 !self.share_private_files && self.settings.is_path_private(path)
1090 }
1091
1092 pub fn fs_is_case_sensitive(&self) -> bool {
1093 self.fs_case_sensitive
1094 }
1095
1096 fn restart_background_scanners(&mut self, cx: &Context<Worktree>) {
1097 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
1098 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
1099 self.scan_requests_tx = scan_requests_tx;
1100 self.path_prefixes_to_scan_tx = path_prefixes_to_scan_tx;
1101
1102 self.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
1103 let always_included_entries = mem::take(&mut self.snapshot.always_included_entries);
1104 log::debug!(
1105 "refreshing entries for the following always included paths: {:?}",
1106 always_included_entries
1107 );
1108
1109 // Cleans up old always included entries to ensure they get updated properly. Otherwise,
1110 // nested always included entries may not get updated and will result in out-of-date info.
1111 self.refresh_entries_for_paths(always_included_entries);
1112 }
1113
1114 fn start_background_scanner(
1115 &mut self,
1116 scan_requests_rx: channel::Receiver<ScanRequest>,
1117 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
1118 cx: &Context<Worktree>,
1119 ) {
1120 let snapshot = self.snapshot();
1121 let share_private_files = self.share_private_files;
1122 let next_entry_id = self.next_entry_id.clone();
1123 let fs = self.fs.clone();
1124 let scanning_enabled = self.scanning_enabled;
1125 let settings = self.settings.clone();
1126 let (scan_states_tx, mut scan_states_rx) = mpsc::unbounded();
1127 let background_scanner = cx.background_spawn({
1128 let abs_path = snapshot.abs_path.as_path().to_path_buf();
1129 let background = cx.background_executor().clone();
1130 async move {
1131 let (events, watcher) = if scanning_enabled {
1132 fs.watch(&abs_path, FS_WATCH_LATENCY).await
1133 } else {
1134 (Box::pin(stream::pending()) as _, Arc::new(NullWatcher) as _)
1135 };
1136 let fs_case_sensitive = fs.is_case_sensitive().await;
1137
1138 let is_single_file = snapshot.snapshot.root_dir().is_none();
1139 let mut scanner = BackgroundScanner {
1140 fs,
1141 fs_case_sensitive,
1142 status_updates_tx: scan_states_tx,
1143 executor: background,
1144 scan_requests_rx,
1145 path_prefixes_to_scan_rx,
1146 next_entry_id,
1147 state: async_lock::Mutex::new(BackgroundScannerState {
1148 prev_snapshot: snapshot.snapshot.clone(),
1149 snapshot,
1150 scanned_dirs: Default::default(),
1151 scanning_enabled,
1152 path_prefixes_to_scan: Default::default(),
1153 paths_to_scan: Default::default(),
1154 removed_entries: Default::default(),
1155 changed_paths: Default::default(),
1156 }),
1157 phase: BackgroundScannerPhase::InitialScan,
1158 share_private_files,
1159 settings,
1160 watcher,
1161 is_single_file,
1162 };
1163
1164 scanner
1165 .run(Box::pin(events.map(|events| events.into_iter().collect())))
1166 .await;
1167 }
1168 });
1169 let scan_state_updater = cx.spawn(async move |this, cx| {
1170 while let Some((state, this)) = scan_states_rx.next().await.zip(this.upgrade()) {
1171 this.update(cx, |this, cx| {
1172 let this = this.as_local_mut().unwrap();
1173 match state {
1174 ScanState::Started => {
1175 *this.is_scanning.0.borrow_mut() = true;
1176 }
1177 ScanState::Updated {
1178 snapshot,
1179 changes,
1180 barrier,
1181 scanning,
1182 } => {
1183 *this.is_scanning.0.borrow_mut() = scanning;
1184 this.set_snapshot(snapshot, changes, cx);
1185 drop(barrier);
1186 }
1187 ScanState::RootUpdated { new_path } => {
1188 this.update_abs_path_and_refresh(new_path, cx);
1189 }
1190 ScanState::RootDeleted => {
1191 log::info!(
1192 "worktree root {} no longer exists, closing worktree",
1193 this.abs_path().display()
1194 );
1195 cx.emit(Event::Deleted);
1196 }
1197 }
1198 });
1199 }
1200 });
1201 self._background_scanner_tasks = vec![background_scanner, scan_state_updater];
1202 *self.is_scanning.0.borrow_mut() = true;
1203 }
1204
1205 fn set_snapshot(
1206 &mut self,
1207 mut new_snapshot: LocalSnapshot,
1208 entry_changes: UpdatedEntriesSet,
1209 cx: &mut Context<Worktree>,
1210 ) {
1211 let repo_changes = self.changed_repos(&self.snapshot, &mut new_snapshot);
1212
1213 new_snapshot.root_repo_common_dir = new_snapshot
1214 .local_repo_for_work_directory_path(RelPath::empty())
1215 .map(|repo| SanitizedPath::from_arc(repo.common_dir_abs_path.clone()));
1216
1217 let root_repo_common_dir_changed =
1218 self.snapshot.root_repo_common_dir != new_snapshot.root_repo_common_dir;
1219 self.snapshot = new_snapshot;
1220
1221 if let Some(share) = self.update_observer.as_mut() {
1222 share
1223 .snapshots_tx
1224 .unbounded_send((self.snapshot.clone(), entry_changes.clone()))
1225 .ok();
1226 }
1227
1228 if !entry_changes.is_empty() {
1229 cx.emit(Event::UpdatedEntries(entry_changes));
1230 }
1231 if !repo_changes.is_empty() {
1232 cx.emit(Event::UpdatedGitRepositories(repo_changes));
1233 }
1234 if root_repo_common_dir_changed {
1235 cx.emit(Event::UpdatedRootRepoCommonDir);
1236 }
1237
1238 while let Some((scan_id, _)) = self.snapshot_subscriptions.front() {
1239 if self.snapshot.completed_scan_id >= *scan_id {
1240 let (_, tx) = self.snapshot_subscriptions.pop_front().unwrap();
1241 tx.send(()).ok();
1242 } else {
1243 break;
1244 }
1245 }
1246 }
1247
1248 fn changed_repos(
1249 &self,
1250 old_snapshot: &LocalSnapshot,
1251 new_snapshot: &mut LocalSnapshot,
1252 ) -> UpdatedGitRepositoriesSet {
1253 let mut changes = Vec::new();
1254 let mut old_repos = old_snapshot.git_repositories.iter().peekable();
1255 let new_repos = new_snapshot.git_repositories.clone();
1256 let mut new_repos = new_repos.iter().peekable();
1257
1258 loop {
1259 match (new_repos.peek().map(clone), old_repos.peek().map(clone)) {
1260 (Some((new_entry_id, new_repo)), Some((old_entry_id, old_repo))) => {
1261 match Ord::cmp(&new_entry_id, &old_entry_id) {
1262 Ordering::Less => {
1263 changes.push(UpdatedGitRepository {
1264 work_directory_id: new_entry_id,
1265 old_work_directory_abs_path: None,
1266 new_work_directory_abs_path: Some(
1267 new_repo.work_directory_abs_path.clone(),
1268 ),
1269 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1270 repository_dir_abs_path: Some(
1271 new_repo.repository_dir_abs_path.clone(),
1272 ),
1273 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1274 });
1275 new_repos.next();
1276 }
1277 Ordering::Equal => {
1278 if new_repo.git_dir_scan_id != old_repo.git_dir_scan_id
1279 || new_repo.work_directory_abs_path
1280 != old_repo.work_directory_abs_path
1281 {
1282 changes.push(UpdatedGitRepository {
1283 work_directory_id: new_entry_id,
1284 old_work_directory_abs_path: Some(
1285 old_repo.work_directory_abs_path.clone(),
1286 ),
1287 new_work_directory_abs_path: Some(
1288 new_repo.work_directory_abs_path.clone(),
1289 ),
1290 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1291 repository_dir_abs_path: Some(
1292 new_repo.repository_dir_abs_path.clone(),
1293 ),
1294 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1295 });
1296 }
1297 new_repos.next();
1298 old_repos.next();
1299 }
1300 Ordering::Greater => {
1301 changes.push(UpdatedGitRepository {
1302 work_directory_id: old_entry_id,
1303 old_work_directory_abs_path: Some(
1304 old_repo.work_directory_abs_path.clone(),
1305 ),
1306 new_work_directory_abs_path: None,
1307 dot_git_abs_path: None,
1308 repository_dir_abs_path: None,
1309 common_dir_abs_path: None,
1310 });
1311 old_repos.next();
1312 }
1313 }
1314 }
1315 (Some((entry_id, repo)), None) => {
1316 changes.push(UpdatedGitRepository {
1317 work_directory_id: entry_id,
1318 old_work_directory_abs_path: None,
1319 new_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1320 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1321 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1322 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1323 });
1324 new_repos.next();
1325 }
1326 (None, Some((entry_id, repo))) => {
1327 changes.push(UpdatedGitRepository {
1328 work_directory_id: entry_id,
1329 old_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1330 new_work_directory_abs_path: None,
1331 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1332 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1333 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1334 });
1335 old_repos.next();
1336 }
1337 (None, None) => break,
1338 }
1339 }
1340
1341 fn clone<T: Clone, U: Clone>(value: &(&T, &U)) -> (T, U) {
1342 (value.0.clone(), value.1.clone())
1343 }
1344
1345 changes.into()
1346 }
1347
1348 pub fn scan_complete(&self) -> impl Future<Output = ()> + use<> {
1349 let mut is_scanning_rx = self.is_scanning.1.clone();
1350 async move {
1351 let mut is_scanning = *is_scanning_rx.borrow();
1352 while is_scanning {
1353 if let Some(value) = is_scanning_rx.recv().await {
1354 is_scanning = value;
1355 } else {
1356 break;
1357 }
1358 }
1359 }
1360 }
1361
1362 pub fn wait_for_snapshot(
1363 &mut self,
1364 scan_id: usize,
1365 ) -> impl Future<Output = Result<()>> + use<> {
1366 let (tx, rx) = oneshot::channel();
1367 if self.snapshot.completed_scan_id >= scan_id {
1368 tx.send(()).ok();
1369 } else {
1370 match self
1371 .snapshot_subscriptions
1372 .binary_search_by_key(&scan_id, |probe| probe.0)
1373 {
1374 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
1375 }
1376 }
1377
1378 async move {
1379 rx.await?;
1380 Ok(())
1381 }
1382 }
1383
1384 pub fn snapshot(&self) -> LocalSnapshot {
1385 self.snapshot.clone()
1386 }
1387
1388 pub fn settings(&self) -> WorktreeSettings {
1389 self.settings.clone()
1390 }
1391
1392 fn load_binary_file(
1393 &self,
1394 path: &RelPath,
1395 cx: &Context<Worktree>,
1396 ) -> Task<Result<LoadedBinaryFile>> {
1397 let path = Arc::from(path);
1398 let abs_path = self.absolutize(&path);
1399 let fs = self.fs.clone();
1400 let entry = self.refresh_entry(path.clone(), None, cx);
1401 let is_private = self.is_path_private(&path);
1402
1403 let worktree = cx.weak_entity();
1404 cx.background_spawn(async move {
1405 let content = fs.load_bytes(&abs_path).await?;
1406
1407 let worktree = worktree.upgrade().context("worktree was dropped")?;
1408 let file = match entry.await? {
1409 Some(entry) => File::for_entry(entry, worktree),
1410 None => {
1411 let metadata = fs
1412 .metadata(&abs_path)
1413 .await
1414 .with_context(|| {
1415 format!("Loading metadata for excluded file {abs_path:?}")
1416 })?
1417 .with_context(|| {
1418 format!("Excluded file {abs_path:?} got removed during loading")
1419 })?;
1420 Arc::new(File {
1421 entry_id: None,
1422 worktree,
1423 path,
1424 disk_state: DiskState::Present {
1425 mtime: metadata.mtime,
1426 size: metadata.len,
1427 },
1428 is_local: true,
1429 is_private,
1430 })
1431 }
1432 };
1433
1434 Ok(LoadedBinaryFile { file, content })
1435 })
1436 }
1437
1438 #[ztracing::instrument(skip_all)]
1439 fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
1440 let path = Arc::from(path);
1441 let abs_path = self.absolutize(&path);
1442 let fs = self.fs.clone();
1443 let entry = self.refresh_entry(path.clone(), None, cx);
1444 let is_private = self.is_path_private(path.as_ref());
1445
1446 let this = cx.weak_entity();
1447 cx.background_spawn(async move {
1448 // WARN: Temporary workaround for #27283.
1449 // We are not efficient with our memory usage per file, and use in excess of 64GB for a 10GB file
1450 // Therefore, as a temporary workaround to prevent system freezes, we just bail before opening a file
1451 // if it is too large
1452 // 5GB seems to be more reasonable, peaking at ~16GB, while 6GB jumps up to >24GB which seems like a
1453 // reasonable limit
1454 {
1455 const FILE_SIZE_MAX: u64 = 6 * 1024 * 1024 * 1024; // 6GB
1456 if let Ok(Some(metadata)) = fs.metadata(&abs_path).await
1457 && metadata.len >= FILE_SIZE_MAX
1458 {
1459 anyhow::bail!("File is too large to load");
1460 }
1461 }
1462 let (text, encoding, has_bom) = decode_file_text(fs.as_ref(), &abs_path).await?;
1463
1464 let worktree = this.upgrade().context("worktree was dropped")?;
1465 let file = match entry.await? {
1466 Some(entry) => File::for_entry(entry, worktree),
1467 None => {
1468 let metadata = fs
1469 .metadata(&abs_path)
1470 .await
1471 .with_context(|| {
1472 format!("Loading metadata for excluded file {abs_path:?}")
1473 })?
1474 .with_context(|| {
1475 format!("Excluded file {abs_path:?} got removed during loading")
1476 })?;
1477 Arc::new(File {
1478 entry_id: None,
1479 worktree,
1480 path,
1481 disk_state: DiskState::Present {
1482 mtime: metadata.mtime,
1483 size: metadata.len,
1484 },
1485 is_local: true,
1486 is_private,
1487 })
1488 }
1489 };
1490
1491 Ok(LoadedFile {
1492 file,
1493 text,
1494 encoding,
1495 has_bom,
1496 })
1497 })
1498 }
1499
1500 /// Find the lowest path in the worktree's datastructures that is an ancestor
1501 fn lowest_ancestor(&self, path: &RelPath) -> Arc<RelPath> {
1502 let mut lowest_ancestor = None;
1503 for path in path.ancestors() {
1504 if self.entry_for_path(path).is_some() {
1505 lowest_ancestor = Some(path.into());
1506 break;
1507 }
1508 }
1509
1510 lowest_ancestor.unwrap_or_else(|| RelPath::empty().into())
1511 }
1512
1513 pub fn create_entry(
1514 &self,
1515 path: Arc<RelPath>,
1516 is_dir: bool,
1517 content: Option<Vec<u8>>,
1518 cx: &Context<Worktree>,
1519 ) -> Task<Result<CreatedEntry>> {
1520 let abs_path = self.absolutize(&path);
1521 let path_excluded = self.settings.is_path_excluded(&path);
1522 let fs = self.fs.clone();
1523 let task_abs_path = abs_path.clone();
1524 let write = cx.background_spawn(async move {
1525 if is_dir {
1526 fs.create_dir(&task_abs_path)
1527 .await
1528 .with_context(|| format!("creating directory {task_abs_path:?}"))
1529 } else {
1530 fs.write(&task_abs_path, content.as_deref().unwrap_or(&[]))
1531 .await
1532 .with_context(|| format!("creating file {task_abs_path:?}"))
1533 }
1534 });
1535
1536 let lowest_ancestor = self.lowest_ancestor(&path);
1537 cx.spawn(async move |this, cx| {
1538 write.await?;
1539 if path_excluded {
1540 return Ok(CreatedEntry::Excluded { abs_path });
1541 }
1542
1543 let (result, refreshes) = this.update(cx, |this, cx| {
1544 let mut refreshes = Vec::new();
1545 let refresh_paths = path.strip_prefix(&lowest_ancestor).unwrap();
1546 for refresh_path in refresh_paths.ancestors() {
1547 if refresh_path == RelPath::empty() {
1548 continue;
1549 }
1550 let refresh_full_path = lowest_ancestor.join(refresh_path);
1551
1552 refreshes.push(this.as_local_mut().unwrap().refresh_entry(
1553 refresh_full_path,
1554 None,
1555 cx,
1556 ));
1557 }
1558 (
1559 this.as_local_mut().unwrap().refresh_entry(path, None, cx),
1560 refreshes,
1561 )
1562 })?;
1563 for refresh in refreshes {
1564 refresh.await.log_err();
1565 }
1566
1567 Ok(result
1568 .await?
1569 .map(CreatedEntry::Included)
1570 .unwrap_or_else(|| CreatedEntry::Excluded { abs_path }))
1571 })
1572 }
1573
1574 pub fn write_file(
1575 &self,
1576 path: Arc<RelPath>,
1577 text: Rope,
1578 line_ending: LineEnding,
1579 encoding: &'static Encoding,
1580 has_bom: bool,
1581 cx: &Context<Worktree>,
1582 ) -> Task<Result<Arc<File>>> {
1583 let fs = self.fs.clone();
1584 let is_private = self.is_path_private(&path);
1585 let abs_path = self.absolutize(&path);
1586
1587 let write = cx.background_spawn({
1588 let fs = fs.clone();
1589 let abs_path = abs_path.clone();
1590 async move {
1591 // For UTF-8, use the optimized `fs.save` which writes Rope chunks directly to disk
1592 // without allocating a contiguous string.
1593 if encoding == encoding_rs::UTF_8 && !has_bom {
1594 return fs.save(&abs_path, &text, line_ending).await;
1595 }
1596
1597 // For legacy encodings (e.g. Shift-JIS), we fall back to converting the entire Rope
1598 // to a String/Bytes in memory before writing.
1599 //
1600 // Note: This is inefficient for very large files compared to the streaming approach above,
1601 // but supporting streaming writes for arbitrary encodings would require a significant
1602 // refactor of the `fs` crate to expose a Writer interface.
1603 let text_string = text.to_string();
1604 let normalized_text = match line_ending {
1605 LineEnding::Unix => text_string,
1606 LineEnding::Windows => text_string.replace('\n', "\r\n"),
1607 };
1608
1609 // Create the byte vector manually for UTF-16 encodings because encoding_rs encodes to UTF-8 by default (per WHATWG standards),
1610 // which is not what we want for saving files.
1611 let bytes = if encoding == encoding_rs::UTF_16BE {
1612 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1613 if has_bom {
1614 data.extend_from_slice(&[0xFE, 0xFF]); // BOM
1615 }
1616 let utf16be_bytes =
1617 normalized_text.encode_utf16().flat_map(|u| u.to_be_bytes());
1618 data.extend(utf16be_bytes);
1619 data.into()
1620 } else if encoding == encoding_rs::UTF_16LE {
1621 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1622 if has_bom {
1623 data.extend_from_slice(&[0xFF, 0xFE]); // BOM
1624 }
1625 let utf16le_bytes =
1626 normalized_text.encode_utf16().flat_map(|u| u.to_le_bytes());
1627 data.extend(utf16le_bytes);
1628 data.into()
1629 } else {
1630 // For other encodings (Shift-JIS, UTF-8 with BOM, etc.), delegate to encoding_rs.
1631 let bom_bytes = if has_bom {
1632 if encoding == encoding_rs::UTF_8 {
1633 vec![0xEF, 0xBB, 0xBF]
1634 } else {
1635 vec![]
1636 }
1637 } else {
1638 vec![]
1639 };
1640 let (cow, _, _) = encoding.encode(&normalized_text);
1641 if !bom_bytes.is_empty() {
1642 let mut bytes = bom_bytes;
1643 bytes.extend_from_slice(&cow);
1644 bytes.into()
1645 } else {
1646 cow
1647 }
1648 };
1649
1650 fs.write(&abs_path, &bytes).await
1651 }
1652 });
1653
1654 cx.spawn(async move |this, cx| {
1655 write.await?;
1656 let entry = this
1657 .update(cx, |this, cx| {
1658 this.as_local_mut()
1659 .unwrap()
1660 .refresh_entry(path.clone(), None, cx)
1661 })?
1662 .await?;
1663 let worktree = this.upgrade().context("worktree dropped")?;
1664 if let Some(entry) = entry {
1665 Ok(File::for_entry(entry, worktree))
1666 } else {
1667 let metadata = fs
1668 .metadata(&abs_path)
1669 .await
1670 .with_context(|| {
1671 format!("Fetching metadata after saving the excluded buffer {abs_path:?}")
1672 })?
1673 .with_context(|| {
1674 format!("Excluded buffer {path:?} got removed during saving")
1675 })?;
1676 Ok(Arc::new(File {
1677 worktree,
1678 path,
1679 disk_state: DiskState::Present {
1680 mtime: metadata.mtime,
1681 size: metadata.len,
1682 },
1683 entry_id: None,
1684 is_local: true,
1685 is_private,
1686 }))
1687 }
1688 })
1689 }
1690
1691 pub fn delete_entry(
1692 &self,
1693 entry_id: ProjectEntryId,
1694 trash: bool,
1695 cx: &Context<Worktree>,
1696 ) -> Option<Task<Result<Option<TrashedEntry>>>> {
1697 let entry = self.entry_for_id(entry_id)?.clone();
1698 let abs_path = self.absolutize(&entry.path);
1699 let fs = self.fs.clone();
1700
1701 let delete = cx.background_spawn(async move {
1702 let trashed_entry = match (entry.is_file(), trash) {
1703 (true, true) => Some(fs.trash(&abs_path, Default::default()).await?),
1704 (false, true) => Some(
1705 fs.trash(
1706 &abs_path,
1707 RemoveOptions {
1708 recursive: true,
1709 ignore_if_not_exists: false,
1710 },
1711 )
1712 .await?,
1713 ),
1714 (true, false) => {
1715 fs.remove_file(&abs_path, Default::default()).await?;
1716 None
1717 }
1718 (false, false) => {
1719 fs.remove_dir(
1720 &abs_path,
1721 RemoveOptions {
1722 recursive: true,
1723 ignore_if_not_exists: false,
1724 },
1725 )
1726 .await?;
1727 None
1728 }
1729 };
1730
1731 anyhow::Ok((trashed_entry, entry.path))
1732 });
1733
1734 Some(cx.spawn(async move |this, cx| {
1735 let (trashed_entry, path) = delete.await?;
1736 this.update(cx, |this, _| {
1737 this.as_local_mut()
1738 .unwrap()
1739 .refresh_entries_for_paths(vec![path])
1740 })?
1741 .recv()
1742 .await;
1743
1744 Ok(trashed_entry)
1745 }))
1746 }
1747
1748 pub async fn restore_entry(
1749 trash_entry: TrashedEntry,
1750 this: Entity<Worktree>,
1751 cx: &mut AsyncApp,
1752 ) -> Result<RelPathBuf> {
1753 let Some((fs, worktree_abs_path, path_style)) = this.read_with(cx, |this, _cx| {
1754 let local_worktree = match this {
1755 Worktree::Local(local_worktree) => local_worktree,
1756 Worktree::Remote(_) => return None,
1757 };
1758
1759 let fs = local_worktree.fs.clone();
1760 let path_style = local_worktree.path_style();
1761 Some((fs, Arc::clone(local_worktree.abs_path()), path_style))
1762 }) else {
1763 return Err(anyhow!("Localworktree should not change into a remote one"));
1764 };
1765
1766 let path_buf = fs.restore(trash_entry).await?;
1767 let path = path_buf
1768 .strip_prefix(worktree_abs_path)
1769 .context("Could not strip prefix")?;
1770 let path = RelPath::new(&path, path_style)?;
1771 let path = path.into_owned();
1772
1773 Ok(path)
1774 }
1775
1776 pub fn copy_external_entries(
1777 &self,
1778 target_directory: Arc<RelPath>,
1779 paths: Vec<Arc<Path>>,
1780 cx: &Context<Worktree>,
1781 ) -> Task<Result<Vec<ProjectEntryId>>> {
1782 let target_directory = self.absolutize(&target_directory);
1783 let worktree_path = self.abs_path().clone();
1784 let fs = self.fs.clone();
1785 let paths = paths
1786 .into_iter()
1787 .filter_map(|source| {
1788 let file_name = source.file_name()?;
1789 let mut target = target_directory.clone();
1790 target.push(file_name);
1791
1792 // Do not allow copying the same file to itself.
1793 if source.as_ref() != target.as_path() {
1794 Some((source, target))
1795 } else {
1796 None
1797 }
1798 })
1799 .collect::<Vec<_>>();
1800
1801 let paths_to_refresh = paths
1802 .iter()
1803 .filter_map(|(_, target)| {
1804 RelPath::new(
1805 target.strip_prefix(&worktree_path).ok()?,
1806 PathStyle::local(),
1807 )
1808 .ok()
1809 .map(|path| path.into_arc())
1810 })
1811 .collect::<Vec<_>>();
1812
1813 cx.spawn(async move |this, cx| {
1814 cx.background_spawn(async move {
1815 for (source, target) in paths {
1816 copy_recursive(
1817 fs.as_ref(),
1818 &source,
1819 &target,
1820 fs::CopyOptions {
1821 overwrite: true,
1822 ..Default::default()
1823 },
1824 )
1825 .await
1826 .with_context(|| {
1827 format!("Failed to copy file from {source:?} to {target:?}")
1828 })?;
1829 }
1830 anyhow::Ok(())
1831 })
1832 .await
1833 .log_err();
1834 let mut refresh = cx.read_entity(
1835 &this.upgrade().with_context(|| "Dropped worktree")?,
1836 |this, _| {
1837 anyhow::Ok::<postage::barrier::Receiver>(
1838 this.as_local()
1839 .with_context(|| "Worktree is not local")?
1840 .refresh_entries_for_paths(paths_to_refresh.clone()),
1841 )
1842 },
1843 )?;
1844
1845 cx.background_spawn(async move {
1846 refresh.next().await;
1847 anyhow::Ok(())
1848 })
1849 .await
1850 .log_err();
1851
1852 let this = this.upgrade().with_context(|| "Dropped worktree")?;
1853 Ok(cx.read_entity(&this, |this, _| {
1854 paths_to_refresh
1855 .iter()
1856 .filter_map(|path| Some(this.entry_for_path(path)?.id))
1857 .collect()
1858 }))
1859 })
1860 }
1861
1862 fn expand_entry(
1863 &self,
1864 entry_id: ProjectEntryId,
1865 cx: &Context<Worktree>,
1866 ) -> Option<Task<Result<()>>> {
1867 let path = self.entry_for_id(entry_id)?.path.clone();
1868 let mut refresh = self.refresh_entries_for_paths(vec![path]);
1869 Some(cx.background_spawn(async move {
1870 refresh.next().await;
1871 Ok(())
1872 }))
1873 }
1874
1875 fn expand_all_for_entry(
1876 &self,
1877 entry_id: ProjectEntryId,
1878 cx: &Context<Worktree>,
1879 ) -> Option<Task<Result<()>>> {
1880 let path = self.entry_for_id(entry_id).unwrap().path.clone();
1881 let mut rx = self.add_path_prefix_to_scan(path);
1882 Some(cx.background_spawn(async move {
1883 rx.next().await;
1884 Ok(())
1885 }))
1886 }
1887
1888 pub fn refresh_entries_for_paths(&self, paths: Vec<Arc<RelPath>>) -> barrier::Receiver {
1889 let (tx, rx) = barrier::channel();
1890 self.scan_requests_tx
1891 .try_send(ScanRequest {
1892 relative_paths: paths,
1893 done: smallvec![tx],
1894 })
1895 .ok();
1896 rx
1897 }
1898
1899 #[cfg(feature = "test-support")]
1900 pub fn manually_refresh_entries_for_paths(
1901 &self,
1902 paths: Vec<Arc<RelPath>>,
1903 ) -> barrier::Receiver {
1904 self.refresh_entries_for_paths(paths)
1905 }
1906
1907 pub fn add_path_prefix_to_scan(&self, path_prefix: Arc<RelPath>) -> barrier::Receiver {
1908 let (tx, rx) = barrier::channel();
1909 self.path_prefixes_to_scan_tx
1910 .try_send(PathPrefixScanRequest {
1911 path: path_prefix,
1912 done: smallvec![tx],
1913 })
1914 .ok();
1915 rx
1916 }
1917
1918 pub fn refresh_entry(
1919 &self,
1920 path: Arc<RelPath>,
1921 old_path: Option<Arc<RelPath>>,
1922 cx: &Context<Worktree>,
1923 ) -> Task<Result<Option<Entry>>> {
1924 if self.settings.is_path_excluded(&path) {
1925 return Task::ready(Ok(None));
1926 }
1927 let paths = if let Some(old_path) = old_path.as_ref() {
1928 vec![old_path.clone(), path.clone()]
1929 } else {
1930 vec![path.clone()]
1931 };
1932 let t0 = Instant::now();
1933 let mut refresh = self.refresh_entries_for_paths(paths);
1934 // todo(lw): Hot foreground spawn
1935 cx.spawn(async move |this, cx| {
1936 refresh.recv().await;
1937 log::trace!("refreshed entry {path:?} in {:?}", t0.elapsed());
1938 let new_entry = this.read_with(cx, |this, _| {
1939 this.entry_for_path(&path).cloned().with_context(|| {
1940 format!("Could not find entry in worktree for {path:?} after refresh")
1941 })
1942 })??;
1943 Ok(Some(new_entry))
1944 })
1945 }
1946
1947 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1948 where
1949 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1950 Fut: 'static + Send + Future<Output = bool>,
1951 {
1952 if let Some(observer) = self.update_observer.as_mut() {
1953 *observer.resume_updates.borrow_mut() = ();
1954 return;
1955 }
1956
1957 let (resume_updates_tx, mut resume_updates_rx) = watch::channel::<()>();
1958 let (snapshots_tx, mut snapshots_rx) =
1959 mpsc::unbounded::<(LocalSnapshot, UpdatedEntriesSet)>();
1960 snapshots_tx
1961 .unbounded_send((self.snapshot(), Arc::default()))
1962 .ok();
1963
1964 let worktree_id = self.id.to_proto();
1965 let _maintain_remote_snapshot = cx.background_spawn(async move {
1966 let mut is_first = true;
1967 while let Some((snapshot, entry_changes)) = snapshots_rx.next().await {
1968 let update = if is_first {
1969 is_first = false;
1970 snapshot.build_initial_update(project_id, worktree_id)
1971 } else {
1972 snapshot.build_update(project_id, worktree_id, entry_changes)
1973 };
1974
1975 for update in proto::split_worktree_update(update) {
1976 let _ = resume_updates_rx.try_recv();
1977 loop {
1978 let result = callback(update.clone());
1979 if result.await {
1980 break;
1981 } else {
1982 log::info!("waiting to resume updates");
1983 if resume_updates_rx.next().await.is_none() {
1984 return Some(());
1985 }
1986 }
1987 }
1988 }
1989 }
1990 Some(())
1991 });
1992
1993 self.update_observer = Some(UpdateObservationState {
1994 snapshots_tx,
1995 resume_updates: resume_updates_tx,
1996 _maintain_remote_snapshot,
1997 });
1998 }
1999
2000 pub fn share_private_files(&mut self, cx: &Context<Worktree>) {
2001 self.share_private_files = true;
2002 self.restart_background_scanners(cx);
2003 }
2004
2005 pub fn update_abs_path_and_refresh(
2006 &mut self,
2007 new_path: Arc<SanitizedPath>,
2008 cx: &Context<Worktree>,
2009 ) {
2010 self.snapshot.git_repositories = Default::default();
2011 self.snapshot.ignores_by_parent_abs_path = Default::default();
2012 let root_name = new_path
2013 .as_path()
2014 .file_name()
2015 .and_then(|f| f.to_str())
2016 .map_or(RelPath::empty().into(), |f| {
2017 RelPath::unix(f).unwrap().into()
2018 });
2019 self.snapshot.update_abs_path(new_path, root_name);
2020 self.restart_background_scanners(cx);
2021 }
2022 #[cfg(feature = "test-support")]
2023 pub fn repositories(&self) -> Vec<Arc<Path>> {
2024 self.git_repositories
2025 .values()
2026 .map(|entry| entry.work_directory_abs_path.clone())
2027 .collect::<Vec<_>>()
2028 }
2029}
2030
2031impl RemoteWorktree {
2032 pub fn project_id(&self) -> u64 {
2033 self.project_id
2034 }
2035
2036 pub fn client(&self) -> AnyProtoClient {
2037 self.client.clone()
2038 }
2039
2040 pub fn disconnected_from_host(&mut self) {
2041 self.updates_tx.take();
2042 self.snapshot_subscriptions.clear();
2043 self.disconnected = true;
2044 }
2045
2046 pub fn update_from_remote(&self, update: proto::UpdateWorktree) {
2047 if let Some(updates_tx) = &self.updates_tx {
2048 updates_tx
2049 .unbounded_send(update)
2050 .expect("consumer runs to completion");
2051 }
2052 }
2053
2054 fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
2055 where
2056 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
2057 Fut: 'static + Send + Future<Output = bool>,
2058 {
2059 let (tx, mut rx) = mpsc::unbounded();
2060 let initial_update = self
2061 .snapshot
2062 .build_initial_update(project_id, self.id().to_proto());
2063 self.update_observer = Some(tx);
2064 cx.spawn(async move |this, cx| {
2065 let mut update = initial_update;
2066 'outer: loop {
2067 // SSH projects use a special project ID of 0, and we need to
2068 // remap it to the correct one here.
2069 update.project_id = project_id;
2070
2071 for chunk in split_worktree_update(update) {
2072 if !callback(chunk).await {
2073 break 'outer;
2074 }
2075 }
2076
2077 if let Some(next_update) = rx.next().await {
2078 update = next_update;
2079 } else {
2080 break;
2081 }
2082 }
2083 this.update(cx, |this, _| {
2084 let this = this.as_remote_mut().unwrap();
2085 this.update_observer.take();
2086 })
2087 })
2088 .detach();
2089 }
2090
2091 fn observed_snapshot(&self, scan_id: usize) -> bool {
2092 self.completed_scan_id >= scan_id
2093 }
2094
2095 pub fn wait_for_snapshot(
2096 &mut self,
2097 scan_id: usize,
2098 ) -> impl Future<Output = Result<()>> + use<> {
2099 let (tx, rx) = oneshot::channel();
2100 if self.observed_snapshot(scan_id) {
2101 let _ = tx.send(());
2102 } else if self.disconnected {
2103 drop(tx);
2104 } else {
2105 match self
2106 .snapshot_subscriptions
2107 .binary_search_by_key(&scan_id, |probe| probe.0)
2108 {
2109 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
2110 }
2111 }
2112
2113 async move {
2114 rx.await?;
2115 Ok(())
2116 }
2117 }
2118
2119 pub fn insert_entry(
2120 &mut self,
2121 entry: proto::Entry,
2122 scan_id: usize,
2123 cx: &Context<Worktree>,
2124 ) -> Task<Result<Entry>> {
2125 let wait_for_snapshot = self.wait_for_snapshot(scan_id);
2126 cx.spawn(async move |this, cx| {
2127 wait_for_snapshot.await?;
2128 this.update(cx, |worktree, _| {
2129 let worktree = worktree.as_remote_mut().unwrap();
2130 let snapshot = &mut worktree.background_snapshot.lock().0;
2131 let entry = snapshot.insert_entry(entry, &worktree.file_scan_inclusions);
2132 worktree.snapshot = snapshot.clone();
2133 entry
2134 })?
2135 })
2136 }
2137
2138 fn delete_entry(
2139 &self,
2140 entry_id: ProjectEntryId,
2141 trash: bool,
2142 cx: &Context<Worktree>,
2143 ) -> Option<Task<Result<Option<TrashedEntry>>>> {
2144 let response = self.client.request(proto::DeleteProjectEntry {
2145 project_id: self.project_id,
2146 entry_id: entry_id.to_proto(),
2147 use_trash: trash,
2148 });
2149 Some(cx.spawn(async move |this, cx| {
2150 let response = response.await?;
2151 let scan_id = response.worktree_scan_id as usize;
2152
2153 this.update(cx, move |this, _| {
2154 this.as_remote_mut().unwrap().wait_for_snapshot(scan_id)
2155 })?
2156 .await?;
2157
2158 this.update(cx, |this, _| {
2159 let this = this.as_remote_mut().unwrap();
2160 let snapshot = &mut this.background_snapshot.lock().0;
2161 snapshot.delete_entry(entry_id);
2162 this.snapshot = snapshot.clone();
2163
2164 // TODO: How can we actually track the deleted entry when
2165 // working in remote? We likely only need to keep this
2166 // information on the remote side in order to support restoring
2167 // the trashed file.
2168 None
2169 })
2170 }))
2171 }
2172
2173 // fn rename_entry(
2174 // &self,
2175 // entry_id: ProjectEntryId,
2176 // new_path: impl Into<Arc<RelPath>>,
2177 // cx: &Context<Worktree>,
2178 // ) -> Task<Result<CreatedEntry>> {
2179 // let new_path: Arc<RelPath> = new_path.into();
2180 // let response = self.client.request(proto::RenameProjectEntry {
2181 // project_id: self.project_id,
2182 // entry_id: entry_id.to_proto(),
2183 // new_worktree_id: new_path.worktree_id,
2184 // new_path: new_path.as_ref().to_proto(),
2185 // });
2186 // cx.spawn(async move |this, cx| {
2187 // let response = response.await?;
2188 // match response.entry {
2189 // Some(entry) => this
2190 // .update(cx, |this, cx| {
2191 // this.as_remote_mut().unwrap().insert_entry(
2192 // entry,
2193 // response.worktree_scan_id as usize,
2194 // cx,
2195 // )
2196 // })?
2197 // .await
2198 // .map(CreatedEntry::Included),
2199 // None => {
2200 // let abs_path =
2201 // this.read_with(cx, |worktree, _| worktree.absolutize(&new_path))?;
2202 // Ok(CreatedEntry::Excluded { abs_path })
2203 // }
2204 // }
2205 // })
2206 // }
2207
2208 fn copy_external_entries(
2209 &self,
2210 target_directory: Arc<RelPath>,
2211 paths_to_copy: Vec<Arc<Path>>,
2212 local_fs: Arc<dyn Fs>,
2213 cx: &Context<Worktree>,
2214 ) -> Task<anyhow::Result<Vec<ProjectEntryId>>> {
2215 let client = self.client.clone();
2216 let worktree_id = self.id().to_proto();
2217 let project_id = self.project_id;
2218
2219 cx.background_spawn(async move {
2220 let mut requests = Vec::new();
2221 for root_path_to_copy in paths_to_copy {
2222 let Some(filename) = root_path_to_copy
2223 .file_name()
2224 .and_then(|name| name.to_str())
2225 .and_then(|filename| RelPath::unix(filename).ok())
2226 else {
2227 continue;
2228 };
2229 for (abs_path, is_directory) in
2230 read_dir_items(local_fs.as_ref(), &root_path_to_copy).await?
2231 {
2232 let Some(relative_path) = abs_path
2233 .strip_prefix(&root_path_to_copy)
2234 .map_err(|e| anyhow::Error::from(e))
2235 .and_then(|relative_path| RelPath::new(relative_path, PathStyle::local()))
2236 .log_err()
2237 else {
2238 continue;
2239 };
2240 let content = if is_directory {
2241 None
2242 } else {
2243 Some(local_fs.load_bytes(&abs_path).await?)
2244 };
2245
2246 let mut target_path = target_directory.join(filename);
2247 if relative_path.file_name().is_some() {
2248 target_path = target_path.join(&relative_path);
2249 }
2250
2251 requests.push(proto::CreateProjectEntry {
2252 project_id,
2253 worktree_id,
2254 path: target_path.to_proto(),
2255 is_directory,
2256 content,
2257 });
2258 }
2259 }
2260 requests.sort_unstable_by(|a, b| a.path.cmp(&b.path));
2261 requests.dedup();
2262
2263 let mut copied_entry_ids = Vec::new();
2264 for request in requests {
2265 let response = client.request(request).await?;
2266 copied_entry_ids.extend(response.entry.map(|e| ProjectEntryId::from_proto(e.id)));
2267 }
2268
2269 Ok(copied_entry_ids)
2270 })
2271 }
2272}
2273
2274impl Snapshot {
2275 pub fn new(
2276 id: WorktreeId,
2277 root_name: Arc<RelPath>,
2278 abs_path: Arc<Path>,
2279 path_style: PathStyle,
2280 ) -> Self {
2281 Snapshot {
2282 id,
2283 abs_path: SanitizedPath::from_arc(abs_path),
2284 path_style,
2285 root_char_bag: root_name
2286 .as_unix_str()
2287 .chars()
2288 .map(|c| c.to_ascii_lowercase())
2289 .collect(),
2290 root_name,
2291 always_included_entries: Default::default(),
2292 entries_by_path: Default::default(),
2293 entries_by_id: Default::default(),
2294 root_repo_common_dir: None,
2295 scan_id: 1,
2296 completed_scan_id: 0,
2297 }
2298 }
2299
2300 pub fn id(&self) -> WorktreeId {
2301 self.id
2302 }
2303
2304 // TODO:
2305 // Consider the following:
2306 //
2307 // ```rust
2308 // let abs_path: Arc<Path> = snapshot.abs_path(); // e.g. "C:\Users\user\Desktop\project"
2309 // let some_non_trimmed_path = Path::new("\\\\?\\C:\\Users\\user\\Desktop\\project\\main.rs");
2310 // // The caller perform some actions here:
2311 // some_non_trimmed_path.strip_prefix(abs_path); // This fails
2312 // some_non_trimmed_path.starts_with(abs_path); // This fails too
2313 // ```
2314 //
2315 // This is definitely a bug, but it's not clear if we should handle it here or not.
2316 pub fn abs_path(&self) -> &Arc<Path> {
2317 SanitizedPath::cast_arc_ref(&self.abs_path)
2318 }
2319
2320 pub fn root_repo_common_dir(&self) -> Option<&Arc<Path>> {
2321 self.root_repo_common_dir
2322 .as_ref()
2323 .map(SanitizedPath::cast_arc_ref)
2324 }
2325
2326 fn build_initial_update(&self, project_id: u64, worktree_id: u64) -> proto::UpdateWorktree {
2327 let mut updated_entries = self
2328 .entries_by_path
2329 .iter()
2330 .map(proto::Entry::from)
2331 .collect::<Vec<_>>();
2332 updated_entries.sort_unstable_by_key(|e| e.id);
2333
2334 proto::UpdateWorktree {
2335 project_id,
2336 worktree_id,
2337 abs_path: self.abs_path().to_string_lossy().into_owned(),
2338 root_name: self.root_name().to_proto(),
2339 root_repo_common_dir: self
2340 .root_repo_common_dir()
2341 .map(|p| p.to_string_lossy().into_owned()),
2342 updated_entries,
2343 removed_entries: Vec::new(),
2344 scan_id: self.scan_id as u64,
2345 is_last_update: self.completed_scan_id == self.scan_id,
2346 // Sent in separate messages.
2347 updated_repositories: Vec::new(),
2348 removed_repositories: Vec::new(),
2349 }
2350 }
2351
2352 pub fn work_directory_abs_path(&self, work_directory: &WorkDirectory) -> PathBuf {
2353 match work_directory {
2354 WorkDirectory::InProject { relative_path } => self.absolutize(relative_path),
2355 WorkDirectory::AboveProject { absolute_path, .. } => absolute_path.as_ref().to_owned(),
2356 }
2357 }
2358
2359 pub fn absolutize(&self, path: &RelPath) -> PathBuf {
2360 if path.file_name().is_some() {
2361 let mut abs_path = self.abs_path.to_string();
2362 for component in path.components() {
2363 if !abs_path.ends_with(self.path_style.primary_separator()) {
2364 abs_path.push_str(self.path_style.primary_separator());
2365 }
2366 abs_path.push_str(component);
2367 }
2368 PathBuf::from(abs_path)
2369 } else {
2370 self.abs_path.as_path().to_path_buf()
2371 }
2372 }
2373
2374 pub fn contains_entry(&self, entry_id: ProjectEntryId) -> bool {
2375 self.entries_by_id.get(&entry_id, ()).is_some()
2376 }
2377
2378 fn insert_entry(
2379 &mut self,
2380 entry: proto::Entry,
2381 always_included_paths: &PathMatcher,
2382 ) -> Result<Entry> {
2383 let entry = Entry::try_from((&self.root_char_bag, always_included_paths, entry))?;
2384 let old_entry = self.entries_by_id.insert_or_replace(
2385 PathEntry {
2386 id: entry.id,
2387 path: entry.path.clone(),
2388 is_ignored: entry.is_ignored,
2389 scan_id: 0,
2390 },
2391 (),
2392 );
2393 if let Some(old_entry) = old_entry {
2394 self.entries_by_path.remove(&PathKey(old_entry.path), ());
2395 }
2396 self.entries_by_path.insert_or_replace(entry.clone(), ());
2397 Ok(entry)
2398 }
2399
2400 fn delete_entry(&mut self, entry_id: ProjectEntryId) -> Option<Arc<RelPath>> {
2401 let removed_entry = self.entries_by_id.remove(&entry_id, ())?;
2402 self.entries_by_path = {
2403 let mut cursor = self.entries_by_path.cursor::<TraversalProgress>(());
2404 let mut new_entries_by_path =
2405 cursor.slice(&TraversalTarget::path(&removed_entry.path), Bias::Left);
2406 while let Some(entry) = cursor.item() {
2407 if entry.path.starts_with(&removed_entry.path) {
2408 self.entries_by_id.remove(&entry.id, ());
2409 cursor.next();
2410 } else {
2411 break;
2412 }
2413 }
2414 new_entries_by_path.append(cursor.suffix(), ());
2415 new_entries_by_path
2416 };
2417
2418 Some(removed_entry.path)
2419 }
2420
2421 fn update_abs_path(&mut self, abs_path: Arc<SanitizedPath>, root_name: Arc<RelPath>) {
2422 self.abs_path = abs_path;
2423 if root_name != self.root_name {
2424 self.root_char_bag = root_name
2425 .as_unix_str()
2426 .chars()
2427 .map(|c| c.to_ascii_lowercase())
2428 .collect();
2429 self.root_name = root_name;
2430 }
2431 }
2432
2433 pub fn apply_remote_update(
2434 &mut self,
2435 update: proto::UpdateWorktree,
2436 always_included_paths: &PathMatcher,
2437 ) {
2438 log::debug!(
2439 "applying remote worktree update. {} entries updated, {} removed",
2440 update.updated_entries.len(),
2441 update.removed_entries.len()
2442 );
2443 if let Some(root_name) = RelPath::from_proto(&update.root_name).log_err() {
2444 self.update_abs_path(
2445 SanitizedPath::new_arc(&Path::new(&update.abs_path)),
2446 root_name,
2447 );
2448 }
2449
2450 let mut entries_by_path_edits = Vec::new();
2451 let mut entries_by_id_edits = Vec::new();
2452
2453 for entry_id in update.removed_entries {
2454 let entry_id = ProjectEntryId::from_proto(entry_id);
2455 entries_by_id_edits.push(Edit::Remove(entry_id));
2456 if let Some(entry) = self.entry_for_id(entry_id) {
2457 entries_by_path_edits.push(Edit::Remove(PathKey(entry.path.clone())));
2458 }
2459 }
2460
2461 for entry in update.updated_entries {
2462 let Some(entry) =
2463 Entry::try_from((&self.root_char_bag, always_included_paths, entry)).log_err()
2464 else {
2465 continue;
2466 };
2467 if let Some(PathEntry { path, .. }) = self.entries_by_id.get(&entry.id, ()) {
2468 entries_by_path_edits.push(Edit::Remove(PathKey(path.clone())));
2469 }
2470 if let Some(old_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2471 && old_entry.id != entry.id
2472 {
2473 entries_by_id_edits.push(Edit::Remove(old_entry.id));
2474 }
2475 entries_by_id_edits.push(Edit::Insert(PathEntry {
2476 id: entry.id,
2477 path: entry.path.clone(),
2478 is_ignored: entry.is_ignored,
2479 scan_id: 0,
2480 }));
2481 entries_by_path_edits.push(Edit::Insert(entry));
2482 }
2483
2484 self.entries_by_path.edit(entries_by_path_edits, ());
2485 self.entries_by_id.edit(entries_by_id_edits, ());
2486
2487 self.root_repo_common_dir = update
2488 .root_repo_common_dir
2489 .map(|p| SanitizedPath::new_arc(Path::new(&p)));
2490
2491 self.scan_id = update.scan_id as usize;
2492 if update.is_last_update {
2493 self.completed_scan_id = update.scan_id as usize;
2494 }
2495 }
2496
2497 pub fn entry_count(&self) -> usize {
2498 self.entries_by_path.summary().count
2499 }
2500
2501 pub fn visible_entry_count(&self) -> usize {
2502 self.entries_by_path.summary().non_ignored_count
2503 }
2504
2505 pub fn dir_count(&self) -> usize {
2506 let summary = self.entries_by_path.summary();
2507 summary.count - summary.file_count
2508 }
2509
2510 pub fn visible_dir_count(&self) -> usize {
2511 let summary = self.entries_by_path.summary();
2512 summary.non_ignored_count - summary.non_ignored_file_count
2513 }
2514
2515 pub fn file_count(&self) -> usize {
2516 self.entries_by_path.summary().file_count
2517 }
2518
2519 pub fn visible_file_count(&self) -> usize {
2520 self.entries_by_path.summary().non_ignored_file_count
2521 }
2522
2523 fn traverse_from_offset(
2524 &self,
2525 include_files: bool,
2526 include_dirs: bool,
2527 include_ignored: bool,
2528 start_offset: usize,
2529 ) -> Traversal<'_> {
2530 let mut cursor = self.entries_by_path.cursor(());
2531 cursor.seek(
2532 &TraversalTarget::Count {
2533 count: start_offset,
2534 include_files,
2535 include_dirs,
2536 include_ignored,
2537 },
2538 Bias::Right,
2539 );
2540 Traversal {
2541 snapshot: self,
2542 cursor,
2543 include_files,
2544 include_dirs,
2545 include_ignored,
2546 }
2547 }
2548
2549 pub fn traverse_from_path(
2550 &self,
2551 include_files: bool,
2552 include_dirs: bool,
2553 include_ignored: bool,
2554 path: &RelPath,
2555 ) -> Traversal<'_> {
2556 Traversal::new(self, include_files, include_dirs, include_ignored, path)
2557 }
2558
2559 pub fn files(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2560 self.traverse_from_offset(true, false, include_ignored, start)
2561 }
2562
2563 pub fn directories(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2564 self.traverse_from_offset(false, true, include_ignored, start)
2565 }
2566
2567 pub fn entries(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2568 self.traverse_from_offset(true, true, include_ignored, start)
2569 }
2570
2571 pub fn paths(&self) -> impl Iterator<Item = &RelPath> {
2572 self.entries_by_path
2573 .cursor::<()>(())
2574 .filter(move |entry| !entry.path.is_empty())
2575 .map(|entry| entry.path.as_ref())
2576 }
2577
2578 pub fn child_entries<'a>(&'a self, parent_path: &'a RelPath) -> ChildEntriesIter<'a> {
2579 let options = ChildEntriesOptions {
2580 include_files: true,
2581 include_dirs: true,
2582 include_ignored: true,
2583 };
2584 self.child_entries_with_options(parent_path, options)
2585 }
2586
2587 pub fn child_entries_with_options<'a>(
2588 &'a self,
2589 parent_path: &'a RelPath,
2590 options: ChildEntriesOptions,
2591 ) -> ChildEntriesIter<'a> {
2592 let mut cursor = self.entries_by_path.cursor(());
2593 cursor.seek(&TraversalTarget::path(parent_path), Bias::Right);
2594 let traversal = Traversal {
2595 snapshot: self,
2596 cursor,
2597 include_files: options.include_files,
2598 include_dirs: options.include_dirs,
2599 include_ignored: options.include_ignored,
2600 };
2601 ChildEntriesIter {
2602 traversal,
2603 parent_path,
2604 }
2605 }
2606
2607 pub fn root_entry(&self) -> Option<&Entry> {
2608 self.entries_by_path.first()
2609 }
2610
2611 /// Returns `None` for a single file worktree, or `Some(self.abs_path())` if
2612 /// it is a directory.
2613 pub fn root_dir(&self) -> Option<Arc<Path>> {
2614 self.root_entry()
2615 .filter(|entry| entry.is_dir())
2616 .map(|_| self.abs_path().clone())
2617 }
2618
2619 pub fn root_name(&self) -> &RelPath {
2620 &self.root_name
2621 }
2622
2623 pub fn root_name_str(&self) -> &str {
2624 self.root_name.as_unix_str()
2625 }
2626
2627 pub fn scan_id(&self) -> usize {
2628 self.scan_id
2629 }
2630
2631 pub fn entry_for_path(&self, path: &RelPath) -> Option<&Entry> {
2632 let entry = self.traverse_from_path(true, true, true, path).entry();
2633 entry.and_then(|entry| {
2634 if entry.path.as_ref() == path {
2635 Some(entry)
2636 } else {
2637 None
2638 }
2639 })
2640 }
2641
2642 /// Resolves a path to an executable using the following heuristics:
2643 ///
2644 /// 1. If the path starts with `~`, it is expanded to the user's home directory.
2645 /// 2. If the path is relative and contains more than one component,
2646 /// it is joined to the worktree root path.
2647 /// 3. If the path is relative and exists in the worktree
2648 /// (even if falls under an exclusion filter),
2649 /// it is joined to the worktree root path.
2650 /// 4. Otherwise the path is returned unmodified.
2651 ///
2652 /// Relative paths that do not exist in the worktree may
2653 /// still be found using the `PATH` environment variable.
2654 pub fn resolve_relative_path(&self, path: PathBuf) -> PathBuf {
2655 if let Some(path_str) = path.to_str() {
2656 if let Some(remaining_path) = path_str.strip_prefix("~/") {
2657 return home_dir().join(remaining_path);
2658 } else if path_str == "~" {
2659 return home_dir().to_path_buf();
2660 }
2661 }
2662
2663 if let Ok(rel_path) = RelPath::new(&path, self.path_style)
2664 && (path.components().count() > 1 || self.entry_for_path(&rel_path).is_some())
2665 {
2666 self.abs_path().join(path)
2667 } else {
2668 path
2669 }
2670 }
2671
2672 pub fn entry_for_id(&self, id: ProjectEntryId) -> Option<&Entry> {
2673 let entry = self.entries_by_id.get(&id, ())?;
2674 self.entry_for_path(&entry.path)
2675 }
2676
2677 pub fn path_style(&self) -> PathStyle {
2678 self.path_style
2679 }
2680}
2681
2682impl LocalSnapshot {
2683 fn local_repo_for_work_directory_path(&self, path: &RelPath) -> Option<&LocalRepositoryEntry> {
2684 self.git_repositories
2685 .iter()
2686 .map(|(_, entry)| entry)
2687 .find(|entry| entry.work_directory.path_key() == PathKey(path.into()))
2688 }
2689
2690 fn build_update(
2691 &self,
2692 project_id: u64,
2693 worktree_id: u64,
2694 entry_changes: UpdatedEntriesSet,
2695 ) -> proto::UpdateWorktree {
2696 let mut updated_entries = Vec::new();
2697 let mut removed_entries = Vec::new();
2698
2699 for (_, entry_id, path_change) in entry_changes.iter() {
2700 if let PathChange::Removed = path_change {
2701 removed_entries.push(entry_id.0 as u64);
2702 } else if let Some(entry) = self.entry_for_id(*entry_id) {
2703 updated_entries.push(proto::Entry::from(entry));
2704 }
2705 }
2706
2707 removed_entries.sort_unstable();
2708 updated_entries.sort_unstable_by_key(|e| e.id);
2709
2710 // TODO - optimize, knowing that removed_entries are sorted.
2711 removed_entries.retain(|id| updated_entries.binary_search_by_key(id, |e| e.id).is_err());
2712
2713 proto::UpdateWorktree {
2714 project_id,
2715 worktree_id,
2716 abs_path: self.abs_path().to_string_lossy().into_owned(),
2717 root_name: self.root_name().to_proto(),
2718 root_repo_common_dir: self
2719 .root_repo_common_dir()
2720 .map(|p| p.to_string_lossy().into_owned()),
2721 updated_entries,
2722 removed_entries,
2723 scan_id: self.scan_id as u64,
2724 is_last_update: self.completed_scan_id == self.scan_id,
2725 // Sent in separate messages.
2726 updated_repositories: Vec::new(),
2727 removed_repositories: Vec::new(),
2728 }
2729 }
2730
2731 async fn insert_entry(&mut self, mut entry: Entry, fs: &dyn Fs) -> Entry {
2732 log::trace!("insert entry {:?}", entry.path);
2733 if entry.is_file() && entry.path.file_name() == Some(&GITIGNORE) {
2734 let abs_path = self.absolutize(&entry.path);
2735 match build_gitignore(&abs_path, fs).await {
2736 Ok(ignore) => {
2737 self.ignores_by_parent_abs_path
2738 .insert(abs_path.parent().unwrap().into(), (Arc::new(ignore), true));
2739 }
2740 Err(error) => {
2741 log::error!(
2742 "error loading .gitignore file {:?} - {:?}",
2743 &entry.path,
2744 error
2745 );
2746 }
2747 }
2748 }
2749
2750 if entry.kind == EntryKind::PendingDir
2751 && let Some(existing_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2752 {
2753 entry.kind = existing_entry.kind;
2754 }
2755
2756 let scan_id = self.scan_id;
2757 let removed = self.entries_by_path.insert_or_replace(entry.clone(), ());
2758 if let Some(removed) = removed
2759 && removed.id != entry.id
2760 {
2761 self.entries_by_id.remove(&removed.id, ());
2762 }
2763 self.entries_by_id.insert_or_replace(
2764 PathEntry {
2765 id: entry.id,
2766 path: entry.path.clone(),
2767 is_ignored: entry.is_ignored,
2768 scan_id,
2769 },
2770 (),
2771 );
2772
2773 entry
2774 }
2775
2776 fn ancestor_inodes_for_path(&self, path: &RelPath) -> TreeSet<u64> {
2777 let mut inodes = TreeSet::default();
2778 for ancestor in path.ancestors().skip(1) {
2779 if let Some(entry) = self.entry_for_path(ancestor) {
2780 inodes.insert(entry.inode);
2781 }
2782 }
2783 inodes
2784 }
2785
2786 async fn ignore_stack_for_abs_path(
2787 &self,
2788 abs_path: &Path,
2789 is_dir: bool,
2790 fs: &dyn Fs,
2791 ) -> IgnoreStack {
2792 let mut new_ignores = Vec::new();
2793 let mut repo_root = None;
2794 for (index, ancestor) in abs_path.ancestors().enumerate() {
2795 if index > 0 {
2796 if let Some((ignore, _)) = self.ignores_by_parent_abs_path.get(ancestor) {
2797 new_ignores.push((ancestor, Some(ignore.clone())));
2798 } else {
2799 new_ignores.push((ancestor, None));
2800 }
2801 }
2802
2803 let metadata = fs.metadata(&ancestor.join(DOT_GIT)).await.ok().flatten();
2804 if metadata.is_some() {
2805 repo_root = Some(Arc::from(ancestor));
2806 break;
2807 }
2808 }
2809
2810 let mut ignore_stack = if let Some(global_gitignore) = self.global_gitignore.clone() {
2811 IgnoreStack::global(global_gitignore)
2812 } else {
2813 IgnoreStack::none()
2814 };
2815
2816 if let Some((repo_exclude, _)) = repo_root
2817 .as_ref()
2818 .and_then(|abs_path| self.repo_exclude_by_work_dir_abs_path.get(abs_path))
2819 {
2820 ignore_stack = ignore_stack.append(IgnoreKind::RepoExclude, repo_exclude.clone());
2821 }
2822 ignore_stack.repo_root = repo_root;
2823 for (parent_abs_path, ignore) in new_ignores.into_iter().rev() {
2824 if ignore_stack.is_abs_path_ignored(parent_abs_path, true) {
2825 ignore_stack = IgnoreStack::all();
2826 break;
2827 } else if let Some(ignore) = ignore {
2828 ignore_stack =
2829 ignore_stack.append(IgnoreKind::Gitignore(parent_abs_path.into()), ignore);
2830 }
2831 }
2832
2833 if ignore_stack.is_abs_path_ignored(abs_path, is_dir) {
2834 ignore_stack = IgnoreStack::all();
2835 }
2836
2837 ignore_stack
2838 }
2839
2840 #[cfg(feature = "test-support")]
2841 pub fn expanded_entries(&self) -> impl Iterator<Item = &Entry> {
2842 self.entries_by_path
2843 .cursor::<()>(())
2844 .filter(|entry| entry.kind == EntryKind::Dir && (entry.is_external || entry.is_ignored))
2845 }
2846
2847 #[cfg(feature = "test-support")]
2848 pub fn check_invariants(&self, git_state: bool) {
2849 use pretty_assertions::assert_eq;
2850
2851 assert_eq!(
2852 self.entries_by_path
2853 .cursor::<()>(())
2854 .map(|e| (&e.path, e.id))
2855 .collect::<Vec<_>>(),
2856 self.entries_by_id
2857 .cursor::<()>(())
2858 .map(|e| (&e.path, e.id))
2859 .collect::<collections::BTreeSet<_>>()
2860 .into_iter()
2861 .collect::<Vec<_>>(),
2862 "entries_by_path and entries_by_id are inconsistent"
2863 );
2864
2865 let mut files = self.files(true, 0);
2866 let mut visible_files = self.files(false, 0);
2867 for entry in self.entries_by_path.cursor::<()>(()) {
2868 if entry.is_file() {
2869 assert_eq!(files.next().unwrap().inode, entry.inode);
2870 if !entry.is_ignored || entry.is_always_included {
2871 assert_eq!(visible_files.next().unwrap().inode, entry.inode);
2872 }
2873 }
2874 }
2875
2876 assert!(files.next().is_none());
2877 assert!(visible_files.next().is_none());
2878
2879 let mut bfs_paths = Vec::new();
2880 let mut stack = self
2881 .root_entry()
2882 .map(|e| e.path.as_ref())
2883 .into_iter()
2884 .collect::<Vec<_>>();
2885 while let Some(path) = stack.pop() {
2886 bfs_paths.push(path);
2887 let ix = stack.len();
2888 for child_entry in self.child_entries(path) {
2889 stack.insert(ix, &child_entry.path);
2890 }
2891 }
2892
2893 let dfs_paths_via_iter = self
2894 .entries_by_path
2895 .cursor::<()>(())
2896 .map(|e| e.path.as_ref())
2897 .collect::<Vec<_>>();
2898 assert_eq!(bfs_paths, dfs_paths_via_iter);
2899
2900 let dfs_paths_via_traversal = self
2901 .entries(true, 0)
2902 .map(|e| e.path.as_ref())
2903 .collect::<Vec<_>>();
2904
2905 assert_eq!(dfs_paths_via_traversal, dfs_paths_via_iter);
2906
2907 if git_state {
2908 for ignore_parent_abs_path in self.ignores_by_parent_abs_path.keys() {
2909 let ignore_parent_path = &RelPath::new(
2910 ignore_parent_abs_path
2911 .strip_prefix(self.abs_path.as_path())
2912 .unwrap(),
2913 PathStyle::local(),
2914 )
2915 .unwrap();
2916 assert!(self.entry_for_path(ignore_parent_path).is_some());
2917 assert!(
2918 self.entry_for_path(
2919 &ignore_parent_path.join(RelPath::unix(GITIGNORE).unwrap())
2920 )
2921 .is_some()
2922 );
2923 }
2924 }
2925 }
2926
2927 #[cfg(feature = "test-support")]
2928 pub fn entries_without_ids(&self, include_ignored: bool) -> Vec<(&RelPath, u64, bool)> {
2929 let mut paths = Vec::new();
2930 for entry in self.entries_by_path.cursor::<()>(()) {
2931 if include_ignored || !entry.is_ignored {
2932 paths.push((entry.path.as_ref(), entry.inode, entry.is_ignored));
2933 }
2934 }
2935 paths.sort_by(|a, b| a.0.cmp(b.0));
2936 paths
2937 }
2938}
2939
2940impl BackgroundScannerState {
2941 fn should_scan_directory(&self, entry: &Entry) -> bool {
2942 (self.scanning_enabled && !entry.is_external && (!entry.is_ignored || entry.is_always_included))
2943 || entry.path.file_name() == Some(DOT_GIT)
2944 || entry.path.file_name() == Some(local_settings_folder_name())
2945 || entry.path.file_name() == Some(local_vscode_folder_name())
2946 || self.scanned_dirs.contains(&entry.id) // If we've ever scanned it, keep scanning
2947 || self
2948 .paths_to_scan
2949 .iter()
2950 .any(|p| p.starts_with(&entry.path))
2951 || self
2952 .path_prefixes_to_scan
2953 .iter()
2954 .any(|p| entry.path.starts_with(p))
2955 }
2956
2957 async fn enqueue_scan_dir(
2958 &self,
2959 abs_path: Arc<Path>,
2960 entry: &Entry,
2961 scan_job_tx: &Sender<ScanJob>,
2962 fs: &dyn Fs,
2963 ) {
2964 let path = entry.path.clone();
2965 let ignore_stack = self
2966 .snapshot
2967 .ignore_stack_for_abs_path(&abs_path, true, fs)
2968 .await;
2969 let mut ancestor_inodes = self.snapshot.ancestor_inodes_for_path(&path);
2970
2971 if !ancestor_inodes.contains(&entry.inode) {
2972 ancestor_inodes.insert(entry.inode);
2973 scan_job_tx
2974 .try_send(ScanJob {
2975 abs_path,
2976 path,
2977 ignore_stack,
2978 scan_queue: scan_job_tx.clone(),
2979 ancestor_inodes,
2980 is_external: entry.is_external,
2981 })
2982 .unwrap();
2983 }
2984 }
2985
2986 fn reuse_entry_id(&mut self, entry: &mut Entry) {
2987 if let Some(mtime) = entry.mtime {
2988 // If an entry with the same inode was removed from the worktree during this scan,
2989 // then it *might* represent the same file or directory. But the OS might also have
2990 // re-used the inode for a completely different file or directory.
2991 //
2992 // Conditionally reuse the old entry's id:
2993 // * if the mtime is the same, the file was probably been renamed.
2994 // * if the path is the same, the file may just have been updated
2995 if let Some(removed_entry) = self.removed_entries.remove(&entry.inode) {
2996 if removed_entry.mtime == Some(mtime) || removed_entry.path == entry.path {
2997 entry.id = removed_entry.id;
2998 }
2999 } else if let Some(existing_entry) = self.snapshot.entry_for_path(&entry.path) {
3000 entry.id = existing_entry.id;
3001 }
3002 }
3003 }
3004
3005 fn entry_id_for(
3006 &mut self,
3007 next_entry_id: &AtomicUsize,
3008 path: &RelPath,
3009 metadata: &fs::Metadata,
3010 ) -> ProjectEntryId {
3011 // If an entry with the same inode was removed from the worktree during this scan,
3012 // then it *might* represent the same file or directory. But the OS might also have
3013 // re-used the inode for a completely different file or directory.
3014 //
3015 // Conditionally reuse the old entry's id:
3016 // * if the mtime is the same, the file was probably been renamed.
3017 // * if the path is the same, the file may just have been updated
3018 if let Some(removed_entry) = self.removed_entries.remove(&metadata.inode) {
3019 if removed_entry.mtime == Some(metadata.mtime) || *removed_entry.path == *path {
3020 return removed_entry.id;
3021 }
3022 } else if let Some(existing_entry) = self.snapshot.entry_for_path(path) {
3023 return existing_entry.id;
3024 }
3025 ProjectEntryId::new(next_entry_id)
3026 }
3027
3028 async fn insert_entry(&mut self, entry: Entry, fs: &dyn Fs, watcher: &dyn Watcher) -> Entry {
3029 let entry = self.snapshot.insert_entry(entry, fs).await;
3030 if entry.path.file_name() == Some(&DOT_GIT) {
3031 self.insert_git_repository(entry.path.clone(), fs, watcher)
3032 .await;
3033 }
3034
3035 #[cfg(feature = "test-support")]
3036 self.snapshot.check_invariants(false);
3037
3038 entry
3039 }
3040
3041 fn populate_dir(
3042 &mut self,
3043 parent_path: Arc<RelPath>,
3044 entries: impl IntoIterator<Item = Entry>,
3045 ignore: Option<Arc<Gitignore>>,
3046 ) {
3047 let mut parent_entry = if let Some(parent_entry) = self
3048 .snapshot
3049 .entries_by_path
3050 .get(&PathKey(parent_path.clone()), ())
3051 {
3052 parent_entry.clone()
3053 } else {
3054 log::warn!(
3055 "populating a directory {:?} that has been removed",
3056 parent_path
3057 );
3058 return;
3059 };
3060
3061 match parent_entry.kind {
3062 EntryKind::PendingDir | EntryKind::UnloadedDir => parent_entry.kind = EntryKind::Dir,
3063 EntryKind::Dir => {}
3064 _ => return,
3065 }
3066
3067 if let Some(ignore) = ignore {
3068 let abs_parent_path = self
3069 .snapshot
3070 .abs_path
3071 .as_path()
3072 .join(parent_path.as_std_path())
3073 .into();
3074 self.snapshot
3075 .ignores_by_parent_abs_path
3076 .insert(abs_parent_path, (ignore, false));
3077 }
3078
3079 let parent_entry_id = parent_entry.id;
3080 self.scanned_dirs.insert(parent_entry_id);
3081 let mut entries_by_path_edits = vec![Edit::Insert(parent_entry)];
3082 let mut entries_by_id_edits = Vec::new();
3083
3084 for entry in entries {
3085 entries_by_id_edits.push(Edit::Insert(PathEntry {
3086 id: entry.id,
3087 path: entry.path.clone(),
3088 is_ignored: entry.is_ignored,
3089 scan_id: self.snapshot.scan_id,
3090 }));
3091 entries_by_path_edits.push(Edit::Insert(entry));
3092 }
3093
3094 self.snapshot
3095 .entries_by_path
3096 .edit(entries_by_path_edits, ());
3097 self.snapshot.entries_by_id.edit(entries_by_id_edits, ());
3098
3099 if let Err(ix) = self.changed_paths.binary_search(&parent_path) {
3100 self.changed_paths.insert(ix, parent_path.clone());
3101 }
3102
3103 #[cfg(feature = "test-support")]
3104 self.snapshot.check_invariants(false);
3105 }
3106
3107 fn remove_path(&mut self, path: &RelPath, watcher: &dyn Watcher) {
3108 log::trace!("background scanner removing path {path:?}");
3109 let mut new_entries;
3110 let removed_entries;
3111 {
3112 let mut cursor = self
3113 .snapshot
3114 .entries_by_path
3115 .cursor::<TraversalProgress>(());
3116 new_entries = cursor.slice(&TraversalTarget::path(path), Bias::Left);
3117 removed_entries = cursor.slice(&TraversalTarget::successor(path), Bias::Left);
3118 new_entries.append(cursor.suffix(), ());
3119 }
3120 self.snapshot.entries_by_path = new_entries;
3121
3122 let mut removed_ids = Vec::with_capacity(removed_entries.summary().count);
3123 let mut removed_dir_abs_paths = Vec::new();
3124 for entry in removed_entries.cursor::<()>(()) {
3125 if entry.is_dir() {
3126 removed_dir_abs_paths.push(self.snapshot.absolutize(&entry.path));
3127 }
3128
3129 match self.removed_entries.entry(entry.inode) {
3130 hash_map::Entry::Occupied(mut e) => {
3131 let prev_removed_entry = e.get_mut();
3132 if entry.id > prev_removed_entry.id {
3133 *prev_removed_entry = entry.clone();
3134 }
3135 }
3136 hash_map::Entry::Vacant(e) => {
3137 e.insert(entry.clone());
3138 }
3139 }
3140
3141 if entry.path.file_name() == Some(GITIGNORE) {
3142 let abs_parent_path = self.snapshot.absolutize(&entry.path.parent().unwrap());
3143 if let Some((_, needs_update)) = self
3144 .snapshot
3145 .ignores_by_parent_abs_path
3146 .get_mut(abs_parent_path.as_path())
3147 {
3148 *needs_update = true;
3149 }
3150 }
3151
3152 if let Err(ix) = removed_ids.binary_search(&entry.id) {
3153 removed_ids.insert(ix, entry.id);
3154 }
3155 }
3156
3157 self.snapshot
3158 .entries_by_id
3159 .edit(removed_ids.iter().map(|&id| Edit::Remove(id)).collect(), ());
3160 self.snapshot
3161 .git_repositories
3162 .retain(|id, _| removed_ids.binary_search(id).is_err());
3163
3164 for removed_dir_abs_path in removed_dir_abs_paths {
3165 watcher.remove(&removed_dir_abs_path).log_err();
3166 }
3167
3168 #[cfg(feature = "test-support")]
3169 self.snapshot.check_invariants(false);
3170 }
3171
3172 async fn insert_git_repository(
3173 &mut self,
3174 dot_git_path: Arc<RelPath>,
3175 fs: &dyn Fs,
3176 watcher: &dyn Watcher,
3177 ) {
3178 let work_dir_path: Arc<RelPath> = match dot_git_path.parent() {
3179 Some(parent_dir) => {
3180 // Guard against repositories inside the repository metadata
3181 if parent_dir
3182 .components()
3183 .any(|component| component == DOT_GIT)
3184 {
3185 log::debug!(
3186 "not building git repository for nested `.git` directory, `.git` path in the worktree: {dot_git_path:?}"
3187 );
3188 return;
3189 };
3190
3191 parent_dir.into()
3192 }
3193 None => {
3194 // `dot_git_path.parent().is_none()` means `.git` directory is the opened worktree itself,
3195 // no files inside that directory are tracked by git, so no need to build the repo around it
3196 log::debug!(
3197 "not building git repository for the worktree itself, `.git` path in the worktree: {dot_git_path:?}"
3198 );
3199 return;
3200 }
3201 };
3202
3203 let dot_git_abs_path = Arc::from(self.snapshot.absolutize(&dot_git_path).as_ref());
3204
3205 self.insert_git_repository_for_path(
3206 WorkDirectory::InProject {
3207 relative_path: work_dir_path,
3208 },
3209 dot_git_abs_path,
3210 fs,
3211 watcher,
3212 )
3213 .await
3214 .log_err();
3215 }
3216
3217 async fn insert_git_repository_for_path(
3218 &mut self,
3219 work_directory: WorkDirectory,
3220 dot_git_abs_path: Arc<Path>,
3221 fs: &dyn Fs,
3222 watcher: &dyn Watcher,
3223 ) -> Result<LocalRepositoryEntry> {
3224 let work_dir_entry = self
3225 .snapshot
3226 .entry_for_path(&work_directory.path_key().0)
3227 .with_context(|| {
3228 format!(
3229 "working directory `{}` not indexed",
3230 work_directory
3231 .path_key()
3232 .0
3233 .display(self.snapshot.path_style)
3234 )
3235 })?;
3236 let work_directory_abs_path = self.snapshot.work_directory_abs_path(&work_directory);
3237
3238 let (repository_dir_abs_path, common_dir_abs_path) =
3239 discover_git_paths(&dot_git_abs_path, fs).await;
3240 watcher
3241 .add(&common_dir_abs_path)
3242 .context("failed to add common directory to watcher")
3243 .log_err();
3244 watcher
3245 .add(&repository_dir_abs_path)
3246 .context("failed to add repository directory to watcher")
3247 .log_err();
3248
3249 let work_directory_id = work_dir_entry.id;
3250
3251 let local_repository = LocalRepositoryEntry {
3252 work_directory_id,
3253 work_directory,
3254 work_directory_abs_path: work_directory_abs_path.as_path().into(),
3255 git_dir_scan_id: 0,
3256 dot_git_abs_path,
3257 common_dir_abs_path,
3258 repository_dir_abs_path,
3259 };
3260
3261 self.snapshot
3262 .git_repositories
3263 .insert(work_directory_id, local_repository.clone());
3264
3265 log::trace!("inserting new local git repository");
3266 Ok(local_repository)
3267 }
3268}
3269
3270async fn is_git_dir(path: &Path, fs: &dyn Fs) -> bool {
3271 if let Some(file_name) = path.file_name()
3272 && file_name == DOT_GIT
3273 {
3274 return true;
3275 }
3276
3277 // If we're in a bare repository, we are not inside a `.git` folder. In a
3278 // bare repository, the root folder contains what would normally be in the
3279 // `.git` folder.
3280 let head_metadata = fs.metadata(&path.join("HEAD")).await;
3281 if !matches!(head_metadata, Ok(Some(_))) {
3282 return false;
3283 }
3284 let config_metadata = fs.metadata(&path.join("config")).await;
3285 matches!(config_metadata, Ok(Some(_)))
3286}
3287
3288async fn build_gitignore(abs_path: &Path, fs: &dyn Fs) -> Result<Gitignore> {
3289 let contents = fs
3290 .load(abs_path)
3291 .await
3292 .with_context(|| format!("failed to load gitignore file at {}", abs_path.display()))?;
3293 let parent = abs_path.parent().unwrap_or_else(|| Path::new("/"));
3294 let mut builder = GitignoreBuilder::new(parent);
3295 for line in contents.lines() {
3296 builder.add_line(Some(abs_path.into()), line)?;
3297 }
3298 Ok(builder.build()?)
3299}
3300
3301impl Deref for Worktree {
3302 type Target = Snapshot;
3303
3304 fn deref(&self) -> &Self::Target {
3305 match self {
3306 Worktree::Local(worktree) => &worktree.snapshot,
3307 Worktree::Remote(worktree) => &worktree.snapshot,
3308 }
3309 }
3310}
3311
3312impl Deref for LocalWorktree {
3313 type Target = LocalSnapshot;
3314
3315 fn deref(&self) -> &Self::Target {
3316 &self.snapshot
3317 }
3318}
3319
3320impl Deref for RemoteWorktree {
3321 type Target = Snapshot;
3322
3323 fn deref(&self) -> &Self::Target {
3324 &self.snapshot
3325 }
3326}
3327
3328impl fmt::Debug for LocalWorktree {
3329 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3330 self.snapshot.fmt(f)
3331 }
3332}
3333
3334impl fmt::Debug for Snapshot {
3335 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3336 struct EntriesById<'a>(&'a SumTree<PathEntry>);
3337 struct EntriesByPath<'a>(&'a SumTree<Entry>);
3338
3339 impl fmt::Debug for EntriesByPath<'_> {
3340 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3341 f.debug_map()
3342 .entries(self.0.iter().map(|entry| (&entry.path, entry.id)))
3343 .finish()
3344 }
3345 }
3346
3347 impl fmt::Debug for EntriesById<'_> {
3348 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3349 f.debug_list().entries(self.0.iter()).finish()
3350 }
3351 }
3352
3353 f.debug_struct("Snapshot")
3354 .field("id", &self.id)
3355 .field("root_name", &self.root_name)
3356 .field("entries_by_path", &EntriesByPath(&self.entries_by_path))
3357 .field("entries_by_id", &EntriesById(&self.entries_by_id))
3358 .finish()
3359 }
3360}
3361
3362#[derive(Debug, Clone, PartialEq)]
3363pub struct File {
3364 pub worktree: Entity<Worktree>,
3365 pub path: Arc<RelPath>,
3366 pub disk_state: DiskState,
3367 pub entry_id: Option<ProjectEntryId>,
3368 pub is_local: bool,
3369 pub is_private: bool,
3370}
3371
3372impl language::File for File {
3373 fn as_local(&self) -> Option<&dyn language::LocalFile> {
3374 if self.is_local { Some(self) } else { None }
3375 }
3376
3377 fn disk_state(&self) -> DiskState {
3378 self.disk_state
3379 }
3380
3381 fn path(&self) -> &Arc<RelPath> {
3382 &self.path
3383 }
3384
3385 fn full_path(&self, cx: &App) -> PathBuf {
3386 self.worktree.read(cx).full_path(&self.path)
3387 }
3388
3389 /// Returns the last component of this handle's absolute path. If this handle refers to the root
3390 /// of its worktree, then this method will return the name of the worktree itself.
3391 fn file_name<'a>(&'a self, cx: &'a App) -> &'a str {
3392 self.path
3393 .file_name()
3394 .unwrap_or_else(|| self.worktree.read(cx).root_name_str())
3395 }
3396
3397 fn worktree_id(&self, cx: &App) -> WorktreeId {
3398 self.worktree.read(cx).id()
3399 }
3400
3401 fn to_proto(&self, cx: &App) -> rpc::proto::File {
3402 rpc::proto::File {
3403 worktree_id: self.worktree.read(cx).id().to_proto(),
3404 entry_id: self.entry_id.map(|id| id.to_proto()),
3405 path: self.path.as_ref().to_proto(),
3406 mtime: self.disk_state.mtime().map(|time| time.into()),
3407 is_deleted: self.disk_state.is_deleted(),
3408 is_historic: matches!(self.disk_state, DiskState::Historic { .. }),
3409 }
3410 }
3411
3412 fn is_private(&self) -> bool {
3413 self.is_private
3414 }
3415
3416 fn path_style(&self, cx: &App) -> PathStyle {
3417 self.worktree.read(cx).path_style()
3418 }
3419
3420 fn can_open(&self) -> bool {
3421 true
3422 }
3423}
3424
3425impl language::LocalFile for File {
3426 fn abs_path(&self, cx: &App) -> PathBuf {
3427 self.worktree.read(cx).absolutize(&self.path)
3428 }
3429
3430 fn load(&self, cx: &App) -> Task<Result<String>> {
3431 let worktree = self.worktree.read(cx).as_local().unwrap();
3432 let abs_path = worktree.absolutize(&self.path);
3433 let fs = worktree.fs.clone();
3434 cx.background_spawn(async move { fs.load(&abs_path).await })
3435 }
3436
3437 fn load_bytes(&self, cx: &App) -> Task<Result<Vec<u8>>> {
3438 let worktree = self.worktree.read(cx).as_local().unwrap();
3439 let abs_path = worktree.absolutize(&self.path);
3440 let fs = worktree.fs.clone();
3441 cx.background_spawn(async move { fs.load_bytes(&abs_path).await })
3442 }
3443}
3444
3445impl File {
3446 pub fn for_entry(entry: Entry, worktree: Entity<Worktree>) -> Arc<Self> {
3447 Arc::new(Self {
3448 worktree,
3449 path: entry.path.clone(),
3450 disk_state: if let Some(mtime) = entry.mtime {
3451 DiskState::Present {
3452 mtime,
3453 size: entry.size,
3454 }
3455 } else {
3456 DiskState::New
3457 },
3458 entry_id: Some(entry.id),
3459 is_local: true,
3460 is_private: entry.is_private,
3461 })
3462 }
3463
3464 pub fn from_proto(
3465 proto: rpc::proto::File,
3466 worktree: Entity<Worktree>,
3467 cx: &App,
3468 ) -> Result<Self> {
3469 let worktree_id = worktree.read(cx).as_remote().context("not remote")?.id();
3470
3471 anyhow::ensure!(
3472 worktree_id.to_proto() == proto.worktree_id,
3473 "worktree id does not match file"
3474 );
3475
3476 let disk_state = if proto.is_historic {
3477 DiskState::Historic {
3478 was_deleted: proto.is_deleted,
3479 }
3480 } else if proto.is_deleted {
3481 DiskState::Deleted
3482 } else if let Some(mtime) = proto.mtime.map(&Into::into) {
3483 DiskState::Present { mtime, size: 0 }
3484 } else {
3485 DiskState::New
3486 };
3487
3488 Ok(Self {
3489 worktree,
3490 path: RelPath::from_proto(&proto.path).context("invalid path in file protobuf")?,
3491 disk_state,
3492 entry_id: proto.entry_id.map(ProjectEntryId::from_proto),
3493 is_local: false,
3494 is_private: false,
3495 })
3496 }
3497
3498 pub fn from_dyn(file: Option<&Arc<dyn language::File>>) -> Option<&Self> {
3499 file.and_then(|f| {
3500 let f: &dyn language::File = f.borrow();
3501 let f: &dyn Any = f;
3502 f.downcast_ref()
3503 })
3504 }
3505
3506 pub fn worktree_id(&self, cx: &App) -> WorktreeId {
3507 self.worktree.read(cx).id()
3508 }
3509
3510 pub fn project_entry_id(&self) -> Option<ProjectEntryId> {
3511 match self.disk_state {
3512 DiskState::Deleted => None,
3513 _ => self.entry_id,
3514 }
3515 }
3516}
3517
3518#[derive(Clone, Debug, PartialEq, Eq)]
3519pub struct Entry {
3520 pub id: ProjectEntryId,
3521 pub kind: EntryKind,
3522 pub path: Arc<RelPath>,
3523 pub inode: u64,
3524 pub mtime: Option<MTime>,
3525
3526 pub canonical_path: Option<Arc<Path>>,
3527 /// Whether this entry is ignored by Git.
3528 ///
3529 /// We only scan ignored entries once the directory is expanded and
3530 /// exclude them from searches.
3531 pub is_ignored: bool,
3532
3533 /// Whether this entry is hidden or inside hidden directory.
3534 ///
3535 /// We only scan hidden entries once the directory is expanded.
3536 pub is_hidden: bool,
3537
3538 /// Whether this entry is always included in searches.
3539 ///
3540 /// This is used for entries that are always included in searches, even
3541 /// if they are ignored by git. Overridden by file_scan_exclusions.
3542 pub is_always_included: bool,
3543
3544 /// Whether this entry's canonical path is outside of the worktree.
3545 /// This means the entry is only accessible from the worktree root via a
3546 /// symlink.
3547 ///
3548 /// We only scan entries outside of the worktree once the symlinked
3549 /// directory is expanded.
3550 pub is_external: bool,
3551
3552 /// Whether this entry is considered to be a `.env` file.
3553 pub is_private: bool,
3554 /// The entry's size on disk, in bytes.
3555 pub size: u64,
3556 pub char_bag: CharBag,
3557 pub is_fifo: bool,
3558}
3559
3560#[derive(Clone, Copy, Debug, PartialEq, Eq)]
3561pub enum EntryKind {
3562 UnloadedDir,
3563 PendingDir,
3564 Dir,
3565 File,
3566}
3567
3568#[derive(Clone, Copy, Debug, PartialEq)]
3569pub enum PathChange {
3570 /// A filesystem entry was was created.
3571 Added,
3572 /// A filesystem entry was removed.
3573 Removed,
3574 /// A filesystem entry was updated.
3575 Updated,
3576 /// A filesystem entry was either updated or added. We don't know
3577 /// whether or not it already existed, because the path had not
3578 /// been loaded before the event.
3579 AddedOrUpdated,
3580 /// A filesystem entry was found during the initial scan of the worktree.
3581 Loaded,
3582}
3583
3584#[derive(Clone, Debug, PartialEq, Eq)]
3585pub struct UpdatedGitRepository {
3586 /// ID of the repository's working directory.
3587 ///
3588 /// For a repo that's above the worktree root, this is the ID of the worktree root, and hence not unique.
3589 /// It's included here to aid the GitStore in detecting when a repository's working directory is renamed.
3590 pub work_directory_id: ProjectEntryId,
3591 pub old_work_directory_abs_path: Option<Arc<Path>>,
3592 pub new_work_directory_abs_path: Option<Arc<Path>>,
3593 /// For a normal git repository checkout, the absolute path to the .git directory.
3594 /// For a worktree, the absolute path to the worktree's subdirectory inside the .git directory.
3595 pub dot_git_abs_path: Option<Arc<Path>>,
3596 pub repository_dir_abs_path: Option<Arc<Path>>,
3597 pub common_dir_abs_path: Option<Arc<Path>>,
3598}
3599
3600pub type UpdatedEntriesSet = Arc<[(Arc<RelPath>, ProjectEntryId, PathChange)]>;
3601pub type UpdatedGitRepositoriesSet = Arc<[UpdatedGitRepository]>;
3602
3603#[derive(Clone, Debug)]
3604pub struct PathProgress<'a> {
3605 pub max_path: &'a RelPath,
3606}
3607
3608#[derive(Clone, Debug)]
3609pub struct PathSummary<S> {
3610 pub max_path: Arc<RelPath>,
3611 pub item_summary: S,
3612}
3613
3614impl<S: Summary> Summary for PathSummary<S> {
3615 type Context<'a> = S::Context<'a>;
3616
3617 fn zero(cx: Self::Context<'_>) -> Self {
3618 Self {
3619 max_path: RelPath::empty().into(),
3620 item_summary: S::zero(cx),
3621 }
3622 }
3623
3624 fn add_summary(&mut self, rhs: &Self, cx: Self::Context<'_>) {
3625 self.max_path = rhs.max_path.clone();
3626 self.item_summary.add_summary(&rhs.item_summary, cx);
3627 }
3628}
3629
3630impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathProgress<'a> {
3631 fn zero(_: <PathSummary<S> as Summary>::Context<'_>) -> Self {
3632 Self {
3633 max_path: RelPath::empty(),
3634 }
3635 }
3636
3637 fn add_summary(
3638 &mut self,
3639 summary: &'a PathSummary<S>,
3640 _: <PathSummary<S> as Summary>::Context<'_>,
3641 ) {
3642 self.max_path = summary.max_path.as_ref()
3643 }
3644}
3645
3646impl<'a> sum_tree::Dimension<'a, PathSummary<GitSummary>> for GitSummary {
3647 fn zero(_cx: ()) -> Self {
3648 Default::default()
3649 }
3650
3651 fn add_summary(&mut self, summary: &'a PathSummary<GitSummary>, _: ()) {
3652 *self += summary.item_summary
3653 }
3654}
3655
3656impl<'a>
3657 sum_tree::SeekTarget<'a, PathSummary<GitSummary>, Dimensions<TraversalProgress<'a>, GitSummary>>
3658 for PathTarget<'_>
3659{
3660 fn cmp(
3661 &self,
3662 cursor_location: &Dimensions<TraversalProgress<'a>, GitSummary>,
3663 _: (),
3664 ) -> Ordering {
3665 self.cmp_path(cursor_location.0.max_path)
3666 }
3667}
3668
3669impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathKey {
3670 fn zero(_: S::Context<'_>) -> Self {
3671 Default::default()
3672 }
3673
3674 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3675 self.0 = summary.max_path.clone();
3676 }
3677}
3678
3679impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for TraversalProgress<'a> {
3680 fn zero(_cx: S::Context<'_>) -> Self {
3681 Default::default()
3682 }
3683
3684 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3685 self.max_path = summary.max_path.as_ref();
3686 }
3687}
3688
3689impl Entry {
3690 fn new(
3691 path: Arc<RelPath>,
3692 metadata: &fs::Metadata,
3693 id: ProjectEntryId,
3694 root_char_bag: CharBag,
3695 canonical_path: Option<Arc<Path>>,
3696 ) -> Self {
3697 let char_bag = char_bag_for_path(root_char_bag, &path);
3698 Self {
3699 id,
3700 kind: if metadata.is_dir {
3701 EntryKind::PendingDir
3702 } else {
3703 EntryKind::File
3704 },
3705 path,
3706 inode: metadata.inode,
3707 mtime: Some(metadata.mtime),
3708 size: metadata.len,
3709 canonical_path,
3710 is_ignored: false,
3711 is_hidden: false,
3712 is_always_included: false,
3713 is_external: false,
3714 is_private: false,
3715 char_bag,
3716 is_fifo: metadata.is_fifo,
3717 }
3718 }
3719
3720 pub fn is_created(&self) -> bool {
3721 self.mtime.is_some()
3722 }
3723
3724 pub fn is_dir(&self) -> bool {
3725 self.kind.is_dir()
3726 }
3727
3728 pub fn is_file(&self) -> bool {
3729 self.kind.is_file()
3730 }
3731}
3732
3733impl EntryKind {
3734 pub fn is_dir(&self) -> bool {
3735 matches!(
3736 self,
3737 EntryKind::Dir | EntryKind::PendingDir | EntryKind::UnloadedDir
3738 )
3739 }
3740
3741 pub fn is_unloaded(&self) -> bool {
3742 matches!(self, EntryKind::UnloadedDir)
3743 }
3744
3745 pub fn is_file(&self) -> bool {
3746 matches!(self, EntryKind::File)
3747 }
3748}
3749
3750impl sum_tree::Item for Entry {
3751 type Summary = EntrySummary;
3752
3753 fn summary(&self, _cx: ()) -> Self::Summary {
3754 let non_ignored_count = if self.is_ignored && !self.is_always_included {
3755 0
3756 } else {
3757 1
3758 };
3759 let file_count;
3760 let non_ignored_file_count;
3761 if self.is_file() {
3762 file_count = 1;
3763 non_ignored_file_count = non_ignored_count;
3764 } else {
3765 file_count = 0;
3766 non_ignored_file_count = 0;
3767 }
3768
3769 EntrySummary {
3770 max_path: self.path.clone(),
3771 count: 1,
3772 non_ignored_count,
3773 file_count,
3774 non_ignored_file_count,
3775 }
3776 }
3777}
3778
3779impl sum_tree::KeyedItem for Entry {
3780 type Key = PathKey;
3781
3782 fn key(&self) -> Self::Key {
3783 PathKey(self.path.clone())
3784 }
3785}
3786
3787#[derive(Clone, Debug)]
3788pub struct EntrySummary {
3789 max_path: Arc<RelPath>,
3790 count: usize,
3791 non_ignored_count: usize,
3792 file_count: usize,
3793 non_ignored_file_count: usize,
3794}
3795
3796impl Default for EntrySummary {
3797 fn default() -> Self {
3798 Self {
3799 max_path: Arc::from(RelPath::empty()),
3800 count: 0,
3801 non_ignored_count: 0,
3802 file_count: 0,
3803 non_ignored_file_count: 0,
3804 }
3805 }
3806}
3807
3808impl sum_tree::ContextLessSummary for EntrySummary {
3809 fn zero() -> Self {
3810 Default::default()
3811 }
3812
3813 fn add_summary(&mut self, rhs: &Self) {
3814 self.max_path = rhs.max_path.clone();
3815 self.count += rhs.count;
3816 self.non_ignored_count += rhs.non_ignored_count;
3817 self.file_count += rhs.file_count;
3818 self.non_ignored_file_count += rhs.non_ignored_file_count;
3819 }
3820}
3821
3822#[derive(Clone, Debug)]
3823struct PathEntry {
3824 id: ProjectEntryId,
3825 path: Arc<RelPath>,
3826 is_ignored: bool,
3827 scan_id: usize,
3828}
3829
3830impl sum_tree::Item for PathEntry {
3831 type Summary = PathEntrySummary;
3832
3833 fn summary(&self, _cx: ()) -> Self::Summary {
3834 PathEntrySummary { max_id: self.id }
3835 }
3836}
3837
3838impl sum_tree::KeyedItem for PathEntry {
3839 type Key = ProjectEntryId;
3840
3841 fn key(&self) -> Self::Key {
3842 self.id
3843 }
3844}
3845
3846#[derive(Clone, Debug, Default)]
3847struct PathEntrySummary {
3848 max_id: ProjectEntryId,
3849}
3850
3851impl sum_tree::ContextLessSummary for PathEntrySummary {
3852 fn zero() -> Self {
3853 Default::default()
3854 }
3855
3856 fn add_summary(&mut self, summary: &Self) {
3857 self.max_id = summary.max_id;
3858 }
3859}
3860
3861impl<'a> sum_tree::Dimension<'a, PathEntrySummary> for ProjectEntryId {
3862 fn zero(_cx: ()) -> Self {
3863 Default::default()
3864 }
3865
3866 fn add_summary(&mut self, summary: &'a PathEntrySummary, _: ()) {
3867 *self = summary.max_id;
3868 }
3869}
3870
3871#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
3872pub struct PathKey(pub Arc<RelPath>);
3873
3874impl Default for PathKey {
3875 fn default() -> Self {
3876 Self(RelPath::empty().into())
3877 }
3878}
3879
3880impl<'a> sum_tree::Dimension<'a, EntrySummary> for PathKey {
3881 fn zero(_cx: ()) -> Self {
3882 Default::default()
3883 }
3884
3885 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
3886 self.0 = summary.max_path.clone();
3887 }
3888}
3889
3890struct BackgroundScanner {
3891 state: async_lock::Mutex<BackgroundScannerState>,
3892 fs: Arc<dyn Fs>,
3893 fs_case_sensitive: bool,
3894 status_updates_tx: UnboundedSender<ScanState>,
3895 executor: BackgroundExecutor,
3896 scan_requests_rx: channel::Receiver<ScanRequest>,
3897 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
3898 next_entry_id: Arc<AtomicUsize>,
3899 phase: BackgroundScannerPhase,
3900 watcher: Arc<dyn Watcher>,
3901 settings: WorktreeSettings,
3902 share_private_files: bool,
3903 /// Whether this is a single-file worktree (root is a file, not a directory).
3904 /// Used to determine if we should give up after repeated canonicalization failures.
3905 is_single_file: bool,
3906}
3907
3908#[derive(Copy, Clone, PartialEq)]
3909enum BackgroundScannerPhase {
3910 InitialScan,
3911 EventsReceivedDuringInitialScan,
3912 Events,
3913}
3914
3915impl BackgroundScanner {
3916 async fn run(&mut self, mut fs_events_rx: Pin<Box<dyn Send + Stream<Item = Vec<PathEvent>>>>) {
3917 let root_abs_path;
3918 let scanning_enabled;
3919 {
3920 let state = self.state.lock().await;
3921 root_abs_path = state.snapshot.abs_path.clone();
3922 scanning_enabled = state.scanning_enabled;
3923 }
3924
3925 // If the worktree root does not contain a git repository, then find
3926 // the git repository in an ancestor directory. Find any gitignore files
3927 // in ancestor directories.
3928 let repo = if scanning_enabled {
3929 let (ignores, exclude, repo) =
3930 discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await;
3931 self.state
3932 .lock()
3933 .await
3934 .snapshot
3935 .ignores_by_parent_abs_path
3936 .extend(ignores);
3937 if let Some(exclude) = exclude {
3938 self.state
3939 .lock()
3940 .await
3941 .snapshot
3942 .repo_exclude_by_work_dir_abs_path
3943 .insert(root_abs_path.as_path().into(), (exclude, false));
3944 }
3945
3946 repo
3947 } else {
3948 None
3949 };
3950
3951 let containing_git_repository = if let Some((ancestor_dot_git, work_directory)) = repo
3952 && scanning_enabled
3953 {
3954 maybe!(async {
3955 self.state
3956 .lock()
3957 .await
3958 .insert_git_repository_for_path(
3959 work_directory,
3960 ancestor_dot_git.clone().into(),
3961 self.fs.as_ref(),
3962 self.watcher.as_ref(),
3963 )
3964 .await
3965 .log_err()?;
3966 Some(ancestor_dot_git)
3967 })
3968 .await
3969 } else {
3970 None
3971 };
3972
3973 log::trace!("containing git repository: {containing_git_repository:?}");
3974
3975 let global_gitignore_file = paths::global_gitignore_path();
3976 let mut global_gitignore_events = if let Some(global_gitignore_path) =
3977 &global_gitignore_file
3978 && scanning_enabled
3979 {
3980 let is_file = self.fs.is_file(&global_gitignore_path).await;
3981 self.state.lock().await.snapshot.global_gitignore = if is_file {
3982 build_gitignore(global_gitignore_path, self.fs.as_ref())
3983 .await
3984 .ok()
3985 .map(Arc::new)
3986 } else {
3987 None
3988 };
3989 if is_file {
3990 self.fs
3991 .watch(global_gitignore_path, FS_WATCH_LATENCY)
3992 .await
3993 .0
3994 } else {
3995 Box::pin(futures::stream::pending())
3996 }
3997 } else {
3998 self.state.lock().await.snapshot.global_gitignore = None;
3999 Box::pin(futures::stream::pending())
4000 };
4001
4002 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4003 {
4004 let mut state = self.state.lock().await;
4005 state.snapshot.scan_id += 1;
4006 if let Some(mut root_entry) = state.snapshot.root_entry().cloned() {
4007 let ignore_stack = state
4008 .snapshot
4009 .ignore_stack_for_abs_path(root_abs_path.as_path(), true, self.fs.as_ref())
4010 .await;
4011 if ignore_stack.is_abs_path_ignored(root_abs_path.as_path(), true) {
4012 root_entry.is_ignored = true;
4013 let mut root_entry = root_entry.clone();
4014 state.reuse_entry_id(&mut root_entry);
4015 state
4016 .insert_entry(root_entry, self.fs.as_ref(), self.watcher.as_ref())
4017 .await;
4018 }
4019 if root_entry.is_dir() && state.scanning_enabled {
4020 state
4021 .enqueue_scan_dir(
4022 root_abs_path.as_path().into(),
4023 &root_entry,
4024 &scan_job_tx,
4025 self.fs.as_ref(),
4026 )
4027 .await;
4028 }
4029 }
4030 };
4031
4032 // Perform an initial scan of the directory.
4033 drop(scan_job_tx);
4034 self.scan_dirs(true, scan_job_rx).await;
4035 {
4036 let mut state = self.state.lock().await;
4037 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4038 }
4039
4040 self.send_status_update(false, SmallVec::new(), &[]).await;
4041
4042 // Process any any FS events that occurred while performing the initial scan.
4043 // For these events, update events cannot be as precise, because we didn't
4044 // have the previous state loaded yet.
4045 self.phase = BackgroundScannerPhase::EventsReceivedDuringInitialScan;
4046 if let Poll::Ready(Some(mut paths)) = futures::poll!(fs_events_rx.next()) {
4047 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
4048 paths.extend(more_paths);
4049 }
4050 self.process_events(
4051 paths
4052 .into_iter()
4053 .filter(|event| event.kind.is_some())
4054 .collect(),
4055 )
4056 .await;
4057 }
4058 if let Some(abs_path) = containing_git_repository {
4059 self.process_events(vec![PathEvent {
4060 path: abs_path,
4061 kind: Some(fs::PathEventKind::Changed),
4062 }])
4063 .await;
4064 }
4065
4066 // Continue processing events until the worktree is dropped.
4067 self.phase = BackgroundScannerPhase::Events;
4068
4069 loop {
4070 select_biased! {
4071 // Process any path refresh requests from the worktree. Prioritize
4072 // these before handling changes reported by the filesystem.
4073 request = self.next_scan_request().fuse() => {
4074 let Ok(request) = request else { break };
4075 if !self.process_scan_request(request, false).await {
4076 return;
4077 }
4078 }
4079
4080 path_prefix_request = self.path_prefixes_to_scan_rx.recv().fuse() => {
4081 let Ok(request) = path_prefix_request else { break };
4082 log::trace!("adding path prefix {:?}", request.path);
4083
4084 let did_scan = self.forcibly_load_paths(std::slice::from_ref(&request.path)).await;
4085 if did_scan {
4086 let abs_path =
4087 {
4088 let mut state = self.state.lock().await;
4089 state.path_prefixes_to_scan.insert(request.path.clone());
4090 state.snapshot.absolutize(&request.path)
4091 };
4092
4093 if let Some(abs_path) = self.fs.canonicalize(&abs_path).await.log_err() {
4094 self.process_events(vec![PathEvent {
4095 path: abs_path,
4096 kind: Some(fs::PathEventKind::Changed),
4097 }])
4098 .await;
4099 }
4100 }
4101 self.send_status_update(false, request.done, &[]).await;
4102 }
4103
4104 paths = fs_events_rx.next().fuse() => {
4105 let Some(mut paths) = paths else { break };
4106 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
4107 paths.extend(more_paths);
4108 }
4109 self.process_events(paths.into_iter().filter(|event| event.kind.is_some()).collect()).await;
4110 }
4111
4112 _ = global_gitignore_events.next().fuse() => {
4113 if let Some(path) = &global_gitignore_file {
4114 self.update_global_gitignore(&path).await;
4115 }
4116 }
4117 }
4118 }
4119 }
4120
4121 async fn process_scan_request(&self, mut request: ScanRequest, scanning: bool) -> bool {
4122 log::debug!("rescanning paths {:?}", request.relative_paths);
4123
4124 request.relative_paths.sort_unstable();
4125 self.forcibly_load_paths(&request.relative_paths).await;
4126
4127 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4128 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4129 let root_canonical_path = match &root_canonical_path {
4130 Ok(path) => SanitizedPath::new(path),
4131 Err(err) => {
4132 log::error!("failed to canonicalize root path {root_path:?}: {err:#}");
4133 return true;
4134 }
4135 };
4136 let abs_paths = request
4137 .relative_paths
4138 .iter()
4139 .map(|path| {
4140 if path.file_name().is_some() {
4141 root_canonical_path.as_path().join(path.as_std_path())
4142 } else {
4143 root_canonical_path.as_path().to_path_buf()
4144 }
4145 })
4146 .collect::<Vec<_>>();
4147
4148 {
4149 let mut state = self.state.lock().await;
4150 let is_idle = state.snapshot.completed_scan_id == state.snapshot.scan_id;
4151 state.snapshot.scan_id += 1;
4152 if is_idle {
4153 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4154 }
4155 }
4156
4157 self.reload_entries_for_paths(
4158 &root_path,
4159 &root_canonical_path,
4160 &request.relative_paths,
4161 abs_paths,
4162 None,
4163 )
4164 .await;
4165
4166 self.send_status_update(scanning, request.done, &[]).await
4167 }
4168
4169 async fn process_events(&self, mut events: Vec<PathEvent>) {
4170 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4171 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4172 let root_canonical_path = match &root_canonical_path {
4173 Ok(path) => SanitizedPath::new(path),
4174 Err(err) => {
4175 let new_path = self
4176 .state
4177 .lock()
4178 .await
4179 .snapshot
4180 .root_file_handle
4181 .clone()
4182 .and_then(|handle| match handle.current_path(&self.fs) {
4183 Ok(new_path) => Some(new_path),
4184 Err(e) => {
4185 log::error!("Failed to refresh worktree root path: {e:#}");
4186 None
4187 }
4188 })
4189 .map(|path| SanitizedPath::new_arc(&path))
4190 .filter(|new_path| *new_path != root_path);
4191
4192 if let Some(new_path) = new_path {
4193 log::info!(
4194 "root renamed from {:?} to {:?}",
4195 root_path.as_path(),
4196 new_path.as_path(),
4197 );
4198 self.status_updates_tx
4199 .unbounded_send(ScanState::RootUpdated { new_path })
4200 .ok();
4201 } else {
4202 log::error!("root path could not be canonicalized: {err:#}");
4203
4204 // For single-file worktrees, if we can't canonicalize and the file handle
4205 // fallback also failed, the file is gone - close the worktree
4206 if self.is_single_file {
4207 log::info!(
4208 "single-file worktree root {:?} no longer exists, marking as deleted",
4209 root_path.as_path()
4210 );
4211 self.status_updates_tx
4212 .unbounded_send(ScanState::RootDeleted)
4213 .ok();
4214 }
4215 }
4216 return;
4217 }
4218 };
4219
4220 // Certain directories may have FS changes, but do not lead to git data changes that Zed cares about.
4221 // Ignore these, to avoid Zed unnecessarily rescanning git metadata.
4222 let skipped_files_in_dot_git = [COMMIT_MESSAGE, INDEX_LOCK];
4223 let skipped_dirs_in_dot_git = [FSMONITOR_DAEMON, LFS_DIR];
4224
4225 let mut relative_paths = Vec::with_capacity(events.len());
4226 let mut dot_git_abs_paths = Vec::new();
4227 let mut work_dirs_needing_exclude_update = Vec::new();
4228 events.sort_unstable_by(|left, right| left.path.cmp(&right.path));
4229 events.dedup_by(|left, right| {
4230 if left.path == right.path {
4231 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4232 right.kind = left.kind;
4233 }
4234 true
4235 } else if left.path.starts_with(&right.path) {
4236 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4237 right.kind = left.kind;
4238 }
4239 true
4240 } else {
4241 false
4242 }
4243 });
4244 {
4245 let snapshot = &self.state.lock().await.snapshot;
4246
4247 let mut ranges_to_drop = SmallVec::<[Range<usize>; 4]>::new();
4248
4249 fn skip_ix(ranges: &mut SmallVec<[Range<usize>; 4]>, ix: usize) {
4250 if let Some(last_range) = ranges.last_mut()
4251 && last_range.end == ix
4252 {
4253 last_range.end += 1;
4254 } else {
4255 ranges.push(ix..ix + 1);
4256 }
4257 }
4258
4259 for (ix, event) in events.iter().enumerate() {
4260 let abs_path = SanitizedPath::new(&event.path);
4261
4262 let mut is_git_related = false;
4263 let mut dot_git_paths = None;
4264
4265 for ancestor in abs_path.as_path().ancestors() {
4266 if is_git_dir(ancestor, self.fs.as_ref()).await {
4267 let path_in_git_dir = abs_path
4268 .as_path()
4269 .strip_prefix(ancestor)
4270 .expect("stripping off the ancestor");
4271 dot_git_paths = Some((ancestor.to_owned(), path_in_git_dir.to_owned()));
4272 break;
4273 }
4274 }
4275
4276 if let Some((dot_git_abs_path, path_in_git_dir)) = dot_git_paths {
4277 // We ignore `""` as well, as that is going to be the
4278 // `.git` folder itself. WE do not care about it, if
4279 // there are changes within we will see them, we need
4280 // this ignore to prevent us from accidentally observing
4281 // the ignored created file due to the events not being
4282 // empty after filtering.
4283
4284 let is_dot_git_changed = {
4285 path_in_git_dir == Path::new("")
4286 && event.kind == Some(PathEventKind::Changed)
4287 && abs_path
4288 .strip_prefix(root_canonical_path)
4289 .ok()
4290 .and_then(|it| RelPath::new(it, PathStyle::local()).ok())
4291 .is_some_and(|it| {
4292 snapshot
4293 .entry_for_path(&it)
4294 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4295 })
4296 };
4297 let condition = skipped_files_in_dot_git.iter().any(|skipped| {
4298 OsStr::new(skipped) == path_in_git_dir.as_path().as_os_str()
4299 }) || skipped_dirs_in_dot_git
4300 .iter()
4301 .any(|skipped_git_subdir| path_in_git_dir.starts_with(skipped_git_subdir))
4302 || is_dot_git_changed;
4303 if condition {
4304 log::debug!(
4305 "ignoring event {abs_path:?} as it's in the .git directory among skipped files or directories"
4306 );
4307 skip_ix(&mut ranges_to_drop, ix);
4308 continue;
4309 }
4310
4311 is_git_related = true;
4312 if !dot_git_abs_paths.contains(&dot_git_abs_path) {
4313 dot_git_abs_paths.push(dot_git_abs_path);
4314 }
4315 }
4316
4317 let relative_path = if let Ok(path) = abs_path.strip_prefix(&root_canonical_path)
4318 && let Ok(path) = RelPath::new(path, PathStyle::local())
4319 {
4320 path
4321 } else {
4322 if is_git_related {
4323 log::debug!(
4324 "ignoring event {abs_path:?}, since it's in git dir outside of root path {root_canonical_path:?}",
4325 );
4326 } else {
4327 log::error!(
4328 "ignoring event {abs_path:?} outside of root path {root_canonical_path:?}",
4329 );
4330 }
4331 skip_ix(&mut ranges_to_drop, ix);
4332 continue;
4333 };
4334
4335 let absolute_path = abs_path.to_path_buf();
4336 if absolute_path.ends_with(Path::new(DOT_GIT).join(REPO_EXCLUDE)) {
4337 if let Some(repository) = snapshot
4338 .git_repositories
4339 .values()
4340 .find(|repo| repo.common_dir_abs_path.join(REPO_EXCLUDE) == absolute_path)
4341 {
4342 work_dirs_needing_exclude_update
4343 .push(repository.work_directory_abs_path.clone());
4344 }
4345 }
4346
4347 if abs_path.file_name() == Some(OsStr::new(GITIGNORE)) {
4348 for (_, repo) in snapshot
4349 .git_repositories
4350 .iter()
4351 .filter(|(_, repo)| repo.directory_contains(&relative_path))
4352 {
4353 if !dot_git_abs_paths.iter().any(|dot_git_abs_path| {
4354 dot_git_abs_path == repo.common_dir_abs_path.as_ref()
4355 }) {
4356 dot_git_abs_paths.push(repo.common_dir_abs_path.to_path_buf());
4357 }
4358 }
4359 }
4360
4361 let parent_dir_is_loaded = relative_path.parent().is_none_or(|parent| {
4362 snapshot
4363 .entry_for_path(parent)
4364 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4365 });
4366 if !parent_dir_is_loaded {
4367 log::debug!("ignoring event {relative_path:?} within unloaded directory");
4368 skip_ix(&mut ranges_to_drop, ix);
4369 continue;
4370 }
4371
4372 if self.settings.is_path_excluded(&relative_path) {
4373 if !is_git_related {
4374 log::debug!("ignoring FS event for excluded path {relative_path:?}");
4375 }
4376 skip_ix(&mut ranges_to_drop, ix);
4377 continue;
4378 }
4379
4380 relative_paths.push(EventRoot {
4381 path: relative_path.into_arc(),
4382 was_rescanned: matches!(event.kind, Some(fs::PathEventKind::Rescan)),
4383 });
4384 }
4385
4386 for range_to_drop in ranges_to_drop.into_iter().rev() {
4387 events.drain(range_to_drop);
4388 }
4389 }
4390
4391 if relative_paths.is_empty() && dot_git_abs_paths.is_empty() {
4392 return;
4393 }
4394
4395 if !work_dirs_needing_exclude_update.is_empty() {
4396 let mut state = self.state.lock().await;
4397 for work_dir_abs_path in work_dirs_needing_exclude_update {
4398 if let Some((_, needs_update)) = state
4399 .snapshot
4400 .repo_exclude_by_work_dir_abs_path
4401 .get_mut(&work_dir_abs_path)
4402 {
4403 *needs_update = true;
4404 }
4405 }
4406 }
4407
4408 self.state.lock().await.snapshot.scan_id += 1;
4409
4410 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4411 log::debug!(
4412 "received fs events {:?}",
4413 relative_paths
4414 .iter()
4415 .map(|event_root| &event_root.path)
4416 .collect::<Vec<_>>()
4417 );
4418 self.reload_entries_for_paths(
4419 &root_path,
4420 &root_canonical_path,
4421 &relative_paths
4422 .iter()
4423 .map(|event_root| event_root.path.clone())
4424 .collect::<Vec<_>>(),
4425 events
4426 .into_iter()
4427 .map(|event| event.path)
4428 .collect::<Vec<_>>(),
4429 Some(scan_job_tx.clone()),
4430 )
4431 .await;
4432
4433 let affected_repo_roots = if !dot_git_abs_paths.is_empty() {
4434 self.update_git_repositories(dot_git_abs_paths).await
4435 } else {
4436 Vec::new()
4437 };
4438
4439 {
4440 let mut ignores_to_update = self.ignores_needing_update().await;
4441 ignores_to_update.extend(affected_repo_roots);
4442 let ignores_to_update = self.order_ignores(ignores_to_update).await;
4443 let snapshot = self.state.lock().await.snapshot.clone();
4444 self.update_ignore_statuses_for_paths(scan_job_tx, snapshot, ignores_to_update)
4445 .await;
4446 self.scan_dirs(false, scan_job_rx).await;
4447 }
4448
4449 {
4450 let mut state = self.state.lock().await;
4451 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4452 for (_, entry) in mem::take(&mut state.removed_entries) {
4453 state.scanned_dirs.remove(&entry.id);
4454 }
4455 }
4456 self.send_status_update(false, SmallVec::new(), &relative_paths)
4457 .await;
4458 }
4459
4460 async fn update_global_gitignore(&self, abs_path: &Path) {
4461 let ignore = build_gitignore(abs_path, self.fs.as_ref())
4462 .await
4463 .log_err()
4464 .map(Arc::new);
4465 let (prev_snapshot, ignore_stack, abs_path) = {
4466 let mut state = self.state.lock().await;
4467 state.snapshot.global_gitignore = ignore;
4468 let abs_path = state.snapshot.abs_path().clone();
4469 let ignore_stack = state
4470 .snapshot
4471 .ignore_stack_for_abs_path(&abs_path, true, self.fs.as_ref())
4472 .await;
4473 (state.snapshot.clone(), ignore_stack, abs_path)
4474 };
4475 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4476 self.update_ignore_statuses_for_paths(
4477 scan_job_tx,
4478 prev_snapshot,
4479 vec![(abs_path, ignore_stack)],
4480 )
4481 .await;
4482 self.scan_dirs(false, scan_job_rx).await;
4483 self.send_status_update(false, SmallVec::new(), &[]).await;
4484 }
4485
4486 async fn forcibly_load_paths(&self, paths: &[Arc<RelPath>]) -> bool {
4487 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4488 {
4489 let mut state = self.state.lock().await;
4490 let root_path = state.snapshot.abs_path.clone();
4491 for path in paths {
4492 for ancestor in path.ancestors() {
4493 if let Some(entry) = state.snapshot.entry_for_path(ancestor)
4494 && entry.kind == EntryKind::UnloadedDir
4495 {
4496 let abs_path = root_path.join(ancestor.as_std_path());
4497 state
4498 .enqueue_scan_dir(
4499 abs_path.into(),
4500 entry,
4501 &scan_job_tx,
4502 self.fs.as_ref(),
4503 )
4504 .await;
4505 state.paths_to_scan.insert(path.clone());
4506 break;
4507 }
4508 }
4509 }
4510 drop(scan_job_tx);
4511 }
4512 while let Ok(job) = scan_job_rx.recv().await {
4513 self.scan_dir(&job).await.log_err();
4514 }
4515
4516 !mem::take(&mut self.state.lock().await.paths_to_scan).is_empty()
4517 }
4518
4519 async fn scan_dirs(
4520 &self,
4521 enable_progress_updates: bool,
4522 scan_jobs_rx: channel::Receiver<ScanJob>,
4523 ) {
4524 if self
4525 .status_updates_tx
4526 .unbounded_send(ScanState::Started)
4527 .is_err()
4528 {
4529 return;
4530 }
4531
4532 let progress_update_count = AtomicUsize::new(0);
4533 self.executor
4534 .scoped_priority(Priority::Low, |scope| {
4535 for _ in 0..self.executor.num_cpus() {
4536 scope.spawn(async {
4537 let mut last_progress_update_count = 0;
4538 let progress_update_timer = self.progress_timer(enable_progress_updates).fuse();
4539 futures::pin_mut!(progress_update_timer);
4540
4541 loop {
4542 select_biased! {
4543 // Process any path refresh requests before moving on to process
4544 // the scan queue, so that user operations are prioritized.
4545 request = self.next_scan_request().fuse() => {
4546 let Ok(request) = request else { break };
4547 if !self.process_scan_request(request, true).await {
4548 return;
4549 }
4550 }
4551
4552 // Send periodic progress updates to the worktree. Use an atomic counter
4553 // to ensure that only one of the workers sends a progress update after
4554 // the update interval elapses.
4555 _ = progress_update_timer => {
4556 match progress_update_count.compare_exchange(
4557 last_progress_update_count,
4558 last_progress_update_count + 1,
4559 SeqCst,
4560 SeqCst
4561 ) {
4562 Ok(_) => {
4563 last_progress_update_count += 1;
4564 self.send_status_update(true, SmallVec::new(), &[])
4565 .await;
4566 }
4567 Err(count) => {
4568 last_progress_update_count = count;
4569 }
4570 }
4571 progress_update_timer.set(self.progress_timer(enable_progress_updates).fuse());
4572 }
4573
4574 // Recursively load directories from the file system.
4575 job = scan_jobs_rx.recv().fuse() => {
4576 let Ok(job) = job else { break };
4577 if let Err(err) = self.scan_dir(&job).await
4578 && job.path.is_empty() {
4579 log::error!("error scanning directory {:?}: {}", job.abs_path, err);
4580 }
4581 }
4582 }
4583 }
4584 });
4585 }
4586 })
4587 .await;
4588 }
4589
4590 async fn send_status_update(
4591 &self,
4592 scanning: bool,
4593 barrier: SmallVec<[barrier::Sender; 1]>,
4594 event_roots: &[EventRoot],
4595 ) -> bool {
4596 let mut state = self.state.lock().await;
4597 if state.changed_paths.is_empty() && event_roots.is_empty() && scanning {
4598 return true;
4599 }
4600
4601 let merged_event_roots = merge_event_roots(&state.changed_paths, event_roots);
4602
4603 let new_snapshot = state.snapshot.clone();
4604 let old_snapshot = mem::replace(&mut state.prev_snapshot, new_snapshot.snapshot.clone());
4605 let changes = build_diff(
4606 self.phase,
4607 &old_snapshot,
4608 &new_snapshot,
4609 &merged_event_roots,
4610 );
4611 state.changed_paths.clear();
4612
4613 self.status_updates_tx
4614 .unbounded_send(ScanState::Updated {
4615 snapshot: new_snapshot,
4616 changes,
4617 scanning,
4618 barrier,
4619 })
4620 .is_ok()
4621 }
4622
4623 async fn scan_dir(&self, job: &ScanJob) -> Result<()> {
4624 let root_abs_path;
4625 let root_char_bag;
4626 {
4627 let snapshot = &self.state.lock().await.snapshot;
4628 if self.settings.is_path_excluded(&job.path) {
4629 log::error!("skipping excluded directory {:?}", job.path);
4630 return Ok(());
4631 }
4632 log::trace!("scanning directory {:?}", job.path);
4633 root_abs_path = snapshot.abs_path().clone();
4634 root_char_bag = snapshot.root_char_bag;
4635 }
4636
4637 let next_entry_id = self.next_entry_id.clone();
4638 let mut ignore_stack = job.ignore_stack.clone();
4639 let mut new_ignore = None;
4640 let mut root_canonical_path = None;
4641 let mut new_entries: Vec<Entry> = Vec::new();
4642 let mut new_jobs: Vec<Option<ScanJob>> = Vec::new();
4643 let mut child_paths = self
4644 .fs
4645 .read_dir(&job.abs_path)
4646 .await?
4647 .filter_map(|entry| async {
4648 match entry {
4649 Ok(entry) => Some(entry),
4650 Err(error) => {
4651 log::error!("error processing entry {:?}", error);
4652 None
4653 }
4654 }
4655 })
4656 .collect::<Vec<_>>()
4657 .await;
4658
4659 // Ensure that .git and .gitignore are processed first.
4660 swap_to_front(&mut child_paths, GITIGNORE);
4661 swap_to_front(&mut child_paths, DOT_GIT);
4662
4663 if let Some(path) = child_paths.first()
4664 && path.ends_with(DOT_GIT)
4665 {
4666 ignore_stack.repo_root = Some(job.abs_path.clone());
4667 }
4668
4669 for child_abs_path in child_paths {
4670 let child_abs_path: Arc<Path> = child_abs_path.into();
4671 let child_name = child_abs_path.file_name().unwrap();
4672 let Some(child_path) = child_name
4673 .to_str()
4674 .and_then(|name| Some(job.path.join(RelPath::unix(name).ok()?)))
4675 else {
4676 continue;
4677 };
4678
4679 if child_name == DOT_GIT {
4680 let mut state = self.state.lock().await;
4681 state
4682 .insert_git_repository(
4683 child_path.clone(),
4684 self.fs.as_ref(),
4685 self.watcher.as_ref(),
4686 )
4687 .await;
4688 } else if child_name == GITIGNORE {
4689 match build_gitignore(&child_abs_path, self.fs.as_ref()).await {
4690 Ok(ignore) => {
4691 let ignore = Arc::new(ignore);
4692 ignore_stack = ignore_stack
4693 .append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4694 new_ignore = Some(ignore);
4695 }
4696 Err(error) => {
4697 log::error!(
4698 "error loading .gitignore file {:?} - {:?}",
4699 child_name,
4700 error
4701 );
4702 }
4703 }
4704 }
4705
4706 if self.settings.is_path_excluded(&child_path) {
4707 log::debug!("skipping excluded child entry {child_path:?}");
4708 self.state
4709 .lock()
4710 .await
4711 .remove_path(&child_path, self.watcher.as_ref());
4712 continue;
4713 }
4714
4715 let child_metadata = match self.fs.metadata(&child_abs_path).await {
4716 Ok(Some(metadata)) => metadata,
4717 Ok(None) => continue,
4718 Err(err) => {
4719 log::error!("error processing {:?}: {err:#}", child_abs_path.display());
4720 continue;
4721 }
4722 };
4723
4724 let mut child_entry = Entry::new(
4725 child_path.clone(),
4726 &child_metadata,
4727 ProjectEntryId::new(&next_entry_id),
4728 root_char_bag,
4729 None,
4730 );
4731
4732 if job.is_external {
4733 child_entry.is_external = true;
4734 } else if child_metadata.is_symlink {
4735 let canonical_path = match self.fs.canonicalize(&child_abs_path).await {
4736 Ok(path) => path,
4737 Err(err) => {
4738 log::error!("error reading target of symlink {child_abs_path:?}: {err:#}",);
4739 continue;
4740 }
4741 };
4742
4743 // lazily canonicalize the root path in order to determine if
4744 // symlinks point outside of the worktree.
4745 let root_canonical_path = match &root_canonical_path {
4746 Some(path) => path,
4747 None => match self.fs.canonicalize(&root_abs_path).await {
4748 Ok(path) => root_canonical_path.insert(path),
4749 Err(err) => {
4750 log::error!("error canonicalizing root {:?}: {:?}", root_abs_path, err);
4751 continue;
4752 }
4753 },
4754 };
4755
4756 if !canonical_path.starts_with(root_canonical_path) {
4757 child_entry.is_external = true;
4758 }
4759
4760 child_entry.canonical_path = Some(canonical_path.into());
4761 }
4762
4763 if child_entry.is_dir() {
4764 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, true);
4765 child_entry.is_always_included =
4766 self.settings.is_path_always_included(&child_path, true);
4767
4768 // Avoid recursing until crash in the case of a recursive symlink
4769 if job.ancestor_inodes.contains(&child_entry.inode) {
4770 new_jobs.push(None);
4771 } else {
4772 let mut ancestor_inodes = job.ancestor_inodes.clone();
4773 ancestor_inodes.insert(child_entry.inode);
4774
4775 new_jobs.push(Some(ScanJob {
4776 abs_path: child_abs_path.clone(),
4777 path: child_path,
4778 is_external: child_entry.is_external,
4779 ignore_stack: if child_entry.is_ignored {
4780 IgnoreStack::all()
4781 } else {
4782 ignore_stack.clone()
4783 },
4784 ancestor_inodes,
4785 scan_queue: job.scan_queue.clone(),
4786 }));
4787 }
4788 } else {
4789 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, false);
4790 child_entry.is_always_included =
4791 self.settings.is_path_always_included(&child_path, false);
4792 }
4793
4794 {
4795 let relative_path = job
4796 .path
4797 .join(RelPath::unix(child_name.to_str().unwrap()).unwrap());
4798 if self.is_path_private(&relative_path) {
4799 log::debug!("detected private file: {relative_path:?}");
4800 child_entry.is_private = true;
4801 }
4802 if self.settings.is_path_hidden(&relative_path) {
4803 log::debug!("detected hidden file: {relative_path:?}");
4804 child_entry.is_hidden = true;
4805 }
4806 }
4807
4808 new_entries.push(child_entry);
4809 }
4810
4811 let mut state = self.state.lock().await;
4812
4813 // Identify any subdirectories that should not be scanned.
4814 let mut job_ix = 0;
4815 for entry in &mut new_entries {
4816 state.reuse_entry_id(entry);
4817 if entry.is_dir() {
4818 if state.should_scan_directory(entry) {
4819 job_ix += 1;
4820 } else {
4821 log::debug!("defer scanning directory {:?}", entry.path);
4822 entry.kind = EntryKind::UnloadedDir;
4823 new_jobs.remove(job_ix);
4824 }
4825 }
4826 if entry.is_always_included {
4827 state
4828 .snapshot
4829 .always_included_entries
4830 .push(entry.path.clone());
4831 }
4832 }
4833
4834 state.populate_dir(job.path.clone(), new_entries, new_ignore);
4835 self.watcher.add(job.abs_path.as_ref()).log_err();
4836
4837 for new_job in new_jobs.into_iter().flatten() {
4838 job.scan_queue
4839 .try_send(new_job)
4840 .expect("channel is unbounded");
4841 }
4842
4843 Ok(())
4844 }
4845
4846 /// All list arguments should be sorted before calling this function
4847 async fn reload_entries_for_paths(
4848 &self,
4849 root_abs_path: &SanitizedPath,
4850 root_canonical_path: &SanitizedPath,
4851 relative_paths: &[Arc<RelPath>],
4852 abs_paths: Vec<PathBuf>,
4853 scan_queue_tx: Option<Sender<ScanJob>>,
4854 ) {
4855 // grab metadata for all requested paths
4856 let metadata = futures::future::join_all(
4857 abs_paths
4858 .iter()
4859 .map(|abs_path| async move {
4860 let metadata = self.fs.metadata(abs_path).await?;
4861 if let Some(metadata) = metadata {
4862 let canonical_path = self.fs.canonicalize(abs_path).await?;
4863
4864 // If we're on a case-insensitive filesystem (default on macOS), we want
4865 // to only ignore metadata for non-symlink files if their absolute-path matches
4866 // the canonical-path.
4867 // Because if not, this might be a case-only-renaming (`mv test.txt TEST.TXT`)
4868 // and we want to ignore the metadata for the old path (`test.txt`) so it's
4869 // treated as removed.
4870 if !self.fs_case_sensitive && !metadata.is_symlink {
4871 let canonical_file_name = canonical_path.file_name();
4872 let file_name = abs_path.file_name();
4873 if canonical_file_name != file_name {
4874 return Ok(None);
4875 }
4876 }
4877
4878 anyhow::Ok(Some((metadata, SanitizedPath::new_arc(&canonical_path))))
4879 } else {
4880 Ok(None)
4881 }
4882 })
4883 .collect::<Vec<_>>(),
4884 )
4885 .await;
4886
4887 let mut new_ancestor_repo = if relative_paths.iter().any(|path| path.is_empty()) {
4888 Some(discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await)
4889 } else {
4890 None
4891 };
4892
4893 let mut state = self.state.lock().await;
4894 let doing_recursive_update = scan_queue_tx.is_some();
4895
4896 // Remove any entries for paths that no longer exist or are being recursively
4897 // refreshed. Do this before adding any new entries, so that renames can be
4898 // detected regardless of the order of the paths.
4899 for (path, metadata) in relative_paths.iter().zip(metadata.iter()) {
4900 if matches!(metadata, Ok(None)) || doing_recursive_update {
4901 state.remove_path(path, self.watcher.as_ref());
4902 }
4903 }
4904
4905 for (path, metadata) in relative_paths.iter().zip(metadata.into_iter()) {
4906 let abs_path: Arc<Path> = root_abs_path.join(path.as_std_path()).into();
4907 match metadata {
4908 Ok(Some((metadata, canonical_path))) => {
4909 let ignore_stack = state
4910 .snapshot
4911 .ignore_stack_for_abs_path(&abs_path, metadata.is_dir, self.fs.as_ref())
4912 .await;
4913 let is_external = !canonical_path.starts_with(&root_canonical_path);
4914 let entry_id = state.entry_id_for(self.next_entry_id.as_ref(), path, &metadata);
4915 let mut fs_entry = Entry::new(
4916 path.clone(),
4917 &metadata,
4918 entry_id,
4919 state.snapshot.root_char_bag,
4920 if metadata.is_symlink {
4921 Some(canonical_path.as_path().to_path_buf().into())
4922 } else {
4923 None
4924 },
4925 );
4926
4927 let is_dir = fs_entry.is_dir();
4928 fs_entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, is_dir);
4929 fs_entry.is_external = is_external;
4930 fs_entry.is_private = self.is_path_private(path);
4931 fs_entry.is_always_included =
4932 self.settings.is_path_always_included(path, is_dir);
4933 fs_entry.is_hidden = self.settings.is_path_hidden(path);
4934
4935 if let (Some(scan_queue_tx), true) = (&scan_queue_tx, is_dir) {
4936 if state.should_scan_directory(&fs_entry)
4937 || (fs_entry.path.is_empty()
4938 && abs_path.file_name() == Some(OsStr::new(DOT_GIT)))
4939 {
4940 state
4941 .enqueue_scan_dir(
4942 abs_path,
4943 &fs_entry,
4944 scan_queue_tx,
4945 self.fs.as_ref(),
4946 )
4947 .await;
4948 } else {
4949 fs_entry.kind = EntryKind::UnloadedDir;
4950 }
4951 }
4952
4953 state
4954 .insert_entry(fs_entry.clone(), self.fs.as_ref(), self.watcher.as_ref())
4955 .await;
4956
4957 if path.is_empty()
4958 && let Some((ignores, exclude, repo)) = new_ancestor_repo.take()
4959 {
4960 log::trace!("updating ancestor git repository");
4961 state.snapshot.ignores_by_parent_abs_path.extend(ignores);
4962 if let Some((ancestor_dot_git, work_directory)) = repo {
4963 if let Some(exclude) = exclude {
4964 let work_directory_abs_path = self
4965 .state
4966 .lock()
4967 .await
4968 .snapshot
4969 .work_directory_abs_path(&work_directory);
4970
4971 state
4972 .snapshot
4973 .repo_exclude_by_work_dir_abs_path
4974 .insert(work_directory_abs_path.into(), (exclude, false));
4975 }
4976 state
4977 .insert_git_repository_for_path(
4978 work_directory,
4979 ancestor_dot_git.into(),
4980 self.fs.as_ref(),
4981 self.watcher.as_ref(),
4982 )
4983 .await
4984 .log_err();
4985 }
4986 }
4987 }
4988 Ok(None) => {
4989 self.remove_repo_path(path.clone(), &mut state.snapshot);
4990 }
4991 Err(err) => {
4992 log::error!("error reading file {abs_path:?} on event: {err:#}");
4993 }
4994 }
4995 }
4996
4997 util::extend_sorted(
4998 &mut state.changed_paths,
4999 relative_paths.iter().cloned(),
5000 usize::MAX,
5001 Ord::cmp,
5002 );
5003 }
5004
5005 fn remove_repo_path(&self, path: Arc<RelPath>, snapshot: &mut LocalSnapshot) -> Option<()> {
5006 if !path.components().any(|component| component == DOT_GIT)
5007 && let Some(local_repo) = snapshot.local_repo_for_work_directory_path(&path)
5008 {
5009 let id = local_repo.work_directory_id;
5010 log::debug!("remove repo path: {:?}", path);
5011 snapshot.git_repositories.remove(&id);
5012 return Some(());
5013 }
5014
5015 Some(())
5016 }
5017
5018 async fn update_ignore_statuses_for_paths(
5019 &self,
5020 scan_job_tx: Sender<ScanJob>,
5021 prev_snapshot: LocalSnapshot,
5022 ignores_to_update: Vec<(Arc<Path>, IgnoreStack)>,
5023 ) {
5024 let (ignore_queue_tx, ignore_queue_rx) = channel::unbounded();
5025 {
5026 for (parent_abs_path, ignore_stack) in ignores_to_update {
5027 ignore_queue_tx
5028 .send_blocking(UpdateIgnoreStatusJob {
5029 abs_path: parent_abs_path,
5030 ignore_stack,
5031 ignore_queue: ignore_queue_tx.clone(),
5032 scan_queue: scan_job_tx.clone(),
5033 })
5034 .unwrap();
5035 }
5036 }
5037 drop(ignore_queue_tx);
5038
5039 self.executor
5040 .scoped(|scope| {
5041 for _ in 0..self.executor.num_cpus() {
5042 scope.spawn(async {
5043 loop {
5044 select_biased! {
5045 // Process any path refresh requests before moving on to process
5046 // the queue of ignore statuses.
5047 request = self.next_scan_request().fuse() => {
5048 let Ok(request) = request else { break };
5049 if !self.process_scan_request(request, true).await {
5050 return;
5051 }
5052 }
5053
5054 // Recursively process directories whose ignores have changed.
5055 job = ignore_queue_rx.recv().fuse() => {
5056 let Ok(job) = job else { break };
5057 self.update_ignore_status(job, &prev_snapshot).await;
5058 }
5059 }
5060 }
5061 });
5062 }
5063 })
5064 .await;
5065 }
5066
5067 async fn ignores_needing_update(&self) -> Vec<Arc<Path>> {
5068 let mut ignores_to_update = Vec::new();
5069 let mut excludes_to_load: Vec<(Arc<Path>, PathBuf)> = Vec::new();
5070
5071 // First pass: collect updates and drop stale entries without awaiting.
5072 {
5073 let snapshot = &mut self.state.lock().await.snapshot;
5074 let abs_path = snapshot.abs_path.clone();
5075 let mut repo_exclude_keys_to_remove: Vec<Arc<Path>> = Vec::new();
5076
5077 for (work_dir_abs_path, (_, needs_update)) in
5078 snapshot.repo_exclude_by_work_dir_abs_path.iter_mut()
5079 {
5080 let repository = snapshot
5081 .git_repositories
5082 .iter()
5083 .find(|(_, repo)| &repo.work_directory_abs_path == work_dir_abs_path);
5084
5085 if *needs_update {
5086 *needs_update = false;
5087 ignores_to_update.push(work_dir_abs_path.clone());
5088
5089 if let Some((_, repository)) = repository {
5090 let exclude_abs_path = repository.common_dir_abs_path.join(REPO_EXCLUDE);
5091 excludes_to_load.push((work_dir_abs_path.clone(), exclude_abs_path));
5092 }
5093 }
5094
5095 if repository.is_none() {
5096 repo_exclude_keys_to_remove.push(work_dir_abs_path.clone());
5097 }
5098 }
5099
5100 for key in repo_exclude_keys_to_remove {
5101 snapshot.repo_exclude_by_work_dir_abs_path.remove(&key);
5102 }
5103
5104 snapshot
5105 .ignores_by_parent_abs_path
5106 .retain(|parent_abs_path, (_, needs_update)| {
5107 if let Ok(parent_path) = parent_abs_path.strip_prefix(abs_path.as_path())
5108 && let Some(parent_path) =
5109 RelPath::new(&parent_path, PathStyle::local()).log_err()
5110 {
5111 if *needs_update {
5112 *needs_update = false;
5113 if snapshot.snapshot.entry_for_path(&parent_path).is_some() {
5114 ignores_to_update.push(parent_abs_path.clone());
5115 }
5116 }
5117
5118 let ignore_path = parent_path.join(RelPath::unix(GITIGNORE).unwrap());
5119 if snapshot.snapshot.entry_for_path(&ignore_path).is_none() {
5120 return false;
5121 }
5122 }
5123 true
5124 });
5125 }
5126
5127 // Load gitignores asynchronously (outside the lock)
5128 let mut loaded_excludes: Vec<(Arc<Path>, Arc<Gitignore>)> = Vec::new();
5129 for (work_dir_abs_path, exclude_abs_path) in excludes_to_load {
5130 if let Ok(current_exclude) = build_gitignore(&exclude_abs_path, self.fs.as_ref()).await
5131 {
5132 loaded_excludes.push((work_dir_abs_path, Arc::new(current_exclude)));
5133 }
5134 }
5135
5136 // Second pass: apply updates.
5137 if !loaded_excludes.is_empty() {
5138 let snapshot = &mut self.state.lock().await.snapshot;
5139
5140 for (work_dir_abs_path, exclude) in loaded_excludes {
5141 if let Some((existing_exclude, _)) = snapshot
5142 .repo_exclude_by_work_dir_abs_path
5143 .get_mut(&work_dir_abs_path)
5144 {
5145 *existing_exclude = exclude;
5146 }
5147 }
5148 }
5149
5150 ignores_to_update
5151 }
5152
5153 async fn order_ignores(&self, mut ignores: Vec<Arc<Path>>) -> Vec<(Arc<Path>, IgnoreStack)> {
5154 let fs = self.fs.clone();
5155 let snapshot = self.state.lock().await.snapshot.clone();
5156 ignores.sort_unstable();
5157 let mut ignores_to_update = ignores.into_iter().peekable();
5158
5159 let mut result = vec![];
5160 while let Some(parent_abs_path) = ignores_to_update.next() {
5161 while ignores_to_update
5162 .peek()
5163 .map_or(false, |p| p.starts_with(&parent_abs_path))
5164 {
5165 ignores_to_update.next().unwrap();
5166 }
5167 let ignore_stack = snapshot
5168 .ignore_stack_for_abs_path(&parent_abs_path, true, fs.as_ref())
5169 .await;
5170 result.push((parent_abs_path, ignore_stack));
5171 }
5172
5173 result
5174 }
5175
5176 async fn update_ignore_status(&self, job: UpdateIgnoreStatusJob, snapshot: &LocalSnapshot) {
5177 log::trace!("update ignore status {:?}", job.abs_path);
5178
5179 let mut ignore_stack = job.ignore_stack;
5180 if let Some((ignore, _)) = snapshot.ignores_by_parent_abs_path.get(&job.abs_path) {
5181 ignore_stack =
5182 ignore_stack.append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
5183 }
5184
5185 let mut entries_by_id_edits = Vec::new();
5186 let mut entries_by_path_edits = Vec::new();
5187 let Some(path) = job
5188 .abs_path
5189 .strip_prefix(snapshot.abs_path.as_path())
5190 .map_err(|_| {
5191 anyhow::anyhow!(
5192 "Failed to strip prefix '{}' from path '{}'",
5193 snapshot.abs_path.as_path().display(),
5194 job.abs_path.display()
5195 )
5196 })
5197 .log_err()
5198 else {
5199 return;
5200 };
5201
5202 let Some(path) = RelPath::new(&path, PathStyle::local()).log_err() else {
5203 return;
5204 };
5205
5206 if let Ok(Some(metadata)) = self.fs.metadata(&job.abs_path.join(DOT_GIT)).await
5207 && metadata.is_dir
5208 {
5209 ignore_stack.repo_root = Some(job.abs_path.clone());
5210 }
5211
5212 for mut entry in snapshot.child_entries(&path).cloned() {
5213 let was_ignored = entry.is_ignored;
5214 let abs_path: Arc<Path> = snapshot.absolutize(&entry.path).into();
5215 entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, entry.is_dir());
5216
5217 if entry.is_dir() {
5218 let child_ignore_stack = if entry.is_ignored {
5219 IgnoreStack::all()
5220 } else {
5221 ignore_stack.clone()
5222 };
5223
5224 // Scan any directories that were previously ignored and weren't previously scanned.
5225 if was_ignored && !entry.is_ignored && entry.kind.is_unloaded() {
5226 let state = self.state.lock().await;
5227 if state.should_scan_directory(&entry) {
5228 state
5229 .enqueue_scan_dir(
5230 abs_path.clone(),
5231 &entry,
5232 &job.scan_queue,
5233 self.fs.as_ref(),
5234 )
5235 .await;
5236 }
5237 }
5238
5239 job.ignore_queue
5240 .send(UpdateIgnoreStatusJob {
5241 abs_path: abs_path.clone(),
5242 ignore_stack: child_ignore_stack,
5243 ignore_queue: job.ignore_queue.clone(),
5244 scan_queue: job.scan_queue.clone(),
5245 })
5246 .await
5247 .unwrap();
5248 }
5249
5250 if entry.is_ignored != was_ignored {
5251 let mut path_entry = snapshot.entries_by_id.get(&entry.id, ()).unwrap().clone();
5252 path_entry.scan_id = snapshot.scan_id;
5253 path_entry.is_ignored = entry.is_ignored;
5254 entries_by_id_edits.push(Edit::Insert(path_entry));
5255 entries_by_path_edits.push(Edit::Insert(entry));
5256 }
5257 }
5258
5259 let state = &mut self.state.lock().await;
5260 for edit in &entries_by_path_edits {
5261 if let Edit::Insert(entry) = edit
5262 && let Err(ix) = state.changed_paths.binary_search(&entry.path)
5263 {
5264 state.changed_paths.insert(ix, entry.path.clone());
5265 }
5266 }
5267
5268 state
5269 .snapshot
5270 .entries_by_path
5271 .edit(entries_by_path_edits, ());
5272 state.snapshot.entries_by_id.edit(entries_by_id_edits, ());
5273 }
5274
5275 async fn update_git_repositories(&self, dot_git_paths: Vec<PathBuf>) -> Vec<Arc<Path>> {
5276 log::trace!("reloading repositories: {dot_git_paths:?}");
5277 let mut state = self.state.lock().await;
5278 let scan_id = state.snapshot.scan_id;
5279 let mut affected_repo_roots = Vec::new();
5280 for dot_git_dir in dot_git_paths {
5281 let existing_repository_entry =
5282 state
5283 .snapshot
5284 .git_repositories
5285 .iter()
5286 .find_map(|(_, repo)| {
5287 let dot_git_dir = SanitizedPath::new(&dot_git_dir);
5288 if SanitizedPath::new(repo.common_dir_abs_path.as_ref()) == dot_git_dir
5289 || SanitizedPath::new(repo.repository_dir_abs_path.as_ref())
5290 == dot_git_dir
5291 {
5292 Some(repo.clone())
5293 } else {
5294 None
5295 }
5296 });
5297
5298 match existing_repository_entry {
5299 None => {
5300 let Ok(relative) = dot_git_dir.strip_prefix(state.snapshot.abs_path()) else {
5301 debug_panic!(
5302 "update_git_repositories called with .git directory outside the worktree root"
5303 );
5304 return Vec::new();
5305 };
5306 affected_repo_roots.push(dot_git_dir.parent().unwrap().into());
5307 state
5308 .insert_git_repository(
5309 RelPath::new(relative, PathStyle::local())
5310 .unwrap()
5311 .into_arc(),
5312 self.fs.as_ref(),
5313 self.watcher.as_ref(),
5314 )
5315 .await;
5316 }
5317 Some(local_repository) => {
5318 state.snapshot.git_repositories.update(
5319 &local_repository.work_directory_id,
5320 |entry| {
5321 entry.git_dir_scan_id = scan_id;
5322 },
5323 );
5324 }
5325 };
5326 }
5327
5328 // Remove any git repositories whose .git entry no longer exists.
5329 let snapshot = &mut state.snapshot;
5330 let mut ids_to_preserve = HashSet::default();
5331 for (&work_directory_id, entry) in snapshot.git_repositories.iter() {
5332 let exists_in_snapshot =
5333 snapshot
5334 .entry_for_id(work_directory_id)
5335 .is_some_and(|entry| {
5336 snapshot
5337 .entry_for_path(&entry.path.join(RelPath::unix(DOT_GIT).unwrap()))
5338 .is_some()
5339 });
5340
5341 if exists_in_snapshot
5342 || matches!(
5343 self.fs.metadata(&entry.common_dir_abs_path).await,
5344 Ok(Some(_))
5345 )
5346 {
5347 ids_to_preserve.insert(work_directory_id);
5348 }
5349 }
5350
5351 snapshot
5352 .git_repositories
5353 .retain(|work_directory_id, entry| {
5354 let preserve = ids_to_preserve.contains(work_directory_id);
5355 if !preserve {
5356 affected_repo_roots.push(entry.dot_git_abs_path.parent().unwrap().into());
5357 snapshot
5358 .repo_exclude_by_work_dir_abs_path
5359 .remove(&entry.work_directory_abs_path);
5360 }
5361 preserve
5362 });
5363
5364 affected_repo_roots
5365 }
5366
5367 async fn progress_timer(&self, running: bool) {
5368 if !running {
5369 return futures::future::pending().await;
5370 }
5371
5372 #[cfg(feature = "test-support")]
5373 if self.fs.is_fake() {
5374 return self.executor.simulate_random_delay().await;
5375 }
5376
5377 self.executor.timer(FS_WATCH_LATENCY).await
5378 }
5379
5380 fn is_path_private(&self, path: &RelPath) -> bool {
5381 !self.share_private_files && self.settings.is_path_private(path)
5382 }
5383
5384 async fn next_scan_request(&self) -> Result<ScanRequest> {
5385 let mut request = self.scan_requests_rx.recv().await?;
5386 while let Ok(next_request) = self.scan_requests_rx.try_recv() {
5387 request.relative_paths.extend(next_request.relative_paths);
5388 request.done.extend(next_request.done);
5389 }
5390 Ok(request)
5391 }
5392}
5393
5394async fn discover_ancestor_git_repo(
5395 fs: Arc<dyn Fs>,
5396 root_abs_path: &SanitizedPath,
5397) -> (
5398 HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
5399 Option<Arc<Gitignore>>,
5400 Option<(PathBuf, WorkDirectory)>,
5401) {
5402 let mut exclude = None;
5403 let mut ignores = HashMap::default();
5404 for (index, ancestor) in root_abs_path.as_path().ancestors().enumerate() {
5405 if index != 0 {
5406 if ancestor == paths::home_dir() {
5407 // Unless $HOME is itself the worktree root, don't consider it as a
5408 // containing git repository---expensive and likely unwanted.
5409 break;
5410 } else if let Ok(ignore) = build_gitignore(&ancestor.join(GITIGNORE), fs.as_ref()).await
5411 {
5412 ignores.insert(ancestor.into(), (ignore.into(), false));
5413 }
5414 }
5415
5416 let ancestor_dot_git = ancestor.join(DOT_GIT);
5417 log::trace!("considering ancestor: {ancestor_dot_git:?}");
5418 // Check whether the directory or file called `.git` exists (in the
5419 // case of worktrees it's a file.)
5420 if fs
5421 .metadata(&ancestor_dot_git)
5422 .await
5423 .is_ok_and(|metadata| metadata.is_some())
5424 {
5425 if index != 0 {
5426 // We canonicalize, since the FS events use the canonicalized path.
5427 if let Some(ancestor_dot_git) = fs.canonicalize(&ancestor_dot_git).await.log_err() {
5428 let location_in_repo = root_abs_path
5429 .as_path()
5430 .strip_prefix(ancestor)
5431 .unwrap()
5432 .into();
5433 log::info!("inserting parent git repo for this worktree: {location_in_repo:?}");
5434 // We associate the external git repo with our root folder and
5435 // also mark where in the git repo the root folder is located.
5436 return (
5437 ignores,
5438 exclude,
5439 Some((
5440 ancestor_dot_git,
5441 WorkDirectory::AboveProject {
5442 absolute_path: ancestor.into(),
5443 location_in_repo,
5444 },
5445 )),
5446 );
5447 };
5448 }
5449
5450 let repo_exclude_abs_path = ancestor_dot_git.join(REPO_EXCLUDE);
5451 if let Ok(repo_exclude) = build_gitignore(&repo_exclude_abs_path, fs.as_ref()).await {
5452 exclude = Some(Arc::new(repo_exclude));
5453 }
5454
5455 // Reached root of git repository.
5456 break;
5457 }
5458 }
5459
5460 (ignores, exclude, None)
5461}
5462
5463fn merge_event_roots(changed_paths: &[Arc<RelPath>], event_roots: &[EventRoot]) -> Vec<EventRoot> {
5464 let mut merged_event_roots = Vec::with_capacity(changed_paths.len() + event_roots.len());
5465 let mut changed_paths = changed_paths.iter().peekable();
5466 let mut event_roots = event_roots.iter().peekable();
5467 while let (Some(path), Some(event_root)) = (changed_paths.peek(), event_roots.peek()) {
5468 match path.cmp(&&event_root.path) {
5469 Ordering::Less => {
5470 merged_event_roots.push(EventRoot {
5471 path: (*changed_paths.next().expect("peeked changed path")).clone(),
5472 was_rescanned: false,
5473 });
5474 }
5475 Ordering::Equal => {
5476 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5477 changed_paths.next();
5478 }
5479 Ordering::Greater => {
5480 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5481 }
5482 }
5483 }
5484 merged_event_roots.extend(changed_paths.map(|path| EventRoot {
5485 path: path.clone(),
5486 was_rescanned: false,
5487 }));
5488 merged_event_roots.extend(event_roots.cloned());
5489 merged_event_roots
5490}
5491
5492fn build_diff(
5493 phase: BackgroundScannerPhase,
5494 old_snapshot: &Snapshot,
5495 new_snapshot: &Snapshot,
5496 event_roots: &[EventRoot],
5497) -> UpdatedEntriesSet {
5498 use BackgroundScannerPhase::*;
5499 use PathChange::{Added, AddedOrUpdated, Loaded, Removed, Updated};
5500
5501 // Identify which paths have changed. Use the known set of changed
5502 // parent paths to optimize the search.
5503 let mut changes = Vec::new();
5504
5505 let mut old_paths = old_snapshot.entries_by_path.cursor::<PathKey>(());
5506 let mut new_paths = new_snapshot.entries_by_path.cursor::<PathKey>(());
5507 let mut last_newly_loaded_dir_path = None;
5508 old_paths.next();
5509 new_paths.next();
5510 for event_root in event_roots {
5511 let path = PathKey(event_root.path.clone());
5512 if old_paths.item().is_some_and(|e| e.path < path.0) {
5513 old_paths.seek_forward(&path, Bias::Left);
5514 }
5515 if new_paths.item().is_some_and(|e| e.path < path.0) {
5516 new_paths.seek_forward(&path, Bias::Left);
5517 }
5518 loop {
5519 match (old_paths.item(), new_paths.item()) {
5520 (Some(old_entry), Some(new_entry)) => {
5521 if old_entry.path > path.0
5522 && new_entry.path > path.0
5523 && !old_entry.path.starts_with(&path.0)
5524 && !new_entry.path.starts_with(&path.0)
5525 {
5526 break;
5527 }
5528
5529 match Ord::cmp(&old_entry.path, &new_entry.path) {
5530 Ordering::Less => {
5531 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5532 old_paths.next();
5533 }
5534 Ordering::Equal => {
5535 if phase == EventsReceivedDuringInitialScan {
5536 if old_entry.id != new_entry.id {
5537 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5538 }
5539 // If the worktree was not fully initialized when this event was generated,
5540 // we can't know whether this entry was added during the scan or whether
5541 // it was merely updated.
5542 changes.push((
5543 new_entry.path.clone(),
5544 new_entry.id,
5545 AddedOrUpdated,
5546 ));
5547 } else if old_entry.id != new_entry.id {
5548 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5549 changes.push((new_entry.path.clone(), new_entry.id, Added));
5550 } else if old_entry != new_entry {
5551 if old_entry.kind.is_unloaded() {
5552 last_newly_loaded_dir_path = Some(&new_entry.path);
5553 changes.push((new_entry.path.clone(), new_entry.id, Loaded));
5554 } else {
5555 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5556 }
5557 } else if event_root.was_rescanned {
5558 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5559 }
5560 old_paths.next();
5561 new_paths.next();
5562 }
5563 Ordering::Greater => {
5564 let is_newly_loaded = phase == InitialScan
5565 || last_newly_loaded_dir_path
5566 .as_ref()
5567 .is_some_and(|dir| new_entry.path.starts_with(dir));
5568 changes.push((
5569 new_entry.path.clone(),
5570 new_entry.id,
5571 if is_newly_loaded { Loaded } else { Added },
5572 ));
5573 new_paths.next();
5574 }
5575 }
5576 }
5577 (Some(old_entry), None) => {
5578 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5579 old_paths.next();
5580 }
5581 (None, Some(new_entry)) => {
5582 let is_newly_loaded = phase == InitialScan
5583 || last_newly_loaded_dir_path
5584 .as_ref()
5585 .is_some_and(|dir| new_entry.path.starts_with(dir));
5586 changes.push((
5587 new_entry.path.clone(),
5588 new_entry.id,
5589 if is_newly_loaded { Loaded } else { Added },
5590 ));
5591 new_paths.next();
5592 }
5593 (None, None) => break,
5594 }
5595 }
5596 }
5597
5598 changes.into()
5599}
5600
5601fn swap_to_front(child_paths: &mut Vec<PathBuf>, file: &str) {
5602 let position = child_paths
5603 .iter()
5604 .position(|path| path.file_name().unwrap() == file);
5605 if let Some(position) = position {
5606 let temp = child_paths.remove(position);
5607 child_paths.insert(0, temp);
5608 }
5609}
5610
5611fn char_bag_for_path(root_char_bag: CharBag, path: &RelPath) -> CharBag {
5612 let mut result = root_char_bag;
5613 result.extend(path.as_unix_str().chars().map(|c| c.to_ascii_lowercase()));
5614 result
5615}
5616
5617#[derive(Debug)]
5618struct ScanJob {
5619 abs_path: Arc<Path>,
5620 path: Arc<RelPath>,
5621 ignore_stack: IgnoreStack,
5622 scan_queue: Sender<ScanJob>,
5623 ancestor_inodes: TreeSet<u64>,
5624 is_external: bool,
5625}
5626
5627struct UpdateIgnoreStatusJob {
5628 abs_path: Arc<Path>,
5629 ignore_stack: IgnoreStack,
5630 ignore_queue: Sender<UpdateIgnoreStatusJob>,
5631 scan_queue: Sender<ScanJob>,
5632}
5633
5634pub trait WorktreeModelHandle {
5635 #[cfg(feature = "test-support")]
5636 fn flush_fs_events<'a>(
5637 &self,
5638 cx: &'a mut gpui::TestAppContext,
5639 ) -> futures::future::LocalBoxFuture<'a, ()>;
5640
5641 #[cfg(feature = "test-support")]
5642 fn flush_fs_events_in_root_git_repository<'a>(
5643 &self,
5644 cx: &'a mut gpui::TestAppContext,
5645 ) -> futures::future::LocalBoxFuture<'a, ()>;
5646}
5647
5648impl WorktreeModelHandle for Entity<Worktree> {
5649 // When the worktree's FS event stream sometimes delivers "redundant" events for FS changes that
5650 // occurred before the worktree was constructed. These events can cause the worktree to perform
5651 // extra directory scans, and emit extra scan-state notifications.
5652 //
5653 // This function mutates the worktree's directory and waits for those mutations to be picked up,
5654 // to ensure that all redundant FS events have already been processed.
5655 #[cfg(feature = "test-support")]
5656 fn flush_fs_events<'a>(
5657 &self,
5658 cx: &'a mut gpui::TestAppContext,
5659 ) -> futures::future::LocalBoxFuture<'a, ()> {
5660 let file_name = "fs-event-sentinel";
5661
5662 let tree = self.clone();
5663 let (fs, root_path) = self.read_with(cx, |tree, _| {
5664 let tree = tree.as_local().unwrap();
5665 (tree.fs.clone(), tree.abs_path.clone())
5666 });
5667
5668 async move {
5669 // Subscribe to events BEFORE creating the file to avoid race condition
5670 // where events fire before subscription is set up
5671 let mut events = cx.events(&tree);
5672
5673 fs.create_file(&root_path.join(file_name), Default::default())
5674 .await
5675 .unwrap();
5676
5677 // Check if condition is already met before waiting for events
5678 let file_exists = || {
5679 tree.read_with(cx, |tree, _| {
5680 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5681 .is_some()
5682 })
5683 };
5684
5685 // Use select to avoid blocking indefinitely if events are delayed
5686 while !file_exists() {
5687 futures::select_biased! {
5688 _ = events.next() => {}
5689 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5690 }
5691 }
5692
5693 fs.remove_file(&root_path.join(file_name), Default::default())
5694 .await
5695 .unwrap();
5696
5697 // Check if condition is already met before waiting for events
5698 let file_gone = || {
5699 tree.read_with(cx, |tree, _| {
5700 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5701 .is_none()
5702 })
5703 };
5704
5705 // Use select to avoid blocking indefinitely if events are delayed
5706 while !file_gone() {
5707 futures::select_biased! {
5708 _ = events.next() => {}
5709 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5710 }
5711 }
5712
5713 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5714 .await;
5715 }
5716 .boxed_local()
5717 }
5718
5719 // This function is similar to flush_fs_events, except that it waits for events to be flushed in
5720 // the .git folder of the root repository.
5721 // The reason for its existence is that a repository's .git folder might live *outside* of the
5722 // worktree and thus its FS events might go through a different path.
5723 // In order to flush those, we need to create artificial events in the .git folder and wait
5724 // for the repository to be reloaded.
5725 #[cfg(feature = "test-support")]
5726 fn flush_fs_events_in_root_git_repository<'a>(
5727 &self,
5728 cx: &'a mut gpui::TestAppContext,
5729 ) -> futures::future::LocalBoxFuture<'a, ()> {
5730 let file_name = "fs-event-sentinel";
5731
5732 let tree = self.clone();
5733 let (fs, root_path, mut git_dir_scan_id) = self.read_with(cx, |tree, _| {
5734 let tree = tree.as_local().unwrap();
5735 let local_repo_entry = tree
5736 .git_repositories
5737 .values()
5738 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5739 .unwrap();
5740 (
5741 tree.fs.clone(),
5742 local_repo_entry.common_dir_abs_path.clone(),
5743 local_repo_entry.git_dir_scan_id,
5744 )
5745 });
5746
5747 let scan_id_increased = |tree: &mut Worktree, git_dir_scan_id: &mut usize| {
5748 let tree = tree.as_local().unwrap();
5749 // let repository = tree.repositories.first().unwrap();
5750 let local_repo_entry = tree
5751 .git_repositories
5752 .values()
5753 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5754 .unwrap();
5755
5756 if local_repo_entry.git_dir_scan_id > *git_dir_scan_id {
5757 *git_dir_scan_id = local_repo_entry.git_dir_scan_id;
5758 true
5759 } else {
5760 false
5761 }
5762 };
5763
5764 async move {
5765 // Subscribe to events BEFORE creating the file to avoid race condition
5766 // where events fire before subscription is set up
5767 let mut events = cx.events(&tree);
5768
5769 fs.create_file(&root_path.join(file_name), Default::default())
5770 .await
5771 .unwrap();
5772
5773 // Use select to avoid blocking indefinitely if events are delayed
5774 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5775 futures::select_biased! {
5776 _ = events.next() => {}
5777 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5778 }
5779 }
5780
5781 fs.remove_file(&root_path.join(file_name), Default::default())
5782 .await
5783 .unwrap();
5784
5785 // Use select to avoid blocking indefinitely if events are delayed
5786 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5787 futures::select_biased! {
5788 _ = events.next() => {}
5789 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5790 }
5791 }
5792
5793 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5794 .await;
5795 }
5796 .boxed_local()
5797 }
5798}
5799
5800#[derive(Clone, Debug)]
5801struct TraversalProgress<'a> {
5802 max_path: &'a RelPath,
5803 count: usize,
5804 non_ignored_count: usize,
5805 file_count: usize,
5806 non_ignored_file_count: usize,
5807}
5808
5809impl TraversalProgress<'_> {
5810 fn count(&self, include_files: bool, include_dirs: bool, include_ignored: bool) -> usize {
5811 match (include_files, include_dirs, include_ignored) {
5812 (true, true, true) => self.count,
5813 (true, true, false) => self.non_ignored_count,
5814 (true, false, true) => self.file_count,
5815 (true, false, false) => self.non_ignored_file_count,
5816 (false, true, true) => self.count - self.file_count,
5817 (false, true, false) => self.non_ignored_count - self.non_ignored_file_count,
5818 (false, false, _) => 0,
5819 }
5820 }
5821}
5822
5823impl<'a> sum_tree::Dimension<'a, EntrySummary> for TraversalProgress<'a> {
5824 fn zero(_cx: ()) -> Self {
5825 Default::default()
5826 }
5827
5828 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
5829 self.max_path = summary.max_path.as_ref();
5830 self.count += summary.count;
5831 self.non_ignored_count += summary.non_ignored_count;
5832 self.file_count += summary.file_count;
5833 self.non_ignored_file_count += summary.non_ignored_file_count;
5834 }
5835}
5836
5837impl Default for TraversalProgress<'_> {
5838 fn default() -> Self {
5839 Self {
5840 max_path: RelPath::empty(),
5841 count: 0,
5842 non_ignored_count: 0,
5843 file_count: 0,
5844 non_ignored_file_count: 0,
5845 }
5846 }
5847}
5848
5849#[derive(Debug)]
5850pub struct Traversal<'a> {
5851 snapshot: &'a Snapshot,
5852 cursor: sum_tree::Cursor<'a, 'static, Entry, TraversalProgress<'a>>,
5853 include_ignored: bool,
5854 include_files: bool,
5855 include_dirs: bool,
5856}
5857
5858impl<'a> Traversal<'a> {
5859 fn new(
5860 snapshot: &'a Snapshot,
5861 include_files: bool,
5862 include_dirs: bool,
5863 include_ignored: bool,
5864 start_path: &RelPath,
5865 ) -> Self {
5866 let mut cursor = snapshot.entries_by_path.cursor(());
5867 cursor.seek(&TraversalTarget::path(start_path), Bias::Left);
5868 let mut traversal = Self {
5869 snapshot,
5870 cursor,
5871 include_files,
5872 include_dirs,
5873 include_ignored,
5874 };
5875 if traversal.end_offset() == traversal.start_offset() {
5876 traversal.next();
5877 }
5878 traversal
5879 }
5880
5881 pub fn advance(&mut self) -> bool {
5882 self.advance_by(1)
5883 }
5884
5885 pub fn advance_by(&mut self, count: usize) -> bool {
5886 self.cursor.seek_forward(
5887 &TraversalTarget::Count {
5888 count: self.end_offset() + count,
5889 include_dirs: self.include_dirs,
5890 include_files: self.include_files,
5891 include_ignored: self.include_ignored,
5892 },
5893 Bias::Left,
5894 )
5895 }
5896
5897 pub fn advance_to_sibling(&mut self) -> bool {
5898 while let Some(entry) = self.cursor.item() {
5899 self.cursor
5900 .seek_forward(&TraversalTarget::successor(&entry.path), Bias::Left);
5901 if let Some(entry) = self.cursor.item()
5902 && (self.include_files || !entry.is_file())
5903 && (self.include_dirs || !entry.is_dir())
5904 && (self.include_ignored || !entry.is_ignored || entry.is_always_included)
5905 {
5906 return true;
5907 }
5908 }
5909 false
5910 }
5911
5912 pub fn back_to_parent(&mut self) -> bool {
5913 let Some(parent_path) = self.cursor.item().and_then(|entry| entry.path.parent()) else {
5914 return false;
5915 };
5916 self.cursor
5917 .seek(&TraversalTarget::path(parent_path), Bias::Left)
5918 }
5919
5920 pub fn entry(&self) -> Option<&'a Entry> {
5921 self.cursor.item()
5922 }
5923
5924 pub fn snapshot(&self) -> &'a Snapshot {
5925 self.snapshot
5926 }
5927
5928 pub fn start_offset(&self) -> usize {
5929 self.cursor
5930 .start()
5931 .count(self.include_files, self.include_dirs, self.include_ignored)
5932 }
5933
5934 pub fn end_offset(&self) -> usize {
5935 self.cursor
5936 .end()
5937 .count(self.include_files, self.include_dirs, self.include_ignored)
5938 }
5939}
5940
5941impl<'a> Iterator for Traversal<'a> {
5942 type Item = &'a Entry;
5943
5944 fn next(&mut self) -> Option<Self::Item> {
5945 if let Some(item) = self.entry() {
5946 self.advance();
5947 Some(item)
5948 } else {
5949 None
5950 }
5951 }
5952}
5953
5954#[derive(Debug, Clone, Copy)]
5955pub enum PathTarget<'a> {
5956 Path(&'a RelPath),
5957 Successor(&'a RelPath),
5958}
5959
5960impl PathTarget<'_> {
5961 fn cmp_path(&self, other: &RelPath) -> Ordering {
5962 match self {
5963 PathTarget::Path(path) => path.cmp(&other),
5964 PathTarget::Successor(path) => {
5965 if other.starts_with(path) {
5966 Ordering::Greater
5967 } else {
5968 Ordering::Equal
5969 }
5970 }
5971 }
5972 }
5973}
5974
5975impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, PathProgress<'a>> for PathTarget<'_> {
5976 fn cmp(&self, cursor_location: &PathProgress<'a>, _: S::Context<'_>) -> Ordering {
5977 self.cmp_path(cursor_location.max_path)
5978 }
5979}
5980
5981impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, TraversalProgress<'a>> for PathTarget<'_> {
5982 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: S::Context<'_>) -> Ordering {
5983 self.cmp_path(cursor_location.max_path)
5984 }
5985}
5986
5987#[derive(Debug)]
5988enum TraversalTarget<'a> {
5989 Path(PathTarget<'a>),
5990 Count {
5991 count: usize,
5992 include_files: bool,
5993 include_ignored: bool,
5994 include_dirs: bool,
5995 },
5996}
5997
5998impl<'a> TraversalTarget<'a> {
5999 fn path(path: &'a RelPath) -> Self {
6000 Self::Path(PathTarget::Path(path))
6001 }
6002
6003 fn successor(path: &'a RelPath) -> Self {
6004 Self::Path(PathTarget::Successor(path))
6005 }
6006
6007 fn cmp_progress(&self, progress: &TraversalProgress) -> Ordering {
6008 match self {
6009 TraversalTarget::Path(path) => path.cmp_path(progress.max_path),
6010 TraversalTarget::Count {
6011 count,
6012 include_files,
6013 include_dirs,
6014 include_ignored,
6015 } => Ord::cmp(
6016 count,
6017 &progress.count(*include_files, *include_dirs, *include_ignored),
6018 ),
6019 }
6020 }
6021}
6022
6023impl<'a> SeekTarget<'a, EntrySummary, TraversalProgress<'a>> for TraversalTarget<'_> {
6024 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
6025 self.cmp_progress(cursor_location)
6026 }
6027}
6028
6029impl<'a> SeekTarget<'a, PathSummary<sum_tree::NoSummary>, TraversalProgress<'a>>
6030 for TraversalTarget<'_>
6031{
6032 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
6033 self.cmp_progress(cursor_location)
6034 }
6035}
6036
6037pub struct ChildEntriesOptions {
6038 pub include_files: bool,
6039 pub include_dirs: bool,
6040 pub include_ignored: bool,
6041}
6042
6043pub struct ChildEntriesIter<'a> {
6044 parent_path: &'a RelPath,
6045 traversal: Traversal<'a>,
6046}
6047
6048impl<'a> Iterator for ChildEntriesIter<'a> {
6049 type Item = &'a Entry;
6050
6051 fn next(&mut self) -> Option<Self::Item> {
6052 if let Some(item) = self.traversal.entry()
6053 && item.path.starts_with(self.parent_path)
6054 {
6055 self.traversal.advance_to_sibling();
6056 return Some(item);
6057 }
6058 None
6059 }
6060}
6061
6062impl<'a> From<&'a Entry> for proto::Entry {
6063 fn from(entry: &'a Entry) -> Self {
6064 Self {
6065 id: entry.id.to_proto(),
6066 is_dir: entry.is_dir(),
6067 path: entry.path.as_ref().to_proto(),
6068 inode: entry.inode,
6069 mtime: entry.mtime.map(|time| time.into()),
6070 is_ignored: entry.is_ignored,
6071 is_hidden: entry.is_hidden,
6072 is_external: entry.is_external,
6073 is_fifo: entry.is_fifo,
6074 size: Some(entry.size),
6075 canonical_path: entry
6076 .canonical_path
6077 .as_ref()
6078 .map(|path| path.to_string_lossy().into_owned()),
6079 }
6080 }
6081}
6082
6083impl TryFrom<(&CharBag, &PathMatcher, proto::Entry)> for Entry {
6084 type Error = anyhow::Error;
6085
6086 fn try_from(
6087 (root_char_bag, always_included, entry): (&CharBag, &PathMatcher, proto::Entry),
6088 ) -> Result<Self> {
6089 let kind = if entry.is_dir {
6090 EntryKind::Dir
6091 } else {
6092 EntryKind::File
6093 };
6094
6095 let path =
6096 RelPath::from_proto(&entry.path).context("invalid relative path in proto message")?;
6097 let char_bag = char_bag_for_path(*root_char_bag, &path);
6098 let is_always_included = always_included.is_match(&path);
6099 Ok(Entry {
6100 id: ProjectEntryId::from_proto(entry.id),
6101 kind,
6102 path,
6103 inode: entry.inode,
6104 mtime: entry.mtime.map(|time| time.into()),
6105 size: entry.size.unwrap_or(0),
6106 canonical_path: entry
6107 .canonical_path
6108 .map(|path_string| Arc::from(PathBuf::from(path_string))),
6109 is_ignored: entry.is_ignored,
6110 is_hidden: entry.is_hidden,
6111 is_always_included,
6112 is_external: entry.is_external,
6113 is_private: false,
6114 char_bag,
6115 is_fifo: entry.is_fifo,
6116 })
6117 }
6118}
6119
6120#[derive(Clone, Copy, Debug, Default, Hash, PartialEq, Eq, PartialOrd, Ord)]
6121pub struct ProjectEntryId(usize);
6122
6123impl ProjectEntryId {
6124 pub const MAX: Self = Self(usize::MAX);
6125 pub const MIN: Self = Self(usize::MIN);
6126
6127 pub fn new(counter: &AtomicUsize) -> Self {
6128 Self(counter.fetch_add(1, SeqCst))
6129 }
6130
6131 pub fn from_proto(id: u64) -> Self {
6132 Self(id as usize)
6133 }
6134
6135 pub fn to_proto(self) -> u64 {
6136 self.0 as u64
6137 }
6138
6139 pub fn from_usize(id: usize) -> Self {
6140 ProjectEntryId(id)
6141 }
6142
6143 pub fn to_usize(self) -> usize {
6144 self.0
6145 }
6146}
6147
6148#[cfg(feature = "test-support")]
6149impl CreatedEntry {
6150 pub fn into_included(self) -> Option<Entry> {
6151 match self {
6152 CreatedEntry::Included(entry) => Some(entry),
6153 CreatedEntry::Excluded { .. } => None,
6154 }
6155 }
6156}
6157
6158fn parse_gitfile(content: &str) -> anyhow::Result<&Path> {
6159 let path = content
6160 .strip_prefix("gitdir:")
6161 .with_context(|| format!("parsing gitfile content {content:?}"))?;
6162 Ok(Path::new(path.trim()))
6163}
6164
6165pub async fn discover_root_repo_common_dir(root_abs_path: &Path, fs: &dyn Fs) -> Option<Arc<Path>> {
6166 let root_dot_git = root_abs_path.join(DOT_GIT);
6167 if !fs.metadata(&root_dot_git).await.is_ok_and(|m| m.is_some()) {
6168 return None;
6169 }
6170 let dot_git_path: Arc<Path> = root_dot_git.into();
6171 let (_, common_dir) = discover_git_paths(&dot_git_path, fs).await;
6172 Some(common_dir)
6173}
6174
6175async fn discover_git_paths(dot_git_abs_path: &Arc<Path>, fs: &dyn Fs) -> (Arc<Path>, Arc<Path>) {
6176 let mut repository_dir_abs_path = dot_git_abs_path.clone();
6177 let mut common_dir_abs_path = dot_git_abs_path.clone();
6178
6179 if let Some(path) = fs
6180 .load(dot_git_abs_path)
6181 .await
6182 .ok()
6183 .as_ref()
6184 .and_then(|contents| parse_gitfile(contents).log_err())
6185 {
6186 let path = dot_git_abs_path
6187 .parent()
6188 .unwrap_or(Path::new(""))
6189 .join(path);
6190 if let Some(path) = fs.canonicalize(&path).await.log_err() {
6191 repository_dir_abs_path = Path::new(&path).into();
6192 common_dir_abs_path = repository_dir_abs_path.clone();
6193
6194 if let Some(commondir_contents) = fs.load(&path.join("commondir")).await.ok()
6195 && let Some(commondir_path) = fs
6196 .canonicalize(&path.join(commondir_contents.trim()))
6197 .await
6198 .log_err()
6199 {
6200 common_dir_abs_path = commondir_path.as_path().into();
6201 }
6202 }
6203 };
6204 (repository_dir_abs_path, common_dir_abs_path)
6205}
6206
6207struct NullWatcher;
6208
6209impl fs::Watcher for NullWatcher {
6210 fn add(&self, _path: &Path) -> Result<()> {
6211 Ok(())
6212 }
6213
6214 fn remove(&self, _path: &Path) -> Result<()> {
6215 Ok(())
6216 }
6217}
6218
6219const FILE_ANALYSIS_BYTES: usize = 1024;
6220
6221async fn decode_file_text(
6222 fs: &dyn Fs,
6223 abs_path: &Path,
6224) -> Result<(String, &'static Encoding, bool)> {
6225 let mut file = fs
6226 .open_sync(&abs_path)
6227 .await
6228 .with_context(|| format!("opening file {abs_path:?}"))?;
6229
6230 // First, read the beginning of the file to determine its kind and encoding.
6231 // We do not want to load an entire large blob into memory only to discard it.
6232 let mut file_first_bytes = Vec::with_capacity(FILE_ANALYSIS_BYTES);
6233 let mut buf = [0u8; FILE_ANALYSIS_BYTES];
6234 let mut reached_eof = false;
6235 loop {
6236 if file_first_bytes.len() >= FILE_ANALYSIS_BYTES {
6237 break;
6238 }
6239 let n = file
6240 .read(&mut buf)
6241 .with_context(|| format!("reading bytes of the file {abs_path:?}"))?;
6242 if n == 0 {
6243 reached_eof = true;
6244 break;
6245 }
6246 file_first_bytes.extend_from_slice(&buf[..n]);
6247 }
6248 let (bom_encoding, byte_content) = decode_byte_header(&file_first_bytes);
6249 anyhow::ensure!(
6250 byte_content != ByteContent::Binary,
6251 "Binary files are not supported"
6252 );
6253
6254 // If the file is eligible for opening, read the rest of the file.
6255 let mut content = file_first_bytes;
6256 if !reached_eof {
6257 let mut buf = [0u8; 8 * 1024];
6258 loop {
6259 let n = file
6260 .read(&mut buf)
6261 .with_context(|| format!("reading remaining bytes of the file {abs_path:?}"))?;
6262 if n == 0 {
6263 break;
6264 }
6265 content.extend_from_slice(&buf[..n]);
6266 }
6267 }
6268 decode_byte_full(content, bom_encoding, byte_content)
6269}
6270
6271fn decode_byte_header(prefix: &[u8]) -> (Option<&'static Encoding>, ByteContent) {
6272 if let Some((encoding, _bom_len)) = Encoding::for_bom(prefix) {
6273 return (Some(encoding), ByteContent::Unknown);
6274 }
6275 (None, analyze_byte_content(prefix))
6276}
6277
6278fn decode_byte_full(
6279 bytes: Vec<u8>,
6280 bom_encoding: Option<&'static Encoding>,
6281 byte_content: ByteContent,
6282) -> Result<(String, &'static Encoding, bool)> {
6283 if let Some(encoding) = bom_encoding {
6284 let (cow, _) = encoding.decode_with_bom_removal(&bytes);
6285 return Ok((cow.into_owned(), encoding, true));
6286 }
6287
6288 match byte_content {
6289 ByteContent::Utf16Le => {
6290 let encoding = encoding_rs::UTF_16LE;
6291 let (cow, _, _) = encoding.decode(&bytes);
6292 return Ok((cow.into_owned(), encoding, false));
6293 }
6294 ByteContent::Utf16Be => {
6295 let encoding = encoding_rs::UTF_16BE;
6296 let (cow, _, _) = encoding.decode(&bytes);
6297 return Ok((cow.into_owned(), encoding, false));
6298 }
6299 ByteContent::Binary => {
6300 anyhow::bail!("Binary files are not supported");
6301 }
6302 ByteContent::Unknown => {}
6303 }
6304
6305 fn detect_encoding(bytes: Vec<u8>) -> (String, &'static Encoding) {
6306 let mut detector = EncodingDetector::new();
6307 detector.feed(&bytes, true);
6308
6309 let encoding = detector.guess(None, true); // Use None for TLD hint to ensure neutral detection logic.
6310
6311 let (cow, _, _) = encoding.decode(&bytes);
6312 (cow.into_owned(), encoding)
6313 }
6314
6315 match String::from_utf8(bytes) {
6316 Ok(text) => {
6317 // ISO-2022-JP (and other ISO-2022 variants) consists entirely of 7-bit ASCII bytes,
6318 // so it is valid UTF-8. However, it contains escape sequences starting with '\x1b'.
6319 // If we find an escape character, we double-check the encoding to prevent
6320 // displaying raw escape sequences instead of the correct characters.
6321 if text.contains('\x1b') {
6322 let (s, enc) = detect_encoding(text.into_bytes());
6323 Ok((s, enc, false))
6324 } else {
6325 Ok((text, encoding_rs::UTF_8, false))
6326 }
6327 }
6328 Err(e) => {
6329 let (s, enc) = detect_encoding(e.into_bytes());
6330 Ok((s, enc, false))
6331 }
6332 }
6333}
6334
6335#[derive(Debug, PartialEq)]
6336enum ByteContent {
6337 Utf16Le,
6338 Utf16Be,
6339 Binary,
6340 Unknown,
6341}
6342
6343// Heuristic check using null byte distribution plus a generic text-likeness
6344// heuristic. This prefers UTF-16 when many bytes are NUL and otherwise
6345// distinguishes between text-like and binary-like content.
6346fn analyze_byte_content(bytes: &[u8]) -> ByteContent {
6347 if bytes.len() < 2 {
6348 return ByteContent::Unknown;
6349 }
6350
6351 if is_known_binary_header(bytes) {
6352 return ByteContent::Binary;
6353 }
6354
6355 let limit = bytes.len().min(FILE_ANALYSIS_BYTES);
6356 let mut even_null_count = 0usize;
6357 let mut odd_null_count = 0usize;
6358 let mut non_text_like_count = 0usize;
6359
6360 for (i, &byte) in bytes[..limit].iter().enumerate() {
6361 if byte == 0 {
6362 if i % 2 == 0 {
6363 even_null_count += 1;
6364 } else {
6365 odd_null_count += 1;
6366 }
6367 non_text_like_count += 1;
6368 continue;
6369 }
6370
6371 let is_text_like = match byte {
6372 b'\t' | b'\n' | b'\r' | 0x0C => true,
6373 0x20..=0x7E => true,
6374 // Treat bytes that are likely part of UTF-8 or single-byte encodings as text-like.
6375 0x80..=0xBF | 0xC2..=0xF4 => true,
6376 _ => false,
6377 };
6378
6379 if !is_text_like {
6380 non_text_like_count += 1;
6381 }
6382 }
6383
6384 let total_null_count = even_null_count + odd_null_count;
6385
6386 // If there are no NUL bytes at all, this is overwhelmingly likely to be text.
6387 if total_null_count == 0 {
6388 return ByteContent::Unknown;
6389 }
6390
6391 let has_significant_nulls = total_null_count >= limit / 16;
6392 let nulls_skew_to_even = even_null_count > odd_null_count * 4;
6393 let nulls_skew_to_odd = odd_null_count > even_null_count * 4;
6394
6395 if has_significant_nulls {
6396 let sample = &bytes[..limit];
6397
6398 // UTF-16BE ASCII: [0x00, char] — nulls at even positions (high byte first)
6399 // UTF-16LE ASCII: [char, 0x00] — nulls at odd positions (low byte first)
6400
6401 if nulls_skew_to_even && is_plausible_utf16_text(sample, false) {
6402 return ByteContent::Utf16Be;
6403 }
6404
6405 if nulls_skew_to_odd && is_plausible_utf16_text(sample, true) {
6406 return ByteContent::Utf16Le;
6407 }
6408
6409 return ByteContent::Binary;
6410 }
6411
6412 if non_text_like_count * 100 < limit * 8 {
6413 ByteContent::Unknown
6414 } else {
6415 ByteContent::Binary
6416 }
6417}
6418
6419fn is_known_binary_header(bytes: &[u8]) -> bool {
6420 bytes.starts_with(b"%PDF-") // PDF
6421 || bytes.starts_with(b"PK\x03\x04") // ZIP local header
6422 || bytes.starts_with(b"PK\x05\x06") // ZIP end of central directory
6423 || bytes.starts_with(b"PK\x07\x08") // ZIP spanning/splitting
6424 || bytes.starts_with(b"\x89PNG\r\n\x1a\n") // PNG
6425 || bytes.starts_with(b"\xFF\xD8\xFF") // JPEG
6426 || bytes.starts_with(b"GIF87a") // GIF87a
6427 || bytes.starts_with(b"GIF89a") // GIF89a
6428 || bytes.starts_with(b"IWAD") // Doom IWAD archive
6429 || bytes.starts_with(b"PWAD") // Doom PWAD archive
6430 || bytes.starts_with(b"RIFF") // WAV, AVI, WebP
6431 || bytes.starts_with(b"OggS") // OGG (Vorbis, Opus, FLAC)
6432 || bytes.starts_with(b"fLaC") // FLAC
6433 || bytes.starts_with(b"ID3") // MP3 with ID3v2 tag
6434 || bytes.starts_with(b"\xFF\xFB") // MP3 frame sync (MPEG1 Layer3)
6435 || bytes.starts_with(b"\xFF\xFA") // MP3 frame sync (MPEG1 Layer3)
6436 || bytes.starts_with(b"\xFF\xF3") // MP3 frame sync (MPEG2 Layer3)
6437 || bytes.starts_with(b"\xFF\xF2") // MP3 frame sync (MPEG2 Layer3)
6438}
6439
6440// Null byte skew alone is not enough to identify UTF-16 -- binary formats with
6441// small 16-bit values (like PCM audio) produce the same pattern. Decode the
6442// bytes as UTF-16 and reject if too many code units land in control character
6443// ranges or form unpaired surrogates, which real text almost never contains.
6444fn is_plausible_utf16_text(bytes: &[u8], little_endian: bool) -> bool {
6445 let mut suspicious_count = 0usize;
6446 let mut total = 0usize;
6447
6448 let mut i = 0;
6449 while let Some(code_unit) = read_u16(bytes, i, little_endian) {
6450 total += 1;
6451
6452 match code_unit {
6453 0x0009 | 0x000A | 0x000C | 0x000D => {}
6454 // C0/C1 control characters and non-characters
6455 0x0000..=0x001F | 0x007F..=0x009F | 0xFFFE | 0xFFFF => suspicious_count += 1,
6456 0xD800..=0xDBFF => {
6457 let next_offset = i + 2;
6458 let has_low_surrogate = read_u16(bytes, next_offset, little_endian)
6459 .is_some_and(|next| (0xDC00..=0xDFFF).contains(&next));
6460 if has_low_surrogate {
6461 total += 1;
6462 i += 2;
6463 } else {
6464 suspicious_count += 1;
6465 }
6466 }
6467 // Lone low surrogate without a preceding high surrogate
6468 0xDC00..=0xDFFF => suspicious_count += 1,
6469 _ => {}
6470 }
6471
6472 i += 2;
6473 }
6474
6475 if total == 0 {
6476 return false;
6477 }
6478
6479 // Real UTF-16 text has near-zero control characters; binary data with
6480 // small 16-bit values typically exceeds 5%. 2% provides a safe margin.
6481 suspicious_count * 100 < total * 2
6482}
6483
6484fn read_u16(bytes: &[u8], offset: usize, little_endian: bool) -> Option<u16> {
6485 let pair = [*bytes.get(offset)?, *bytes.get(offset + 1)?];
6486 if little_endian {
6487 return Some(u16::from_le_bytes(pair));
6488 }
6489 Some(u16::from_be_bytes(pair))
6490}
6491
6492#[cfg(test)]
6493mod tests {
6494 use super::*;
6495
6496 /// reproduction of issue #50785
6497 fn build_pcm16_wav_bytes() -> Vec<u8> {
6498 let header: Vec<u8> = vec![
6499 /* RIFF header */
6500 0x52, 0x49, 0x46, 0x46, // "RIFF"
6501 0xc6, 0xcf, 0x00, 0x00, // file size: 8
6502 0x57, 0x41, 0x56, 0x45, // "WAVE"
6503 /* fmt chunk */
6504 0x66, 0x6d, 0x74, 0x20, // "fmt "
6505 0x10, 0x00, 0x00, 0x00, // chunk size: 16
6506 0x01, 0x00, // format: PCM (1)
6507 0x01, 0x00, // channels: 1 (mono)
6508 0x80, 0x3e, 0x00, 0x00, // sample rate: 16000
6509 0x00, 0x7d, 0x00, 0x00, // byte rate: 32000
6510 0x02, 0x00, // block align: 2
6511 0x10, 0x00, // bits per sample: 16
6512 /* LIST chunk */
6513 0x4c, 0x49, 0x53, 0x54, // "LIST"
6514 0x1a, 0x00, 0x00, 0x00, // chunk size: 26
6515 0x49, 0x4e, 0x46, 0x4f, // "INFO"
6516 0x49, 0x53, 0x46, 0x54, // "ISFT"
6517 0x0d, 0x00, 0x00, 0x00, // sub-chunk size: 13
6518 0x4c, 0x61, 0x76, 0x66, 0x36, 0x32, 0x2e, 0x33, // "Lavf62.3"
6519 0x2e, 0x31, 0x30, 0x30, 0x00, // ".100\0"
6520 /* padding byte for word alignment */
6521 0x00, // data chunk header
6522 0x64, 0x61, 0x74, 0x61, // "data"
6523 0x80, 0xcf, 0x00, 0x00, // chunk size
6524 ];
6525
6526 let mut bytes = header;
6527
6528 // fill remaining space up to `FILE_ANALYSIS_BYTES` with synthetic PCM
6529 let audio_bytes_needed = FILE_ANALYSIS_BYTES - bytes.len();
6530 for i in 0..(audio_bytes_needed / 2) {
6531 let sample = (i & 0xFF) as u8;
6532 bytes.push(sample); // low byte: varies
6533 bytes.push(0x00); // high byte: zero for small values
6534 }
6535
6536 bytes
6537 }
6538
6539 #[test]
6540 fn test_pcm16_wav_detected_as_binary() {
6541 let wav_bytes = build_pcm16_wav_bytes();
6542 assert_eq!(wav_bytes.len(), FILE_ANALYSIS_BYTES);
6543
6544 let result = analyze_byte_content(&wav_bytes);
6545 assert_eq!(
6546 result,
6547 ByteContent::Binary,
6548 "PCM 16-bit WAV should be detected as Binary via RIFF header"
6549 );
6550 }
6551
6552 #[test]
6553 fn test_le16_binary_not_misdetected_as_utf16le() {
6554 let mut bytes = b"FAKE".to_vec();
6555 while bytes.len() < FILE_ANALYSIS_BYTES {
6556 let sample = (bytes.len() & 0xFF) as u8;
6557 bytes.push(sample);
6558 bytes.push(0x00);
6559 }
6560 bytes.truncate(FILE_ANALYSIS_BYTES);
6561
6562 let result = analyze_byte_content(&bytes);
6563 assert_eq!(
6564 result,
6565 ByteContent::Binary,
6566 "LE 16-bit binary with control characters should be detected as Binary"
6567 );
6568 }
6569
6570 #[test]
6571 fn test_be16_binary_not_misdetected_as_utf16be() {
6572 let mut bytes = b"FAKE".to_vec();
6573 while bytes.len() < FILE_ANALYSIS_BYTES {
6574 bytes.push(0x00);
6575 let sample = (bytes.len() & 0xFF) as u8;
6576 bytes.push(sample);
6577 }
6578 bytes.truncate(FILE_ANALYSIS_BYTES);
6579
6580 let result = analyze_byte_content(&bytes);
6581 assert_eq!(
6582 result,
6583 ByteContent::Binary,
6584 "BE 16-bit binary with control characters should be detected as Binary"
6585 );
6586 }
6587
6588 #[test]
6589 fn test_utf16le_text_detected_as_utf16le() {
6590 let text = "Hello, world! This is a UTF-16 test string. ";
6591 let mut bytes = Vec::new();
6592 while bytes.len() < FILE_ANALYSIS_BYTES {
6593 bytes.extend(text.encode_utf16().flat_map(|u| u.to_le_bytes()));
6594 }
6595 bytes.truncate(FILE_ANALYSIS_BYTES);
6596
6597 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Le);
6598 }
6599
6600 #[test]
6601 fn test_utf16be_text_detected_as_utf16be() {
6602 let text = "Hello, world! This is a UTF-16 test string. ";
6603 let mut bytes = Vec::new();
6604 while bytes.len() < FILE_ANALYSIS_BYTES {
6605 bytes.extend(text.encode_utf16().flat_map(|u| u.to_be_bytes()));
6606 }
6607 bytes.truncate(FILE_ANALYSIS_BYTES);
6608
6609 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Be);
6610 }
6611
6612 #[test]
6613 fn test_known_binary_headers() {
6614 let cases: &[(&[u8], &str)] = &[
6615 (b"RIFF\x00\x00\x00\x00WAVE", "WAV"),
6616 (b"RIFF\x00\x00\x00\x00AVI ", "AVI"),
6617 (b"OggS\x00\x02", "OGG"),
6618 (b"fLaC\x00\x00", "FLAC"),
6619 (b"ID3\x03\x00", "MP3 ID3v2"),
6620 (b"\xFF\xFB\x90\x00", "MP3 MPEG1 Layer3"),
6621 (b"\xFF\xF3\x90\x00", "MP3 MPEG2 Layer3"),
6622 ];
6623
6624 for (header, label) in cases {
6625 let mut bytes = header.to_vec();
6626 bytes.resize(FILE_ANALYSIS_BYTES, 0x41); // pad with 'A'
6627 assert_eq!(
6628 analyze_byte_content(&bytes),
6629 ByteContent::Binary,
6630 "{label} should be detected as Binary"
6631 );
6632 }
6633 }
6634}