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