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