1mod ignore;
2mod worktree_settings;
3#[cfg(test)]
4mod worktree_tests;
5
6use ::ignore::gitignore::{Gitignore, GitignoreBuilder};
7use anyhow::{Context as _, Result, anyhow};
8use chardetng::EncodingDetector;
9use clock::ReplicaId;
10use collections::{HashMap, HashSet, VecDeque};
11use encoding_rs::Encoding;
12use fs::{Fs, MTime, PathEvent, RemoveOptions, Watcher, copy_recursive, read_dir_items};
13use futures::{
14 FutureExt as _, Stream, StreamExt,
15 channel::{
16 mpsc::{self, UnboundedSender},
17 oneshot,
18 },
19 select_biased, stream,
20 task::Poll,
21};
22use fuzzy::CharBag;
23use git::{
24 COMMIT_MESSAGE, DOT_GIT, FSMONITOR_DAEMON, GITIGNORE, INDEX_LOCK, LFS_DIR, REPO_EXCLUDE,
25 status::GitSummary,
26};
27use gpui::{
28 App, AppContext as _, AsyncApp, BackgroundExecutor, Context, Entity, EventEmitter, Priority,
29 Task,
30};
31use ignore::IgnoreStack;
32use language::DiskState;
33
34use parking_lot::Mutex;
35use paths::{local_settings_folder_name, local_vscode_folder_name};
36use postage::{
37 barrier,
38 prelude::{Sink as _, Stream as _},
39 watch,
40};
41use rpc::{
42 AnyProtoClient,
43 proto::{self, split_worktree_update},
44};
45pub use settings::WorktreeId;
46use settings::{Settings, SettingsLocation, SettingsStore};
47use smallvec::{SmallVec, smallvec};
48use smol::channel::{self, Sender};
49use std::{
50 any::Any,
51 borrow::Borrow as _,
52 cmp::Ordering,
53 collections::hash_map,
54 convert::TryFrom,
55 ffi::OsStr,
56 fmt,
57 future::Future,
58 mem::{self},
59 ops::{Deref, DerefMut, Range},
60 path::{Path, PathBuf},
61 pin::Pin,
62 sync::{
63 Arc,
64 atomic::{AtomicUsize, Ordering::SeqCst},
65 },
66 time::{Duration, Instant},
67};
68use sum_tree::{Bias, Dimensions, Edit, KeyedItem, SeekTarget, SumTree, Summary, TreeMap, TreeSet};
69use text::{LineEnding, Rope};
70use util::{
71 ResultExt, debug_panic, maybe,
72 paths::{PathMatcher, PathStyle, SanitizedPath, home_dir},
73 rel_path::RelPath,
74};
75pub use worktree_settings::WorktreeSettings;
76
77use crate::ignore::IgnoreKind;
78
79pub const FS_WATCH_LATENCY: Duration = Duration::from_millis(100);
80
81/// A set of local or remote files that are being opened as part of a project.
82/// Responsible for tracking related FS (for local)/collab (for remote) events and corresponding updates.
83/// Stores git repositories data and the diagnostics for the file(s).
84///
85/// Has an absolute path, and may be set to be visible in Zed UI or not.
86/// May correspond to a directory or a single file.
87/// Possible examples:
88/// * a drag and dropped file — may be added as an invisible, "ephemeral" entry to the current worktree
89/// * a directory opened in Zed — may be added as a visible entry to the current worktree
90///
91/// Uses [`Entry`] to track the state of each file/directory, can look up absolute paths for entries.
92pub enum Worktree {
93 Local(LocalWorktree),
94 Remote(RemoteWorktree),
95}
96
97/// An entry, created in the worktree.
98#[derive(Debug)]
99pub enum CreatedEntry {
100 /// Got created and indexed by the worktree, receiving a corresponding entry.
101 Included(Entry),
102 /// Got created, but not indexed due to falling under exclusion filters.
103 Excluded { abs_path: PathBuf },
104}
105
106#[derive(Debug)]
107pub struct LoadedFile {
108 pub file: Arc<File>,
109 pub text: String,
110 pub encoding: &'static Encoding,
111 pub has_bom: bool,
112}
113
114pub struct LoadedBinaryFile {
115 pub file: Arc<File>,
116 pub content: Vec<u8>,
117}
118
119impl fmt::Debug for LoadedBinaryFile {
120 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
121 f.debug_struct("LoadedBinaryFile")
122 .field("file", &self.file)
123 .field("content_bytes", &self.content.len())
124 .finish()
125 }
126}
127
128pub struct LocalWorktree {
129 snapshot: LocalSnapshot,
130 scan_requests_tx: channel::Sender<ScanRequest>,
131 path_prefixes_to_scan_tx: channel::Sender<PathPrefixScanRequest>,
132 is_scanning: (watch::Sender<bool>, watch::Receiver<bool>),
133 _background_scanner_tasks: Vec<Task<()>>,
134 update_observer: Option<UpdateObservationState>,
135 fs: Arc<dyn Fs>,
136 fs_case_sensitive: bool,
137 visible: bool,
138 next_entry_id: Arc<AtomicUsize>,
139 settings: WorktreeSettings,
140 share_private_files: bool,
141 scanning_enabled: bool,
142}
143
144pub struct PathPrefixScanRequest {
145 path: Arc<RelPath>,
146 done: SmallVec<[barrier::Sender; 1]>,
147}
148
149struct ScanRequest {
150 relative_paths: Vec<Arc<RelPath>>,
151 done: SmallVec<[barrier::Sender; 1]>,
152}
153
154pub struct RemoteWorktree {
155 snapshot: Snapshot,
156 background_snapshot: Arc<Mutex<(Snapshot, Vec<proto::UpdateWorktree>)>>,
157 project_id: u64,
158 client: AnyProtoClient,
159 file_scan_inclusions: PathMatcher,
160 updates_tx: Option<UnboundedSender<proto::UpdateWorktree>>,
161 update_observer: Option<mpsc::UnboundedSender<proto::UpdateWorktree>>,
162 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
163 replica_id: ReplicaId,
164 visible: bool,
165 disconnected: bool,
166}
167
168#[derive(Clone)]
169pub struct Snapshot {
170 id: WorktreeId,
171 /// The absolute path of the worktree root.
172 abs_path: Arc<SanitizedPath>,
173 path_style: PathStyle,
174 root_name: Arc<RelPath>,
175 root_char_bag: CharBag,
176 entries_by_path: SumTree<Entry>,
177 entries_by_id: SumTree<PathEntry>,
178 always_included_entries: Vec<Arc<RelPath>>,
179
180 /// A number that increases every time the worktree begins scanning
181 /// a set of paths from the filesystem. This scanning could be caused
182 /// by some operation performed on the worktree, such as reading or
183 /// writing a file, or by an event reported by the filesystem.
184 scan_id: usize,
185
186 /// The latest scan id that has completed, and whose preceding scans
187 /// have all completed. The current `scan_id` could be more than one
188 /// greater than the `completed_scan_id` if operations are performed
189 /// on the worktree while it is processing a file-system event.
190 completed_scan_id: usize,
191}
192
193/// This path corresponds to the 'content path' of a repository in relation
194/// to Zed's project root.
195/// In the majority of the cases, this is the folder that contains the .git folder.
196/// But if a sub-folder of a git repository is opened, this corresponds to the
197/// project root and the .git folder is located in a parent directory.
198#[derive(Clone, Debug, Ord, PartialOrd, Eq, PartialEq, Hash)]
199pub enum WorkDirectory {
200 InProject {
201 relative_path: Arc<RelPath>,
202 },
203 AboveProject {
204 absolute_path: Arc<Path>,
205 location_in_repo: Arc<Path>,
206 },
207}
208
209impl WorkDirectory {
210 fn path_key(&self) -> PathKey {
211 match self {
212 WorkDirectory::InProject { relative_path } => PathKey(relative_path.clone()),
213 WorkDirectory::AboveProject { .. } => PathKey(RelPath::empty().into()),
214 }
215 }
216
217 /// Returns true if the given path is a child of the work directory.
218 ///
219 /// Note that the path may not be a member of this repository, if there
220 /// is a repository in a directory between these two paths
221 /// external .git folder in a parent folder of the project root.
222 #[track_caller]
223 pub fn directory_contains(&self, path: &RelPath) -> bool {
224 match self {
225 WorkDirectory::InProject { relative_path } => path.starts_with(relative_path),
226 WorkDirectory::AboveProject { .. } => true,
227 }
228 }
229}
230
231impl Default for WorkDirectory {
232 fn default() -> Self {
233 Self::InProject {
234 relative_path: Arc::from(RelPath::empty()),
235 }
236 }
237}
238
239#[derive(Clone)]
240pub struct LocalSnapshot {
241 snapshot: Snapshot,
242 global_gitignore: Option<Arc<Gitignore>>,
243 /// Exclude files for all git repositories in the worktree, indexed by their absolute path.
244 /// The boolean indicates whether the gitignore needs to be updated.
245 repo_exclude_by_work_dir_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
246 /// All of the gitignore files in the worktree, indexed by their absolute path.
247 /// The boolean indicates whether the gitignore needs to be updated.
248 ignores_by_parent_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
249 /// All of the git repositories in the worktree, indexed by the project entry
250 /// id of their parent directory.
251 git_repositories: TreeMap<ProjectEntryId, LocalRepositoryEntry>,
252 /// The file handle of the worktree root
253 /// (so we can find it after it's been moved)
254 root_file_handle: Option<Arc<dyn fs::FileHandle>>,
255 executor: BackgroundExecutor,
256}
257
258struct BackgroundScannerState {
259 snapshot: LocalSnapshot,
260 scanned_dirs: HashSet<ProjectEntryId>,
261 path_prefixes_to_scan: HashSet<Arc<RelPath>>,
262 paths_to_scan: HashSet<Arc<RelPath>>,
263 /// The ids of all of the entries that were removed from the snapshot
264 /// as part of the current update. These entry ids may be re-used
265 /// if the same inode is discovered at a new path, or if the given
266 /// path is re-created after being deleted.
267 removed_entries: HashMap<u64, Entry>,
268 changed_paths: Vec<Arc<RelPath>>,
269 prev_snapshot: Snapshot,
270}
271
272#[derive(Debug, Clone)]
273struct LocalRepositoryEntry {
274 work_directory_id: ProjectEntryId,
275 work_directory: WorkDirectory,
276 work_directory_abs_path: Arc<Path>,
277 git_dir_scan_id: usize,
278 /// Absolute path to the original .git entry that caused us to create this repository.
279 ///
280 /// This is normally a directory, but may be a "gitfile" that points to a directory elsewhere
281 /// (whose path we then store in `repository_dir_abs_path`).
282 dot_git_abs_path: Arc<Path>,
283 /// Absolute path to the "commondir" for this repository.
284 ///
285 /// This is always a directory. For a normal repository, this is the same as dot_git_abs_path,
286 /// but in the case of a submodule or a worktree it is the path to the "parent" .git directory
287 /// from which the submodule/worktree was derived.
288 common_dir_abs_path: Arc<Path>,
289 /// Absolute path to the directory holding the repository's state.
290 ///
291 /// For a normal repository, this is a directory and coincides with `dot_git_abs_path` and
292 /// `common_dir_abs_path`. For a submodule or worktree, this is some subdirectory of the
293 /// commondir like `/project/.git/modules/foo`.
294 repository_dir_abs_path: Arc<Path>,
295}
296
297impl sum_tree::Item for LocalRepositoryEntry {
298 type Summary = PathSummary<sum_tree::NoSummary>;
299
300 fn summary(&self, _: <Self::Summary as Summary>::Context<'_>) -> Self::Summary {
301 PathSummary {
302 max_path: self.work_directory.path_key().0,
303 item_summary: sum_tree::NoSummary,
304 }
305 }
306}
307
308impl KeyedItem for LocalRepositoryEntry {
309 type Key = PathKey;
310
311 fn key(&self) -> Self::Key {
312 self.work_directory.path_key()
313 }
314}
315
316impl Deref for LocalRepositoryEntry {
317 type Target = WorkDirectory;
318
319 fn deref(&self) -> &Self::Target {
320 &self.work_directory
321 }
322}
323
324impl Deref for LocalSnapshot {
325 type Target = Snapshot;
326
327 fn deref(&self) -> &Self::Target {
328 &self.snapshot
329 }
330}
331
332impl DerefMut for LocalSnapshot {
333 fn deref_mut(&mut self) -> &mut Self::Target {
334 &mut self.snapshot
335 }
336}
337
338enum ScanState {
339 Started,
340 Updated {
341 snapshot: LocalSnapshot,
342 changes: UpdatedEntriesSet,
343 barrier: SmallVec<[barrier::Sender; 1]>,
344 scanning: bool,
345 },
346 RootUpdated {
347 new_path: Arc<SanitizedPath>,
348 },
349}
350
351struct UpdateObservationState {
352 snapshots_tx: mpsc::UnboundedSender<(LocalSnapshot, UpdatedEntriesSet)>,
353 resume_updates: watch::Sender<()>,
354 _maintain_remote_snapshot: Task<Option<()>>,
355}
356
357#[derive(Clone)]
358pub enum Event {
359 UpdatedEntries(UpdatedEntriesSet),
360 UpdatedGitRepositories(UpdatedGitRepositoriesSet),
361 DeletedEntry(ProjectEntryId),
362}
363
364impl EventEmitter<Event> for Worktree {}
365
366impl Worktree {
367 pub async fn local(
368 path: impl Into<Arc<Path>>,
369 visible: bool,
370 fs: Arc<dyn Fs>,
371 next_entry_id: Arc<AtomicUsize>,
372 scanning_enabled: bool,
373 cx: &mut AsyncApp,
374 ) -> Result<Entity<Self>> {
375 let abs_path = path.into();
376 let metadata = fs
377 .metadata(&abs_path)
378 .await
379 .context("failed to stat worktree path")?;
380
381 let fs_case_sensitive = fs.is_case_sensitive().await.unwrap_or_else(|e| {
382 log::error!(
383 "Failed to determine whether filesystem is case sensitive (falling back to true) due to error: {e:#}"
384 );
385 true
386 });
387
388 let root_file_handle = if metadata.as_ref().is_some() {
389 fs.open_handle(&abs_path)
390 .await
391 .with_context(|| {
392 format!(
393 "failed to open local worktree root at {}",
394 abs_path.display()
395 )
396 })
397 .log_err()
398 } else {
399 None
400 };
401
402 cx.new(move |cx: &mut Context<Worktree>| {
403 let mut snapshot = LocalSnapshot {
404 ignores_by_parent_abs_path: Default::default(),
405 global_gitignore: Default::default(),
406 repo_exclude_by_work_dir_abs_path: Default::default(),
407 git_repositories: Default::default(),
408 snapshot: Snapshot::new(
409 cx.entity_id().as_u64(),
410 abs_path
411 .file_name()
412 .and_then(|f| f.to_str())
413 .map_or(RelPath::empty().into(), |f| {
414 RelPath::unix(f).unwrap().into()
415 }),
416 abs_path.clone(),
417 PathStyle::local(),
418 ),
419 root_file_handle,
420 executor: cx.background_executor().clone(),
421 };
422
423 let worktree_id = snapshot.id();
424 let settings_location = Some(SettingsLocation {
425 worktree_id,
426 path: RelPath::empty(),
427 });
428
429 let settings = WorktreeSettings::get(settings_location, cx).clone();
430 cx.observe_global::<SettingsStore>(move |this, cx| {
431 if let Self::Local(this) = this {
432 let settings = WorktreeSettings::get(settings_location, cx).clone();
433 if this.settings != settings {
434 this.settings = settings;
435 this.restart_background_scanners(cx);
436 }
437 }
438 })
439 .detach();
440
441 let share_private_files = false;
442 if let Some(metadata) = metadata {
443 let mut entry = Entry::new(
444 RelPath::empty().into(),
445 &metadata,
446 ProjectEntryId::new(&next_entry_id),
447 snapshot.root_char_bag,
448 None,
449 );
450 if !metadata.is_dir {
451 if let Some(file_name) = abs_path.file_name()
452 && let Some(file_name) = file_name.to_str()
453 && let Ok(path) = RelPath::unix(file_name)
454 {
455 entry.is_private = !share_private_files && settings.is_path_private(path);
456 entry.is_hidden = settings.is_path_hidden(path);
457 }
458 }
459 snapshot.insert_entry(entry, fs.as_ref());
460 }
461
462 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
463 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
464 let mut worktree = LocalWorktree {
465 share_private_files,
466 next_entry_id,
467 snapshot,
468 is_scanning: watch::channel_with(true),
469 update_observer: None,
470 scan_requests_tx,
471 path_prefixes_to_scan_tx,
472 _background_scanner_tasks: Vec::new(),
473 fs,
474 fs_case_sensitive,
475 visible,
476 settings,
477 scanning_enabled,
478 };
479 worktree.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
480 Worktree::Local(worktree)
481 })
482 }
483
484 pub fn remote(
485 project_id: u64,
486 replica_id: ReplicaId,
487 worktree: proto::WorktreeMetadata,
488 client: AnyProtoClient,
489 path_style: PathStyle,
490 cx: &mut App,
491 ) -> Entity<Self> {
492 cx.new(|cx: &mut Context<Self>| {
493 let snapshot = Snapshot::new(
494 worktree.id,
495 RelPath::from_proto(&worktree.root_name)
496 .unwrap_or_else(|_| RelPath::empty().into()),
497 Path::new(&worktree.abs_path).into(),
498 path_style,
499 );
500
501 let background_snapshot = Arc::new(Mutex::new((
502 snapshot.clone(),
503 Vec::<proto::UpdateWorktree>::new(),
504 )));
505 let (background_updates_tx, mut background_updates_rx) =
506 mpsc::unbounded::<proto::UpdateWorktree>();
507 let (mut snapshot_updated_tx, mut snapshot_updated_rx) = watch::channel();
508
509 let worktree_id = snapshot.id();
510 let settings_location = Some(SettingsLocation {
511 worktree_id,
512 path: RelPath::empty(),
513 });
514
515 let settings = WorktreeSettings::get(settings_location, cx).clone();
516 let worktree = RemoteWorktree {
517 client,
518 project_id,
519 replica_id,
520 snapshot,
521 file_scan_inclusions: settings.parent_dir_scan_inclusions.clone(),
522 background_snapshot: background_snapshot.clone(),
523 updates_tx: Some(background_updates_tx),
524 update_observer: None,
525 snapshot_subscriptions: Default::default(),
526 visible: worktree.visible,
527 disconnected: false,
528 };
529
530 // Apply updates to a separate snapshot in a background task, then
531 // send them to a foreground task which updates the model.
532 cx.background_spawn(async move {
533 while let Some(update) = background_updates_rx.next().await {
534 {
535 let mut lock = background_snapshot.lock();
536 lock.0.apply_remote_update(
537 update.clone(),
538 &settings.parent_dir_scan_inclusions,
539 );
540 lock.1.push(update);
541 }
542 snapshot_updated_tx.send(()).await.ok();
543 }
544 })
545 .detach();
546
547 // On the foreground task, update to the latest snapshot and notify
548 // any update observer of all updates that led to that snapshot.
549 cx.spawn(async move |this, cx| {
550 while (snapshot_updated_rx.recv().await).is_some() {
551 this.update(cx, |this, cx| {
552 let mut entries_changed = false;
553 let this = this.as_remote_mut().unwrap();
554 {
555 let mut lock = this.background_snapshot.lock();
556 this.snapshot = lock.0.clone();
557 for update in lock.1.drain(..) {
558 entries_changed |= !update.updated_entries.is_empty()
559 || !update.removed_entries.is_empty();
560 if let Some(tx) = &this.update_observer {
561 tx.unbounded_send(update).ok();
562 }
563 }
564 };
565
566 if entries_changed {
567 cx.emit(Event::UpdatedEntries(Arc::default()));
568 }
569 cx.notify();
570 while let Some((scan_id, _)) = this.snapshot_subscriptions.front() {
571 if this.observed_snapshot(*scan_id) {
572 let (_, tx) = this.snapshot_subscriptions.pop_front().unwrap();
573 let _ = tx.send(());
574 } else {
575 break;
576 }
577 }
578 })?;
579 }
580 anyhow::Ok(())
581 })
582 .detach();
583
584 Worktree::Remote(worktree)
585 })
586 }
587
588 pub fn as_local(&self) -> Option<&LocalWorktree> {
589 if let Worktree::Local(worktree) = self {
590 Some(worktree)
591 } else {
592 None
593 }
594 }
595
596 pub fn as_remote(&self) -> Option<&RemoteWorktree> {
597 if let Worktree::Remote(worktree) = self {
598 Some(worktree)
599 } else {
600 None
601 }
602 }
603
604 pub fn as_local_mut(&mut self) -> Option<&mut LocalWorktree> {
605 if let Worktree::Local(worktree) = self {
606 Some(worktree)
607 } else {
608 None
609 }
610 }
611
612 pub fn as_remote_mut(&mut self) -> Option<&mut RemoteWorktree> {
613 if let Worktree::Remote(worktree) = self {
614 Some(worktree)
615 } else {
616 None
617 }
618 }
619
620 pub fn is_local(&self) -> bool {
621 matches!(self, Worktree::Local(_))
622 }
623
624 pub fn is_remote(&self) -> bool {
625 !self.is_local()
626 }
627
628 pub fn settings_location(&self, _: &Context<Self>) -> SettingsLocation<'static> {
629 SettingsLocation {
630 worktree_id: self.id(),
631 path: RelPath::empty(),
632 }
633 }
634
635 pub fn snapshot(&self) -> Snapshot {
636 match self {
637 Worktree::Local(worktree) => worktree.snapshot.snapshot.clone(),
638 Worktree::Remote(worktree) => worktree.snapshot.clone(),
639 }
640 }
641
642 pub fn scan_id(&self) -> usize {
643 match self {
644 Worktree::Local(worktree) => worktree.snapshot.scan_id,
645 Worktree::Remote(worktree) => worktree.snapshot.scan_id,
646 }
647 }
648
649 pub fn metadata_proto(&self) -> proto::WorktreeMetadata {
650 proto::WorktreeMetadata {
651 id: self.id().to_proto(),
652 root_name: self.root_name().to_proto(),
653 visible: self.is_visible(),
654 abs_path: self.abs_path().to_string_lossy().into_owned(),
655 }
656 }
657
658 pub fn completed_scan_id(&self) -> usize {
659 match self {
660 Worktree::Local(worktree) => worktree.snapshot.completed_scan_id,
661 Worktree::Remote(worktree) => worktree.snapshot.completed_scan_id,
662 }
663 }
664
665 pub fn is_visible(&self) -> bool {
666 match self {
667 Worktree::Local(worktree) => worktree.visible,
668 Worktree::Remote(worktree) => worktree.visible,
669 }
670 }
671
672 pub fn replica_id(&self) -> ReplicaId {
673 match self {
674 Worktree::Local(_) => ReplicaId::LOCAL,
675 Worktree::Remote(worktree) => worktree.replica_id,
676 }
677 }
678
679 pub fn abs_path(&self) -> Arc<Path> {
680 match self {
681 Worktree::Local(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
682 Worktree::Remote(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
683 }
684 }
685
686 pub fn root_file(&self, cx: &Context<Self>) -> Option<Arc<File>> {
687 let entry = self.root_entry()?;
688 Some(File::for_entry(entry.clone(), cx.entity()))
689 }
690
691 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
692 where
693 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
694 Fut: 'static + Send + Future<Output = bool>,
695 {
696 match self {
697 Worktree::Local(this) => this.observe_updates(project_id, cx, callback),
698 Worktree::Remote(this) => this.observe_updates(project_id, cx, callback),
699 }
700 }
701
702 pub fn stop_observing_updates(&mut self) {
703 match self {
704 Worktree::Local(this) => {
705 this.update_observer.take();
706 }
707 Worktree::Remote(this) => {
708 this.update_observer.take();
709 }
710 }
711 }
712
713 #[cfg(any(test, feature = "test-support"))]
714 pub fn has_update_observer(&self) -> bool {
715 match self {
716 Worktree::Local(this) => this.update_observer.is_some(),
717 Worktree::Remote(this) => this.update_observer.is_some(),
718 }
719 }
720
721 pub fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
722 match self {
723 Worktree::Local(this) => this.load_file(path, cx),
724 Worktree::Remote(_) => {
725 Task::ready(Err(anyhow!("remote worktrees can't yet load files")))
726 }
727 }
728 }
729
730 pub fn load_binary_file(
731 &self,
732 path: &RelPath,
733 cx: &Context<Worktree>,
734 ) -> Task<Result<LoadedBinaryFile>> {
735 match self {
736 Worktree::Local(this) => this.load_binary_file(path, cx),
737 Worktree::Remote(_) => {
738 Task::ready(Err(anyhow!("remote worktrees can't yet load binary files")))
739 }
740 }
741 }
742
743 pub fn write_file(
744 &self,
745 path: Arc<RelPath>,
746 text: Rope,
747 line_ending: LineEnding,
748 encoding: &'static Encoding,
749 has_bom: bool,
750 cx: &Context<Worktree>,
751 ) -> Task<Result<Arc<File>>> {
752 match self {
753 Worktree::Local(this) => {
754 this.write_file(path, text, line_ending, encoding, has_bom, cx)
755 }
756 Worktree::Remote(_) => {
757 Task::ready(Err(anyhow!("remote worktree can't yet write files")))
758 }
759 }
760 }
761
762 pub fn create_entry(
763 &mut self,
764 path: Arc<RelPath>,
765 is_directory: bool,
766 content: Option<Vec<u8>>,
767 cx: &Context<Worktree>,
768 ) -> Task<Result<CreatedEntry>> {
769 let worktree_id = self.id();
770 match self {
771 Worktree::Local(this) => this.create_entry(path, is_directory, content, cx),
772 Worktree::Remote(this) => {
773 let project_id = this.project_id;
774 let request = this.client.request(proto::CreateProjectEntry {
775 worktree_id: worktree_id.to_proto(),
776 project_id,
777 path: path.as_ref().to_proto(),
778 content,
779 is_directory,
780 });
781 cx.spawn(async move |this, cx| {
782 let response = request.await?;
783 match response.entry {
784 Some(entry) => this
785 .update(cx, |worktree, cx| {
786 worktree.as_remote_mut().unwrap().insert_entry(
787 entry,
788 response.worktree_scan_id as usize,
789 cx,
790 )
791 })?
792 .await
793 .map(CreatedEntry::Included),
794 None => {
795 let abs_path =
796 this.read_with(cx, |worktree, _| worktree.absolutize(&path))?;
797 Ok(CreatedEntry::Excluded { abs_path })
798 }
799 }
800 })
801 }
802 }
803 }
804
805 pub fn delete_entry(
806 &mut self,
807 entry_id: ProjectEntryId,
808 trash: bool,
809 cx: &mut Context<Worktree>,
810 ) -> Option<Task<Result<()>>> {
811 let task = match self {
812 Worktree::Local(this) => this.delete_entry(entry_id, trash, cx),
813 Worktree::Remote(this) => this.delete_entry(entry_id, trash, cx),
814 }?;
815
816 let entry = match &*self {
817 Worktree::Local(this) => this.entry_for_id(entry_id),
818 Worktree::Remote(this) => this.entry_for_id(entry_id),
819 }?;
820
821 let mut ids = vec![entry_id];
822 let path = &*entry.path;
823
824 self.get_children_ids_recursive(path, &mut ids);
825
826 for id in ids {
827 cx.emit(Event::DeletedEntry(id));
828 }
829 Some(task)
830 }
831
832 fn get_children_ids_recursive(&self, path: &RelPath, ids: &mut Vec<ProjectEntryId>) {
833 let children_iter = self.child_entries(path);
834 for child in children_iter {
835 ids.push(child.id);
836 self.get_children_ids_recursive(&child.path, ids);
837 }
838 }
839
840 // pub fn rename_entry(
841 // &mut self,
842 // entry_id: ProjectEntryId,
843 // new_path: Arc<RelPath>,
844 // cx: &Context<Self>,
845 // ) -> Task<Result<CreatedEntry>> {
846 // match self {
847 // Worktree::Local(this) => this.rename_entry(entry_id, new_path, cx),
848 // Worktree::Remote(this) => this.rename_entry(entry_id, new_path, cx),
849 // }
850 // }
851
852 pub fn copy_external_entries(
853 &mut self,
854 target_directory: Arc<RelPath>,
855 paths: Vec<Arc<Path>>,
856 fs: Arc<dyn Fs>,
857 cx: &Context<Worktree>,
858 ) -> Task<Result<Vec<ProjectEntryId>>> {
859 match self {
860 Worktree::Local(this) => this.copy_external_entries(target_directory, paths, cx),
861 Worktree::Remote(this) => this.copy_external_entries(target_directory, paths, fs, cx),
862 }
863 }
864
865 pub fn expand_entry(
866 &mut self,
867 entry_id: ProjectEntryId,
868 cx: &Context<Worktree>,
869 ) -> Option<Task<Result<()>>> {
870 match self {
871 Worktree::Local(this) => this.expand_entry(entry_id, cx),
872 Worktree::Remote(this) => {
873 let response = this.client.request(proto::ExpandProjectEntry {
874 project_id: this.project_id,
875 entry_id: entry_id.to_proto(),
876 });
877 Some(cx.spawn(async move |this, cx| {
878 let response = response.await?;
879 this.update(cx, |this, _| {
880 this.as_remote_mut()
881 .unwrap()
882 .wait_for_snapshot(response.worktree_scan_id as usize)
883 })?
884 .await?;
885 Ok(())
886 }))
887 }
888 }
889 }
890
891 pub fn expand_all_for_entry(
892 &mut self,
893 entry_id: ProjectEntryId,
894 cx: &Context<Worktree>,
895 ) -> Option<Task<Result<()>>> {
896 match self {
897 Worktree::Local(this) => this.expand_all_for_entry(entry_id, cx),
898 Worktree::Remote(this) => {
899 let response = this.client.request(proto::ExpandAllForProjectEntry {
900 project_id: this.project_id,
901 entry_id: entry_id.to_proto(),
902 });
903 Some(cx.spawn(async move |this, cx| {
904 let response = response.await?;
905 this.update(cx, |this, _| {
906 this.as_remote_mut()
907 .unwrap()
908 .wait_for_snapshot(response.worktree_scan_id as usize)
909 })?
910 .await?;
911 Ok(())
912 }))
913 }
914 }
915 }
916
917 pub async fn handle_create_entry(
918 this: Entity<Self>,
919 request: proto::CreateProjectEntry,
920 mut cx: AsyncApp,
921 ) -> Result<proto::ProjectEntryResponse> {
922 let (scan_id, entry) = this.update(&mut cx, |this, cx| {
923 anyhow::Ok((
924 this.scan_id(),
925 this.create_entry(
926 RelPath::from_proto(&request.path).with_context(|| {
927 format!("received invalid relative path {:?}", request.path)
928 })?,
929 request.is_directory,
930 request.content,
931 cx,
932 ),
933 ))
934 })??;
935 Ok(proto::ProjectEntryResponse {
936 entry: match &entry.await? {
937 CreatedEntry::Included(entry) => Some(entry.into()),
938 CreatedEntry::Excluded { .. } => None,
939 },
940 worktree_scan_id: scan_id as u64,
941 })
942 }
943
944 pub async fn handle_delete_entry(
945 this: Entity<Self>,
946 request: proto::DeleteProjectEntry,
947 mut cx: AsyncApp,
948 ) -> Result<proto::ProjectEntryResponse> {
949 let (scan_id, task) = this.update(&mut cx, |this, cx| {
950 (
951 this.scan_id(),
952 this.delete_entry(
953 ProjectEntryId::from_proto(request.entry_id),
954 request.use_trash,
955 cx,
956 ),
957 )
958 })?;
959 task.context("invalid entry")?.await?;
960 Ok(proto::ProjectEntryResponse {
961 entry: None,
962 worktree_scan_id: scan_id as u64,
963 })
964 }
965
966 pub async fn handle_expand_entry(
967 this: Entity<Self>,
968 request: proto::ExpandProjectEntry,
969 mut cx: AsyncApp,
970 ) -> Result<proto::ExpandProjectEntryResponse> {
971 let task = this.update(&mut cx, |this, cx| {
972 this.expand_entry(ProjectEntryId::from_proto(request.entry_id), cx)
973 })?;
974 task.context("no such entry")?.await?;
975 let scan_id = this.read_with(&cx, |this, _| this.scan_id())?;
976 Ok(proto::ExpandProjectEntryResponse {
977 worktree_scan_id: scan_id as u64,
978 })
979 }
980
981 pub async fn handle_expand_all_for_entry(
982 this: Entity<Self>,
983 request: proto::ExpandAllForProjectEntry,
984 mut cx: AsyncApp,
985 ) -> Result<proto::ExpandAllForProjectEntryResponse> {
986 let task = this.update(&mut cx, |this, cx| {
987 this.expand_all_for_entry(ProjectEntryId::from_proto(request.entry_id), cx)
988 })?;
989 task.context("no such entry")?.await?;
990 let scan_id = this.read_with(&cx, |this, _| this.scan_id())?;
991 Ok(proto::ExpandAllForProjectEntryResponse {
992 worktree_scan_id: scan_id as u64,
993 })
994 }
995
996 pub fn is_single_file(&self) -> bool {
997 self.root_dir().is_none()
998 }
999
1000 /// For visible worktrees, returns the path with the worktree name as the first component.
1001 /// Otherwise, returns an absolute path.
1002 pub fn full_path(&self, worktree_relative_path: &RelPath) -> PathBuf {
1003 if self.is_visible() {
1004 self.root_name()
1005 .join(worktree_relative_path)
1006 .display(self.path_style)
1007 .to_string()
1008 .into()
1009 } else {
1010 let full_path = self.abs_path();
1011 let mut full_path_string = if self.is_local()
1012 && let Ok(stripped) = full_path.strip_prefix(home_dir())
1013 {
1014 self.path_style
1015 .join("~", &*stripped.to_string_lossy())
1016 .unwrap()
1017 } else {
1018 full_path.to_string_lossy().into_owned()
1019 };
1020
1021 if worktree_relative_path.components().next().is_some() {
1022 full_path_string.push_str(self.path_style.primary_separator());
1023 full_path_string.push_str(&worktree_relative_path.display(self.path_style));
1024 }
1025
1026 full_path_string.into()
1027 }
1028 }
1029}
1030
1031impl LocalWorktree {
1032 pub fn fs(&self) -> &Arc<dyn Fs> {
1033 &self.fs
1034 }
1035
1036 pub fn is_path_private(&self, path: &RelPath) -> bool {
1037 !self.share_private_files && self.settings.is_path_private(path)
1038 }
1039
1040 pub fn fs_is_case_sensitive(&self) -> bool {
1041 self.fs_case_sensitive
1042 }
1043
1044 fn restart_background_scanners(&mut self, cx: &Context<Worktree>) {
1045 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
1046 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
1047 self.scan_requests_tx = scan_requests_tx;
1048 self.path_prefixes_to_scan_tx = path_prefixes_to_scan_tx;
1049
1050 self.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
1051 let always_included_entries = mem::take(&mut self.snapshot.always_included_entries);
1052 log::debug!(
1053 "refreshing entries for the following always included paths: {:?}",
1054 always_included_entries
1055 );
1056
1057 // Cleans up old always included entries to ensure they get updated properly. Otherwise,
1058 // nested always included entries may not get updated and will result in out-of-date info.
1059 self.refresh_entries_for_paths(always_included_entries);
1060 }
1061
1062 fn start_background_scanner(
1063 &mut self,
1064 scan_requests_rx: channel::Receiver<ScanRequest>,
1065 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
1066 cx: &Context<Worktree>,
1067 ) {
1068 let snapshot = self.snapshot();
1069 let share_private_files = self.share_private_files;
1070 let next_entry_id = self.next_entry_id.clone();
1071 let fs = self.fs.clone();
1072 let scanning_enabled = self.scanning_enabled;
1073 let settings = self.settings.clone();
1074 let (scan_states_tx, mut scan_states_rx) = mpsc::unbounded();
1075 let background_scanner = cx.background_spawn({
1076 let abs_path = snapshot.abs_path.as_path().to_path_buf();
1077 let background = cx.background_executor().clone();
1078 async move {
1079 let (events, watcher) = if scanning_enabled {
1080 fs.watch(&abs_path, FS_WATCH_LATENCY).await
1081 } else {
1082 (Box::pin(stream::pending()) as _, Arc::new(NullWatcher) as _)
1083 };
1084 let fs_case_sensitive = fs.is_case_sensitive().await.unwrap_or_else(|e| {
1085 log::error!("Failed to determine whether filesystem is case sensitive: {e:#}");
1086 true
1087 });
1088
1089 let mut scanner = BackgroundScanner {
1090 fs,
1091 fs_case_sensitive,
1092 status_updates_tx: scan_states_tx,
1093 executor: background,
1094 scan_requests_rx,
1095 path_prefixes_to_scan_rx,
1096 next_entry_id,
1097 state: async_lock::Mutex::new(BackgroundScannerState {
1098 prev_snapshot: snapshot.snapshot.clone(),
1099 snapshot,
1100 scanned_dirs: Default::default(),
1101 path_prefixes_to_scan: Default::default(),
1102 paths_to_scan: Default::default(),
1103 removed_entries: Default::default(),
1104 changed_paths: Default::default(),
1105 }),
1106 phase: BackgroundScannerPhase::InitialScan,
1107 share_private_files,
1108 scanning_enabled,
1109 settings,
1110 watcher,
1111 };
1112
1113 scanner
1114 .run(Box::pin(events.map(|events| events.into_iter().collect())))
1115 .await;
1116 }
1117 });
1118 let scan_state_updater = cx.spawn(async move |this, cx| {
1119 while let Some((state, this)) = scan_states_rx.next().await.zip(this.upgrade()) {
1120 this.update(cx, |this, cx| {
1121 let this = this.as_local_mut().unwrap();
1122 match state {
1123 ScanState::Started => {
1124 *this.is_scanning.0.borrow_mut() = true;
1125 }
1126 ScanState::Updated {
1127 snapshot,
1128 changes,
1129 barrier,
1130 scanning,
1131 } => {
1132 *this.is_scanning.0.borrow_mut() = scanning;
1133 this.set_snapshot(snapshot, changes, cx);
1134 drop(barrier);
1135 }
1136 ScanState::RootUpdated { new_path } => {
1137 this.update_abs_path_and_refresh(new_path, cx);
1138 }
1139 }
1140 })
1141 .ok();
1142 }
1143 });
1144 self._background_scanner_tasks = vec![background_scanner, scan_state_updater];
1145 *self.is_scanning.0.borrow_mut() = true;
1146 }
1147
1148 fn set_snapshot(
1149 &mut self,
1150 mut new_snapshot: LocalSnapshot,
1151 entry_changes: UpdatedEntriesSet,
1152 cx: &mut Context<Worktree>,
1153 ) {
1154 let repo_changes = self.changed_repos(&self.snapshot, &mut new_snapshot);
1155 self.snapshot = new_snapshot;
1156
1157 if let Some(share) = self.update_observer.as_mut() {
1158 share
1159 .snapshots_tx
1160 .unbounded_send((self.snapshot.clone(), entry_changes.clone()))
1161 .ok();
1162 }
1163
1164 if !entry_changes.is_empty() {
1165 cx.emit(Event::UpdatedEntries(entry_changes));
1166 }
1167 if !repo_changes.is_empty() {
1168 cx.emit(Event::UpdatedGitRepositories(repo_changes));
1169 }
1170 }
1171
1172 fn changed_repos(
1173 &self,
1174 old_snapshot: &LocalSnapshot,
1175 new_snapshot: &mut LocalSnapshot,
1176 ) -> UpdatedGitRepositoriesSet {
1177 let mut changes = Vec::new();
1178 let mut old_repos = old_snapshot.git_repositories.iter().peekable();
1179 let new_repos = new_snapshot.git_repositories.clone();
1180 let mut new_repos = new_repos.iter().peekable();
1181
1182 loop {
1183 match (new_repos.peek().map(clone), old_repos.peek().map(clone)) {
1184 (Some((new_entry_id, new_repo)), Some((old_entry_id, old_repo))) => {
1185 match Ord::cmp(&new_entry_id, &old_entry_id) {
1186 Ordering::Less => {
1187 changes.push(UpdatedGitRepository {
1188 work_directory_id: new_entry_id,
1189 old_work_directory_abs_path: None,
1190 new_work_directory_abs_path: Some(
1191 new_repo.work_directory_abs_path.clone(),
1192 ),
1193 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1194 repository_dir_abs_path: Some(
1195 new_repo.repository_dir_abs_path.clone(),
1196 ),
1197 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1198 });
1199 new_repos.next();
1200 }
1201 Ordering::Equal => {
1202 if new_repo.git_dir_scan_id != old_repo.git_dir_scan_id
1203 || new_repo.work_directory_abs_path
1204 != old_repo.work_directory_abs_path
1205 {
1206 changes.push(UpdatedGitRepository {
1207 work_directory_id: new_entry_id,
1208 old_work_directory_abs_path: Some(
1209 old_repo.work_directory_abs_path.clone(),
1210 ),
1211 new_work_directory_abs_path: Some(
1212 new_repo.work_directory_abs_path.clone(),
1213 ),
1214 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1215 repository_dir_abs_path: Some(
1216 new_repo.repository_dir_abs_path.clone(),
1217 ),
1218 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1219 });
1220 }
1221 new_repos.next();
1222 old_repos.next();
1223 }
1224 Ordering::Greater => {
1225 changes.push(UpdatedGitRepository {
1226 work_directory_id: old_entry_id,
1227 old_work_directory_abs_path: Some(
1228 old_repo.work_directory_abs_path.clone(),
1229 ),
1230 new_work_directory_abs_path: None,
1231 dot_git_abs_path: None,
1232 repository_dir_abs_path: None,
1233 common_dir_abs_path: None,
1234 });
1235 old_repos.next();
1236 }
1237 }
1238 }
1239 (Some((entry_id, repo)), None) => {
1240 changes.push(UpdatedGitRepository {
1241 work_directory_id: entry_id,
1242 old_work_directory_abs_path: None,
1243 new_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1244 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1245 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1246 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1247 });
1248 new_repos.next();
1249 }
1250 (None, Some((entry_id, repo))) => {
1251 changes.push(UpdatedGitRepository {
1252 work_directory_id: entry_id,
1253 old_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1254 new_work_directory_abs_path: None,
1255 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1256 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1257 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1258 });
1259 old_repos.next();
1260 }
1261 (None, None) => break,
1262 }
1263 }
1264
1265 fn clone<T: Clone, U: Clone>(value: &(&T, &U)) -> (T, U) {
1266 (value.0.clone(), value.1.clone())
1267 }
1268
1269 changes.into()
1270 }
1271
1272 pub fn scan_complete(&self) -> impl Future<Output = ()> + use<> {
1273 let mut is_scanning_rx = self.is_scanning.1.clone();
1274 async move {
1275 let mut is_scanning = *is_scanning_rx.borrow();
1276 while is_scanning {
1277 if let Some(value) = is_scanning_rx.recv().await {
1278 is_scanning = value;
1279 } else {
1280 break;
1281 }
1282 }
1283 }
1284 }
1285
1286 pub fn snapshot(&self) -> LocalSnapshot {
1287 self.snapshot.clone()
1288 }
1289
1290 pub fn settings(&self) -> WorktreeSettings {
1291 self.settings.clone()
1292 }
1293
1294 fn load_binary_file(
1295 &self,
1296 path: &RelPath,
1297 cx: &Context<Worktree>,
1298 ) -> Task<Result<LoadedBinaryFile>> {
1299 let path = Arc::from(path);
1300 let abs_path = self.absolutize(&path);
1301 let fs = self.fs.clone();
1302 let entry = self.refresh_entry(path.clone(), None, cx);
1303 let is_private = self.is_path_private(&path);
1304
1305 let worktree = cx.weak_entity();
1306 cx.background_spawn(async move {
1307 let content = fs.load_bytes(&abs_path).await?;
1308
1309 let worktree = worktree.upgrade().context("worktree was dropped")?;
1310 let file = match entry.await? {
1311 Some(entry) => File::for_entry(entry, worktree),
1312 None => {
1313 let metadata = fs
1314 .metadata(&abs_path)
1315 .await
1316 .with_context(|| {
1317 format!("Loading metadata for excluded file {abs_path:?}")
1318 })?
1319 .with_context(|| {
1320 format!("Excluded file {abs_path:?} got removed during loading")
1321 })?;
1322 Arc::new(File {
1323 entry_id: None,
1324 worktree,
1325 path,
1326 disk_state: DiskState::Present {
1327 mtime: metadata.mtime,
1328 },
1329 is_local: true,
1330 is_private,
1331 })
1332 }
1333 };
1334
1335 Ok(LoadedBinaryFile { file, content })
1336 })
1337 }
1338
1339 fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
1340 let path = Arc::from(path);
1341 let abs_path = self.absolutize(&path);
1342 let fs = self.fs.clone();
1343 let entry = self.refresh_entry(path.clone(), None, cx);
1344 let is_private = self.is_path_private(path.as_ref());
1345
1346 let this = cx.weak_entity();
1347 cx.background_spawn(async move {
1348 // WARN: Temporary workaround for #27283.
1349 // We are not efficient with our memory usage per file, and use in excess of 64GB for a 10GB file
1350 // Therefore, as a temporary workaround to prevent system freezes, we just bail before opening a file
1351 // if it is too large
1352 // 5GB seems to be more reasonable, peaking at ~16GB, while 6GB jumps up to >24GB which seems like a
1353 // reasonable limit
1354 {
1355 const FILE_SIZE_MAX: u64 = 6 * 1024 * 1024 * 1024; // 6GB
1356 if let Ok(Some(metadata)) = fs.metadata(&abs_path).await
1357 && metadata.len >= FILE_SIZE_MAX
1358 {
1359 anyhow::bail!("File is too large to load");
1360 }
1361 }
1362 let (text, encoding, has_bom) = decode_file_text(fs.as_ref(), &abs_path).await?;
1363
1364 let worktree = this.upgrade().context("worktree was dropped")?;
1365 let file = match entry.await? {
1366 Some(entry) => File::for_entry(entry, worktree),
1367 None => {
1368 let metadata = fs
1369 .metadata(&abs_path)
1370 .await
1371 .with_context(|| {
1372 format!("Loading metadata for excluded file {abs_path:?}")
1373 })?
1374 .with_context(|| {
1375 format!("Excluded file {abs_path:?} got removed during loading")
1376 })?;
1377 Arc::new(File {
1378 entry_id: None,
1379 worktree,
1380 path,
1381 disk_state: DiskState::Present {
1382 mtime: metadata.mtime,
1383 },
1384 is_local: true,
1385 is_private,
1386 })
1387 }
1388 };
1389
1390 Ok(LoadedFile {
1391 file,
1392 text,
1393 encoding,
1394 has_bom,
1395 })
1396 })
1397 }
1398
1399 /// Find the lowest path in the worktree's datastructures that is an ancestor
1400 fn lowest_ancestor(&self, path: &RelPath) -> Arc<RelPath> {
1401 let mut lowest_ancestor = None;
1402 for path in path.ancestors() {
1403 if self.entry_for_path(path).is_some() {
1404 lowest_ancestor = Some(path.into());
1405 break;
1406 }
1407 }
1408
1409 lowest_ancestor.unwrap_or_else(|| RelPath::empty().into())
1410 }
1411
1412 fn create_entry(
1413 &self,
1414 path: Arc<RelPath>,
1415 is_dir: bool,
1416 content: Option<Vec<u8>>,
1417 cx: &Context<Worktree>,
1418 ) -> Task<Result<CreatedEntry>> {
1419 let abs_path = self.absolutize(&path);
1420 let path_excluded = self.settings.is_path_excluded(&path);
1421 let fs = self.fs.clone();
1422 let task_abs_path = abs_path.clone();
1423 let write = cx.background_spawn(async move {
1424 if is_dir {
1425 fs.create_dir(&task_abs_path)
1426 .await
1427 .with_context(|| format!("creating directory {task_abs_path:?}"))
1428 } else {
1429 fs.write(&task_abs_path, content.as_deref().unwrap_or(&[]))
1430 .await
1431 .with_context(|| format!("creating file {task_abs_path:?}"))
1432 }
1433 });
1434
1435 let lowest_ancestor = self.lowest_ancestor(&path);
1436 cx.spawn(async move |this, cx| {
1437 write.await?;
1438 if path_excluded {
1439 return Ok(CreatedEntry::Excluded { abs_path });
1440 }
1441
1442 let (result, refreshes) = this.update(cx, |this, cx| {
1443 let mut refreshes = Vec::new();
1444 let refresh_paths = path.strip_prefix(&lowest_ancestor).unwrap();
1445 for refresh_path in refresh_paths.ancestors() {
1446 if refresh_path == RelPath::empty() {
1447 continue;
1448 }
1449 let refresh_full_path = lowest_ancestor.join(refresh_path);
1450
1451 refreshes.push(this.as_local_mut().unwrap().refresh_entry(
1452 refresh_full_path,
1453 None,
1454 cx,
1455 ));
1456 }
1457 (
1458 this.as_local_mut().unwrap().refresh_entry(path, None, cx),
1459 refreshes,
1460 )
1461 })?;
1462 for refresh in refreshes {
1463 refresh.await.log_err();
1464 }
1465
1466 Ok(result
1467 .await?
1468 .map(CreatedEntry::Included)
1469 .unwrap_or_else(|| CreatedEntry::Excluded { abs_path }))
1470 })
1471 }
1472
1473 fn write_file(
1474 &self,
1475 path: Arc<RelPath>,
1476 text: Rope,
1477 line_ending: LineEnding,
1478 encoding: &'static Encoding,
1479 has_bom: bool,
1480 cx: &Context<Worktree>,
1481 ) -> Task<Result<Arc<File>>> {
1482 let fs = self.fs.clone();
1483 let is_private = self.is_path_private(&path);
1484 let abs_path = self.absolutize(&path);
1485
1486 let write = cx.background_spawn({
1487 let fs = fs.clone();
1488 let abs_path = abs_path.clone();
1489 async move {
1490 // For UTF-8, use the optimized `fs.save` which writes Rope chunks directly to disk
1491 // without allocating a contiguous string.
1492 if encoding == encoding_rs::UTF_8 && !has_bom {
1493 return fs.save(&abs_path, &text, line_ending).await;
1494 }
1495
1496 // For legacy encodings (e.g. Shift-JIS), we fall back to converting the entire Rope
1497 // to a String/Bytes in memory before writing.
1498 //
1499 // Note: This is inefficient for very large files compared to the streaming approach above,
1500 // but supporting streaming writes for arbitrary encodings would require a significant
1501 // refactor of the `fs` crate to expose a Writer interface.
1502 let text_string = text.to_string();
1503 let normalized_text = match line_ending {
1504 LineEnding::Unix => text_string,
1505 LineEnding::Windows => text_string.replace('\n', "\r\n"),
1506 };
1507
1508 // Create the byte vector manually for UTF-16 encodings because encoding_rs encodes to UTF-8 by default (per WHATWG standards),
1509 // which is not what we want for saving files.
1510 let bytes = if encoding == encoding_rs::UTF_16BE {
1511 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1512 if has_bom {
1513 data.extend_from_slice(&[0xFE, 0xFF]); // BOM
1514 }
1515 let utf16be_bytes =
1516 normalized_text.encode_utf16().flat_map(|u| u.to_be_bytes());
1517 data.extend(utf16be_bytes);
1518 data.into()
1519 } else if encoding == encoding_rs::UTF_16LE {
1520 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1521 if has_bom {
1522 data.extend_from_slice(&[0xFF, 0xFE]); // BOM
1523 }
1524 let utf16le_bytes =
1525 normalized_text.encode_utf16().flat_map(|u| u.to_le_bytes());
1526 data.extend(utf16le_bytes);
1527 data.into()
1528 } else {
1529 // For other encodings (Shift-JIS, UTF-8 with BOM, etc.), delegate to encoding_rs.
1530 let bom_bytes = if has_bom {
1531 if encoding == encoding_rs::UTF_8 {
1532 vec![0xEF, 0xBB, 0xBF]
1533 } else {
1534 vec![]
1535 }
1536 } else {
1537 vec![]
1538 };
1539 let (cow, _, _) = encoding.encode(&normalized_text);
1540 if !bom_bytes.is_empty() {
1541 let mut bytes = bom_bytes;
1542 bytes.extend_from_slice(&cow);
1543 bytes.into()
1544 } else {
1545 cow
1546 }
1547 };
1548
1549 fs.write(&abs_path, &bytes).await
1550 }
1551 });
1552
1553 cx.spawn(async move |this, cx| {
1554 write.await?;
1555 let entry = this
1556 .update(cx, |this, cx| {
1557 this.as_local_mut()
1558 .unwrap()
1559 .refresh_entry(path.clone(), None, cx)
1560 })?
1561 .await?;
1562 let worktree = this.upgrade().context("worktree dropped")?;
1563 if let Some(entry) = entry {
1564 Ok(File::for_entry(entry, worktree))
1565 } else {
1566 let metadata = fs
1567 .metadata(&abs_path)
1568 .await
1569 .with_context(|| {
1570 format!("Fetching metadata after saving the excluded buffer {abs_path:?}")
1571 })?
1572 .with_context(|| {
1573 format!("Excluded buffer {path:?} got removed during saving")
1574 })?;
1575 Ok(Arc::new(File {
1576 worktree,
1577 path,
1578 disk_state: DiskState::Present {
1579 mtime: metadata.mtime,
1580 },
1581 entry_id: None,
1582 is_local: true,
1583 is_private,
1584 }))
1585 }
1586 })
1587 }
1588
1589 fn delete_entry(
1590 &self,
1591 entry_id: ProjectEntryId,
1592 trash: bool,
1593 cx: &Context<Worktree>,
1594 ) -> Option<Task<Result<()>>> {
1595 let entry = self.entry_for_id(entry_id)?.clone();
1596 let abs_path = self.absolutize(&entry.path);
1597 let fs = self.fs.clone();
1598
1599 let delete = cx.background_spawn(async move {
1600 if entry.is_file() {
1601 if trash {
1602 fs.trash_file(&abs_path, Default::default()).await?;
1603 } else {
1604 fs.remove_file(&abs_path, Default::default()).await?;
1605 }
1606 } else if trash {
1607 fs.trash_dir(
1608 &abs_path,
1609 RemoveOptions {
1610 recursive: true,
1611 ignore_if_not_exists: false,
1612 },
1613 )
1614 .await?;
1615 } else {
1616 fs.remove_dir(
1617 &abs_path,
1618 RemoveOptions {
1619 recursive: true,
1620 ignore_if_not_exists: false,
1621 },
1622 )
1623 .await?;
1624 }
1625 anyhow::Ok(entry.path)
1626 });
1627
1628 Some(cx.spawn(async move |this, cx| {
1629 let path = delete.await?;
1630 this.update(cx, |this, _| {
1631 this.as_local_mut()
1632 .unwrap()
1633 .refresh_entries_for_paths(vec![path])
1634 })?
1635 .recv()
1636 .await;
1637 Ok(())
1638 }))
1639 }
1640
1641 pub fn copy_external_entries(
1642 &self,
1643 target_directory: Arc<RelPath>,
1644 paths: Vec<Arc<Path>>,
1645 cx: &Context<Worktree>,
1646 ) -> Task<Result<Vec<ProjectEntryId>>> {
1647 let target_directory = self.absolutize(&target_directory);
1648 let worktree_path = self.abs_path().clone();
1649 let fs = self.fs.clone();
1650 let paths = paths
1651 .into_iter()
1652 .filter_map(|source| {
1653 let file_name = source.file_name()?;
1654 let mut target = target_directory.clone();
1655 target.push(file_name);
1656
1657 // Do not allow copying the same file to itself.
1658 if source.as_ref() != target.as_path() {
1659 Some((source, target))
1660 } else {
1661 None
1662 }
1663 })
1664 .collect::<Vec<_>>();
1665
1666 let paths_to_refresh = paths
1667 .iter()
1668 .filter_map(|(_, target)| {
1669 RelPath::new(
1670 target.strip_prefix(&worktree_path).ok()?,
1671 PathStyle::local(),
1672 )
1673 .ok()
1674 .map(|path| path.into_arc())
1675 })
1676 .collect::<Vec<_>>();
1677
1678 cx.spawn(async move |this, cx| {
1679 cx.background_spawn(async move {
1680 for (source, target) in paths {
1681 copy_recursive(
1682 fs.as_ref(),
1683 &source,
1684 &target,
1685 fs::CopyOptions {
1686 overwrite: true,
1687 ..Default::default()
1688 },
1689 )
1690 .await
1691 .with_context(|| {
1692 format!("Failed to copy file from {source:?} to {target:?}")
1693 })?;
1694 }
1695 anyhow::Ok(())
1696 })
1697 .await
1698 .log_err();
1699 let mut refresh = cx.read_entity(
1700 &this.upgrade().with_context(|| "Dropped worktree")?,
1701 |this, _| {
1702 anyhow::Ok::<postage::barrier::Receiver>(
1703 this.as_local()
1704 .with_context(|| "Worktree is not local")?
1705 .refresh_entries_for_paths(paths_to_refresh.clone()),
1706 )
1707 },
1708 )??;
1709
1710 cx.background_spawn(async move {
1711 refresh.next().await;
1712 anyhow::Ok(())
1713 })
1714 .await
1715 .log_err();
1716
1717 let this = this.upgrade().with_context(|| "Dropped worktree")?;
1718 cx.read_entity(&this, |this, _| {
1719 paths_to_refresh
1720 .iter()
1721 .filter_map(|path| Some(this.entry_for_path(path)?.id))
1722 .collect()
1723 })
1724 })
1725 }
1726
1727 fn expand_entry(
1728 &self,
1729 entry_id: ProjectEntryId,
1730 cx: &Context<Worktree>,
1731 ) -> Option<Task<Result<()>>> {
1732 let path = self.entry_for_id(entry_id)?.path.clone();
1733 let mut refresh = self.refresh_entries_for_paths(vec![path]);
1734 Some(cx.background_spawn(async move {
1735 refresh.next().await;
1736 Ok(())
1737 }))
1738 }
1739
1740 fn expand_all_for_entry(
1741 &self,
1742 entry_id: ProjectEntryId,
1743 cx: &Context<Worktree>,
1744 ) -> Option<Task<Result<()>>> {
1745 let path = self.entry_for_id(entry_id).unwrap().path.clone();
1746 let mut rx = self.add_path_prefix_to_scan(path);
1747 Some(cx.background_spawn(async move {
1748 rx.next().await;
1749 Ok(())
1750 }))
1751 }
1752
1753 pub fn refresh_entries_for_paths(&self, paths: Vec<Arc<RelPath>>) -> barrier::Receiver {
1754 let (tx, rx) = barrier::channel();
1755 self.scan_requests_tx
1756 .try_send(ScanRequest {
1757 relative_paths: paths,
1758 done: smallvec![tx],
1759 })
1760 .ok();
1761 rx
1762 }
1763
1764 #[cfg(feature = "test-support")]
1765 pub fn manually_refresh_entries_for_paths(
1766 &self,
1767 paths: Vec<Arc<RelPath>>,
1768 ) -> barrier::Receiver {
1769 self.refresh_entries_for_paths(paths)
1770 }
1771
1772 pub fn add_path_prefix_to_scan(&self, path_prefix: Arc<RelPath>) -> barrier::Receiver {
1773 let (tx, rx) = barrier::channel();
1774 self.path_prefixes_to_scan_tx
1775 .try_send(PathPrefixScanRequest {
1776 path: path_prefix,
1777 done: smallvec![tx],
1778 })
1779 .ok();
1780 rx
1781 }
1782
1783 pub fn refresh_entry(
1784 &self,
1785 path: Arc<RelPath>,
1786 old_path: Option<Arc<RelPath>>,
1787 cx: &Context<Worktree>,
1788 ) -> Task<Result<Option<Entry>>> {
1789 if self.settings.is_path_excluded(&path) {
1790 return Task::ready(Ok(None));
1791 }
1792 let paths = if let Some(old_path) = old_path.as_ref() {
1793 vec![old_path.clone(), path.clone()]
1794 } else {
1795 vec![path.clone()]
1796 };
1797 let t0 = Instant::now();
1798 let mut refresh = self.refresh_entries_for_paths(paths);
1799 // todo(lw): Hot foreground spawn
1800 cx.spawn(async move |this, cx| {
1801 refresh.recv().await;
1802 log::trace!("refreshed entry {path:?} in {:?}", t0.elapsed());
1803 let new_entry = this.read_with(cx, |this, _| {
1804 this.entry_for_path(&path).cloned().with_context(|| {
1805 format!("Could not find entry in worktree for {path:?} after refresh")
1806 })
1807 })??;
1808 Ok(Some(new_entry))
1809 })
1810 }
1811
1812 fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1813 where
1814 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1815 Fut: 'static + Send + Future<Output = bool>,
1816 {
1817 if let Some(observer) = self.update_observer.as_mut() {
1818 *observer.resume_updates.borrow_mut() = ();
1819 return;
1820 }
1821
1822 let (resume_updates_tx, mut resume_updates_rx) = watch::channel::<()>();
1823 let (snapshots_tx, mut snapshots_rx) =
1824 mpsc::unbounded::<(LocalSnapshot, UpdatedEntriesSet)>();
1825 snapshots_tx
1826 .unbounded_send((self.snapshot(), Arc::default()))
1827 .ok();
1828
1829 let worktree_id = cx.entity_id().as_u64();
1830 let _maintain_remote_snapshot = cx.background_spawn(async move {
1831 let mut is_first = true;
1832 while let Some((snapshot, entry_changes)) = snapshots_rx.next().await {
1833 let update = if is_first {
1834 is_first = false;
1835 snapshot.build_initial_update(project_id, worktree_id)
1836 } else {
1837 snapshot.build_update(project_id, worktree_id, entry_changes)
1838 };
1839
1840 for update in proto::split_worktree_update(update) {
1841 let _ = resume_updates_rx.try_recv();
1842 loop {
1843 let result = callback(update.clone());
1844 if result.await {
1845 break;
1846 } else {
1847 log::info!("waiting to resume updates");
1848 if resume_updates_rx.next().await.is_none() {
1849 return Some(());
1850 }
1851 }
1852 }
1853 }
1854 }
1855 Some(())
1856 });
1857
1858 self.update_observer = Some(UpdateObservationState {
1859 snapshots_tx,
1860 resume_updates: resume_updates_tx,
1861 _maintain_remote_snapshot,
1862 });
1863 }
1864
1865 pub fn share_private_files(&mut self, cx: &Context<Worktree>) {
1866 self.share_private_files = true;
1867 self.restart_background_scanners(cx);
1868 }
1869
1870 pub fn update_abs_path_and_refresh(
1871 &mut self,
1872 new_path: Arc<SanitizedPath>,
1873 cx: &Context<Worktree>,
1874 ) {
1875 self.snapshot.git_repositories = Default::default();
1876 self.snapshot.ignores_by_parent_abs_path = Default::default();
1877 let root_name = new_path
1878 .as_path()
1879 .file_name()
1880 .and_then(|f| f.to_str())
1881 .map_or(RelPath::empty().into(), |f| {
1882 RelPath::unix(f).unwrap().into()
1883 });
1884 self.snapshot.update_abs_path(new_path, root_name);
1885 self.restart_background_scanners(cx);
1886 }
1887}
1888
1889impl RemoteWorktree {
1890 pub fn project_id(&self) -> u64 {
1891 self.project_id
1892 }
1893
1894 pub fn client(&self) -> AnyProtoClient {
1895 self.client.clone()
1896 }
1897
1898 pub fn disconnected_from_host(&mut self) {
1899 self.updates_tx.take();
1900 self.snapshot_subscriptions.clear();
1901 self.disconnected = true;
1902 }
1903
1904 pub fn update_from_remote(&self, update: proto::UpdateWorktree) {
1905 if let Some(updates_tx) = &self.updates_tx {
1906 updates_tx
1907 .unbounded_send(update)
1908 .expect("consumer runs to completion");
1909 }
1910 }
1911
1912 fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1913 where
1914 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1915 Fut: 'static + Send + Future<Output = bool>,
1916 {
1917 let (tx, mut rx) = mpsc::unbounded();
1918 let initial_update = self
1919 .snapshot
1920 .build_initial_update(project_id, self.id().to_proto());
1921 self.update_observer = Some(tx);
1922 cx.spawn(async move |this, cx| {
1923 let mut update = initial_update;
1924 'outer: loop {
1925 // SSH projects use a special project ID of 0, and we need to
1926 // remap it to the correct one here.
1927 update.project_id = project_id;
1928
1929 for chunk in split_worktree_update(update) {
1930 if !callback(chunk).await {
1931 break 'outer;
1932 }
1933 }
1934
1935 if let Some(next_update) = rx.next().await {
1936 update = next_update;
1937 } else {
1938 break;
1939 }
1940 }
1941 this.update(cx, |this, _| {
1942 let this = this.as_remote_mut().unwrap();
1943 this.update_observer.take();
1944 })
1945 })
1946 .detach();
1947 }
1948
1949 fn observed_snapshot(&self, scan_id: usize) -> bool {
1950 self.completed_scan_id >= scan_id
1951 }
1952
1953 pub fn wait_for_snapshot(
1954 &mut self,
1955 scan_id: usize,
1956 ) -> impl Future<Output = Result<()>> + use<> {
1957 let (tx, rx) = oneshot::channel();
1958 if self.observed_snapshot(scan_id) {
1959 let _ = tx.send(());
1960 } else if self.disconnected {
1961 drop(tx);
1962 } else {
1963 match self
1964 .snapshot_subscriptions
1965 .binary_search_by_key(&scan_id, |probe| probe.0)
1966 {
1967 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
1968 }
1969 }
1970
1971 async move {
1972 rx.await?;
1973 Ok(())
1974 }
1975 }
1976
1977 pub fn insert_entry(
1978 &mut self,
1979 entry: proto::Entry,
1980 scan_id: usize,
1981 cx: &Context<Worktree>,
1982 ) -> Task<Result<Entry>> {
1983 let wait_for_snapshot = self.wait_for_snapshot(scan_id);
1984 cx.spawn(async move |this, cx| {
1985 wait_for_snapshot.await?;
1986 this.update(cx, |worktree, _| {
1987 let worktree = worktree.as_remote_mut().unwrap();
1988 let snapshot = &mut worktree.background_snapshot.lock().0;
1989 let entry = snapshot.insert_entry(entry, &worktree.file_scan_inclusions);
1990 worktree.snapshot = snapshot.clone();
1991 entry
1992 })?
1993 })
1994 }
1995
1996 fn delete_entry(
1997 &self,
1998 entry_id: ProjectEntryId,
1999 trash: bool,
2000 cx: &Context<Worktree>,
2001 ) -> Option<Task<Result<()>>> {
2002 let response = self.client.request(proto::DeleteProjectEntry {
2003 project_id: self.project_id,
2004 entry_id: entry_id.to_proto(),
2005 use_trash: trash,
2006 });
2007 Some(cx.spawn(async move |this, cx| {
2008 let response = response.await?;
2009 let scan_id = response.worktree_scan_id as usize;
2010
2011 this.update(cx, move |this, _| {
2012 this.as_remote_mut().unwrap().wait_for_snapshot(scan_id)
2013 })?
2014 .await?;
2015
2016 this.update(cx, |this, _| {
2017 let this = this.as_remote_mut().unwrap();
2018 let snapshot = &mut this.background_snapshot.lock().0;
2019 snapshot.delete_entry(entry_id);
2020 this.snapshot = snapshot.clone();
2021 })
2022 }))
2023 }
2024
2025 // fn rename_entry(
2026 // &self,
2027 // entry_id: ProjectEntryId,
2028 // new_path: impl Into<Arc<RelPath>>,
2029 // cx: &Context<Worktree>,
2030 // ) -> Task<Result<CreatedEntry>> {
2031 // let new_path: Arc<RelPath> = new_path.into();
2032 // let response = self.client.request(proto::RenameProjectEntry {
2033 // project_id: self.project_id,
2034 // entry_id: entry_id.to_proto(),
2035 // new_worktree_id: new_path.worktree_id,
2036 // new_path: new_path.as_ref().to_proto(),
2037 // });
2038 // cx.spawn(async move |this, cx| {
2039 // let response = response.await?;
2040 // match response.entry {
2041 // Some(entry) => this
2042 // .update(cx, |this, cx| {
2043 // this.as_remote_mut().unwrap().insert_entry(
2044 // entry,
2045 // response.worktree_scan_id as usize,
2046 // cx,
2047 // )
2048 // })?
2049 // .await
2050 // .map(CreatedEntry::Included),
2051 // None => {
2052 // let abs_path =
2053 // this.read_with(cx, |worktree, _| worktree.absolutize(&new_path))?;
2054 // Ok(CreatedEntry::Excluded { abs_path })
2055 // }
2056 // }
2057 // })
2058 // }
2059
2060 fn copy_external_entries(
2061 &self,
2062 target_directory: Arc<RelPath>,
2063 paths_to_copy: Vec<Arc<Path>>,
2064 local_fs: Arc<dyn Fs>,
2065 cx: &Context<Worktree>,
2066 ) -> Task<anyhow::Result<Vec<ProjectEntryId>>> {
2067 let client = self.client.clone();
2068 let worktree_id = self.id().to_proto();
2069 let project_id = self.project_id;
2070
2071 cx.background_spawn(async move {
2072 let mut requests = Vec::new();
2073 for root_path_to_copy in paths_to_copy {
2074 let Some(filename) = root_path_to_copy
2075 .file_name()
2076 .and_then(|name| name.to_str())
2077 .and_then(|filename| RelPath::unix(filename).ok())
2078 else {
2079 continue;
2080 };
2081 for (abs_path, is_directory) in
2082 read_dir_items(local_fs.as_ref(), &root_path_to_copy).await?
2083 {
2084 let Some(relative_path) = abs_path
2085 .strip_prefix(&root_path_to_copy)
2086 .map_err(|e| anyhow::Error::from(e))
2087 .and_then(|relative_path| RelPath::new(relative_path, PathStyle::local()))
2088 .log_err()
2089 else {
2090 continue;
2091 };
2092 let content = if is_directory {
2093 None
2094 } else {
2095 Some(local_fs.load_bytes(&abs_path).await?)
2096 };
2097
2098 let mut target_path = target_directory.join(filename);
2099 if relative_path.file_name().is_some() {
2100 target_path = target_path.join(&relative_path);
2101 }
2102
2103 requests.push(proto::CreateProjectEntry {
2104 project_id,
2105 worktree_id,
2106 path: target_path.to_proto(),
2107 is_directory,
2108 content,
2109 });
2110 }
2111 }
2112 requests.sort_unstable_by(|a, b| a.path.cmp(&b.path));
2113 requests.dedup();
2114
2115 let mut copied_entry_ids = Vec::new();
2116 for request in requests {
2117 let response = client.request(request).await?;
2118 copied_entry_ids.extend(response.entry.map(|e| ProjectEntryId::from_proto(e.id)));
2119 }
2120
2121 Ok(copied_entry_ids)
2122 })
2123 }
2124}
2125
2126impl Snapshot {
2127 pub fn new(
2128 id: u64,
2129 root_name: Arc<RelPath>,
2130 abs_path: Arc<Path>,
2131 path_style: PathStyle,
2132 ) -> Self {
2133 Snapshot {
2134 id: WorktreeId::from_usize(id as usize),
2135 abs_path: SanitizedPath::from_arc(abs_path),
2136 path_style,
2137 root_char_bag: root_name
2138 .as_unix_str()
2139 .chars()
2140 .map(|c| c.to_ascii_lowercase())
2141 .collect(),
2142 root_name,
2143 always_included_entries: Default::default(),
2144 entries_by_path: Default::default(),
2145 entries_by_id: Default::default(),
2146 scan_id: 1,
2147 completed_scan_id: 0,
2148 }
2149 }
2150
2151 pub fn id(&self) -> WorktreeId {
2152 self.id
2153 }
2154
2155 // TODO:
2156 // Consider the following:
2157 //
2158 // ```rust
2159 // let abs_path: Arc<Path> = snapshot.abs_path(); // e.g. "C:\Users\user\Desktop\project"
2160 // let some_non_trimmed_path = Path::new("\\\\?\\C:\\Users\\user\\Desktop\\project\\main.rs");
2161 // // The caller perform some actions here:
2162 // some_non_trimmed_path.strip_prefix(abs_path); // This fails
2163 // some_non_trimmed_path.starts_with(abs_path); // This fails too
2164 // ```
2165 //
2166 // This is definitely a bug, but it's not clear if we should handle it here or not.
2167 pub fn abs_path(&self) -> &Arc<Path> {
2168 SanitizedPath::cast_arc_ref(&self.abs_path)
2169 }
2170
2171 fn build_initial_update(&self, project_id: u64, worktree_id: u64) -> proto::UpdateWorktree {
2172 let mut updated_entries = self
2173 .entries_by_path
2174 .iter()
2175 .map(proto::Entry::from)
2176 .collect::<Vec<_>>();
2177 updated_entries.sort_unstable_by_key(|e| e.id);
2178
2179 proto::UpdateWorktree {
2180 project_id,
2181 worktree_id,
2182 abs_path: self.abs_path().to_string_lossy().into_owned(),
2183 root_name: self.root_name().to_proto(),
2184 updated_entries,
2185 removed_entries: Vec::new(),
2186 scan_id: self.scan_id as u64,
2187 is_last_update: self.completed_scan_id == self.scan_id,
2188 // Sent in separate messages.
2189 updated_repositories: Vec::new(),
2190 removed_repositories: Vec::new(),
2191 }
2192 }
2193
2194 pub fn work_directory_abs_path(&self, work_directory: &WorkDirectory) -> PathBuf {
2195 match work_directory {
2196 WorkDirectory::InProject { relative_path } => self.absolutize(relative_path),
2197 WorkDirectory::AboveProject { absolute_path, .. } => absolute_path.as_ref().to_owned(),
2198 }
2199 }
2200
2201 pub fn absolutize(&self, path: &RelPath) -> PathBuf {
2202 if path.file_name().is_some() {
2203 let mut abs_path = self.abs_path.to_string();
2204 for component in path.components() {
2205 if !abs_path.ends_with(self.path_style.primary_separator()) {
2206 abs_path.push_str(self.path_style.primary_separator());
2207 }
2208 abs_path.push_str(component);
2209 }
2210 PathBuf::from(abs_path)
2211 } else {
2212 self.abs_path.as_path().to_path_buf()
2213 }
2214 }
2215
2216 pub fn contains_entry(&self, entry_id: ProjectEntryId) -> bool {
2217 self.entries_by_id.get(&entry_id, ()).is_some()
2218 }
2219
2220 fn insert_entry(
2221 &mut self,
2222 entry: proto::Entry,
2223 always_included_paths: &PathMatcher,
2224 ) -> Result<Entry> {
2225 let entry = Entry::try_from((&self.root_char_bag, always_included_paths, entry))?;
2226 let old_entry = self.entries_by_id.insert_or_replace(
2227 PathEntry {
2228 id: entry.id,
2229 path: entry.path.clone(),
2230 is_ignored: entry.is_ignored,
2231 scan_id: 0,
2232 },
2233 (),
2234 );
2235 if let Some(old_entry) = old_entry {
2236 self.entries_by_path.remove(&PathKey(old_entry.path), ());
2237 }
2238 self.entries_by_path.insert_or_replace(entry.clone(), ());
2239 Ok(entry)
2240 }
2241
2242 fn delete_entry(&mut self, entry_id: ProjectEntryId) -> Option<Arc<RelPath>> {
2243 let removed_entry = self.entries_by_id.remove(&entry_id, ())?;
2244 self.entries_by_path = {
2245 let mut cursor = self.entries_by_path.cursor::<TraversalProgress>(());
2246 let mut new_entries_by_path =
2247 cursor.slice(&TraversalTarget::path(&removed_entry.path), Bias::Left);
2248 while let Some(entry) = cursor.item() {
2249 if entry.path.starts_with(&removed_entry.path) {
2250 self.entries_by_id.remove(&entry.id, ());
2251 cursor.next();
2252 } else {
2253 break;
2254 }
2255 }
2256 new_entries_by_path.append(cursor.suffix(), ());
2257 new_entries_by_path
2258 };
2259
2260 Some(removed_entry.path)
2261 }
2262
2263 fn update_abs_path(&mut self, abs_path: Arc<SanitizedPath>, root_name: Arc<RelPath>) {
2264 self.abs_path = abs_path;
2265 if root_name != self.root_name {
2266 self.root_char_bag = root_name
2267 .as_unix_str()
2268 .chars()
2269 .map(|c| c.to_ascii_lowercase())
2270 .collect();
2271 self.root_name = root_name;
2272 }
2273 }
2274
2275 fn apply_remote_update(
2276 &mut self,
2277 update: proto::UpdateWorktree,
2278 always_included_paths: &PathMatcher,
2279 ) {
2280 log::debug!(
2281 "applying remote worktree update. {} entries updated, {} removed",
2282 update.updated_entries.len(),
2283 update.removed_entries.len()
2284 );
2285 if let Some(root_name) = RelPath::from_proto(&update.root_name).log_err() {
2286 self.update_abs_path(
2287 SanitizedPath::new_arc(&Path::new(&update.abs_path)),
2288 root_name,
2289 );
2290 }
2291
2292 let mut entries_by_path_edits = Vec::new();
2293 let mut entries_by_id_edits = Vec::new();
2294
2295 for entry_id in update.removed_entries {
2296 let entry_id = ProjectEntryId::from_proto(entry_id);
2297 entries_by_id_edits.push(Edit::Remove(entry_id));
2298 if let Some(entry) = self.entry_for_id(entry_id) {
2299 entries_by_path_edits.push(Edit::Remove(PathKey(entry.path.clone())));
2300 }
2301 }
2302
2303 for entry in update.updated_entries {
2304 let Some(entry) =
2305 Entry::try_from((&self.root_char_bag, always_included_paths, entry)).log_err()
2306 else {
2307 continue;
2308 };
2309 if let Some(PathEntry { path, .. }) = self.entries_by_id.get(&entry.id, ()) {
2310 entries_by_path_edits.push(Edit::Remove(PathKey(path.clone())));
2311 }
2312 if let Some(old_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2313 && old_entry.id != entry.id
2314 {
2315 entries_by_id_edits.push(Edit::Remove(old_entry.id));
2316 }
2317 entries_by_id_edits.push(Edit::Insert(PathEntry {
2318 id: entry.id,
2319 path: entry.path.clone(),
2320 is_ignored: entry.is_ignored,
2321 scan_id: 0,
2322 }));
2323 entries_by_path_edits.push(Edit::Insert(entry));
2324 }
2325
2326 self.entries_by_path.edit(entries_by_path_edits, ());
2327 self.entries_by_id.edit(entries_by_id_edits, ());
2328
2329 self.scan_id = update.scan_id as usize;
2330 if update.is_last_update {
2331 self.completed_scan_id = update.scan_id as usize;
2332 }
2333 }
2334
2335 pub fn entry_count(&self) -> usize {
2336 self.entries_by_path.summary().count
2337 }
2338
2339 pub fn visible_entry_count(&self) -> usize {
2340 self.entries_by_path.summary().non_ignored_count
2341 }
2342
2343 pub fn dir_count(&self) -> usize {
2344 let summary = self.entries_by_path.summary();
2345 summary.count - summary.file_count
2346 }
2347
2348 pub fn visible_dir_count(&self) -> usize {
2349 let summary = self.entries_by_path.summary();
2350 summary.non_ignored_count - summary.non_ignored_file_count
2351 }
2352
2353 pub fn file_count(&self) -> usize {
2354 self.entries_by_path.summary().file_count
2355 }
2356
2357 pub fn visible_file_count(&self) -> usize {
2358 self.entries_by_path.summary().non_ignored_file_count
2359 }
2360
2361 fn traverse_from_offset(
2362 &self,
2363 include_files: bool,
2364 include_dirs: bool,
2365 include_ignored: bool,
2366 start_offset: usize,
2367 ) -> Traversal<'_> {
2368 let mut cursor = self.entries_by_path.cursor(());
2369 cursor.seek(
2370 &TraversalTarget::Count {
2371 count: start_offset,
2372 include_files,
2373 include_dirs,
2374 include_ignored,
2375 },
2376 Bias::Right,
2377 );
2378 Traversal {
2379 snapshot: self,
2380 cursor,
2381 include_files,
2382 include_dirs,
2383 include_ignored,
2384 }
2385 }
2386
2387 pub fn traverse_from_path(
2388 &self,
2389 include_files: bool,
2390 include_dirs: bool,
2391 include_ignored: bool,
2392 path: &RelPath,
2393 ) -> Traversal<'_> {
2394 Traversal::new(self, include_files, include_dirs, include_ignored, path)
2395 }
2396
2397 pub fn files(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2398 self.traverse_from_offset(true, false, include_ignored, start)
2399 }
2400
2401 pub fn directories(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2402 self.traverse_from_offset(false, true, include_ignored, start)
2403 }
2404
2405 pub fn entries(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2406 self.traverse_from_offset(true, true, include_ignored, start)
2407 }
2408
2409 pub fn paths(&self) -> impl Iterator<Item = &RelPath> {
2410 self.entries_by_path
2411 .cursor::<()>(())
2412 .filter(move |entry| !entry.path.is_empty())
2413 .map(|entry| entry.path.as_ref())
2414 }
2415
2416 pub fn child_entries<'a>(&'a self, parent_path: &'a RelPath) -> ChildEntriesIter<'a> {
2417 let options = ChildEntriesOptions {
2418 include_files: true,
2419 include_dirs: true,
2420 include_ignored: true,
2421 };
2422 self.child_entries_with_options(parent_path, options)
2423 }
2424
2425 pub fn child_entries_with_options<'a>(
2426 &'a self,
2427 parent_path: &'a RelPath,
2428 options: ChildEntriesOptions,
2429 ) -> ChildEntriesIter<'a> {
2430 let mut cursor = self.entries_by_path.cursor(());
2431 cursor.seek(&TraversalTarget::path(parent_path), Bias::Right);
2432 let traversal = Traversal {
2433 snapshot: self,
2434 cursor,
2435 include_files: options.include_files,
2436 include_dirs: options.include_dirs,
2437 include_ignored: options.include_ignored,
2438 };
2439 ChildEntriesIter {
2440 traversal,
2441 parent_path,
2442 }
2443 }
2444
2445 pub fn root_entry(&self) -> Option<&Entry> {
2446 self.entries_by_path.first()
2447 }
2448
2449 /// Returns `None` for a single file worktree, or `Some(self.abs_path())` if
2450 /// it is a directory.
2451 pub fn root_dir(&self) -> Option<Arc<Path>> {
2452 self.root_entry()
2453 .filter(|entry| entry.is_dir())
2454 .map(|_| self.abs_path().clone())
2455 }
2456
2457 pub fn root_name(&self) -> &RelPath {
2458 &self.root_name
2459 }
2460
2461 pub fn root_name_str(&self) -> &str {
2462 self.root_name.as_unix_str()
2463 }
2464
2465 pub fn scan_id(&self) -> usize {
2466 self.scan_id
2467 }
2468
2469 pub fn entry_for_path(&self, path: &RelPath) -> Option<&Entry> {
2470 self.traverse_from_path(true, true, true, path)
2471 .entry()
2472 .and_then(|entry| {
2473 if entry.path.as_ref() == path {
2474 Some(entry)
2475 } else {
2476 None
2477 }
2478 })
2479 }
2480
2481 /// Resolves a path to an executable using the following heuristics:
2482 ///
2483 /// 1. If the path starts with `~`, it is expanded to the user's home directory.
2484 /// 2. If the path is relative and contains more than one component,
2485 /// it is joined to the worktree root path.
2486 /// 3. If the path is relative and exists in the worktree
2487 /// (even if falls under an exclusion filter),
2488 /// it is joined to the worktree root path.
2489 /// 4. Otherwise the path is returned unmodified.
2490 ///
2491 /// Relative paths that do not exist in the worktree may
2492 /// still be found using the `PATH` environment variable.
2493 pub fn resolve_executable_path(&self, path: PathBuf) -> PathBuf {
2494 if let Some(path_str) = path.to_str() {
2495 if let Some(remaining_path) = path_str.strip_prefix("~/") {
2496 return home_dir().join(remaining_path);
2497 } else if path_str == "~" {
2498 return home_dir().to_path_buf();
2499 }
2500 }
2501
2502 if let Ok(rel_path) = RelPath::new(&path, self.path_style)
2503 && (path.components().count() > 1 || self.entry_for_path(&rel_path).is_some())
2504 {
2505 self.abs_path().join(path)
2506 } else {
2507 path
2508 }
2509 }
2510
2511 pub fn entry_for_id(&self, id: ProjectEntryId) -> Option<&Entry> {
2512 let entry = self.entries_by_id.get(&id, ())?;
2513 self.entry_for_path(&entry.path)
2514 }
2515
2516 pub fn path_style(&self) -> PathStyle {
2517 self.path_style
2518 }
2519}
2520
2521impl LocalSnapshot {
2522 fn local_repo_for_work_directory_path(&self, path: &RelPath) -> Option<&LocalRepositoryEntry> {
2523 self.git_repositories
2524 .iter()
2525 .map(|(_, entry)| entry)
2526 .find(|entry| entry.work_directory.path_key() == PathKey(path.into()))
2527 }
2528
2529 fn build_update(
2530 &self,
2531 project_id: u64,
2532 worktree_id: u64,
2533 entry_changes: UpdatedEntriesSet,
2534 ) -> proto::UpdateWorktree {
2535 let mut updated_entries = Vec::new();
2536 let mut removed_entries = Vec::new();
2537
2538 for (_, entry_id, path_change) in entry_changes.iter() {
2539 if let PathChange::Removed = path_change {
2540 removed_entries.push(entry_id.0 as u64);
2541 } else if let Some(entry) = self.entry_for_id(*entry_id) {
2542 updated_entries.push(proto::Entry::from(entry));
2543 }
2544 }
2545
2546 removed_entries.sort_unstable();
2547 updated_entries.sort_unstable_by_key(|e| e.id);
2548
2549 // TODO - optimize, knowing that removed_entries are sorted.
2550 removed_entries.retain(|id| updated_entries.binary_search_by_key(id, |e| e.id).is_err());
2551
2552 proto::UpdateWorktree {
2553 project_id,
2554 worktree_id,
2555 abs_path: self.abs_path().to_string_lossy().into_owned(),
2556 root_name: self.root_name().to_proto(),
2557 updated_entries,
2558 removed_entries,
2559 scan_id: self.scan_id as u64,
2560 is_last_update: self.completed_scan_id == self.scan_id,
2561 // Sent in separate messages.
2562 updated_repositories: Vec::new(),
2563 removed_repositories: Vec::new(),
2564 }
2565 }
2566
2567 fn insert_entry(&mut self, mut entry: Entry, fs: &dyn Fs) -> Entry {
2568 log::trace!("insert entry {:?}", entry.path);
2569 if entry.is_file() && entry.path.file_name() == Some(&GITIGNORE) {
2570 let abs_path = self.absolutize(&entry.path);
2571 match self.executor.block(build_gitignore(&abs_path, fs)) {
2572 Ok(ignore) => {
2573 self.ignores_by_parent_abs_path
2574 .insert(abs_path.parent().unwrap().into(), (Arc::new(ignore), true));
2575 }
2576 Err(error) => {
2577 log::error!(
2578 "error loading .gitignore file {:?} - {:?}",
2579 &entry.path,
2580 error
2581 );
2582 }
2583 }
2584 }
2585
2586 if entry.kind == EntryKind::PendingDir
2587 && let Some(existing_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2588 {
2589 entry.kind = existing_entry.kind;
2590 }
2591
2592 let scan_id = self.scan_id;
2593 let removed = self.entries_by_path.insert_or_replace(entry.clone(), ());
2594 if let Some(removed) = removed
2595 && removed.id != entry.id
2596 {
2597 self.entries_by_id.remove(&removed.id, ());
2598 }
2599 self.entries_by_id.insert_or_replace(
2600 PathEntry {
2601 id: entry.id,
2602 path: entry.path.clone(),
2603 is_ignored: entry.is_ignored,
2604 scan_id,
2605 },
2606 (),
2607 );
2608
2609 entry
2610 }
2611
2612 fn ancestor_inodes_for_path(&self, path: &RelPath) -> TreeSet<u64> {
2613 let mut inodes = TreeSet::default();
2614 for ancestor in path.ancestors().skip(1) {
2615 if let Some(entry) = self.entry_for_path(ancestor) {
2616 inodes.insert(entry.inode);
2617 }
2618 }
2619 inodes
2620 }
2621
2622 async fn ignore_stack_for_abs_path(
2623 &self,
2624 abs_path: &Path,
2625 is_dir: bool,
2626 fs: &dyn Fs,
2627 ) -> IgnoreStack {
2628 let mut new_ignores = Vec::new();
2629 let mut repo_root = None;
2630 for (index, ancestor) in abs_path.ancestors().enumerate() {
2631 if index > 0 {
2632 if let Some((ignore, _)) = self.ignores_by_parent_abs_path.get(ancestor) {
2633 new_ignores.push((ancestor, Some(ignore.clone())));
2634 } else {
2635 new_ignores.push((ancestor, None));
2636 }
2637 }
2638
2639 let metadata = fs.metadata(&ancestor.join(DOT_GIT)).await.ok().flatten();
2640 if metadata.is_some() {
2641 repo_root = Some(Arc::from(ancestor));
2642 break;
2643 }
2644 }
2645
2646 let mut ignore_stack = if let Some(global_gitignore) = self.global_gitignore.clone() {
2647 IgnoreStack::global(global_gitignore)
2648 } else {
2649 IgnoreStack::none()
2650 };
2651
2652 if let Some((repo_exclude, _)) = repo_root
2653 .as_ref()
2654 .and_then(|abs_path| self.repo_exclude_by_work_dir_abs_path.get(abs_path))
2655 {
2656 ignore_stack = ignore_stack.append(IgnoreKind::RepoExclude, repo_exclude.clone());
2657 }
2658 ignore_stack.repo_root = repo_root;
2659 for (parent_abs_path, ignore) in new_ignores.into_iter().rev() {
2660 if ignore_stack.is_abs_path_ignored(parent_abs_path, true) {
2661 ignore_stack = IgnoreStack::all();
2662 break;
2663 } else if let Some(ignore) = ignore {
2664 ignore_stack =
2665 ignore_stack.append(IgnoreKind::Gitignore(parent_abs_path.into()), ignore);
2666 }
2667 }
2668
2669 if ignore_stack.is_abs_path_ignored(abs_path, is_dir) {
2670 ignore_stack = IgnoreStack::all();
2671 }
2672
2673 ignore_stack
2674 }
2675
2676 #[cfg(test)]
2677 fn expanded_entries(&self) -> impl Iterator<Item = &Entry> {
2678 self.entries_by_path
2679 .cursor::<()>(())
2680 .filter(|entry| entry.kind == EntryKind::Dir && (entry.is_external || entry.is_ignored))
2681 }
2682
2683 #[cfg(test)]
2684 pub fn check_invariants(&self, git_state: bool) {
2685 use pretty_assertions::assert_eq;
2686
2687 assert_eq!(
2688 self.entries_by_path
2689 .cursor::<()>(())
2690 .map(|e| (&e.path, e.id))
2691 .collect::<Vec<_>>(),
2692 self.entries_by_id
2693 .cursor::<()>(())
2694 .map(|e| (&e.path, e.id))
2695 .collect::<collections::BTreeSet<_>>()
2696 .into_iter()
2697 .collect::<Vec<_>>(),
2698 "entries_by_path and entries_by_id are inconsistent"
2699 );
2700
2701 let mut files = self.files(true, 0);
2702 let mut visible_files = self.files(false, 0);
2703 for entry in self.entries_by_path.cursor::<()>(()) {
2704 if entry.is_file() {
2705 assert_eq!(files.next().unwrap().inode, entry.inode);
2706 if (!entry.is_ignored && !entry.is_external) || entry.is_always_included {
2707 assert_eq!(visible_files.next().unwrap().inode, entry.inode);
2708 }
2709 }
2710 }
2711
2712 assert!(files.next().is_none());
2713 assert!(visible_files.next().is_none());
2714
2715 let mut bfs_paths = Vec::new();
2716 let mut stack = self
2717 .root_entry()
2718 .map(|e| e.path.as_ref())
2719 .into_iter()
2720 .collect::<Vec<_>>();
2721 while let Some(path) = stack.pop() {
2722 bfs_paths.push(path);
2723 let ix = stack.len();
2724 for child_entry in self.child_entries(path) {
2725 stack.insert(ix, &child_entry.path);
2726 }
2727 }
2728
2729 let dfs_paths_via_iter = self
2730 .entries_by_path
2731 .cursor::<()>(())
2732 .map(|e| e.path.as_ref())
2733 .collect::<Vec<_>>();
2734 assert_eq!(bfs_paths, dfs_paths_via_iter);
2735
2736 let dfs_paths_via_traversal = self
2737 .entries(true, 0)
2738 .map(|e| e.path.as_ref())
2739 .collect::<Vec<_>>();
2740
2741 assert_eq!(dfs_paths_via_traversal, dfs_paths_via_iter);
2742
2743 if git_state {
2744 for ignore_parent_abs_path in self.ignores_by_parent_abs_path.keys() {
2745 let ignore_parent_path = &RelPath::new(
2746 ignore_parent_abs_path
2747 .strip_prefix(self.abs_path.as_path())
2748 .unwrap(),
2749 PathStyle::local(),
2750 )
2751 .unwrap();
2752 assert!(self.entry_for_path(ignore_parent_path).is_some());
2753 assert!(
2754 self.entry_for_path(
2755 &ignore_parent_path.join(RelPath::unix(GITIGNORE).unwrap())
2756 )
2757 .is_some()
2758 );
2759 }
2760 }
2761 }
2762
2763 #[cfg(test)]
2764 pub fn entries_without_ids(&self, include_ignored: bool) -> Vec<(&RelPath, u64, bool)> {
2765 let mut paths = Vec::new();
2766 for entry in self.entries_by_path.cursor::<()>(()) {
2767 if include_ignored || !entry.is_ignored {
2768 paths.push((entry.path.as_ref(), entry.inode, entry.is_ignored));
2769 }
2770 }
2771 paths.sort_by(|a, b| a.0.cmp(b.0));
2772 paths
2773 }
2774}
2775
2776impl BackgroundScannerState {
2777 fn should_scan_directory(&self, entry: &Entry) -> bool {
2778 (!entry.is_external && (!entry.is_ignored || entry.is_always_included))
2779 || entry.path.file_name() == Some(DOT_GIT)
2780 || entry.path.file_name() == Some(local_settings_folder_name())
2781 || entry.path.file_name() == Some(local_vscode_folder_name())
2782 || self.scanned_dirs.contains(&entry.id) // If we've ever scanned it, keep scanning
2783 || self
2784 .paths_to_scan
2785 .iter()
2786 .any(|p| p.starts_with(&entry.path))
2787 || self
2788 .path_prefixes_to_scan
2789 .iter()
2790 .any(|p| entry.path.starts_with(p))
2791 }
2792
2793 async fn enqueue_scan_dir(
2794 &self,
2795 abs_path: Arc<Path>,
2796 entry: &Entry,
2797 scan_job_tx: &Sender<ScanJob>,
2798 fs: &dyn Fs,
2799 ) {
2800 let path = entry.path.clone();
2801 let ignore_stack = self
2802 .snapshot
2803 .ignore_stack_for_abs_path(&abs_path, true, fs)
2804 .await;
2805 let mut ancestor_inodes = self.snapshot.ancestor_inodes_for_path(&path);
2806
2807 if !ancestor_inodes.contains(&entry.inode) {
2808 ancestor_inodes.insert(entry.inode);
2809 scan_job_tx
2810 .try_send(ScanJob {
2811 abs_path,
2812 path,
2813 ignore_stack,
2814 scan_queue: scan_job_tx.clone(),
2815 ancestor_inodes,
2816 is_external: entry.is_external,
2817 })
2818 .unwrap();
2819 }
2820 }
2821
2822 fn reuse_entry_id(&mut self, entry: &mut Entry) {
2823 if let Some(mtime) = entry.mtime {
2824 // If an entry with the same inode was removed from the worktree during this scan,
2825 // then it *might* represent the same file or directory. But the OS might also have
2826 // re-used the inode for a completely different file or directory.
2827 //
2828 // Conditionally reuse the old entry's id:
2829 // * if the mtime is the same, the file was probably been renamed.
2830 // * if the path is the same, the file may just have been updated
2831 if let Some(removed_entry) = self.removed_entries.remove(&entry.inode) {
2832 if removed_entry.mtime == Some(mtime) || removed_entry.path == entry.path {
2833 entry.id = removed_entry.id;
2834 }
2835 } else if let Some(existing_entry) = self.snapshot.entry_for_path(&entry.path) {
2836 entry.id = existing_entry.id;
2837 }
2838 }
2839 }
2840
2841 fn entry_id_for(
2842 &mut self,
2843 next_entry_id: &AtomicUsize,
2844 path: &RelPath,
2845 metadata: &fs::Metadata,
2846 ) -> ProjectEntryId {
2847 // If an entry with the same inode was removed from the worktree during this scan,
2848 // then it *might* represent the same file or directory. But the OS might also have
2849 // re-used the inode for a completely different file or directory.
2850 //
2851 // Conditionally reuse the old entry's id:
2852 // * if the mtime is the same, the file was probably been renamed.
2853 // * if the path is the same, the file may just have been updated
2854 if let Some(removed_entry) = self.removed_entries.remove(&metadata.inode) {
2855 if removed_entry.mtime == Some(metadata.mtime) || *removed_entry.path == *path {
2856 return removed_entry.id;
2857 }
2858 } else if let Some(existing_entry) = self.snapshot.entry_for_path(path) {
2859 return existing_entry.id;
2860 }
2861 ProjectEntryId::new(next_entry_id)
2862 }
2863
2864 async fn insert_entry(&mut self, entry: Entry, fs: &dyn Fs, watcher: &dyn Watcher) -> Entry {
2865 let entry = self.snapshot.insert_entry(entry, fs);
2866 if entry.path.file_name() == Some(&DOT_GIT) {
2867 self.insert_git_repository(entry.path.clone(), fs, watcher)
2868 .await;
2869 }
2870
2871 #[cfg(test)]
2872 self.snapshot.check_invariants(false);
2873
2874 entry
2875 }
2876
2877 fn populate_dir(
2878 &mut self,
2879 parent_path: Arc<RelPath>,
2880 entries: impl IntoIterator<Item = Entry>,
2881 ignore: Option<Arc<Gitignore>>,
2882 ) {
2883 let mut parent_entry = if let Some(parent_entry) = self
2884 .snapshot
2885 .entries_by_path
2886 .get(&PathKey(parent_path.clone()), ())
2887 {
2888 parent_entry.clone()
2889 } else {
2890 log::warn!(
2891 "populating a directory {:?} that has been removed",
2892 parent_path
2893 );
2894 return;
2895 };
2896
2897 match parent_entry.kind {
2898 EntryKind::PendingDir | EntryKind::UnloadedDir => parent_entry.kind = EntryKind::Dir,
2899 EntryKind::Dir => {}
2900 _ => return,
2901 }
2902
2903 if let Some(ignore) = ignore {
2904 let abs_parent_path = self
2905 .snapshot
2906 .abs_path
2907 .as_path()
2908 .join(parent_path.as_std_path())
2909 .into();
2910 self.snapshot
2911 .ignores_by_parent_abs_path
2912 .insert(abs_parent_path, (ignore, false));
2913 }
2914
2915 let parent_entry_id = parent_entry.id;
2916 self.scanned_dirs.insert(parent_entry_id);
2917 let mut entries_by_path_edits = vec![Edit::Insert(parent_entry)];
2918 let mut entries_by_id_edits = Vec::new();
2919
2920 for entry in entries {
2921 entries_by_id_edits.push(Edit::Insert(PathEntry {
2922 id: entry.id,
2923 path: entry.path.clone(),
2924 is_ignored: entry.is_ignored,
2925 scan_id: self.snapshot.scan_id,
2926 }));
2927 entries_by_path_edits.push(Edit::Insert(entry));
2928 }
2929
2930 self.snapshot
2931 .entries_by_path
2932 .edit(entries_by_path_edits, ());
2933 self.snapshot.entries_by_id.edit(entries_by_id_edits, ());
2934
2935 if let Err(ix) = self.changed_paths.binary_search(&parent_path) {
2936 self.changed_paths.insert(ix, parent_path.clone());
2937 }
2938
2939 #[cfg(test)]
2940 self.snapshot.check_invariants(false);
2941 }
2942
2943 fn remove_path(&mut self, path: &RelPath) {
2944 log::trace!("background scanner removing path {path:?}");
2945 let mut new_entries;
2946 let removed_entries;
2947 {
2948 let mut cursor = self
2949 .snapshot
2950 .entries_by_path
2951 .cursor::<TraversalProgress>(());
2952 new_entries = cursor.slice(&TraversalTarget::path(path), Bias::Left);
2953 removed_entries = cursor.slice(&TraversalTarget::successor(path), Bias::Left);
2954 new_entries.append(cursor.suffix(), ());
2955 }
2956 self.snapshot.entries_by_path = new_entries;
2957
2958 let mut removed_ids = Vec::with_capacity(removed_entries.summary().count);
2959 for entry in removed_entries.cursor::<()>(()) {
2960 match self.removed_entries.entry(entry.inode) {
2961 hash_map::Entry::Occupied(mut e) => {
2962 let prev_removed_entry = e.get_mut();
2963 if entry.id > prev_removed_entry.id {
2964 *prev_removed_entry = entry.clone();
2965 }
2966 }
2967 hash_map::Entry::Vacant(e) => {
2968 e.insert(entry.clone());
2969 }
2970 }
2971
2972 if entry.path.file_name() == Some(GITIGNORE) {
2973 let abs_parent_path = self.snapshot.absolutize(&entry.path.parent().unwrap());
2974 if let Some((_, needs_update)) = self
2975 .snapshot
2976 .ignores_by_parent_abs_path
2977 .get_mut(abs_parent_path.as_path())
2978 {
2979 *needs_update = true;
2980 }
2981 }
2982
2983 if let Err(ix) = removed_ids.binary_search(&entry.id) {
2984 removed_ids.insert(ix, entry.id);
2985 }
2986 }
2987
2988 self.snapshot
2989 .entries_by_id
2990 .edit(removed_ids.iter().map(|&id| Edit::Remove(id)).collect(), ());
2991 self.snapshot
2992 .git_repositories
2993 .retain(|id, _| removed_ids.binary_search(id).is_err());
2994
2995 #[cfg(test)]
2996 self.snapshot.check_invariants(false);
2997 }
2998
2999 async fn insert_git_repository(
3000 &mut self,
3001 dot_git_path: Arc<RelPath>,
3002 fs: &dyn Fs,
3003 watcher: &dyn Watcher,
3004 ) {
3005 let work_dir_path: Arc<RelPath> = match dot_git_path.parent() {
3006 Some(parent_dir) => {
3007 // Guard against repositories inside the repository metadata
3008 if parent_dir
3009 .components()
3010 .any(|component| component == DOT_GIT)
3011 {
3012 log::debug!(
3013 "not building git repository for nested `.git` directory, `.git` path in the worktree: {dot_git_path:?}"
3014 );
3015 return;
3016 };
3017
3018 parent_dir.into()
3019 }
3020 None => {
3021 // `dot_git_path.parent().is_none()` means `.git` directory is the opened worktree itself,
3022 // no files inside that directory are tracked by git, so no need to build the repo around it
3023 log::debug!(
3024 "not building git repository for the worktree itself, `.git` path in the worktree: {dot_git_path:?}"
3025 );
3026 return;
3027 }
3028 };
3029
3030 let dot_git_abs_path = Arc::from(self.snapshot.absolutize(&dot_git_path).as_ref());
3031
3032 self.insert_git_repository_for_path(
3033 WorkDirectory::InProject {
3034 relative_path: work_dir_path,
3035 },
3036 dot_git_abs_path,
3037 fs,
3038 watcher,
3039 )
3040 .await
3041 .log_err();
3042 }
3043
3044 async fn insert_git_repository_for_path(
3045 &mut self,
3046 work_directory: WorkDirectory,
3047 dot_git_abs_path: Arc<Path>,
3048 fs: &dyn Fs,
3049 watcher: &dyn Watcher,
3050 ) -> Result<LocalRepositoryEntry> {
3051 let work_dir_entry = self
3052 .snapshot
3053 .entry_for_path(&work_directory.path_key().0)
3054 .with_context(|| {
3055 format!(
3056 "working directory `{}` not indexed",
3057 work_directory
3058 .path_key()
3059 .0
3060 .display(self.snapshot.path_style)
3061 )
3062 })?;
3063 let work_directory_abs_path = self.snapshot.work_directory_abs_path(&work_directory);
3064
3065 let (repository_dir_abs_path, common_dir_abs_path) =
3066 discover_git_paths(&dot_git_abs_path, fs).await;
3067 watcher
3068 .add(&common_dir_abs_path)
3069 .context("failed to add common directory to watcher")
3070 .log_err();
3071 if !repository_dir_abs_path.starts_with(&common_dir_abs_path) {
3072 watcher
3073 .add(&repository_dir_abs_path)
3074 .context("failed to add repository directory to watcher")
3075 .log_err();
3076 }
3077
3078 let work_directory_id = work_dir_entry.id;
3079
3080 let local_repository = LocalRepositoryEntry {
3081 work_directory_id,
3082 work_directory,
3083 work_directory_abs_path: work_directory_abs_path.as_path().into(),
3084 git_dir_scan_id: 0,
3085 dot_git_abs_path,
3086 common_dir_abs_path,
3087 repository_dir_abs_path,
3088 };
3089
3090 self.snapshot
3091 .git_repositories
3092 .insert(work_directory_id, local_repository.clone());
3093
3094 log::trace!("inserting new local git repository");
3095 Ok(local_repository)
3096 }
3097}
3098
3099async fn is_git_dir(path: &Path, fs: &dyn Fs) -> bool {
3100 if let Some(file_name) = path.file_name()
3101 && file_name == DOT_GIT
3102 {
3103 return true;
3104 }
3105
3106 // If we're in a bare repository, we are not inside a `.git` folder. In a
3107 // bare repository, the root folder contains what would normally be in the
3108 // `.git` folder.
3109 let head_metadata = fs.metadata(&path.join("HEAD")).await;
3110 if !matches!(head_metadata, Ok(Some(_))) {
3111 return false;
3112 }
3113 let config_metadata = fs.metadata(&path.join("config")).await;
3114 matches!(config_metadata, Ok(Some(_)))
3115}
3116
3117async fn build_gitignore(abs_path: &Path, fs: &dyn Fs) -> Result<Gitignore> {
3118 let contents = fs
3119 .load(abs_path)
3120 .await
3121 .with_context(|| format!("failed to load gitignore file at {}", abs_path.display()))?;
3122 let parent = abs_path.parent().unwrap_or_else(|| Path::new("/"));
3123 let mut builder = GitignoreBuilder::new(parent);
3124 for line in contents.lines() {
3125 builder.add_line(Some(abs_path.into()), line)?;
3126 }
3127 Ok(builder.build()?)
3128}
3129
3130impl Deref for Worktree {
3131 type Target = Snapshot;
3132
3133 fn deref(&self) -> &Self::Target {
3134 match self {
3135 Worktree::Local(worktree) => &worktree.snapshot,
3136 Worktree::Remote(worktree) => &worktree.snapshot,
3137 }
3138 }
3139}
3140
3141impl Deref for LocalWorktree {
3142 type Target = LocalSnapshot;
3143
3144 fn deref(&self) -> &Self::Target {
3145 &self.snapshot
3146 }
3147}
3148
3149impl Deref for RemoteWorktree {
3150 type Target = Snapshot;
3151
3152 fn deref(&self) -> &Self::Target {
3153 &self.snapshot
3154 }
3155}
3156
3157impl fmt::Debug for LocalWorktree {
3158 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3159 self.snapshot.fmt(f)
3160 }
3161}
3162
3163impl fmt::Debug for Snapshot {
3164 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3165 struct EntriesById<'a>(&'a SumTree<PathEntry>);
3166 struct EntriesByPath<'a>(&'a SumTree<Entry>);
3167
3168 impl fmt::Debug for EntriesByPath<'_> {
3169 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3170 f.debug_map()
3171 .entries(self.0.iter().map(|entry| (&entry.path, entry.id)))
3172 .finish()
3173 }
3174 }
3175
3176 impl fmt::Debug for EntriesById<'_> {
3177 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3178 f.debug_list().entries(self.0.iter()).finish()
3179 }
3180 }
3181
3182 f.debug_struct("Snapshot")
3183 .field("id", &self.id)
3184 .field("root_name", &self.root_name)
3185 .field("entries_by_path", &EntriesByPath(&self.entries_by_path))
3186 .field("entries_by_id", &EntriesById(&self.entries_by_id))
3187 .finish()
3188 }
3189}
3190
3191#[derive(Debug, Clone, PartialEq)]
3192pub struct File {
3193 pub worktree: Entity<Worktree>,
3194 pub path: Arc<RelPath>,
3195 pub disk_state: DiskState,
3196 pub entry_id: Option<ProjectEntryId>,
3197 pub is_local: bool,
3198 pub is_private: bool,
3199}
3200
3201impl language::File for File {
3202 fn as_local(&self) -> Option<&dyn language::LocalFile> {
3203 if self.is_local { Some(self) } else { None }
3204 }
3205
3206 fn disk_state(&self) -> DiskState {
3207 self.disk_state
3208 }
3209
3210 fn path(&self) -> &Arc<RelPath> {
3211 &self.path
3212 }
3213
3214 fn full_path(&self, cx: &App) -> PathBuf {
3215 self.worktree.read(cx).full_path(&self.path)
3216 }
3217
3218 /// Returns the last component of this handle's absolute path. If this handle refers to the root
3219 /// of its worktree, then this method will return the name of the worktree itself.
3220 fn file_name<'a>(&'a self, cx: &'a App) -> &'a str {
3221 self.path
3222 .file_name()
3223 .unwrap_or_else(|| self.worktree.read(cx).root_name_str())
3224 }
3225
3226 fn worktree_id(&self, cx: &App) -> WorktreeId {
3227 self.worktree.read(cx).id()
3228 }
3229
3230 fn to_proto(&self, cx: &App) -> rpc::proto::File {
3231 rpc::proto::File {
3232 worktree_id: self.worktree.read(cx).id().to_proto(),
3233 entry_id: self.entry_id.map(|id| id.to_proto()),
3234 path: self.path.as_ref().to_proto(),
3235 mtime: self.disk_state.mtime().map(|time| time.into()),
3236 is_deleted: self.disk_state.is_deleted(),
3237 is_historic: matches!(self.disk_state, DiskState::Historic { .. }),
3238 }
3239 }
3240
3241 fn is_private(&self) -> bool {
3242 self.is_private
3243 }
3244
3245 fn path_style(&self, cx: &App) -> PathStyle {
3246 self.worktree.read(cx).path_style()
3247 }
3248}
3249
3250impl language::LocalFile for File {
3251 fn abs_path(&self, cx: &App) -> PathBuf {
3252 self.worktree.read(cx).absolutize(&self.path)
3253 }
3254
3255 fn load(&self, cx: &App) -> Task<Result<String>> {
3256 let worktree = self.worktree.read(cx).as_local().unwrap();
3257 let abs_path = worktree.absolutize(&self.path);
3258 let fs = worktree.fs.clone();
3259 cx.background_spawn(async move { fs.load(&abs_path).await })
3260 }
3261
3262 fn load_bytes(&self, cx: &App) -> Task<Result<Vec<u8>>> {
3263 let worktree = self.worktree.read(cx).as_local().unwrap();
3264 let abs_path = worktree.absolutize(&self.path);
3265 let fs = worktree.fs.clone();
3266 cx.background_spawn(async move { fs.load_bytes(&abs_path).await })
3267 }
3268}
3269
3270impl File {
3271 pub fn for_entry(entry: Entry, worktree: Entity<Worktree>) -> Arc<Self> {
3272 Arc::new(Self {
3273 worktree,
3274 path: entry.path.clone(),
3275 disk_state: if let Some(mtime) = entry.mtime {
3276 DiskState::Present { mtime }
3277 } else {
3278 DiskState::New
3279 },
3280 entry_id: Some(entry.id),
3281 is_local: true,
3282 is_private: entry.is_private,
3283 })
3284 }
3285
3286 pub fn from_proto(
3287 proto: rpc::proto::File,
3288 worktree: Entity<Worktree>,
3289 cx: &App,
3290 ) -> Result<Self> {
3291 let worktree_id = worktree.read(cx).as_remote().context("not remote")?.id();
3292
3293 anyhow::ensure!(
3294 worktree_id.to_proto() == proto.worktree_id,
3295 "worktree id does not match file"
3296 );
3297
3298 let disk_state = if proto.is_historic {
3299 DiskState::Historic {
3300 was_deleted: proto.is_deleted,
3301 }
3302 } else if proto.is_deleted {
3303 DiskState::Deleted
3304 } else if let Some(mtime) = proto.mtime.map(&Into::into) {
3305 DiskState::Present { mtime }
3306 } else {
3307 DiskState::New
3308 };
3309
3310 Ok(Self {
3311 worktree,
3312 path: RelPath::from_proto(&proto.path).context("invalid path in file protobuf")?,
3313 disk_state,
3314 entry_id: proto.entry_id.map(ProjectEntryId::from_proto),
3315 is_local: false,
3316 is_private: false,
3317 })
3318 }
3319
3320 pub fn from_dyn(file: Option<&Arc<dyn language::File>>) -> Option<&Self> {
3321 file.and_then(|f| {
3322 let f: &dyn language::File = f.borrow();
3323 let f: &dyn Any = f;
3324 f.downcast_ref()
3325 })
3326 }
3327
3328 pub fn worktree_id(&self, cx: &App) -> WorktreeId {
3329 self.worktree.read(cx).id()
3330 }
3331
3332 pub fn project_entry_id(&self) -> Option<ProjectEntryId> {
3333 match self.disk_state {
3334 DiskState::Deleted => None,
3335 _ => self.entry_id,
3336 }
3337 }
3338}
3339
3340#[derive(Clone, Debug, PartialEq, Eq)]
3341pub struct Entry {
3342 pub id: ProjectEntryId,
3343 pub kind: EntryKind,
3344 pub path: Arc<RelPath>,
3345 pub inode: u64,
3346 pub mtime: Option<MTime>,
3347
3348 pub canonical_path: Option<Arc<Path>>,
3349 /// Whether this entry is ignored by Git.
3350 ///
3351 /// We only scan ignored entries once the directory is expanded and
3352 /// exclude them from searches.
3353 pub is_ignored: bool,
3354
3355 /// Whether this entry is hidden or inside hidden directory.
3356 ///
3357 /// We only scan hidden entries once the directory is expanded.
3358 pub is_hidden: bool,
3359
3360 /// Whether this entry is always included in searches.
3361 ///
3362 /// This is used for entries that are always included in searches, even
3363 /// if they are ignored by git. Overridden by file_scan_exclusions.
3364 pub is_always_included: bool,
3365
3366 /// Whether this entry's canonical path is outside of the worktree.
3367 /// This means the entry is only accessible from the worktree root via a
3368 /// symlink.
3369 ///
3370 /// We only scan entries outside of the worktree once the symlinked
3371 /// directory is expanded. External entries are treated like gitignored
3372 /// entries in that they are not included in searches.
3373 pub is_external: bool,
3374
3375 /// Whether this entry is considered to be a `.env` file.
3376 pub is_private: bool,
3377 /// The entry's size on disk, in bytes.
3378 pub size: u64,
3379 pub char_bag: CharBag,
3380 pub is_fifo: bool,
3381}
3382
3383#[derive(Clone, Copy, Debug, PartialEq, Eq)]
3384pub enum EntryKind {
3385 UnloadedDir,
3386 PendingDir,
3387 Dir,
3388 File,
3389}
3390
3391#[derive(Clone, Copy, Debug, PartialEq)]
3392pub enum PathChange {
3393 /// A filesystem entry was was created.
3394 Added,
3395 /// A filesystem entry was removed.
3396 Removed,
3397 /// A filesystem entry was updated.
3398 Updated,
3399 /// A filesystem entry was either updated or added. We don't know
3400 /// whether or not it already existed, because the path had not
3401 /// been loaded before the event.
3402 AddedOrUpdated,
3403 /// A filesystem entry was found during the initial scan of the worktree.
3404 Loaded,
3405}
3406
3407#[derive(Clone, Debug, PartialEq, Eq)]
3408pub struct UpdatedGitRepository {
3409 /// ID of the repository's working directory.
3410 ///
3411 /// For a repo that's above the worktree root, this is the ID of the worktree root, and hence not unique.
3412 /// It's included here to aid the GitStore in detecting when a repository's working directory is renamed.
3413 pub work_directory_id: ProjectEntryId,
3414 pub old_work_directory_abs_path: Option<Arc<Path>>,
3415 pub new_work_directory_abs_path: Option<Arc<Path>>,
3416 /// For a normal git repository checkout, the absolute path to the .git directory.
3417 /// For a worktree, the absolute path to the worktree's subdirectory inside the .git directory.
3418 pub dot_git_abs_path: Option<Arc<Path>>,
3419 pub repository_dir_abs_path: Option<Arc<Path>>,
3420 pub common_dir_abs_path: Option<Arc<Path>>,
3421}
3422
3423pub type UpdatedEntriesSet = Arc<[(Arc<RelPath>, ProjectEntryId, PathChange)]>;
3424pub type UpdatedGitRepositoriesSet = Arc<[UpdatedGitRepository]>;
3425
3426#[derive(Clone, Debug)]
3427pub struct PathProgress<'a> {
3428 pub max_path: &'a RelPath,
3429}
3430
3431#[derive(Clone, Debug)]
3432pub struct PathSummary<S> {
3433 pub max_path: Arc<RelPath>,
3434 pub item_summary: S,
3435}
3436
3437impl<S: Summary> Summary for PathSummary<S> {
3438 type Context<'a> = S::Context<'a>;
3439
3440 fn zero(cx: Self::Context<'_>) -> Self {
3441 Self {
3442 max_path: RelPath::empty().into(),
3443 item_summary: S::zero(cx),
3444 }
3445 }
3446
3447 fn add_summary(&mut self, rhs: &Self, cx: Self::Context<'_>) {
3448 self.max_path = rhs.max_path.clone();
3449 self.item_summary.add_summary(&rhs.item_summary, cx);
3450 }
3451}
3452
3453impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathProgress<'a> {
3454 fn zero(_: <PathSummary<S> as Summary>::Context<'_>) -> Self {
3455 Self {
3456 max_path: RelPath::empty(),
3457 }
3458 }
3459
3460 fn add_summary(
3461 &mut self,
3462 summary: &'a PathSummary<S>,
3463 _: <PathSummary<S> as Summary>::Context<'_>,
3464 ) {
3465 self.max_path = summary.max_path.as_ref()
3466 }
3467}
3468
3469impl<'a> sum_tree::Dimension<'a, PathSummary<GitSummary>> for GitSummary {
3470 fn zero(_cx: ()) -> Self {
3471 Default::default()
3472 }
3473
3474 fn add_summary(&mut self, summary: &'a PathSummary<GitSummary>, _: ()) {
3475 *self += summary.item_summary
3476 }
3477}
3478
3479impl<'a>
3480 sum_tree::SeekTarget<'a, PathSummary<GitSummary>, Dimensions<TraversalProgress<'a>, GitSummary>>
3481 for PathTarget<'_>
3482{
3483 fn cmp(
3484 &self,
3485 cursor_location: &Dimensions<TraversalProgress<'a>, GitSummary>,
3486 _: (),
3487 ) -> Ordering {
3488 self.cmp_path(cursor_location.0.max_path)
3489 }
3490}
3491
3492impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathKey {
3493 fn zero(_: S::Context<'_>) -> Self {
3494 Default::default()
3495 }
3496
3497 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3498 self.0 = summary.max_path.clone();
3499 }
3500}
3501
3502impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for TraversalProgress<'a> {
3503 fn zero(_cx: S::Context<'_>) -> Self {
3504 Default::default()
3505 }
3506
3507 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3508 self.max_path = summary.max_path.as_ref();
3509 }
3510}
3511
3512impl Entry {
3513 fn new(
3514 path: Arc<RelPath>,
3515 metadata: &fs::Metadata,
3516 id: ProjectEntryId,
3517 root_char_bag: CharBag,
3518 canonical_path: Option<Arc<Path>>,
3519 ) -> Self {
3520 let char_bag = char_bag_for_path(root_char_bag, &path);
3521 Self {
3522 id,
3523 kind: if metadata.is_dir {
3524 EntryKind::PendingDir
3525 } else {
3526 EntryKind::File
3527 },
3528 path,
3529 inode: metadata.inode,
3530 mtime: Some(metadata.mtime),
3531 size: metadata.len,
3532 canonical_path,
3533 is_ignored: false,
3534 is_hidden: false,
3535 is_always_included: false,
3536 is_external: false,
3537 is_private: false,
3538 char_bag,
3539 is_fifo: metadata.is_fifo,
3540 }
3541 }
3542
3543 pub fn is_created(&self) -> bool {
3544 self.mtime.is_some()
3545 }
3546
3547 pub fn is_dir(&self) -> bool {
3548 self.kind.is_dir()
3549 }
3550
3551 pub fn is_file(&self) -> bool {
3552 self.kind.is_file()
3553 }
3554}
3555
3556impl EntryKind {
3557 pub fn is_dir(&self) -> bool {
3558 matches!(
3559 self,
3560 EntryKind::Dir | EntryKind::PendingDir | EntryKind::UnloadedDir
3561 )
3562 }
3563
3564 pub fn is_unloaded(&self) -> bool {
3565 matches!(self, EntryKind::UnloadedDir)
3566 }
3567
3568 pub fn is_file(&self) -> bool {
3569 matches!(self, EntryKind::File)
3570 }
3571}
3572
3573impl sum_tree::Item for Entry {
3574 type Summary = EntrySummary;
3575
3576 fn summary(&self, _cx: ()) -> Self::Summary {
3577 let non_ignored_count = if (self.is_ignored || self.is_external) && !self.is_always_included
3578 {
3579 0
3580 } else {
3581 1
3582 };
3583 let file_count;
3584 let non_ignored_file_count;
3585 if self.is_file() {
3586 file_count = 1;
3587 non_ignored_file_count = non_ignored_count;
3588 } else {
3589 file_count = 0;
3590 non_ignored_file_count = 0;
3591 }
3592
3593 EntrySummary {
3594 max_path: self.path.clone(),
3595 count: 1,
3596 non_ignored_count,
3597 file_count,
3598 non_ignored_file_count,
3599 }
3600 }
3601}
3602
3603impl sum_tree::KeyedItem for Entry {
3604 type Key = PathKey;
3605
3606 fn key(&self) -> Self::Key {
3607 PathKey(self.path.clone())
3608 }
3609}
3610
3611#[derive(Clone, Debug)]
3612pub struct EntrySummary {
3613 max_path: Arc<RelPath>,
3614 count: usize,
3615 non_ignored_count: usize,
3616 file_count: usize,
3617 non_ignored_file_count: usize,
3618}
3619
3620impl Default for EntrySummary {
3621 fn default() -> Self {
3622 Self {
3623 max_path: Arc::from(RelPath::empty()),
3624 count: 0,
3625 non_ignored_count: 0,
3626 file_count: 0,
3627 non_ignored_file_count: 0,
3628 }
3629 }
3630}
3631
3632impl sum_tree::ContextLessSummary for EntrySummary {
3633 fn zero() -> Self {
3634 Default::default()
3635 }
3636
3637 fn add_summary(&mut self, rhs: &Self) {
3638 self.max_path = rhs.max_path.clone();
3639 self.count += rhs.count;
3640 self.non_ignored_count += rhs.non_ignored_count;
3641 self.file_count += rhs.file_count;
3642 self.non_ignored_file_count += rhs.non_ignored_file_count;
3643 }
3644}
3645
3646#[derive(Clone, Debug)]
3647struct PathEntry {
3648 id: ProjectEntryId,
3649 path: Arc<RelPath>,
3650 is_ignored: bool,
3651 scan_id: usize,
3652}
3653
3654impl sum_tree::Item for PathEntry {
3655 type Summary = PathEntrySummary;
3656
3657 fn summary(&self, _cx: ()) -> Self::Summary {
3658 PathEntrySummary { max_id: self.id }
3659 }
3660}
3661
3662impl sum_tree::KeyedItem for PathEntry {
3663 type Key = ProjectEntryId;
3664
3665 fn key(&self) -> Self::Key {
3666 self.id
3667 }
3668}
3669
3670#[derive(Clone, Debug, Default)]
3671struct PathEntrySummary {
3672 max_id: ProjectEntryId,
3673}
3674
3675impl sum_tree::ContextLessSummary for PathEntrySummary {
3676 fn zero() -> Self {
3677 Default::default()
3678 }
3679
3680 fn add_summary(&mut self, summary: &Self) {
3681 self.max_id = summary.max_id;
3682 }
3683}
3684
3685impl<'a> sum_tree::Dimension<'a, PathEntrySummary> for ProjectEntryId {
3686 fn zero(_cx: ()) -> Self {
3687 Default::default()
3688 }
3689
3690 fn add_summary(&mut self, summary: &'a PathEntrySummary, _: ()) {
3691 *self = summary.max_id;
3692 }
3693}
3694
3695#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
3696pub struct PathKey(pub Arc<RelPath>);
3697
3698impl Default for PathKey {
3699 fn default() -> Self {
3700 Self(RelPath::empty().into())
3701 }
3702}
3703
3704impl<'a> sum_tree::Dimension<'a, EntrySummary> for PathKey {
3705 fn zero(_cx: ()) -> Self {
3706 Default::default()
3707 }
3708
3709 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
3710 self.0 = summary.max_path.clone();
3711 }
3712}
3713
3714struct BackgroundScanner {
3715 state: async_lock::Mutex<BackgroundScannerState>,
3716 fs: Arc<dyn Fs>,
3717 fs_case_sensitive: bool,
3718 status_updates_tx: UnboundedSender<ScanState>,
3719 executor: BackgroundExecutor,
3720 scan_requests_rx: channel::Receiver<ScanRequest>,
3721 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
3722 next_entry_id: Arc<AtomicUsize>,
3723 phase: BackgroundScannerPhase,
3724 watcher: Arc<dyn Watcher>,
3725 settings: WorktreeSettings,
3726 share_private_files: bool,
3727 scanning_enabled: bool,
3728}
3729
3730#[derive(Copy, Clone, PartialEq)]
3731enum BackgroundScannerPhase {
3732 InitialScan,
3733 EventsReceivedDuringInitialScan,
3734 Events,
3735}
3736
3737impl BackgroundScanner {
3738 async fn run(&mut self, mut fs_events_rx: Pin<Box<dyn Send + Stream<Item = Vec<PathEvent>>>>) {
3739 // If the worktree root does not contain a git repository, then find
3740 // the git repository in an ancestor directory. Find any gitignore files
3741 // in ancestor directories.
3742 let root_abs_path = self.state.lock().await.snapshot.abs_path.clone();
3743
3744 let repo = if self.scanning_enabled {
3745 let (ignores, exclude, repo) =
3746 discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await;
3747 self.state
3748 .lock()
3749 .await
3750 .snapshot
3751 .ignores_by_parent_abs_path
3752 .extend(ignores);
3753 if let Some(exclude) = exclude {
3754 self.state
3755 .lock()
3756 .await
3757 .snapshot
3758 .repo_exclude_by_work_dir_abs_path
3759 .insert(root_abs_path.as_path().into(), (exclude, false));
3760 }
3761
3762 repo
3763 } else {
3764 None
3765 };
3766
3767 let containing_git_repository = if let Some((ancestor_dot_git, work_directory)) = repo
3768 && self.scanning_enabled
3769 {
3770 maybe!(async {
3771 self.state
3772 .lock()
3773 .await
3774 .insert_git_repository_for_path(
3775 work_directory,
3776 ancestor_dot_git.clone().into(),
3777 self.fs.as_ref(),
3778 self.watcher.as_ref(),
3779 )
3780 .await
3781 .log_err()?;
3782 Some(ancestor_dot_git)
3783 })
3784 .await
3785 } else {
3786 None
3787 };
3788
3789 log::trace!("containing git repository: {containing_git_repository:?}");
3790
3791 let mut global_gitignore_events = if let Some(global_gitignore_path) =
3792 &paths::global_gitignore_path()
3793 && self.scanning_enabled
3794 {
3795 let is_file = self.fs.is_file(&global_gitignore_path).await;
3796 self.state.lock().await.snapshot.global_gitignore = if is_file {
3797 build_gitignore(global_gitignore_path, self.fs.as_ref())
3798 .await
3799 .ok()
3800 .map(Arc::new)
3801 } else {
3802 None
3803 };
3804 if is_file
3805 || matches!(global_gitignore_path.parent(), Some(path) if self.fs.is_dir(path).await)
3806 {
3807 self.fs
3808 .watch(global_gitignore_path, FS_WATCH_LATENCY)
3809 .await
3810 .0
3811 } else {
3812 Box::pin(futures::stream::pending())
3813 }
3814 } else {
3815 self.state.lock().await.snapshot.global_gitignore = None;
3816 Box::pin(futures::stream::pending())
3817 };
3818
3819 let (scan_job_tx, scan_job_rx) = channel::unbounded();
3820 {
3821 let mut state = self.state.lock().await;
3822 state.snapshot.scan_id += 1;
3823 if let Some(mut root_entry) = state.snapshot.root_entry().cloned() {
3824 let ignore_stack = state
3825 .snapshot
3826 .ignore_stack_for_abs_path(root_abs_path.as_path(), true, self.fs.as_ref())
3827 .await;
3828 if ignore_stack.is_abs_path_ignored(root_abs_path.as_path(), true) {
3829 root_entry.is_ignored = true;
3830 let mut root_entry = root_entry.clone();
3831 state.reuse_entry_id(&mut root_entry);
3832 state
3833 .insert_entry(root_entry, self.fs.as_ref(), self.watcher.as_ref())
3834 .await;
3835 }
3836 if root_entry.is_dir() && self.scanning_enabled {
3837 state
3838 .enqueue_scan_dir(
3839 root_abs_path.as_path().into(),
3840 &root_entry,
3841 &scan_job_tx,
3842 self.fs.as_ref(),
3843 )
3844 .await;
3845 }
3846 }
3847 };
3848
3849 // Perform an initial scan of the directory.
3850 drop(scan_job_tx);
3851 self.scan_dirs(true, scan_job_rx).await;
3852 {
3853 let mut state = self.state.lock().await;
3854 state.snapshot.completed_scan_id = state.snapshot.scan_id;
3855 }
3856
3857 self.send_status_update(false, SmallVec::new()).await;
3858
3859 // Process any any FS events that occurred while performing the initial scan.
3860 // For these events, update events cannot be as precise, because we didn't
3861 // have the previous state loaded yet.
3862 self.phase = BackgroundScannerPhase::EventsReceivedDuringInitialScan;
3863 if let Poll::Ready(Some(mut paths)) = futures::poll!(fs_events_rx.next()) {
3864 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
3865 paths.extend(more_paths);
3866 }
3867 self.process_events(
3868 paths
3869 .into_iter()
3870 .filter(|e| e.kind.is_some())
3871 .map(Into::into)
3872 .collect(),
3873 )
3874 .await;
3875 }
3876 if let Some(abs_path) = containing_git_repository {
3877 self.process_events(vec![abs_path]).await;
3878 }
3879
3880 // Continue processing events until the worktree is dropped.
3881 self.phase = BackgroundScannerPhase::Events;
3882
3883 loop {
3884 select_biased! {
3885 // Process any path refresh requests from the worktree. Prioritize
3886 // these before handling changes reported by the filesystem.
3887 request = self.next_scan_request().fuse() => {
3888 let Ok(request) = request else { break };
3889 if !self.process_scan_request(request, false).await {
3890 return;
3891 }
3892 }
3893
3894 path_prefix_request = self.path_prefixes_to_scan_rx.recv().fuse() => {
3895 let Ok(request) = path_prefix_request else { break };
3896 log::trace!("adding path prefix {:?}", request.path);
3897
3898 let did_scan = self.forcibly_load_paths(std::slice::from_ref(&request.path)).await;
3899 if did_scan {
3900 let abs_path =
3901 {
3902 let mut state = self.state.lock().await;
3903 state.path_prefixes_to_scan.insert(request.path.clone());
3904 state.snapshot.absolutize(&request.path)
3905 };
3906
3907 if let Some(abs_path) = self.fs.canonicalize(&abs_path).await.log_err() {
3908 self.process_events(vec![abs_path]).await;
3909 }
3910 }
3911 self.send_status_update(false, request.done).await;
3912 }
3913
3914 paths = fs_events_rx.next().fuse() => {
3915 let Some(mut paths) = paths else { break };
3916 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
3917 paths.extend(more_paths);
3918 }
3919 self.process_events(paths.into_iter().filter(|e| e.kind.is_some()).map(Into::into).collect()).await;
3920 }
3921
3922 paths = global_gitignore_events.next().fuse() => {
3923 match paths.as_deref() {
3924 Some([event, ..]) => {
3925 self.update_global_gitignore(&event.path).await;
3926 }
3927 _ => (),
3928 }
3929 }
3930 }
3931 }
3932 }
3933
3934 async fn process_scan_request(&self, mut request: ScanRequest, scanning: bool) -> bool {
3935 log::debug!("rescanning paths {:?}", request.relative_paths);
3936
3937 request.relative_paths.sort_unstable();
3938 self.forcibly_load_paths(&request.relative_paths).await;
3939
3940 let root_path = self.state.lock().await.snapshot.abs_path.clone();
3941 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
3942 let root_canonical_path = match &root_canonical_path {
3943 Ok(path) => SanitizedPath::new(path),
3944 Err(err) => {
3945 log::error!("failed to canonicalize root path {root_path:?}: {err:#}");
3946 return true;
3947 }
3948 };
3949 let abs_paths = request
3950 .relative_paths
3951 .iter()
3952 .map(|path| {
3953 if path.file_name().is_some() {
3954 root_canonical_path.as_path().join(path.as_std_path())
3955 } else {
3956 root_canonical_path.as_path().to_path_buf()
3957 }
3958 })
3959 .collect::<Vec<_>>();
3960
3961 {
3962 let mut state = self.state.lock().await;
3963 let is_idle = state.snapshot.completed_scan_id == state.snapshot.scan_id;
3964 state.snapshot.scan_id += 1;
3965 if is_idle {
3966 state.snapshot.completed_scan_id = state.snapshot.scan_id;
3967 }
3968 }
3969
3970 self.reload_entries_for_paths(
3971 &root_path,
3972 &root_canonical_path,
3973 &request.relative_paths,
3974 abs_paths,
3975 None,
3976 )
3977 .await;
3978
3979 self.send_status_update(scanning, request.done).await
3980 }
3981
3982 async fn process_events(&self, mut abs_paths: Vec<PathBuf>) {
3983 log::trace!("process events: {abs_paths:?}");
3984 let root_path = self.state.lock().await.snapshot.abs_path.clone();
3985 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
3986 let root_canonical_path = match &root_canonical_path {
3987 Ok(path) => SanitizedPath::new(path),
3988 Err(err) => {
3989 let new_path = self
3990 .state
3991 .lock()
3992 .await
3993 .snapshot
3994 .root_file_handle
3995 .clone()
3996 .and_then(|handle| match handle.current_path(&self.fs) {
3997 Ok(new_path) => Some(new_path),
3998 Err(e) => {
3999 log::error!("Failed to refresh worktree root path: {e:#}");
4000 None
4001 }
4002 })
4003 .map(|path| SanitizedPath::new_arc(&path))
4004 .filter(|new_path| *new_path != root_path);
4005
4006 if let Some(new_path) = new_path {
4007 log::info!(
4008 "root renamed from {:?} to {:?}",
4009 root_path.as_path(),
4010 new_path.as_path(),
4011 );
4012 self.status_updates_tx
4013 .unbounded_send(ScanState::RootUpdated { new_path })
4014 .ok();
4015 } else {
4016 log::error!("root path could not be canonicalized: {err:#}");
4017 }
4018 return;
4019 }
4020 };
4021
4022 // Certain directories may have FS changes, but do not lead to git data changes that Zed cares about.
4023 // Ignore these, to avoid Zed unnecessarily rescanning git metadata.
4024 let skipped_files_in_dot_git = [COMMIT_MESSAGE, INDEX_LOCK];
4025 let skipped_dirs_in_dot_git = [FSMONITOR_DAEMON, LFS_DIR];
4026
4027 let mut relative_paths = Vec::with_capacity(abs_paths.len());
4028 let mut dot_git_abs_paths = Vec::new();
4029 let mut work_dirs_needing_exclude_update = Vec::new();
4030 abs_paths.sort_unstable();
4031 abs_paths.dedup_by(|a, b| a.starts_with(b));
4032 {
4033 let snapshot = &self.state.lock().await.snapshot;
4034
4035 let mut ranges_to_drop = SmallVec::<[Range<usize>; 4]>::new();
4036
4037 fn skip_ix(ranges: &mut SmallVec<[Range<usize>; 4]>, ix: usize) {
4038 if let Some(last_range) = ranges.last_mut()
4039 && last_range.end == ix
4040 {
4041 last_range.end += 1;
4042 } else {
4043 ranges.push(ix..ix + 1);
4044 }
4045 }
4046
4047 for (ix, abs_path) in abs_paths.iter().enumerate() {
4048 let abs_path = &SanitizedPath::new(&abs_path);
4049
4050 let mut is_git_related = false;
4051 let mut dot_git_paths = None;
4052
4053 for ancestor in abs_path.as_path().ancestors() {
4054 if is_git_dir(ancestor, self.fs.as_ref()).await {
4055 let path_in_git_dir = abs_path
4056 .as_path()
4057 .strip_prefix(ancestor)
4058 .expect("stripping off the ancestor");
4059 dot_git_paths = Some((ancestor.to_owned(), path_in_git_dir.to_owned()));
4060 break;
4061 }
4062 }
4063
4064 if let Some((dot_git_abs_path, path_in_git_dir)) = dot_git_paths {
4065 if skipped_files_in_dot_git
4066 .iter()
4067 .any(|skipped| OsStr::new(skipped) == path_in_git_dir.as_path().as_os_str())
4068 || skipped_dirs_in_dot_git.iter().any(|skipped_git_subdir| {
4069 path_in_git_dir.starts_with(skipped_git_subdir)
4070 })
4071 {
4072 log::debug!(
4073 "ignoring event {abs_path:?} as it's in the .git directory among skipped files or directories"
4074 );
4075 skip_ix(&mut ranges_to_drop, ix);
4076 continue;
4077 }
4078
4079 is_git_related = true;
4080 if !dot_git_abs_paths.contains(&dot_git_abs_path) {
4081 dot_git_abs_paths.push(dot_git_abs_path);
4082 }
4083 }
4084
4085 let relative_path = if let Ok(path) = abs_path.strip_prefix(&root_canonical_path)
4086 && let Ok(path) = RelPath::new(path, PathStyle::local())
4087 {
4088 path
4089 } else {
4090 if is_git_related {
4091 log::debug!(
4092 "ignoring event {abs_path:?}, since it's in git dir outside of root path {root_canonical_path:?}",
4093 );
4094 } else {
4095 log::error!(
4096 "ignoring event {abs_path:?} outside of root path {root_canonical_path:?}",
4097 );
4098 }
4099 skip_ix(&mut ranges_to_drop, ix);
4100 continue;
4101 };
4102
4103 let absolute_path = abs_path.to_path_buf();
4104 if absolute_path.ends_with(Path::new(DOT_GIT).join(REPO_EXCLUDE)) {
4105 if let Some(repository) = snapshot
4106 .git_repositories
4107 .values()
4108 .find(|repo| repo.common_dir_abs_path.join(REPO_EXCLUDE) == absolute_path)
4109 {
4110 work_dirs_needing_exclude_update
4111 .push(repository.work_directory_abs_path.clone());
4112 }
4113 }
4114
4115 if abs_path.file_name() == Some(OsStr::new(GITIGNORE)) {
4116 for (_, repo) in snapshot
4117 .git_repositories
4118 .iter()
4119 .filter(|(_, repo)| repo.directory_contains(&relative_path))
4120 {
4121 if !dot_git_abs_paths.iter().any(|dot_git_abs_path| {
4122 dot_git_abs_path == repo.common_dir_abs_path.as_ref()
4123 }) {
4124 dot_git_abs_paths.push(repo.common_dir_abs_path.to_path_buf());
4125 }
4126 }
4127 }
4128
4129 let parent_dir_is_loaded = relative_path.parent().is_none_or(|parent| {
4130 snapshot
4131 .entry_for_path(parent)
4132 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4133 });
4134 if !parent_dir_is_loaded {
4135 log::debug!("ignoring event {relative_path:?} within unloaded directory");
4136 skip_ix(&mut ranges_to_drop, ix);
4137 continue;
4138 }
4139
4140 if self.settings.is_path_excluded(&relative_path) {
4141 if !is_git_related {
4142 log::debug!("ignoring FS event for excluded path {relative_path:?}");
4143 }
4144 skip_ix(&mut ranges_to_drop, ix);
4145 continue;
4146 }
4147
4148 relative_paths.push(relative_path.into_arc());
4149 }
4150
4151 for range_to_drop in ranges_to_drop.into_iter().rev() {
4152 abs_paths.drain(range_to_drop);
4153 }
4154 }
4155
4156 if relative_paths.is_empty() && dot_git_abs_paths.is_empty() {
4157 return;
4158 }
4159
4160 if !work_dirs_needing_exclude_update.is_empty() {
4161 let mut state = self.state.lock().await;
4162 for work_dir_abs_path in work_dirs_needing_exclude_update {
4163 if let Some((_, needs_update)) = state
4164 .snapshot
4165 .repo_exclude_by_work_dir_abs_path
4166 .get_mut(&work_dir_abs_path)
4167 {
4168 *needs_update = true;
4169 }
4170 }
4171 }
4172
4173 self.state.lock().await.snapshot.scan_id += 1;
4174
4175 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4176 log::debug!("received fs events {:?}", relative_paths);
4177 self.reload_entries_for_paths(
4178 &root_path,
4179 &root_canonical_path,
4180 &relative_paths,
4181 abs_paths,
4182 Some(scan_job_tx.clone()),
4183 )
4184 .await;
4185
4186 let affected_repo_roots = if !dot_git_abs_paths.is_empty() {
4187 self.update_git_repositories(dot_git_abs_paths).await
4188 } else {
4189 Vec::new()
4190 };
4191
4192 {
4193 let mut ignores_to_update = self.ignores_needing_update().await;
4194 ignores_to_update.extend(affected_repo_roots);
4195 let ignores_to_update = self.order_ignores(ignores_to_update).await;
4196 let snapshot = self.state.lock().await.snapshot.clone();
4197 self.update_ignore_statuses_for_paths(scan_job_tx, snapshot, ignores_to_update)
4198 .await;
4199 self.scan_dirs(false, scan_job_rx).await;
4200 }
4201
4202 {
4203 let mut state = self.state.lock().await;
4204 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4205 for (_, entry) in mem::take(&mut state.removed_entries) {
4206 state.scanned_dirs.remove(&entry.id);
4207 }
4208 }
4209 self.send_status_update(false, SmallVec::new()).await;
4210 }
4211
4212 async fn update_global_gitignore(&self, abs_path: &Path) {
4213 let ignore = build_gitignore(abs_path, self.fs.as_ref())
4214 .await
4215 .log_err()
4216 .map(Arc::new);
4217 let (prev_snapshot, ignore_stack, abs_path) = {
4218 let mut state = self.state.lock().await;
4219 state.snapshot.global_gitignore = ignore;
4220 let abs_path = state.snapshot.abs_path().clone();
4221 let ignore_stack = state
4222 .snapshot
4223 .ignore_stack_for_abs_path(&abs_path, true, self.fs.as_ref())
4224 .await;
4225 (state.snapshot.clone(), ignore_stack, abs_path)
4226 };
4227 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4228 self.update_ignore_statuses_for_paths(
4229 scan_job_tx,
4230 prev_snapshot,
4231 vec![(abs_path, ignore_stack)],
4232 )
4233 .await;
4234 self.scan_dirs(false, scan_job_rx).await;
4235 self.send_status_update(false, SmallVec::new()).await;
4236 }
4237
4238 async fn forcibly_load_paths(&self, paths: &[Arc<RelPath>]) -> bool {
4239 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4240 {
4241 let mut state = self.state.lock().await;
4242 let root_path = state.snapshot.abs_path.clone();
4243 for path in paths {
4244 for ancestor in path.ancestors() {
4245 if let Some(entry) = state.snapshot.entry_for_path(ancestor)
4246 && entry.kind == EntryKind::UnloadedDir
4247 {
4248 let abs_path = root_path.join(ancestor.as_std_path());
4249 state
4250 .enqueue_scan_dir(
4251 abs_path.into(),
4252 entry,
4253 &scan_job_tx,
4254 self.fs.as_ref(),
4255 )
4256 .await;
4257 state.paths_to_scan.insert(path.clone());
4258 break;
4259 }
4260 }
4261 }
4262 drop(scan_job_tx);
4263 }
4264 while let Ok(job) = scan_job_rx.recv().await {
4265 self.scan_dir(&job).await.log_err();
4266 }
4267
4268 !mem::take(&mut self.state.lock().await.paths_to_scan).is_empty()
4269 }
4270
4271 async fn scan_dirs(
4272 &self,
4273 enable_progress_updates: bool,
4274 scan_jobs_rx: channel::Receiver<ScanJob>,
4275 ) {
4276 if self
4277 .status_updates_tx
4278 .unbounded_send(ScanState::Started)
4279 .is_err()
4280 {
4281 return;
4282 }
4283
4284 let progress_update_count = AtomicUsize::new(0);
4285 self.executor
4286 .scoped_priority(Priority::Low, |scope| {
4287 for _ in 0..self.executor.num_cpus() {
4288 scope.spawn(async {
4289 let mut last_progress_update_count = 0;
4290 let progress_update_timer = self.progress_timer(enable_progress_updates).fuse();
4291 futures::pin_mut!(progress_update_timer);
4292
4293 loop {
4294 select_biased! {
4295 // Process any path refresh requests before moving on to process
4296 // the scan queue, so that user operations are prioritized.
4297 request = self.next_scan_request().fuse() => {
4298 let Ok(request) = request else { break };
4299 if !self.process_scan_request(request, true).await {
4300 return;
4301 }
4302 }
4303
4304 // Send periodic progress updates to the worktree. Use an atomic counter
4305 // to ensure that only one of the workers sends a progress update after
4306 // the update interval elapses.
4307 _ = progress_update_timer => {
4308 match progress_update_count.compare_exchange(
4309 last_progress_update_count,
4310 last_progress_update_count + 1,
4311 SeqCst,
4312 SeqCst
4313 ) {
4314 Ok(_) => {
4315 last_progress_update_count += 1;
4316 self.send_status_update(true, SmallVec::new()).await;
4317 }
4318 Err(count) => {
4319 last_progress_update_count = count;
4320 }
4321 }
4322 progress_update_timer.set(self.progress_timer(enable_progress_updates).fuse());
4323 }
4324
4325 // Recursively load directories from the file system.
4326 job = scan_jobs_rx.recv().fuse() => {
4327 let Ok(job) = job else { break };
4328 if let Err(err) = self.scan_dir(&job).await
4329 && job.path.is_empty() {
4330 log::error!("error scanning directory {:?}: {}", job.abs_path, err);
4331 }
4332 }
4333 }
4334 }
4335 });
4336 }
4337 })
4338 .await;
4339 }
4340
4341 async fn send_status_update(
4342 &self,
4343 scanning: bool,
4344 barrier: SmallVec<[barrier::Sender; 1]>,
4345 ) -> bool {
4346 let mut state = self.state.lock().await;
4347 if state.changed_paths.is_empty() && scanning {
4348 return true;
4349 }
4350
4351 let new_snapshot = state.snapshot.clone();
4352 let old_snapshot = mem::replace(&mut state.prev_snapshot, new_snapshot.snapshot.clone());
4353 let changes = build_diff(
4354 self.phase,
4355 &old_snapshot,
4356 &new_snapshot,
4357 &state.changed_paths,
4358 );
4359 state.changed_paths.clear();
4360
4361 self.status_updates_tx
4362 .unbounded_send(ScanState::Updated {
4363 snapshot: new_snapshot,
4364 changes,
4365 scanning,
4366 barrier,
4367 })
4368 .is_ok()
4369 }
4370
4371 async fn scan_dir(&self, job: &ScanJob) -> Result<()> {
4372 let root_abs_path;
4373 let root_char_bag;
4374 {
4375 let snapshot = &self.state.lock().await.snapshot;
4376 if self.settings.is_path_excluded(&job.path) {
4377 log::error!("skipping excluded directory {:?}", job.path);
4378 return Ok(());
4379 }
4380 log::trace!("scanning directory {:?}", job.path);
4381 root_abs_path = snapshot.abs_path().clone();
4382 root_char_bag = snapshot.root_char_bag;
4383 }
4384
4385 let next_entry_id = self.next_entry_id.clone();
4386 let mut ignore_stack = job.ignore_stack.clone();
4387 let mut new_ignore = None;
4388 let mut root_canonical_path = None;
4389 let mut new_entries: Vec<Entry> = Vec::new();
4390 let mut new_jobs: Vec<Option<ScanJob>> = Vec::new();
4391 let mut child_paths = self
4392 .fs
4393 .read_dir(&job.abs_path)
4394 .await?
4395 .filter_map(|entry| async {
4396 match entry {
4397 Ok(entry) => Some(entry),
4398 Err(error) => {
4399 log::error!("error processing entry {:?}", error);
4400 None
4401 }
4402 }
4403 })
4404 .collect::<Vec<_>>()
4405 .await;
4406
4407 // Ensure that .git and .gitignore are processed first.
4408 swap_to_front(&mut child_paths, GITIGNORE);
4409 swap_to_front(&mut child_paths, DOT_GIT);
4410
4411 if let Some(path) = child_paths.first()
4412 && path.ends_with(DOT_GIT)
4413 {
4414 ignore_stack.repo_root = Some(job.abs_path.clone());
4415 }
4416
4417 for child_abs_path in child_paths {
4418 let child_abs_path: Arc<Path> = child_abs_path.into();
4419 let child_name = child_abs_path.file_name().unwrap();
4420 let Some(child_path) = child_name
4421 .to_str()
4422 .and_then(|name| Some(job.path.join(RelPath::unix(name).ok()?)))
4423 else {
4424 continue;
4425 };
4426
4427 if child_name == DOT_GIT {
4428 let mut state = self.state.lock().await;
4429 state
4430 .insert_git_repository(
4431 child_path.clone(),
4432 self.fs.as_ref(),
4433 self.watcher.as_ref(),
4434 )
4435 .await;
4436 } else if child_name == GITIGNORE {
4437 match build_gitignore(&child_abs_path, self.fs.as_ref()).await {
4438 Ok(ignore) => {
4439 let ignore = Arc::new(ignore);
4440 ignore_stack = ignore_stack
4441 .append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4442 new_ignore = Some(ignore);
4443 }
4444 Err(error) => {
4445 log::error!(
4446 "error loading .gitignore file {:?} - {:?}",
4447 child_name,
4448 error
4449 );
4450 }
4451 }
4452 }
4453
4454 if self.settings.is_path_excluded(&child_path) {
4455 log::debug!("skipping excluded child entry {child_path:?}");
4456 self.state.lock().await.remove_path(&child_path);
4457 continue;
4458 }
4459
4460 let child_metadata = match self.fs.metadata(&child_abs_path).await {
4461 Ok(Some(metadata)) => metadata,
4462 Ok(None) => continue,
4463 Err(err) => {
4464 log::error!("error processing {child_abs_path:?}: {err:#}");
4465 continue;
4466 }
4467 };
4468
4469 let mut child_entry = Entry::new(
4470 child_path.clone(),
4471 &child_metadata,
4472 ProjectEntryId::new(&next_entry_id),
4473 root_char_bag,
4474 None,
4475 );
4476
4477 if job.is_external {
4478 child_entry.is_external = true;
4479 } else if child_metadata.is_symlink {
4480 let canonical_path = match self.fs.canonicalize(&child_abs_path).await {
4481 Ok(path) => path,
4482 Err(err) => {
4483 log::error!("error reading target of symlink {child_abs_path:?}: {err:#}",);
4484 continue;
4485 }
4486 };
4487
4488 // lazily canonicalize the root path in order to determine if
4489 // symlinks point outside of the worktree.
4490 let root_canonical_path = match &root_canonical_path {
4491 Some(path) => path,
4492 None => match self.fs.canonicalize(&root_abs_path).await {
4493 Ok(path) => root_canonical_path.insert(path),
4494 Err(err) => {
4495 log::error!("error canonicalizing root {:?}: {:?}", root_abs_path, err);
4496 continue;
4497 }
4498 },
4499 };
4500
4501 if !canonical_path.starts_with(root_canonical_path) {
4502 child_entry.is_external = true;
4503 }
4504
4505 child_entry.canonical_path = Some(canonical_path.into());
4506 }
4507
4508 if child_entry.is_dir() {
4509 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, true);
4510 child_entry.is_always_included =
4511 self.settings.is_path_always_included(&child_path, true);
4512
4513 // Avoid recursing until crash in the case of a recursive symlink
4514 if job.ancestor_inodes.contains(&child_entry.inode) {
4515 new_jobs.push(None);
4516 } else {
4517 let mut ancestor_inodes = job.ancestor_inodes.clone();
4518 ancestor_inodes.insert(child_entry.inode);
4519
4520 new_jobs.push(Some(ScanJob {
4521 abs_path: child_abs_path.clone(),
4522 path: child_path,
4523 is_external: child_entry.is_external,
4524 ignore_stack: if child_entry.is_ignored {
4525 IgnoreStack::all()
4526 } else {
4527 ignore_stack.clone()
4528 },
4529 ancestor_inodes,
4530 scan_queue: job.scan_queue.clone(),
4531 }));
4532 }
4533 } else {
4534 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, false);
4535 child_entry.is_always_included =
4536 self.settings.is_path_always_included(&child_path, false);
4537 }
4538
4539 {
4540 let relative_path = job
4541 .path
4542 .join(RelPath::unix(child_name.to_str().unwrap()).unwrap());
4543 if self.is_path_private(&relative_path) {
4544 log::debug!("detected private file: {relative_path:?}");
4545 child_entry.is_private = true;
4546 }
4547 if self.settings.is_path_hidden(&relative_path) {
4548 log::debug!("detected hidden file: {relative_path:?}");
4549 child_entry.is_hidden = true;
4550 }
4551 }
4552
4553 new_entries.push(child_entry);
4554 }
4555
4556 let mut state = self.state.lock().await;
4557
4558 // Identify any subdirectories that should not be scanned.
4559 let mut job_ix = 0;
4560 for entry in &mut new_entries {
4561 state.reuse_entry_id(entry);
4562 if entry.is_dir() {
4563 if state.should_scan_directory(entry) {
4564 job_ix += 1;
4565 } else {
4566 log::debug!("defer scanning directory {:?}", entry.path);
4567 entry.kind = EntryKind::UnloadedDir;
4568 new_jobs.remove(job_ix);
4569 }
4570 }
4571 if entry.is_always_included {
4572 state
4573 .snapshot
4574 .always_included_entries
4575 .push(entry.path.clone());
4576 }
4577 }
4578
4579 state.populate_dir(job.path.clone(), new_entries, new_ignore);
4580 self.watcher.add(job.abs_path.as_ref()).log_err();
4581
4582 for new_job in new_jobs.into_iter().flatten() {
4583 job.scan_queue
4584 .try_send(new_job)
4585 .expect("channel is unbounded");
4586 }
4587
4588 Ok(())
4589 }
4590
4591 /// All list arguments should be sorted before calling this function
4592 async fn reload_entries_for_paths(
4593 &self,
4594 root_abs_path: &SanitizedPath,
4595 root_canonical_path: &SanitizedPath,
4596 relative_paths: &[Arc<RelPath>],
4597 abs_paths: Vec<PathBuf>,
4598 scan_queue_tx: Option<Sender<ScanJob>>,
4599 ) {
4600 // grab metadata for all requested paths
4601 let metadata = futures::future::join_all(
4602 abs_paths
4603 .iter()
4604 .map(|abs_path| async move {
4605 let metadata = self.fs.metadata(abs_path).await?;
4606 if let Some(metadata) = metadata {
4607 let canonical_path = self.fs.canonicalize(abs_path).await?;
4608
4609 // If we're on a case-insensitive filesystem (default on macOS), we want
4610 // to only ignore metadata for non-symlink files if their absolute-path matches
4611 // the canonical-path.
4612 // Because if not, this might be a case-only-renaming (`mv test.txt TEST.TXT`)
4613 // and we want to ignore the metadata for the old path (`test.txt`) so it's
4614 // treated as removed.
4615 if !self.fs_case_sensitive && !metadata.is_symlink {
4616 let canonical_file_name = canonical_path.file_name();
4617 let file_name = abs_path.file_name();
4618 if canonical_file_name != file_name {
4619 return Ok(None);
4620 }
4621 }
4622
4623 anyhow::Ok(Some((metadata, SanitizedPath::new_arc(&canonical_path))))
4624 } else {
4625 Ok(None)
4626 }
4627 })
4628 .collect::<Vec<_>>(),
4629 )
4630 .await;
4631
4632 let mut new_ancestor_repo = if relative_paths.iter().any(|path| path.is_empty()) {
4633 Some(discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await)
4634 } else {
4635 None
4636 };
4637
4638 let mut state = self.state.lock().await;
4639 let doing_recursive_update = scan_queue_tx.is_some();
4640
4641 // Remove any entries for paths that no longer exist or are being recursively
4642 // refreshed. Do this before adding any new entries, so that renames can be
4643 // detected regardless of the order of the paths.
4644 for (path, metadata) in relative_paths.iter().zip(metadata.iter()) {
4645 if matches!(metadata, Ok(None)) || doing_recursive_update {
4646 state.remove_path(path);
4647 }
4648 }
4649
4650 for (path, metadata) in relative_paths.iter().zip(metadata.into_iter()) {
4651 let abs_path: Arc<Path> = root_abs_path.join(path.as_std_path()).into();
4652 match metadata {
4653 Ok(Some((metadata, canonical_path))) => {
4654 let ignore_stack = state
4655 .snapshot
4656 .ignore_stack_for_abs_path(&abs_path, metadata.is_dir, self.fs.as_ref())
4657 .await;
4658 let is_external = !canonical_path.starts_with(&root_canonical_path);
4659 let entry_id = state.entry_id_for(self.next_entry_id.as_ref(), path, &metadata);
4660 let mut fs_entry = Entry::new(
4661 path.clone(),
4662 &metadata,
4663 entry_id,
4664 state.snapshot.root_char_bag,
4665 if metadata.is_symlink {
4666 Some(canonical_path.as_path().to_path_buf().into())
4667 } else {
4668 None
4669 },
4670 );
4671
4672 let is_dir = fs_entry.is_dir();
4673 fs_entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, is_dir);
4674 fs_entry.is_external = is_external;
4675 fs_entry.is_private = self.is_path_private(path);
4676 fs_entry.is_always_included =
4677 self.settings.is_path_always_included(path, is_dir);
4678 fs_entry.is_hidden = self.settings.is_path_hidden(path);
4679
4680 if let (Some(scan_queue_tx), true) = (&scan_queue_tx, is_dir) {
4681 if state.should_scan_directory(&fs_entry)
4682 || (fs_entry.path.is_empty()
4683 && abs_path.file_name() == Some(OsStr::new(DOT_GIT)))
4684 {
4685 state
4686 .enqueue_scan_dir(
4687 abs_path,
4688 &fs_entry,
4689 scan_queue_tx,
4690 self.fs.as_ref(),
4691 )
4692 .await;
4693 } else {
4694 fs_entry.kind = EntryKind::UnloadedDir;
4695 }
4696 }
4697
4698 state
4699 .insert_entry(fs_entry.clone(), self.fs.as_ref(), self.watcher.as_ref())
4700 .await;
4701
4702 if path.is_empty()
4703 && let Some((ignores, exclude, repo)) = new_ancestor_repo.take()
4704 {
4705 log::trace!("updating ancestor git repository");
4706 state.snapshot.ignores_by_parent_abs_path.extend(ignores);
4707 if let Some((ancestor_dot_git, work_directory)) = repo {
4708 if let Some(exclude) = exclude {
4709 let work_directory_abs_path = self
4710 .state
4711 .lock()
4712 .await
4713 .snapshot
4714 .work_directory_abs_path(&work_directory);
4715
4716 state
4717 .snapshot
4718 .repo_exclude_by_work_dir_abs_path
4719 .insert(work_directory_abs_path.into(), (exclude, false));
4720 }
4721 state
4722 .insert_git_repository_for_path(
4723 work_directory,
4724 ancestor_dot_git.into(),
4725 self.fs.as_ref(),
4726 self.watcher.as_ref(),
4727 )
4728 .await
4729 .log_err();
4730 }
4731 }
4732 }
4733 Ok(None) => {
4734 self.remove_repo_path(path.clone(), &mut state.snapshot);
4735 }
4736 Err(err) => {
4737 log::error!("error reading file {abs_path:?} on event: {err:#}");
4738 }
4739 }
4740 }
4741
4742 util::extend_sorted(
4743 &mut state.changed_paths,
4744 relative_paths.iter().cloned(),
4745 usize::MAX,
4746 Ord::cmp,
4747 );
4748 }
4749
4750 fn remove_repo_path(&self, path: Arc<RelPath>, snapshot: &mut LocalSnapshot) -> Option<()> {
4751 if !path.components().any(|component| component == DOT_GIT)
4752 && let Some(local_repo) = snapshot.local_repo_for_work_directory_path(&path)
4753 {
4754 let id = local_repo.work_directory_id;
4755 log::debug!("remove repo path: {:?}", path);
4756 snapshot.git_repositories.remove(&id);
4757 return Some(());
4758 }
4759
4760 Some(())
4761 }
4762
4763 async fn update_ignore_statuses_for_paths(
4764 &self,
4765 scan_job_tx: Sender<ScanJob>,
4766 prev_snapshot: LocalSnapshot,
4767 ignores_to_update: Vec<(Arc<Path>, IgnoreStack)>,
4768 ) {
4769 let (ignore_queue_tx, ignore_queue_rx) = channel::unbounded();
4770 {
4771 for (parent_abs_path, ignore_stack) in ignores_to_update {
4772 ignore_queue_tx
4773 .send_blocking(UpdateIgnoreStatusJob {
4774 abs_path: parent_abs_path,
4775 ignore_stack,
4776 ignore_queue: ignore_queue_tx.clone(),
4777 scan_queue: scan_job_tx.clone(),
4778 })
4779 .unwrap();
4780 }
4781 }
4782 drop(ignore_queue_tx);
4783
4784 self.executor
4785 .scoped(|scope| {
4786 for _ in 0..self.executor.num_cpus() {
4787 scope.spawn(async {
4788 loop {
4789 select_biased! {
4790 // Process any path refresh requests before moving on to process
4791 // the queue of ignore statuses.
4792 request = self.next_scan_request().fuse() => {
4793 let Ok(request) = request else { break };
4794 if !self.process_scan_request(request, true).await {
4795 return;
4796 }
4797 }
4798
4799 // Recursively process directories whose ignores have changed.
4800 job = ignore_queue_rx.recv().fuse() => {
4801 let Ok(job) = job else { break };
4802 self.update_ignore_status(job, &prev_snapshot).await;
4803 }
4804 }
4805 }
4806 });
4807 }
4808 })
4809 .await;
4810 }
4811
4812 async fn ignores_needing_update(&self) -> Vec<Arc<Path>> {
4813 let mut ignores_to_update = Vec::new();
4814
4815 {
4816 let snapshot = &mut self.state.lock().await.snapshot;
4817 let abs_path = snapshot.abs_path.clone();
4818
4819 snapshot.repo_exclude_by_work_dir_abs_path.retain(
4820 |work_dir_abs_path, (exclude, needs_update)| {
4821 if *needs_update {
4822 *needs_update = false;
4823 ignores_to_update.push(work_dir_abs_path.clone());
4824
4825 if let Some((_, repository)) = snapshot
4826 .git_repositories
4827 .iter()
4828 .find(|(_, repo)| &repo.work_directory_abs_path == work_dir_abs_path)
4829 {
4830 let exclude_abs_path =
4831 repository.common_dir_abs_path.join(REPO_EXCLUDE);
4832 if let Ok(current_exclude) = self
4833 .executor
4834 .block(build_gitignore(&exclude_abs_path, self.fs.as_ref()))
4835 {
4836 *exclude = Arc::new(current_exclude);
4837 }
4838 }
4839 }
4840
4841 snapshot
4842 .git_repositories
4843 .iter()
4844 .any(|(_, repo)| &repo.work_directory_abs_path == work_dir_abs_path)
4845 },
4846 );
4847
4848 snapshot
4849 .ignores_by_parent_abs_path
4850 .retain(|parent_abs_path, (_, needs_update)| {
4851 if let Ok(parent_path) = parent_abs_path.strip_prefix(abs_path.as_path())
4852 && let Some(parent_path) =
4853 RelPath::new(&parent_path, PathStyle::local()).log_err()
4854 {
4855 if *needs_update {
4856 *needs_update = false;
4857 if snapshot.snapshot.entry_for_path(&parent_path).is_some() {
4858 ignores_to_update.push(parent_abs_path.clone());
4859 }
4860 }
4861
4862 let ignore_path = parent_path.join(RelPath::unix(GITIGNORE).unwrap());
4863 if snapshot.snapshot.entry_for_path(&ignore_path).is_none() {
4864 return false;
4865 }
4866 }
4867 true
4868 });
4869 }
4870
4871 ignores_to_update
4872 }
4873
4874 async fn order_ignores(&self, mut ignores: Vec<Arc<Path>>) -> Vec<(Arc<Path>, IgnoreStack)> {
4875 let fs = self.fs.clone();
4876 let snapshot = self.state.lock().await.snapshot.clone();
4877 ignores.sort_unstable();
4878 let mut ignores_to_update = ignores.into_iter().peekable();
4879
4880 let mut result = vec![];
4881 while let Some(parent_abs_path) = ignores_to_update.next() {
4882 while ignores_to_update
4883 .peek()
4884 .map_or(false, |p| p.starts_with(&parent_abs_path))
4885 {
4886 ignores_to_update.next().unwrap();
4887 }
4888 let ignore_stack = snapshot
4889 .ignore_stack_for_abs_path(&parent_abs_path, true, fs.as_ref())
4890 .await;
4891 result.push((parent_abs_path, ignore_stack));
4892 }
4893
4894 result
4895 }
4896
4897 async fn update_ignore_status(&self, job: UpdateIgnoreStatusJob, snapshot: &LocalSnapshot) {
4898 log::trace!("update ignore status {:?}", job.abs_path);
4899
4900 let mut ignore_stack = job.ignore_stack;
4901 if let Some((ignore, _)) = snapshot.ignores_by_parent_abs_path.get(&job.abs_path) {
4902 ignore_stack =
4903 ignore_stack.append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4904 }
4905
4906 let mut entries_by_id_edits = Vec::new();
4907 let mut entries_by_path_edits = Vec::new();
4908 let Some(path) = job
4909 .abs_path
4910 .strip_prefix(snapshot.abs_path.as_path())
4911 .map_err(|_| {
4912 anyhow::anyhow!(
4913 "Failed to strip prefix '{}' from path '{}'",
4914 snapshot.abs_path.as_path().display(),
4915 job.abs_path.display()
4916 )
4917 })
4918 .log_err()
4919 else {
4920 return;
4921 };
4922
4923 let Some(path) = RelPath::new(&path, PathStyle::local()).log_err() else {
4924 return;
4925 };
4926
4927 if let Ok(Some(metadata)) = self.fs.metadata(&job.abs_path.join(DOT_GIT)).await
4928 && metadata.is_dir
4929 {
4930 ignore_stack.repo_root = Some(job.abs_path.clone());
4931 }
4932
4933 for mut entry in snapshot.child_entries(&path).cloned() {
4934 let was_ignored = entry.is_ignored;
4935 let abs_path: Arc<Path> = snapshot.absolutize(&entry.path).into();
4936 entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, entry.is_dir());
4937
4938 if entry.is_dir() {
4939 let child_ignore_stack = if entry.is_ignored {
4940 IgnoreStack::all()
4941 } else {
4942 ignore_stack.clone()
4943 };
4944
4945 // Scan any directories that were previously ignored and weren't previously scanned.
4946 if was_ignored && !entry.is_ignored && entry.kind.is_unloaded() {
4947 let state = self.state.lock().await;
4948 if state.should_scan_directory(&entry) {
4949 state
4950 .enqueue_scan_dir(
4951 abs_path.clone(),
4952 &entry,
4953 &job.scan_queue,
4954 self.fs.as_ref(),
4955 )
4956 .await;
4957 }
4958 }
4959
4960 job.ignore_queue
4961 .send(UpdateIgnoreStatusJob {
4962 abs_path: abs_path.clone(),
4963 ignore_stack: child_ignore_stack,
4964 ignore_queue: job.ignore_queue.clone(),
4965 scan_queue: job.scan_queue.clone(),
4966 })
4967 .await
4968 .unwrap();
4969 }
4970
4971 if entry.is_ignored != was_ignored {
4972 let mut path_entry = snapshot.entries_by_id.get(&entry.id, ()).unwrap().clone();
4973 path_entry.scan_id = snapshot.scan_id;
4974 path_entry.is_ignored = entry.is_ignored;
4975 entries_by_id_edits.push(Edit::Insert(path_entry));
4976 entries_by_path_edits.push(Edit::Insert(entry));
4977 }
4978 }
4979
4980 let state = &mut self.state.lock().await;
4981 for edit in &entries_by_path_edits {
4982 if let Edit::Insert(entry) = edit
4983 && let Err(ix) = state.changed_paths.binary_search(&entry.path)
4984 {
4985 state.changed_paths.insert(ix, entry.path.clone());
4986 }
4987 }
4988
4989 state
4990 .snapshot
4991 .entries_by_path
4992 .edit(entries_by_path_edits, ());
4993 state.snapshot.entries_by_id.edit(entries_by_id_edits, ());
4994 }
4995
4996 async fn update_git_repositories(&self, dot_git_paths: Vec<PathBuf>) -> Vec<Arc<Path>> {
4997 log::trace!("reloading repositories: {dot_git_paths:?}");
4998 let mut state = self.state.lock().await;
4999 let scan_id = state.snapshot.scan_id;
5000 let mut affected_repo_roots = Vec::new();
5001 for dot_git_dir in dot_git_paths {
5002 let existing_repository_entry =
5003 state
5004 .snapshot
5005 .git_repositories
5006 .iter()
5007 .find_map(|(_, repo)| {
5008 let dot_git_dir = SanitizedPath::new(&dot_git_dir);
5009 if SanitizedPath::new(repo.common_dir_abs_path.as_ref()) == dot_git_dir
5010 || SanitizedPath::new(repo.repository_dir_abs_path.as_ref())
5011 == dot_git_dir
5012 {
5013 Some(repo.clone())
5014 } else {
5015 None
5016 }
5017 });
5018
5019 match existing_repository_entry {
5020 None => {
5021 let Ok(relative) = dot_git_dir.strip_prefix(state.snapshot.abs_path()) else {
5022 debug_panic!(
5023 "update_git_repositories called with .git directory outside the worktree root"
5024 );
5025 return Vec::new();
5026 };
5027 affected_repo_roots.push(dot_git_dir.parent().unwrap().into());
5028 state
5029 .insert_git_repository(
5030 RelPath::new(relative, PathStyle::local())
5031 .unwrap()
5032 .into_arc(),
5033 self.fs.as_ref(),
5034 self.watcher.as_ref(),
5035 )
5036 .await;
5037 }
5038 Some(local_repository) => {
5039 state.snapshot.git_repositories.update(
5040 &local_repository.work_directory_id,
5041 |entry| {
5042 entry.git_dir_scan_id = scan_id;
5043 },
5044 );
5045 }
5046 };
5047 }
5048
5049 // Remove any git repositories whose .git entry no longer exists.
5050 let snapshot = &mut state.snapshot;
5051 let mut ids_to_preserve = HashSet::default();
5052 for (&work_directory_id, entry) in snapshot.git_repositories.iter() {
5053 let exists_in_snapshot =
5054 snapshot
5055 .entry_for_id(work_directory_id)
5056 .is_some_and(|entry| {
5057 snapshot
5058 .entry_for_path(&entry.path.join(RelPath::unix(DOT_GIT).unwrap()))
5059 .is_some()
5060 });
5061
5062 if exists_in_snapshot
5063 || matches!(
5064 self.fs.metadata(&entry.common_dir_abs_path).await,
5065 Ok(Some(_))
5066 )
5067 {
5068 ids_to_preserve.insert(work_directory_id);
5069 }
5070 }
5071
5072 snapshot
5073 .git_repositories
5074 .retain(|work_directory_id, entry| {
5075 let preserve = ids_to_preserve.contains(work_directory_id);
5076 if !preserve {
5077 affected_repo_roots.push(entry.dot_git_abs_path.parent().unwrap().into());
5078 snapshot
5079 .repo_exclude_by_work_dir_abs_path
5080 .remove(&entry.work_directory_abs_path);
5081 }
5082 preserve
5083 });
5084
5085 affected_repo_roots
5086 }
5087
5088 async fn progress_timer(&self, running: bool) {
5089 if !running {
5090 return futures::future::pending().await;
5091 }
5092
5093 #[cfg(any(test, feature = "test-support"))]
5094 if self.fs.is_fake() {
5095 return self.executor.simulate_random_delay().await;
5096 }
5097
5098 smol::Timer::after(FS_WATCH_LATENCY).await;
5099 }
5100
5101 fn is_path_private(&self, path: &RelPath) -> bool {
5102 !self.share_private_files && self.settings.is_path_private(path)
5103 }
5104
5105 async fn next_scan_request(&self) -> Result<ScanRequest> {
5106 let mut request = self.scan_requests_rx.recv().await?;
5107 while let Ok(next_request) = self.scan_requests_rx.try_recv() {
5108 request.relative_paths.extend(next_request.relative_paths);
5109 request.done.extend(next_request.done);
5110 }
5111 Ok(request)
5112 }
5113}
5114
5115async fn discover_ancestor_git_repo(
5116 fs: Arc<dyn Fs>,
5117 root_abs_path: &SanitizedPath,
5118) -> (
5119 HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
5120 Option<Arc<Gitignore>>,
5121 Option<(PathBuf, WorkDirectory)>,
5122) {
5123 let mut exclude = None;
5124 let mut ignores = HashMap::default();
5125 for (index, ancestor) in root_abs_path.as_path().ancestors().enumerate() {
5126 if index != 0 {
5127 if ancestor == paths::home_dir() {
5128 // Unless $HOME is itself the worktree root, don't consider it as a
5129 // containing git repository---expensive and likely unwanted.
5130 break;
5131 } else if let Ok(ignore) = build_gitignore(&ancestor.join(GITIGNORE), fs.as_ref()).await
5132 {
5133 ignores.insert(ancestor.into(), (ignore.into(), false));
5134 }
5135 }
5136
5137 let ancestor_dot_git = ancestor.join(DOT_GIT);
5138 log::trace!("considering ancestor: {ancestor_dot_git:?}");
5139 // Check whether the directory or file called `.git` exists (in the
5140 // case of worktrees it's a file.)
5141 if fs
5142 .metadata(&ancestor_dot_git)
5143 .await
5144 .is_ok_and(|metadata| metadata.is_some())
5145 {
5146 if index != 0 {
5147 // We canonicalize, since the FS events use the canonicalized path.
5148 if let Some(ancestor_dot_git) = fs.canonicalize(&ancestor_dot_git).await.log_err() {
5149 let location_in_repo = root_abs_path
5150 .as_path()
5151 .strip_prefix(ancestor)
5152 .unwrap()
5153 .into();
5154 log::info!("inserting parent git repo for this worktree: {location_in_repo:?}");
5155 // We associate the external git repo with our root folder and
5156 // also mark where in the git repo the root folder is located.
5157 return (
5158 ignores,
5159 exclude,
5160 Some((
5161 ancestor_dot_git,
5162 WorkDirectory::AboveProject {
5163 absolute_path: ancestor.into(),
5164 location_in_repo,
5165 },
5166 )),
5167 );
5168 };
5169 }
5170
5171 let repo_exclude_abs_path = ancestor_dot_git.join(REPO_EXCLUDE);
5172 if let Ok(repo_exclude) = build_gitignore(&repo_exclude_abs_path, fs.as_ref()).await {
5173 exclude = Some(Arc::new(repo_exclude));
5174 }
5175
5176 // Reached root of git repository.
5177 break;
5178 }
5179 }
5180
5181 (ignores, exclude, None)
5182}
5183
5184fn build_diff(
5185 phase: BackgroundScannerPhase,
5186 old_snapshot: &Snapshot,
5187 new_snapshot: &Snapshot,
5188 event_paths: &[Arc<RelPath>],
5189) -> UpdatedEntriesSet {
5190 use BackgroundScannerPhase::*;
5191 use PathChange::{Added, AddedOrUpdated, Loaded, Removed, Updated};
5192
5193 // Identify which paths have changed. Use the known set of changed
5194 // parent paths to optimize the search.
5195 let mut changes = Vec::new();
5196 let mut old_paths = old_snapshot.entries_by_path.cursor::<PathKey>(());
5197 let mut new_paths = new_snapshot.entries_by_path.cursor::<PathKey>(());
5198 let mut last_newly_loaded_dir_path = None;
5199 old_paths.next();
5200 new_paths.next();
5201 for path in event_paths {
5202 let path = PathKey(path.clone());
5203 if old_paths.item().is_some_and(|e| e.path < path.0) {
5204 old_paths.seek_forward(&path, Bias::Left);
5205 }
5206 if new_paths.item().is_some_and(|e| e.path < path.0) {
5207 new_paths.seek_forward(&path, Bias::Left);
5208 }
5209 loop {
5210 match (old_paths.item(), new_paths.item()) {
5211 (Some(old_entry), Some(new_entry)) => {
5212 if old_entry.path > path.0
5213 && new_entry.path > path.0
5214 && !old_entry.path.starts_with(&path.0)
5215 && !new_entry.path.starts_with(&path.0)
5216 {
5217 break;
5218 }
5219
5220 match Ord::cmp(&old_entry.path, &new_entry.path) {
5221 Ordering::Less => {
5222 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5223 old_paths.next();
5224 }
5225 Ordering::Equal => {
5226 if phase == EventsReceivedDuringInitialScan {
5227 if old_entry.id != new_entry.id {
5228 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5229 }
5230 // If the worktree was not fully initialized when this event was generated,
5231 // we can't know whether this entry was added during the scan or whether
5232 // it was merely updated.
5233 changes.push((
5234 new_entry.path.clone(),
5235 new_entry.id,
5236 AddedOrUpdated,
5237 ));
5238 } else if old_entry.id != new_entry.id {
5239 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5240 changes.push((new_entry.path.clone(), new_entry.id, Added));
5241 } else if old_entry != new_entry {
5242 if old_entry.kind.is_unloaded() {
5243 last_newly_loaded_dir_path = Some(&new_entry.path);
5244 changes.push((new_entry.path.clone(), new_entry.id, Loaded));
5245 } else {
5246 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5247 }
5248 }
5249 old_paths.next();
5250 new_paths.next();
5251 }
5252 Ordering::Greater => {
5253 let is_newly_loaded = phase == InitialScan
5254 || last_newly_loaded_dir_path
5255 .as_ref()
5256 .is_some_and(|dir| new_entry.path.starts_with(dir));
5257 changes.push((
5258 new_entry.path.clone(),
5259 new_entry.id,
5260 if is_newly_loaded { Loaded } else { Added },
5261 ));
5262 new_paths.next();
5263 }
5264 }
5265 }
5266 (Some(old_entry), None) => {
5267 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5268 old_paths.next();
5269 }
5270 (None, Some(new_entry)) => {
5271 let is_newly_loaded = phase == InitialScan
5272 || last_newly_loaded_dir_path
5273 .as_ref()
5274 .is_some_and(|dir| new_entry.path.starts_with(dir));
5275 changes.push((
5276 new_entry.path.clone(),
5277 new_entry.id,
5278 if is_newly_loaded { Loaded } else { Added },
5279 ));
5280 new_paths.next();
5281 }
5282 (None, None) => break,
5283 }
5284 }
5285 }
5286
5287 changes.into()
5288}
5289
5290fn swap_to_front(child_paths: &mut Vec<PathBuf>, file: &str) {
5291 let position = child_paths
5292 .iter()
5293 .position(|path| path.file_name().unwrap() == file);
5294 if let Some(position) = position {
5295 let temp = child_paths.remove(position);
5296 child_paths.insert(0, temp);
5297 }
5298}
5299
5300fn char_bag_for_path(root_char_bag: CharBag, path: &RelPath) -> CharBag {
5301 let mut result = root_char_bag;
5302 result.extend(path.as_unix_str().chars().map(|c| c.to_ascii_lowercase()));
5303 result
5304}
5305
5306#[derive(Debug)]
5307struct ScanJob {
5308 abs_path: Arc<Path>,
5309 path: Arc<RelPath>,
5310 ignore_stack: IgnoreStack,
5311 scan_queue: Sender<ScanJob>,
5312 ancestor_inodes: TreeSet<u64>,
5313 is_external: bool,
5314}
5315
5316struct UpdateIgnoreStatusJob {
5317 abs_path: Arc<Path>,
5318 ignore_stack: IgnoreStack,
5319 ignore_queue: Sender<UpdateIgnoreStatusJob>,
5320 scan_queue: Sender<ScanJob>,
5321}
5322
5323pub trait WorktreeModelHandle {
5324 #[cfg(any(test, feature = "test-support"))]
5325 fn flush_fs_events<'a>(
5326 &self,
5327 cx: &'a mut gpui::TestAppContext,
5328 ) -> futures::future::LocalBoxFuture<'a, ()>;
5329
5330 #[cfg(any(test, feature = "test-support"))]
5331 fn flush_fs_events_in_root_git_repository<'a>(
5332 &self,
5333 cx: &'a mut gpui::TestAppContext,
5334 ) -> futures::future::LocalBoxFuture<'a, ()>;
5335}
5336
5337impl WorktreeModelHandle for Entity<Worktree> {
5338 // When the worktree's FS event stream sometimes delivers "redundant" events for FS changes that
5339 // occurred before the worktree was constructed. These events can cause the worktree to perform
5340 // extra directory scans, and emit extra scan-state notifications.
5341 //
5342 // This function mutates the worktree's directory and waits for those mutations to be picked up,
5343 // to ensure that all redundant FS events have already been processed.
5344 #[cfg(any(test, feature = "test-support"))]
5345 fn flush_fs_events<'a>(
5346 &self,
5347 cx: &'a mut gpui::TestAppContext,
5348 ) -> futures::future::LocalBoxFuture<'a, ()> {
5349 let file_name = "fs-event-sentinel";
5350
5351 let tree = self.clone();
5352 let (fs, root_path) = self.read_with(cx, |tree, _| {
5353 let tree = tree.as_local().unwrap();
5354 (tree.fs.clone(), tree.abs_path.clone())
5355 });
5356
5357 async move {
5358 fs.create_file(&root_path.join(file_name), Default::default())
5359 .await
5360 .unwrap();
5361
5362 let mut events = cx.events(&tree);
5363 while events.next().await.is_some() {
5364 if tree.read_with(cx, |tree, _| {
5365 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5366 .is_some()
5367 }) {
5368 break;
5369 }
5370 }
5371
5372 fs.remove_file(&root_path.join(file_name), Default::default())
5373 .await
5374 .unwrap();
5375 while events.next().await.is_some() {
5376 if tree.read_with(cx, |tree, _| {
5377 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5378 .is_none()
5379 }) {
5380 break;
5381 }
5382 }
5383
5384 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5385 .await;
5386 }
5387 .boxed_local()
5388 }
5389
5390 // This function is similar to flush_fs_events, except that it waits for events to be flushed in
5391 // the .git folder of the root repository.
5392 // The reason for its existence is that a repository's .git folder might live *outside* of the
5393 // worktree and thus its FS events might go through a different path.
5394 // In order to flush those, we need to create artificial events in the .git folder and wait
5395 // for the repository to be reloaded.
5396 #[cfg(any(test, feature = "test-support"))]
5397 fn flush_fs_events_in_root_git_repository<'a>(
5398 &self,
5399 cx: &'a mut gpui::TestAppContext,
5400 ) -> futures::future::LocalBoxFuture<'a, ()> {
5401 let file_name = "fs-event-sentinel";
5402
5403 let tree = self.clone();
5404 let (fs, root_path, mut git_dir_scan_id) = self.read_with(cx, |tree, _| {
5405 let tree = tree.as_local().unwrap();
5406 let local_repo_entry = tree
5407 .git_repositories
5408 .values()
5409 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5410 .unwrap();
5411 (
5412 tree.fs.clone(),
5413 local_repo_entry.common_dir_abs_path.clone(),
5414 local_repo_entry.git_dir_scan_id,
5415 )
5416 });
5417
5418 let scan_id_increased = |tree: &mut Worktree, git_dir_scan_id: &mut usize| {
5419 let tree = tree.as_local().unwrap();
5420 // let repository = tree.repositories.first().unwrap();
5421 let local_repo_entry = tree
5422 .git_repositories
5423 .values()
5424 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5425 .unwrap();
5426
5427 if local_repo_entry.git_dir_scan_id > *git_dir_scan_id {
5428 *git_dir_scan_id = local_repo_entry.git_dir_scan_id;
5429 true
5430 } else {
5431 false
5432 }
5433 };
5434
5435 async move {
5436 fs.create_file(&root_path.join(file_name), Default::default())
5437 .await
5438 .unwrap();
5439
5440 let mut events = cx.events(&tree);
5441 while events.next().await.is_some() {
5442 if tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5443 break;
5444 }
5445 }
5446
5447 fs.remove_file(&root_path.join(file_name), Default::default())
5448 .await
5449 .unwrap();
5450
5451 while events.next().await.is_some() {
5452 if tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5453 break;
5454 }
5455 }
5456
5457 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5458 .await;
5459 }
5460 .boxed_local()
5461 }
5462}
5463
5464#[derive(Clone, Debug)]
5465struct TraversalProgress<'a> {
5466 max_path: &'a RelPath,
5467 count: usize,
5468 non_ignored_count: usize,
5469 file_count: usize,
5470 non_ignored_file_count: usize,
5471}
5472
5473impl TraversalProgress<'_> {
5474 fn count(&self, include_files: bool, include_dirs: bool, include_ignored: bool) -> usize {
5475 match (include_files, include_dirs, include_ignored) {
5476 (true, true, true) => self.count,
5477 (true, true, false) => self.non_ignored_count,
5478 (true, false, true) => self.file_count,
5479 (true, false, false) => self.non_ignored_file_count,
5480 (false, true, true) => self.count - self.file_count,
5481 (false, true, false) => self.non_ignored_count - self.non_ignored_file_count,
5482 (false, false, _) => 0,
5483 }
5484 }
5485}
5486
5487impl<'a> sum_tree::Dimension<'a, EntrySummary> for TraversalProgress<'a> {
5488 fn zero(_cx: ()) -> Self {
5489 Default::default()
5490 }
5491
5492 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
5493 self.max_path = summary.max_path.as_ref();
5494 self.count += summary.count;
5495 self.non_ignored_count += summary.non_ignored_count;
5496 self.file_count += summary.file_count;
5497 self.non_ignored_file_count += summary.non_ignored_file_count;
5498 }
5499}
5500
5501impl Default for TraversalProgress<'_> {
5502 fn default() -> Self {
5503 Self {
5504 max_path: RelPath::empty(),
5505 count: 0,
5506 non_ignored_count: 0,
5507 file_count: 0,
5508 non_ignored_file_count: 0,
5509 }
5510 }
5511}
5512
5513#[derive(Debug)]
5514pub struct Traversal<'a> {
5515 snapshot: &'a Snapshot,
5516 cursor: sum_tree::Cursor<'a, 'static, Entry, TraversalProgress<'a>>,
5517 include_ignored: bool,
5518 include_files: bool,
5519 include_dirs: bool,
5520}
5521
5522impl<'a> Traversal<'a> {
5523 fn new(
5524 snapshot: &'a Snapshot,
5525 include_files: bool,
5526 include_dirs: bool,
5527 include_ignored: bool,
5528 start_path: &RelPath,
5529 ) -> Self {
5530 let mut cursor = snapshot.entries_by_path.cursor(());
5531 cursor.seek(&TraversalTarget::path(start_path), Bias::Left);
5532 let mut traversal = Self {
5533 snapshot,
5534 cursor,
5535 include_files,
5536 include_dirs,
5537 include_ignored,
5538 };
5539 if traversal.end_offset() == traversal.start_offset() {
5540 traversal.next();
5541 }
5542 traversal
5543 }
5544
5545 pub fn advance(&mut self) -> bool {
5546 self.advance_by(1)
5547 }
5548
5549 pub fn advance_by(&mut self, count: usize) -> bool {
5550 self.cursor.seek_forward(
5551 &TraversalTarget::Count {
5552 count: self.end_offset() + count,
5553 include_dirs: self.include_dirs,
5554 include_files: self.include_files,
5555 include_ignored: self.include_ignored,
5556 },
5557 Bias::Left,
5558 )
5559 }
5560
5561 pub fn advance_to_sibling(&mut self) -> bool {
5562 while let Some(entry) = self.cursor.item() {
5563 self.cursor
5564 .seek_forward(&TraversalTarget::successor(&entry.path), Bias::Left);
5565 if let Some(entry) = self.cursor.item()
5566 && (self.include_files || !entry.is_file())
5567 && (self.include_dirs || !entry.is_dir())
5568 && (self.include_ignored || !entry.is_ignored || entry.is_always_included)
5569 {
5570 return true;
5571 }
5572 }
5573 false
5574 }
5575
5576 pub fn back_to_parent(&mut self) -> bool {
5577 let Some(parent_path) = self.cursor.item().and_then(|entry| entry.path.parent()) else {
5578 return false;
5579 };
5580 self.cursor
5581 .seek(&TraversalTarget::path(parent_path), Bias::Left)
5582 }
5583
5584 pub fn entry(&self) -> Option<&'a Entry> {
5585 self.cursor.item()
5586 }
5587
5588 pub fn snapshot(&self) -> &'a Snapshot {
5589 self.snapshot
5590 }
5591
5592 pub fn start_offset(&self) -> usize {
5593 self.cursor
5594 .start()
5595 .count(self.include_files, self.include_dirs, self.include_ignored)
5596 }
5597
5598 pub fn end_offset(&self) -> usize {
5599 self.cursor
5600 .end()
5601 .count(self.include_files, self.include_dirs, self.include_ignored)
5602 }
5603}
5604
5605impl<'a> Iterator for Traversal<'a> {
5606 type Item = &'a Entry;
5607
5608 fn next(&mut self) -> Option<Self::Item> {
5609 if let Some(item) = self.entry() {
5610 self.advance();
5611 Some(item)
5612 } else {
5613 None
5614 }
5615 }
5616}
5617
5618#[derive(Debug, Clone, Copy)]
5619pub enum PathTarget<'a> {
5620 Path(&'a RelPath),
5621 Successor(&'a RelPath),
5622}
5623
5624impl PathTarget<'_> {
5625 fn cmp_path(&self, other: &RelPath) -> Ordering {
5626 match self {
5627 PathTarget::Path(path) => path.cmp(&other),
5628 PathTarget::Successor(path) => {
5629 if other.starts_with(path) {
5630 Ordering::Greater
5631 } else {
5632 Ordering::Equal
5633 }
5634 }
5635 }
5636 }
5637}
5638
5639impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, PathProgress<'a>> for PathTarget<'_> {
5640 fn cmp(&self, cursor_location: &PathProgress<'a>, _: S::Context<'_>) -> Ordering {
5641 self.cmp_path(cursor_location.max_path)
5642 }
5643}
5644
5645impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, TraversalProgress<'a>> for PathTarget<'_> {
5646 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: S::Context<'_>) -> Ordering {
5647 self.cmp_path(cursor_location.max_path)
5648 }
5649}
5650
5651#[derive(Debug)]
5652enum TraversalTarget<'a> {
5653 Path(PathTarget<'a>),
5654 Count {
5655 count: usize,
5656 include_files: bool,
5657 include_ignored: bool,
5658 include_dirs: bool,
5659 },
5660}
5661
5662impl<'a> TraversalTarget<'a> {
5663 fn path(path: &'a RelPath) -> Self {
5664 Self::Path(PathTarget::Path(path))
5665 }
5666
5667 fn successor(path: &'a RelPath) -> Self {
5668 Self::Path(PathTarget::Successor(path))
5669 }
5670
5671 fn cmp_progress(&self, progress: &TraversalProgress) -> Ordering {
5672 match self {
5673 TraversalTarget::Path(path) => path.cmp_path(progress.max_path),
5674 TraversalTarget::Count {
5675 count,
5676 include_files,
5677 include_dirs,
5678 include_ignored,
5679 } => Ord::cmp(
5680 count,
5681 &progress.count(*include_files, *include_dirs, *include_ignored),
5682 ),
5683 }
5684 }
5685}
5686
5687impl<'a> SeekTarget<'a, EntrySummary, TraversalProgress<'a>> for TraversalTarget<'_> {
5688 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5689 self.cmp_progress(cursor_location)
5690 }
5691}
5692
5693impl<'a> SeekTarget<'a, PathSummary<sum_tree::NoSummary>, TraversalProgress<'a>>
5694 for TraversalTarget<'_>
5695{
5696 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5697 self.cmp_progress(cursor_location)
5698 }
5699}
5700
5701pub struct ChildEntriesOptions {
5702 pub include_files: bool,
5703 pub include_dirs: bool,
5704 pub include_ignored: bool,
5705}
5706
5707pub struct ChildEntriesIter<'a> {
5708 parent_path: &'a RelPath,
5709 traversal: Traversal<'a>,
5710}
5711
5712impl<'a> Iterator for ChildEntriesIter<'a> {
5713 type Item = &'a Entry;
5714
5715 fn next(&mut self) -> Option<Self::Item> {
5716 if let Some(item) = self.traversal.entry()
5717 && item.path.starts_with(self.parent_path)
5718 {
5719 self.traversal.advance_to_sibling();
5720 return Some(item);
5721 }
5722 None
5723 }
5724}
5725
5726impl<'a> From<&'a Entry> for proto::Entry {
5727 fn from(entry: &'a Entry) -> Self {
5728 Self {
5729 id: entry.id.to_proto(),
5730 is_dir: entry.is_dir(),
5731 path: entry.path.as_ref().to_proto(),
5732 inode: entry.inode,
5733 mtime: entry.mtime.map(|time| time.into()),
5734 is_ignored: entry.is_ignored,
5735 is_hidden: entry.is_hidden,
5736 is_external: entry.is_external,
5737 is_fifo: entry.is_fifo,
5738 size: Some(entry.size),
5739 canonical_path: entry
5740 .canonical_path
5741 .as_ref()
5742 .map(|path| path.to_string_lossy().into_owned()),
5743 }
5744 }
5745}
5746
5747impl TryFrom<(&CharBag, &PathMatcher, proto::Entry)> for Entry {
5748 type Error = anyhow::Error;
5749
5750 fn try_from(
5751 (root_char_bag, always_included, entry): (&CharBag, &PathMatcher, proto::Entry),
5752 ) -> Result<Self> {
5753 let kind = if entry.is_dir {
5754 EntryKind::Dir
5755 } else {
5756 EntryKind::File
5757 };
5758
5759 let path =
5760 RelPath::from_proto(&entry.path).context("invalid relative path in proto message")?;
5761 let char_bag = char_bag_for_path(*root_char_bag, &path);
5762 let is_always_included = always_included.is_match(&path);
5763 Ok(Entry {
5764 id: ProjectEntryId::from_proto(entry.id),
5765 kind,
5766 path,
5767 inode: entry.inode,
5768 mtime: entry.mtime.map(|time| time.into()),
5769 size: entry.size.unwrap_or(0),
5770 canonical_path: entry
5771 .canonical_path
5772 .map(|path_string| Arc::from(PathBuf::from(path_string))),
5773 is_ignored: entry.is_ignored,
5774 is_hidden: entry.is_hidden,
5775 is_always_included,
5776 is_external: entry.is_external,
5777 is_private: false,
5778 char_bag,
5779 is_fifo: entry.is_fifo,
5780 })
5781 }
5782}
5783
5784#[derive(Clone, Copy, Debug, Default, Hash, PartialEq, Eq, PartialOrd, Ord)]
5785pub struct ProjectEntryId(usize);
5786
5787impl ProjectEntryId {
5788 pub const MAX: Self = Self(usize::MAX);
5789 pub const MIN: Self = Self(usize::MIN);
5790
5791 pub fn new(counter: &AtomicUsize) -> Self {
5792 Self(counter.fetch_add(1, SeqCst))
5793 }
5794
5795 pub fn from_proto(id: u64) -> Self {
5796 Self(id as usize)
5797 }
5798
5799 pub fn to_proto(self) -> u64 {
5800 self.0 as u64
5801 }
5802
5803 pub fn from_usize(id: usize) -> Self {
5804 ProjectEntryId(id)
5805 }
5806
5807 pub fn to_usize(self) -> usize {
5808 self.0
5809 }
5810}
5811
5812#[cfg(any(test, feature = "test-support"))]
5813impl CreatedEntry {
5814 pub fn into_included(self) -> Option<Entry> {
5815 match self {
5816 CreatedEntry::Included(entry) => Some(entry),
5817 CreatedEntry::Excluded { .. } => None,
5818 }
5819 }
5820}
5821
5822fn parse_gitfile(content: &str) -> anyhow::Result<&Path> {
5823 let path = content
5824 .strip_prefix("gitdir:")
5825 .with_context(|| format!("parsing gitfile content {content:?}"))?;
5826 Ok(Path::new(path.trim()))
5827}
5828
5829async fn discover_git_paths(dot_git_abs_path: &Arc<Path>, fs: &dyn Fs) -> (Arc<Path>, Arc<Path>) {
5830 let mut repository_dir_abs_path = dot_git_abs_path.clone();
5831 let mut common_dir_abs_path = dot_git_abs_path.clone();
5832
5833 if let Some(path) = fs
5834 .load(dot_git_abs_path)
5835 .await
5836 .ok()
5837 .as_ref()
5838 .and_then(|contents| parse_gitfile(contents).log_err())
5839 {
5840 let path = dot_git_abs_path
5841 .parent()
5842 .unwrap_or(Path::new(""))
5843 .join(path);
5844 if let Some(path) = fs.canonicalize(&path).await.log_err() {
5845 repository_dir_abs_path = Path::new(&path).into();
5846 common_dir_abs_path = repository_dir_abs_path.clone();
5847
5848 if let Some(commondir_contents) = fs.load(&path.join("commondir")).await.ok()
5849 && let Some(commondir_path) = fs
5850 .canonicalize(&path.join(commondir_contents.trim()))
5851 .await
5852 .log_err()
5853 {
5854 common_dir_abs_path = commondir_path.as_path().into();
5855 }
5856 }
5857 };
5858 (repository_dir_abs_path, common_dir_abs_path)
5859}
5860
5861struct NullWatcher;
5862
5863impl fs::Watcher for NullWatcher {
5864 fn add(&self, _path: &Path) -> Result<()> {
5865 Ok(())
5866 }
5867
5868 fn remove(&self, _path: &Path) -> Result<()> {
5869 Ok(())
5870 }
5871}
5872
5873const FILE_ANALYSIS_BYTES: usize = 1024;
5874
5875async fn decode_file_text(
5876 fs: &dyn Fs,
5877 abs_path: &Path,
5878) -> Result<(String, &'static Encoding, bool)> {
5879 let mut file = fs
5880 .open_sync(&abs_path)
5881 .await
5882 .with_context(|| format!("opening file {abs_path:?}"))?;
5883
5884 // First, read the beginning of the file to determine its kind and encoding.
5885 // We do not want to load an entire large blob into memory only to discard it.
5886 let mut file_first_bytes = Vec::with_capacity(FILE_ANALYSIS_BYTES);
5887 let mut buf = [0u8; FILE_ANALYSIS_BYTES];
5888 let mut reached_eof = false;
5889 loop {
5890 if file_first_bytes.len() >= FILE_ANALYSIS_BYTES {
5891 break;
5892 }
5893 let n = file
5894 .read(&mut buf)
5895 .with_context(|| format!("reading bytes of the file {abs_path:?}"))?;
5896 if n == 0 {
5897 reached_eof = true;
5898 break;
5899 }
5900 file_first_bytes.extend_from_slice(&buf[..n]);
5901 }
5902 let (bom_encoding, byte_content) = decode_byte_header(&file_first_bytes);
5903 anyhow::ensure!(
5904 byte_content != ByteContent::Binary,
5905 "Binary files are not supported"
5906 );
5907
5908 // If the file is eligible for opening, read the rest of the file.
5909 let mut content = file_first_bytes;
5910 if !reached_eof {
5911 let mut buf = [0u8; 8 * 1024];
5912 loop {
5913 let n = file
5914 .read(&mut buf)
5915 .with_context(|| format!("reading remaining bytes of the file {abs_path:?}"))?;
5916 if n == 0 {
5917 break;
5918 }
5919 content.extend_from_slice(&buf[..n]);
5920 }
5921 }
5922 decode_byte_full(content, bom_encoding, byte_content)
5923}
5924
5925fn decode_byte_header(prefix: &[u8]) -> (Option<&'static Encoding>, ByteContent) {
5926 if let Some((encoding, _bom_len)) = Encoding::for_bom(prefix) {
5927 return (Some(encoding), ByteContent::Unknown);
5928 }
5929 (None, analyze_byte_content(prefix))
5930}
5931
5932fn decode_byte_full(
5933 bytes: Vec<u8>,
5934 bom_encoding: Option<&'static Encoding>,
5935 byte_content: ByteContent,
5936) -> Result<(String, &'static Encoding, bool)> {
5937 if let Some(encoding) = bom_encoding {
5938 let (cow, _) = encoding.decode_with_bom_removal(&bytes);
5939 return Ok((cow.into_owned(), encoding, true));
5940 }
5941
5942 match byte_content {
5943 ByteContent::Utf16Le => {
5944 let encoding = encoding_rs::UTF_16LE;
5945 let (cow, _, _) = encoding.decode(&bytes);
5946 return Ok((cow.into_owned(), encoding, false));
5947 }
5948 ByteContent::Utf16Be => {
5949 let encoding = encoding_rs::UTF_16BE;
5950 let (cow, _, _) = encoding.decode(&bytes);
5951 return Ok((cow.into_owned(), encoding, false));
5952 }
5953 ByteContent::Binary => {
5954 anyhow::bail!("Binary files are not supported");
5955 }
5956 ByteContent::Unknown => {}
5957 }
5958
5959 fn detect_encoding(bytes: Vec<u8>) -> (String, &'static Encoding) {
5960 let mut detector = EncodingDetector::new();
5961 detector.feed(&bytes, true);
5962
5963 let encoding = detector.guess(None, true); // Use None for TLD hint to ensure neutral detection logic.
5964
5965 let (cow, _, _) = encoding.decode(&bytes);
5966 (cow.into_owned(), encoding)
5967 }
5968
5969 match String::from_utf8(bytes) {
5970 Ok(text) => {
5971 // ISO-2022-JP (and other ISO-2022 variants) consists entirely of 7-bit ASCII bytes,
5972 // so it is valid UTF-8. However, it contains escape sequences starting with '\x1b'.
5973 // If we find an escape character, we double-check the encoding to prevent
5974 // displaying raw escape sequences instead of the correct characters.
5975 if text.contains('\x1b') {
5976 let (s, enc) = detect_encoding(text.into_bytes());
5977 Ok((s, enc, false))
5978 } else {
5979 Ok((text, encoding_rs::UTF_8, false))
5980 }
5981 }
5982 Err(e) => {
5983 let (s, enc) = detect_encoding(e.into_bytes());
5984 Ok((s, enc, false))
5985 }
5986 }
5987}
5988
5989#[derive(PartialEq)]
5990enum ByteContent {
5991 Utf16Le,
5992 Utf16Be,
5993 Binary,
5994 Unknown,
5995}
5996// Heuristic check using null byte distribution.
5997// NOTE: This relies on the presence of ASCII characters (which become `0x00` in UTF-16).
5998// Files consisting purely of non-ASCII characters (like Japanese) may not be detected here
5999// and will result in `Unknown`.
6000fn analyze_byte_content(bytes: &[u8]) -> ByteContent {
6001 if bytes.len() < 2 {
6002 return ByteContent::Unknown;
6003 }
6004
6005 let check_len = bytes.len().min(FILE_ANALYSIS_BYTES);
6006 let sample = &bytes[..check_len];
6007
6008 if !sample.contains(&0) {
6009 return ByteContent::Unknown;
6010 }
6011
6012 let mut even_nulls = 0;
6013 let mut odd_nulls = 0;
6014
6015 for (i, &byte) in sample.iter().enumerate() {
6016 if byte == 0 {
6017 if i % 2 == 0 {
6018 even_nulls += 1;
6019 } else {
6020 odd_nulls += 1;
6021 }
6022 }
6023 }
6024
6025 let total_nulls = even_nulls + odd_nulls;
6026 if total_nulls < check_len / 10 {
6027 return ByteContent::Unknown;
6028 }
6029
6030 if even_nulls > odd_nulls * 4 {
6031 return ByteContent::Utf16Be;
6032 }
6033
6034 if odd_nulls > even_nulls * 4 {
6035 return ByteContent::Utf16Le;
6036 }
6037
6038 ByteContent::Binary
6039}