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}
256
257struct BackgroundScannerState {
258 snapshot: LocalSnapshot,
259 scanned_dirs: HashSet<ProjectEntryId>,
260 path_prefixes_to_scan: HashSet<Arc<RelPath>>,
261 paths_to_scan: HashSet<Arc<RelPath>>,
262 /// The ids of all of the entries that were removed from the snapshot
263 /// as part of the current update. These entry ids may be re-used
264 /// if the same inode is discovered at a new path, or if the given
265 /// path is re-created after being deleted.
266 removed_entries: HashMap<u64, Entry>,
267 changed_paths: Vec<Arc<RelPath>>,
268 prev_snapshot: Snapshot,
269 scanning_enabled: bool,
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(Debug, 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 Ok(cx.new(move |cx: &mut Context<Worktree>| {
403 let mut snapshot = LocalSnapshot {
404 ignores_by_parent_abs_path: Default::default(),
405 global_gitignore: Default::default(),
406 repo_exclude_by_work_dir_abs_path: Default::default(),
407 git_repositories: Default::default(),
408 snapshot: Snapshot::new(
409 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 };
421
422 let worktree_id = snapshot.id();
423 let settings_location = Some(SettingsLocation {
424 worktree_id,
425 path: RelPath::empty(),
426 });
427
428 let settings = WorktreeSettings::get(settings_location, cx).clone();
429 cx.observe_global::<SettingsStore>(move |this, cx| {
430 if let Self::Local(this) = this {
431 let settings = WorktreeSettings::get(settings_location, cx).clone();
432 if this.settings != settings {
433 this.settings = settings;
434 this.restart_background_scanners(cx);
435 }
436 }
437 })
438 .detach();
439
440 let share_private_files = false;
441 if let Some(metadata) = metadata {
442 let mut entry = Entry::new(
443 RelPath::empty().into(),
444 &metadata,
445 ProjectEntryId::new(&next_entry_id),
446 snapshot.root_char_bag,
447 None,
448 );
449 if metadata.is_dir {
450 if !scanning_enabled {
451 entry.kind = EntryKind::UnloadedDir;
452 }
453 } else {
454 if let Some(file_name) = abs_path.file_name()
455 && let Some(file_name) = file_name.to_str()
456 && let Ok(path) = RelPath::unix(file_name)
457 {
458 entry.is_private = !share_private_files && settings.is_path_private(path);
459 entry.is_hidden = settings.is_path_hidden(path);
460 }
461 }
462 cx.foreground_executor()
463 .block_on(snapshot.insert_entry(entry, fs.as_ref()));
464 }
465
466 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
467 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
468 let mut worktree = LocalWorktree {
469 share_private_files,
470 next_entry_id,
471 snapshot,
472 is_scanning: watch::channel_with(true),
473 update_observer: None,
474 scan_requests_tx,
475 path_prefixes_to_scan_tx,
476 _background_scanner_tasks: Vec::new(),
477 fs,
478 fs_case_sensitive,
479 visible,
480 settings,
481 scanning_enabled,
482 };
483 worktree.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
484 Worktree::Local(worktree)
485 }))
486 }
487
488 pub fn remote(
489 project_id: u64,
490 replica_id: ReplicaId,
491 worktree: proto::WorktreeMetadata,
492 client: AnyProtoClient,
493 path_style: PathStyle,
494 cx: &mut App,
495 ) -> Entity<Self> {
496 cx.new(|cx: &mut Context<Self>| {
497 let snapshot = Snapshot::new(
498 worktree.id,
499 RelPath::from_proto(&worktree.root_name)
500 .unwrap_or_else(|_| RelPath::empty().into()),
501 Path::new(&worktree.abs_path).into(),
502 path_style,
503 );
504
505 let background_snapshot = Arc::new(Mutex::new((
506 snapshot.clone(),
507 Vec::<proto::UpdateWorktree>::new(),
508 )));
509 let (background_updates_tx, mut background_updates_rx) =
510 mpsc::unbounded::<proto::UpdateWorktree>();
511 let (mut snapshot_updated_tx, mut snapshot_updated_rx) = watch::channel();
512
513 let worktree_id = snapshot.id();
514 let settings_location = Some(SettingsLocation {
515 worktree_id,
516 path: RelPath::empty(),
517 });
518
519 let settings = WorktreeSettings::get(settings_location, cx).clone();
520 let worktree = RemoteWorktree {
521 client,
522 project_id,
523 replica_id,
524 snapshot,
525 file_scan_inclusions: settings.parent_dir_scan_inclusions.clone(),
526 background_snapshot: background_snapshot.clone(),
527 updates_tx: Some(background_updates_tx),
528 update_observer: None,
529 snapshot_subscriptions: Default::default(),
530 visible: worktree.visible,
531 disconnected: false,
532 };
533
534 // Apply updates to a separate snapshot in a background task, then
535 // send them to a foreground task which updates the model.
536 cx.background_spawn(async move {
537 while let Some(update) = background_updates_rx.next().await {
538 {
539 let mut lock = background_snapshot.lock();
540 lock.0.apply_remote_update(
541 update.clone(),
542 &settings.parent_dir_scan_inclusions,
543 );
544 lock.1.push(update);
545 }
546 snapshot_updated_tx.send(()).await.ok();
547 }
548 })
549 .detach();
550
551 // On the foreground task, update to the latest snapshot and notify
552 // any update observer of all updates that led to that snapshot.
553 cx.spawn(async move |this, cx| {
554 while (snapshot_updated_rx.recv().await).is_some() {
555 this.update(cx, |this, cx| {
556 let mut entries_changed = false;
557 let this = this.as_remote_mut().unwrap();
558 {
559 let mut lock = this.background_snapshot.lock();
560 this.snapshot = lock.0.clone();
561 for update in lock.1.drain(..) {
562 entries_changed |= !update.updated_entries.is_empty()
563 || !update.removed_entries.is_empty();
564 if let Some(tx) = &this.update_observer {
565 tx.unbounded_send(update).ok();
566 }
567 }
568 };
569
570 if entries_changed {
571 cx.emit(Event::UpdatedEntries(Arc::default()));
572 }
573 cx.notify();
574 while let Some((scan_id, _)) = this.snapshot_subscriptions.front() {
575 if this.observed_snapshot(*scan_id) {
576 let (_, tx) = this.snapshot_subscriptions.pop_front().unwrap();
577 let _ = tx.send(());
578 } else {
579 break;
580 }
581 }
582 })?;
583 }
584 anyhow::Ok(())
585 })
586 .detach();
587
588 Worktree::Remote(worktree)
589 })
590 }
591
592 pub fn as_local(&self) -> Option<&LocalWorktree> {
593 if let Worktree::Local(worktree) = self {
594 Some(worktree)
595 } else {
596 None
597 }
598 }
599
600 pub fn as_remote(&self) -> Option<&RemoteWorktree> {
601 if let Worktree::Remote(worktree) = self {
602 Some(worktree)
603 } else {
604 None
605 }
606 }
607
608 pub fn as_local_mut(&mut self) -> Option<&mut LocalWorktree> {
609 if let Worktree::Local(worktree) = self {
610 Some(worktree)
611 } else {
612 None
613 }
614 }
615
616 pub fn as_remote_mut(&mut self) -> Option<&mut RemoteWorktree> {
617 if let Worktree::Remote(worktree) = self {
618 Some(worktree)
619 } else {
620 None
621 }
622 }
623
624 pub fn is_local(&self) -> bool {
625 matches!(self, Worktree::Local(_))
626 }
627
628 pub fn is_remote(&self) -> bool {
629 !self.is_local()
630 }
631
632 pub fn settings_location(&self, _: &Context<Self>) -> SettingsLocation<'static> {
633 SettingsLocation {
634 worktree_id: self.id(),
635 path: RelPath::empty(),
636 }
637 }
638
639 pub fn snapshot(&self) -> Snapshot {
640 match self {
641 Worktree::Local(worktree) => worktree.snapshot.snapshot.clone(),
642 Worktree::Remote(worktree) => worktree.snapshot.clone(),
643 }
644 }
645
646 pub fn scan_id(&self) -> usize {
647 match self {
648 Worktree::Local(worktree) => worktree.snapshot.scan_id,
649 Worktree::Remote(worktree) => worktree.snapshot.scan_id,
650 }
651 }
652
653 pub fn metadata_proto(&self) -> proto::WorktreeMetadata {
654 proto::WorktreeMetadata {
655 id: self.id().to_proto(),
656 root_name: self.root_name().to_proto(),
657 visible: self.is_visible(),
658 abs_path: self.abs_path().to_string_lossy().into_owned(),
659 }
660 }
661
662 pub fn completed_scan_id(&self) -> usize {
663 match self {
664 Worktree::Local(worktree) => worktree.snapshot.completed_scan_id,
665 Worktree::Remote(worktree) => worktree.snapshot.completed_scan_id,
666 }
667 }
668
669 pub fn is_visible(&self) -> bool {
670 match self {
671 Worktree::Local(worktree) => worktree.visible,
672 Worktree::Remote(worktree) => worktree.visible,
673 }
674 }
675
676 pub fn replica_id(&self) -> ReplicaId {
677 match self {
678 Worktree::Local(_) => ReplicaId::LOCAL,
679 Worktree::Remote(worktree) => worktree.replica_id,
680 }
681 }
682
683 pub fn abs_path(&self) -> Arc<Path> {
684 match self {
685 Worktree::Local(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
686 Worktree::Remote(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
687 }
688 }
689
690 pub fn root_file(&self, cx: &Context<Self>) -> Option<Arc<File>> {
691 let entry = self.root_entry()?;
692 Some(File::for_entry(entry.clone(), cx.entity()))
693 }
694
695 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
696 where
697 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
698 Fut: 'static + Send + Future<Output = bool>,
699 {
700 match self {
701 Worktree::Local(this) => this.observe_updates(project_id, cx, callback),
702 Worktree::Remote(this) => this.observe_updates(project_id, cx, callback),
703 }
704 }
705
706 pub fn stop_observing_updates(&mut self) {
707 match self {
708 Worktree::Local(this) => {
709 this.update_observer.take();
710 }
711 Worktree::Remote(this) => {
712 this.update_observer.take();
713 }
714 }
715 }
716
717 #[cfg(any(test, feature = "test-support"))]
718 pub fn has_update_observer(&self) -> bool {
719 match self {
720 Worktree::Local(this) => this.update_observer.is_some(),
721 Worktree::Remote(this) => this.update_observer.is_some(),
722 }
723 }
724
725 pub fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
726 match self {
727 Worktree::Local(this) => this.load_file(path, cx),
728 Worktree::Remote(_) => {
729 Task::ready(Err(anyhow!("remote worktrees can't yet load files")))
730 }
731 }
732 }
733
734 pub fn load_binary_file(
735 &self,
736 path: &RelPath,
737 cx: &Context<Worktree>,
738 ) -> Task<Result<LoadedBinaryFile>> {
739 match self {
740 Worktree::Local(this) => this.load_binary_file(path, cx),
741 Worktree::Remote(_) => {
742 Task::ready(Err(anyhow!("remote worktrees can't yet load binary files")))
743 }
744 }
745 }
746
747 pub fn write_file(
748 &self,
749 path: Arc<RelPath>,
750 text: Rope,
751 line_ending: LineEnding,
752 encoding: &'static Encoding,
753 has_bom: bool,
754 cx: &Context<Worktree>,
755 ) -> Task<Result<Arc<File>>> {
756 match self {
757 Worktree::Local(this) => {
758 this.write_file(path, text, line_ending, encoding, has_bom, cx)
759 }
760 Worktree::Remote(_) => {
761 Task::ready(Err(anyhow!("remote worktree can't yet write files")))
762 }
763 }
764 }
765
766 pub fn create_entry(
767 &mut self,
768 path: Arc<RelPath>,
769 is_directory: bool,
770 content: Option<Vec<u8>>,
771 cx: &Context<Worktree>,
772 ) -> Task<Result<CreatedEntry>> {
773 let worktree_id = self.id();
774 match self {
775 Worktree::Local(this) => this.create_entry(path, is_directory, content, cx),
776 Worktree::Remote(this) => {
777 let project_id = this.project_id;
778 let request = this.client.request(proto::CreateProjectEntry {
779 worktree_id: worktree_id.to_proto(),
780 project_id,
781 path: path.as_ref().to_proto(),
782 content,
783 is_directory,
784 });
785 cx.spawn(async move |this, cx| {
786 let response = request.await?;
787 match response.entry {
788 Some(entry) => this
789 .update(cx, |worktree, cx| {
790 worktree.as_remote_mut().unwrap().insert_entry(
791 entry,
792 response.worktree_scan_id as usize,
793 cx,
794 )
795 })?
796 .await
797 .map(CreatedEntry::Included),
798 None => {
799 let abs_path =
800 this.read_with(cx, |worktree, _| worktree.absolutize(&path))?;
801 Ok(CreatedEntry::Excluded { abs_path })
802 }
803 }
804 })
805 }
806 }
807 }
808
809 pub fn delete_entry(
810 &mut self,
811 entry_id: ProjectEntryId,
812 trash: bool,
813 cx: &mut Context<Worktree>,
814 ) -> Option<Task<Result<()>>> {
815 let task = match self {
816 Worktree::Local(this) => this.delete_entry(entry_id, trash, cx),
817 Worktree::Remote(this) => this.delete_entry(entry_id, trash, cx),
818 }?;
819
820 let entry = match &*self {
821 Worktree::Local(this) => this.entry_for_id(entry_id),
822 Worktree::Remote(this) => this.entry_for_id(entry_id),
823 }?;
824
825 let mut ids = vec![entry_id];
826 let path = &*entry.path;
827
828 self.get_children_ids_recursive(path, &mut ids);
829
830 for id in ids {
831 cx.emit(Event::DeletedEntry(id));
832 }
833 Some(task)
834 }
835
836 fn get_children_ids_recursive(&self, path: &RelPath, ids: &mut Vec<ProjectEntryId>) {
837 let children_iter = self.child_entries(path);
838 for child in children_iter {
839 ids.push(child.id);
840 self.get_children_ids_recursive(&child.path, ids);
841 }
842 }
843
844 // pub fn rename_entry(
845 // &mut self,
846 // entry_id: ProjectEntryId,
847 // new_path: Arc<RelPath>,
848 // cx: &Context<Self>,
849 // ) -> Task<Result<CreatedEntry>> {
850 // match self {
851 // Worktree::Local(this) => this.rename_entry(entry_id, new_path, cx),
852 // Worktree::Remote(this) => this.rename_entry(entry_id, new_path, cx),
853 // }
854 // }
855
856 pub fn copy_external_entries(
857 &mut self,
858 target_directory: Arc<RelPath>,
859 paths: Vec<Arc<Path>>,
860 fs: Arc<dyn Fs>,
861 cx: &Context<Worktree>,
862 ) -> Task<Result<Vec<ProjectEntryId>>> {
863 match self {
864 Worktree::Local(this) => this.copy_external_entries(target_directory, paths, cx),
865 Worktree::Remote(this) => this.copy_external_entries(target_directory, paths, fs, cx),
866 }
867 }
868
869 pub fn expand_entry(
870 &mut self,
871 entry_id: ProjectEntryId,
872 cx: &Context<Worktree>,
873 ) -> Option<Task<Result<()>>> {
874 match self {
875 Worktree::Local(this) => this.expand_entry(entry_id, cx),
876 Worktree::Remote(this) => {
877 let response = this.client.request(proto::ExpandProjectEntry {
878 project_id: this.project_id,
879 entry_id: entry_id.to_proto(),
880 });
881 Some(cx.spawn(async move |this, cx| {
882 let response = response.await?;
883 this.update(cx, |this, _| {
884 this.as_remote_mut()
885 .unwrap()
886 .wait_for_snapshot(response.worktree_scan_id as usize)
887 })?
888 .await?;
889 Ok(())
890 }))
891 }
892 }
893 }
894
895 pub fn expand_all_for_entry(
896 &mut self,
897 entry_id: ProjectEntryId,
898 cx: &Context<Worktree>,
899 ) -> Option<Task<Result<()>>> {
900 match self {
901 Worktree::Local(this) => this.expand_all_for_entry(entry_id, cx),
902 Worktree::Remote(this) => {
903 let response = this.client.request(proto::ExpandAllForProjectEntry {
904 project_id: this.project_id,
905 entry_id: entry_id.to_proto(),
906 });
907 Some(cx.spawn(async move |this, cx| {
908 let response = response.await?;
909 this.update(cx, |this, _| {
910 this.as_remote_mut()
911 .unwrap()
912 .wait_for_snapshot(response.worktree_scan_id as usize)
913 })?
914 .await?;
915 Ok(())
916 }))
917 }
918 }
919 }
920
921 pub async fn handle_create_entry(
922 this: Entity<Self>,
923 request: proto::CreateProjectEntry,
924 mut cx: AsyncApp,
925 ) -> Result<proto::ProjectEntryResponse> {
926 let (scan_id, entry) = this.update(&mut cx, |this, cx| {
927 anyhow::Ok((
928 this.scan_id(),
929 this.create_entry(
930 RelPath::from_proto(&request.path).with_context(|| {
931 format!("received invalid relative path {:?}", request.path)
932 })?,
933 request.is_directory,
934 request.content,
935 cx,
936 ),
937 ))
938 })?;
939 Ok(proto::ProjectEntryResponse {
940 entry: match &entry.await? {
941 CreatedEntry::Included(entry) => Some(entry.into()),
942 CreatedEntry::Excluded { .. } => None,
943 },
944 worktree_scan_id: scan_id as u64,
945 })
946 }
947
948 pub async fn handle_delete_entry(
949 this: Entity<Self>,
950 request: proto::DeleteProjectEntry,
951 mut cx: AsyncApp,
952 ) -> Result<proto::ProjectEntryResponse> {
953 let (scan_id, task) = this.update(&mut cx, |this, cx| {
954 (
955 this.scan_id(),
956 this.delete_entry(
957 ProjectEntryId::from_proto(request.entry_id),
958 request.use_trash,
959 cx,
960 ),
961 )
962 });
963 task.ok_or_else(|| anyhow::anyhow!("invalid entry"))?
964 .await?;
965 Ok(proto::ProjectEntryResponse {
966 entry: None,
967 worktree_scan_id: scan_id as u64,
968 })
969 }
970
971 pub async fn handle_expand_entry(
972 this: Entity<Self>,
973 request: proto::ExpandProjectEntry,
974 mut cx: AsyncApp,
975 ) -> Result<proto::ExpandProjectEntryResponse> {
976 let task = this.update(&mut cx, |this, cx| {
977 this.expand_entry(ProjectEntryId::from_proto(request.entry_id), cx)
978 });
979 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
980 .await?;
981 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
982 Ok(proto::ExpandProjectEntryResponse {
983 worktree_scan_id: scan_id as u64,
984 })
985 }
986
987 pub async fn handle_expand_all_for_entry(
988 this: Entity<Self>,
989 request: proto::ExpandAllForProjectEntry,
990 mut cx: AsyncApp,
991 ) -> Result<proto::ExpandAllForProjectEntryResponse> {
992 let task = this.update(&mut cx, |this, cx| {
993 this.expand_all_for_entry(ProjectEntryId::from_proto(request.entry_id), cx)
994 });
995 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
996 .await?;
997 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
998 Ok(proto::ExpandAllForProjectEntryResponse {
999 worktree_scan_id: scan_id as u64,
1000 })
1001 }
1002
1003 pub fn is_single_file(&self) -> bool {
1004 self.root_dir().is_none()
1005 }
1006
1007 /// For visible worktrees, returns the path with the worktree name as the first component.
1008 /// Otherwise, returns an absolute path.
1009 pub fn full_path(&self, worktree_relative_path: &RelPath) -> PathBuf {
1010 if self.is_visible() {
1011 self.root_name()
1012 .join(worktree_relative_path)
1013 .display(self.path_style)
1014 .to_string()
1015 .into()
1016 } else {
1017 let full_path = self.abs_path();
1018 let mut full_path_string = if self.is_local()
1019 && let Ok(stripped) = full_path.strip_prefix(home_dir())
1020 {
1021 self.path_style
1022 .join("~", &*stripped.to_string_lossy())
1023 .unwrap()
1024 } else {
1025 full_path.to_string_lossy().into_owned()
1026 };
1027
1028 if worktree_relative_path.components().next().is_some() {
1029 full_path_string.push_str(self.path_style.primary_separator());
1030 full_path_string.push_str(&worktree_relative_path.display(self.path_style));
1031 }
1032
1033 full_path_string.into()
1034 }
1035 }
1036}
1037
1038impl LocalWorktree {
1039 pub fn fs(&self) -> &Arc<dyn Fs> {
1040 &self.fs
1041 }
1042
1043 pub fn is_path_private(&self, path: &RelPath) -> bool {
1044 !self.share_private_files && self.settings.is_path_private(path)
1045 }
1046
1047 pub fn fs_is_case_sensitive(&self) -> bool {
1048 self.fs_case_sensitive
1049 }
1050
1051 fn restart_background_scanners(&mut self, cx: &Context<Worktree>) {
1052 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
1053 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
1054 self.scan_requests_tx = scan_requests_tx;
1055 self.path_prefixes_to_scan_tx = path_prefixes_to_scan_tx;
1056
1057 self.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
1058 let always_included_entries = mem::take(&mut self.snapshot.always_included_entries);
1059 log::debug!(
1060 "refreshing entries for the following always included paths: {:?}",
1061 always_included_entries
1062 );
1063
1064 // Cleans up old always included entries to ensure they get updated properly. Otherwise,
1065 // nested always included entries may not get updated and will result in out-of-date info.
1066 self.refresh_entries_for_paths(always_included_entries);
1067 }
1068
1069 fn start_background_scanner(
1070 &mut self,
1071 scan_requests_rx: channel::Receiver<ScanRequest>,
1072 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
1073 cx: &Context<Worktree>,
1074 ) {
1075 let snapshot = self.snapshot();
1076 let share_private_files = self.share_private_files;
1077 let next_entry_id = self.next_entry_id.clone();
1078 let fs = self.fs.clone();
1079 let scanning_enabled = self.scanning_enabled;
1080 let settings = self.settings.clone();
1081 let (scan_states_tx, mut scan_states_rx) = mpsc::unbounded();
1082 let background_scanner = cx.background_spawn({
1083 let abs_path = snapshot.abs_path.as_path().to_path_buf();
1084 let background = cx.background_executor().clone();
1085 async move {
1086 let (events, watcher) = if scanning_enabled {
1087 fs.watch(&abs_path, FS_WATCH_LATENCY).await
1088 } else {
1089 (Box::pin(stream::pending()) as _, Arc::new(NullWatcher) as _)
1090 };
1091 let fs_case_sensitive = fs.is_case_sensitive().await.unwrap_or_else(|e| {
1092 log::error!("Failed to determine whether filesystem is case sensitive: {e:#}");
1093 true
1094 });
1095
1096 let mut scanner = BackgroundScanner {
1097 fs,
1098 fs_case_sensitive,
1099 status_updates_tx: scan_states_tx,
1100 executor: background,
1101 scan_requests_rx,
1102 path_prefixes_to_scan_rx,
1103 next_entry_id,
1104 state: async_lock::Mutex::new(BackgroundScannerState {
1105 prev_snapshot: snapshot.snapshot.clone(),
1106 snapshot,
1107 scanned_dirs: Default::default(),
1108 scanning_enabled,
1109 path_prefixes_to_scan: Default::default(),
1110 paths_to_scan: Default::default(),
1111 removed_entries: Default::default(),
1112 changed_paths: Default::default(),
1113 }),
1114 phase: BackgroundScannerPhase::InitialScan,
1115 share_private_files,
1116 settings,
1117 watcher,
1118 };
1119
1120 scanner
1121 .run(Box::pin(events.map(|events| events.into_iter().collect())))
1122 .await;
1123 }
1124 });
1125 let scan_state_updater = cx.spawn(async move |this, cx| {
1126 while let Some((state, this)) = scan_states_rx.next().await.zip(this.upgrade()) {
1127 this.update(cx, |this, cx| {
1128 let this = this.as_local_mut().unwrap();
1129 match state {
1130 ScanState::Started => {
1131 *this.is_scanning.0.borrow_mut() = true;
1132 }
1133 ScanState::Updated {
1134 snapshot,
1135 changes,
1136 barrier,
1137 scanning,
1138 } => {
1139 *this.is_scanning.0.borrow_mut() = scanning;
1140 this.set_snapshot(snapshot, changes, cx);
1141 drop(barrier);
1142 }
1143 ScanState::RootUpdated { new_path } => {
1144 this.update_abs_path_and_refresh(new_path, cx);
1145 }
1146 }
1147 });
1148 }
1149 });
1150 self._background_scanner_tasks = vec![background_scanner, scan_state_updater];
1151 *self.is_scanning.0.borrow_mut() = true;
1152 }
1153
1154 fn set_snapshot(
1155 &mut self,
1156 mut new_snapshot: LocalSnapshot,
1157 entry_changes: UpdatedEntriesSet,
1158 cx: &mut Context<Worktree>,
1159 ) {
1160 let repo_changes = self.changed_repos(&self.snapshot, &mut new_snapshot);
1161 self.snapshot = new_snapshot;
1162
1163 if let Some(share) = self.update_observer.as_mut() {
1164 share
1165 .snapshots_tx
1166 .unbounded_send((self.snapshot.clone(), entry_changes.clone()))
1167 .ok();
1168 }
1169
1170 if !entry_changes.is_empty() {
1171 cx.emit(Event::UpdatedEntries(entry_changes));
1172 }
1173 if !repo_changes.is_empty() {
1174 cx.emit(Event::UpdatedGitRepositories(repo_changes));
1175 }
1176 }
1177
1178 fn changed_repos(
1179 &self,
1180 old_snapshot: &LocalSnapshot,
1181 new_snapshot: &mut LocalSnapshot,
1182 ) -> UpdatedGitRepositoriesSet {
1183 let mut changes = Vec::new();
1184 let mut old_repos = old_snapshot.git_repositories.iter().peekable();
1185 let new_repos = new_snapshot.git_repositories.clone();
1186 let mut new_repos = new_repos.iter().peekable();
1187
1188 loop {
1189 match (new_repos.peek().map(clone), old_repos.peek().map(clone)) {
1190 (Some((new_entry_id, new_repo)), Some((old_entry_id, old_repo))) => {
1191 match Ord::cmp(&new_entry_id, &old_entry_id) {
1192 Ordering::Less => {
1193 changes.push(UpdatedGitRepository {
1194 work_directory_id: new_entry_id,
1195 old_work_directory_abs_path: None,
1196 new_work_directory_abs_path: Some(
1197 new_repo.work_directory_abs_path.clone(),
1198 ),
1199 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1200 repository_dir_abs_path: Some(
1201 new_repo.repository_dir_abs_path.clone(),
1202 ),
1203 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1204 });
1205 new_repos.next();
1206 }
1207 Ordering::Equal => {
1208 if new_repo.git_dir_scan_id != old_repo.git_dir_scan_id
1209 || new_repo.work_directory_abs_path
1210 != old_repo.work_directory_abs_path
1211 {
1212 changes.push(UpdatedGitRepository {
1213 work_directory_id: new_entry_id,
1214 old_work_directory_abs_path: Some(
1215 old_repo.work_directory_abs_path.clone(),
1216 ),
1217 new_work_directory_abs_path: Some(
1218 new_repo.work_directory_abs_path.clone(),
1219 ),
1220 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1221 repository_dir_abs_path: Some(
1222 new_repo.repository_dir_abs_path.clone(),
1223 ),
1224 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1225 });
1226 }
1227 new_repos.next();
1228 old_repos.next();
1229 }
1230 Ordering::Greater => {
1231 changes.push(UpdatedGitRepository {
1232 work_directory_id: old_entry_id,
1233 old_work_directory_abs_path: Some(
1234 old_repo.work_directory_abs_path.clone(),
1235 ),
1236 new_work_directory_abs_path: None,
1237 dot_git_abs_path: None,
1238 repository_dir_abs_path: None,
1239 common_dir_abs_path: None,
1240 });
1241 old_repos.next();
1242 }
1243 }
1244 }
1245 (Some((entry_id, repo)), None) => {
1246 changes.push(UpdatedGitRepository {
1247 work_directory_id: entry_id,
1248 old_work_directory_abs_path: None,
1249 new_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1250 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1251 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1252 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1253 });
1254 new_repos.next();
1255 }
1256 (None, Some((entry_id, repo))) => {
1257 changes.push(UpdatedGitRepository {
1258 work_directory_id: entry_id,
1259 old_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1260 new_work_directory_abs_path: None,
1261 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1262 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1263 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1264 });
1265 old_repos.next();
1266 }
1267 (None, None) => break,
1268 }
1269 }
1270
1271 fn clone<T: Clone, U: Clone>(value: &(&T, &U)) -> (T, U) {
1272 (value.0.clone(), value.1.clone())
1273 }
1274
1275 changes.into()
1276 }
1277
1278 pub fn scan_complete(&self) -> impl Future<Output = ()> + use<> {
1279 let mut is_scanning_rx = self.is_scanning.1.clone();
1280 async move {
1281 let mut is_scanning = *is_scanning_rx.borrow();
1282 while is_scanning {
1283 if let Some(value) = is_scanning_rx.recv().await {
1284 is_scanning = value;
1285 } else {
1286 break;
1287 }
1288 }
1289 }
1290 }
1291
1292 pub fn snapshot(&self) -> LocalSnapshot {
1293 self.snapshot.clone()
1294 }
1295
1296 pub fn settings(&self) -> WorktreeSettings {
1297 self.settings.clone()
1298 }
1299
1300 fn load_binary_file(
1301 &self,
1302 path: &RelPath,
1303 cx: &Context<Worktree>,
1304 ) -> Task<Result<LoadedBinaryFile>> {
1305 let path = Arc::from(path);
1306 let abs_path = self.absolutize(&path);
1307 let fs = self.fs.clone();
1308 let entry = self.refresh_entry(path.clone(), None, cx);
1309 let is_private = self.is_path_private(&path);
1310
1311 let worktree = cx.weak_entity();
1312 cx.background_spawn(async move {
1313 let content = fs.load_bytes(&abs_path).await?;
1314
1315 let worktree = worktree.upgrade().context("worktree was dropped")?;
1316 let file = match entry.await? {
1317 Some(entry) => File::for_entry(entry, worktree),
1318 None => {
1319 let metadata = fs
1320 .metadata(&abs_path)
1321 .await
1322 .with_context(|| {
1323 format!("Loading metadata for excluded file {abs_path:?}")
1324 })?
1325 .with_context(|| {
1326 format!("Excluded file {abs_path:?} got removed during loading")
1327 })?;
1328 Arc::new(File {
1329 entry_id: None,
1330 worktree,
1331 path,
1332 disk_state: DiskState::Present {
1333 mtime: metadata.mtime,
1334 },
1335 is_local: true,
1336 is_private,
1337 })
1338 }
1339 };
1340
1341 Ok(LoadedBinaryFile { file, content })
1342 })
1343 }
1344
1345 #[ztracing::instrument(skip_all)]
1346 fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
1347 let path = Arc::from(path);
1348 let abs_path = self.absolutize(&path);
1349 let fs = self.fs.clone();
1350 let entry = self.refresh_entry(path.clone(), None, cx);
1351 let is_private = self.is_path_private(path.as_ref());
1352
1353 let this = cx.weak_entity();
1354 cx.background_spawn(async move {
1355 // WARN: Temporary workaround for #27283.
1356 // We are not efficient with our memory usage per file, and use in excess of 64GB for a 10GB file
1357 // Therefore, as a temporary workaround to prevent system freezes, we just bail before opening a file
1358 // if it is too large
1359 // 5GB seems to be more reasonable, peaking at ~16GB, while 6GB jumps up to >24GB which seems like a
1360 // reasonable limit
1361 {
1362 const FILE_SIZE_MAX: u64 = 6 * 1024 * 1024 * 1024; // 6GB
1363 if let Ok(Some(metadata)) = fs.metadata(&abs_path).await
1364 && metadata.len >= FILE_SIZE_MAX
1365 {
1366 anyhow::bail!("File is too large to load");
1367 }
1368 }
1369 let (text, encoding, has_bom) = decode_file_text(fs.as_ref(), &abs_path).await?;
1370
1371 let worktree = this.upgrade().context("worktree was dropped")?;
1372 let file = match entry.await? {
1373 Some(entry) => File::for_entry(entry, worktree),
1374 None => {
1375 let metadata = fs
1376 .metadata(&abs_path)
1377 .await
1378 .with_context(|| {
1379 format!("Loading metadata for excluded file {abs_path:?}")
1380 })?
1381 .with_context(|| {
1382 format!("Excluded file {abs_path:?} got removed during loading")
1383 })?;
1384 Arc::new(File {
1385 entry_id: None,
1386 worktree,
1387 path,
1388 disk_state: DiskState::Present {
1389 mtime: metadata.mtime,
1390 },
1391 is_local: true,
1392 is_private,
1393 })
1394 }
1395 };
1396
1397 Ok(LoadedFile {
1398 file,
1399 text,
1400 encoding,
1401 has_bom,
1402 })
1403 })
1404 }
1405
1406 /// Find the lowest path in the worktree's datastructures that is an ancestor
1407 fn lowest_ancestor(&self, path: &RelPath) -> Arc<RelPath> {
1408 let mut lowest_ancestor = None;
1409 for path in path.ancestors() {
1410 if self.entry_for_path(path).is_some() {
1411 lowest_ancestor = Some(path.into());
1412 break;
1413 }
1414 }
1415
1416 lowest_ancestor.unwrap_or_else(|| RelPath::empty().into())
1417 }
1418
1419 fn create_entry(
1420 &self,
1421 path: Arc<RelPath>,
1422 is_dir: bool,
1423 content: Option<Vec<u8>>,
1424 cx: &Context<Worktree>,
1425 ) -> Task<Result<CreatedEntry>> {
1426 let abs_path = self.absolutize(&path);
1427 let path_excluded = self.settings.is_path_excluded(&path);
1428 let fs = self.fs.clone();
1429 let task_abs_path = abs_path.clone();
1430 let write = cx.background_spawn(async move {
1431 if is_dir {
1432 fs.create_dir(&task_abs_path)
1433 .await
1434 .with_context(|| format!("creating directory {task_abs_path:?}"))
1435 } else {
1436 fs.write(&task_abs_path, content.as_deref().unwrap_or(&[]))
1437 .await
1438 .with_context(|| format!("creating file {task_abs_path:?}"))
1439 }
1440 });
1441
1442 let lowest_ancestor = self.lowest_ancestor(&path);
1443 cx.spawn(async move |this, cx| {
1444 write.await?;
1445 if path_excluded {
1446 return Ok(CreatedEntry::Excluded { abs_path });
1447 }
1448
1449 let (result, refreshes) = this.update(cx, |this, cx| {
1450 let mut refreshes = Vec::new();
1451 let refresh_paths = path.strip_prefix(&lowest_ancestor).unwrap();
1452 for refresh_path in refresh_paths.ancestors() {
1453 if refresh_path == RelPath::empty() {
1454 continue;
1455 }
1456 let refresh_full_path = lowest_ancestor.join(refresh_path);
1457
1458 refreshes.push(this.as_local_mut().unwrap().refresh_entry(
1459 refresh_full_path,
1460 None,
1461 cx,
1462 ));
1463 }
1464 (
1465 this.as_local_mut().unwrap().refresh_entry(path, None, cx),
1466 refreshes,
1467 )
1468 })?;
1469 for refresh in refreshes {
1470 refresh.await.log_err();
1471 }
1472
1473 Ok(result
1474 .await?
1475 .map(CreatedEntry::Included)
1476 .unwrap_or_else(|| CreatedEntry::Excluded { abs_path }))
1477 })
1478 }
1479
1480 fn write_file(
1481 &self,
1482 path: Arc<RelPath>,
1483 text: Rope,
1484 line_ending: LineEnding,
1485 encoding: &'static Encoding,
1486 has_bom: bool,
1487 cx: &Context<Worktree>,
1488 ) -> Task<Result<Arc<File>>> {
1489 let fs = self.fs.clone();
1490 let is_private = self.is_path_private(&path);
1491 let abs_path = self.absolutize(&path);
1492
1493 let write = cx.background_spawn({
1494 let fs = fs.clone();
1495 let abs_path = abs_path.clone();
1496 async move {
1497 // For UTF-8, use the optimized `fs.save` which writes Rope chunks directly to disk
1498 // without allocating a contiguous string.
1499 if encoding == encoding_rs::UTF_8 && !has_bom {
1500 return fs.save(&abs_path, &text, line_ending).await;
1501 }
1502
1503 // For legacy encodings (e.g. Shift-JIS), we fall back to converting the entire Rope
1504 // to a String/Bytes in memory before writing.
1505 //
1506 // Note: This is inefficient for very large files compared to the streaming approach above,
1507 // but supporting streaming writes for arbitrary encodings would require a significant
1508 // refactor of the `fs` crate to expose a Writer interface.
1509 let text_string = text.to_string();
1510 let normalized_text = match line_ending {
1511 LineEnding::Unix => text_string,
1512 LineEnding::Windows => text_string.replace('\n', "\r\n"),
1513 };
1514
1515 // Create the byte vector manually for UTF-16 encodings because encoding_rs encodes to UTF-8 by default (per WHATWG standards),
1516 // which is not what we want for saving files.
1517 let bytes = if encoding == encoding_rs::UTF_16BE {
1518 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1519 if has_bom {
1520 data.extend_from_slice(&[0xFE, 0xFF]); // BOM
1521 }
1522 let utf16be_bytes =
1523 normalized_text.encode_utf16().flat_map(|u| u.to_be_bytes());
1524 data.extend(utf16be_bytes);
1525 data.into()
1526 } else if encoding == encoding_rs::UTF_16LE {
1527 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1528 if has_bom {
1529 data.extend_from_slice(&[0xFF, 0xFE]); // BOM
1530 }
1531 let utf16le_bytes =
1532 normalized_text.encode_utf16().flat_map(|u| u.to_le_bytes());
1533 data.extend(utf16le_bytes);
1534 data.into()
1535 } else {
1536 // For other encodings (Shift-JIS, UTF-8 with BOM, etc.), delegate to encoding_rs.
1537 let bom_bytes = if has_bom {
1538 if encoding == encoding_rs::UTF_8 {
1539 vec![0xEF, 0xBB, 0xBF]
1540 } else {
1541 vec![]
1542 }
1543 } else {
1544 vec![]
1545 };
1546 let (cow, _, _) = encoding.encode(&normalized_text);
1547 if !bom_bytes.is_empty() {
1548 let mut bytes = bom_bytes;
1549 bytes.extend_from_slice(&cow);
1550 bytes.into()
1551 } else {
1552 cow
1553 }
1554 };
1555
1556 fs.write(&abs_path, &bytes).await
1557 }
1558 });
1559
1560 cx.spawn(async move |this, cx| {
1561 write.await?;
1562 let entry = this
1563 .update(cx, |this, cx| {
1564 this.as_local_mut()
1565 .unwrap()
1566 .refresh_entry(path.clone(), None, cx)
1567 })?
1568 .await?;
1569 let worktree = this.upgrade().context("worktree dropped")?;
1570 if let Some(entry) = entry {
1571 Ok(File::for_entry(entry, worktree))
1572 } else {
1573 let metadata = fs
1574 .metadata(&abs_path)
1575 .await
1576 .with_context(|| {
1577 format!("Fetching metadata after saving the excluded buffer {abs_path:?}")
1578 })?
1579 .with_context(|| {
1580 format!("Excluded buffer {path:?} got removed during saving")
1581 })?;
1582 Ok(Arc::new(File {
1583 worktree,
1584 path,
1585 disk_state: DiskState::Present {
1586 mtime: metadata.mtime,
1587 },
1588 entry_id: None,
1589 is_local: true,
1590 is_private,
1591 }))
1592 }
1593 })
1594 }
1595
1596 fn delete_entry(
1597 &self,
1598 entry_id: ProjectEntryId,
1599 trash: bool,
1600 cx: &Context<Worktree>,
1601 ) -> Option<Task<Result<()>>> {
1602 let entry = self.entry_for_id(entry_id)?.clone();
1603 let abs_path = self.absolutize(&entry.path);
1604 let fs = self.fs.clone();
1605
1606 let delete = cx.background_spawn(async move {
1607 if entry.is_file() {
1608 if trash {
1609 fs.trash_file(&abs_path, Default::default()).await?;
1610 } else {
1611 fs.remove_file(&abs_path, Default::default()).await?;
1612 }
1613 } else if trash {
1614 fs.trash_dir(
1615 &abs_path,
1616 RemoveOptions {
1617 recursive: true,
1618 ignore_if_not_exists: false,
1619 },
1620 )
1621 .await?;
1622 } else {
1623 fs.remove_dir(
1624 &abs_path,
1625 RemoveOptions {
1626 recursive: true,
1627 ignore_if_not_exists: false,
1628 },
1629 )
1630 .await?;
1631 }
1632 anyhow::Ok(entry.path)
1633 });
1634
1635 Some(cx.spawn(async move |this, cx| {
1636 let path = delete.await?;
1637 this.update(cx, |this, _| {
1638 this.as_local_mut()
1639 .unwrap()
1640 .refresh_entries_for_paths(vec![path])
1641 })?
1642 .recv()
1643 .await;
1644 Ok(())
1645 }))
1646 }
1647
1648 pub fn copy_external_entries(
1649 &self,
1650 target_directory: Arc<RelPath>,
1651 paths: Vec<Arc<Path>>,
1652 cx: &Context<Worktree>,
1653 ) -> Task<Result<Vec<ProjectEntryId>>> {
1654 let target_directory = self.absolutize(&target_directory);
1655 let worktree_path = self.abs_path().clone();
1656 let fs = self.fs.clone();
1657 let paths = paths
1658 .into_iter()
1659 .filter_map(|source| {
1660 let file_name = source.file_name()?;
1661 let mut target = target_directory.clone();
1662 target.push(file_name);
1663
1664 // Do not allow copying the same file to itself.
1665 if source.as_ref() != target.as_path() {
1666 Some((source, target))
1667 } else {
1668 None
1669 }
1670 })
1671 .collect::<Vec<_>>();
1672
1673 let paths_to_refresh = paths
1674 .iter()
1675 .filter_map(|(_, target)| {
1676 RelPath::new(
1677 target.strip_prefix(&worktree_path).ok()?,
1678 PathStyle::local(),
1679 )
1680 .ok()
1681 .map(|path| path.into_arc())
1682 })
1683 .collect::<Vec<_>>();
1684
1685 cx.spawn(async move |this, cx| {
1686 cx.background_spawn(async move {
1687 for (source, target) in paths {
1688 copy_recursive(
1689 fs.as_ref(),
1690 &source,
1691 &target,
1692 fs::CopyOptions {
1693 overwrite: true,
1694 ..Default::default()
1695 },
1696 )
1697 .await
1698 .with_context(|| {
1699 format!("Failed to copy file from {source:?} to {target:?}")
1700 })?;
1701 }
1702 anyhow::Ok(())
1703 })
1704 .await
1705 .log_err();
1706 let mut refresh = cx.read_entity(
1707 &this.upgrade().with_context(|| "Dropped worktree")?,
1708 |this, _| {
1709 anyhow::Ok::<postage::barrier::Receiver>(
1710 this.as_local()
1711 .with_context(|| "Worktree is not local")?
1712 .refresh_entries_for_paths(paths_to_refresh.clone()),
1713 )
1714 },
1715 )?;
1716
1717 cx.background_spawn(async move {
1718 refresh.next().await;
1719 anyhow::Ok(())
1720 })
1721 .await
1722 .log_err();
1723
1724 let this = this.upgrade().with_context(|| "Dropped worktree")?;
1725 Ok(cx.read_entity(&this, |this, _| {
1726 paths_to_refresh
1727 .iter()
1728 .filter_map(|path| Some(this.entry_for_path(path)?.id))
1729 .collect()
1730 }))
1731 })
1732 }
1733
1734 fn expand_entry(
1735 &self,
1736 entry_id: ProjectEntryId,
1737 cx: &Context<Worktree>,
1738 ) -> Option<Task<Result<()>>> {
1739 let path = self.entry_for_id(entry_id)?.path.clone();
1740 let mut refresh = self.refresh_entries_for_paths(vec![path]);
1741 Some(cx.background_spawn(async move {
1742 refresh.next().await;
1743 Ok(())
1744 }))
1745 }
1746
1747 fn expand_all_for_entry(
1748 &self,
1749 entry_id: ProjectEntryId,
1750 cx: &Context<Worktree>,
1751 ) -> Option<Task<Result<()>>> {
1752 let path = self.entry_for_id(entry_id).unwrap().path.clone();
1753 let mut rx = self.add_path_prefix_to_scan(path);
1754 Some(cx.background_spawn(async move {
1755 rx.next().await;
1756 Ok(())
1757 }))
1758 }
1759
1760 pub fn refresh_entries_for_paths(&self, paths: Vec<Arc<RelPath>>) -> barrier::Receiver {
1761 let (tx, rx) = barrier::channel();
1762 self.scan_requests_tx
1763 .try_send(ScanRequest {
1764 relative_paths: paths,
1765 done: smallvec![tx],
1766 })
1767 .ok();
1768 rx
1769 }
1770
1771 #[cfg(feature = "test-support")]
1772 pub fn manually_refresh_entries_for_paths(
1773 &self,
1774 paths: Vec<Arc<RelPath>>,
1775 ) -> barrier::Receiver {
1776 self.refresh_entries_for_paths(paths)
1777 }
1778
1779 pub fn add_path_prefix_to_scan(&self, path_prefix: Arc<RelPath>) -> barrier::Receiver {
1780 let (tx, rx) = barrier::channel();
1781 self.path_prefixes_to_scan_tx
1782 .try_send(PathPrefixScanRequest {
1783 path: path_prefix,
1784 done: smallvec![tx],
1785 })
1786 .ok();
1787 rx
1788 }
1789
1790 pub fn refresh_entry(
1791 &self,
1792 path: Arc<RelPath>,
1793 old_path: Option<Arc<RelPath>>,
1794 cx: &Context<Worktree>,
1795 ) -> Task<Result<Option<Entry>>> {
1796 if self.settings.is_path_excluded(&path) {
1797 return Task::ready(Ok(None));
1798 }
1799 let paths = if let Some(old_path) = old_path.as_ref() {
1800 vec![old_path.clone(), path.clone()]
1801 } else {
1802 vec![path.clone()]
1803 };
1804 let t0 = Instant::now();
1805 let mut refresh = self.refresh_entries_for_paths(paths);
1806 // todo(lw): Hot foreground spawn
1807 cx.spawn(async move |this, cx| {
1808 refresh.recv().await;
1809 log::trace!("refreshed entry {path:?} in {:?}", t0.elapsed());
1810 let new_entry = this.read_with(cx, |this, _| {
1811 this.entry_for_path(&path).cloned().with_context(|| {
1812 format!("Could not find entry in worktree for {path:?} after refresh")
1813 })
1814 })??;
1815 Ok(Some(new_entry))
1816 })
1817 }
1818
1819 fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1820 where
1821 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1822 Fut: 'static + Send + Future<Output = bool>,
1823 {
1824 if let Some(observer) = self.update_observer.as_mut() {
1825 *observer.resume_updates.borrow_mut() = ();
1826 return;
1827 }
1828
1829 let (resume_updates_tx, mut resume_updates_rx) = watch::channel::<()>();
1830 let (snapshots_tx, mut snapshots_rx) =
1831 mpsc::unbounded::<(LocalSnapshot, UpdatedEntriesSet)>();
1832 snapshots_tx
1833 .unbounded_send((self.snapshot(), Arc::default()))
1834 .ok();
1835
1836 let worktree_id = cx.entity_id().as_u64();
1837 let _maintain_remote_snapshot = cx.background_spawn(async move {
1838 let mut is_first = true;
1839 while let Some((snapshot, entry_changes)) = snapshots_rx.next().await {
1840 let update = if is_first {
1841 is_first = false;
1842 snapshot.build_initial_update(project_id, worktree_id)
1843 } else {
1844 snapshot.build_update(project_id, worktree_id, entry_changes)
1845 };
1846
1847 for update in proto::split_worktree_update(update) {
1848 let _ = resume_updates_rx.try_recv();
1849 loop {
1850 let result = callback(update.clone());
1851 if result.await {
1852 break;
1853 } else {
1854 log::info!("waiting to resume updates");
1855 if resume_updates_rx.next().await.is_none() {
1856 return Some(());
1857 }
1858 }
1859 }
1860 }
1861 }
1862 Some(())
1863 });
1864
1865 self.update_observer = Some(UpdateObservationState {
1866 snapshots_tx,
1867 resume_updates: resume_updates_tx,
1868 _maintain_remote_snapshot,
1869 });
1870 }
1871
1872 pub fn share_private_files(&mut self, cx: &Context<Worktree>) {
1873 self.share_private_files = true;
1874 self.restart_background_scanners(cx);
1875 }
1876
1877 pub fn update_abs_path_and_refresh(
1878 &mut self,
1879 new_path: Arc<SanitizedPath>,
1880 cx: &Context<Worktree>,
1881 ) {
1882 self.snapshot.git_repositories = Default::default();
1883 self.snapshot.ignores_by_parent_abs_path = Default::default();
1884 let root_name = new_path
1885 .as_path()
1886 .file_name()
1887 .and_then(|f| f.to_str())
1888 .map_or(RelPath::empty().into(), |f| {
1889 RelPath::unix(f).unwrap().into()
1890 });
1891 self.snapshot.update_abs_path(new_path, root_name);
1892 self.restart_background_scanners(cx);
1893 }
1894}
1895
1896impl RemoteWorktree {
1897 pub fn project_id(&self) -> u64 {
1898 self.project_id
1899 }
1900
1901 pub fn client(&self) -> AnyProtoClient {
1902 self.client.clone()
1903 }
1904
1905 pub fn disconnected_from_host(&mut self) {
1906 self.updates_tx.take();
1907 self.snapshot_subscriptions.clear();
1908 self.disconnected = true;
1909 }
1910
1911 pub fn update_from_remote(&self, update: proto::UpdateWorktree) {
1912 if let Some(updates_tx) = &self.updates_tx {
1913 updates_tx
1914 .unbounded_send(update)
1915 .expect("consumer runs to completion");
1916 }
1917 }
1918
1919 fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1920 where
1921 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1922 Fut: 'static + Send + Future<Output = bool>,
1923 {
1924 let (tx, mut rx) = mpsc::unbounded();
1925 let initial_update = self
1926 .snapshot
1927 .build_initial_update(project_id, self.id().to_proto());
1928 self.update_observer = Some(tx);
1929 cx.spawn(async move |this, cx| {
1930 let mut update = initial_update;
1931 'outer: loop {
1932 // SSH projects use a special project ID of 0, and we need to
1933 // remap it to the correct one here.
1934 update.project_id = project_id;
1935
1936 for chunk in split_worktree_update(update) {
1937 if !callback(chunk).await {
1938 break 'outer;
1939 }
1940 }
1941
1942 if let Some(next_update) = rx.next().await {
1943 update = next_update;
1944 } else {
1945 break;
1946 }
1947 }
1948 this.update(cx, |this, _| {
1949 let this = this.as_remote_mut().unwrap();
1950 this.update_observer.take();
1951 })
1952 })
1953 .detach();
1954 }
1955
1956 fn observed_snapshot(&self, scan_id: usize) -> bool {
1957 self.completed_scan_id >= scan_id
1958 }
1959
1960 pub fn wait_for_snapshot(
1961 &mut self,
1962 scan_id: usize,
1963 ) -> impl Future<Output = Result<()>> + use<> {
1964 let (tx, rx) = oneshot::channel();
1965 if self.observed_snapshot(scan_id) {
1966 let _ = tx.send(());
1967 } else if self.disconnected {
1968 drop(tx);
1969 } else {
1970 match self
1971 .snapshot_subscriptions
1972 .binary_search_by_key(&scan_id, |probe| probe.0)
1973 {
1974 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
1975 }
1976 }
1977
1978 async move {
1979 rx.await?;
1980 Ok(())
1981 }
1982 }
1983
1984 pub fn insert_entry(
1985 &mut self,
1986 entry: proto::Entry,
1987 scan_id: usize,
1988 cx: &Context<Worktree>,
1989 ) -> Task<Result<Entry>> {
1990 let wait_for_snapshot = self.wait_for_snapshot(scan_id);
1991 cx.spawn(async move |this, cx| {
1992 wait_for_snapshot.await?;
1993 this.update(cx, |worktree, _| {
1994 let worktree = worktree.as_remote_mut().unwrap();
1995 let snapshot = &mut worktree.background_snapshot.lock().0;
1996 let entry = snapshot.insert_entry(entry, &worktree.file_scan_inclusions);
1997 worktree.snapshot = snapshot.clone();
1998 entry
1999 })?
2000 })
2001 }
2002
2003 fn delete_entry(
2004 &self,
2005 entry_id: ProjectEntryId,
2006 trash: bool,
2007 cx: &Context<Worktree>,
2008 ) -> Option<Task<Result<()>>> {
2009 let response = self.client.request(proto::DeleteProjectEntry {
2010 project_id: self.project_id,
2011 entry_id: entry_id.to_proto(),
2012 use_trash: trash,
2013 });
2014 Some(cx.spawn(async move |this, cx| {
2015 let response = response.await?;
2016 let scan_id = response.worktree_scan_id as usize;
2017
2018 this.update(cx, move |this, _| {
2019 this.as_remote_mut().unwrap().wait_for_snapshot(scan_id)
2020 })?
2021 .await?;
2022
2023 this.update(cx, |this, _| {
2024 let this = this.as_remote_mut().unwrap();
2025 let snapshot = &mut this.background_snapshot.lock().0;
2026 snapshot.delete_entry(entry_id);
2027 this.snapshot = snapshot.clone();
2028 })
2029 }))
2030 }
2031
2032 // fn rename_entry(
2033 // &self,
2034 // entry_id: ProjectEntryId,
2035 // new_path: impl Into<Arc<RelPath>>,
2036 // cx: &Context<Worktree>,
2037 // ) -> Task<Result<CreatedEntry>> {
2038 // let new_path: Arc<RelPath> = new_path.into();
2039 // let response = self.client.request(proto::RenameProjectEntry {
2040 // project_id: self.project_id,
2041 // entry_id: entry_id.to_proto(),
2042 // new_worktree_id: new_path.worktree_id,
2043 // new_path: new_path.as_ref().to_proto(),
2044 // });
2045 // cx.spawn(async move |this, cx| {
2046 // let response = response.await?;
2047 // match response.entry {
2048 // Some(entry) => this
2049 // .update(cx, |this, cx| {
2050 // this.as_remote_mut().unwrap().insert_entry(
2051 // entry,
2052 // response.worktree_scan_id as usize,
2053 // cx,
2054 // )
2055 // })?
2056 // .await
2057 // .map(CreatedEntry::Included),
2058 // None => {
2059 // let abs_path =
2060 // this.read_with(cx, |worktree, _| worktree.absolutize(&new_path))?;
2061 // Ok(CreatedEntry::Excluded { abs_path })
2062 // }
2063 // }
2064 // })
2065 // }
2066
2067 fn copy_external_entries(
2068 &self,
2069 target_directory: Arc<RelPath>,
2070 paths_to_copy: Vec<Arc<Path>>,
2071 local_fs: Arc<dyn Fs>,
2072 cx: &Context<Worktree>,
2073 ) -> Task<anyhow::Result<Vec<ProjectEntryId>>> {
2074 let client = self.client.clone();
2075 let worktree_id = self.id().to_proto();
2076 let project_id = self.project_id;
2077
2078 cx.background_spawn(async move {
2079 let mut requests = Vec::new();
2080 for root_path_to_copy in paths_to_copy {
2081 let Some(filename) = root_path_to_copy
2082 .file_name()
2083 .and_then(|name| name.to_str())
2084 .and_then(|filename| RelPath::unix(filename).ok())
2085 else {
2086 continue;
2087 };
2088 for (abs_path, is_directory) in
2089 read_dir_items(local_fs.as_ref(), &root_path_to_copy).await?
2090 {
2091 let Some(relative_path) = abs_path
2092 .strip_prefix(&root_path_to_copy)
2093 .map_err(|e| anyhow::Error::from(e))
2094 .and_then(|relative_path| RelPath::new(relative_path, PathStyle::local()))
2095 .log_err()
2096 else {
2097 continue;
2098 };
2099 let content = if is_directory {
2100 None
2101 } else {
2102 Some(local_fs.load_bytes(&abs_path).await?)
2103 };
2104
2105 let mut target_path = target_directory.join(filename);
2106 if relative_path.file_name().is_some() {
2107 target_path = target_path.join(&relative_path);
2108 }
2109
2110 requests.push(proto::CreateProjectEntry {
2111 project_id,
2112 worktree_id,
2113 path: target_path.to_proto(),
2114 is_directory,
2115 content,
2116 });
2117 }
2118 }
2119 requests.sort_unstable_by(|a, b| a.path.cmp(&b.path));
2120 requests.dedup();
2121
2122 let mut copied_entry_ids = Vec::new();
2123 for request in requests {
2124 let response = client.request(request).await?;
2125 copied_entry_ids.extend(response.entry.map(|e| ProjectEntryId::from_proto(e.id)));
2126 }
2127
2128 Ok(copied_entry_ids)
2129 })
2130 }
2131}
2132
2133impl Snapshot {
2134 pub fn new(
2135 id: u64,
2136 root_name: Arc<RelPath>,
2137 abs_path: Arc<Path>,
2138 path_style: PathStyle,
2139 ) -> Self {
2140 Snapshot {
2141 id: WorktreeId::from_usize(id as usize),
2142 abs_path: SanitizedPath::from_arc(abs_path),
2143 path_style,
2144 root_char_bag: root_name
2145 .as_unix_str()
2146 .chars()
2147 .map(|c| c.to_ascii_lowercase())
2148 .collect(),
2149 root_name,
2150 always_included_entries: Default::default(),
2151 entries_by_path: Default::default(),
2152 entries_by_id: Default::default(),
2153 scan_id: 1,
2154 completed_scan_id: 0,
2155 }
2156 }
2157
2158 pub fn id(&self) -> WorktreeId {
2159 self.id
2160 }
2161
2162 // TODO:
2163 // Consider the following:
2164 //
2165 // ```rust
2166 // let abs_path: Arc<Path> = snapshot.abs_path(); // e.g. "C:\Users\user\Desktop\project"
2167 // let some_non_trimmed_path = Path::new("\\\\?\\C:\\Users\\user\\Desktop\\project\\main.rs");
2168 // // The caller perform some actions here:
2169 // some_non_trimmed_path.strip_prefix(abs_path); // This fails
2170 // some_non_trimmed_path.starts_with(abs_path); // This fails too
2171 // ```
2172 //
2173 // This is definitely a bug, but it's not clear if we should handle it here or not.
2174 pub fn abs_path(&self) -> &Arc<Path> {
2175 SanitizedPath::cast_arc_ref(&self.abs_path)
2176 }
2177
2178 fn build_initial_update(&self, project_id: u64, worktree_id: u64) -> proto::UpdateWorktree {
2179 let mut updated_entries = self
2180 .entries_by_path
2181 .iter()
2182 .map(proto::Entry::from)
2183 .collect::<Vec<_>>();
2184 updated_entries.sort_unstable_by_key(|e| e.id);
2185
2186 proto::UpdateWorktree {
2187 project_id,
2188 worktree_id,
2189 abs_path: self.abs_path().to_string_lossy().into_owned(),
2190 root_name: self.root_name().to_proto(),
2191 updated_entries,
2192 removed_entries: Vec::new(),
2193 scan_id: self.scan_id as u64,
2194 is_last_update: self.completed_scan_id == self.scan_id,
2195 // Sent in separate messages.
2196 updated_repositories: Vec::new(),
2197 removed_repositories: Vec::new(),
2198 }
2199 }
2200
2201 pub fn work_directory_abs_path(&self, work_directory: &WorkDirectory) -> PathBuf {
2202 match work_directory {
2203 WorkDirectory::InProject { relative_path } => self.absolutize(relative_path),
2204 WorkDirectory::AboveProject { absolute_path, .. } => absolute_path.as_ref().to_owned(),
2205 }
2206 }
2207
2208 pub fn absolutize(&self, path: &RelPath) -> PathBuf {
2209 if path.file_name().is_some() {
2210 let mut abs_path = self.abs_path.to_string();
2211 for component in path.components() {
2212 if !abs_path.ends_with(self.path_style.primary_separator()) {
2213 abs_path.push_str(self.path_style.primary_separator());
2214 }
2215 abs_path.push_str(component);
2216 }
2217 PathBuf::from(abs_path)
2218 } else {
2219 self.abs_path.as_path().to_path_buf()
2220 }
2221 }
2222
2223 pub fn contains_entry(&self, entry_id: ProjectEntryId) -> bool {
2224 self.entries_by_id.get(&entry_id, ()).is_some()
2225 }
2226
2227 fn insert_entry(
2228 &mut self,
2229 entry: proto::Entry,
2230 always_included_paths: &PathMatcher,
2231 ) -> Result<Entry> {
2232 let entry = Entry::try_from((&self.root_char_bag, always_included_paths, entry))?;
2233 let old_entry = self.entries_by_id.insert_or_replace(
2234 PathEntry {
2235 id: entry.id,
2236 path: entry.path.clone(),
2237 is_ignored: entry.is_ignored,
2238 scan_id: 0,
2239 },
2240 (),
2241 );
2242 if let Some(old_entry) = old_entry {
2243 self.entries_by_path.remove(&PathKey(old_entry.path), ());
2244 }
2245 self.entries_by_path.insert_or_replace(entry.clone(), ());
2246 Ok(entry)
2247 }
2248
2249 fn delete_entry(&mut self, entry_id: ProjectEntryId) -> Option<Arc<RelPath>> {
2250 let removed_entry = self.entries_by_id.remove(&entry_id, ())?;
2251 self.entries_by_path = {
2252 let mut cursor = self.entries_by_path.cursor::<TraversalProgress>(());
2253 let mut new_entries_by_path =
2254 cursor.slice(&TraversalTarget::path(&removed_entry.path), Bias::Left);
2255 while let Some(entry) = cursor.item() {
2256 if entry.path.starts_with(&removed_entry.path) {
2257 self.entries_by_id.remove(&entry.id, ());
2258 cursor.next();
2259 } else {
2260 break;
2261 }
2262 }
2263 new_entries_by_path.append(cursor.suffix(), ());
2264 new_entries_by_path
2265 };
2266
2267 Some(removed_entry.path)
2268 }
2269
2270 fn update_abs_path(&mut self, abs_path: Arc<SanitizedPath>, root_name: Arc<RelPath>) {
2271 self.abs_path = abs_path;
2272 if root_name != self.root_name {
2273 self.root_char_bag = root_name
2274 .as_unix_str()
2275 .chars()
2276 .map(|c| c.to_ascii_lowercase())
2277 .collect();
2278 self.root_name = root_name;
2279 }
2280 }
2281
2282 fn apply_remote_update(
2283 &mut self,
2284 update: proto::UpdateWorktree,
2285 always_included_paths: &PathMatcher,
2286 ) {
2287 log::debug!(
2288 "applying remote worktree update. {} entries updated, {} removed",
2289 update.updated_entries.len(),
2290 update.removed_entries.len()
2291 );
2292 if let Some(root_name) = RelPath::from_proto(&update.root_name).log_err() {
2293 self.update_abs_path(
2294 SanitizedPath::new_arc(&Path::new(&update.abs_path)),
2295 root_name,
2296 );
2297 }
2298
2299 let mut entries_by_path_edits = Vec::new();
2300 let mut entries_by_id_edits = Vec::new();
2301
2302 for entry_id in update.removed_entries {
2303 let entry_id = ProjectEntryId::from_proto(entry_id);
2304 entries_by_id_edits.push(Edit::Remove(entry_id));
2305 if let Some(entry) = self.entry_for_id(entry_id) {
2306 entries_by_path_edits.push(Edit::Remove(PathKey(entry.path.clone())));
2307 }
2308 }
2309
2310 for entry in update.updated_entries {
2311 let Some(entry) =
2312 Entry::try_from((&self.root_char_bag, always_included_paths, entry)).log_err()
2313 else {
2314 continue;
2315 };
2316 if let Some(PathEntry { path, .. }) = self.entries_by_id.get(&entry.id, ()) {
2317 entries_by_path_edits.push(Edit::Remove(PathKey(path.clone())));
2318 }
2319 if let Some(old_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2320 && old_entry.id != entry.id
2321 {
2322 entries_by_id_edits.push(Edit::Remove(old_entry.id));
2323 }
2324 entries_by_id_edits.push(Edit::Insert(PathEntry {
2325 id: entry.id,
2326 path: entry.path.clone(),
2327 is_ignored: entry.is_ignored,
2328 scan_id: 0,
2329 }));
2330 entries_by_path_edits.push(Edit::Insert(entry));
2331 }
2332
2333 self.entries_by_path.edit(entries_by_path_edits, ());
2334 self.entries_by_id.edit(entries_by_id_edits, ());
2335
2336 self.scan_id = update.scan_id as usize;
2337 if update.is_last_update {
2338 self.completed_scan_id = update.scan_id as usize;
2339 }
2340 }
2341
2342 pub fn entry_count(&self) -> usize {
2343 self.entries_by_path.summary().count
2344 }
2345
2346 pub fn visible_entry_count(&self) -> usize {
2347 self.entries_by_path.summary().non_ignored_count
2348 }
2349
2350 pub fn dir_count(&self) -> usize {
2351 let summary = self.entries_by_path.summary();
2352 summary.count - summary.file_count
2353 }
2354
2355 pub fn visible_dir_count(&self) -> usize {
2356 let summary = self.entries_by_path.summary();
2357 summary.non_ignored_count - summary.non_ignored_file_count
2358 }
2359
2360 pub fn file_count(&self) -> usize {
2361 self.entries_by_path.summary().file_count
2362 }
2363
2364 pub fn visible_file_count(&self) -> usize {
2365 self.entries_by_path.summary().non_ignored_file_count
2366 }
2367
2368 fn traverse_from_offset(
2369 &self,
2370 include_files: bool,
2371 include_dirs: bool,
2372 include_ignored: bool,
2373 start_offset: usize,
2374 ) -> Traversal<'_> {
2375 let mut cursor = self.entries_by_path.cursor(());
2376 cursor.seek(
2377 &TraversalTarget::Count {
2378 count: start_offset,
2379 include_files,
2380 include_dirs,
2381 include_ignored,
2382 },
2383 Bias::Right,
2384 );
2385 Traversal {
2386 snapshot: self,
2387 cursor,
2388 include_files,
2389 include_dirs,
2390 include_ignored,
2391 }
2392 }
2393
2394 pub fn traverse_from_path(
2395 &self,
2396 include_files: bool,
2397 include_dirs: bool,
2398 include_ignored: bool,
2399 path: &RelPath,
2400 ) -> Traversal<'_> {
2401 Traversal::new(self, include_files, include_dirs, include_ignored, path)
2402 }
2403
2404 pub fn files(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2405 self.traverse_from_offset(true, false, include_ignored, start)
2406 }
2407
2408 pub fn directories(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2409 self.traverse_from_offset(false, true, include_ignored, start)
2410 }
2411
2412 pub fn entries(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2413 self.traverse_from_offset(true, true, include_ignored, start)
2414 }
2415
2416 pub fn paths(&self) -> impl Iterator<Item = &RelPath> {
2417 self.entries_by_path
2418 .cursor::<()>(())
2419 .filter(move |entry| !entry.path.is_empty())
2420 .map(|entry| entry.path.as_ref())
2421 }
2422
2423 pub fn child_entries<'a>(&'a self, parent_path: &'a RelPath) -> ChildEntriesIter<'a> {
2424 let options = ChildEntriesOptions {
2425 include_files: true,
2426 include_dirs: true,
2427 include_ignored: true,
2428 };
2429 self.child_entries_with_options(parent_path, options)
2430 }
2431
2432 pub fn child_entries_with_options<'a>(
2433 &'a self,
2434 parent_path: &'a RelPath,
2435 options: ChildEntriesOptions,
2436 ) -> ChildEntriesIter<'a> {
2437 let mut cursor = self.entries_by_path.cursor(());
2438 cursor.seek(&TraversalTarget::path(parent_path), Bias::Right);
2439 let traversal = Traversal {
2440 snapshot: self,
2441 cursor,
2442 include_files: options.include_files,
2443 include_dirs: options.include_dirs,
2444 include_ignored: options.include_ignored,
2445 };
2446 ChildEntriesIter {
2447 traversal,
2448 parent_path,
2449 }
2450 }
2451
2452 pub fn root_entry(&self) -> Option<&Entry> {
2453 self.entries_by_path.first()
2454 }
2455
2456 /// Returns `None` for a single file worktree, or `Some(self.abs_path())` if
2457 /// it is a directory.
2458 pub fn root_dir(&self) -> Option<Arc<Path>> {
2459 self.root_entry()
2460 .filter(|entry| entry.is_dir())
2461 .map(|_| self.abs_path().clone())
2462 }
2463
2464 pub fn root_name(&self) -> &RelPath {
2465 &self.root_name
2466 }
2467
2468 pub fn root_name_str(&self) -> &str {
2469 self.root_name.as_unix_str()
2470 }
2471
2472 pub fn scan_id(&self) -> usize {
2473 self.scan_id
2474 }
2475
2476 pub fn entry_for_path(&self, path: &RelPath) -> Option<&Entry> {
2477 self.traverse_from_path(true, true, true, path)
2478 .entry()
2479 .and_then(|entry| {
2480 if entry.path.as_ref() == path {
2481 Some(entry)
2482 } else {
2483 None
2484 }
2485 })
2486 }
2487
2488 /// Resolves a path to an executable using the following heuristics:
2489 ///
2490 /// 1. If the path starts with `~`, it is expanded to the user's home directory.
2491 /// 2. If the path is relative and contains more than one component,
2492 /// it is joined to the worktree root path.
2493 /// 3. If the path is relative and exists in the worktree
2494 /// (even if falls under an exclusion filter),
2495 /// it is joined to the worktree root path.
2496 /// 4. Otherwise the path is returned unmodified.
2497 ///
2498 /// Relative paths that do not exist in the worktree may
2499 /// still be found using the `PATH` environment variable.
2500 pub fn resolve_executable_path(&self, path: PathBuf) -> PathBuf {
2501 if let Some(path_str) = path.to_str() {
2502 if let Some(remaining_path) = path_str.strip_prefix("~/") {
2503 return home_dir().join(remaining_path);
2504 } else if path_str == "~" {
2505 return home_dir().to_path_buf();
2506 }
2507 }
2508
2509 if let Ok(rel_path) = RelPath::new(&path, self.path_style)
2510 && (path.components().count() > 1 || self.entry_for_path(&rel_path).is_some())
2511 {
2512 self.abs_path().join(path)
2513 } else {
2514 path
2515 }
2516 }
2517
2518 pub fn entry_for_id(&self, id: ProjectEntryId) -> Option<&Entry> {
2519 let entry = self.entries_by_id.get(&id, ())?;
2520 self.entry_for_path(&entry.path)
2521 }
2522
2523 pub fn path_style(&self) -> PathStyle {
2524 self.path_style
2525 }
2526}
2527
2528impl LocalSnapshot {
2529 fn local_repo_for_work_directory_path(&self, path: &RelPath) -> Option<&LocalRepositoryEntry> {
2530 self.git_repositories
2531 .iter()
2532 .map(|(_, entry)| entry)
2533 .find(|entry| entry.work_directory.path_key() == PathKey(path.into()))
2534 }
2535
2536 fn build_update(
2537 &self,
2538 project_id: u64,
2539 worktree_id: u64,
2540 entry_changes: UpdatedEntriesSet,
2541 ) -> proto::UpdateWorktree {
2542 let mut updated_entries = Vec::new();
2543 let mut removed_entries = Vec::new();
2544
2545 for (_, entry_id, path_change) in entry_changes.iter() {
2546 if let PathChange::Removed = path_change {
2547 removed_entries.push(entry_id.0 as u64);
2548 } else if let Some(entry) = self.entry_for_id(*entry_id) {
2549 updated_entries.push(proto::Entry::from(entry));
2550 }
2551 }
2552
2553 removed_entries.sort_unstable();
2554 updated_entries.sort_unstable_by_key(|e| e.id);
2555
2556 // TODO - optimize, knowing that removed_entries are sorted.
2557 removed_entries.retain(|id| updated_entries.binary_search_by_key(id, |e| e.id).is_err());
2558
2559 proto::UpdateWorktree {
2560 project_id,
2561 worktree_id,
2562 abs_path: self.abs_path().to_string_lossy().into_owned(),
2563 root_name: self.root_name().to_proto(),
2564 updated_entries,
2565 removed_entries,
2566 scan_id: self.scan_id as u64,
2567 is_last_update: self.completed_scan_id == self.scan_id,
2568 // Sent in separate messages.
2569 updated_repositories: Vec::new(),
2570 removed_repositories: Vec::new(),
2571 }
2572 }
2573
2574 async fn insert_entry(&mut self, mut entry: Entry, fs: &dyn Fs) -> Entry {
2575 log::trace!("insert entry {:?}", entry.path);
2576 if entry.is_file() && entry.path.file_name() == Some(&GITIGNORE) {
2577 let abs_path = self.absolutize(&entry.path);
2578 match build_gitignore(&abs_path, fs).await {
2579 Ok(ignore) => {
2580 self.ignores_by_parent_abs_path
2581 .insert(abs_path.parent().unwrap().into(), (Arc::new(ignore), true));
2582 }
2583 Err(error) => {
2584 log::error!(
2585 "error loading .gitignore file {:?} - {:?}",
2586 &entry.path,
2587 error
2588 );
2589 }
2590 }
2591 }
2592
2593 if entry.kind == EntryKind::PendingDir
2594 && let Some(existing_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2595 {
2596 entry.kind = existing_entry.kind;
2597 }
2598
2599 let scan_id = self.scan_id;
2600 let removed = self.entries_by_path.insert_or_replace(entry.clone(), ());
2601 if let Some(removed) = removed
2602 && removed.id != entry.id
2603 {
2604 self.entries_by_id.remove(&removed.id, ());
2605 }
2606 self.entries_by_id.insert_or_replace(
2607 PathEntry {
2608 id: entry.id,
2609 path: entry.path.clone(),
2610 is_ignored: entry.is_ignored,
2611 scan_id,
2612 },
2613 (),
2614 );
2615
2616 entry
2617 }
2618
2619 fn ancestor_inodes_for_path(&self, path: &RelPath) -> TreeSet<u64> {
2620 let mut inodes = TreeSet::default();
2621 for ancestor in path.ancestors().skip(1) {
2622 if let Some(entry) = self.entry_for_path(ancestor) {
2623 inodes.insert(entry.inode);
2624 }
2625 }
2626 inodes
2627 }
2628
2629 async fn ignore_stack_for_abs_path(
2630 &self,
2631 abs_path: &Path,
2632 is_dir: bool,
2633 fs: &dyn Fs,
2634 ) -> IgnoreStack {
2635 let mut new_ignores = Vec::new();
2636 let mut repo_root = None;
2637 for (index, ancestor) in abs_path.ancestors().enumerate() {
2638 if index > 0 {
2639 if let Some((ignore, _)) = self.ignores_by_parent_abs_path.get(ancestor) {
2640 new_ignores.push((ancestor, Some(ignore.clone())));
2641 } else {
2642 new_ignores.push((ancestor, None));
2643 }
2644 }
2645
2646 let metadata = fs.metadata(&ancestor.join(DOT_GIT)).await.ok().flatten();
2647 if metadata.is_some() {
2648 repo_root = Some(Arc::from(ancestor));
2649 break;
2650 }
2651 }
2652
2653 let mut ignore_stack = if let Some(global_gitignore) = self.global_gitignore.clone() {
2654 IgnoreStack::global(global_gitignore)
2655 } else {
2656 IgnoreStack::none()
2657 };
2658
2659 if let Some((repo_exclude, _)) = repo_root
2660 .as_ref()
2661 .and_then(|abs_path| self.repo_exclude_by_work_dir_abs_path.get(abs_path))
2662 {
2663 ignore_stack = ignore_stack.append(IgnoreKind::RepoExclude, repo_exclude.clone());
2664 }
2665 ignore_stack.repo_root = repo_root;
2666 for (parent_abs_path, ignore) in new_ignores.into_iter().rev() {
2667 if ignore_stack.is_abs_path_ignored(parent_abs_path, true) {
2668 ignore_stack = IgnoreStack::all();
2669 break;
2670 } else if let Some(ignore) = ignore {
2671 ignore_stack =
2672 ignore_stack.append(IgnoreKind::Gitignore(parent_abs_path.into()), ignore);
2673 }
2674 }
2675
2676 if ignore_stack.is_abs_path_ignored(abs_path, is_dir) {
2677 ignore_stack = IgnoreStack::all();
2678 }
2679
2680 ignore_stack
2681 }
2682
2683 #[cfg(test)]
2684 fn expanded_entries(&self) -> impl Iterator<Item = &Entry> {
2685 self.entries_by_path
2686 .cursor::<()>(())
2687 .filter(|entry| entry.kind == EntryKind::Dir && (entry.is_external || entry.is_ignored))
2688 }
2689
2690 #[cfg(test)]
2691 pub fn check_invariants(&self, git_state: bool) {
2692 use pretty_assertions::assert_eq;
2693
2694 assert_eq!(
2695 self.entries_by_path
2696 .cursor::<()>(())
2697 .map(|e| (&e.path, e.id))
2698 .collect::<Vec<_>>(),
2699 self.entries_by_id
2700 .cursor::<()>(())
2701 .map(|e| (&e.path, e.id))
2702 .collect::<collections::BTreeSet<_>>()
2703 .into_iter()
2704 .collect::<Vec<_>>(),
2705 "entries_by_path and entries_by_id are inconsistent"
2706 );
2707
2708 let mut files = self.files(true, 0);
2709 let mut visible_files = self.files(false, 0);
2710 for entry in self.entries_by_path.cursor::<()>(()) {
2711 if entry.is_file() {
2712 assert_eq!(files.next().unwrap().inode, entry.inode);
2713 if (!entry.is_ignored && !entry.is_external) || entry.is_always_included {
2714 assert_eq!(visible_files.next().unwrap().inode, entry.inode);
2715 }
2716 }
2717 }
2718
2719 assert!(files.next().is_none());
2720 assert!(visible_files.next().is_none());
2721
2722 let mut bfs_paths = Vec::new();
2723 let mut stack = self
2724 .root_entry()
2725 .map(|e| e.path.as_ref())
2726 .into_iter()
2727 .collect::<Vec<_>>();
2728 while let Some(path) = stack.pop() {
2729 bfs_paths.push(path);
2730 let ix = stack.len();
2731 for child_entry in self.child_entries(path) {
2732 stack.insert(ix, &child_entry.path);
2733 }
2734 }
2735
2736 let dfs_paths_via_iter = self
2737 .entries_by_path
2738 .cursor::<()>(())
2739 .map(|e| e.path.as_ref())
2740 .collect::<Vec<_>>();
2741 assert_eq!(bfs_paths, dfs_paths_via_iter);
2742
2743 let dfs_paths_via_traversal = self
2744 .entries(true, 0)
2745 .map(|e| e.path.as_ref())
2746 .collect::<Vec<_>>();
2747
2748 assert_eq!(dfs_paths_via_traversal, dfs_paths_via_iter);
2749
2750 if git_state {
2751 for ignore_parent_abs_path in self.ignores_by_parent_abs_path.keys() {
2752 let ignore_parent_path = &RelPath::new(
2753 ignore_parent_abs_path
2754 .strip_prefix(self.abs_path.as_path())
2755 .unwrap(),
2756 PathStyle::local(),
2757 )
2758 .unwrap();
2759 assert!(self.entry_for_path(ignore_parent_path).is_some());
2760 assert!(
2761 self.entry_for_path(
2762 &ignore_parent_path.join(RelPath::unix(GITIGNORE).unwrap())
2763 )
2764 .is_some()
2765 );
2766 }
2767 }
2768 }
2769
2770 #[cfg(test)]
2771 pub fn entries_without_ids(&self, include_ignored: bool) -> Vec<(&RelPath, u64, bool)> {
2772 let mut paths = Vec::new();
2773 for entry in self.entries_by_path.cursor::<()>(()) {
2774 if include_ignored || !entry.is_ignored {
2775 paths.push((entry.path.as_ref(), entry.inode, entry.is_ignored));
2776 }
2777 }
2778 paths.sort_by(|a, b| a.0.cmp(b.0));
2779 paths
2780 }
2781}
2782
2783impl BackgroundScannerState {
2784 fn should_scan_directory(&self, entry: &Entry) -> bool {
2785 (self.scanning_enabled && !entry.is_external && (!entry.is_ignored || entry.is_always_included))
2786 || entry.path.file_name() == Some(DOT_GIT)
2787 || entry.path.file_name() == Some(local_settings_folder_name())
2788 || entry.path.file_name() == Some(local_vscode_folder_name())
2789 || self.scanned_dirs.contains(&entry.id) // If we've ever scanned it, keep scanning
2790 || self
2791 .paths_to_scan
2792 .iter()
2793 .any(|p| p.starts_with(&entry.path))
2794 || self
2795 .path_prefixes_to_scan
2796 .iter()
2797 .any(|p| entry.path.starts_with(p))
2798 }
2799
2800 async fn enqueue_scan_dir(
2801 &self,
2802 abs_path: Arc<Path>,
2803 entry: &Entry,
2804 scan_job_tx: &Sender<ScanJob>,
2805 fs: &dyn Fs,
2806 ) {
2807 let path = entry.path.clone();
2808 let ignore_stack = self
2809 .snapshot
2810 .ignore_stack_for_abs_path(&abs_path, true, fs)
2811 .await;
2812 let mut ancestor_inodes = self.snapshot.ancestor_inodes_for_path(&path);
2813
2814 if !ancestor_inodes.contains(&entry.inode) {
2815 ancestor_inodes.insert(entry.inode);
2816 scan_job_tx
2817 .try_send(ScanJob {
2818 abs_path,
2819 path,
2820 ignore_stack,
2821 scan_queue: scan_job_tx.clone(),
2822 ancestor_inodes,
2823 is_external: entry.is_external,
2824 })
2825 .unwrap();
2826 }
2827 }
2828
2829 fn reuse_entry_id(&mut self, entry: &mut Entry) {
2830 if let Some(mtime) = entry.mtime {
2831 // If an entry with the same inode was removed from the worktree during this scan,
2832 // then it *might* represent the same file or directory. But the OS might also have
2833 // re-used the inode for a completely different file or directory.
2834 //
2835 // Conditionally reuse the old entry's id:
2836 // * if the mtime is the same, the file was probably been renamed.
2837 // * if the path is the same, the file may just have been updated
2838 if let Some(removed_entry) = self.removed_entries.remove(&entry.inode) {
2839 if removed_entry.mtime == Some(mtime) || removed_entry.path == entry.path {
2840 entry.id = removed_entry.id;
2841 }
2842 } else if let Some(existing_entry) = self.snapshot.entry_for_path(&entry.path) {
2843 entry.id = existing_entry.id;
2844 }
2845 }
2846 }
2847
2848 fn entry_id_for(
2849 &mut self,
2850 next_entry_id: &AtomicUsize,
2851 path: &RelPath,
2852 metadata: &fs::Metadata,
2853 ) -> ProjectEntryId {
2854 // If an entry with the same inode was removed from the worktree during this scan,
2855 // then it *might* represent the same file or directory. But the OS might also have
2856 // re-used the inode for a completely different file or directory.
2857 //
2858 // Conditionally reuse the old entry's id:
2859 // * if the mtime is the same, the file was probably been renamed.
2860 // * if the path is the same, the file may just have been updated
2861 if let Some(removed_entry) = self.removed_entries.remove(&metadata.inode) {
2862 if removed_entry.mtime == Some(metadata.mtime) || *removed_entry.path == *path {
2863 return removed_entry.id;
2864 }
2865 } else if let Some(existing_entry) = self.snapshot.entry_for_path(path) {
2866 return existing_entry.id;
2867 }
2868 ProjectEntryId::new(next_entry_id)
2869 }
2870
2871 async fn insert_entry(&mut self, entry: Entry, fs: &dyn Fs, watcher: &dyn Watcher) -> Entry {
2872 let entry = self.snapshot.insert_entry(entry, fs).await;
2873 if entry.path.file_name() == Some(&DOT_GIT) {
2874 self.insert_git_repository(entry.path.clone(), fs, watcher)
2875 .await;
2876 }
2877
2878 #[cfg(test)]
2879 self.snapshot.check_invariants(false);
2880
2881 entry
2882 }
2883
2884 fn populate_dir(
2885 &mut self,
2886 parent_path: Arc<RelPath>,
2887 entries: impl IntoIterator<Item = Entry>,
2888 ignore: Option<Arc<Gitignore>>,
2889 ) {
2890 let mut parent_entry = if let Some(parent_entry) = self
2891 .snapshot
2892 .entries_by_path
2893 .get(&PathKey(parent_path.clone()), ())
2894 {
2895 parent_entry.clone()
2896 } else {
2897 log::warn!(
2898 "populating a directory {:?} that has been removed",
2899 parent_path
2900 );
2901 return;
2902 };
2903
2904 match parent_entry.kind {
2905 EntryKind::PendingDir | EntryKind::UnloadedDir => parent_entry.kind = EntryKind::Dir,
2906 EntryKind::Dir => {}
2907 _ => return,
2908 }
2909
2910 if let Some(ignore) = ignore {
2911 let abs_parent_path = self
2912 .snapshot
2913 .abs_path
2914 .as_path()
2915 .join(parent_path.as_std_path())
2916 .into();
2917 self.snapshot
2918 .ignores_by_parent_abs_path
2919 .insert(abs_parent_path, (ignore, false));
2920 }
2921
2922 let parent_entry_id = parent_entry.id;
2923 self.scanned_dirs.insert(parent_entry_id);
2924 let mut entries_by_path_edits = vec![Edit::Insert(parent_entry)];
2925 let mut entries_by_id_edits = Vec::new();
2926
2927 for entry in entries {
2928 entries_by_id_edits.push(Edit::Insert(PathEntry {
2929 id: entry.id,
2930 path: entry.path.clone(),
2931 is_ignored: entry.is_ignored,
2932 scan_id: self.snapshot.scan_id,
2933 }));
2934 entries_by_path_edits.push(Edit::Insert(entry));
2935 }
2936
2937 self.snapshot
2938 .entries_by_path
2939 .edit(entries_by_path_edits, ());
2940 self.snapshot.entries_by_id.edit(entries_by_id_edits, ());
2941
2942 if let Err(ix) = self.changed_paths.binary_search(&parent_path) {
2943 self.changed_paths.insert(ix, parent_path.clone());
2944 }
2945
2946 #[cfg(test)]
2947 self.snapshot.check_invariants(false);
2948 }
2949
2950 fn remove_path(&mut self, path: &RelPath) {
2951 log::trace!("background scanner removing path {path:?}");
2952 let mut new_entries;
2953 let removed_entries;
2954 {
2955 let mut cursor = self
2956 .snapshot
2957 .entries_by_path
2958 .cursor::<TraversalProgress>(());
2959 new_entries = cursor.slice(&TraversalTarget::path(path), Bias::Left);
2960 removed_entries = cursor.slice(&TraversalTarget::successor(path), Bias::Left);
2961 new_entries.append(cursor.suffix(), ());
2962 }
2963 self.snapshot.entries_by_path = new_entries;
2964
2965 let mut removed_ids = Vec::with_capacity(removed_entries.summary().count);
2966 for entry in removed_entries.cursor::<()>(()) {
2967 match self.removed_entries.entry(entry.inode) {
2968 hash_map::Entry::Occupied(mut e) => {
2969 let prev_removed_entry = e.get_mut();
2970 if entry.id > prev_removed_entry.id {
2971 *prev_removed_entry = entry.clone();
2972 }
2973 }
2974 hash_map::Entry::Vacant(e) => {
2975 e.insert(entry.clone());
2976 }
2977 }
2978
2979 if entry.path.file_name() == Some(GITIGNORE) {
2980 let abs_parent_path = self.snapshot.absolutize(&entry.path.parent().unwrap());
2981 if let Some((_, needs_update)) = self
2982 .snapshot
2983 .ignores_by_parent_abs_path
2984 .get_mut(abs_parent_path.as_path())
2985 {
2986 *needs_update = true;
2987 }
2988 }
2989
2990 if let Err(ix) = removed_ids.binary_search(&entry.id) {
2991 removed_ids.insert(ix, entry.id);
2992 }
2993 }
2994
2995 self.snapshot
2996 .entries_by_id
2997 .edit(removed_ids.iter().map(|&id| Edit::Remove(id)).collect(), ());
2998 self.snapshot
2999 .git_repositories
3000 .retain(|id, _| removed_ids.binary_search(id).is_err());
3001
3002 #[cfg(test)]
3003 self.snapshot.check_invariants(false);
3004 }
3005
3006 async fn insert_git_repository(
3007 &mut self,
3008 dot_git_path: Arc<RelPath>,
3009 fs: &dyn Fs,
3010 watcher: &dyn Watcher,
3011 ) {
3012 let work_dir_path: Arc<RelPath> = match dot_git_path.parent() {
3013 Some(parent_dir) => {
3014 // Guard against repositories inside the repository metadata
3015 if parent_dir
3016 .components()
3017 .any(|component| component == DOT_GIT)
3018 {
3019 log::debug!(
3020 "not building git repository for nested `.git` directory, `.git` path in the worktree: {dot_git_path:?}"
3021 );
3022 return;
3023 };
3024
3025 parent_dir.into()
3026 }
3027 None => {
3028 // `dot_git_path.parent().is_none()` means `.git` directory is the opened worktree itself,
3029 // no files inside that directory are tracked by git, so no need to build the repo around it
3030 log::debug!(
3031 "not building git repository for the worktree itself, `.git` path in the worktree: {dot_git_path:?}"
3032 );
3033 return;
3034 }
3035 };
3036
3037 let dot_git_abs_path = Arc::from(self.snapshot.absolutize(&dot_git_path).as_ref());
3038
3039 self.insert_git_repository_for_path(
3040 WorkDirectory::InProject {
3041 relative_path: work_dir_path,
3042 },
3043 dot_git_abs_path,
3044 fs,
3045 watcher,
3046 )
3047 .await
3048 .log_err();
3049 }
3050
3051 async fn insert_git_repository_for_path(
3052 &mut self,
3053 work_directory: WorkDirectory,
3054 dot_git_abs_path: Arc<Path>,
3055 fs: &dyn Fs,
3056 watcher: &dyn Watcher,
3057 ) -> Result<LocalRepositoryEntry> {
3058 let work_dir_entry = self
3059 .snapshot
3060 .entry_for_path(&work_directory.path_key().0)
3061 .with_context(|| {
3062 format!(
3063 "working directory `{}` not indexed",
3064 work_directory
3065 .path_key()
3066 .0
3067 .display(self.snapshot.path_style)
3068 )
3069 })?;
3070 let work_directory_abs_path = self.snapshot.work_directory_abs_path(&work_directory);
3071
3072 let (repository_dir_abs_path, common_dir_abs_path) =
3073 discover_git_paths(&dot_git_abs_path, fs).await;
3074 watcher
3075 .add(&common_dir_abs_path)
3076 .context("failed to add common directory to watcher")
3077 .log_err();
3078 if !repository_dir_abs_path.starts_with(&common_dir_abs_path) {
3079 watcher
3080 .add(&repository_dir_abs_path)
3081 .context("failed to add repository directory to watcher")
3082 .log_err();
3083 }
3084
3085 let work_directory_id = work_dir_entry.id;
3086
3087 let local_repository = LocalRepositoryEntry {
3088 work_directory_id,
3089 work_directory,
3090 work_directory_abs_path: work_directory_abs_path.as_path().into(),
3091 git_dir_scan_id: 0,
3092 dot_git_abs_path,
3093 common_dir_abs_path,
3094 repository_dir_abs_path,
3095 };
3096
3097 self.snapshot
3098 .git_repositories
3099 .insert(work_directory_id, local_repository.clone());
3100
3101 log::trace!("inserting new local git repository");
3102 Ok(local_repository)
3103 }
3104}
3105
3106async fn is_git_dir(path: &Path, fs: &dyn Fs) -> bool {
3107 if let Some(file_name) = path.file_name()
3108 && file_name == DOT_GIT
3109 {
3110 return true;
3111 }
3112
3113 // If we're in a bare repository, we are not inside a `.git` folder. In a
3114 // bare repository, the root folder contains what would normally be in the
3115 // `.git` folder.
3116 let head_metadata = fs.metadata(&path.join("HEAD")).await;
3117 if !matches!(head_metadata, Ok(Some(_))) {
3118 return false;
3119 }
3120 let config_metadata = fs.metadata(&path.join("config")).await;
3121 matches!(config_metadata, Ok(Some(_)))
3122}
3123
3124async fn build_gitignore(abs_path: &Path, fs: &dyn Fs) -> Result<Gitignore> {
3125 let contents = fs
3126 .load(abs_path)
3127 .await
3128 .with_context(|| format!("failed to load gitignore file at {}", abs_path.display()))?;
3129 let parent = abs_path.parent().unwrap_or_else(|| Path::new("/"));
3130 let mut builder = GitignoreBuilder::new(parent);
3131 for line in contents.lines() {
3132 builder.add_line(Some(abs_path.into()), line)?;
3133 }
3134 Ok(builder.build()?)
3135}
3136
3137impl Deref for Worktree {
3138 type Target = Snapshot;
3139
3140 fn deref(&self) -> &Self::Target {
3141 match self {
3142 Worktree::Local(worktree) => &worktree.snapshot,
3143 Worktree::Remote(worktree) => &worktree.snapshot,
3144 }
3145 }
3146}
3147
3148impl Deref for LocalWorktree {
3149 type Target = LocalSnapshot;
3150
3151 fn deref(&self) -> &Self::Target {
3152 &self.snapshot
3153 }
3154}
3155
3156impl Deref for RemoteWorktree {
3157 type Target = Snapshot;
3158
3159 fn deref(&self) -> &Self::Target {
3160 &self.snapshot
3161 }
3162}
3163
3164impl fmt::Debug for LocalWorktree {
3165 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3166 self.snapshot.fmt(f)
3167 }
3168}
3169
3170impl fmt::Debug for Snapshot {
3171 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3172 struct EntriesById<'a>(&'a SumTree<PathEntry>);
3173 struct EntriesByPath<'a>(&'a SumTree<Entry>);
3174
3175 impl fmt::Debug for EntriesByPath<'_> {
3176 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3177 f.debug_map()
3178 .entries(self.0.iter().map(|entry| (&entry.path, entry.id)))
3179 .finish()
3180 }
3181 }
3182
3183 impl fmt::Debug for EntriesById<'_> {
3184 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3185 f.debug_list().entries(self.0.iter()).finish()
3186 }
3187 }
3188
3189 f.debug_struct("Snapshot")
3190 .field("id", &self.id)
3191 .field("root_name", &self.root_name)
3192 .field("entries_by_path", &EntriesByPath(&self.entries_by_path))
3193 .field("entries_by_id", &EntriesById(&self.entries_by_id))
3194 .finish()
3195 }
3196}
3197
3198#[derive(Debug, Clone, PartialEq)]
3199pub struct File {
3200 pub worktree: Entity<Worktree>,
3201 pub path: Arc<RelPath>,
3202 pub disk_state: DiskState,
3203 pub entry_id: Option<ProjectEntryId>,
3204 pub is_local: bool,
3205 pub is_private: bool,
3206}
3207
3208impl language::File for File {
3209 fn as_local(&self) -> Option<&dyn language::LocalFile> {
3210 if self.is_local { Some(self) } else { None }
3211 }
3212
3213 fn disk_state(&self) -> DiskState {
3214 self.disk_state
3215 }
3216
3217 fn path(&self) -> &Arc<RelPath> {
3218 &self.path
3219 }
3220
3221 fn full_path(&self, cx: &App) -> PathBuf {
3222 self.worktree.read(cx).full_path(&self.path)
3223 }
3224
3225 /// Returns the last component of this handle's absolute path. If this handle refers to the root
3226 /// of its worktree, then this method will return the name of the worktree itself.
3227 fn file_name<'a>(&'a self, cx: &'a App) -> &'a str {
3228 self.path
3229 .file_name()
3230 .unwrap_or_else(|| self.worktree.read(cx).root_name_str())
3231 }
3232
3233 fn worktree_id(&self, cx: &App) -> WorktreeId {
3234 self.worktree.read(cx).id()
3235 }
3236
3237 fn to_proto(&self, cx: &App) -> rpc::proto::File {
3238 rpc::proto::File {
3239 worktree_id: self.worktree.read(cx).id().to_proto(),
3240 entry_id: self.entry_id.map(|id| id.to_proto()),
3241 path: self.path.as_ref().to_proto(),
3242 mtime: self.disk_state.mtime().map(|time| time.into()),
3243 is_deleted: self.disk_state.is_deleted(),
3244 is_historic: matches!(self.disk_state, DiskState::Historic { .. }),
3245 }
3246 }
3247
3248 fn is_private(&self) -> bool {
3249 self.is_private
3250 }
3251
3252 fn path_style(&self, cx: &App) -> PathStyle {
3253 self.worktree.read(cx).path_style()
3254 }
3255
3256 fn can_open(&self) -> bool {
3257 true
3258 }
3259}
3260
3261impl language::LocalFile for File {
3262 fn abs_path(&self, cx: &App) -> PathBuf {
3263 self.worktree.read(cx).absolutize(&self.path)
3264 }
3265
3266 fn load(&self, cx: &App) -> Task<Result<String>> {
3267 let worktree = self.worktree.read(cx).as_local().unwrap();
3268 let abs_path = worktree.absolutize(&self.path);
3269 let fs = worktree.fs.clone();
3270 cx.background_spawn(async move { fs.load(&abs_path).await })
3271 }
3272
3273 fn load_bytes(&self, cx: &App) -> Task<Result<Vec<u8>>> {
3274 let worktree = self.worktree.read(cx).as_local().unwrap();
3275 let abs_path = worktree.absolutize(&self.path);
3276 let fs = worktree.fs.clone();
3277 cx.background_spawn(async move { fs.load_bytes(&abs_path).await })
3278 }
3279}
3280
3281impl File {
3282 pub fn for_entry(entry: Entry, worktree: Entity<Worktree>) -> Arc<Self> {
3283 Arc::new(Self {
3284 worktree,
3285 path: entry.path.clone(),
3286 disk_state: if let Some(mtime) = entry.mtime {
3287 DiskState::Present { mtime }
3288 } else {
3289 DiskState::New
3290 },
3291 entry_id: Some(entry.id),
3292 is_local: true,
3293 is_private: entry.is_private,
3294 })
3295 }
3296
3297 pub fn from_proto(
3298 proto: rpc::proto::File,
3299 worktree: Entity<Worktree>,
3300 cx: &App,
3301 ) -> Result<Self> {
3302 let worktree_id = worktree.read(cx).as_remote().context("not remote")?.id();
3303
3304 anyhow::ensure!(
3305 worktree_id.to_proto() == proto.worktree_id,
3306 "worktree id does not match file"
3307 );
3308
3309 let disk_state = if proto.is_historic {
3310 DiskState::Historic {
3311 was_deleted: proto.is_deleted,
3312 }
3313 } else if proto.is_deleted {
3314 DiskState::Deleted
3315 } else if let Some(mtime) = proto.mtime.map(&Into::into) {
3316 DiskState::Present { mtime }
3317 } else {
3318 DiskState::New
3319 };
3320
3321 Ok(Self {
3322 worktree,
3323 path: RelPath::from_proto(&proto.path).context("invalid path in file protobuf")?,
3324 disk_state,
3325 entry_id: proto.entry_id.map(ProjectEntryId::from_proto),
3326 is_local: false,
3327 is_private: false,
3328 })
3329 }
3330
3331 pub fn from_dyn(file: Option<&Arc<dyn language::File>>) -> Option<&Self> {
3332 file.and_then(|f| {
3333 let f: &dyn language::File = f.borrow();
3334 let f: &dyn Any = f;
3335 f.downcast_ref()
3336 })
3337 }
3338
3339 pub fn worktree_id(&self, cx: &App) -> WorktreeId {
3340 self.worktree.read(cx).id()
3341 }
3342
3343 pub fn project_entry_id(&self) -> Option<ProjectEntryId> {
3344 match self.disk_state {
3345 DiskState::Deleted => None,
3346 _ => self.entry_id,
3347 }
3348 }
3349}
3350
3351#[derive(Clone, Debug, PartialEq, Eq)]
3352pub struct Entry {
3353 pub id: ProjectEntryId,
3354 pub kind: EntryKind,
3355 pub path: Arc<RelPath>,
3356 pub inode: u64,
3357 pub mtime: Option<MTime>,
3358
3359 pub canonical_path: Option<Arc<Path>>,
3360 /// Whether this entry is ignored by Git.
3361 ///
3362 /// We only scan ignored entries once the directory is expanded and
3363 /// exclude them from searches.
3364 pub is_ignored: bool,
3365
3366 /// Whether this entry is hidden or inside hidden directory.
3367 ///
3368 /// We only scan hidden entries once the directory is expanded.
3369 pub is_hidden: bool,
3370
3371 /// Whether this entry is always included in searches.
3372 ///
3373 /// This is used for entries that are always included in searches, even
3374 /// if they are ignored by git. Overridden by file_scan_exclusions.
3375 pub is_always_included: bool,
3376
3377 /// Whether this entry's canonical path is outside of the worktree.
3378 /// This means the entry is only accessible from the worktree root via a
3379 /// symlink.
3380 ///
3381 /// We only scan entries outside of the worktree once the symlinked
3382 /// directory is expanded. External entries are treated like gitignored
3383 /// entries in that they are not included in searches.
3384 pub is_external: bool,
3385
3386 /// Whether this entry is considered to be a `.env` file.
3387 pub is_private: bool,
3388 /// The entry's size on disk, in bytes.
3389 pub size: u64,
3390 pub char_bag: CharBag,
3391 pub is_fifo: bool,
3392}
3393
3394#[derive(Clone, Copy, Debug, PartialEq, Eq)]
3395pub enum EntryKind {
3396 UnloadedDir,
3397 PendingDir,
3398 Dir,
3399 File,
3400}
3401
3402#[derive(Clone, Copy, Debug, PartialEq)]
3403pub enum PathChange {
3404 /// A filesystem entry was was created.
3405 Added,
3406 /// A filesystem entry was removed.
3407 Removed,
3408 /// A filesystem entry was updated.
3409 Updated,
3410 /// A filesystem entry was either updated or added. We don't know
3411 /// whether or not it already existed, because the path had not
3412 /// been loaded before the event.
3413 AddedOrUpdated,
3414 /// A filesystem entry was found during the initial scan of the worktree.
3415 Loaded,
3416}
3417
3418#[derive(Clone, Debug, PartialEq, Eq)]
3419pub struct UpdatedGitRepository {
3420 /// ID of the repository's working directory.
3421 ///
3422 /// For a repo that's above the worktree root, this is the ID of the worktree root, and hence not unique.
3423 /// It's included here to aid the GitStore in detecting when a repository's working directory is renamed.
3424 pub work_directory_id: ProjectEntryId,
3425 pub old_work_directory_abs_path: Option<Arc<Path>>,
3426 pub new_work_directory_abs_path: Option<Arc<Path>>,
3427 /// For a normal git repository checkout, the absolute path to the .git directory.
3428 /// For a worktree, the absolute path to the worktree's subdirectory inside the .git directory.
3429 pub dot_git_abs_path: Option<Arc<Path>>,
3430 pub repository_dir_abs_path: Option<Arc<Path>>,
3431 pub common_dir_abs_path: Option<Arc<Path>>,
3432}
3433
3434pub type UpdatedEntriesSet = Arc<[(Arc<RelPath>, ProjectEntryId, PathChange)]>;
3435pub type UpdatedGitRepositoriesSet = Arc<[UpdatedGitRepository]>;
3436
3437#[derive(Clone, Debug)]
3438pub struct PathProgress<'a> {
3439 pub max_path: &'a RelPath,
3440}
3441
3442#[derive(Clone, Debug)]
3443pub struct PathSummary<S> {
3444 pub max_path: Arc<RelPath>,
3445 pub item_summary: S,
3446}
3447
3448impl<S: Summary> Summary for PathSummary<S> {
3449 type Context<'a> = S::Context<'a>;
3450
3451 fn zero(cx: Self::Context<'_>) -> Self {
3452 Self {
3453 max_path: RelPath::empty().into(),
3454 item_summary: S::zero(cx),
3455 }
3456 }
3457
3458 fn add_summary(&mut self, rhs: &Self, cx: Self::Context<'_>) {
3459 self.max_path = rhs.max_path.clone();
3460 self.item_summary.add_summary(&rhs.item_summary, cx);
3461 }
3462}
3463
3464impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathProgress<'a> {
3465 fn zero(_: <PathSummary<S> as Summary>::Context<'_>) -> Self {
3466 Self {
3467 max_path: RelPath::empty(),
3468 }
3469 }
3470
3471 fn add_summary(
3472 &mut self,
3473 summary: &'a PathSummary<S>,
3474 _: <PathSummary<S> as Summary>::Context<'_>,
3475 ) {
3476 self.max_path = summary.max_path.as_ref()
3477 }
3478}
3479
3480impl<'a> sum_tree::Dimension<'a, PathSummary<GitSummary>> for GitSummary {
3481 fn zero(_cx: ()) -> Self {
3482 Default::default()
3483 }
3484
3485 fn add_summary(&mut self, summary: &'a PathSummary<GitSummary>, _: ()) {
3486 *self += summary.item_summary
3487 }
3488}
3489
3490impl<'a>
3491 sum_tree::SeekTarget<'a, PathSummary<GitSummary>, Dimensions<TraversalProgress<'a>, GitSummary>>
3492 for PathTarget<'_>
3493{
3494 fn cmp(
3495 &self,
3496 cursor_location: &Dimensions<TraversalProgress<'a>, GitSummary>,
3497 _: (),
3498 ) -> Ordering {
3499 self.cmp_path(cursor_location.0.max_path)
3500 }
3501}
3502
3503impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathKey {
3504 fn zero(_: S::Context<'_>) -> Self {
3505 Default::default()
3506 }
3507
3508 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3509 self.0 = summary.max_path.clone();
3510 }
3511}
3512
3513impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for TraversalProgress<'a> {
3514 fn zero(_cx: S::Context<'_>) -> Self {
3515 Default::default()
3516 }
3517
3518 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3519 self.max_path = summary.max_path.as_ref();
3520 }
3521}
3522
3523impl Entry {
3524 fn new(
3525 path: Arc<RelPath>,
3526 metadata: &fs::Metadata,
3527 id: ProjectEntryId,
3528 root_char_bag: CharBag,
3529 canonical_path: Option<Arc<Path>>,
3530 ) -> Self {
3531 let char_bag = char_bag_for_path(root_char_bag, &path);
3532 Self {
3533 id,
3534 kind: if metadata.is_dir {
3535 EntryKind::PendingDir
3536 } else {
3537 EntryKind::File
3538 },
3539 path,
3540 inode: metadata.inode,
3541 mtime: Some(metadata.mtime),
3542 size: metadata.len,
3543 canonical_path,
3544 is_ignored: false,
3545 is_hidden: false,
3546 is_always_included: false,
3547 is_external: false,
3548 is_private: false,
3549 char_bag,
3550 is_fifo: metadata.is_fifo,
3551 }
3552 }
3553
3554 pub fn is_created(&self) -> bool {
3555 self.mtime.is_some()
3556 }
3557
3558 pub fn is_dir(&self) -> bool {
3559 self.kind.is_dir()
3560 }
3561
3562 pub fn is_file(&self) -> bool {
3563 self.kind.is_file()
3564 }
3565}
3566
3567impl EntryKind {
3568 pub fn is_dir(&self) -> bool {
3569 matches!(
3570 self,
3571 EntryKind::Dir | EntryKind::PendingDir | EntryKind::UnloadedDir
3572 )
3573 }
3574
3575 pub fn is_unloaded(&self) -> bool {
3576 matches!(self, EntryKind::UnloadedDir)
3577 }
3578
3579 pub fn is_file(&self) -> bool {
3580 matches!(self, EntryKind::File)
3581 }
3582}
3583
3584impl sum_tree::Item for Entry {
3585 type Summary = EntrySummary;
3586
3587 fn summary(&self, _cx: ()) -> Self::Summary {
3588 let non_ignored_count = if (self.is_ignored || self.is_external) && !self.is_always_included
3589 {
3590 0
3591 } else {
3592 1
3593 };
3594 let file_count;
3595 let non_ignored_file_count;
3596 if self.is_file() {
3597 file_count = 1;
3598 non_ignored_file_count = non_ignored_count;
3599 } else {
3600 file_count = 0;
3601 non_ignored_file_count = 0;
3602 }
3603
3604 EntrySummary {
3605 max_path: self.path.clone(),
3606 count: 1,
3607 non_ignored_count,
3608 file_count,
3609 non_ignored_file_count,
3610 }
3611 }
3612}
3613
3614impl sum_tree::KeyedItem for Entry {
3615 type Key = PathKey;
3616
3617 fn key(&self) -> Self::Key {
3618 PathKey(self.path.clone())
3619 }
3620}
3621
3622#[derive(Clone, Debug)]
3623pub struct EntrySummary {
3624 max_path: Arc<RelPath>,
3625 count: usize,
3626 non_ignored_count: usize,
3627 file_count: usize,
3628 non_ignored_file_count: usize,
3629}
3630
3631impl Default for EntrySummary {
3632 fn default() -> Self {
3633 Self {
3634 max_path: Arc::from(RelPath::empty()),
3635 count: 0,
3636 non_ignored_count: 0,
3637 file_count: 0,
3638 non_ignored_file_count: 0,
3639 }
3640 }
3641}
3642
3643impl sum_tree::ContextLessSummary for EntrySummary {
3644 fn zero() -> Self {
3645 Default::default()
3646 }
3647
3648 fn add_summary(&mut self, rhs: &Self) {
3649 self.max_path = rhs.max_path.clone();
3650 self.count += rhs.count;
3651 self.non_ignored_count += rhs.non_ignored_count;
3652 self.file_count += rhs.file_count;
3653 self.non_ignored_file_count += rhs.non_ignored_file_count;
3654 }
3655}
3656
3657#[derive(Clone, Debug)]
3658struct PathEntry {
3659 id: ProjectEntryId,
3660 path: Arc<RelPath>,
3661 is_ignored: bool,
3662 scan_id: usize,
3663}
3664
3665impl sum_tree::Item for PathEntry {
3666 type Summary = PathEntrySummary;
3667
3668 fn summary(&self, _cx: ()) -> Self::Summary {
3669 PathEntrySummary { max_id: self.id }
3670 }
3671}
3672
3673impl sum_tree::KeyedItem for PathEntry {
3674 type Key = ProjectEntryId;
3675
3676 fn key(&self) -> Self::Key {
3677 self.id
3678 }
3679}
3680
3681#[derive(Clone, Debug, Default)]
3682struct PathEntrySummary {
3683 max_id: ProjectEntryId,
3684}
3685
3686impl sum_tree::ContextLessSummary for PathEntrySummary {
3687 fn zero() -> Self {
3688 Default::default()
3689 }
3690
3691 fn add_summary(&mut self, summary: &Self) {
3692 self.max_id = summary.max_id;
3693 }
3694}
3695
3696impl<'a> sum_tree::Dimension<'a, PathEntrySummary> for ProjectEntryId {
3697 fn zero(_cx: ()) -> Self {
3698 Default::default()
3699 }
3700
3701 fn add_summary(&mut self, summary: &'a PathEntrySummary, _: ()) {
3702 *self = summary.max_id;
3703 }
3704}
3705
3706#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
3707pub struct PathKey(pub Arc<RelPath>);
3708
3709impl Default for PathKey {
3710 fn default() -> Self {
3711 Self(RelPath::empty().into())
3712 }
3713}
3714
3715impl<'a> sum_tree::Dimension<'a, EntrySummary> for PathKey {
3716 fn zero(_cx: ()) -> Self {
3717 Default::default()
3718 }
3719
3720 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
3721 self.0 = summary.max_path.clone();
3722 }
3723}
3724
3725struct BackgroundScanner {
3726 state: async_lock::Mutex<BackgroundScannerState>,
3727 fs: Arc<dyn Fs>,
3728 fs_case_sensitive: bool,
3729 status_updates_tx: UnboundedSender<ScanState>,
3730 executor: BackgroundExecutor,
3731 scan_requests_rx: channel::Receiver<ScanRequest>,
3732 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
3733 next_entry_id: Arc<AtomicUsize>,
3734 phase: BackgroundScannerPhase,
3735 watcher: Arc<dyn Watcher>,
3736 settings: WorktreeSettings,
3737 share_private_files: bool,
3738}
3739
3740#[derive(Copy, Clone, PartialEq)]
3741enum BackgroundScannerPhase {
3742 InitialScan,
3743 EventsReceivedDuringInitialScan,
3744 Events,
3745}
3746
3747impl BackgroundScanner {
3748 async fn run(&mut self, mut fs_events_rx: Pin<Box<dyn Send + Stream<Item = Vec<PathEvent>>>>) {
3749 let root_abs_path;
3750 let scanning_enabled;
3751 {
3752 let state = self.state.lock().await;
3753 root_abs_path = state.snapshot.abs_path.clone();
3754 scanning_enabled = state.scanning_enabled;
3755 }
3756
3757 // If the worktree root does not contain a git repository, then find
3758 // the git repository in an ancestor directory. Find any gitignore files
3759 // in ancestor directories.
3760 let repo = if scanning_enabled {
3761 let (ignores, exclude, repo) =
3762 discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await;
3763 self.state
3764 .lock()
3765 .await
3766 .snapshot
3767 .ignores_by_parent_abs_path
3768 .extend(ignores);
3769 if let Some(exclude) = exclude {
3770 self.state
3771 .lock()
3772 .await
3773 .snapshot
3774 .repo_exclude_by_work_dir_abs_path
3775 .insert(root_abs_path.as_path().into(), (exclude, false));
3776 }
3777
3778 repo
3779 } else {
3780 None
3781 };
3782
3783 let containing_git_repository = if let Some((ancestor_dot_git, work_directory)) = repo
3784 && scanning_enabled
3785 {
3786 maybe!(async {
3787 self.state
3788 .lock()
3789 .await
3790 .insert_git_repository_for_path(
3791 work_directory,
3792 ancestor_dot_git.clone().into(),
3793 self.fs.as_ref(),
3794 self.watcher.as_ref(),
3795 )
3796 .await
3797 .log_err()?;
3798 Some(ancestor_dot_git)
3799 })
3800 .await
3801 } else {
3802 None
3803 };
3804
3805 log::trace!("containing git repository: {containing_git_repository:?}");
3806
3807 let mut global_gitignore_events = if let Some(global_gitignore_path) =
3808 &paths::global_gitignore_path()
3809 && scanning_enabled
3810 {
3811 let is_file = self.fs.is_file(&global_gitignore_path).await;
3812 self.state.lock().await.snapshot.global_gitignore = if is_file {
3813 build_gitignore(global_gitignore_path, self.fs.as_ref())
3814 .await
3815 .ok()
3816 .map(Arc::new)
3817 } else {
3818 None
3819 };
3820 if is_file
3821 || matches!(global_gitignore_path.parent(), Some(path) if self.fs.is_dir(path).await)
3822 {
3823 self.fs
3824 .watch(global_gitignore_path, FS_WATCH_LATENCY)
3825 .await
3826 .0
3827 } else {
3828 Box::pin(futures::stream::pending())
3829 }
3830 } else {
3831 self.state.lock().await.snapshot.global_gitignore = None;
3832 Box::pin(futures::stream::pending())
3833 };
3834
3835 let (scan_job_tx, scan_job_rx) = channel::unbounded();
3836 {
3837 let mut state = self.state.lock().await;
3838 state.snapshot.scan_id += 1;
3839 if let Some(mut root_entry) = state.snapshot.root_entry().cloned() {
3840 let ignore_stack = state
3841 .snapshot
3842 .ignore_stack_for_abs_path(root_abs_path.as_path(), true, self.fs.as_ref())
3843 .await;
3844 if ignore_stack.is_abs_path_ignored(root_abs_path.as_path(), true) {
3845 root_entry.is_ignored = true;
3846 let mut root_entry = root_entry.clone();
3847 state.reuse_entry_id(&mut root_entry);
3848 state
3849 .insert_entry(root_entry, self.fs.as_ref(), self.watcher.as_ref())
3850 .await;
3851 }
3852 if root_entry.is_dir() && state.scanning_enabled {
3853 state
3854 .enqueue_scan_dir(
3855 root_abs_path.as_path().into(),
3856 &root_entry,
3857 &scan_job_tx,
3858 self.fs.as_ref(),
3859 )
3860 .await;
3861 }
3862 }
3863 };
3864
3865 // Perform an initial scan of the directory.
3866 drop(scan_job_tx);
3867 self.scan_dirs(true, scan_job_rx).await;
3868 {
3869 let mut state = self.state.lock().await;
3870 state.snapshot.completed_scan_id = state.snapshot.scan_id;
3871 }
3872
3873 self.send_status_update(false, SmallVec::new()).await;
3874
3875 // Process any any FS events that occurred while performing the initial scan.
3876 // For these events, update events cannot be as precise, because we didn't
3877 // have the previous state loaded yet.
3878 self.phase = BackgroundScannerPhase::EventsReceivedDuringInitialScan;
3879 if let Poll::Ready(Some(mut paths)) = futures::poll!(fs_events_rx.next()) {
3880 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
3881 paths.extend(more_paths);
3882 }
3883 self.process_events(
3884 paths
3885 .into_iter()
3886 .filter(|e| e.kind.is_some())
3887 .map(Into::into)
3888 .collect(),
3889 )
3890 .await;
3891 }
3892 if let Some(abs_path) = containing_git_repository {
3893 self.process_events(vec![abs_path]).await;
3894 }
3895
3896 // Continue processing events until the worktree is dropped.
3897 self.phase = BackgroundScannerPhase::Events;
3898
3899 loop {
3900 select_biased! {
3901 // Process any path refresh requests from the worktree. Prioritize
3902 // these before handling changes reported by the filesystem.
3903 request = self.next_scan_request().fuse() => {
3904 let Ok(request) = request else { break };
3905 if !self.process_scan_request(request, false).await {
3906 return;
3907 }
3908 }
3909
3910 path_prefix_request = self.path_prefixes_to_scan_rx.recv().fuse() => {
3911 let Ok(request) = path_prefix_request else { break };
3912 log::trace!("adding path prefix {:?}", request.path);
3913
3914 let did_scan = self.forcibly_load_paths(std::slice::from_ref(&request.path)).await;
3915 if did_scan {
3916 let abs_path =
3917 {
3918 let mut state = self.state.lock().await;
3919 state.path_prefixes_to_scan.insert(request.path.clone());
3920 state.snapshot.absolutize(&request.path)
3921 };
3922
3923 if let Some(abs_path) = self.fs.canonicalize(&abs_path).await.log_err() {
3924 self.process_events(vec![abs_path]).await;
3925 }
3926 }
3927 self.send_status_update(false, request.done).await;
3928 }
3929
3930 paths = fs_events_rx.next().fuse() => {
3931 let Some(mut paths) = paths else { break };
3932 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
3933 paths.extend(more_paths);
3934 }
3935 self.process_events(paths.into_iter().filter(|e| e.kind.is_some()).map(Into::into).collect()).await;
3936 }
3937
3938 paths = global_gitignore_events.next().fuse() => {
3939 match paths.as_deref() {
3940 Some([event, ..]) => {
3941 self.update_global_gitignore(&event.path).await;
3942 }
3943 _ => (),
3944 }
3945 }
3946 }
3947 }
3948 }
3949
3950 async fn process_scan_request(&self, mut request: ScanRequest, scanning: bool) -> bool {
3951 log::debug!("rescanning paths {:?}", request.relative_paths);
3952
3953 request.relative_paths.sort_unstable();
3954 self.forcibly_load_paths(&request.relative_paths).await;
3955
3956 let root_path = self.state.lock().await.snapshot.abs_path.clone();
3957 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
3958 let root_canonical_path = match &root_canonical_path {
3959 Ok(path) => SanitizedPath::new(path),
3960 Err(err) => {
3961 log::error!("failed to canonicalize root path {root_path:?}: {err:#}");
3962 return true;
3963 }
3964 };
3965 let abs_paths = request
3966 .relative_paths
3967 .iter()
3968 .map(|path| {
3969 if path.file_name().is_some() {
3970 root_canonical_path.as_path().join(path.as_std_path())
3971 } else {
3972 root_canonical_path.as_path().to_path_buf()
3973 }
3974 })
3975 .collect::<Vec<_>>();
3976
3977 {
3978 let mut state = self.state.lock().await;
3979 let is_idle = state.snapshot.completed_scan_id == state.snapshot.scan_id;
3980 state.snapshot.scan_id += 1;
3981 if is_idle {
3982 state.snapshot.completed_scan_id = state.snapshot.scan_id;
3983 }
3984 }
3985
3986 self.reload_entries_for_paths(
3987 &root_path,
3988 &root_canonical_path,
3989 &request.relative_paths,
3990 abs_paths,
3991 None,
3992 )
3993 .await;
3994
3995 self.send_status_update(scanning, request.done).await
3996 }
3997
3998 async fn process_events(&self, mut abs_paths: Vec<PathBuf>) {
3999 log::trace!("process events: {abs_paths:?}");
4000 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4001 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4002 let root_canonical_path = match &root_canonical_path {
4003 Ok(path) => SanitizedPath::new(path),
4004 Err(err) => {
4005 let new_path = self
4006 .state
4007 .lock()
4008 .await
4009 .snapshot
4010 .root_file_handle
4011 .clone()
4012 .and_then(|handle| match handle.current_path(&self.fs) {
4013 Ok(new_path) => Some(new_path),
4014 Err(e) => {
4015 log::error!("Failed to refresh worktree root path: {e:#}");
4016 None
4017 }
4018 })
4019 .map(|path| SanitizedPath::new_arc(&path))
4020 .filter(|new_path| *new_path != root_path);
4021
4022 if let Some(new_path) = new_path {
4023 log::info!(
4024 "root renamed from {:?} to {:?}",
4025 root_path.as_path(),
4026 new_path.as_path(),
4027 );
4028 self.status_updates_tx
4029 .unbounded_send(ScanState::RootUpdated { new_path })
4030 .ok();
4031 } else {
4032 log::error!("root path could not be canonicalized: {err:#}");
4033 }
4034 return;
4035 }
4036 };
4037
4038 // Certain directories may have FS changes, but do not lead to git data changes that Zed cares about.
4039 // Ignore these, to avoid Zed unnecessarily rescanning git metadata.
4040 let skipped_files_in_dot_git = [COMMIT_MESSAGE, INDEX_LOCK];
4041 let skipped_dirs_in_dot_git = [FSMONITOR_DAEMON, LFS_DIR];
4042
4043 let mut relative_paths = Vec::with_capacity(abs_paths.len());
4044 let mut dot_git_abs_paths = Vec::new();
4045 let mut work_dirs_needing_exclude_update = Vec::new();
4046 abs_paths.sort_unstable();
4047 abs_paths.dedup_by(|a, b| a.starts_with(b));
4048 {
4049 let snapshot = &self.state.lock().await.snapshot;
4050
4051 let mut ranges_to_drop = SmallVec::<[Range<usize>; 4]>::new();
4052
4053 fn skip_ix(ranges: &mut SmallVec<[Range<usize>; 4]>, ix: usize) {
4054 if let Some(last_range) = ranges.last_mut()
4055 && last_range.end == ix
4056 {
4057 last_range.end += 1;
4058 } else {
4059 ranges.push(ix..ix + 1);
4060 }
4061 }
4062
4063 for (ix, abs_path) in abs_paths.iter().enumerate() {
4064 let abs_path = &SanitizedPath::new(&abs_path);
4065
4066 let mut is_git_related = false;
4067 let mut dot_git_paths = None;
4068
4069 for ancestor in abs_path.as_path().ancestors() {
4070 if is_git_dir(ancestor, self.fs.as_ref()).await {
4071 let path_in_git_dir = abs_path
4072 .as_path()
4073 .strip_prefix(ancestor)
4074 .expect("stripping off the ancestor");
4075 dot_git_paths = Some((ancestor.to_owned(), path_in_git_dir.to_owned()));
4076 break;
4077 }
4078 }
4079
4080 if let Some((dot_git_abs_path, path_in_git_dir)) = dot_git_paths {
4081 if skipped_files_in_dot_git
4082 .iter()
4083 .any(|skipped| OsStr::new(skipped) == path_in_git_dir.as_path().as_os_str())
4084 || skipped_dirs_in_dot_git.iter().any(|skipped_git_subdir| {
4085 path_in_git_dir.starts_with(skipped_git_subdir)
4086 })
4087 {
4088 log::debug!(
4089 "ignoring event {abs_path:?} as it's in the .git directory among skipped files or directories"
4090 );
4091 skip_ix(&mut ranges_to_drop, ix);
4092 continue;
4093 }
4094
4095 is_git_related = true;
4096 if !dot_git_abs_paths.contains(&dot_git_abs_path) {
4097 dot_git_abs_paths.push(dot_git_abs_path);
4098 }
4099 }
4100
4101 let relative_path = if let Ok(path) = abs_path.strip_prefix(&root_canonical_path)
4102 && let Ok(path) = RelPath::new(path, PathStyle::local())
4103 {
4104 path
4105 } else {
4106 if is_git_related {
4107 log::debug!(
4108 "ignoring event {abs_path:?}, since it's in git dir outside of root path {root_canonical_path:?}",
4109 );
4110 } else {
4111 log::error!(
4112 "ignoring event {abs_path:?} outside of root path {root_canonical_path:?}",
4113 );
4114 }
4115 skip_ix(&mut ranges_to_drop, ix);
4116 continue;
4117 };
4118
4119 let absolute_path = abs_path.to_path_buf();
4120 if absolute_path.ends_with(Path::new(DOT_GIT).join(REPO_EXCLUDE)) {
4121 if let Some(repository) = snapshot
4122 .git_repositories
4123 .values()
4124 .find(|repo| repo.common_dir_abs_path.join(REPO_EXCLUDE) == absolute_path)
4125 {
4126 work_dirs_needing_exclude_update
4127 .push(repository.work_directory_abs_path.clone());
4128 }
4129 }
4130
4131 if abs_path.file_name() == Some(OsStr::new(GITIGNORE)) {
4132 for (_, repo) in snapshot
4133 .git_repositories
4134 .iter()
4135 .filter(|(_, repo)| repo.directory_contains(&relative_path))
4136 {
4137 if !dot_git_abs_paths.iter().any(|dot_git_abs_path| {
4138 dot_git_abs_path == repo.common_dir_abs_path.as_ref()
4139 }) {
4140 dot_git_abs_paths.push(repo.common_dir_abs_path.to_path_buf());
4141 }
4142 }
4143 }
4144
4145 let parent_dir_is_loaded = relative_path.parent().is_none_or(|parent| {
4146 snapshot
4147 .entry_for_path(parent)
4148 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4149 });
4150 if !parent_dir_is_loaded {
4151 log::debug!("ignoring event {relative_path:?} within unloaded directory");
4152 skip_ix(&mut ranges_to_drop, ix);
4153 continue;
4154 }
4155
4156 if self.settings.is_path_excluded(&relative_path) {
4157 if !is_git_related {
4158 log::debug!("ignoring FS event for excluded path {relative_path:?}");
4159 }
4160 skip_ix(&mut ranges_to_drop, ix);
4161 continue;
4162 }
4163
4164 relative_paths.push(relative_path.into_arc());
4165 }
4166
4167 for range_to_drop in ranges_to_drop.into_iter().rev() {
4168 abs_paths.drain(range_to_drop);
4169 }
4170 }
4171
4172 if relative_paths.is_empty() && dot_git_abs_paths.is_empty() {
4173 return;
4174 }
4175
4176 if !work_dirs_needing_exclude_update.is_empty() {
4177 let mut state = self.state.lock().await;
4178 for work_dir_abs_path in work_dirs_needing_exclude_update {
4179 if let Some((_, needs_update)) = state
4180 .snapshot
4181 .repo_exclude_by_work_dir_abs_path
4182 .get_mut(&work_dir_abs_path)
4183 {
4184 *needs_update = true;
4185 }
4186 }
4187 }
4188
4189 self.state.lock().await.snapshot.scan_id += 1;
4190
4191 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4192 log::debug!("received fs events {:?}", relative_paths);
4193 self.reload_entries_for_paths(
4194 &root_path,
4195 &root_canonical_path,
4196 &relative_paths,
4197 abs_paths,
4198 Some(scan_job_tx.clone()),
4199 )
4200 .await;
4201
4202 let affected_repo_roots = if !dot_git_abs_paths.is_empty() {
4203 self.update_git_repositories(dot_git_abs_paths).await
4204 } else {
4205 Vec::new()
4206 };
4207
4208 {
4209 let mut ignores_to_update = self.ignores_needing_update().await;
4210 ignores_to_update.extend(affected_repo_roots);
4211 let ignores_to_update = self.order_ignores(ignores_to_update).await;
4212 let snapshot = self.state.lock().await.snapshot.clone();
4213 self.update_ignore_statuses_for_paths(scan_job_tx, snapshot, ignores_to_update)
4214 .await;
4215 self.scan_dirs(false, scan_job_rx).await;
4216 }
4217
4218 {
4219 let mut state = self.state.lock().await;
4220 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4221 for (_, entry) in mem::take(&mut state.removed_entries) {
4222 state.scanned_dirs.remove(&entry.id);
4223 }
4224 }
4225 self.send_status_update(false, SmallVec::new()).await;
4226 }
4227
4228 async fn update_global_gitignore(&self, abs_path: &Path) {
4229 let ignore = build_gitignore(abs_path, self.fs.as_ref())
4230 .await
4231 .log_err()
4232 .map(Arc::new);
4233 let (prev_snapshot, ignore_stack, abs_path) = {
4234 let mut state = self.state.lock().await;
4235 state.snapshot.global_gitignore = ignore;
4236 let abs_path = state.snapshot.abs_path().clone();
4237 let ignore_stack = state
4238 .snapshot
4239 .ignore_stack_for_abs_path(&abs_path, true, self.fs.as_ref())
4240 .await;
4241 (state.snapshot.clone(), ignore_stack, abs_path)
4242 };
4243 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4244 self.update_ignore_statuses_for_paths(
4245 scan_job_tx,
4246 prev_snapshot,
4247 vec![(abs_path, ignore_stack)],
4248 )
4249 .await;
4250 self.scan_dirs(false, scan_job_rx).await;
4251 self.send_status_update(false, SmallVec::new()).await;
4252 }
4253
4254 async fn forcibly_load_paths(&self, paths: &[Arc<RelPath>]) -> bool {
4255 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4256 {
4257 let mut state = self.state.lock().await;
4258 let root_path = state.snapshot.abs_path.clone();
4259 for path in paths {
4260 for ancestor in path.ancestors() {
4261 if let Some(entry) = state.snapshot.entry_for_path(ancestor)
4262 && entry.kind == EntryKind::UnloadedDir
4263 {
4264 let abs_path = root_path.join(ancestor.as_std_path());
4265 state
4266 .enqueue_scan_dir(
4267 abs_path.into(),
4268 entry,
4269 &scan_job_tx,
4270 self.fs.as_ref(),
4271 )
4272 .await;
4273 state.paths_to_scan.insert(path.clone());
4274 break;
4275 }
4276 }
4277 }
4278 drop(scan_job_tx);
4279 }
4280 while let Ok(job) = scan_job_rx.recv().await {
4281 self.scan_dir(&job).await.log_err();
4282 }
4283
4284 !mem::take(&mut self.state.lock().await.paths_to_scan).is_empty()
4285 }
4286
4287 async fn scan_dirs(
4288 &self,
4289 enable_progress_updates: bool,
4290 scan_jobs_rx: channel::Receiver<ScanJob>,
4291 ) {
4292 if self
4293 .status_updates_tx
4294 .unbounded_send(ScanState::Started)
4295 .is_err()
4296 {
4297 return;
4298 }
4299
4300 let progress_update_count = AtomicUsize::new(0);
4301 self.executor
4302 .scoped_priority(Priority::Low, |scope| {
4303 for _ in 0..self.executor.num_cpus() {
4304 scope.spawn(async {
4305 let mut last_progress_update_count = 0;
4306 let progress_update_timer = self.progress_timer(enable_progress_updates).fuse();
4307 futures::pin_mut!(progress_update_timer);
4308
4309 loop {
4310 select_biased! {
4311 // Process any path refresh requests before moving on to process
4312 // the scan queue, so that user operations are prioritized.
4313 request = self.next_scan_request().fuse() => {
4314 let Ok(request) = request else { break };
4315 if !self.process_scan_request(request, true).await {
4316 return;
4317 }
4318 }
4319
4320 // Send periodic progress updates to the worktree. Use an atomic counter
4321 // to ensure that only one of the workers sends a progress update after
4322 // the update interval elapses.
4323 _ = progress_update_timer => {
4324 match progress_update_count.compare_exchange(
4325 last_progress_update_count,
4326 last_progress_update_count + 1,
4327 SeqCst,
4328 SeqCst
4329 ) {
4330 Ok(_) => {
4331 last_progress_update_count += 1;
4332 self.send_status_update(true, SmallVec::new()).await;
4333 }
4334 Err(count) => {
4335 last_progress_update_count = count;
4336 }
4337 }
4338 progress_update_timer.set(self.progress_timer(enable_progress_updates).fuse());
4339 }
4340
4341 // Recursively load directories from the file system.
4342 job = scan_jobs_rx.recv().fuse() => {
4343 let Ok(job) = job else { break };
4344 if let Err(err) = self.scan_dir(&job).await
4345 && job.path.is_empty() {
4346 log::error!("error scanning directory {:?}: {}", job.abs_path, err);
4347 }
4348 }
4349 }
4350 }
4351 });
4352 }
4353 })
4354 .await;
4355 }
4356
4357 async fn send_status_update(
4358 &self,
4359 scanning: bool,
4360 barrier: SmallVec<[barrier::Sender; 1]>,
4361 ) -> bool {
4362 let mut state = self.state.lock().await;
4363 if state.changed_paths.is_empty() && scanning {
4364 return true;
4365 }
4366
4367 let new_snapshot = state.snapshot.clone();
4368 let old_snapshot = mem::replace(&mut state.prev_snapshot, new_snapshot.snapshot.clone());
4369 let changes = build_diff(
4370 self.phase,
4371 &old_snapshot,
4372 &new_snapshot,
4373 &state.changed_paths,
4374 );
4375 state.changed_paths.clear();
4376
4377 self.status_updates_tx
4378 .unbounded_send(ScanState::Updated {
4379 snapshot: new_snapshot,
4380 changes,
4381 scanning,
4382 barrier,
4383 })
4384 .is_ok()
4385 }
4386
4387 async fn scan_dir(&self, job: &ScanJob) -> Result<()> {
4388 let root_abs_path;
4389 let root_char_bag;
4390 {
4391 let snapshot = &self.state.lock().await.snapshot;
4392 if self.settings.is_path_excluded(&job.path) {
4393 log::error!("skipping excluded directory {:?}", job.path);
4394 return Ok(());
4395 }
4396 log::trace!("scanning directory {:?}", job.path);
4397 root_abs_path = snapshot.abs_path().clone();
4398 root_char_bag = snapshot.root_char_bag;
4399 }
4400
4401 let next_entry_id = self.next_entry_id.clone();
4402 let mut ignore_stack = job.ignore_stack.clone();
4403 let mut new_ignore = None;
4404 let mut root_canonical_path = None;
4405 let mut new_entries: Vec<Entry> = Vec::new();
4406 let mut new_jobs: Vec<Option<ScanJob>> = Vec::new();
4407 let mut child_paths = self
4408 .fs
4409 .read_dir(&job.abs_path)
4410 .await?
4411 .filter_map(|entry| async {
4412 match entry {
4413 Ok(entry) => Some(entry),
4414 Err(error) => {
4415 log::error!("error processing entry {:?}", error);
4416 None
4417 }
4418 }
4419 })
4420 .collect::<Vec<_>>()
4421 .await;
4422
4423 // Ensure that .git and .gitignore are processed first.
4424 swap_to_front(&mut child_paths, GITIGNORE);
4425 swap_to_front(&mut child_paths, DOT_GIT);
4426
4427 if let Some(path) = child_paths.first()
4428 && path.ends_with(DOT_GIT)
4429 {
4430 ignore_stack.repo_root = Some(job.abs_path.clone());
4431 }
4432
4433 for child_abs_path in child_paths {
4434 let child_abs_path: Arc<Path> = child_abs_path.into();
4435 let child_name = child_abs_path.file_name().unwrap();
4436 let Some(child_path) = child_name
4437 .to_str()
4438 .and_then(|name| Some(job.path.join(RelPath::unix(name).ok()?)))
4439 else {
4440 continue;
4441 };
4442
4443 if child_name == DOT_GIT {
4444 let mut state = self.state.lock().await;
4445 state
4446 .insert_git_repository(
4447 child_path.clone(),
4448 self.fs.as_ref(),
4449 self.watcher.as_ref(),
4450 )
4451 .await;
4452 } else if child_name == GITIGNORE {
4453 match build_gitignore(&child_abs_path, self.fs.as_ref()).await {
4454 Ok(ignore) => {
4455 let ignore = Arc::new(ignore);
4456 ignore_stack = ignore_stack
4457 .append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4458 new_ignore = Some(ignore);
4459 }
4460 Err(error) => {
4461 log::error!(
4462 "error loading .gitignore file {:?} - {:?}",
4463 child_name,
4464 error
4465 );
4466 }
4467 }
4468 }
4469
4470 if self.settings.is_path_excluded(&child_path) {
4471 log::debug!("skipping excluded child entry {child_path:?}");
4472 self.state.lock().await.remove_path(&child_path);
4473 continue;
4474 }
4475
4476 let child_metadata = match self.fs.metadata(&child_abs_path).await {
4477 Ok(Some(metadata)) => metadata,
4478 Ok(None) => continue,
4479 Err(err) => {
4480 log::error!("error processing {child_abs_path:?}: {err:#}");
4481 continue;
4482 }
4483 };
4484
4485 let mut child_entry = Entry::new(
4486 child_path.clone(),
4487 &child_metadata,
4488 ProjectEntryId::new(&next_entry_id),
4489 root_char_bag,
4490 None,
4491 );
4492
4493 if job.is_external {
4494 child_entry.is_external = true;
4495 } else if child_metadata.is_symlink {
4496 let canonical_path = match self.fs.canonicalize(&child_abs_path).await {
4497 Ok(path) => path,
4498 Err(err) => {
4499 log::error!("error reading target of symlink {child_abs_path:?}: {err:#}",);
4500 continue;
4501 }
4502 };
4503
4504 // lazily canonicalize the root path in order to determine if
4505 // symlinks point outside of the worktree.
4506 let root_canonical_path = match &root_canonical_path {
4507 Some(path) => path,
4508 None => match self.fs.canonicalize(&root_abs_path).await {
4509 Ok(path) => root_canonical_path.insert(path),
4510 Err(err) => {
4511 log::error!("error canonicalizing root {:?}: {:?}", root_abs_path, err);
4512 continue;
4513 }
4514 },
4515 };
4516
4517 if !canonical_path.starts_with(root_canonical_path) {
4518 child_entry.is_external = true;
4519 }
4520
4521 child_entry.canonical_path = Some(canonical_path.into());
4522 }
4523
4524 if child_entry.is_dir() {
4525 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, true);
4526 child_entry.is_always_included =
4527 self.settings.is_path_always_included(&child_path, true);
4528
4529 // Avoid recursing until crash in the case of a recursive symlink
4530 if job.ancestor_inodes.contains(&child_entry.inode) {
4531 new_jobs.push(None);
4532 } else {
4533 let mut ancestor_inodes = job.ancestor_inodes.clone();
4534 ancestor_inodes.insert(child_entry.inode);
4535
4536 new_jobs.push(Some(ScanJob {
4537 abs_path: child_abs_path.clone(),
4538 path: child_path,
4539 is_external: child_entry.is_external,
4540 ignore_stack: if child_entry.is_ignored {
4541 IgnoreStack::all()
4542 } else {
4543 ignore_stack.clone()
4544 },
4545 ancestor_inodes,
4546 scan_queue: job.scan_queue.clone(),
4547 }));
4548 }
4549 } else {
4550 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, false);
4551 child_entry.is_always_included =
4552 self.settings.is_path_always_included(&child_path, false);
4553 }
4554
4555 {
4556 let relative_path = job
4557 .path
4558 .join(RelPath::unix(child_name.to_str().unwrap()).unwrap());
4559 if self.is_path_private(&relative_path) {
4560 log::debug!("detected private file: {relative_path:?}");
4561 child_entry.is_private = true;
4562 }
4563 if self.settings.is_path_hidden(&relative_path) {
4564 log::debug!("detected hidden file: {relative_path:?}");
4565 child_entry.is_hidden = true;
4566 }
4567 }
4568
4569 new_entries.push(child_entry);
4570 }
4571
4572 let mut state = self.state.lock().await;
4573
4574 // Identify any subdirectories that should not be scanned.
4575 let mut job_ix = 0;
4576 for entry in &mut new_entries {
4577 state.reuse_entry_id(entry);
4578 if entry.is_dir() {
4579 if state.should_scan_directory(entry) {
4580 job_ix += 1;
4581 } else {
4582 log::debug!("defer scanning directory {:?}", entry.path);
4583 entry.kind = EntryKind::UnloadedDir;
4584 new_jobs.remove(job_ix);
4585 }
4586 }
4587 if entry.is_always_included {
4588 state
4589 .snapshot
4590 .always_included_entries
4591 .push(entry.path.clone());
4592 }
4593 }
4594
4595 state.populate_dir(job.path.clone(), new_entries, new_ignore);
4596 self.watcher.add(job.abs_path.as_ref()).log_err();
4597
4598 for new_job in new_jobs.into_iter().flatten() {
4599 job.scan_queue
4600 .try_send(new_job)
4601 .expect("channel is unbounded");
4602 }
4603
4604 Ok(())
4605 }
4606
4607 /// All list arguments should be sorted before calling this function
4608 async fn reload_entries_for_paths(
4609 &self,
4610 root_abs_path: &SanitizedPath,
4611 root_canonical_path: &SanitizedPath,
4612 relative_paths: &[Arc<RelPath>],
4613 abs_paths: Vec<PathBuf>,
4614 scan_queue_tx: Option<Sender<ScanJob>>,
4615 ) {
4616 // grab metadata for all requested paths
4617 let metadata = futures::future::join_all(
4618 abs_paths
4619 .iter()
4620 .map(|abs_path| async move {
4621 let metadata = self.fs.metadata(abs_path).await?;
4622 if let Some(metadata) = metadata {
4623 let canonical_path = self.fs.canonicalize(abs_path).await?;
4624
4625 // If we're on a case-insensitive filesystem (default on macOS), we want
4626 // to only ignore metadata for non-symlink files if their absolute-path matches
4627 // the canonical-path.
4628 // Because if not, this might be a case-only-renaming (`mv test.txt TEST.TXT`)
4629 // and we want to ignore the metadata for the old path (`test.txt`) so it's
4630 // treated as removed.
4631 if !self.fs_case_sensitive && !metadata.is_symlink {
4632 let canonical_file_name = canonical_path.file_name();
4633 let file_name = abs_path.file_name();
4634 if canonical_file_name != file_name {
4635 return Ok(None);
4636 }
4637 }
4638
4639 anyhow::Ok(Some((metadata, SanitizedPath::new_arc(&canonical_path))))
4640 } else {
4641 Ok(None)
4642 }
4643 })
4644 .collect::<Vec<_>>(),
4645 )
4646 .await;
4647
4648 let mut new_ancestor_repo = if relative_paths.iter().any(|path| path.is_empty()) {
4649 Some(discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await)
4650 } else {
4651 None
4652 };
4653
4654 let mut state = self.state.lock().await;
4655 let doing_recursive_update = scan_queue_tx.is_some();
4656
4657 // Remove any entries for paths that no longer exist or are being recursively
4658 // refreshed. Do this before adding any new entries, so that renames can be
4659 // detected regardless of the order of the paths.
4660 for (path, metadata) in relative_paths.iter().zip(metadata.iter()) {
4661 if matches!(metadata, Ok(None)) || doing_recursive_update {
4662 state.remove_path(path);
4663 }
4664 }
4665
4666 for (path, metadata) in relative_paths.iter().zip(metadata.into_iter()) {
4667 let abs_path: Arc<Path> = root_abs_path.join(path.as_std_path()).into();
4668 match metadata {
4669 Ok(Some((metadata, canonical_path))) => {
4670 let ignore_stack = state
4671 .snapshot
4672 .ignore_stack_for_abs_path(&abs_path, metadata.is_dir, self.fs.as_ref())
4673 .await;
4674 let is_external = !canonical_path.starts_with(&root_canonical_path);
4675 let entry_id = state.entry_id_for(self.next_entry_id.as_ref(), path, &metadata);
4676 let mut fs_entry = Entry::new(
4677 path.clone(),
4678 &metadata,
4679 entry_id,
4680 state.snapshot.root_char_bag,
4681 if metadata.is_symlink {
4682 Some(canonical_path.as_path().to_path_buf().into())
4683 } else {
4684 None
4685 },
4686 );
4687
4688 let is_dir = fs_entry.is_dir();
4689 fs_entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, is_dir);
4690 fs_entry.is_external = is_external;
4691 fs_entry.is_private = self.is_path_private(path);
4692 fs_entry.is_always_included =
4693 self.settings.is_path_always_included(path, is_dir);
4694 fs_entry.is_hidden = self.settings.is_path_hidden(path);
4695
4696 if let (Some(scan_queue_tx), true) = (&scan_queue_tx, is_dir) {
4697 if state.should_scan_directory(&fs_entry)
4698 || (fs_entry.path.is_empty()
4699 && abs_path.file_name() == Some(OsStr::new(DOT_GIT)))
4700 {
4701 state
4702 .enqueue_scan_dir(
4703 abs_path,
4704 &fs_entry,
4705 scan_queue_tx,
4706 self.fs.as_ref(),
4707 )
4708 .await;
4709 } else {
4710 fs_entry.kind = EntryKind::UnloadedDir;
4711 }
4712 }
4713
4714 state
4715 .insert_entry(fs_entry.clone(), self.fs.as_ref(), self.watcher.as_ref())
4716 .await;
4717
4718 if path.is_empty()
4719 && let Some((ignores, exclude, repo)) = new_ancestor_repo.take()
4720 {
4721 log::trace!("updating ancestor git repository");
4722 state.snapshot.ignores_by_parent_abs_path.extend(ignores);
4723 if let Some((ancestor_dot_git, work_directory)) = repo {
4724 if let Some(exclude) = exclude {
4725 let work_directory_abs_path = self
4726 .state
4727 .lock()
4728 .await
4729 .snapshot
4730 .work_directory_abs_path(&work_directory);
4731
4732 state
4733 .snapshot
4734 .repo_exclude_by_work_dir_abs_path
4735 .insert(work_directory_abs_path.into(), (exclude, false));
4736 }
4737 state
4738 .insert_git_repository_for_path(
4739 work_directory,
4740 ancestor_dot_git.into(),
4741 self.fs.as_ref(),
4742 self.watcher.as_ref(),
4743 )
4744 .await
4745 .log_err();
4746 }
4747 }
4748 }
4749 Ok(None) => {
4750 self.remove_repo_path(path.clone(), &mut state.snapshot);
4751 }
4752 Err(err) => {
4753 log::error!("error reading file {abs_path:?} on event: {err:#}");
4754 }
4755 }
4756 }
4757
4758 util::extend_sorted(
4759 &mut state.changed_paths,
4760 relative_paths.iter().cloned(),
4761 usize::MAX,
4762 Ord::cmp,
4763 );
4764 }
4765
4766 fn remove_repo_path(&self, path: Arc<RelPath>, snapshot: &mut LocalSnapshot) -> Option<()> {
4767 if !path.components().any(|component| component == DOT_GIT)
4768 && let Some(local_repo) = snapshot.local_repo_for_work_directory_path(&path)
4769 {
4770 let id = local_repo.work_directory_id;
4771 log::debug!("remove repo path: {:?}", path);
4772 snapshot.git_repositories.remove(&id);
4773 return Some(());
4774 }
4775
4776 Some(())
4777 }
4778
4779 async fn update_ignore_statuses_for_paths(
4780 &self,
4781 scan_job_tx: Sender<ScanJob>,
4782 prev_snapshot: LocalSnapshot,
4783 ignores_to_update: Vec<(Arc<Path>, IgnoreStack)>,
4784 ) {
4785 let (ignore_queue_tx, ignore_queue_rx) = channel::unbounded();
4786 {
4787 for (parent_abs_path, ignore_stack) in ignores_to_update {
4788 ignore_queue_tx
4789 .send_blocking(UpdateIgnoreStatusJob {
4790 abs_path: parent_abs_path,
4791 ignore_stack,
4792 ignore_queue: ignore_queue_tx.clone(),
4793 scan_queue: scan_job_tx.clone(),
4794 })
4795 .unwrap();
4796 }
4797 }
4798 drop(ignore_queue_tx);
4799
4800 self.executor
4801 .scoped(|scope| {
4802 for _ in 0..self.executor.num_cpus() {
4803 scope.spawn(async {
4804 loop {
4805 select_biased! {
4806 // Process any path refresh requests before moving on to process
4807 // the queue of ignore statuses.
4808 request = self.next_scan_request().fuse() => {
4809 let Ok(request) = request else { break };
4810 if !self.process_scan_request(request, true).await {
4811 return;
4812 }
4813 }
4814
4815 // Recursively process directories whose ignores have changed.
4816 job = ignore_queue_rx.recv().fuse() => {
4817 let Ok(job) = job else { break };
4818 self.update_ignore_status(job, &prev_snapshot).await;
4819 }
4820 }
4821 }
4822 });
4823 }
4824 })
4825 .await;
4826 }
4827
4828 async fn ignores_needing_update(&self) -> Vec<Arc<Path>> {
4829 let mut ignores_to_update = Vec::new();
4830 let mut excludes_to_load: Vec<(Arc<Path>, PathBuf)> = Vec::new();
4831
4832 // First pass: collect updates and drop stale entries without awaiting.
4833 {
4834 let snapshot = &mut self.state.lock().await.snapshot;
4835 let abs_path = snapshot.abs_path.clone();
4836 let mut repo_exclude_keys_to_remove: Vec<Arc<Path>> = Vec::new();
4837
4838 for (work_dir_abs_path, (_, needs_update)) in
4839 snapshot.repo_exclude_by_work_dir_abs_path.iter_mut()
4840 {
4841 let repository = snapshot
4842 .git_repositories
4843 .iter()
4844 .find(|(_, repo)| &repo.work_directory_abs_path == work_dir_abs_path);
4845
4846 if *needs_update {
4847 *needs_update = false;
4848 ignores_to_update.push(work_dir_abs_path.clone());
4849
4850 if let Some((_, repository)) = repository {
4851 let exclude_abs_path = repository.common_dir_abs_path.join(REPO_EXCLUDE);
4852 excludes_to_load.push((work_dir_abs_path.clone(), exclude_abs_path));
4853 }
4854 }
4855
4856 if repository.is_none() {
4857 repo_exclude_keys_to_remove.push(work_dir_abs_path.clone());
4858 }
4859 }
4860
4861 for key in repo_exclude_keys_to_remove {
4862 snapshot.repo_exclude_by_work_dir_abs_path.remove(&key);
4863 }
4864
4865 snapshot
4866 .ignores_by_parent_abs_path
4867 .retain(|parent_abs_path, (_, needs_update)| {
4868 if let Ok(parent_path) = parent_abs_path.strip_prefix(abs_path.as_path())
4869 && let Some(parent_path) =
4870 RelPath::new(&parent_path, PathStyle::local()).log_err()
4871 {
4872 if *needs_update {
4873 *needs_update = false;
4874 if snapshot.snapshot.entry_for_path(&parent_path).is_some() {
4875 ignores_to_update.push(parent_abs_path.clone());
4876 }
4877 }
4878
4879 let ignore_path = parent_path.join(RelPath::unix(GITIGNORE).unwrap());
4880 if snapshot.snapshot.entry_for_path(&ignore_path).is_none() {
4881 return false;
4882 }
4883 }
4884 true
4885 });
4886 }
4887
4888 // Load gitignores asynchronously (outside the lock)
4889 let mut loaded_excludes: Vec<(Arc<Path>, Arc<Gitignore>)> = Vec::new();
4890 for (work_dir_abs_path, exclude_abs_path) in excludes_to_load {
4891 if let Ok(current_exclude) = build_gitignore(&exclude_abs_path, self.fs.as_ref()).await
4892 {
4893 loaded_excludes.push((work_dir_abs_path, Arc::new(current_exclude)));
4894 }
4895 }
4896
4897 // Second pass: apply updates.
4898 if !loaded_excludes.is_empty() {
4899 let snapshot = &mut self.state.lock().await.snapshot;
4900
4901 for (work_dir_abs_path, exclude) in loaded_excludes {
4902 if let Some((existing_exclude, _)) = snapshot
4903 .repo_exclude_by_work_dir_abs_path
4904 .get_mut(&work_dir_abs_path)
4905 {
4906 *existing_exclude = exclude;
4907 }
4908 }
4909 }
4910
4911 ignores_to_update
4912 }
4913
4914 async fn order_ignores(&self, mut ignores: Vec<Arc<Path>>) -> Vec<(Arc<Path>, IgnoreStack)> {
4915 let fs = self.fs.clone();
4916 let snapshot = self.state.lock().await.snapshot.clone();
4917 ignores.sort_unstable();
4918 let mut ignores_to_update = ignores.into_iter().peekable();
4919
4920 let mut result = vec![];
4921 while let Some(parent_abs_path) = ignores_to_update.next() {
4922 while ignores_to_update
4923 .peek()
4924 .map_or(false, |p| p.starts_with(&parent_abs_path))
4925 {
4926 ignores_to_update.next().unwrap();
4927 }
4928 let ignore_stack = snapshot
4929 .ignore_stack_for_abs_path(&parent_abs_path, true, fs.as_ref())
4930 .await;
4931 result.push((parent_abs_path, ignore_stack));
4932 }
4933
4934 result
4935 }
4936
4937 async fn update_ignore_status(&self, job: UpdateIgnoreStatusJob, snapshot: &LocalSnapshot) {
4938 log::trace!("update ignore status {:?}", job.abs_path);
4939
4940 let mut ignore_stack = job.ignore_stack;
4941 if let Some((ignore, _)) = snapshot.ignores_by_parent_abs_path.get(&job.abs_path) {
4942 ignore_stack =
4943 ignore_stack.append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4944 }
4945
4946 let mut entries_by_id_edits = Vec::new();
4947 let mut entries_by_path_edits = Vec::new();
4948 let Some(path) = job
4949 .abs_path
4950 .strip_prefix(snapshot.abs_path.as_path())
4951 .map_err(|_| {
4952 anyhow::anyhow!(
4953 "Failed to strip prefix '{}' from path '{}'",
4954 snapshot.abs_path.as_path().display(),
4955 job.abs_path.display()
4956 )
4957 })
4958 .log_err()
4959 else {
4960 return;
4961 };
4962
4963 let Some(path) = RelPath::new(&path, PathStyle::local()).log_err() else {
4964 return;
4965 };
4966
4967 if let Ok(Some(metadata)) = self.fs.metadata(&job.abs_path.join(DOT_GIT)).await
4968 && metadata.is_dir
4969 {
4970 ignore_stack.repo_root = Some(job.abs_path.clone());
4971 }
4972
4973 for mut entry in snapshot.child_entries(&path).cloned() {
4974 let was_ignored = entry.is_ignored;
4975 let abs_path: Arc<Path> = snapshot.absolutize(&entry.path).into();
4976 entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, entry.is_dir());
4977
4978 if entry.is_dir() {
4979 let child_ignore_stack = if entry.is_ignored {
4980 IgnoreStack::all()
4981 } else {
4982 ignore_stack.clone()
4983 };
4984
4985 // Scan any directories that were previously ignored and weren't previously scanned.
4986 if was_ignored && !entry.is_ignored && entry.kind.is_unloaded() {
4987 let state = self.state.lock().await;
4988 if state.should_scan_directory(&entry) {
4989 state
4990 .enqueue_scan_dir(
4991 abs_path.clone(),
4992 &entry,
4993 &job.scan_queue,
4994 self.fs.as_ref(),
4995 )
4996 .await;
4997 }
4998 }
4999
5000 job.ignore_queue
5001 .send(UpdateIgnoreStatusJob {
5002 abs_path: abs_path.clone(),
5003 ignore_stack: child_ignore_stack,
5004 ignore_queue: job.ignore_queue.clone(),
5005 scan_queue: job.scan_queue.clone(),
5006 })
5007 .await
5008 .unwrap();
5009 }
5010
5011 if entry.is_ignored != was_ignored {
5012 let mut path_entry = snapshot.entries_by_id.get(&entry.id, ()).unwrap().clone();
5013 path_entry.scan_id = snapshot.scan_id;
5014 path_entry.is_ignored = entry.is_ignored;
5015 entries_by_id_edits.push(Edit::Insert(path_entry));
5016 entries_by_path_edits.push(Edit::Insert(entry));
5017 }
5018 }
5019
5020 let state = &mut self.state.lock().await;
5021 for edit in &entries_by_path_edits {
5022 if let Edit::Insert(entry) = edit
5023 && let Err(ix) = state.changed_paths.binary_search(&entry.path)
5024 {
5025 state.changed_paths.insert(ix, entry.path.clone());
5026 }
5027 }
5028
5029 state
5030 .snapshot
5031 .entries_by_path
5032 .edit(entries_by_path_edits, ());
5033 state.snapshot.entries_by_id.edit(entries_by_id_edits, ());
5034 }
5035
5036 async fn update_git_repositories(&self, dot_git_paths: Vec<PathBuf>) -> Vec<Arc<Path>> {
5037 log::trace!("reloading repositories: {dot_git_paths:?}");
5038 let mut state = self.state.lock().await;
5039 let scan_id = state.snapshot.scan_id;
5040 let mut affected_repo_roots = Vec::new();
5041 for dot_git_dir in dot_git_paths {
5042 let existing_repository_entry =
5043 state
5044 .snapshot
5045 .git_repositories
5046 .iter()
5047 .find_map(|(_, repo)| {
5048 let dot_git_dir = SanitizedPath::new(&dot_git_dir);
5049 if SanitizedPath::new(repo.common_dir_abs_path.as_ref()) == dot_git_dir
5050 || SanitizedPath::new(repo.repository_dir_abs_path.as_ref())
5051 == dot_git_dir
5052 {
5053 Some(repo.clone())
5054 } else {
5055 None
5056 }
5057 });
5058
5059 match existing_repository_entry {
5060 None => {
5061 let Ok(relative) = dot_git_dir.strip_prefix(state.snapshot.abs_path()) else {
5062 debug_panic!(
5063 "update_git_repositories called with .git directory outside the worktree root"
5064 );
5065 return Vec::new();
5066 };
5067 affected_repo_roots.push(dot_git_dir.parent().unwrap().into());
5068 state
5069 .insert_git_repository(
5070 RelPath::new(relative, PathStyle::local())
5071 .unwrap()
5072 .into_arc(),
5073 self.fs.as_ref(),
5074 self.watcher.as_ref(),
5075 )
5076 .await;
5077 }
5078 Some(local_repository) => {
5079 state.snapshot.git_repositories.update(
5080 &local_repository.work_directory_id,
5081 |entry| {
5082 entry.git_dir_scan_id = scan_id;
5083 },
5084 );
5085 }
5086 };
5087 }
5088
5089 // Remove any git repositories whose .git entry no longer exists.
5090 let snapshot = &mut state.snapshot;
5091 let mut ids_to_preserve = HashSet::default();
5092 for (&work_directory_id, entry) in snapshot.git_repositories.iter() {
5093 let exists_in_snapshot =
5094 snapshot
5095 .entry_for_id(work_directory_id)
5096 .is_some_and(|entry| {
5097 snapshot
5098 .entry_for_path(&entry.path.join(RelPath::unix(DOT_GIT).unwrap()))
5099 .is_some()
5100 });
5101
5102 if exists_in_snapshot
5103 || matches!(
5104 self.fs.metadata(&entry.common_dir_abs_path).await,
5105 Ok(Some(_))
5106 )
5107 {
5108 ids_to_preserve.insert(work_directory_id);
5109 }
5110 }
5111
5112 snapshot
5113 .git_repositories
5114 .retain(|work_directory_id, entry| {
5115 let preserve = ids_to_preserve.contains(work_directory_id);
5116 if !preserve {
5117 affected_repo_roots.push(entry.dot_git_abs_path.parent().unwrap().into());
5118 snapshot
5119 .repo_exclude_by_work_dir_abs_path
5120 .remove(&entry.work_directory_abs_path);
5121 }
5122 preserve
5123 });
5124
5125 affected_repo_roots
5126 }
5127
5128 async fn progress_timer(&self, running: bool) {
5129 if !running {
5130 return futures::future::pending().await;
5131 }
5132
5133 #[cfg(any(test, feature = "test-support"))]
5134 if self.fs.is_fake() {
5135 return self.executor.simulate_random_delay().await;
5136 }
5137
5138 self.executor.timer(FS_WATCH_LATENCY).await
5139 }
5140
5141 fn is_path_private(&self, path: &RelPath) -> bool {
5142 !self.share_private_files && self.settings.is_path_private(path)
5143 }
5144
5145 async fn next_scan_request(&self) -> Result<ScanRequest> {
5146 let mut request = self.scan_requests_rx.recv().await?;
5147 while let Ok(next_request) = self.scan_requests_rx.try_recv() {
5148 request.relative_paths.extend(next_request.relative_paths);
5149 request.done.extend(next_request.done);
5150 }
5151 Ok(request)
5152 }
5153}
5154
5155async fn discover_ancestor_git_repo(
5156 fs: Arc<dyn Fs>,
5157 root_abs_path: &SanitizedPath,
5158) -> (
5159 HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
5160 Option<Arc<Gitignore>>,
5161 Option<(PathBuf, WorkDirectory)>,
5162) {
5163 let mut exclude = None;
5164 let mut ignores = HashMap::default();
5165 for (index, ancestor) in root_abs_path.as_path().ancestors().enumerate() {
5166 if index != 0 {
5167 if ancestor == paths::home_dir() {
5168 // Unless $HOME is itself the worktree root, don't consider it as a
5169 // containing git repository---expensive and likely unwanted.
5170 break;
5171 } else if let Ok(ignore) = build_gitignore(&ancestor.join(GITIGNORE), fs.as_ref()).await
5172 {
5173 ignores.insert(ancestor.into(), (ignore.into(), false));
5174 }
5175 }
5176
5177 let ancestor_dot_git = ancestor.join(DOT_GIT);
5178 log::trace!("considering ancestor: {ancestor_dot_git:?}");
5179 // Check whether the directory or file called `.git` exists (in the
5180 // case of worktrees it's a file.)
5181 if fs
5182 .metadata(&ancestor_dot_git)
5183 .await
5184 .is_ok_and(|metadata| metadata.is_some())
5185 {
5186 if index != 0 {
5187 // We canonicalize, since the FS events use the canonicalized path.
5188 if let Some(ancestor_dot_git) = fs.canonicalize(&ancestor_dot_git).await.log_err() {
5189 let location_in_repo = root_abs_path
5190 .as_path()
5191 .strip_prefix(ancestor)
5192 .unwrap()
5193 .into();
5194 log::info!("inserting parent git repo for this worktree: {location_in_repo:?}");
5195 // We associate the external git repo with our root folder and
5196 // also mark where in the git repo the root folder is located.
5197 return (
5198 ignores,
5199 exclude,
5200 Some((
5201 ancestor_dot_git,
5202 WorkDirectory::AboveProject {
5203 absolute_path: ancestor.into(),
5204 location_in_repo,
5205 },
5206 )),
5207 );
5208 };
5209 }
5210
5211 let repo_exclude_abs_path = ancestor_dot_git.join(REPO_EXCLUDE);
5212 if let Ok(repo_exclude) = build_gitignore(&repo_exclude_abs_path, fs.as_ref()).await {
5213 exclude = Some(Arc::new(repo_exclude));
5214 }
5215
5216 // Reached root of git repository.
5217 break;
5218 }
5219 }
5220
5221 (ignores, exclude, None)
5222}
5223
5224fn build_diff(
5225 phase: BackgroundScannerPhase,
5226 old_snapshot: &Snapshot,
5227 new_snapshot: &Snapshot,
5228 event_paths: &[Arc<RelPath>],
5229) -> UpdatedEntriesSet {
5230 use BackgroundScannerPhase::*;
5231 use PathChange::{Added, AddedOrUpdated, Loaded, Removed, Updated};
5232
5233 // Identify which paths have changed. Use the known set of changed
5234 // parent paths to optimize the search.
5235 let mut changes = Vec::new();
5236 let mut old_paths = old_snapshot.entries_by_path.cursor::<PathKey>(());
5237 let mut new_paths = new_snapshot.entries_by_path.cursor::<PathKey>(());
5238 let mut last_newly_loaded_dir_path = None;
5239 old_paths.next();
5240 new_paths.next();
5241 for path in event_paths {
5242 let path = PathKey(path.clone());
5243 if old_paths.item().is_some_and(|e| e.path < path.0) {
5244 old_paths.seek_forward(&path, Bias::Left);
5245 }
5246 if new_paths.item().is_some_and(|e| e.path < path.0) {
5247 new_paths.seek_forward(&path, Bias::Left);
5248 }
5249 loop {
5250 match (old_paths.item(), new_paths.item()) {
5251 (Some(old_entry), Some(new_entry)) => {
5252 if old_entry.path > path.0
5253 && new_entry.path > path.0
5254 && !old_entry.path.starts_with(&path.0)
5255 && !new_entry.path.starts_with(&path.0)
5256 {
5257 break;
5258 }
5259
5260 match Ord::cmp(&old_entry.path, &new_entry.path) {
5261 Ordering::Less => {
5262 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5263 old_paths.next();
5264 }
5265 Ordering::Equal => {
5266 if phase == EventsReceivedDuringInitialScan {
5267 if old_entry.id != new_entry.id {
5268 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5269 }
5270 // If the worktree was not fully initialized when this event was generated,
5271 // we can't know whether this entry was added during the scan or whether
5272 // it was merely updated.
5273 changes.push((
5274 new_entry.path.clone(),
5275 new_entry.id,
5276 AddedOrUpdated,
5277 ));
5278 } else if old_entry.id != new_entry.id {
5279 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5280 changes.push((new_entry.path.clone(), new_entry.id, Added));
5281 } else if old_entry != new_entry {
5282 if old_entry.kind.is_unloaded() {
5283 last_newly_loaded_dir_path = Some(&new_entry.path);
5284 changes.push((new_entry.path.clone(), new_entry.id, Loaded));
5285 } else {
5286 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5287 }
5288 }
5289 old_paths.next();
5290 new_paths.next();
5291 }
5292 Ordering::Greater => {
5293 let is_newly_loaded = phase == InitialScan
5294 || last_newly_loaded_dir_path
5295 .as_ref()
5296 .is_some_and(|dir| new_entry.path.starts_with(dir));
5297 changes.push((
5298 new_entry.path.clone(),
5299 new_entry.id,
5300 if is_newly_loaded { Loaded } else { Added },
5301 ));
5302 new_paths.next();
5303 }
5304 }
5305 }
5306 (Some(old_entry), None) => {
5307 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5308 old_paths.next();
5309 }
5310 (None, Some(new_entry)) => {
5311 let is_newly_loaded = phase == InitialScan
5312 || last_newly_loaded_dir_path
5313 .as_ref()
5314 .is_some_and(|dir| new_entry.path.starts_with(dir));
5315 changes.push((
5316 new_entry.path.clone(),
5317 new_entry.id,
5318 if is_newly_loaded { Loaded } else { Added },
5319 ));
5320 new_paths.next();
5321 }
5322 (None, None) => break,
5323 }
5324 }
5325 }
5326
5327 changes.into()
5328}
5329
5330fn swap_to_front(child_paths: &mut Vec<PathBuf>, file: &str) {
5331 let position = child_paths
5332 .iter()
5333 .position(|path| path.file_name().unwrap() == file);
5334 if let Some(position) = position {
5335 let temp = child_paths.remove(position);
5336 child_paths.insert(0, temp);
5337 }
5338}
5339
5340fn char_bag_for_path(root_char_bag: CharBag, path: &RelPath) -> CharBag {
5341 let mut result = root_char_bag;
5342 result.extend(path.as_unix_str().chars().map(|c| c.to_ascii_lowercase()));
5343 result
5344}
5345
5346#[derive(Debug)]
5347struct ScanJob {
5348 abs_path: Arc<Path>,
5349 path: Arc<RelPath>,
5350 ignore_stack: IgnoreStack,
5351 scan_queue: Sender<ScanJob>,
5352 ancestor_inodes: TreeSet<u64>,
5353 is_external: bool,
5354}
5355
5356struct UpdateIgnoreStatusJob {
5357 abs_path: Arc<Path>,
5358 ignore_stack: IgnoreStack,
5359 ignore_queue: Sender<UpdateIgnoreStatusJob>,
5360 scan_queue: Sender<ScanJob>,
5361}
5362
5363pub trait WorktreeModelHandle {
5364 #[cfg(any(test, feature = "test-support"))]
5365 fn flush_fs_events<'a>(
5366 &self,
5367 cx: &'a mut gpui::TestAppContext,
5368 ) -> futures::future::LocalBoxFuture<'a, ()>;
5369
5370 #[cfg(any(test, feature = "test-support"))]
5371 fn flush_fs_events_in_root_git_repository<'a>(
5372 &self,
5373 cx: &'a mut gpui::TestAppContext,
5374 ) -> futures::future::LocalBoxFuture<'a, ()>;
5375}
5376
5377impl WorktreeModelHandle for Entity<Worktree> {
5378 // When the worktree's FS event stream sometimes delivers "redundant" events for FS changes that
5379 // occurred before the worktree was constructed. These events can cause the worktree to perform
5380 // extra directory scans, and emit extra scan-state notifications.
5381 //
5382 // This function mutates the worktree's directory and waits for those mutations to be picked up,
5383 // to ensure that all redundant FS events have already been processed.
5384 #[cfg(any(test, feature = "test-support"))]
5385 fn flush_fs_events<'a>(
5386 &self,
5387 cx: &'a mut gpui::TestAppContext,
5388 ) -> futures::future::LocalBoxFuture<'a, ()> {
5389 let file_name = "fs-event-sentinel";
5390
5391 let tree = self.clone();
5392 let (fs, root_path) = self.read_with(cx, |tree, _| {
5393 let tree = tree.as_local().unwrap();
5394 (tree.fs.clone(), tree.abs_path.clone())
5395 });
5396
5397 async move {
5398 // Subscribe to events BEFORE creating the file to avoid race condition
5399 // where events fire before subscription is set up
5400 let mut events = cx.events(&tree);
5401
5402 fs.create_file(&root_path.join(file_name), Default::default())
5403 .await
5404 .unwrap();
5405
5406 // Check if condition is already met before waiting for events
5407 let file_exists = || {
5408 tree.read_with(cx, |tree, _| {
5409 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5410 .is_some()
5411 })
5412 };
5413
5414 // Use select to avoid blocking indefinitely if events are delayed
5415 while !file_exists() {
5416 futures::select_biased! {
5417 _ = events.next() => {}
5418 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5419 }
5420 }
5421
5422 fs.remove_file(&root_path.join(file_name), Default::default())
5423 .await
5424 .unwrap();
5425
5426 // Check if condition is already met before waiting for events
5427 let file_gone = || {
5428 tree.read_with(cx, |tree, _| {
5429 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5430 .is_none()
5431 })
5432 };
5433
5434 // Use select to avoid blocking indefinitely if events are delayed
5435 while !file_gone() {
5436 futures::select_biased! {
5437 _ = events.next() => {}
5438 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5439 }
5440 }
5441
5442 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5443 .await;
5444 }
5445 .boxed_local()
5446 }
5447
5448 // This function is similar to flush_fs_events, except that it waits for events to be flushed in
5449 // the .git folder of the root repository.
5450 // The reason for its existence is that a repository's .git folder might live *outside* of the
5451 // worktree and thus its FS events might go through a different path.
5452 // In order to flush those, we need to create artificial events in the .git folder and wait
5453 // for the repository to be reloaded.
5454 #[cfg(any(test, feature = "test-support"))]
5455 fn flush_fs_events_in_root_git_repository<'a>(
5456 &self,
5457 cx: &'a mut gpui::TestAppContext,
5458 ) -> futures::future::LocalBoxFuture<'a, ()> {
5459 let file_name = "fs-event-sentinel";
5460
5461 let tree = self.clone();
5462 let (fs, root_path, mut git_dir_scan_id) = self.read_with(cx, |tree, _| {
5463 let tree = tree.as_local().unwrap();
5464 let local_repo_entry = tree
5465 .git_repositories
5466 .values()
5467 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5468 .unwrap();
5469 (
5470 tree.fs.clone(),
5471 local_repo_entry.common_dir_abs_path.clone(),
5472 local_repo_entry.git_dir_scan_id,
5473 )
5474 });
5475
5476 let scan_id_increased = |tree: &mut Worktree, git_dir_scan_id: &mut usize| {
5477 let tree = tree.as_local().unwrap();
5478 // let repository = tree.repositories.first().unwrap();
5479 let local_repo_entry = tree
5480 .git_repositories
5481 .values()
5482 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5483 .unwrap();
5484
5485 if local_repo_entry.git_dir_scan_id > *git_dir_scan_id {
5486 *git_dir_scan_id = local_repo_entry.git_dir_scan_id;
5487 true
5488 } else {
5489 false
5490 }
5491 };
5492
5493 async move {
5494 // Subscribe to events BEFORE creating the file to avoid race condition
5495 // where events fire before subscription is set up
5496 let mut events = cx.events(&tree);
5497
5498 fs.create_file(&root_path.join(file_name), Default::default())
5499 .await
5500 .unwrap();
5501
5502 // Use select to avoid blocking indefinitely if events are delayed
5503 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5504 futures::select_biased! {
5505 _ = events.next() => {}
5506 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5507 }
5508 }
5509
5510 fs.remove_file(&root_path.join(file_name), Default::default())
5511 .await
5512 .unwrap();
5513
5514 // Use select to avoid blocking indefinitely if events are delayed
5515 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5516 futures::select_biased! {
5517 _ = events.next() => {}
5518 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5519 }
5520 }
5521
5522 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5523 .await;
5524 }
5525 .boxed_local()
5526 }
5527}
5528
5529#[derive(Clone, Debug)]
5530struct TraversalProgress<'a> {
5531 max_path: &'a RelPath,
5532 count: usize,
5533 non_ignored_count: usize,
5534 file_count: usize,
5535 non_ignored_file_count: usize,
5536}
5537
5538impl TraversalProgress<'_> {
5539 fn count(&self, include_files: bool, include_dirs: bool, include_ignored: bool) -> usize {
5540 match (include_files, include_dirs, include_ignored) {
5541 (true, true, true) => self.count,
5542 (true, true, false) => self.non_ignored_count,
5543 (true, false, true) => self.file_count,
5544 (true, false, false) => self.non_ignored_file_count,
5545 (false, true, true) => self.count - self.file_count,
5546 (false, true, false) => self.non_ignored_count - self.non_ignored_file_count,
5547 (false, false, _) => 0,
5548 }
5549 }
5550}
5551
5552impl<'a> sum_tree::Dimension<'a, EntrySummary> for TraversalProgress<'a> {
5553 fn zero(_cx: ()) -> Self {
5554 Default::default()
5555 }
5556
5557 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
5558 self.max_path = summary.max_path.as_ref();
5559 self.count += summary.count;
5560 self.non_ignored_count += summary.non_ignored_count;
5561 self.file_count += summary.file_count;
5562 self.non_ignored_file_count += summary.non_ignored_file_count;
5563 }
5564}
5565
5566impl Default for TraversalProgress<'_> {
5567 fn default() -> Self {
5568 Self {
5569 max_path: RelPath::empty(),
5570 count: 0,
5571 non_ignored_count: 0,
5572 file_count: 0,
5573 non_ignored_file_count: 0,
5574 }
5575 }
5576}
5577
5578#[derive(Debug)]
5579pub struct Traversal<'a> {
5580 snapshot: &'a Snapshot,
5581 cursor: sum_tree::Cursor<'a, 'static, Entry, TraversalProgress<'a>>,
5582 include_ignored: bool,
5583 include_files: bool,
5584 include_dirs: bool,
5585}
5586
5587impl<'a> Traversal<'a> {
5588 fn new(
5589 snapshot: &'a Snapshot,
5590 include_files: bool,
5591 include_dirs: bool,
5592 include_ignored: bool,
5593 start_path: &RelPath,
5594 ) -> Self {
5595 let mut cursor = snapshot.entries_by_path.cursor(());
5596 cursor.seek(&TraversalTarget::path(start_path), Bias::Left);
5597 let mut traversal = Self {
5598 snapshot,
5599 cursor,
5600 include_files,
5601 include_dirs,
5602 include_ignored,
5603 };
5604 if traversal.end_offset() == traversal.start_offset() {
5605 traversal.next();
5606 }
5607 traversal
5608 }
5609
5610 pub fn advance(&mut self) -> bool {
5611 self.advance_by(1)
5612 }
5613
5614 pub fn advance_by(&mut self, count: usize) -> bool {
5615 self.cursor.seek_forward(
5616 &TraversalTarget::Count {
5617 count: self.end_offset() + count,
5618 include_dirs: self.include_dirs,
5619 include_files: self.include_files,
5620 include_ignored: self.include_ignored,
5621 },
5622 Bias::Left,
5623 )
5624 }
5625
5626 pub fn advance_to_sibling(&mut self) -> bool {
5627 while let Some(entry) = self.cursor.item() {
5628 self.cursor
5629 .seek_forward(&TraversalTarget::successor(&entry.path), Bias::Left);
5630 if let Some(entry) = self.cursor.item()
5631 && (self.include_files || !entry.is_file())
5632 && (self.include_dirs || !entry.is_dir())
5633 && (self.include_ignored || !entry.is_ignored || entry.is_always_included)
5634 {
5635 return true;
5636 }
5637 }
5638 false
5639 }
5640
5641 pub fn back_to_parent(&mut self) -> bool {
5642 let Some(parent_path) = self.cursor.item().and_then(|entry| entry.path.parent()) else {
5643 return false;
5644 };
5645 self.cursor
5646 .seek(&TraversalTarget::path(parent_path), Bias::Left)
5647 }
5648
5649 pub fn entry(&self) -> Option<&'a Entry> {
5650 self.cursor.item()
5651 }
5652
5653 pub fn snapshot(&self) -> &'a Snapshot {
5654 self.snapshot
5655 }
5656
5657 pub fn start_offset(&self) -> usize {
5658 self.cursor
5659 .start()
5660 .count(self.include_files, self.include_dirs, self.include_ignored)
5661 }
5662
5663 pub fn end_offset(&self) -> usize {
5664 self.cursor
5665 .end()
5666 .count(self.include_files, self.include_dirs, self.include_ignored)
5667 }
5668}
5669
5670impl<'a> Iterator for Traversal<'a> {
5671 type Item = &'a Entry;
5672
5673 fn next(&mut self) -> Option<Self::Item> {
5674 if let Some(item) = self.entry() {
5675 self.advance();
5676 Some(item)
5677 } else {
5678 None
5679 }
5680 }
5681}
5682
5683#[derive(Debug, Clone, Copy)]
5684pub enum PathTarget<'a> {
5685 Path(&'a RelPath),
5686 Successor(&'a RelPath),
5687}
5688
5689impl PathTarget<'_> {
5690 fn cmp_path(&self, other: &RelPath) -> Ordering {
5691 match self {
5692 PathTarget::Path(path) => path.cmp(&other),
5693 PathTarget::Successor(path) => {
5694 if other.starts_with(path) {
5695 Ordering::Greater
5696 } else {
5697 Ordering::Equal
5698 }
5699 }
5700 }
5701 }
5702}
5703
5704impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, PathProgress<'a>> for PathTarget<'_> {
5705 fn cmp(&self, cursor_location: &PathProgress<'a>, _: S::Context<'_>) -> Ordering {
5706 self.cmp_path(cursor_location.max_path)
5707 }
5708}
5709
5710impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, TraversalProgress<'a>> for PathTarget<'_> {
5711 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: S::Context<'_>) -> Ordering {
5712 self.cmp_path(cursor_location.max_path)
5713 }
5714}
5715
5716#[derive(Debug)]
5717enum TraversalTarget<'a> {
5718 Path(PathTarget<'a>),
5719 Count {
5720 count: usize,
5721 include_files: bool,
5722 include_ignored: bool,
5723 include_dirs: bool,
5724 },
5725}
5726
5727impl<'a> TraversalTarget<'a> {
5728 fn path(path: &'a RelPath) -> Self {
5729 Self::Path(PathTarget::Path(path))
5730 }
5731
5732 fn successor(path: &'a RelPath) -> Self {
5733 Self::Path(PathTarget::Successor(path))
5734 }
5735
5736 fn cmp_progress(&self, progress: &TraversalProgress) -> Ordering {
5737 match self {
5738 TraversalTarget::Path(path) => path.cmp_path(progress.max_path),
5739 TraversalTarget::Count {
5740 count,
5741 include_files,
5742 include_dirs,
5743 include_ignored,
5744 } => Ord::cmp(
5745 count,
5746 &progress.count(*include_files, *include_dirs, *include_ignored),
5747 ),
5748 }
5749 }
5750}
5751
5752impl<'a> SeekTarget<'a, EntrySummary, TraversalProgress<'a>> for TraversalTarget<'_> {
5753 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5754 self.cmp_progress(cursor_location)
5755 }
5756}
5757
5758impl<'a> SeekTarget<'a, PathSummary<sum_tree::NoSummary>, TraversalProgress<'a>>
5759 for TraversalTarget<'_>
5760{
5761 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5762 self.cmp_progress(cursor_location)
5763 }
5764}
5765
5766pub struct ChildEntriesOptions {
5767 pub include_files: bool,
5768 pub include_dirs: bool,
5769 pub include_ignored: bool,
5770}
5771
5772pub struct ChildEntriesIter<'a> {
5773 parent_path: &'a RelPath,
5774 traversal: Traversal<'a>,
5775}
5776
5777impl<'a> Iterator for ChildEntriesIter<'a> {
5778 type Item = &'a Entry;
5779
5780 fn next(&mut self) -> Option<Self::Item> {
5781 if let Some(item) = self.traversal.entry()
5782 && item.path.starts_with(self.parent_path)
5783 {
5784 self.traversal.advance_to_sibling();
5785 return Some(item);
5786 }
5787 None
5788 }
5789}
5790
5791impl<'a> From<&'a Entry> for proto::Entry {
5792 fn from(entry: &'a Entry) -> Self {
5793 Self {
5794 id: entry.id.to_proto(),
5795 is_dir: entry.is_dir(),
5796 path: entry.path.as_ref().to_proto(),
5797 inode: entry.inode,
5798 mtime: entry.mtime.map(|time| time.into()),
5799 is_ignored: entry.is_ignored,
5800 is_hidden: entry.is_hidden,
5801 is_external: entry.is_external,
5802 is_fifo: entry.is_fifo,
5803 size: Some(entry.size),
5804 canonical_path: entry
5805 .canonical_path
5806 .as_ref()
5807 .map(|path| path.to_string_lossy().into_owned()),
5808 }
5809 }
5810}
5811
5812impl TryFrom<(&CharBag, &PathMatcher, proto::Entry)> for Entry {
5813 type Error = anyhow::Error;
5814
5815 fn try_from(
5816 (root_char_bag, always_included, entry): (&CharBag, &PathMatcher, proto::Entry),
5817 ) -> Result<Self> {
5818 let kind = if entry.is_dir {
5819 EntryKind::Dir
5820 } else {
5821 EntryKind::File
5822 };
5823
5824 let path =
5825 RelPath::from_proto(&entry.path).context("invalid relative path in proto message")?;
5826 let char_bag = char_bag_for_path(*root_char_bag, &path);
5827 let is_always_included = always_included.is_match(&path);
5828 Ok(Entry {
5829 id: ProjectEntryId::from_proto(entry.id),
5830 kind,
5831 path,
5832 inode: entry.inode,
5833 mtime: entry.mtime.map(|time| time.into()),
5834 size: entry.size.unwrap_or(0),
5835 canonical_path: entry
5836 .canonical_path
5837 .map(|path_string| Arc::from(PathBuf::from(path_string))),
5838 is_ignored: entry.is_ignored,
5839 is_hidden: entry.is_hidden,
5840 is_always_included,
5841 is_external: entry.is_external,
5842 is_private: false,
5843 char_bag,
5844 is_fifo: entry.is_fifo,
5845 })
5846 }
5847}
5848
5849#[derive(Clone, Copy, Debug, Default, Hash, PartialEq, Eq, PartialOrd, Ord)]
5850pub struct ProjectEntryId(usize);
5851
5852impl ProjectEntryId {
5853 pub const MAX: Self = Self(usize::MAX);
5854 pub const MIN: Self = Self(usize::MIN);
5855
5856 pub fn new(counter: &AtomicUsize) -> Self {
5857 Self(counter.fetch_add(1, SeqCst))
5858 }
5859
5860 pub fn from_proto(id: u64) -> Self {
5861 Self(id as usize)
5862 }
5863
5864 pub fn to_proto(self) -> u64 {
5865 self.0 as u64
5866 }
5867
5868 pub fn from_usize(id: usize) -> Self {
5869 ProjectEntryId(id)
5870 }
5871
5872 pub fn to_usize(self) -> usize {
5873 self.0
5874 }
5875}
5876
5877#[cfg(any(test, feature = "test-support"))]
5878impl CreatedEntry {
5879 pub fn into_included(self) -> Option<Entry> {
5880 match self {
5881 CreatedEntry::Included(entry) => Some(entry),
5882 CreatedEntry::Excluded { .. } => None,
5883 }
5884 }
5885}
5886
5887fn parse_gitfile(content: &str) -> anyhow::Result<&Path> {
5888 let path = content
5889 .strip_prefix("gitdir:")
5890 .with_context(|| format!("parsing gitfile content {content:?}"))?;
5891 Ok(Path::new(path.trim()))
5892}
5893
5894async fn discover_git_paths(dot_git_abs_path: &Arc<Path>, fs: &dyn Fs) -> (Arc<Path>, Arc<Path>) {
5895 let mut repository_dir_abs_path = dot_git_abs_path.clone();
5896 let mut common_dir_abs_path = dot_git_abs_path.clone();
5897
5898 if let Some(path) = fs
5899 .load(dot_git_abs_path)
5900 .await
5901 .ok()
5902 .as_ref()
5903 .and_then(|contents| parse_gitfile(contents).log_err())
5904 {
5905 let path = dot_git_abs_path
5906 .parent()
5907 .unwrap_or(Path::new(""))
5908 .join(path);
5909 if let Some(path) = fs.canonicalize(&path).await.log_err() {
5910 repository_dir_abs_path = Path::new(&path).into();
5911 common_dir_abs_path = repository_dir_abs_path.clone();
5912
5913 if let Some(commondir_contents) = fs.load(&path.join("commondir")).await.ok()
5914 && let Some(commondir_path) = fs
5915 .canonicalize(&path.join(commondir_contents.trim()))
5916 .await
5917 .log_err()
5918 {
5919 common_dir_abs_path = commondir_path.as_path().into();
5920 }
5921 }
5922 };
5923 (repository_dir_abs_path, common_dir_abs_path)
5924}
5925
5926struct NullWatcher;
5927
5928impl fs::Watcher for NullWatcher {
5929 fn add(&self, _path: &Path) -> Result<()> {
5930 Ok(())
5931 }
5932
5933 fn remove(&self, _path: &Path) -> Result<()> {
5934 Ok(())
5935 }
5936}
5937
5938const FILE_ANALYSIS_BYTES: usize = 1024;
5939
5940async fn decode_file_text(
5941 fs: &dyn Fs,
5942 abs_path: &Path,
5943) -> Result<(String, &'static Encoding, bool)> {
5944 let mut file = fs
5945 .open_sync(&abs_path)
5946 .await
5947 .with_context(|| format!("opening file {abs_path:?}"))?;
5948
5949 // First, read the beginning of the file to determine its kind and encoding.
5950 // We do not want to load an entire large blob into memory only to discard it.
5951 let mut file_first_bytes = Vec::with_capacity(FILE_ANALYSIS_BYTES);
5952 let mut buf = [0u8; FILE_ANALYSIS_BYTES];
5953 let mut reached_eof = false;
5954 loop {
5955 if file_first_bytes.len() >= FILE_ANALYSIS_BYTES {
5956 break;
5957 }
5958 let n = file
5959 .read(&mut buf)
5960 .with_context(|| format!("reading bytes of the file {abs_path:?}"))?;
5961 if n == 0 {
5962 reached_eof = true;
5963 break;
5964 }
5965 file_first_bytes.extend_from_slice(&buf[..n]);
5966 }
5967 let (bom_encoding, byte_content) = decode_byte_header(&file_first_bytes);
5968 anyhow::ensure!(
5969 byte_content != ByteContent::Binary,
5970 "Binary files are not supported"
5971 );
5972
5973 // If the file is eligible for opening, read the rest of the file.
5974 let mut content = file_first_bytes;
5975 if !reached_eof {
5976 let mut buf = [0u8; 8 * 1024];
5977 loop {
5978 let n = file
5979 .read(&mut buf)
5980 .with_context(|| format!("reading remaining bytes of the file {abs_path:?}"))?;
5981 if n == 0 {
5982 break;
5983 }
5984 content.extend_from_slice(&buf[..n]);
5985 }
5986 }
5987 decode_byte_full(content, bom_encoding, byte_content)
5988}
5989
5990fn decode_byte_header(prefix: &[u8]) -> (Option<&'static Encoding>, ByteContent) {
5991 if let Some((encoding, _bom_len)) = Encoding::for_bom(prefix) {
5992 return (Some(encoding), ByteContent::Unknown);
5993 }
5994 (None, analyze_byte_content(prefix))
5995}
5996
5997fn decode_byte_full(
5998 bytes: Vec<u8>,
5999 bom_encoding: Option<&'static Encoding>,
6000 byte_content: ByteContent,
6001) -> Result<(String, &'static Encoding, bool)> {
6002 if let Some(encoding) = bom_encoding {
6003 let (cow, _) = encoding.decode_with_bom_removal(&bytes);
6004 return Ok((cow.into_owned(), encoding, true));
6005 }
6006
6007 match byte_content {
6008 ByteContent::Utf16Le => {
6009 let encoding = encoding_rs::UTF_16LE;
6010 let (cow, _, _) = encoding.decode(&bytes);
6011 return Ok((cow.into_owned(), encoding, false));
6012 }
6013 ByteContent::Utf16Be => {
6014 let encoding = encoding_rs::UTF_16BE;
6015 let (cow, _, _) = encoding.decode(&bytes);
6016 return Ok((cow.into_owned(), encoding, false));
6017 }
6018 ByteContent::Binary => {
6019 anyhow::bail!("Binary files are not supported");
6020 }
6021 ByteContent::Unknown => {}
6022 }
6023
6024 fn detect_encoding(bytes: Vec<u8>) -> (String, &'static Encoding) {
6025 let mut detector = EncodingDetector::new();
6026 detector.feed(&bytes, true);
6027
6028 let encoding = detector.guess(None, true); // Use None for TLD hint to ensure neutral detection logic.
6029
6030 let (cow, _, _) = encoding.decode(&bytes);
6031 (cow.into_owned(), encoding)
6032 }
6033
6034 match String::from_utf8(bytes) {
6035 Ok(text) => {
6036 // ISO-2022-JP (and other ISO-2022 variants) consists entirely of 7-bit ASCII bytes,
6037 // so it is valid UTF-8. However, it contains escape sequences starting with '\x1b'.
6038 // If we find an escape character, we double-check the encoding to prevent
6039 // displaying raw escape sequences instead of the correct characters.
6040 if text.contains('\x1b') {
6041 let (s, enc) = detect_encoding(text.into_bytes());
6042 Ok((s, enc, false))
6043 } else {
6044 Ok((text, encoding_rs::UTF_8, false))
6045 }
6046 }
6047 Err(e) => {
6048 let (s, enc) = detect_encoding(e.into_bytes());
6049 Ok((s, enc, false))
6050 }
6051 }
6052}
6053
6054#[derive(PartialEq)]
6055enum ByteContent {
6056 Utf16Le,
6057 Utf16Be,
6058 Binary,
6059 Unknown,
6060}
6061
6062// Heuristic check using null byte distribution plus a generic text-likeness
6063// heuristic. This prefers UTF-16 when many bytes are NUL and otherwise
6064// distinguishes between text-like and binary-like content.
6065fn analyze_byte_content(bytes: &[u8]) -> ByteContent {
6066 if bytes.len() < 2 {
6067 return ByteContent::Unknown;
6068 }
6069
6070 if is_known_binary_header(bytes) {
6071 return ByteContent::Binary;
6072 }
6073
6074 let limit = bytes.len().min(FILE_ANALYSIS_BYTES);
6075 let mut even_null_count = 0usize;
6076 let mut odd_null_count = 0usize;
6077 let mut non_text_like_count = 0usize;
6078
6079 for (i, &byte) in bytes[..limit].iter().enumerate() {
6080 if byte == 0 {
6081 if i % 2 == 0 {
6082 even_null_count += 1;
6083 } else {
6084 odd_null_count += 1;
6085 }
6086 non_text_like_count += 1;
6087 continue;
6088 }
6089
6090 let is_text_like = match byte {
6091 b'\t' | b'\n' | b'\r' | 0x0C => true,
6092 0x20..=0x7E => true,
6093 // Treat bytes that are likely part of UTF-8 or single-byte encodings as text-like.
6094 0x80..=0xBF | 0xC2..=0xF4 => true,
6095 _ => false,
6096 };
6097
6098 if !is_text_like {
6099 non_text_like_count += 1;
6100 }
6101 }
6102
6103 let total_null_count = even_null_count + odd_null_count;
6104
6105 // If there are no NUL bytes at all, this is overwhelmingly likely to be text.
6106 if total_null_count == 0 {
6107 return ByteContent::Unknown;
6108 }
6109
6110 if total_null_count >= limit / 16 {
6111 if even_null_count > odd_null_count * 4 {
6112 return ByteContent::Utf16Be;
6113 }
6114 if odd_null_count > even_null_count * 4 {
6115 return ByteContent::Utf16Le;
6116 }
6117 return ByteContent::Binary;
6118 }
6119
6120 if non_text_like_count * 100 < limit * 8 {
6121 ByteContent::Unknown
6122 } else {
6123 ByteContent::Binary
6124 }
6125}
6126
6127fn is_known_binary_header(bytes: &[u8]) -> bool {
6128 bytes.starts_with(b"%PDF-") // PDF
6129 || bytes.starts_with(b"PK\x03\x04") // ZIP local header
6130 || bytes.starts_with(b"PK\x05\x06") // ZIP end of central directory
6131 || bytes.starts_with(b"PK\x07\x08") // ZIP spanning/splitting
6132 || bytes.starts_with(b"\x89PNG\r\n\x1a\n") // PNG
6133 || bytes.starts_with(b"\xFF\xD8\xFF") // JPEG
6134 || bytes.starts_with(b"GIF87a") // GIF87a
6135 || bytes.starts_with(b"GIF89a") // GIF89a
6136}