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