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, debug_panic, maybe,
72 paths::{PathMatcher, PathStyle, SanitizedPath, home_dir},
73 rel_path::RelPath,
74};
75pub use worktree_settings::WorktreeSettings;
76
77use crate::ignore::IgnoreKind;
78
79pub const FS_WATCH_LATENCY: Duration = Duration::from_millis(100);
80
81/// A set of local or remote files that are being opened as part of a project.
82/// Responsible for tracking related FS (for local)/collab (for remote) events and corresponding updates.
83/// Stores git repositories data and the diagnostics for the file(s).
84///
85/// Has an absolute path, and may be set to be visible in Zed UI or not.
86/// May correspond to a directory or a single file.
87/// Possible examples:
88/// * a drag and dropped file — may be added as an invisible, "ephemeral" entry to the current worktree
89/// * a directory opened in Zed — may be added as a visible entry to the current worktree
90///
91/// Uses [`Entry`] to track the state of each file/directory, can look up absolute paths for entries.
92pub enum Worktree {
93 Local(LocalWorktree),
94 Remote(RemoteWorktree),
95}
96
97/// An entry, created in the worktree.
98#[derive(Debug)]
99pub enum CreatedEntry {
100 /// Got created and indexed by the worktree, receiving a corresponding entry.
101 Included(Entry),
102 /// Got created, but not indexed due to falling under exclusion filters.
103 Excluded { abs_path: PathBuf },
104}
105
106#[derive(Debug)]
107pub struct LoadedFile {
108 pub file: Arc<File>,
109 pub text: String,
110 pub encoding: &'static Encoding,
111 pub has_bom: bool,
112}
113
114pub struct LoadedBinaryFile {
115 pub file: Arc<File>,
116 pub content: Vec<u8>,
117}
118
119impl fmt::Debug for LoadedBinaryFile {
120 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
121 f.debug_struct("LoadedBinaryFile")
122 .field("file", &self.file)
123 .field("content_bytes", &self.content.len())
124 .finish()
125 }
126}
127
128pub struct LocalWorktree {
129 snapshot: LocalSnapshot,
130 scan_requests_tx: channel::Sender<ScanRequest>,
131 path_prefixes_to_scan_tx: channel::Sender<PathPrefixScanRequest>,
132 is_scanning: (watch::Sender<bool>, watch::Receiver<bool>),
133 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 always_included_entries: Vec<Arc<RelPath>>,
180
181 /// A number that increases every time the worktree begins scanning
182 /// a set of paths from the filesystem. This scanning could be caused
183 /// by some operation performed on the worktree, such as reading or
184 /// writing a file, or by an event reported by the filesystem.
185 scan_id: usize,
186
187 /// The latest scan id that has completed, and whose preceding scans
188 /// have all completed. The current `scan_id` could be more than one
189 /// greater than the `completed_scan_id` if operations are performed
190 /// on the worktree while it is processing a file-system event.
191 completed_scan_id: usize,
192}
193
194/// This path corresponds to the 'content path' of a repository in relation
195/// to Zed's project root.
196/// In the majority of the cases, this is the folder that contains the .git folder.
197/// But if a sub-folder of a git repository is opened, this corresponds to the
198/// project root and the .git folder is located in a parent directory.
199#[derive(Clone, Debug, Ord, PartialOrd, Eq, PartialEq, Hash)]
200pub enum WorkDirectory {
201 InProject {
202 relative_path: Arc<RelPath>,
203 },
204 AboveProject {
205 absolute_path: Arc<Path>,
206 location_in_repo: Arc<Path>,
207 },
208}
209
210impl WorkDirectory {
211 fn path_key(&self) -> PathKey {
212 match self {
213 WorkDirectory::InProject { relative_path } => PathKey(relative_path.clone()),
214 WorkDirectory::AboveProject { .. } => PathKey(RelPath::empty().into()),
215 }
216 }
217
218 /// Returns true if the given path is a child of the work directory.
219 ///
220 /// Note that the path may not be a member of this repository, if there
221 /// is a repository in a directory between these two paths
222 /// external .git folder in a parent folder of the project root.
223 #[track_caller]
224 pub fn directory_contains(&self, path: &RelPath) -> bool {
225 match self {
226 WorkDirectory::InProject { relative_path } => path.starts_with(relative_path),
227 WorkDirectory::AboveProject { .. } => true,
228 }
229 }
230}
231
232impl Default for WorkDirectory {
233 fn default() -> Self {
234 Self::InProject {
235 relative_path: Arc::from(RelPath::empty()),
236 }
237 }
238}
239
240#[derive(Clone)]
241pub struct LocalSnapshot {
242 snapshot: Snapshot,
243 global_gitignore: Option<Arc<Gitignore>>,
244 /// Exclude files for all git repositories in the worktree, indexed by their absolute path.
245 /// The boolean indicates whether the gitignore needs to be updated.
246 repo_exclude_by_work_dir_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
247 /// All of the gitignore files in the worktree, indexed by their absolute path.
248 /// The boolean indicates whether the gitignore needs to be updated.
249 ignores_by_parent_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
250 /// All of the git repositories in the worktree, indexed by the project entry
251 /// id of their parent directory.
252 git_repositories: TreeMap<ProjectEntryId, LocalRepositoryEntry>,
253 /// The file handle of the worktree root
254 /// (so we can find it after it's been moved)
255 root_file_handle: Option<Arc<dyn fs::FileHandle>>,
256}
257
258struct BackgroundScannerState {
259 snapshot: LocalSnapshot,
260 scanned_dirs: HashSet<ProjectEntryId>,
261 path_prefixes_to_scan: HashSet<Arc<RelPath>>,
262 paths_to_scan: HashSet<Arc<RelPath>>,
263 /// The ids of all of the entries that were removed from the snapshot
264 /// as part of the current update. These entry ids may be re-used
265 /// if the same inode is discovered at a new path, or if the given
266 /// path is re-created after being deleted.
267 removed_entries: HashMap<u64, Entry>,
268 changed_paths: Vec<Arc<RelPath>>,
269 prev_snapshot: Snapshot,
270 scanning_enabled: bool,
271}
272
273#[derive(Clone, Debug, Eq, PartialEq)]
274struct EventRoot {
275 path: Arc<RelPath>,
276 was_rescanned: bool,
277}
278
279#[derive(Debug, Clone)]
280struct LocalRepositoryEntry {
281 work_directory_id: ProjectEntryId,
282 work_directory: WorkDirectory,
283 work_directory_abs_path: Arc<Path>,
284 git_dir_scan_id: usize,
285 /// Absolute path to the original .git entry that caused us to create this repository.
286 ///
287 /// This is normally a directory, but may be a "gitfile" that points to a directory elsewhere
288 /// (whose path we then store in `repository_dir_abs_path`).
289 dot_git_abs_path: Arc<Path>,
290 /// Absolute path to the "commondir" for this repository.
291 ///
292 /// This is always a directory. For a normal repository, this is the same as
293 /// `dot_git_abs_path`. For a linked worktree, this is the main repo's `.git`
294 /// directory (resolved from the worktree's `commondir` file). For a submodule,
295 /// this equals `repository_dir_abs_path` (submodules don't have a `commondir`
296 /// file).
297 common_dir_abs_path: Arc<Path>,
298 /// Absolute path to the directory holding the repository's state.
299 ///
300 /// For a normal repository, this is a directory and coincides with `dot_git_abs_path` and
301 /// `common_dir_abs_path`. For a submodule or worktree, this is some subdirectory of the
302 /// commondir like `/project/.git/modules/foo`.
303 repository_dir_abs_path: Arc<Path>,
304}
305
306impl sum_tree::Item for LocalRepositoryEntry {
307 type Summary = PathSummary<sum_tree::NoSummary>;
308
309 fn summary(&self, _: <Self::Summary as Summary>::Context<'_>) -> Self::Summary {
310 PathSummary {
311 max_path: self.work_directory.path_key().0,
312 item_summary: sum_tree::NoSummary,
313 }
314 }
315}
316
317impl KeyedItem for LocalRepositoryEntry {
318 type Key = PathKey;
319
320 fn key(&self) -> Self::Key {
321 self.work_directory.path_key()
322 }
323}
324
325impl Deref for LocalRepositoryEntry {
326 type Target = WorkDirectory;
327
328 fn deref(&self) -> &Self::Target {
329 &self.work_directory
330 }
331}
332
333impl Deref for LocalSnapshot {
334 type Target = Snapshot;
335
336 fn deref(&self) -> &Self::Target {
337 &self.snapshot
338 }
339}
340
341impl DerefMut for LocalSnapshot {
342 fn deref_mut(&mut self) -> &mut Self::Target {
343 &mut self.snapshot
344 }
345}
346
347enum ScanState {
348 Started,
349 Updated {
350 snapshot: LocalSnapshot,
351 changes: UpdatedEntriesSet,
352 barrier: SmallVec<[barrier::Sender; 1]>,
353 scanning: bool,
354 },
355 RootUpdated {
356 new_path: Arc<SanitizedPath>,
357 },
358 RootDeleted,
359}
360
361struct UpdateObservationState {
362 snapshots_tx: mpsc::UnboundedSender<(LocalSnapshot, UpdatedEntriesSet)>,
363 resume_updates: watch::Sender<()>,
364 _maintain_remote_snapshot: Task<Option<()>>,
365}
366
367#[derive(Debug, Clone)]
368pub enum Event {
369 UpdatedEntries(UpdatedEntriesSet),
370 UpdatedGitRepositories(UpdatedGitRepositoriesSet),
371 DeletedEntry(ProjectEntryId),
372 /// The worktree root itself has been deleted (for single-file worktrees)
373 Deleted,
374}
375
376impl EventEmitter<Event> for Worktree {}
377
378impl Worktree {
379 pub async fn local(
380 path: impl Into<Arc<Path>>,
381 visible: bool,
382 fs: Arc<dyn Fs>,
383 next_entry_id: Arc<AtomicUsize>,
384 scanning_enabled: bool,
385 worktree_id: WorktreeId,
386 cx: &mut AsyncApp,
387 ) -> Result<Entity<Self>> {
388 let abs_path = path.into();
389 let metadata = fs
390 .metadata(&abs_path)
391 .await
392 .context("failed to stat worktree path")?;
393
394 let fs_case_sensitive = fs.is_case_sensitive().await;
395
396 let root_file_handle = if metadata.as_ref().is_some() {
397 fs.open_handle(&abs_path)
398 .await
399 .with_context(|| {
400 format!(
401 "failed to open local worktree root at {}",
402 abs_path.display()
403 )
404 })
405 .log_err()
406 } else {
407 None
408 };
409
410 Ok(cx.new(move |cx: &mut Context<Worktree>| {
411 let mut snapshot = LocalSnapshot {
412 ignores_by_parent_abs_path: Default::default(),
413 global_gitignore: Default::default(),
414 repo_exclude_by_work_dir_abs_path: Default::default(),
415 git_repositories: Default::default(),
416 snapshot: Snapshot::new(
417 worktree_id,
418 abs_path
419 .file_name()
420 .and_then(|f| f.to_str())
421 .map_or(RelPath::empty().into(), |f| {
422 RelPath::unix(f).unwrap().into()
423 }),
424 abs_path.clone(),
425 PathStyle::local(),
426 ),
427 root_file_handle,
428 };
429
430 let worktree_id = snapshot.id();
431 let settings_location = Some(SettingsLocation {
432 worktree_id,
433 path: RelPath::empty(),
434 });
435
436 let settings = WorktreeSettings::get(settings_location, cx).clone();
437 cx.observe_global::<SettingsStore>(move |this, cx| {
438 if let Self::Local(this) = this {
439 let settings = WorktreeSettings::get(settings_location, cx).clone();
440 if this.settings != settings {
441 this.settings = settings;
442 this.restart_background_scanners(cx);
443 }
444 }
445 })
446 .detach();
447
448 let share_private_files = false;
449 if let Some(metadata) = metadata {
450 let mut entry = Entry::new(
451 RelPath::empty().into(),
452 &metadata,
453 ProjectEntryId::new(&next_entry_id),
454 snapshot.root_char_bag,
455 None,
456 );
457 if metadata.is_dir {
458 if !scanning_enabled {
459 entry.kind = EntryKind::UnloadedDir;
460 }
461 } else {
462 if let Some(file_name) = abs_path.file_name()
463 && let Some(file_name) = file_name.to_str()
464 && let Ok(path) = RelPath::unix(file_name)
465 {
466 entry.is_private = !share_private_files && settings.is_path_private(path);
467 entry.is_hidden = settings.is_path_hidden(path);
468 }
469 }
470 cx.foreground_executor()
471 .block_on(snapshot.insert_entry(entry, fs.as_ref()));
472 }
473
474 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
475 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
476 let mut worktree = LocalWorktree {
477 share_private_files,
478 next_entry_id,
479 snapshot,
480 is_scanning: watch::channel_with(true),
481 snapshot_subscriptions: Default::default(),
482 update_observer: None,
483 scan_requests_tx,
484 path_prefixes_to_scan_tx,
485 _background_scanner_tasks: Vec::new(),
486 fs,
487 fs_case_sensitive,
488 visible,
489 settings,
490 scanning_enabled,
491 };
492 worktree.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
493 Worktree::Local(worktree)
494 }))
495 }
496
497 pub fn remote(
498 project_id: u64,
499 replica_id: ReplicaId,
500 worktree: proto::WorktreeMetadata,
501 client: AnyProtoClient,
502 path_style: PathStyle,
503 cx: &mut App,
504 ) -> Entity<Self> {
505 cx.new(|cx: &mut Context<Self>| {
506 let snapshot = Snapshot::new(
507 WorktreeId::from_proto(worktree.id),
508 RelPath::from_proto(&worktree.root_name)
509 .unwrap_or_else(|_| RelPath::empty().into()),
510 Path::new(&worktree.abs_path).into(),
511 path_style,
512 );
513
514 let background_snapshot = Arc::new(Mutex::new((
515 snapshot.clone(),
516 Vec::<proto::UpdateWorktree>::new(),
517 )));
518 let (background_updates_tx, mut background_updates_rx) =
519 mpsc::unbounded::<proto::UpdateWorktree>();
520 let (mut snapshot_updated_tx, mut snapshot_updated_rx) = watch::channel();
521
522 let worktree_id = snapshot.id();
523 let settings_location = Some(SettingsLocation {
524 worktree_id,
525 path: RelPath::empty(),
526 });
527
528 let settings = WorktreeSettings::get(settings_location, cx).clone();
529 let worktree = RemoteWorktree {
530 client,
531 project_id,
532 replica_id,
533 snapshot,
534 file_scan_inclusions: settings.parent_dir_scan_inclusions.clone(),
535 background_snapshot: background_snapshot.clone(),
536 updates_tx: Some(background_updates_tx),
537 update_observer: None,
538 snapshot_subscriptions: Default::default(),
539 visible: worktree.visible,
540 disconnected: false,
541 };
542
543 // Apply updates to a separate snapshot in a background task, then
544 // send them to a foreground task which updates the model.
545 cx.background_spawn(async move {
546 while let Some(update) = background_updates_rx.next().await {
547 {
548 let mut lock = background_snapshot.lock();
549 lock.0.apply_remote_update(
550 update.clone(),
551 &settings.parent_dir_scan_inclusions,
552 );
553 lock.1.push(update);
554 }
555 snapshot_updated_tx.send(()).await.ok();
556 }
557 })
558 .detach();
559
560 // On the foreground task, update to the latest snapshot and notify
561 // any update observer of all updates that led to that snapshot.
562 cx.spawn(async move |this, cx| {
563 while (snapshot_updated_rx.recv().await).is_some() {
564 this.update(cx, |this, cx| {
565 let mut entries_changed = false;
566 let this = this.as_remote_mut().unwrap();
567 {
568 let mut lock = this.background_snapshot.lock();
569 this.snapshot = lock.0.clone();
570 for update in lock.1.drain(..) {
571 entries_changed |= !update.updated_entries.is_empty()
572 || !update.removed_entries.is_empty();
573 if let Some(tx) = &this.update_observer {
574 tx.unbounded_send(update).ok();
575 }
576 }
577 };
578
579 if entries_changed {
580 cx.emit(Event::UpdatedEntries(Arc::default()));
581 }
582 cx.notify();
583 while let Some((scan_id, _)) = this.snapshot_subscriptions.front() {
584 if this.observed_snapshot(*scan_id) {
585 let (_, tx) = this.snapshot_subscriptions.pop_front().unwrap();
586 let _ = tx.send(());
587 } else {
588 break;
589 }
590 }
591 })?;
592 }
593 anyhow::Ok(())
594 })
595 .detach();
596
597 Worktree::Remote(worktree)
598 })
599 }
600
601 pub fn as_local(&self) -> Option<&LocalWorktree> {
602 if let Worktree::Local(worktree) = self {
603 Some(worktree)
604 } else {
605 None
606 }
607 }
608
609 pub fn as_remote(&self) -> Option<&RemoteWorktree> {
610 if let Worktree::Remote(worktree) = self {
611 Some(worktree)
612 } else {
613 None
614 }
615 }
616
617 pub fn as_local_mut(&mut self) -> Option<&mut LocalWorktree> {
618 if let Worktree::Local(worktree) = self {
619 Some(worktree)
620 } else {
621 None
622 }
623 }
624
625 pub fn as_remote_mut(&mut self) -> Option<&mut RemoteWorktree> {
626 if let Worktree::Remote(worktree) = self {
627 Some(worktree)
628 } else {
629 None
630 }
631 }
632
633 pub fn is_local(&self) -> bool {
634 matches!(self, Worktree::Local(_))
635 }
636
637 pub fn is_remote(&self) -> bool {
638 !self.is_local()
639 }
640
641 pub fn settings_location(&self, _: &Context<Self>) -> SettingsLocation<'static> {
642 SettingsLocation {
643 worktree_id: self.id(),
644 path: RelPath::empty(),
645 }
646 }
647
648 pub fn snapshot(&self) -> Snapshot {
649 match self {
650 Worktree::Local(worktree) => worktree.snapshot.snapshot.clone(),
651 Worktree::Remote(worktree) => worktree.snapshot.clone(),
652 }
653 }
654
655 pub fn scan_id(&self) -> usize {
656 match self {
657 Worktree::Local(worktree) => worktree.snapshot.scan_id,
658 Worktree::Remote(worktree) => worktree.snapshot.scan_id,
659 }
660 }
661
662 pub fn metadata_proto(&self) -> proto::WorktreeMetadata {
663 proto::WorktreeMetadata {
664 id: self.id().to_proto(),
665 root_name: self.root_name().to_proto(),
666 visible: self.is_visible(),
667 abs_path: self.abs_path().to_string_lossy().into_owned(),
668 }
669 }
670
671 pub fn completed_scan_id(&self) -> usize {
672 match self {
673 Worktree::Local(worktree) => worktree.snapshot.completed_scan_id,
674 Worktree::Remote(worktree) => worktree.snapshot.completed_scan_id,
675 }
676 }
677
678 pub fn is_visible(&self) -> bool {
679 match self {
680 Worktree::Local(worktree) => worktree.visible,
681 Worktree::Remote(worktree) => worktree.visible,
682 }
683 }
684
685 pub fn replica_id(&self) -> ReplicaId {
686 match self {
687 Worktree::Local(_) => ReplicaId::LOCAL,
688 Worktree::Remote(worktree) => worktree.replica_id,
689 }
690 }
691
692 pub fn abs_path(&self) -> Arc<Path> {
693 match self {
694 Worktree::Local(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
695 Worktree::Remote(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
696 }
697 }
698
699 pub fn root_file(&self, cx: &Context<Self>) -> Option<Arc<File>> {
700 let entry = self.root_entry()?;
701 Some(File::for_entry(entry.clone(), cx.entity()))
702 }
703
704 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
705 where
706 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
707 Fut: 'static + Send + Future<Output = bool>,
708 {
709 match self {
710 Worktree::Local(this) => this.observe_updates(project_id, cx, callback),
711 Worktree::Remote(this) => this.observe_updates(project_id, cx, callback),
712 }
713 }
714
715 pub fn stop_observing_updates(&mut self) {
716 match self {
717 Worktree::Local(this) => {
718 this.update_observer.take();
719 }
720 Worktree::Remote(this) => {
721 this.update_observer.take();
722 }
723 }
724 }
725
726 pub fn wait_for_snapshot(
727 &mut self,
728 scan_id: usize,
729 ) -> impl Future<Output = Result<()>> + use<> {
730 match self {
731 Worktree::Local(this) => this.wait_for_snapshot(scan_id).boxed(),
732 Worktree::Remote(this) => this.wait_for_snapshot(scan_id).boxed(),
733 }
734 }
735
736 #[cfg(feature = "test-support")]
737 pub fn has_update_observer(&self) -> bool {
738 match self {
739 Worktree::Local(this) => this.update_observer.is_some(),
740 Worktree::Remote(this) => this.update_observer.is_some(),
741 }
742 }
743
744 pub fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
745 match self {
746 Worktree::Local(this) => this.load_file(path, cx),
747 Worktree::Remote(_) => {
748 Task::ready(Err(anyhow!("remote worktrees can't yet load files")))
749 }
750 }
751 }
752
753 pub fn load_binary_file(
754 &self,
755 path: &RelPath,
756 cx: &Context<Worktree>,
757 ) -> Task<Result<LoadedBinaryFile>> {
758 match self {
759 Worktree::Local(this) => this.load_binary_file(path, cx),
760 Worktree::Remote(_) => {
761 Task::ready(Err(anyhow!("remote worktrees can't yet load binary files")))
762 }
763 }
764 }
765
766 pub fn write_file(
767 &self,
768 path: Arc<RelPath>,
769 text: Rope,
770 line_ending: LineEnding,
771 encoding: &'static Encoding,
772 has_bom: bool,
773 cx: &Context<Worktree>,
774 ) -> Task<Result<Arc<File>>> {
775 match self {
776 Worktree::Local(this) => {
777 this.write_file(path, text, line_ending, encoding, has_bom, cx)
778 }
779 Worktree::Remote(_) => {
780 Task::ready(Err(anyhow!("remote worktree can't yet write files")))
781 }
782 }
783 }
784
785 pub fn create_entry(
786 &mut self,
787 path: Arc<RelPath>,
788 is_directory: bool,
789 content: Option<Vec<u8>>,
790 cx: &Context<Worktree>,
791 ) -> Task<Result<CreatedEntry>> {
792 let worktree_id = self.id();
793 match self {
794 Worktree::Local(this) => this.create_entry(path, is_directory, content, cx),
795 Worktree::Remote(this) => {
796 let project_id = this.project_id;
797 let request = this.client.request(proto::CreateProjectEntry {
798 worktree_id: worktree_id.to_proto(),
799 project_id,
800 path: path.as_ref().to_proto(),
801 content,
802 is_directory,
803 });
804 cx.spawn(async move |this, cx| {
805 let response = request.await?;
806 match response.entry {
807 Some(entry) => this
808 .update(cx, |worktree, cx| {
809 worktree.as_remote_mut().unwrap().insert_entry(
810 entry,
811 response.worktree_scan_id as usize,
812 cx,
813 )
814 })?
815 .await
816 .map(CreatedEntry::Included),
817 None => {
818 let abs_path =
819 this.read_with(cx, |worktree, _| worktree.absolutize(&path))?;
820 Ok(CreatedEntry::Excluded { abs_path })
821 }
822 }
823 })
824 }
825 }
826 }
827
828 pub fn delete_entry(
829 &mut self,
830 entry_id: ProjectEntryId,
831 trash: bool,
832 cx: &mut Context<Worktree>,
833 ) -> Option<Task<Result<()>>> {
834 let task = match self {
835 Worktree::Local(this) => this.delete_entry(entry_id, trash, cx),
836 Worktree::Remote(this) => this.delete_entry(entry_id, trash, cx),
837 }?;
838
839 let entry = match &*self {
840 Worktree::Local(this) => this.entry_for_id(entry_id),
841 Worktree::Remote(this) => this.entry_for_id(entry_id),
842 }?;
843
844 let mut ids = vec![entry_id];
845 let path = &*entry.path;
846
847 self.get_children_ids_recursive(path, &mut ids);
848
849 for id in ids {
850 cx.emit(Event::DeletedEntry(id));
851 }
852 Some(task)
853 }
854
855 fn get_children_ids_recursive(&self, path: &RelPath, ids: &mut Vec<ProjectEntryId>) {
856 let children_iter = self.child_entries(path);
857 for child in children_iter {
858 ids.push(child.id);
859 self.get_children_ids_recursive(&child.path, ids);
860 }
861 }
862
863 // pub fn rename_entry(
864 // &mut self,
865 // entry_id: ProjectEntryId,
866 // new_path: Arc<RelPath>,
867 // cx: &Context<Self>,
868 // ) -> Task<Result<CreatedEntry>> {
869 // match self {
870 // Worktree::Local(this) => this.rename_entry(entry_id, new_path, cx),
871 // Worktree::Remote(this) => this.rename_entry(entry_id, new_path, cx),
872 // }
873 // }
874
875 pub fn copy_external_entries(
876 &mut self,
877 target_directory: Arc<RelPath>,
878 paths: Vec<Arc<Path>>,
879 fs: Arc<dyn Fs>,
880 cx: &Context<Worktree>,
881 ) -> Task<Result<Vec<ProjectEntryId>>> {
882 match self {
883 Worktree::Local(this) => this.copy_external_entries(target_directory, paths, cx),
884 Worktree::Remote(this) => this.copy_external_entries(target_directory, paths, fs, cx),
885 }
886 }
887
888 pub fn expand_entry(
889 &mut self,
890 entry_id: ProjectEntryId,
891 cx: &Context<Worktree>,
892 ) -> Option<Task<Result<()>>> {
893 match self {
894 Worktree::Local(this) => this.expand_entry(entry_id, cx),
895 Worktree::Remote(this) => {
896 let response = this.client.request(proto::ExpandProjectEntry {
897 project_id: this.project_id,
898 entry_id: entry_id.to_proto(),
899 });
900 Some(cx.spawn(async move |this, cx| {
901 let response = response.await?;
902 this.update(cx, |this, _| {
903 this.as_remote_mut()
904 .unwrap()
905 .wait_for_snapshot(response.worktree_scan_id as usize)
906 })?
907 .await?;
908 Ok(())
909 }))
910 }
911 }
912 }
913
914 pub fn expand_all_for_entry(
915 &mut self,
916 entry_id: ProjectEntryId,
917 cx: &Context<Worktree>,
918 ) -> Option<Task<Result<()>>> {
919 match self {
920 Worktree::Local(this) => this.expand_all_for_entry(entry_id, cx),
921 Worktree::Remote(this) => {
922 let response = this.client.request(proto::ExpandAllForProjectEntry {
923 project_id: this.project_id,
924 entry_id: entry_id.to_proto(),
925 });
926 Some(cx.spawn(async move |this, cx| {
927 let response = response.await?;
928 this.update(cx, |this, _| {
929 this.as_remote_mut()
930 .unwrap()
931 .wait_for_snapshot(response.worktree_scan_id as usize)
932 })?
933 .await?;
934 Ok(())
935 }))
936 }
937 }
938 }
939
940 pub async fn handle_create_entry(
941 this: Entity<Self>,
942 request: proto::CreateProjectEntry,
943 mut cx: AsyncApp,
944 ) -> Result<proto::ProjectEntryResponse> {
945 let (scan_id, entry) = this.update(&mut cx, |this, cx| {
946 anyhow::Ok((
947 this.scan_id(),
948 this.create_entry(
949 RelPath::from_proto(&request.path).with_context(|| {
950 format!("received invalid relative path {:?}", request.path)
951 })?,
952 request.is_directory,
953 request.content,
954 cx,
955 ),
956 ))
957 })?;
958 Ok(proto::ProjectEntryResponse {
959 entry: match &entry.await? {
960 CreatedEntry::Included(entry) => Some(entry.into()),
961 CreatedEntry::Excluded { .. } => None,
962 },
963 worktree_scan_id: scan_id as u64,
964 })
965 }
966
967 pub async fn handle_delete_entry(
968 this: Entity<Self>,
969 request: proto::DeleteProjectEntry,
970 mut cx: AsyncApp,
971 ) -> Result<proto::ProjectEntryResponse> {
972 let (scan_id, task) = this.update(&mut cx, |this, cx| {
973 (
974 this.scan_id(),
975 this.delete_entry(
976 ProjectEntryId::from_proto(request.entry_id),
977 request.use_trash,
978 cx,
979 ),
980 )
981 });
982 task.ok_or_else(|| anyhow::anyhow!("invalid entry"))?
983 .await?;
984 Ok(proto::ProjectEntryResponse {
985 entry: None,
986 worktree_scan_id: scan_id as u64,
987 })
988 }
989
990 pub async fn handle_expand_entry(
991 this: Entity<Self>,
992 request: proto::ExpandProjectEntry,
993 mut cx: AsyncApp,
994 ) -> Result<proto::ExpandProjectEntryResponse> {
995 let task = this.update(&mut cx, |this, cx| {
996 this.expand_entry(ProjectEntryId::from_proto(request.entry_id), cx)
997 });
998 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
999 .await?;
1000 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
1001 Ok(proto::ExpandProjectEntryResponse {
1002 worktree_scan_id: scan_id as u64,
1003 })
1004 }
1005
1006 pub async fn handle_expand_all_for_entry(
1007 this: Entity<Self>,
1008 request: proto::ExpandAllForProjectEntry,
1009 mut cx: AsyncApp,
1010 ) -> Result<proto::ExpandAllForProjectEntryResponse> {
1011 let task = this.update(&mut cx, |this, cx| {
1012 this.expand_all_for_entry(ProjectEntryId::from_proto(request.entry_id), cx)
1013 });
1014 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
1015 .await?;
1016 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
1017 Ok(proto::ExpandAllForProjectEntryResponse {
1018 worktree_scan_id: scan_id as u64,
1019 })
1020 }
1021
1022 pub fn is_single_file(&self) -> bool {
1023 self.root_dir().is_none()
1024 }
1025
1026 /// For visible worktrees, returns the path with the worktree name as the first component.
1027 /// Otherwise, returns an absolute path.
1028 pub fn full_path(&self, worktree_relative_path: &RelPath) -> PathBuf {
1029 if self.is_visible() {
1030 self.root_name()
1031 .join(worktree_relative_path)
1032 .display(self.path_style)
1033 .to_string()
1034 .into()
1035 } else {
1036 let full_path = self.abs_path();
1037 let mut full_path_string = if self.is_local()
1038 && let Ok(stripped) = full_path.strip_prefix(home_dir())
1039 {
1040 self.path_style
1041 .join("~", &*stripped.to_string_lossy())
1042 .unwrap()
1043 } else {
1044 full_path.to_string_lossy().into_owned()
1045 };
1046
1047 if worktree_relative_path.components().next().is_some() {
1048 full_path_string.push_str(self.path_style.primary_separator());
1049 full_path_string.push_str(&worktree_relative_path.display(self.path_style));
1050 }
1051
1052 full_path_string.into()
1053 }
1054 }
1055}
1056
1057impl LocalWorktree {
1058 pub fn fs(&self) -> &Arc<dyn Fs> {
1059 &self.fs
1060 }
1061
1062 pub fn is_path_private(&self, path: &RelPath) -> bool {
1063 !self.share_private_files && self.settings.is_path_private(path)
1064 }
1065
1066 pub fn fs_is_case_sensitive(&self) -> bool {
1067 self.fs_case_sensitive
1068 }
1069
1070 fn restart_background_scanners(&mut self, cx: &Context<Worktree>) {
1071 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
1072 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
1073 self.scan_requests_tx = scan_requests_tx;
1074 self.path_prefixes_to_scan_tx = path_prefixes_to_scan_tx;
1075
1076 self.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
1077 let always_included_entries = mem::take(&mut self.snapshot.always_included_entries);
1078 log::debug!(
1079 "refreshing entries for the following always included paths: {:?}",
1080 always_included_entries
1081 );
1082
1083 // Cleans up old always included entries to ensure they get updated properly. Otherwise,
1084 // nested always included entries may not get updated and will result in out-of-date info.
1085 self.refresh_entries_for_paths(always_included_entries);
1086 }
1087
1088 fn start_background_scanner(
1089 &mut self,
1090 scan_requests_rx: channel::Receiver<ScanRequest>,
1091 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
1092 cx: &Context<Worktree>,
1093 ) {
1094 let snapshot = self.snapshot();
1095 let share_private_files = self.share_private_files;
1096 let next_entry_id = self.next_entry_id.clone();
1097 let fs = self.fs.clone();
1098 let scanning_enabled = self.scanning_enabled;
1099 let settings = self.settings.clone();
1100 let (scan_states_tx, mut scan_states_rx) = mpsc::unbounded();
1101 let background_scanner = cx.background_spawn({
1102 let abs_path = snapshot.abs_path.as_path().to_path_buf();
1103 let background = cx.background_executor().clone();
1104 async move {
1105 let (events, watcher) = if scanning_enabled {
1106 fs.watch(&abs_path, FS_WATCH_LATENCY).await
1107 } else {
1108 (Box::pin(stream::pending()) as _, Arc::new(NullWatcher) as _)
1109 };
1110 let fs_case_sensitive = fs.is_case_sensitive().await;
1111
1112 let is_single_file = snapshot.snapshot.root_dir().is_none();
1113 let mut scanner = BackgroundScanner {
1114 fs,
1115 fs_case_sensitive,
1116 status_updates_tx: scan_states_tx,
1117 executor: background,
1118 scan_requests_rx,
1119 path_prefixes_to_scan_rx,
1120 next_entry_id,
1121 state: async_lock::Mutex::new(BackgroundScannerState {
1122 prev_snapshot: snapshot.snapshot.clone(),
1123 snapshot,
1124 scanned_dirs: Default::default(),
1125 scanning_enabled,
1126 path_prefixes_to_scan: Default::default(),
1127 paths_to_scan: Default::default(),
1128 removed_entries: Default::default(),
1129 changed_paths: Default::default(),
1130 }),
1131 phase: BackgroundScannerPhase::InitialScan,
1132 share_private_files,
1133 settings,
1134 watcher,
1135 is_single_file,
1136 };
1137
1138 scanner
1139 .run(Box::pin(events.map(|events| events.into_iter().collect())))
1140 .await;
1141 }
1142 });
1143 let scan_state_updater = cx.spawn(async move |this, cx| {
1144 while let Some((state, this)) = scan_states_rx.next().await.zip(this.upgrade()) {
1145 this.update(cx, |this, cx| {
1146 let this = this.as_local_mut().unwrap();
1147 match state {
1148 ScanState::Started => {
1149 *this.is_scanning.0.borrow_mut() = true;
1150 }
1151 ScanState::Updated {
1152 snapshot,
1153 changes,
1154 barrier,
1155 scanning,
1156 } => {
1157 *this.is_scanning.0.borrow_mut() = scanning;
1158 this.set_snapshot(snapshot, changes, cx);
1159 drop(barrier);
1160 }
1161 ScanState::RootUpdated { new_path } => {
1162 this.update_abs_path_and_refresh(new_path, cx);
1163 }
1164 ScanState::RootDeleted => {
1165 log::info!(
1166 "worktree root {} no longer exists, closing worktree",
1167 this.abs_path().display()
1168 );
1169 cx.emit(Event::Deleted);
1170 }
1171 }
1172 });
1173 }
1174 });
1175 self._background_scanner_tasks = vec![background_scanner, scan_state_updater];
1176 *self.is_scanning.0.borrow_mut() = true;
1177 }
1178
1179 fn set_snapshot(
1180 &mut self,
1181 mut new_snapshot: LocalSnapshot,
1182 entry_changes: UpdatedEntriesSet,
1183 cx: &mut Context<Worktree>,
1184 ) {
1185 let repo_changes = self.changed_repos(&self.snapshot, &mut new_snapshot);
1186 self.snapshot = new_snapshot;
1187
1188 if let Some(share) = self.update_observer.as_mut() {
1189 share
1190 .snapshots_tx
1191 .unbounded_send((self.snapshot.clone(), entry_changes.clone()))
1192 .ok();
1193 }
1194
1195 if !entry_changes.is_empty() {
1196 cx.emit(Event::UpdatedEntries(entry_changes));
1197 }
1198 if !repo_changes.is_empty() {
1199 cx.emit(Event::UpdatedGitRepositories(repo_changes));
1200 }
1201
1202 while let Some((scan_id, _)) = self.snapshot_subscriptions.front() {
1203 if self.snapshot.completed_scan_id >= *scan_id {
1204 let (_, tx) = self.snapshot_subscriptions.pop_front().unwrap();
1205 tx.send(()).ok();
1206 } else {
1207 break;
1208 }
1209 }
1210 }
1211
1212 fn changed_repos(
1213 &self,
1214 old_snapshot: &LocalSnapshot,
1215 new_snapshot: &mut LocalSnapshot,
1216 ) -> UpdatedGitRepositoriesSet {
1217 let mut changes = Vec::new();
1218 let mut old_repos = old_snapshot.git_repositories.iter().peekable();
1219 let new_repos = new_snapshot.git_repositories.clone();
1220 let mut new_repos = new_repos.iter().peekable();
1221
1222 loop {
1223 match (new_repos.peek().map(clone), old_repos.peek().map(clone)) {
1224 (Some((new_entry_id, new_repo)), Some((old_entry_id, old_repo))) => {
1225 match Ord::cmp(&new_entry_id, &old_entry_id) {
1226 Ordering::Less => {
1227 changes.push(UpdatedGitRepository {
1228 work_directory_id: new_entry_id,
1229 old_work_directory_abs_path: None,
1230 new_work_directory_abs_path: Some(
1231 new_repo.work_directory_abs_path.clone(),
1232 ),
1233 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1234 repository_dir_abs_path: Some(
1235 new_repo.repository_dir_abs_path.clone(),
1236 ),
1237 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1238 });
1239 new_repos.next();
1240 }
1241 Ordering::Equal => {
1242 if new_repo.git_dir_scan_id != old_repo.git_dir_scan_id
1243 || new_repo.work_directory_abs_path
1244 != old_repo.work_directory_abs_path
1245 {
1246 changes.push(UpdatedGitRepository {
1247 work_directory_id: new_entry_id,
1248 old_work_directory_abs_path: Some(
1249 old_repo.work_directory_abs_path.clone(),
1250 ),
1251 new_work_directory_abs_path: Some(
1252 new_repo.work_directory_abs_path.clone(),
1253 ),
1254 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1255 repository_dir_abs_path: Some(
1256 new_repo.repository_dir_abs_path.clone(),
1257 ),
1258 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1259 });
1260 }
1261 new_repos.next();
1262 old_repos.next();
1263 }
1264 Ordering::Greater => {
1265 changes.push(UpdatedGitRepository {
1266 work_directory_id: old_entry_id,
1267 old_work_directory_abs_path: Some(
1268 old_repo.work_directory_abs_path.clone(),
1269 ),
1270 new_work_directory_abs_path: None,
1271 dot_git_abs_path: None,
1272 repository_dir_abs_path: None,
1273 common_dir_abs_path: None,
1274 });
1275 old_repos.next();
1276 }
1277 }
1278 }
1279 (Some((entry_id, repo)), None) => {
1280 changes.push(UpdatedGitRepository {
1281 work_directory_id: entry_id,
1282 old_work_directory_abs_path: None,
1283 new_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1284 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1285 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1286 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1287 });
1288 new_repos.next();
1289 }
1290 (None, Some((entry_id, repo))) => {
1291 changes.push(UpdatedGitRepository {
1292 work_directory_id: entry_id,
1293 old_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1294 new_work_directory_abs_path: None,
1295 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1296 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1297 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1298 });
1299 old_repos.next();
1300 }
1301 (None, None) => break,
1302 }
1303 }
1304
1305 fn clone<T: Clone, U: Clone>(value: &(&T, &U)) -> (T, U) {
1306 (value.0.clone(), value.1.clone())
1307 }
1308
1309 changes.into()
1310 }
1311
1312 pub fn scan_complete(&self) -> impl Future<Output = ()> + use<> {
1313 let mut is_scanning_rx = self.is_scanning.1.clone();
1314 async move {
1315 let mut is_scanning = *is_scanning_rx.borrow();
1316 while is_scanning {
1317 if let Some(value) = is_scanning_rx.recv().await {
1318 is_scanning = value;
1319 } else {
1320 break;
1321 }
1322 }
1323 }
1324 }
1325
1326 pub fn wait_for_snapshot(
1327 &mut self,
1328 scan_id: usize,
1329 ) -> impl Future<Output = Result<()>> + use<> {
1330 let (tx, rx) = oneshot::channel();
1331 if self.snapshot.completed_scan_id >= scan_id {
1332 tx.send(()).ok();
1333 } else {
1334 match self
1335 .snapshot_subscriptions
1336 .binary_search_by_key(&scan_id, |probe| probe.0)
1337 {
1338 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
1339 }
1340 }
1341
1342 async move {
1343 rx.await?;
1344 Ok(())
1345 }
1346 }
1347
1348 pub fn snapshot(&self) -> LocalSnapshot {
1349 self.snapshot.clone()
1350 }
1351
1352 pub fn settings(&self) -> WorktreeSettings {
1353 self.settings.clone()
1354 }
1355
1356 fn load_binary_file(
1357 &self,
1358 path: &RelPath,
1359 cx: &Context<Worktree>,
1360 ) -> Task<Result<LoadedBinaryFile>> {
1361 let path = Arc::from(path);
1362 let abs_path = self.absolutize(&path);
1363 let fs = self.fs.clone();
1364 let entry = self.refresh_entry(path.clone(), None, cx);
1365 let is_private = self.is_path_private(&path);
1366
1367 let worktree = cx.weak_entity();
1368 cx.background_spawn(async move {
1369 let content = fs.load_bytes(&abs_path).await?;
1370
1371 let worktree = worktree.upgrade().context("worktree was dropped")?;
1372 let file = match entry.await? {
1373 Some(entry) => File::for_entry(entry, worktree),
1374 None => {
1375 let metadata = fs
1376 .metadata(&abs_path)
1377 .await
1378 .with_context(|| {
1379 format!("Loading metadata for excluded file {abs_path:?}")
1380 })?
1381 .with_context(|| {
1382 format!("Excluded file {abs_path:?} got removed during loading")
1383 })?;
1384 Arc::new(File {
1385 entry_id: None,
1386 worktree,
1387 path,
1388 disk_state: DiskState::Present {
1389 mtime: metadata.mtime,
1390 size: metadata.len,
1391 },
1392 is_local: true,
1393 is_private,
1394 })
1395 }
1396 };
1397
1398 Ok(LoadedBinaryFile { file, content })
1399 })
1400 }
1401
1402 #[ztracing::instrument(skip_all)]
1403 fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
1404 let path = Arc::from(path);
1405 let abs_path = self.absolutize(&path);
1406 let fs = self.fs.clone();
1407 let entry = self.refresh_entry(path.clone(), None, cx);
1408 let is_private = self.is_path_private(path.as_ref());
1409
1410 let this = cx.weak_entity();
1411 cx.background_spawn(async move {
1412 // WARN: Temporary workaround for #27283.
1413 // We are not efficient with our memory usage per file, and use in excess of 64GB for a 10GB file
1414 // Therefore, as a temporary workaround to prevent system freezes, we just bail before opening a file
1415 // if it is too large
1416 // 5GB seems to be more reasonable, peaking at ~16GB, while 6GB jumps up to >24GB which seems like a
1417 // reasonable limit
1418 {
1419 const FILE_SIZE_MAX: u64 = 6 * 1024 * 1024 * 1024; // 6GB
1420 if let Ok(Some(metadata)) = fs.metadata(&abs_path).await
1421 && metadata.len >= FILE_SIZE_MAX
1422 {
1423 anyhow::bail!("File is too large to load");
1424 }
1425 }
1426 let (text, encoding, has_bom) = decode_file_text(fs.as_ref(), &abs_path).await?;
1427
1428 let worktree = this.upgrade().context("worktree was dropped")?;
1429 let file = match entry.await? {
1430 Some(entry) => File::for_entry(entry, worktree),
1431 None => {
1432 let metadata = fs
1433 .metadata(&abs_path)
1434 .await
1435 .with_context(|| {
1436 format!("Loading metadata for excluded file {abs_path:?}")
1437 })?
1438 .with_context(|| {
1439 format!("Excluded file {abs_path:?} got removed during loading")
1440 })?;
1441 Arc::new(File {
1442 entry_id: None,
1443 worktree,
1444 path,
1445 disk_state: DiskState::Present {
1446 mtime: metadata.mtime,
1447 size: metadata.len,
1448 },
1449 is_local: true,
1450 is_private,
1451 })
1452 }
1453 };
1454
1455 Ok(LoadedFile {
1456 file,
1457 text,
1458 encoding,
1459 has_bom,
1460 })
1461 })
1462 }
1463
1464 /// Find the lowest path in the worktree's datastructures that is an ancestor
1465 fn lowest_ancestor(&self, path: &RelPath) -> Arc<RelPath> {
1466 let mut lowest_ancestor = None;
1467 for path in path.ancestors() {
1468 if self.entry_for_path(path).is_some() {
1469 lowest_ancestor = Some(path.into());
1470 break;
1471 }
1472 }
1473
1474 lowest_ancestor.unwrap_or_else(|| RelPath::empty().into())
1475 }
1476
1477 pub fn create_entry(
1478 &self,
1479 path: Arc<RelPath>,
1480 is_dir: bool,
1481 content: Option<Vec<u8>>,
1482 cx: &Context<Worktree>,
1483 ) -> Task<Result<CreatedEntry>> {
1484 let abs_path = self.absolutize(&path);
1485 let path_excluded = self.settings.is_path_excluded(&path);
1486 let fs = self.fs.clone();
1487 let task_abs_path = abs_path.clone();
1488 let write = cx.background_spawn(async move {
1489 if is_dir {
1490 fs.create_dir(&task_abs_path)
1491 .await
1492 .with_context(|| format!("creating directory {task_abs_path:?}"))
1493 } else {
1494 fs.write(&task_abs_path, content.as_deref().unwrap_or(&[]))
1495 .await
1496 .with_context(|| format!("creating file {task_abs_path:?}"))
1497 }
1498 });
1499
1500 let lowest_ancestor = self.lowest_ancestor(&path);
1501 cx.spawn(async move |this, cx| {
1502 write.await?;
1503 if path_excluded {
1504 return Ok(CreatedEntry::Excluded { abs_path });
1505 }
1506
1507 let (result, refreshes) = this.update(cx, |this, cx| {
1508 let mut refreshes = Vec::new();
1509 let refresh_paths = path.strip_prefix(&lowest_ancestor).unwrap();
1510 for refresh_path in refresh_paths.ancestors() {
1511 if refresh_path == RelPath::empty() {
1512 continue;
1513 }
1514 let refresh_full_path = lowest_ancestor.join(refresh_path);
1515
1516 refreshes.push(this.as_local_mut().unwrap().refresh_entry(
1517 refresh_full_path,
1518 None,
1519 cx,
1520 ));
1521 }
1522 (
1523 this.as_local_mut().unwrap().refresh_entry(path, None, cx),
1524 refreshes,
1525 )
1526 })?;
1527 for refresh in refreshes {
1528 refresh.await.log_err();
1529 }
1530
1531 Ok(result
1532 .await?
1533 .map(CreatedEntry::Included)
1534 .unwrap_or_else(|| CreatedEntry::Excluded { abs_path }))
1535 })
1536 }
1537
1538 pub fn write_file(
1539 &self,
1540 path: Arc<RelPath>,
1541 text: Rope,
1542 line_ending: LineEnding,
1543 encoding: &'static Encoding,
1544 has_bom: bool,
1545 cx: &Context<Worktree>,
1546 ) -> Task<Result<Arc<File>>> {
1547 let fs = self.fs.clone();
1548 let is_private = self.is_path_private(&path);
1549 let abs_path = self.absolutize(&path);
1550
1551 let write = cx.background_spawn({
1552 let fs = fs.clone();
1553 let abs_path = abs_path.clone();
1554 async move {
1555 // For UTF-8, use the optimized `fs.save` which writes Rope chunks directly to disk
1556 // without allocating a contiguous string.
1557 if encoding == encoding_rs::UTF_8 && !has_bom {
1558 return fs.save(&abs_path, &text, line_ending).await;
1559 }
1560
1561 // For legacy encodings (e.g. Shift-JIS), we fall back to converting the entire Rope
1562 // to a String/Bytes in memory before writing.
1563 //
1564 // Note: This is inefficient for very large files compared to the streaming approach above,
1565 // but supporting streaming writes for arbitrary encodings would require a significant
1566 // refactor of the `fs` crate to expose a Writer interface.
1567 let text_string = text.to_string();
1568 let normalized_text = match line_ending {
1569 LineEnding::Unix => text_string,
1570 LineEnding::Windows => text_string.replace('\n', "\r\n"),
1571 };
1572
1573 // Create the byte vector manually for UTF-16 encodings because encoding_rs encodes to UTF-8 by default (per WHATWG standards),
1574 // which is not what we want for saving files.
1575 let bytes = if encoding == encoding_rs::UTF_16BE {
1576 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1577 if has_bom {
1578 data.extend_from_slice(&[0xFE, 0xFF]); // BOM
1579 }
1580 let utf16be_bytes =
1581 normalized_text.encode_utf16().flat_map(|u| u.to_be_bytes());
1582 data.extend(utf16be_bytes);
1583 data.into()
1584 } else if encoding == encoding_rs::UTF_16LE {
1585 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1586 if has_bom {
1587 data.extend_from_slice(&[0xFF, 0xFE]); // BOM
1588 }
1589 let utf16le_bytes =
1590 normalized_text.encode_utf16().flat_map(|u| u.to_le_bytes());
1591 data.extend(utf16le_bytes);
1592 data.into()
1593 } else {
1594 // For other encodings (Shift-JIS, UTF-8 with BOM, etc.), delegate to encoding_rs.
1595 let bom_bytes = if has_bom {
1596 if encoding == encoding_rs::UTF_8 {
1597 vec![0xEF, 0xBB, 0xBF]
1598 } else {
1599 vec![]
1600 }
1601 } else {
1602 vec![]
1603 };
1604 let (cow, _, _) = encoding.encode(&normalized_text);
1605 if !bom_bytes.is_empty() {
1606 let mut bytes = bom_bytes;
1607 bytes.extend_from_slice(&cow);
1608 bytes.into()
1609 } else {
1610 cow
1611 }
1612 };
1613
1614 fs.write(&abs_path, &bytes).await
1615 }
1616 });
1617
1618 cx.spawn(async move |this, cx| {
1619 write.await?;
1620 let entry = this
1621 .update(cx, |this, cx| {
1622 this.as_local_mut()
1623 .unwrap()
1624 .refresh_entry(path.clone(), None, cx)
1625 })?
1626 .await?;
1627 let worktree = this.upgrade().context("worktree dropped")?;
1628 if let Some(entry) = entry {
1629 Ok(File::for_entry(entry, worktree))
1630 } else {
1631 let metadata = fs
1632 .metadata(&abs_path)
1633 .await
1634 .with_context(|| {
1635 format!("Fetching metadata after saving the excluded buffer {abs_path:?}")
1636 })?
1637 .with_context(|| {
1638 format!("Excluded buffer {path:?} got removed during saving")
1639 })?;
1640 Ok(Arc::new(File {
1641 worktree,
1642 path,
1643 disk_state: DiskState::Present {
1644 mtime: metadata.mtime,
1645 size: metadata.len,
1646 },
1647 entry_id: None,
1648 is_local: true,
1649 is_private,
1650 }))
1651 }
1652 })
1653 }
1654
1655 pub fn delete_entry(
1656 &self,
1657 entry_id: ProjectEntryId,
1658 trash: bool,
1659 cx: &Context<Worktree>,
1660 ) -> Option<Task<Result<()>>> {
1661 let entry = self.entry_for_id(entry_id)?.clone();
1662 let abs_path = self.absolutize(&entry.path);
1663 let fs = self.fs.clone();
1664
1665 let delete = cx.background_spawn(async move {
1666 if entry.is_file() {
1667 if trash {
1668 fs.trash_file(&abs_path, Default::default()).await?;
1669 } else {
1670 fs.remove_file(&abs_path, Default::default()).await?;
1671 }
1672 } else if trash {
1673 fs.trash_dir(
1674 &abs_path,
1675 RemoveOptions {
1676 recursive: true,
1677 ignore_if_not_exists: false,
1678 },
1679 )
1680 .await?;
1681 } else {
1682 fs.remove_dir(
1683 &abs_path,
1684 RemoveOptions {
1685 recursive: true,
1686 ignore_if_not_exists: false,
1687 },
1688 )
1689 .await?;
1690 }
1691 anyhow::Ok(entry.path)
1692 });
1693
1694 Some(cx.spawn(async move |this, cx| {
1695 let path = delete.await?;
1696 this.update(cx, |this, _| {
1697 this.as_local_mut()
1698 .unwrap()
1699 .refresh_entries_for_paths(vec![path])
1700 })?
1701 .recv()
1702 .await;
1703 Ok(())
1704 }))
1705 }
1706
1707 pub fn copy_external_entries(
1708 &self,
1709 target_directory: Arc<RelPath>,
1710 paths: Vec<Arc<Path>>,
1711 cx: &Context<Worktree>,
1712 ) -> Task<Result<Vec<ProjectEntryId>>> {
1713 let target_directory = self.absolutize(&target_directory);
1714 let worktree_path = self.abs_path().clone();
1715 let fs = self.fs.clone();
1716 let paths = paths
1717 .into_iter()
1718 .filter_map(|source| {
1719 let file_name = source.file_name()?;
1720 let mut target = target_directory.clone();
1721 target.push(file_name);
1722
1723 // Do not allow copying the same file to itself.
1724 if source.as_ref() != target.as_path() {
1725 Some((source, target))
1726 } else {
1727 None
1728 }
1729 })
1730 .collect::<Vec<_>>();
1731
1732 let paths_to_refresh = paths
1733 .iter()
1734 .filter_map(|(_, target)| {
1735 RelPath::new(
1736 target.strip_prefix(&worktree_path).ok()?,
1737 PathStyle::local(),
1738 )
1739 .ok()
1740 .map(|path| path.into_arc())
1741 })
1742 .collect::<Vec<_>>();
1743
1744 cx.spawn(async move |this, cx| {
1745 cx.background_spawn(async move {
1746 for (source, target) in paths {
1747 copy_recursive(
1748 fs.as_ref(),
1749 &source,
1750 &target,
1751 fs::CopyOptions {
1752 overwrite: true,
1753 ..Default::default()
1754 },
1755 )
1756 .await
1757 .with_context(|| {
1758 format!("Failed to copy file from {source:?} to {target:?}")
1759 })?;
1760 }
1761 anyhow::Ok(())
1762 })
1763 .await
1764 .log_err();
1765 let mut refresh = cx.read_entity(
1766 &this.upgrade().with_context(|| "Dropped worktree")?,
1767 |this, _| {
1768 anyhow::Ok::<postage::barrier::Receiver>(
1769 this.as_local()
1770 .with_context(|| "Worktree is not local")?
1771 .refresh_entries_for_paths(paths_to_refresh.clone()),
1772 )
1773 },
1774 )?;
1775
1776 cx.background_spawn(async move {
1777 refresh.next().await;
1778 anyhow::Ok(())
1779 })
1780 .await
1781 .log_err();
1782
1783 let this = this.upgrade().with_context(|| "Dropped worktree")?;
1784 Ok(cx.read_entity(&this, |this, _| {
1785 paths_to_refresh
1786 .iter()
1787 .filter_map(|path| Some(this.entry_for_path(path)?.id))
1788 .collect()
1789 }))
1790 })
1791 }
1792
1793 fn expand_entry(
1794 &self,
1795 entry_id: ProjectEntryId,
1796 cx: &Context<Worktree>,
1797 ) -> Option<Task<Result<()>>> {
1798 let path = self.entry_for_id(entry_id)?.path.clone();
1799 let mut refresh = self.refresh_entries_for_paths(vec![path]);
1800 Some(cx.background_spawn(async move {
1801 refresh.next().await;
1802 Ok(())
1803 }))
1804 }
1805
1806 fn expand_all_for_entry(
1807 &self,
1808 entry_id: ProjectEntryId,
1809 cx: &Context<Worktree>,
1810 ) -> Option<Task<Result<()>>> {
1811 let path = self.entry_for_id(entry_id).unwrap().path.clone();
1812 let mut rx = self.add_path_prefix_to_scan(path);
1813 Some(cx.background_spawn(async move {
1814 rx.next().await;
1815 Ok(())
1816 }))
1817 }
1818
1819 pub fn refresh_entries_for_paths(&self, paths: Vec<Arc<RelPath>>) -> barrier::Receiver {
1820 let (tx, rx) = barrier::channel();
1821 self.scan_requests_tx
1822 .try_send(ScanRequest {
1823 relative_paths: paths,
1824 done: smallvec![tx],
1825 })
1826 .ok();
1827 rx
1828 }
1829
1830 #[cfg(feature = "test-support")]
1831 pub fn manually_refresh_entries_for_paths(
1832 &self,
1833 paths: Vec<Arc<RelPath>>,
1834 ) -> barrier::Receiver {
1835 self.refresh_entries_for_paths(paths)
1836 }
1837
1838 pub fn add_path_prefix_to_scan(&self, path_prefix: Arc<RelPath>) -> barrier::Receiver {
1839 let (tx, rx) = barrier::channel();
1840 self.path_prefixes_to_scan_tx
1841 .try_send(PathPrefixScanRequest {
1842 path: path_prefix,
1843 done: smallvec![tx],
1844 })
1845 .ok();
1846 rx
1847 }
1848
1849 pub fn refresh_entry(
1850 &self,
1851 path: Arc<RelPath>,
1852 old_path: Option<Arc<RelPath>>,
1853 cx: &Context<Worktree>,
1854 ) -> Task<Result<Option<Entry>>> {
1855 if self.settings.is_path_excluded(&path) {
1856 return Task::ready(Ok(None));
1857 }
1858 let paths = if let Some(old_path) = old_path.as_ref() {
1859 vec![old_path.clone(), path.clone()]
1860 } else {
1861 vec![path.clone()]
1862 };
1863 let t0 = Instant::now();
1864 let mut refresh = self.refresh_entries_for_paths(paths);
1865 // todo(lw): Hot foreground spawn
1866 cx.spawn(async move |this, cx| {
1867 refresh.recv().await;
1868 log::trace!("refreshed entry {path:?} in {:?}", t0.elapsed());
1869 let new_entry = this.read_with(cx, |this, _| {
1870 this.entry_for_path(&path).cloned().with_context(|| {
1871 format!("Could not find entry in worktree for {path:?} after refresh")
1872 })
1873 })??;
1874 Ok(Some(new_entry))
1875 })
1876 }
1877
1878 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1879 where
1880 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1881 Fut: 'static + Send + Future<Output = bool>,
1882 {
1883 if let Some(observer) = self.update_observer.as_mut() {
1884 *observer.resume_updates.borrow_mut() = ();
1885 return;
1886 }
1887
1888 let (resume_updates_tx, mut resume_updates_rx) = watch::channel::<()>();
1889 let (snapshots_tx, mut snapshots_rx) =
1890 mpsc::unbounded::<(LocalSnapshot, UpdatedEntriesSet)>();
1891 snapshots_tx
1892 .unbounded_send((self.snapshot(), Arc::default()))
1893 .ok();
1894
1895 let worktree_id = self.id.to_proto();
1896 let _maintain_remote_snapshot = cx.background_spawn(async move {
1897 let mut is_first = true;
1898 while let Some((snapshot, entry_changes)) = snapshots_rx.next().await {
1899 let update = if is_first {
1900 is_first = false;
1901 snapshot.build_initial_update(project_id, worktree_id)
1902 } else {
1903 snapshot.build_update(project_id, worktree_id, entry_changes)
1904 };
1905
1906 for update in proto::split_worktree_update(update) {
1907 let _ = resume_updates_rx.try_recv();
1908 loop {
1909 let result = callback(update.clone());
1910 if result.await {
1911 break;
1912 } else {
1913 log::info!("waiting to resume updates");
1914 if resume_updates_rx.next().await.is_none() {
1915 return Some(());
1916 }
1917 }
1918 }
1919 }
1920 }
1921 Some(())
1922 });
1923
1924 self.update_observer = Some(UpdateObservationState {
1925 snapshots_tx,
1926 resume_updates: resume_updates_tx,
1927 _maintain_remote_snapshot,
1928 });
1929 }
1930
1931 pub fn share_private_files(&mut self, cx: &Context<Worktree>) {
1932 self.share_private_files = true;
1933 self.restart_background_scanners(cx);
1934 }
1935
1936 pub fn update_abs_path_and_refresh(
1937 &mut self,
1938 new_path: Arc<SanitizedPath>,
1939 cx: &Context<Worktree>,
1940 ) {
1941 self.snapshot.git_repositories = Default::default();
1942 self.snapshot.ignores_by_parent_abs_path = Default::default();
1943 let root_name = new_path
1944 .as_path()
1945 .file_name()
1946 .and_then(|f| f.to_str())
1947 .map_or(RelPath::empty().into(), |f| {
1948 RelPath::unix(f).unwrap().into()
1949 });
1950 self.snapshot.update_abs_path(new_path, root_name);
1951 self.restart_background_scanners(cx);
1952 }
1953 #[cfg(feature = "test-support")]
1954 pub fn repositories(&self) -> Vec<Arc<Path>> {
1955 self.git_repositories
1956 .values()
1957 .map(|entry| entry.work_directory_abs_path.clone())
1958 .collect::<Vec<_>>()
1959 }
1960}
1961
1962impl RemoteWorktree {
1963 pub fn project_id(&self) -> u64 {
1964 self.project_id
1965 }
1966
1967 pub fn client(&self) -> AnyProtoClient {
1968 self.client.clone()
1969 }
1970
1971 pub fn disconnected_from_host(&mut self) {
1972 self.updates_tx.take();
1973 self.snapshot_subscriptions.clear();
1974 self.disconnected = true;
1975 }
1976
1977 pub fn update_from_remote(&self, update: proto::UpdateWorktree) {
1978 if let Some(updates_tx) = &self.updates_tx {
1979 updates_tx
1980 .unbounded_send(update)
1981 .expect("consumer runs to completion");
1982 }
1983 }
1984
1985 fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1986 where
1987 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1988 Fut: 'static + Send + Future<Output = bool>,
1989 {
1990 let (tx, mut rx) = mpsc::unbounded();
1991 let initial_update = self
1992 .snapshot
1993 .build_initial_update(project_id, self.id().to_proto());
1994 self.update_observer = Some(tx);
1995 cx.spawn(async move |this, cx| {
1996 let mut update = initial_update;
1997 'outer: loop {
1998 // SSH projects use a special project ID of 0, and we need to
1999 // remap it to the correct one here.
2000 update.project_id = project_id;
2001
2002 for chunk in split_worktree_update(update) {
2003 if !callback(chunk).await {
2004 break 'outer;
2005 }
2006 }
2007
2008 if let Some(next_update) = rx.next().await {
2009 update = next_update;
2010 } else {
2011 break;
2012 }
2013 }
2014 this.update(cx, |this, _| {
2015 let this = this.as_remote_mut().unwrap();
2016 this.update_observer.take();
2017 })
2018 })
2019 .detach();
2020 }
2021
2022 fn observed_snapshot(&self, scan_id: usize) -> bool {
2023 self.completed_scan_id >= scan_id
2024 }
2025
2026 pub fn wait_for_snapshot(
2027 &mut self,
2028 scan_id: usize,
2029 ) -> impl Future<Output = Result<()>> + use<> {
2030 let (tx, rx) = oneshot::channel();
2031 if self.observed_snapshot(scan_id) {
2032 let _ = tx.send(());
2033 } else if self.disconnected {
2034 drop(tx);
2035 } else {
2036 match self
2037 .snapshot_subscriptions
2038 .binary_search_by_key(&scan_id, |probe| probe.0)
2039 {
2040 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
2041 }
2042 }
2043
2044 async move {
2045 rx.await?;
2046 Ok(())
2047 }
2048 }
2049
2050 pub fn insert_entry(
2051 &mut self,
2052 entry: proto::Entry,
2053 scan_id: usize,
2054 cx: &Context<Worktree>,
2055 ) -> Task<Result<Entry>> {
2056 let wait_for_snapshot = self.wait_for_snapshot(scan_id);
2057 cx.spawn(async move |this, cx| {
2058 wait_for_snapshot.await?;
2059 this.update(cx, |worktree, _| {
2060 let worktree = worktree.as_remote_mut().unwrap();
2061 let snapshot = &mut worktree.background_snapshot.lock().0;
2062 let entry = snapshot.insert_entry(entry, &worktree.file_scan_inclusions);
2063 worktree.snapshot = snapshot.clone();
2064 entry
2065 })?
2066 })
2067 }
2068
2069 fn delete_entry(
2070 &self,
2071 entry_id: ProjectEntryId,
2072 trash: bool,
2073 cx: &Context<Worktree>,
2074 ) -> Option<Task<Result<()>>> {
2075 let response = self.client.request(proto::DeleteProjectEntry {
2076 project_id: self.project_id,
2077 entry_id: entry_id.to_proto(),
2078 use_trash: trash,
2079 });
2080 Some(cx.spawn(async move |this, cx| {
2081 let response = response.await?;
2082 let scan_id = response.worktree_scan_id as usize;
2083
2084 this.update(cx, move |this, _| {
2085 this.as_remote_mut().unwrap().wait_for_snapshot(scan_id)
2086 })?
2087 .await?;
2088
2089 this.update(cx, |this, _| {
2090 let this = this.as_remote_mut().unwrap();
2091 let snapshot = &mut this.background_snapshot.lock().0;
2092 snapshot.delete_entry(entry_id);
2093 this.snapshot = snapshot.clone();
2094 })
2095 }))
2096 }
2097
2098 // fn rename_entry(
2099 // &self,
2100 // entry_id: ProjectEntryId,
2101 // new_path: impl Into<Arc<RelPath>>,
2102 // cx: &Context<Worktree>,
2103 // ) -> Task<Result<CreatedEntry>> {
2104 // let new_path: Arc<RelPath> = new_path.into();
2105 // let response = self.client.request(proto::RenameProjectEntry {
2106 // project_id: self.project_id,
2107 // entry_id: entry_id.to_proto(),
2108 // new_worktree_id: new_path.worktree_id,
2109 // new_path: new_path.as_ref().to_proto(),
2110 // });
2111 // cx.spawn(async move |this, cx| {
2112 // let response = response.await?;
2113 // match response.entry {
2114 // Some(entry) => this
2115 // .update(cx, |this, cx| {
2116 // this.as_remote_mut().unwrap().insert_entry(
2117 // entry,
2118 // response.worktree_scan_id as usize,
2119 // cx,
2120 // )
2121 // })?
2122 // .await
2123 // .map(CreatedEntry::Included),
2124 // None => {
2125 // let abs_path =
2126 // this.read_with(cx, |worktree, _| worktree.absolutize(&new_path))?;
2127 // Ok(CreatedEntry::Excluded { abs_path })
2128 // }
2129 // }
2130 // })
2131 // }
2132
2133 fn copy_external_entries(
2134 &self,
2135 target_directory: Arc<RelPath>,
2136 paths_to_copy: Vec<Arc<Path>>,
2137 local_fs: Arc<dyn Fs>,
2138 cx: &Context<Worktree>,
2139 ) -> Task<anyhow::Result<Vec<ProjectEntryId>>> {
2140 let client = self.client.clone();
2141 let worktree_id = self.id().to_proto();
2142 let project_id = self.project_id;
2143
2144 cx.background_spawn(async move {
2145 let mut requests = Vec::new();
2146 for root_path_to_copy in paths_to_copy {
2147 let Some(filename) = root_path_to_copy
2148 .file_name()
2149 .and_then(|name| name.to_str())
2150 .and_then(|filename| RelPath::unix(filename).ok())
2151 else {
2152 continue;
2153 };
2154 for (abs_path, is_directory) in
2155 read_dir_items(local_fs.as_ref(), &root_path_to_copy).await?
2156 {
2157 let Some(relative_path) = abs_path
2158 .strip_prefix(&root_path_to_copy)
2159 .map_err(|e| anyhow::Error::from(e))
2160 .and_then(|relative_path| RelPath::new(relative_path, PathStyle::local()))
2161 .log_err()
2162 else {
2163 continue;
2164 };
2165 let content = if is_directory {
2166 None
2167 } else {
2168 Some(local_fs.load_bytes(&abs_path).await?)
2169 };
2170
2171 let mut target_path = target_directory.join(filename);
2172 if relative_path.file_name().is_some() {
2173 target_path = target_path.join(&relative_path);
2174 }
2175
2176 requests.push(proto::CreateProjectEntry {
2177 project_id,
2178 worktree_id,
2179 path: target_path.to_proto(),
2180 is_directory,
2181 content,
2182 });
2183 }
2184 }
2185 requests.sort_unstable_by(|a, b| a.path.cmp(&b.path));
2186 requests.dedup();
2187
2188 let mut copied_entry_ids = Vec::new();
2189 for request in requests {
2190 let response = client.request(request).await?;
2191 copied_entry_ids.extend(response.entry.map(|e| ProjectEntryId::from_proto(e.id)));
2192 }
2193
2194 Ok(copied_entry_ids)
2195 })
2196 }
2197}
2198
2199impl Snapshot {
2200 pub fn new(
2201 id: WorktreeId,
2202 root_name: Arc<RelPath>,
2203 abs_path: Arc<Path>,
2204 path_style: PathStyle,
2205 ) -> Self {
2206 Snapshot {
2207 id,
2208 abs_path: SanitizedPath::from_arc(abs_path),
2209 path_style,
2210 root_char_bag: root_name
2211 .as_unix_str()
2212 .chars()
2213 .map(|c| c.to_ascii_lowercase())
2214 .collect(),
2215 root_name,
2216 always_included_entries: Default::default(),
2217 entries_by_path: Default::default(),
2218 entries_by_id: Default::default(),
2219 scan_id: 1,
2220 completed_scan_id: 0,
2221 }
2222 }
2223
2224 pub fn id(&self) -> WorktreeId {
2225 self.id
2226 }
2227
2228 // TODO:
2229 // Consider the following:
2230 //
2231 // ```rust
2232 // let abs_path: Arc<Path> = snapshot.abs_path(); // e.g. "C:\Users\user\Desktop\project"
2233 // let some_non_trimmed_path = Path::new("\\\\?\\C:\\Users\\user\\Desktop\\project\\main.rs");
2234 // // The caller perform some actions here:
2235 // some_non_trimmed_path.strip_prefix(abs_path); // This fails
2236 // some_non_trimmed_path.starts_with(abs_path); // This fails too
2237 // ```
2238 //
2239 // This is definitely a bug, but it's not clear if we should handle it here or not.
2240 pub fn abs_path(&self) -> &Arc<Path> {
2241 SanitizedPath::cast_arc_ref(&self.abs_path)
2242 }
2243
2244 fn build_initial_update(&self, project_id: u64, worktree_id: u64) -> proto::UpdateWorktree {
2245 let mut updated_entries = self
2246 .entries_by_path
2247 .iter()
2248 .map(proto::Entry::from)
2249 .collect::<Vec<_>>();
2250 updated_entries.sort_unstable_by_key(|e| e.id);
2251
2252 proto::UpdateWorktree {
2253 project_id,
2254 worktree_id,
2255 abs_path: self.abs_path().to_string_lossy().into_owned(),
2256 root_name: self.root_name().to_proto(),
2257 updated_entries,
2258 removed_entries: Vec::new(),
2259 scan_id: self.scan_id as u64,
2260 is_last_update: self.completed_scan_id == self.scan_id,
2261 // Sent in separate messages.
2262 updated_repositories: Vec::new(),
2263 removed_repositories: Vec::new(),
2264 }
2265 }
2266
2267 pub fn work_directory_abs_path(&self, work_directory: &WorkDirectory) -> PathBuf {
2268 match work_directory {
2269 WorkDirectory::InProject { relative_path } => self.absolutize(relative_path),
2270 WorkDirectory::AboveProject { absolute_path, .. } => absolute_path.as_ref().to_owned(),
2271 }
2272 }
2273
2274 pub fn absolutize(&self, path: &RelPath) -> PathBuf {
2275 if path.file_name().is_some() {
2276 let mut abs_path = self.abs_path.to_string();
2277 for component in path.components() {
2278 if !abs_path.ends_with(self.path_style.primary_separator()) {
2279 abs_path.push_str(self.path_style.primary_separator());
2280 }
2281 abs_path.push_str(component);
2282 }
2283 PathBuf::from(abs_path)
2284 } else {
2285 self.abs_path.as_path().to_path_buf()
2286 }
2287 }
2288
2289 pub fn contains_entry(&self, entry_id: ProjectEntryId) -> bool {
2290 self.entries_by_id.get(&entry_id, ()).is_some()
2291 }
2292
2293 fn insert_entry(
2294 &mut self,
2295 entry: proto::Entry,
2296 always_included_paths: &PathMatcher,
2297 ) -> Result<Entry> {
2298 let entry = Entry::try_from((&self.root_char_bag, always_included_paths, entry))?;
2299 let old_entry = self.entries_by_id.insert_or_replace(
2300 PathEntry {
2301 id: entry.id,
2302 path: entry.path.clone(),
2303 is_ignored: entry.is_ignored,
2304 scan_id: 0,
2305 },
2306 (),
2307 );
2308 if let Some(old_entry) = old_entry {
2309 self.entries_by_path.remove(&PathKey(old_entry.path), ());
2310 }
2311 self.entries_by_path.insert_or_replace(entry.clone(), ());
2312 Ok(entry)
2313 }
2314
2315 fn delete_entry(&mut self, entry_id: ProjectEntryId) -> Option<Arc<RelPath>> {
2316 let removed_entry = self.entries_by_id.remove(&entry_id, ())?;
2317 self.entries_by_path = {
2318 let mut cursor = self.entries_by_path.cursor::<TraversalProgress>(());
2319 let mut new_entries_by_path =
2320 cursor.slice(&TraversalTarget::path(&removed_entry.path), Bias::Left);
2321 while let Some(entry) = cursor.item() {
2322 if entry.path.starts_with(&removed_entry.path) {
2323 self.entries_by_id.remove(&entry.id, ());
2324 cursor.next();
2325 } else {
2326 break;
2327 }
2328 }
2329 new_entries_by_path.append(cursor.suffix(), ());
2330 new_entries_by_path
2331 };
2332
2333 Some(removed_entry.path)
2334 }
2335
2336 fn update_abs_path(&mut self, abs_path: Arc<SanitizedPath>, root_name: Arc<RelPath>) {
2337 self.abs_path = abs_path;
2338 if root_name != self.root_name {
2339 self.root_char_bag = root_name
2340 .as_unix_str()
2341 .chars()
2342 .map(|c| c.to_ascii_lowercase())
2343 .collect();
2344 self.root_name = root_name;
2345 }
2346 }
2347
2348 pub fn apply_remote_update(
2349 &mut self,
2350 update: proto::UpdateWorktree,
2351 always_included_paths: &PathMatcher,
2352 ) {
2353 log::debug!(
2354 "applying remote worktree update. {} entries updated, {} removed",
2355 update.updated_entries.len(),
2356 update.removed_entries.len()
2357 );
2358 if let Some(root_name) = RelPath::from_proto(&update.root_name).log_err() {
2359 self.update_abs_path(
2360 SanitizedPath::new_arc(&Path::new(&update.abs_path)),
2361 root_name,
2362 );
2363 }
2364
2365 let mut entries_by_path_edits = Vec::new();
2366 let mut entries_by_id_edits = Vec::new();
2367
2368 for entry_id in update.removed_entries {
2369 let entry_id = ProjectEntryId::from_proto(entry_id);
2370 entries_by_id_edits.push(Edit::Remove(entry_id));
2371 if let Some(entry) = self.entry_for_id(entry_id) {
2372 entries_by_path_edits.push(Edit::Remove(PathKey(entry.path.clone())));
2373 }
2374 }
2375
2376 for entry in update.updated_entries {
2377 let Some(entry) =
2378 Entry::try_from((&self.root_char_bag, always_included_paths, entry)).log_err()
2379 else {
2380 continue;
2381 };
2382 if let Some(PathEntry { path, .. }) = self.entries_by_id.get(&entry.id, ()) {
2383 entries_by_path_edits.push(Edit::Remove(PathKey(path.clone())));
2384 }
2385 if let Some(old_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2386 && old_entry.id != entry.id
2387 {
2388 entries_by_id_edits.push(Edit::Remove(old_entry.id));
2389 }
2390 entries_by_id_edits.push(Edit::Insert(PathEntry {
2391 id: entry.id,
2392 path: entry.path.clone(),
2393 is_ignored: entry.is_ignored,
2394 scan_id: 0,
2395 }));
2396 entries_by_path_edits.push(Edit::Insert(entry));
2397 }
2398
2399 self.entries_by_path.edit(entries_by_path_edits, ());
2400 self.entries_by_id.edit(entries_by_id_edits, ());
2401
2402 self.scan_id = update.scan_id as usize;
2403 if update.is_last_update {
2404 self.completed_scan_id = update.scan_id as usize;
2405 }
2406 }
2407
2408 pub fn entry_count(&self) -> usize {
2409 self.entries_by_path.summary().count
2410 }
2411
2412 pub fn visible_entry_count(&self) -> usize {
2413 self.entries_by_path.summary().non_ignored_count
2414 }
2415
2416 pub fn dir_count(&self) -> usize {
2417 let summary = self.entries_by_path.summary();
2418 summary.count - summary.file_count
2419 }
2420
2421 pub fn visible_dir_count(&self) -> usize {
2422 let summary = self.entries_by_path.summary();
2423 summary.non_ignored_count - summary.non_ignored_file_count
2424 }
2425
2426 pub fn file_count(&self) -> usize {
2427 self.entries_by_path.summary().file_count
2428 }
2429
2430 pub fn visible_file_count(&self) -> usize {
2431 self.entries_by_path.summary().non_ignored_file_count
2432 }
2433
2434 fn traverse_from_offset(
2435 &self,
2436 include_files: bool,
2437 include_dirs: bool,
2438 include_ignored: bool,
2439 start_offset: usize,
2440 ) -> Traversal<'_> {
2441 let mut cursor = self.entries_by_path.cursor(());
2442 cursor.seek(
2443 &TraversalTarget::Count {
2444 count: start_offset,
2445 include_files,
2446 include_dirs,
2447 include_ignored,
2448 },
2449 Bias::Right,
2450 );
2451 Traversal {
2452 snapshot: self,
2453 cursor,
2454 include_files,
2455 include_dirs,
2456 include_ignored,
2457 }
2458 }
2459
2460 pub fn traverse_from_path(
2461 &self,
2462 include_files: bool,
2463 include_dirs: bool,
2464 include_ignored: bool,
2465 path: &RelPath,
2466 ) -> Traversal<'_> {
2467 Traversal::new(self, include_files, include_dirs, include_ignored, path)
2468 }
2469
2470 pub fn files(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2471 self.traverse_from_offset(true, false, include_ignored, start)
2472 }
2473
2474 pub fn directories(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2475 self.traverse_from_offset(false, true, include_ignored, start)
2476 }
2477
2478 pub fn entries(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2479 self.traverse_from_offset(true, true, include_ignored, start)
2480 }
2481
2482 pub fn paths(&self) -> impl Iterator<Item = &RelPath> {
2483 self.entries_by_path
2484 .cursor::<()>(())
2485 .filter(move |entry| !entry.path.is_empty())
2486 .map(|entry| entry.path.as_ref())
2487 }
2488
2489 pub fn child_entries<'a>(&'a self, parent_path: &'a RelPath) -> ChildEntriesIter<'a> {
2490 let options = ChildEntriesOptions {
2491 include_files: true,
2492 include_dirs: true,
2493 include_ignored: true,
2494 };
2495 self.child_entries_with_options(parent_path, options)
2496 }
2497
2498 pub fn child_entries_with_options<'a>(
2499 &'a self,
2500 parent_path: &'a RelPath,
2501 options: ChildEntriesOptions,
2502 ) -> ChildEntriesIter<'a> {
2503 let mut cursor = self.entries_by_path.cursor(());
2504 cursor.seek(&TraversalTarget::path(parent_path), Bias::Right);
2505 let traversal = Traversal {
2506 snapshot: self,
2507 cursor,
2508 include_files: options.include_files,
2509 include_dirs: options.include_dirs,
2510 include_ignored: options.include_ignored,
2511 };
2512 ChildEntriesIter {
2513 traversal,
2514 parent_path,
2515 }
2516 }
2517
2518 pub fn root_entry(&self) -> Option<&Entry> {
2519 self.entries_by_path.first()
2520 }
2521
2522 /// Returns `None` for a single file worktree, or `Some(self.abs_path())` if
2523 /// it is a directory.
2524 pub fn root_dir(&self) -> Option<Arc<Path>> {
2525 self.root_entry()
2526 .filter(|entry| entry.is_dir())
2527 .map(|_| self.abs_path().clone())
2528 }
2529
2530 pub fn root_name(&self) -> &RelPath {
2531 &self.root_name
2532 }
2533
2534 pub fn root_name_str(&self) -> &str {
2535 self.root_name.as_unix_str()
2536 }
2537
2538 pub fn scan_id(&self) -> usize {
2539 self.scan_id
2540 }
2541
2542 pub fn entry_for_path(&self, path: &RelPath) -> Option<&Entry> {
2543 self.traverse_from_path(true, true, true, path)
2544 .entry()
2545 .and_then(|entry| {
2546 if entry.path.as_ref() == path {
2547 Some(entry)
2548 } else {
2549 None
2550 }
2551 })
2552 }
2553
2554 /// Resolves a path to an executable using the following heuristics:
2555 ///
2556 /// 1. If the path starts with `~`, it is expanded to the user's home directory.
2557 /// 2. If the path is relative and contains more than one component,
2558 /// it is joined to the worktree root path.
2559 /// 3. If the path is relative and exists in the worktree
2560 /// (even if falls under an exclusion filter),
2561 /// it is joined to the worktree root path.
2562 /// 4. Otherwise the path is returned unmodified.
2563 ///
2564 /// Relative paths that do not exist in the worktree may
2565 /// still be found using the `PATH` environment variable.
2566 pub fn resolve_relative_path(&self, path: PathBuf) -> PathBuf {
2567 if let Some(path_str) = path.to_str() {
2568 if let Some(remaining_path) = path_str.strip_prefix("~/") {
2569 return home_dir().join(remaining_path);
2570 } else if path_str == "~" {
2571 return home_dir().to_path_buf();
2572 }
2573 }
2574
2575 if let Ok(rel_path) = RelPath::new(&path, self.path_style)
2576 && (path.components().count() > 1 || self.entry_for_path(&rel_path).is_some())
2577 {
2578 self.abs_path().join(path)
2579 } else {
2580 path
2581 }
2582 }
2583
2584 pub fn entry_for_id(&self, id: ProjectEntryId) -> Option<&Entry> {
2585 let entry = self.entries_by_id.get(&id, ())?;
2586 self.entry_for_path(&entry.path)
2587 }
2588
2589 pub fn path_style(&self) -> PathStyle {
2590 self.path_style
2591 }
2592}
2593
2594impl LocalSnapshot {
2595 fn local_repo_for_work_directory_path(&self, path: &RelPath) -> Option<&LocalRepositoryEntry> {
2596 self.git_repositories
2597 .iter()
2598 .map(|(_, entry)| entry)
2599 .find(|entry| entry.work_directory.path_key() == PathKey(path.into()))
2600 }
2601
2602 fn build_update(
2603 &self,
2604 project_id: u64,
2605 worktree_id: u64,
2606 entry_changes: UpdatedEntriesSet,
2607 ) -> proto::UpdateWorktree {
2608 let mut updated_entries = Vec::new();
2609 let mut removed_entries = Vec::new();
2610
2611 for (_, entry_id, path_change) in entry_changes.iter() {
2612 if let PathChange::Removed = path_change {
2613 removed_entries.push(entry_id.0 as u64);
2614 } else if let Some(entry) = self.entry_for_id(*entry_id) {
2615 updated_entries.push(proto::Entry::from(entry));
2616 }
2617 }
2618
2619 removed_entries.sort_unstable();
2620 updated_entries.sort_unstable_by_key(|e| e.id);
2621
2622 // TODO - optimize, knowing that removed_entries are sorted.
2623 removed_entries.retain(|id| updated_entries.binary_search_by_key(id, |e| e.id).is_err());
2624
2625 proto::UpdateWorktree {
2626 project_id,
2627 worktree_id,
2628 abs_path: self.abs_path().to_string_lossy().into_owned(),
2629 root_name: self.root_name().to_proto(),
2630 updated_entries,
2631 removed_entries,
2632 scan_id: self.scan_id as u64,
2633 is_last_update: self.completed_scan_id == self.scan_id,
2634 // Sent in separate messages.
2635 updated_repositories: Vec::new(),
2636 removed_repositories: Vec::new(),
2637 }
2638 }
2639
2640 async fn insert_entry(&mut self, mut entry: Entry, fs: &dyn Fs) -> Entry {
2641 log::trace!("insert entry {:?}", entry.path);
2642 if entry.is_file() && entry.path.file_name() == Some(&GITIGNORE) {
2643 let abs_path = self.absolutize(&entry.path);
2644 match build_gitignore(&abs_path, fs).await {
2645 Ok(ignore) => {
2646 self.ignores_by_parent_abs_path
2647 .insert(abs_path.parent().unwrap().into(), (Arc::new(ignore), true));
2648 }
2649 Err(error) => {
2650 log::error!(
2651 "error loading .gitignore file {:?} - {:?}",
2652 &entry.path,
2653 error
2654 );
2655 }
2656 }
2657 }
2658
2659 if entry.kind == EntryKind::PendingDir
2660 && let Some(existing_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2661 {
2662 entry.kind = existing_entry.kind;
2663 }
2664
2665 let scan_id = self.scan_id;
2666 let removed = self.entries_by_path.insert_or_replace(entry.clone(), ());
2667 if let Some(removed) = removed
2668 && removed.id != entry.id
2669 {
2670 self.entries_by_id.remove(&removed.id, ());
2671 }
2672 self.entries_by_id.insert_or_replace(
2673 PathEntry {
2674 id: entry.id,
2675 path: entry.path.clone(),
2676 is_ignored: entry.is_ignored,
2677 scan_id,
2678 },
2679 (),
2680 );
2681
2682 entry
2683 }
2684
2685 fn ancestor_inodes_for_path(&self, path: &RelPath) -> TreeSet<u64> {
2686 let mut inodes = TreeSet::default();
2687 for ancestor in path.ancestors().skip(1) {
2688 if let Some(entry) = self.entry_for_path(ancestor) {
2689 inodes.insert(entry.inode);
2690 }
2691 }
2692 inodes
2693 }
2694
2695 async fn ignore_stack_for_abs_path(
2696 &self,
2697 abs_path: &Path,
2698 is_dir: bool,
2699 fs: &dyn Fs,
2700 ) -> IgnoreStack {
2701 let mut new_ignores = Vec::new();
2702 let mut repo_root = None;
2703 for (index, ancestor) in abs_path.ancestors().enumerate() {
2704 if index > 0 {
2705 if let Some((ignore, _)) = self.ignores_by_parent_abs_path.get(ancestor) {
2706 new_ignores.push((ancestor, Some(ignore.clone())));
2707 } else {
2708 new_ignores.push((ancestor, None));
2709 }
2710 }
2711
2712 let metadata = fs.metadata(&ancestor.join(DOT_GIT)).await.ok().flatten();
2713 if metadata.is_some() {
2714 repo_root = Some(Arc::from(ancestor));
2715 break;
2716 }
2717 }
2718
2719 let mut ignore_stack = if let Some(global_gitignore) = self.global_gitignore.clone() {
2720 IgnoreStack::global(global_gitignore)
2721 } else {
2722 IgnoreStack::none()
2723 };
2724
2725 if let Some((repo_exclude, _)) = repo_root
2726 .as_ref()
2727 .and_then(|abs_path| self.repo_exclude_by_work_dir_abs_path.get(abs_path))
2728 {
2729 ignore_stack = ignore_stack.append(IgnoreKind::RepoExclude, repo_exclude.clone());
2730 }
2731 ignore_stack.repo_root = repo_root;
2732 for (parent_abs_path, ignore) in new_ignores.into_iter().rev() {
2733 if ignore_stack.is_abs_path_ignored(parent_abs_path, true) {
2734 ignore_stack = IgnoreStack::all();
2735 break;
2736 } else if let Some(ignore) = ignore {
2737 ignore_stack =
2738 ignore_stack.append(IgnoreKind::Gitignore(parent_abs_path.into()), ignore);
2739 }
2740 }
2741
2742 if ignore_stack.is_abs_path_ignored(abs_path, is_dir) {
2743 ignore_stack = IgnoreStack::all();
2744 }
2745
2746 ignore_stack
2747 }
2748
2749 #[cfg(feature = "test-support")]
2750 pub fn expanded_entries(&self) -> impl Iterator<Item = &Entry> {
2751 self.entries_by_path
2752 .cursor::<()>(())
2753 .filter(|entry| entry.kind == EntryKind::Dir && (entry.is_external || entry.is_ignored))
2754 }
2755
2756 #[cfg(feature = "test-support")]
2757 pub fn check_invariants(&self, git_state: bool) {
2758 use pretty_assertions::assert_eq;
2759
2760 assert_eq!(
2761 self.entries_by_path
2762 .cursor::<()>(())
2763 .map(|e| (&e.path, e.id))
2764 .collect::<Vec<_>>(),
2765 self.entries_by_id
2766 .cursor::<()>(())
2767 .map(|e| (&e.path, e.id))
2768 .collect::<collections::BTreeSet<_>>()
2769 .into_iter()
2770 .collect::<Vec<_>>(),
2771 "entries_by_path and entries_by_id are inconsistent"
2772 );
2773
2774 let mut files = self.files(true, 0);
2775 let mut visible_files = self.files(false, 0);
2776 for entry in self.entries_by_path.cursor::<()>(()) {
2777 if entry.is_file() {
2778 assert_eq!(files.next().unwrap().inode, entry.inode);
2779 if !entry.is_ignored || entry.is_always_included {
2780 assert_eq!(visible_files.next().unwrap().inode, entry.inode);
2781 }
2782 }
2783 }
2784
2785 assert!(files.next().is_none());
2786 assert!(visible_files.next().is_none());
2787
2788 let mut bfs_paths = Vec::new();
2789 let mut stack = self
2790 .root_entry()
2791 .map(|e| e.path.as_ref())
2792 .into_iter()
2793 .collect::<Vec<_>>();
2794 while let Some(path) = stack.pop() {
2795 bfs_paths.push(path);
2796 let ix = stack.len();
2797 for child_entry in self.child_entries(path) {
2798 stack.insert(ix, &child_entry.path);
2799 }
2800 }
2801
2802 let dfs_paths_via_iter = self
2803 .entries_by_path
2804 .cursor::<()>(())
2805 .map(|e| e.path.as_ref())
2806 .collect::<Vec<_>>();
2807 assert_eq!(bfs_paths, dfs_paths_via_iter);
2808
2809 let dfs_paths_via_traversal = self
2810 .entries(true, 0)
2811 .map(|e| e.path.as_ref())
2812 .collect::<Vec<_>>();
2813
2814 assert_eq!(dfs_paths_via_traversal, dfs_paths_via_iter);
2815
2816 if git_state {
2817 for ignore_parent_abs_path in self.ignores_by_parent_abs_path.keys() {
2818 let ignore_parent_path = &RelPath::new(
2819 ignore_parent_abs_path
2820 .strip_prefix(self.abs_path.as_path())
2821 .unwrap(),
2822 PathStyle::local(),
2823 )
2824 .unwrap();
2825 assert!(self.entry_for_path(ignore_parent_path).is_some());
2826 assert!(
2827 self.entry_for_path(
2828 &ignore_parent_path.join(RelPath::unix(GITIGNORE).unwrap())
2829 )
2830 .is_some()
2831 );
2832 }
2833 }
2834 }
2835
2836 #[cfg(feature = "test-support")]
2837 pub fn entries_without_ids(&self, include_ignored: bool) -> Vec<(&RelPath, u64, bool)> {
2838 let mut paths = Vec::new();
2839 for entry in self.entries_by_path.cursor::<()>(()) {
2840 if include_ignored || !entry.is_ignored {
2841 paths.push((entry.path.as_ref(), entry.inode, entry.is_ignored));
2842 }
2843 }
2844 paths.sort_by(|a, b| a.0.cmp(b.0));
2845 paths
2846 }
2847}
2848
2849impl BackgroundScannerState {
2850 fn should_scan_directory(&self, entry: &Entry) -> bool {
2851 (self.scanning_enabled && !entry.is_external && (!entry.is_ignored || entry.is_always_included))
2852 || entry.path.file_name() == Some(DOT_GIT)
2853 || entry.path.file_name() == Some(local_settings_folder_name())
2854 || entry.path.file_name() == Some(local_vscode_folder_name())
2855 || self.scanned_dirs.contains(&entry.id) // If we've ever scanned it, keep scanning
2856 || self
2857 .paths_to_scan
2858 .iter()
2859 .any(|p| p.starts_with(&entry.path))
2860 || self
2861 .path_prefixes_to_scan
2862 .iter()
2863 .any(|p| entry.path.starts_with(p))
2864 }
2865
2866 async fn enqueue_scan_dir(
2867 &self,
2868 abs_path: Arc<Path>,
2869 entry: &Entry,
2870 scan_job_tx: &Sender<ScanJob>,
2871 fs: &dyn Fs,
2872 ) {
2873 let path = entry.path.clone();
2874 let ignore_stack = self
2875 .snapshot
2876 .ignore_stack_for_abs_path(&abs_path, true, fs)
2877 .await;
2878 let mut ancestor_inodes = self.snapshot.ancestor_inodes_for_path(&path);
2879
2880 if !ancestor_inodes.contains(&entry.inode) {
2881 ancestor_inodes.insert(entry.inode);
2882 scan_job_tx
2883 .try_send(ScanJob {
2884 abs_path,
2885 path,
2886 ignore_stack,
2887 scan_queue: scan_job_tx.clone(),
2888 ancestor_inodes,
2889 is_external: entry.is_external,
2890 })
2891 .unwrap();
2892 }
2893 }
2894
2895 fn reuse_entry_id(&mut self, entry: &mut Entry) {
2896 if let Some(mtime) = entry.mtime {
2897 // If an entry with the same inode was removed from the worktree during this scan,
2898 // then it *might* represent the same file or directory. But the OS might also have
2899 // re-used the inode for a completely different file or directory.
2900 //
2901 // Conditionally reuse the old entry's id:
2902 // * if the mtime is the same, the file was probably been renamed.
2903 // * if the path is the same, the file may just have been updated
2904 if let Some(removed_entry) = self.removed_entries.remove(&entry.inode) {
2905 if removed_entry.mtime == Some(mtime) || removed_entry.path == entry.path {
2906 entry.id = removed_entry.id;
2907 }
2908 } else if let Some(existing_entry) = self.snapshot.entry_for_path(&entry.path) {
2909 entry.id = existing_entry.id;
2910 }
2911 }
2912 }
2913
2914 fn entry_id_for(
2915 &mut self,
2916 next_entry_id: &AtomicUsize,
2917 path: &RelPath,
2918 metadata: &fs::Metadata,
2919 ) -> ProjectEntryId {
2920 // If an entry with the same inode was removed from the worktree during this scan,
2921 // then it *might* represent the same file or directory. But the OS might also have
2922 // re-used the inode for a completely different file or directory.
2923 //
2924 // Conditionally reuse the old entry's id:
2925 // * if the mtime is the same, the file was probably been renamed.
2926 // * if the path is the same, the file may just have been updated
2927 if let Some(removed_entry) = self.removed_entries.remove(&metadata.inode) {
2928 if removed_entry.mtime == Some(metadata.mtime) || *removed_entry.path == *path {
2929 return removed_entry.id;
2930 }
2931 } else if let Some(existing_entry) = self.snapshot.entry_for_path(path) {
2932 return existing_entry.id;
2933 }
2934 ProjectEntryId::new(next_entry_id)
2935 }
2936
2937 async fn insert_entry(&mut self, entry: Entry, fs: &dyn Fs, watcher: &dyn Watcher) -> Entry {
2938 let entry = self.snapshot.insert_entry(entry, fs).await;
2939 if entry.path.file_name() == Some(&DOT_GIT) {
2940 self.insert_git_repository(entry.path.clone(), fs, watcher)
2941 .await;
2942 }
2943
2944 #[cfg(feature = "test-support")]
2945 self.snapshot.check_invariants(false);
2946
2947 entry
2948 }
2949
2950 fn populate_dir(
2951 &mut self,
2952 parent_path: Arc<RelPath>,
2953 entries: impl IntoIterator<Item = Entry>,
2954 ignore: Option<Arc<Gitignore>>,
2955 ) {
2956 let mut parent_entry = if let Some(parent_entry) = self
2957 .snapshot
2958 .entries_by_path
2959 .get(&PathKey(parent_path.clone()), ())
2960 {
2961 parent_entry.clone()
2962 } else {
2963 log::warn!(
2964 "populating a directory {:?} that has been removed",
2965 parent_path
2966 );
2967 return;
2968 };
2969
2970 match parent_entry.kind {
2971 EntryKind::PendingDir | EntryKind::UnloadedDir => parent_entry.kind = EntryKind::Dir,
2972 EntryKind::Dir => {}
2973 _ => return,
2974 }
2975
2976 if let Some(ignore) = ignore {
2977 let abs_parent_path = self
2978 .snapshot
2979 .abs_path
2980 .as_path()
2981 .join(parent_path.as_std_path())
2982 .into();
2983 self.snapshot
2984 .ignores_by_parent_abs_path
2985 .insert(abs_parent_path, (ignore, false));
2986 }
2987
2988 let parent_entry_id = parent_entry.id;
2989 self.scanned_dirs.insert(parent_entry_id);
2990 let mut entries_by_path_edits = vec![Edit::Insert(parent_entry)];
2991 let mut entries_by_id_edits = Vec::new();
2992
2993 for entry in entries {
2994 entries_by_id_edits.push(Edit::Insert(PathEntry {
2995 id: entry.id,
2996 path: entry.path.clone(),
2997 is_ignored: entry.is_ignored,
2998 scan_id: self.snapshot.scan_id,
2999 }));
3000 entries_by_path_edits.push(Edit::Insert(entry));
3001 }
3002
3003 self.snapshot
3004 .entries_by_path
3005 .edit(entries_by_path_edits, ());
3006 self.snapshot.entries_by_id.edit(entries_by_id_edits, ());
3007
3008 if let Err(ix) = self.changed_paths.binary_search(&parent_path) {
3009 self.changed_paths.insert(ix, parent_path.clone());
3010 }
3011
3012 #[cfg(feature = "test-support")]
3013 self.snapshot.check_invariants(false);
3014 }
3015
3016 fn remove_path(&mut self, path: &RelPath, watcher: &dyn Watcher) {
3017 log::trace!("background scanner removing path {path:?}");
3018 let mut new_entries;
3019 let removed_entries;
3020 {
3021 let mut cursor = self
3022 .snapshot
3023 .entries_by_path
3024 .cursor::<TraversalProgress>(());
3025 new_entries = cursor.slice(&TraversalTarget::path(path), Bias::Left);
3026 removed_entries = cursor.slice(&TraversalTarget::successor(path), Bias::Left);
3027 new_entries.append(cursor.suffix(), ());
3028 }
3029 self.snapshot.entries_by_path = new_entries;
3030
3031 let mut removed_ids = Vec::with_capacity(removed_entries.summary().count);
3032 let mut removed_dir_abs_paths = Vec::new();
3033 for entry in removed_entries.cursor::<()>(()) {
3034 if entry.is_dir() {
3035 removed_dir_abs_paths.push(self.snapshot.absolutize(&entry.path));
3036 }
3037
3038 match self.removed_entries.entry(entry.inode) {
3039 hash_map::Entry::Occupied(mut e) => {
3040 let prev_removed_entry = e.get_mut();
3041 if entry.id > prev_removed_entry.id {
3042 *prev_removed_entry = entry.clone();
3043 }
3044 }
3045 hash_map::Entry::Vacant(e) => {
3046 e.insert(entry.clone());
3047 }
3048 }
3049
3050 if entry.path.file_name() == Some(GITIGNORE) {
3051 let abs_parent_path = self.snapshot.absolutize(&entry.path.parent().unwrap());
3052 if let Some((_, needs_update)) = self
3053 .snapshot
3054 .ignores_by_parent_abs_path
3055 .get_mut(abs_parent_path.as_path())
3056 {
3057 *needs_update = true;
3058 }
3059 }
3060
3061 if let Err(ix) = removed_ids.binary_search(&entry.id) {
3062 removed_ids.insert(ix, entry.id);
3063 }
3064 }
3065
3066 self.snapshot
3067 .entries_by_id
3068 .edit(removed_ids.iter().map(|&id| Edit::Remove(id)).collect(), ());
3069 self.snapshot
3070 .git_repositories
3071 .retain(|id, _| removed_ids.binary_search(id).is_err());
3072
3073 for removed_dir_abs_path in removed_dir_abs_paths {
3074 watcher.remove(&removed_dir_abs_path).log_err();
3075 }
3076
3077 #[cfg(feature = "test-support")]
3078 self.snapshot.check_invariants(false);
3079 }
3080
3081 async fn insert_git_repository(
3082 &mut self,
3083 dot_git_path: Arc<RelPath>,
3084 fs: &dyn Fs,
3085 watcher: &dyn Watcher,
3086 ) {
3087 let work_dir_path: Arc<RelPath> = match dot_git_path.parent() {
3088 Some(parent_dir) => {
3089 // Guard against repositories inside the repository metadata
3090 if parent_dir
3091 .components()
3092 .any(|component| component == DOT_GIT)
3093 {
3094 log::debug!(
3095 "not building git repository for nested `.git` directory, `.git` path in the worktree: {dot_git_path:?}"
3096 );
3097 return;
3098 };
3099
3100 parent_dir.into()
3101 }
3102 None => {
3103 // `dot_git_path.parent().is_none()` means `.git` directory is the opened worktree itself,
3104 // no files inside that directory are tracked by git, so no need to build the repo around it
3105 log::debug!(
3106 "not building git repository for the worktree itself, `.git` path in the worktree: {dot_git_path:?}"
3107 );
3108 return;
3109 }
3110 };
3111
3112 let dot_git_abs_path = Arc::from(self.snapshot.absolutize(&dot_git_path).as_ref());
3113
3114 self.insert_git_repository_for_path(
3115 WorkDirectory::InProject {
3116 relative_path: work_dir_path,
3117 },
3118 dot_git_abs_path,
3119 fs,
3120 watcher,
3121 )
3122 .await
3123 .log_err();
3124 }
3125
3126 async fn insert_git_repository_for_path(
3127 &mut self,
3128 work_directory: WorkDirectory,
3129 dot_git_abs_path: Arc<Path>,
3130 fs: &dyn Fs,
3131 watcher: &dyn Watcher,
3132 ) -> Result<LocalRepositoryEntry> {
3133 let work_dir_entry = self
3134 .snapshot
3135 .entry_for_path(&work_directory.path_key().0)
3136 .with_context(|| {
3137 format!(
3138 "working directory `{}` not indexed",
3139 work_directory
3140 .path_key()
3141 .0
3142 .display(self.snapshot.path_style)
3143 )
3144 })?;
3145 let work_directory_abs_path = self.snapshot.work_directory_abs_path(&work_directory);
3146
3147 let (repository_dir_abs_path, common_dir_abs_path) =
3148 discover_git_paths(&dot_git_abs_path, fs).await;
3149 watcher
3150 .add(&common_dir_abs_path)
3151 .context("failed to add common directory to watcher")
3152 .log_err();
3153 watcher
3154 .add(&repository_dir_abs_path)
3155 .context("failed to add repository directory to watcher")
3156 .log_err();
3157
3158 let work_directory_id = work_dir_entry.id;
3159
3160 let local_repository = LocalRepositoryEntry {
3161 work_directory_id,
3162 work_directory,
3163 work_directory_abs_path: work_directory_abs_path.as_path().into(),
3164 git_dir_scan_id: 0,
3165 dot_git_abs_path,
3166 common_dir_abs_path,
3167 repository_dir_abs_path,
3168 };
3169
3170 self.snapshot
3171 .git_repositories
3172 .insert(work_directory_id, local_repository.clone());
3173
3174 log::trace!("inserting new local git repository");
3175 Ok(local_repository)
3176 }
3177}
3178
3179async fn is_git_dir(path: &Path, fs: &dyn Fs) -> bool {
3180 if let Some(file_name) = path.file_name()
3181 && file_name == DOT_GIT
3182 {
3183 return true;
3184 }
3185
3186 // If we're in a bare repository, we are not inside a `.git` folder. In a
3187 // bare repository, the root folder contains what would normally be in the
3188 // `.git` folder.
3189 let head_metadata = fs.metadata(&path.join("HEAD")).await;
3190 if !matches!(head_metadata, Ok(Some(_))) {
3191 return false;
3192 }
3193 let config_metadata = fs.metadata(&path.join("config")).await;
3194 matches!(config_metadata, Ok(Some(_)))
3195}
3196
3197async fn build_gitignore(abs_path: &Path, fs: &dyn Fs) -> Result<Gitignore> {
3198 let contents = fs
3199 .load(abs_path)
3200 .await
3201 .with_context(|| format!("failed to load gitignore file at {}", abs_path.display()))?;
3202 let parent = abs_path.parent().unwrap_or_else(|| Path::new("/"));
3203 let mut builder = GitignoreBuilder::new(parent);
3204 for line in contents.lines() {
3205 builder.add_line(Some(abs_path.into()), line)?;
3206 }
3207 Ok(builder.build()?)
3208}
3209
3210impl Deref for Worktree {
3211 type Target = Snapshot;
3212
3213 fn deref(&self) -> &Self::Target {
3214 match self {
3215 Worktree::Local(worktree) => &worktree.snapshot,
3216 Worktree::Remote(worktree) => &worktree.snapshot,
3217 }
3218 }
3219}
3220
3221impl Deref for LocalWorktree {
3222 type Target = LocalSnapshot;
3223
3224 fn deref(&self) -> &Self::Target {
3225 &self.snapshot
3226 }
3227}
3228
3229impl Deref for RemoteWorktree {
3230 type Target = Snapshot;
3231
3232 fn deref(&self) -> &Self::Target {
3233 &self.snapshot
3234 }
3235}
3236
3237impl fmt::Debug for LocalWorktree {
3238 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3239 self.snapshot.fmt(f)
3240 }
3241}
3242
3243impl fmt::Debug for Snapshot {
3244 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3245 struct EntriesById<'a>(&'a SumTree<PathEntry>);
3246 struct EntriesByPath<'a>(&'a SumTree<Entry>);
3247
3248 impl fmt::Debug for EntriesByPath<'_> {
3249 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3250 f.debug_map()
3251 .entries(self.0.iter().map(|entry| (&entry.path, entry.id)))
3252 .finish()
3253 }
3254 }
3255
3256 impl fmt::Debug for EntriesById<'_> {
3257 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3258 f.debug_list().entries(self.0.iter()).finish()
3259 }
3260 }
3261
3262 f.debug_struct("Snapshot")
3263 .field("id", &self.id)
3264 .field("root_name", &self.root_name)
3265 .field("entries_by_path", &EntriesByPath(&self.entries_by_path))
3266 .field("entries_by_id", &EntriesById(&self.entries_by_id))
3267 .finish()
3268 }
3269}
3270
3271#[derive(Debug, Clone, PartialEq)]
3272pub struct File {
3273 pub worktree: Entity<Worktree>,
3274 pub path: Arc<RelPath>,
3275 pub disk_state: DiskState,
3276 pub entry_id: Option<ProjectEntryId>,
3277 pub is_local: bool,
3278 pub is_private: bool,
3279}
3280
3281impl language::File for File {
3282 fn as_local(&self) -> Option<&dyn language::LocalFile> {
3283 if self.is_local { Some(self) } else { None }
3284 }
3285
3286 fn disk_state(&self) -> DiskState {
3287 self.disk_state
3288 }
3289
3290 fn path(&self) -> &Arc<RelPath> {
3291 &self.path
3292 }
3293
3294 fn full_path(&self, cx: &App) -> PathBuf {
3295 self.worktree.read(cx).full_path(&self.path)
3296 }
3297
3298 /// Returns the last component of this handle's absolute path. If this handle refers to the root
3299 /// of its worktree, then this method will return the name of the worktree itself.
3300 fn file_name<'a>(&'a self, cx: &'a App) -> &'a str {
3301 self.path
3302 .file_name()
3303 .unwrap_or_else(|| self.worktree.read(cx).root_name_str())
3304 }
3305
3306 fn worktree_id(&self, cx: &App) -> WorktreeId {
3307 self.worktree.read(cx).id()
3308 }
3309
3310 fn to_proto(&self, cx: &App) -> rpc::proto::File {
3311 rpc::proto::File {
3312 worktree_id: self.worktree.read(cx).id().to_proto(),
3313 entry_id: self.entry_id.map(|id| id.to_proto()),
3314 path: self.path.as_ref().to_proto(),
3315 mtime: self.disk_state.mtime().map(|time| time.into()),
3316 is_deleted: self.disk_state.is_deleted(),
3317 is_historic: matches!(self.disk_state, DiskState::Historic { .. }),
3318 }
3319 }
3320
3321 fn is_private(&self) -> bool {
3322 self.is_private
3323 }
3324
3325 fn path_style(&self, cx: &App) -> PathStyle {
3326 self.worktree.read(cx).path_style()
3327 }
3328
3329 fn can_open(&self) -> bool {
3330 true
3331 }
3332}
3333
3334impl language::LocalFile for File {
3335 fn abs_path(&self, cx: &App) -> PathBuf {
3336 self.worktree.read(cx).absolutize(&self.path)
3337 }
3338
3339 fn load(&self, cx: &App) -> Task<Result<String>> {
3340 let worktree = self.worktree.read(cx).as_local().unwrap();
3341 let abs_path = worktree.absolutize(&self.path);
3342 let fs = worktree.fs.clone();
3343 cx.background_spawn(async move { fs.load(&abs_path).await })
3344 }
3345
3346 fn load_bytes(&self, cx: &App) -> Task<Result<Vec<u8>>> {
3347 let worktree = self.worktree.read(cx).as_local().unwrap();
3348 let abs_path = worktree.absolutize(&self.path);
3349 let fs = worktree.fs.clone();
3350 cx.background_spawn(async move { fs.load_bytes(&abs_path).await })
3351 }
3352}
3353
3354impl File {
3355 pub fn for_entry(entry: Entry, worktree: Entity<Worktree>) -> Arc<Self> {
3356 Arc::new(Self {
3357 worktree,
3358 path: entry.path.clone(),
3359 disk_state: if let Some(mtime) = entry.mtime {
3360 DiskState::Present {
3361 mtime,
3362 size: entry.size,
3363 }
3364 } else {
3365 DiskState::New
3366 },
3367 entry_id: Some(entry.id),
3368 is_local: true,
3369 is_private: entry.is_private,
3370 })
3371 }
3372
3373 pub fn from_proto(
3374 proto: rpc::proto::File,
3375 worktree: Entity<Worktree>,
3376 cx: &App,
3377 ) -> Result<Self> {
3378 let worktree_id = worktree.read(cx).as_remote().context("not remote")?.id();
3379
3380 anyhow::ensure!(
3381 worktree_id.to_proto() == proto.worktree_id,
3382 "worktree id does not match file"
3383 );
3384
3385 let disk_state = if proto.is_historic {
3386 DiskState::Historic {
3387 was_deleted: proto.is_deleted,
3388 }
3389 } else if proto.is_deleted {
3390 DiskState::Deleted
3391 } else if let Some(mtime) = proto.mtime.map(&Into::into) {
3392 DiskState::Present { mtime, size: 0 }
3393 } else {
3394 DiskState::New
3395 };
3396
3397 Ok(Self {
3398 worktree,
3399 path: RelPath::from_proto(&proto.path).context("invalid path in file protobuf")?,
3400 disk_state,
3401 entry_id: proto.entry_id.map(ProjectEntryId::from_proto),
3402 is_local: false,
3403 is_private: false,
3404 })
3405 }
3406
3407 pub fn from_dyn(file: Option<&Arc<dyn language::File>>) -> Option<&Self> {
3408 file.and_then(|f| {
3409 let f: &dyn language::File = f.borrow();
3410 let f: &dyn Any = f;
3411 f.downcast_ref()
3412 })
3413 }
3414
3415 pub fn worktree_id(&self, cx: &App) -> WorktreeId {
3416 self.worktree.read(cx).id()
3417 }
3418
3419 pub fn project_entry_id(&self) -> Option<ProjectEntryId> {
3420 match self.disk_state {
3421 DiskState::Deleted => None,
3422 _ => self.entry_id,
3423 }
3424 }
3425}
3426
3427#[derive(Clone, Debug, PartialEq, Eq)]
3428pub struct Entry {
3429 pub id: ProjectEntryId,
3430 pub kind: EntryKind,
3431 pub path: Arc<RelPath>,
3432 pub inode: u64,
3433 pub mtime: Option<MTime>,
3434
3435 pub canonical_path: Option<Arc<Path>>,
3436 /// Whether this entry is ignored by Git.
3437 ///
3438 /// We only scan ignored entries once the directory is expanded and
3439 /// exclude them from searches.
3440 pub is_ignored: bool,
3441
3442 /// Whether this entry is hidden or inside hidden directory.
3443 ///
3444 /// We only scan hidden entries once the directory is expanded.
3445 pub is_hidden: bool,
3446
3447 /// Whether this entry is always included in searches.
3448 ///
3449 /// This is used for entries that are always included in searches, even
3450 /// if they are ignored by git. Overridden by file_scan_exclusions.
3451 pub is_always_included: bool,
3452
3453 /// Whether this entry's canonical path is outside of the worktree.
3454 /// This means the entry is only accessible from the worktree root via a
3455 /// symlink.
3456 ///
3457 /// We only scan entries outside of the worktree once the symlinked
3458 /// directory is expanded.
3459 pub is_external: bool,
3460
3461 /// Whether this entry is considered to be a `.env` file.
3462 pub is_private: bool,
3463 /// The entry's size on disk, in bytes.
3464 pub size: u64,
3465 pub char_bag: CharBag,
3466 pub is_fifo: bool,
3467}
3468
3469#[derive(Clone, Copy, Debug, PartialEq, Eq)]
3470pub enum EntryKind {
3471 UnloadedDir,
3472 PendingDir,
3473 Dir,
3474 File,
3475}
3476
3477#[derive(Clone, Copy, Debug, PartialEq)]
3478pub enum PathChange {
3479 /// A filesystem entry was was created.
3480 Added,
3481 /// A filesystem entry was removed.
3482 Removed,
3483 /// A filesystem entry was updated.
3484 Updated,
3485 /// A filesystem entry was either updated or added. We don't know
3486 /// whether or not it already existed, because the path had not
3487 /// been loaded before the event.
3488 AddedOrUpdated,
3489 /// A filesystem entry was found during the initial scan of the worktree.
3490 Loaded,
3491}
3492
3493#[derive(Clone, Debug, PartialEq, Eq)]
3494pub struct UpdatedGitRepository {
3495 /// ID of the repository's working directory.
3496 ///
3497 /// For a repo that's above the worktree root, this is the ID of the worktree root, and hence not unique.
3498 /// It's included here to aid the GitStore in detecting when a repository's working directory is renamed.
3499 pub work_directory_id: ProjectEntryId,
3500 pub old_work_directory_abs_path: Option<Arc<Path>>,
3501 pub new_work_directory_abs_path: Option<Arc<Path>>,
3502 /// For a normal git repository checkout, the absolute path to the .git directory.
3503 /// For a worktree, the absolute path to the worktree's subdirectory inside the .git directory.
3504 pub dot_git_abs_path: Option<Arc<Path>>,
3505 pub repository_dir_abs_path: Option<Arc<Path>>,
3506 pub common_dir_abs_path: Option<Arc<Path>>,
3507}
3508
3509pub type UpdatedEntriesSet = Arc<[(Arc<RelPath>, ProjectEntryId, PathChange)]>;
3510pub type UpdatedGitRepositoriesSet = Arc<[UpdatedGitRepository]>;
3511
3512#[derive(Clone, Debug)]
3513pub struct PathProgress<'a> {
3514 pub max_path: &'a RelPath,
3515}
3516
3517#[derive(Clone, Debug)]
3518pub struct PathSummary<S> {
3519 pub max_path: Arc<RelPath>,
3520 pub item_summary: S,
3521}
3522
3523impl<S: Summary> Summary for PathSummary<S> {
3524 type Context<'a> = S::Context<'a>;
3525
3526 fn zero(cx: Self::Context<'_>) -> Self {
3527 Self {
3528 max_path: RelPath::empty().into(),
3529 item_summary: S::zero(cx),
3530 }
3531 }
3532
3533 fn add_summary(&mut self, rhs: &Self, cx: Self::Context<'_>) {
3534 self.max_path = rhs.max_path.clone();
3535 self.item_summary.add_summary(&rhs.item_summary, cx);
3536 }
3537}
3538
3539impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathProgress<'a> {
3540 fn zero(_: <PathSummary<S> as Summary>::Context<'_>) -> Self {
3541 Self {
3542 max_path: RelPath::empty(),
3543 }
3544 }
3545
3546 fn add_summary(
3547 &mut self,
3548 summary: &'a PathSummary<S>,
3549 _: <PathSummary<S> as Summary>::Context<'_>,
3550 ) {
3551 self.max_path = summary.max_path.as_ref()
3552 }
3553}
3554
3555impl<'a> sum_tree::Dimension<'a, PathSummary<GitSummary>> for GitSummary {
3556 fn zero(_cx: ()) -> Self {
3557 Default::default()
3558 }
3559
3560 fn add_summary(&mut self, summary: &'a PathSummary<GitSummary>, _: ()) {
3561 *self += summary.item_summary
3562 }
3563}
3564
3565impl<'a>
3566 sum_tree::SeekTarget<'a, PathSummary<GitSummary>, Dimensions<TraversalProgress<'a>, GitSummary>>
3567 for PathTarget<'_>
3568{
3569 fn cmp(
3570 &self,
3571 cursor_location: &Dimensions<TraversalProgress<'a>, GitSummary>,
3572 _: (),
3573 ) -> Ordering {
3574 self.cmp_path(cursor_location.0.max_path)
3575 }
3576}
3577
3578impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathKey {
3579 fn zero(_: S::Context<'_>) -> Self {
3580 Default::default()
3581 }
3582
3583 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3584 self.0 = summary.max_path.clone();
3585 }
3586}
3587
3588impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for TraversalProgress<'a> {
3589 fn zero(_cx: S::Context<'_>) -> Self {
3590 Default::default()
3591 }
3592
3593 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3594 self.max_path = summary.max_path.as_ref();
3595 }
3596}
3597
3598impl Entry {
3599 fn new(
3600 path: Arc<RelPath>,
3601 metadata: &fs::Metadata,
3602 id: ProjectEntryId,
3603 root_char_bag: CharBag,
3604 canonical_path: Option<Arc<Path>>,
3605 ) -> Self {
3606 let char_bag = char_bag_for_path(root_char_bag, &path);
3607 Self {
3608 id,
3609 kind: if metadata.is_dir {
3610 EntryKind::PendingDir
3611 } else {
3612 EntryKind::File
3613 },
3614 path,
3615 inode: metadata.inode,
3616 mtime: Some(metadata.mtime),
3617 size: metadata.len,
3618 canonical_path,
3619 is_ignored: false,
3620 is_hidden: false,
3621 is_always_included: false,
3622 is_external: false,
3623 is_private: false,
3624 char_bag,
3625 is_fifo: metadata.is_fifo,
3626 }
3627 }
3628
3629 pub fn is_created(&self) -> bool {
3630 self.mtime.is_some()
3631 }
3632
3633 pub fn is_dir(&self) -> bool {
3634 self.kind.is_dir()
3635 }
3636
3637 pub fn is_file(&self) -> bool {
3638 self.kind.is_file()
3639 }
3640}
3641
3642impl EntryKind {
3643 pub fn is_dir(&self) -> bool {
3644 matches!(
3645 self,
3646 EntryKind::Dir | EntryKind::PendingDir | EntryKind::UnloadedDir
3647 )
3648 }
3649
3650 pub fn is_unloaded(&self) -> bool {
3651 matches!(self, EntryKind::UnloadedDir)
3652 }
3653
3654 pub fn is_file(&self) -> bool {
3655 matches!(self, EntryKind::File)
3656 }
3657}
3658
3659impl sum_tree::Item for Entry {
3660 type Summary = EntrySummary;
3661
3662 fn summary(&self, _cx: ()) -> Self::Summary {
3663 let non_ignored_count = if self.is_ignored && !self.is_always_included {
3664 0
3665 } else {
3666 1
3667 };
3668 let file_count;
3669 let non_ignored_file_count;
3670 if self.is_file() {
3671 file_count = 1;
3672 non_ignored_file_count = non_ignored_count;
3673 } else {
3674 file_count = 0;
3675 non_ignored_file_count = 0;
3676 }
3677
3678 EntrySummary {
3679 max_path: self.path.clone(),
3680 count: 1,
3681 non_ignored_count,
3682 file_count,
3683 non_ignored_file_count,
3684 }
3685 }
3686}
3687
3688impl sum_tree::KeyedItem for Entry {
3689 type Key = PathKey;
3690
3691 fn key(&self) -> Self::Key {
3692 PathKey(self.path.clone())
3693 }
3694}
3695
3696#[derive(Clone, Debug)]
3697pub struct EntrySummary {
3698 max_path: Arc<RelPath>,
3699 count: usize,
3700 non_ignored_count: usize,
3701 file_count: usize,
3702 non_ignored_file_count: usize,
3703}
3704
3705impl Default for EntrySummary {
3706 fn default() -> Self {
3707 Self {
3708 max_path: Arc::from(RelPath::empty()),
3709 count: 0,
3710 non_ignored_count: 0,
3711 file_count: 0,
3712 non_ignored_file_count: 0,
3713 }
3714 }
3715}
3716
3717impl sum_tree::ContextLessSummary for EntrySummary {
3718 fn zero() -> Self {
3719 Default::default()
3720 }
3721
3722 fn add_summary(&mut self, rhs: &Self) {
3723 self.max_path = rhs.max_path.clone();
3724 self.count += rhs.count;
3725 self.non_ignored_count += rhs.non_ignored_count;
3726 self.file_count += rhs.file_count;
3727 self.non_ignored_file_count += rhs.non_ignored_file_count;
3728 }
3729}
3730
3731#[derive(Clone, Debug)]
3732struct PathEntry {
3733 id: ProjectEntryId,
3734 path: Arc<RelPath>,
3735 is_ignored: bool,
3736 scan_id: usize,
3737}
3738
3739impl sum_tree::Item for PathEntry {
3740 type Summary = PathEntrySummary;
3741
3742 fn summary(&self, _cx: ()) -> Self::Summary {
3743 PathEntrySummary { max_id: self.id }
3744 }
3745}
3746
3747impl sum_tree::KeyedItem for PathEntry {
3748 type Key = ProjectEntryId;
3749
3750 fn key(&self) -> Self::Key {
3751 self.id
3752 }
3753}
3754
3755#[derive(Clone, Debug, Default)]
3756struct PathEntrySummary {
3757 max_id: ProjectEntryId,
3758}
3759
3760impl sum_tree::ContextLessSummary for PathEntrySummary {
3761 fn zero() -> Self {
3762 Default::default()
3763 }
3764
3765 fn add_summary(&mut self, summary: &Self) {
3766 self.max_id = summary.max_id;
3767 }
3768}
3769
3770impl<'a> sum_tree::Dimension<'a, PathEntrySummary> for ProjectEntryId {
3771 fn zero(_cx: ()) -> Self {
3772 Default::default()
3773 }
3774
3775 fn add_summary(&mut self, summary: &'a PathEntrySummary, _: ()) {
3776 *self = summary.max_id;
3777 }
3778}
3779
3780#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
3781pub struct PathKey(pub Arc<RelPath>);
3782
3783impl Default for PathKey {
3784 fn default() -> Self {
3785 Self(RelPath::empty().into())
3786 }
3787}
3788
3789impl<'a> sum_tree::Dimension<'a, EntrySummary> for PathKey {
3790 fn zero(_cx: ()) -> Self {
3791 Default::default()
3792 }
3793
3794 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
3795 self.0 = summary.max_path.clone();
3796 }
3797}
3798
3799struct BackgroundScanner {
3800 state: async_lock::Mutex<BackgroundScannerState>,
3801 fs: Arc<dyn Fs>,
3802 fs_case_sensitive: bool,
3803 status_updates_tx: UnboundedSender<ScanState>,
3804 executor: BackgroundExecutor,
3805 scan_requests_rx: channel::Receiver<ScanRequest>,
3806 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
3807 next_entry_id: Arc<AtomicUsize>,
3808 phase: BackgroundScannerPhase,
3809 watcher: Arc<dyn Watcher>,
3810 settings: WorktreeSettings,
3811 share_private_files: bool,
3812 /// Whether this is a single-file worktree (root is a file, not a directory).
3813 /// Used to determine if we should give up after repeated canonicalization failures.
3814 is_single_file: bool,
3815}
3816
3817#[derive(Copy, Clone, PartialEq)]
3818enum BackgroundScannerPhase {
3819 InitialScan,
3820 EventsReceivedDuringInitialScan,
3821 Events,
3822}
3823
3824impl BackgroundScanner {
3825 async fn run(&mut self, mut fs_events_rx: Pin<Box<dyn Send + Stream<Item = Vec<PathEvent>>>>) {
3826 let root_abs_path;
3827 let scanning_enabled;
3828 {
3829 let state = self.state.lock().await;
3830 root_abs_path = state.snapshot.abs_path.clone();
3831 scanning_enabled = state.scanning_enabled;
3832 }
3833
3834 // If the worktree root does not contain a git repository, then find
3835 // the git repository in an ancestor directory. Find any gitignore files
3836 // in ancestor directories.
3837 let repo = if scanning_enabled {
3838 let (ignores, exclude, repo) =
3839 discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await;
3840 self.state
3841 .lock()
3842 .await
3843 .snapshot
3844 .ignores_by_parent_abs_path
3845 .extend(ignores);
3846 if let Some(exclude) = exclude {
3847 self.state
3848 .lock()
3849 .await
3850 .snapshot
3851 .repo_exclude_by_work_dir_abs_path
3852 .insert(root_abs_path.as_path().into(), (exclude, false));
3853 }
3854
3855 repo
3856 } else {
3857 None
3858 };
3859
3860 let containing_git_repository = if let Some((ancestor_dot_git, work_directory)) = repo
3861 && scanning_enabled
3862 {
3863 maybe!(async {
3864 self.state
3865 .lock()
3866 .await
3867 .insert_git_repository_for_path(
3868 work_directory,
3869 ancestor_dot_git.clone().into(),
3870 self.fs.as_ref(),
3871 self.watcher.as_ref(),
3872 )
3873 .await
3874 .log_err()?;
3875 Some(ancestor_dot_git)
3876 })
3877 .await
3878 } else {
3879 None
3880 };
3881
3882 log::trace!("containing git repository: {containing_git_repository:?}");
3883
3884 let global_gitignore_file = paths::global_gitignore_path();
3885 let mut global_gitignore_events = if let Some(global_gitignore_path) =
3886 &global_gitignore_file
3887 && scanning_enabled
3888 {
3889 let is_file = self.fs.is_file(&global_gitignore_path).await;
3890 self.state.lock().await.snapshot.global_gitignore = if is_file {
3891 build_gitignore(global_gitignore_path, self.fs.as_ref())
3892 .await
3893 .ok()
3894 .map(Arc::new)
3895 } else {
3896 None
3897 };
3898 if is_file {
3899 self.fs
3900 .watch(global_gitignore_path, FS_WATCH_LATENCY)
3901 .await
3902 .0
3903 } else {
3904 Box::pin(futures::stream::pending())
3905 }
3906 } else {
3907 self.state.lock().await.snapshot.global_gitignore = None;
3908 Box::pin(futures::stream::pending())
3909 };
3910
3911 let (scan_job_tx, scan_job_rx) = channel::unbounded();
3912 {
3913 let mut state = self.state.lock().await;
3914 state.snapshot.scan_id += 1;
3915 if let Some(mut root_entry) = state.snapshot.root_entry().cloned() {
3916 let ignore_stack = state
3917 .snapshot
3918 .ignore_stack_for_abs_path(root_abs_path.as_path(), true, self.fs.as_ref())
3919 .await;
3920 if ignore_stack.is_abs_path_ignored(root_abs_path.as_path(), true) {
3921 root_entry.is_ignored = true;
3922 let mut root_entry = root_entry.clone();
3923 state.reuse_entry_id(&mut root_entry);
3924 state
3925 .insert_entry(root_entry, self.fs.as_ref(), self.watcher.as_ref())
3926 .await;
3927 }
3928 if root_entry.is_dir() && state.scanning_enabled {
3929 state
3930 .enqueue_scan_dir(
3931 root_abs_path.as_path().into(),
3932 &root_entry,
3933 &scan_job_tx,
3934 self.fs.as_ref(),
3935 )
3936 .await;
3937 }
3938 }
3939 };
3940
3941 // Perform an initial scan of the directory.
3942 drop(scan_job_tx);
3943 self.scan_dirs(true, scan_job_rx).await;
3944 {
3945 let mut state = self.state.lock().await;
3946 state.snapshot.completed_scan_id = state.snapshot.scan_id;
3947 }
3948
3949 self.send_status_update(false, SmallVec::new(), &[]).await;
3950
3951 // Process any any FS events that occurred while performing the initial scan.
3952 // For these events, update events cannot be as precise, because we didn't
3953 // have the previous state loaded yet.
3954 self.phase = BackgroundScannerPhase::EventsReceivedDuringInitialScan;
3955 if let Poll::Ready(Some(mut paths)) = futures::poll!(fs_events_rx.next()) {
3956 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
3957 paths.extend(more_paths);
3958 }
3959 self.process_events(
3960 paths
3961 .into_iter()
3962 .filter(|event| event.kind.is_some())
3963 .collect(),
3964 )
3965 .await;
3966 }
3967 if let Some(abs_path) = containing_git_repository {
3968 self.process_events(vec![PathEvent {
3969 path: abs_path,
3970 kind: Some(fs::PathEventKind::Changed),
3971 }])
3972 .await;
3973 }
3974
3975 // Continue processing events until the worktree is dropped.
3976 self.phase = BackgroundScannerPhase::Events;
3977
3978 loop {
3979 select_biased! {
3980 // Process any path refresh requests from the worktree. Prioritize
3981 // these before handling changes reported by the filesystem.
3982 request = self.next_scan_request().fuse() => {
3983 let Ok(request) = request else { break };
3984 if !self.process_scan_request(request, false).await {
3985 return;
3986 }
3987 }
3988
3989 path_prefix_request = self.path_prefixes_to_scan_rx.recv().fuse() => {
3990 let Ok(request) = path_prefix_request else { break };
3991 log::trace!("adding path prefix {:?}", request.path);
3992
3993 let did_scan = self.forcibly_load_paths(std::slice::from_ref(&request.path)).await;
3994 if did_scan {
3995 let abs_path =
3996 {
3997 let mut state = self.state.lock().await;
3998 state.path_prefixes_to_scan.insert(request.path.clone());
3999 state.snapshot.absolutize(&request.path)
4000 };
4001
4002 if let Some(abs_path) = self.fs.canonicalize(&abs_path).await.log_err() {
4003 self.process_events(vec![PathEvent {
4004 path: abs_path,
4005 kind: Some(fs::PathEventKind::Changed),
4006 }])
4007 .await;
4008 }
4009 }
4010 self.send_status_update(false, request.done, &[]).await;
4011 }
4012
4013 paths = fs_events_rx.next().fuse() => {
4014 let Some(mut paths) = paths else { break };
4015 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
4016 paths.extend(more_paths);
4017 }
4018 self.process_events(paths.into_iter().filter(|event| event.kind.is_some()).collect()).await;
4019 }
4020
4021 _ = global_gitignore_events.next().fuse() => {
4022 if let Some(path) = &global_gitignore_file {
4023 self.update_global_gitignore(&path).await;
4024 }
4025 }
4026 }
4027 }
4028 }
4029
4030 async fn process_scan_request(&self, mut request: ScanRequest, scanning: bool) -> bool {
4031 log::debug!("rescanning paths {:?}", request.relative_paths);
4032
4033 request.relative_paths.sort_unstable();
4034 self.forcibly_load_paths(&request.relative_paths).await;
4035
4036 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4037 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4038 let root_canonical_path = match &root_canonical_path {
4039 Ok(path) => SanitizedPath::new(path),
4040 Err(err) => {
4041 log::error!("failed to canonicalize root path {root_path:?}: {err:#}");
4042 return true;
4043 }
4044 };
4045 let abs_paths = request
4046 .relative_paths
4047 .iter()
4048 .map(|path| {
4049 if path.file_name().is_some() {
4050 root_canonical_path.as_path().join(path.as_std_path())
4051 } else {
4052 root_canonical_path.as_path().to_path_buf()
4053 }
4054 })
4055 .collect::<Vec<_>>();
4056
4057 {
4058 let mut state = self.state.lock().await;
4059 let is_idle = state.snapshot.completed_scan_id == state.snapshot.scan_id;
4060 state.snapshot.scan_id += 1;
4061 if is_idle {
4062 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4063 }
4064 }
4065
4066 self.reload_entries_for_paths(
4067 &root_path,
4068 &root_canonical_path,
4069 &request.relative_paths,
4070 abs_paths,
4071 None,
4072 )
4073 .await;
4074
4075 self.send_status_update(scanning, request.done, &[]).await
4076 }
4077
4078 async fn process_events(&self, mut events: Vec<PathEvent>) {
4079 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4080 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4081 let root_canonical_path = match &root_canonical_path {
4082 Ok(path) => SanitizedPath::new(path),
4083 Err(err) => {
4084 let new_path = self
4085 .state
4086 .lock()
4087 .await
4088 .snapshot
4089 .root_file_handle
4090 .clone()
4091 .and_then(|handle| match handle.current_path(&self.fs) {
4092 Ok(new_path) => Some(new_path),
4093 Err(e) => {
4094 log::error!("Failed to refresh worktree root path: {e:#}");
4095 None
4096 }
4097 })
4098 .map(|path| SanitizedPath::new_arc(&path))
4099 .filter(|new_path| *new_path != root_path);
4100
4101 if let Some(new_path) = new_path {
4102 log::info!(
4103 "root renamed from {:?} to {:?}",
4104 root_path.as_path(),
4105 new_path.as_path(),
4106 );
4107 self.status_updates_tx
4108 .unbounded_send(ScanState::RootUpdated { new_path })
4109 .ok();
4110 } else {
4111 log::error!("root path could not be canonicalized: {err:#}");
4112
4113 // For single-file worktrees, if we can't canonicalize and the file handle
4114 // fallback also failed, the file is gone - close the worktree
4115 if self.is_single_file {
4116 log::info!(
4117 "single-file worktree root {:?} no longer exists, marking as deleted",
4118 root_path.as_path()
4119 );
4120 self.status_updates_tx
4121 .unbounded_send(ScanState::RootDeleted)
4122 .ok();
4123 }
4124 }
4125 return;
4126 }
4127 };
4128
4129 // Certain directories may have FS changes, but do not lead to git data changes that Zed cares about.
4130 // Ignore these, to avoid Zed unnecessarily rescanning git metadata.
4131 let skipped_files_in_dot_git = [COMMIT_MESSAGE, INDEX_LOCK];
4132 let skipped_dirs_in_dot_git = [FSMONITOR_DAEMON, LFS_DIR];
4133
4134 let mut relative_paths = Vec::with_capacity(events.len());
4135 let mut dot_git_abs_paths = Vec::new();
4136 let mut work_dirs_needing_exclude_update = Vec::new();
4137 events.sort_unstable_by(|left, right| left.path.cmp(&right.path));
4138 events.dedup_by(|left, right| {
4139 if left.path == right.path {
4140 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4141 right.kind = left.kind;
4142 }
4143 true
4144 } else if left.path.starts_with(&right.path) {
4145 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4146 right.kind = left.kind;
4147 }
4148 true
4149 } else {
4150 false
4151 }
4152 });
4153 {
4154 let snapshot = &self.state.lock().await.snapshot;
4155
4156 let mut ranges_to_drop = SmallVec::<[Range<usize>; 4]>::new();
4157
4158 fn skip_ix(ranges: &mut SmallVec<[Range<usize>; 4]>, ix: usize) {
4159 if let Some(last_range) = ranges.last_mut()
4160 && last_range.end == ix
4161 {
4162 last_range.end += 1;
4163 } else {
4164 ranges.push(ix..ix + 1);
4165 }
4166 }
4167
4168 for (ix, event) in events.iter().enumerate() {
4169 let abs_path = SanitizedPath::new(&event.path);
4170
4171 let mut is_git_related = false;
4172 let mut dot_git_paths = None;
4173
4174 for ancestor in abs_path.as_path().ancestors() {
4175 if is_git_dir(ancestor, self.fs.as_ref()).await {
4176 let path_in_git_dir = abs_path
4177 .as_path()
4178 .strip_prefix(ancestor)
4179 .expect("stripping off the ancestor");
4180 dot_git_paths = Some((ancestor.to_owned(), path_in_git_dir.to_owned()));
4181 break;
4182 }
4183 }
4184
4185 if let Some((dot_git_abs_path, path_in_git_dir)) = dot_git_paths {
4186 // We ignore `""` as well, as that is going to be the
4187 // `.git` folder itself. WE do not care about it, if
4188 // there are changes within we will see them, we need
4189 // this ignore to prevent us from accidentally observing
4190 // the ignored created file due to the events not being
4191 // empty after filtering.
4192
4193 let is_dot_git_changed = {
4194 path_in_git_dir == Path::new("")
4195 && event.kind == Some(PathEventKind::Changed)
4196 && abs_path
4197 .strip_prefix(root_canonical_path)
4198 .ok()
4199 .and_then(|it| RelPath::new(it, PathStyle::local()).ok())
4200 .is_some_and(|it| {
4201 snapshot
4202 .entry_for_path(&it)
4203 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4204 })
4205 };
4206 let condition = skipped_files_in_dot_git.iter().any(|skipped| {
4207 OsStr::new(skipped) == path_in_git_dir.as_path().as_os_str()
4208 }) || skipped_dirs_in_dot_git
4209 .iter()
4210 .any(|skipped_git_subdir| path_in_git_dir.starts_with(skipped_git_subdir))
4211 || is_dot_git_changed;
4212 if condition {
4213 log::debug!(
4214 "ignoring event {abs_path:?} as it's in the .git directory among skipped files or directories"
4215 );
4216 skip_ix(&mut ranges_to_drop, ix);
4217 continue;
4218 }
4219
4220 is_git_related = true;
4221 if !dot_git_abs_paths.contains(&dot_git_abs_path) {
4222 dot_git_abs_paths.push(dot_git_abs_path);
4223 }
4224 }
4225
4226 let relative_path = if let Ok(path) = abs_path.strip_prefix(&root_canonical_path)
4227 && let Ok(path) = RelPath::new(path, PathStyle::local())
4228 {
4229 path
4230 } else {
4231 if is_git_related {
4232 log::debug!(
4233 "ignoring event {abs_path:?}, since it's in git dir outside of root path {root_canonical_path:?}",
4234 );
4235 } else {
4236 log::error!(
4237 "ignoring event {abs_path:?} outside of root path {root_canonical_path:?}",
4238 );
4239 }
4240 skip_ix(&mut ranges_to_drop, ix);
4241 continue;
4242 };
4243
4244 let absolute_path = abs_path.to_path_buf();
4245 if absolute_path.ends_with(Path::new(DOT_GIT).join(REPO_EXCLUDE)) {
4246 if let Some(repository) = snapshot
4247 .git_repositories
4248 .values()
4249 .find(|repo| repo.common_dir_abs_path.join(REPO_EXCLUDE) == absolute_path)
4250 {
4251 work_dirs_needing_exclude_update
4252 .push(repository.work_directory_abs_path.clone());
4253 }
4254 }
4255
4256 if abs_path.file_name() == Some(OsStr::new(GITIGNORE)) {
4257 for (_, repo) in snapshot
4258 .git_repositories
4259 .iter()
4260 .filter(|(_, repo)| repo.directory_contains(&relative_path))
4261 {
4262 if !dot_git_abs_paths.iter().any(|dot_git_abs_path| {
4263 dot_git_abs_path == repo.common_dir_abs_path.as_ref()
4264 }) {
4265 dot_git_abs_paths.push(repo.common_dir_abs_path.to_path_buf());
4266 }
4267 }
4268 }
4269
4270 let parent_dir_is_loaded = relative_path.parent().is_none_or(|parent| {
4271 snapshot
4272 .entry_for_path(parent)
4273 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4274 });
4275 if !parent_dir_is_loaded {
4276 log::debug!("ignoring event {relative_path:?} within unloaded directory");
4277 skip_ix(&mut ranges_to_drop, ix);
4278 continue;
4279 }
4280
4281 if self.settings.is_path_excluded(&relative_path) {
4282 if !is_git_related {
4283 log::debug!("ignoring FS event for excluded path {relative_path:?}");
4284 }
4285 skip_ix(&mut ranges_to_drop, ix);
4286 continue;
4287 }
4288
4289 relative_paths.push(EventRoot {
4290 path: relative_path.into_arc(),
4291 was_rescanned: matches!(event.kind, Some(fs::PathEventKind::Rescan)),
4292 });
4293 }
4294
4295 for range_to_drop in ranges_to_drop.into_iter().rev() {
4296 events.drain(range_to_drop);
4297 }
4298 }
4299
4300 if relative_paths.is_empty() && dot_git_abs_paths.is_empty() {
4301 return;
4302 }
4303
4304 if !work_dirs_needing_exclude_update.is_empty() {
4305 let mut state = self.state.lock().await;
4306 for work_dir_abs_path in work_dirs_needing_exclude_update {
4307 if let Some((_, needs_update)) = state
4308 .snapshot
4309 .repo_exclude_by_work_dir_abs_path
4310 .get_mut(&work_dir_abs_path)
4311 {
4312 *needs_update = true;
4313 }
4314 }
4315 }
4316
4317 self.state.lock().await.snapshot.scan_id += 1;
4318
4319 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4320 log::debug!(
4321 "received fs events {:?}",
4322 relative_paths
4323 .iter()
4324 .map(|event_root| &event_root.path)
4325 .collect::<Vec<_>>()
4326 );
4327 self.reload_entries_for_paths(
4328 &root_path,
4329 &root_canonical_path,
4330 &relative_paths
4331 .iter()
4332 .map(|event_root| event_root.path.clone())
4333 .collect::<Vec<_>>(),
4334 events
4335 .into_iter()
4336 .map(|event| event.path)
4337 .collect::<Vec<_>>(),
4338 Some(scan_job_tx.clone()),
4339 )
4340 .await;
4341
4342 let affected_repo_roots = if !dot_git_abs_paths.is_empty() {
4343 self.update_git_repositories(dot_git_abs_paths).await
4344 } else {
4345 Vec::new()
4346 };
4347
4348 {
4349 let mut ignores_to_update = self.ignores_needing_update().await;
4350 ignores_to_update.extend(affected_repo_roots);
4351 let ignores_to_update = self.order_ignores(ignores_to_update).await;
4352 let snapshot = self.state.lock().await.snapshot.clone();
4353 self.update_ignore_statuses_for_paths(scan_job_tx, snapshot, ignores_to_update)
4354 .await;
4355 self.scan_dirs(false, scan_job_rx).await;
4356 }
4357
4358 {
4359 let mut state = self.state.lock().await;
4360 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4361 for (_, entry) in mem::take(&mut state.removed_entries) {
4362 state.scanned_dirs.remove(&entry.id);
4363 }
4364 }
4365 self.send_status_update(false, SmallVec::new(), &relative_paths)
4366 .await;
4367 }
4368
4369 async fn update_global_gitignore(&self, abs_path: &Path) {
4370 let ignore = build_gitignore(abs_path, self.fs.as_ref())
4371 .await
4372 .log_err()
4373 .map(Arc::new);
4374 let (prev_snapshot, ignore_stack, abs_path) = {
4375 let mut state = self.state.lock().await;
4376 state.snapshot.global_gitignore = ignore;
4377 let abs_path = state.snapshot.abs_path().clone();
4378 let ignore_stack = state
4379 .snapshot
4380 .ignore_stack_for_abs_path(&abs_path, true, self.fs.as_ref())
4381 .await;
4382 (state.snapshot.clone(), ignore_stack, abs_path)
4383 };
4384 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4385 self.update_ignore_statuses_for_paths(
4386 scan_job_tx,
4387 prev_snapshot,
4388 vec![(abs_path, ignore_stack)],
4389 )
4390 .await;
4391 self.scan_dirs(false, scan_job_rx).await;
4392 self.send_status_update(false, SmallVec::new(), &[]).await;
4393 }
4394
4395 async fn forcibly_load_paths(&self, paths: &[Arc<RelPath>]) -> bool {
4396 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4397 {
4398 let mut state = self.state.lock().await;
4399 let root_path = state.snapshot.abs_path.clone();
4400 for path in paths {
4401 for ancestor in path.ancestors() {
4402 if let Some(entry) = state.snapshot.entry_for_path(ancestor)
4403 && entry.kind == EntryKind::UnloadedDir
4404 {
4405 let abs_path = root_path.join(ancestor.as_std_path());
4406 state
4407 .enqueue_scan_dir(
4408 abs_path.into(),
4409 entry,
4410 &scan_job_tx,
4411 self.fs.as_ref(),
4412 )
4413 .await;
4414 state.paths_to_scan.insert(path.clone());
4415 break;
4416 }
4417 }
4418 }
4419 drop(scan_job_tx);
4420 }
4421 while let Ok(job) = scan_job_rx.recv().await {
4422 self.scan_dir(&job).await.log_err();
4423 }
4424
4425 !mem::take(&mut self.state.lock().await.paths_to_scan).is_empty()
4426 }
4427
4428 async fn scan_dirs(
4429 &self,
4430 enable_progress_updates: bool,
4431 scan_jobs_rx: channel::Receiver<ScanJob>,
4432 ) {
4433 if self
4434 .status_updates_tx
4435 .unbounded_send(ScanState::Started)
4436 .is_err()
4437 {
4438 return;
4439 }
4440
4441 let progress_update_count = AtomicUsize::new(0);
4442 self.executor
4443 .scoped_priority(Priority::Low, |scope| {
4444 for _ in 0..self.executor.num_cpus() {
4445 scope.spawn(async {
4446 let mut last_progress_update_count = 0;
4447 let progress_update_timer = self.progress_timer(enable_progress_updates).fuse();
4448 futures::pin_mut!(progress_update_timer);
4449
4450 loop {
4451 select_biased! {
4452 // Process any path refresh requests before moving on to process
4453 // the scan queue, so that user operations are prioritized.
4454 request = self.next_scan_request().fuse() => {
4455 let Ok(request) = request else { break };
4456 if !self.process_scan_request(request, true).await {
4457 return;
4458 }
4459 }
4460
4461 // Send periodic progress updates to the worktree. Use an atomic counter
4462 // to ensure that only one of the workers sends a progress update after
4463 // the update interval elapses.
4464 _ = progress_update_timer => {
4465 match progress_update_count.compare_exchange(
4466 last_progress_update_count,
4467 last_progress_update_count + 1,
4468 SeqCst,
4469 SeqCst
4470 ) {
4471 Ok(_) => {
4472 last_progress_update_count += 1;
4473 self.send_status_update(true, SmallVec::new(), &[])
4474 .await;
4475 }
4476 Err(count) => {
4477 last_progress_update_count = count;
4478 }
4479 }
4480 progress_update_timer.set(self.progress_timer(enable_progress_updates).fuse());
4481 }
4482
4483 // Recursively load directories from the file system.
4484 job = scan_jobs_rx.recv().fuse() => {
4485 let Ok(job) = job else { break };
4486 if let Err(err) = self.scan_dir(&job).await
4487 && job.path.is_empty() {
4488 log::error!("error scanning directory {:?}: {}", job.abs_path, err);
4489 }
4490 }
4491 }
4492 }
4493 });
4494 }
4495 })
4496 .await;
4497 }
4498
4499 async fn send_status_update(
4500 &self,
4501 scanning: bool,
4502 barrier: SmallVec<[barrier::Sender; 1]>,
4503 event_roots: &[EventRoot],
4504 ) -> bool {
4505 let mut state = self.state.lock().await;
4506 if state.changed_paths.is_empty() && event_roots.is_empty() && scanning {
4507 return true;
4508 }
4509
4510 let merged_event_roots = merge_event_roots(&state.changed_paths, event_roots);
4511
4512 let new_snapshot = state.snapshot.clone();
4513 let old_snapshot = mem::replace(&mut state.prev_snapshot, new_snapshot.snapshot.clone());
4514 let changes = build_diff(
4515 self.phase,
4516 &old_snapshot,
4517 &new_snapshot,
4518 &merged_event_roots,
4519 );
4520 state.changed_paths.clear();
4521
4522 self.status_updates_tx
4523 .unbounded_send(ScanState::Updated {
4524 snapshot: new_snapshot,
4525 changes,
4526 scanning,
4527 barrier,
4528 })
4529 .is_ok()
4530 }
4531
4532 async fn scan_dir(&self, job: &ScanJob) -> Result<()> {
4533 let root_abs_path;
4534 let root_char_bag;
4535 {
4536 let snapshot = &self.state.lock().await.snapshot;
4537 if self.settings.is_path_excluded(&job.path) {
4538 log::error!("skipping excluded directory {:?}", job.path);
4539 return Ok(());
4540 }
4541 log::trace!("scanning directory {:?}", job.path);
4542 root_abs_path = snapshot.abs_path().clone();
4543 root_char_bag = snapshot.root_char_bag;
4544 }
4545
4546 let next_entry_id = self.next_entry_id.clone();
4547 let mut ignore_stack = job.ignore_stack.clone();
4548 let mut new_ignore = None;
4549 let mut root_canonical_path = None;
4550 let mut new_entries: Vec<Entry> = Vec::new();
4551 let mut new_jobs: Vec<Option<ScanJob>> = Vec::new();
4552 let mut child_paths = self
4553 .fs
4554 .read_dir(&job.abs_path)
4555 .await?
4556 .filter_map(|entry| async {
4557 match entry {
4558 Ok(entry) => Some(entry),
4559 Err(error) => {
4560 log::error!("error processing entry {:?}", error);
4561 None
4562 }
4563 }
4564 })
4565 .collect::<Vec<_>>()
4566 .await;
4567
4568 // Ensure that .git and .gitignore are processed first.
4569 swap_to_front(&mut child_paths, GITIGNORE);
4570 swap_to_front(&mut child_paths, DOT_GIT);
4571
4572 if let Some(path) = child_paths.first()
4573 && path.ends_with(DOT_GIT)
4574 {
4575 ignore_stack.repo_root = Some(job.abs_path.clone());
4576 }
4577
4578 for child_abs_path in child_paths {
4579 let child_abs_path: Arc<Path> = child_abs_path.into();
4580 let child_name = child_abs_path.file_name().unwrap();
4581 let Some(child_path) = child_name
4582 .to_str()
4583 .and_then(|name| Some(job.path.join(RelPath::unix(name).ok()?)))
4584 else {
4585 continue;
4586 };
4587
4588 if child_name == DOT_GIT {
4589 let mut state = self.state.lock().await;
4590 state
4591 .insert_git_repository(
4592 child_path.clone(),
4593 self.fs.as_ref(),
4594 self.watcher.as_ref(),
4595 )
4596 .await;
4597 } else if child_name == GITIGNORE {
4598 match build_gitignore(&child_abs_path, self.fs.as_ref()).await {
4599 Ok(ignore) => {
4600 let ignore = Arc::new(ignore);
4601 ignore_stack = ignore_stack
4602 .append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4603 new_ignore = Some(ignore);
4604 }
4605 Err(error) => {
4606 log::error!(
4607 "error loading .gitignore file {:?} - {:?}",
4608 child_name,
4609 error
4610 );
4611 }
4612 }
4613 }
4614
4615 if self.settings.is_path_excluded(&child_path) {
4616 log::debug!("skipping excluded child entry {child_path:?}");
4617 self.state
4618 .lock()
4619 .await
4620 .remove_path(&child_path, self.watcher.as_ref());
4621 continue;
4622 }
4623
4624 let child_metadata = match self.fs.metadata(&child_abs_path).await {
4625 Ok(Some(metadata)) => metadata,
4626 Ok(None) => continue,
4627 Err(err) => {
4628 log::error!("error processing {:?}: {err:#}", child_abs_path.display());
4629 continue;
4630 }
4631 };
4632
4633 let mut child_entry = Entry::new(
4634 child_path.clone(),
4635 &child_metadata,
4636 ProjectEntryId::new(&next_entry_id),
4637 root_char_bag,
4638 None,
4639 );
4640
4641 if job.is_external {
4642 child_entry.is_external = true;
4643 } else if child_metadata.is_symlink {
4644 let canonical_path = match self.fs.canonicalize(&child_abs_path).await {
4645 Ok(path) => path,
4646 Err(err) => {
4647 log::error!("error reading target of symlink {child_abs_path:?}: {err:#}",);
4648 continue;
4649 }
4650 };
4651
4652 // lazily canonicalize the root path in order to determine if
4653 // symlinks point outside of the worktree.
4654 let root_canonical_path = match &root_canonical_path {
4655 Some(path) => path,
4656 None => match self.fs.canonicalize(&root_abs_path).await {
4657 Ok(path) => root_canonical_path.insert(path),
4658 Err(err) => {
4659 log::error!("error canonicalizing root {:?}: {:?}", root_abs_path, err);
4660 continue;
4661 }
4662 },
4663 };
4664
4665 if !canonical_path.starts_with(root_canonical_path) {
4666 child_entry.is_external = true;
4667 }
4668
4669 child_entry.canonical_path = Some(canonical_path.into());
4670 }
4671
4672 if child_entry.is_dir() {
4673 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, true);
4674 child_entry.is_always_included =
4675 self.settings.is_path_always_included(&child_path, true);
4676
4677 // Avoid recursing until crash in the case of a recursive symlink
4678 if job.ancestor_inodes.contains(&child_entry.inode) {
4679 new_jobs.push(None);
4680 } else {
4681 let mut ancestor_inodes = job.ancestor_inodes.clone();
4682 ancestor_inodes.insert(child_entry.inode);
4683
4684 new_jobs.push(Some(ScanJob {
4685 abs_path: child_abs_path.clone(),
4686 path: child_path,
4687 is_external: child_entry.is_external,
4688 ignore_stack: if child_entry.is_ignored {
4689 IgnoreStack::all()
4690 } else {
4691 ignore_stack.clone()
4692 },
4693 ancestor_inodes,
4694 scan_queue: job.scan_queue.clone(),
4695 }));
4696 }
4697 } else {
4698 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, false);
4699 child_entry.is_always_included =
4700 self.settings.is_path_always_included(&child_path, false);
4701 }
4702
4703 {
4704 let relative_path = job
4705 .path
4706 .join(RelPath::unix(child_name.to_str().unwrap()).unwrap());
4707 if self.is_path_private(&relative_path) {
4708 log::debug!("detected private file: {relative_path:?}");
4709 child_entry.is_private = true;
4710 }
4711 if self.settings.is_path_hidden(&relative_path) {
4712 log::debug!("detected hidden file: {relative_path:?}");
4713 child_entry.is_hidden = true;
4714 }
4715 }
4716
4717 new_entries.push(child_entry);
4718 }
4719
4720 let mut state = self.state.lock().await;
4721
4722 // Identify any subdirectories that should not be scanned.
4723 let mut job_ix = 0;
4724 for entry in &mut new_entries {
4725 state.reuse_entry_id(entry);
4726 if entry.is_dir() {
4727 if state.should_scan_directory(entry) {
4728 job_ix += 1;
4729 } else {
4730 log::debug!("defer scanning directory {:?}", entry.path);
4731 entry.kind = EntryKind::UnloadedDir;
4732 new_jobs.remove(job_ix);
4733 }
4734 }
4735 if entry.is_always_included {
4736 state
4737 .snapshot
4738 .always_included_entries
4739 .push(entry.path.clone());
4740 }
4741 }
4742
4743 state.populate_dir(job.path.clone(), new_entries, new_ignore);
4744 self.watcher.add(job.abs_path.as_ref()).log_err();
4745
4746 for new_job in new_jobs.into_iter().flatten() {
4747 job.scan_queue
4748 .try_send(new_job)
4749 .expect("channel is unbounded");
4750 }
4751
4752 Ok(())
4753 }
4754
4755 /// All list arguments should be sorted before calling this function
4756 async fn reload_entries_for_paths(
4757 &self,
4758 root_abs_path: &SanitizedPath,
4759 root_canonical_path: &SanitizedPath,
4760 relative_paths: &[Arc<RelPath>],
4761 abs_paths: Vec<PathBuf>,
4762 scan_queue_tx: Option<Sender<ScanJob>>,
4763 ) {
4764 // grab metadata for all requested paths
4765 let metadata = futures::future::join_all(
4766 abs_paths
4767 .iter()
4768 .map(|abs_path| async move {
4769 let metadata = self.fs.metadata(abs_path).await?;
4770 if let Some(metadata) = metadata {
4771 let canonical_path = self.fs.canonicalize(abs_path).await?;
4772
4773 // If we're on a case-insensitive filesystem (default on macOS), we want
4774 // to only ignore metadata for non-symlink files if their absolute-path matches
4775 // the canonical-path.
4776 // Because if not, this might be a case-only-renaming (`mv test.txt TEST.TXT`)
4777 // and we want to ignore the metadata for the old path (`test.txt`) so it's
4778 // treated as removed.
4779 if !self.fs_case_sensitive && !metadata.is_symlink {
4780 let canonical_file_name = canonical_path.file_name();
4781 let file_name = abs_path.file_name();
4782 if canonical_file_name != file_name {
4783 return Ok(None);
4784 }
4785 }
4786
4787 anyhow::Ok(Some((metadata, SanitizedPath::new_arc(&canonical_path))))
4788 } else {
4789 Ok(None)
4790 }
4791 })
4792 .collect::<Vec<_>>(),
4793 )
4794 .await;
4795
4796 let mut new_ancestor_repo = if relative_paths.iter().any(|path| path.is_empty()) {
4797 Some(discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await)
4798 } else {
4799 None
4800 };
4801
4802 let mut state = self.state.lock().await;
4803 let doing_recursive_update = scan_queue_tx.is_some();
4804
4805 // Remove any entries for paths that no longer exist or are being recursively
4806 // refreshed. Do this before adding any new entries, so that renames can be
4807 // detected regardless of the order of the paths.
4808 for (path, metadata) in relative_paths.iter().zip(metadata.iter()) {
4809 if matches!(metadata, Ok(None)) || doing_recursive_update {
4810 state.remove_path(path, self.watcher.as_ref());
4811 }
4812 }
4813
4814 for (path, metadata) in relative_paths.iter().zip(metadata.into_iter()) {
4815 let abs_path: Arc<Path> = root_abs_path.join(path.as_std_path()).into();
4816 match metadata {
4817 Ok(Some((metadata, canonical_path))) => {
4818 let ignore_stack = state
4819 .snapshot
4820 .ignore_stack_for_abs_path(&abs_path, metadata.is_dir, self.fs.as_ref())
4821 .await;
4822 let is_external = !canonical_path.starts_with(&root_canonical_path);
4823 let entry_id = state.entry_id_for(self.next_entry_id.as_ref(), path, &metadata);
4824 let mut fs_entry = Entry::new(
4825 path.clone(),
4826 &metadata,
4827 entry_id,
4828 state.snapshot.root_char_bag,
4829 if metadata.is_symlink {
4830 Some(canonical_path.as_path().to_path_buf().into())
4831 } else {
4832 None
4833 },
4834 );
4835
4836 let is_dir = fs_entry.is_dir();
4837 fs_entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, is_dir);
4838 fs_entry.is_external = is_external;
4839 fs_entry.is_private = self.is_path_private(path);
4840 fs_entry.is_always_included =
4841 self.settings.is_path_always_included(path, is_dir);
4842 fs_entry.is_hidden = self.settings.is_path_hidden(path);
4843
4844 if let (Some(scan_queue_tx), true) = (&scan_queue_tx, is_dir) {
4845 if state.should_scan_directory(&fs_entry)
4846 || (fs_entry.path.is_empty()
4847 && abs_path.file_name() == Some(OsStr::new(DOT_GIT)))
4848 {
4849 state
4850 .enqueue_scan_dir(
4851 abs_path,
4852 &fs_entry,
4853 scan_queue_tx,
4854 self.fs.as_ref(),
4855 )
4856 .await;
4857 } else {
4858 fs_entry.kind = EntryKind::UnloadedDir;
4859 }
4860 }
4861
4862 state
4863 .insert_entry(fs_entry.clone(), self.fs.as_ref(), self.watcher.as_ref())
4864 .await;
4865
4866 if path.is_empty()
4867 && let Some((ignores, exclude, repo)) = new_ancestor_repo.take()
4868 {
4869 log::trace!("updating ancestor git repository");
4870 state.snapshot.ignores_by_parent_abs_path.extend(ignores);
4871 if let Some((ancestor_dot_git, work_directory)) = repo {
4872 if let Some(exclude) = exclude {
4873 let work_directory_abs_path = self
4874 .state
4875 .lock()
4876 .await
4877 .snapshot
4878 .work_directory_abs_path(&work_directory);
4879
4880 state
4881 .snapshot
4882 .repo_exclude_by_work_dir_abs_path
4883 .insert(work_directory_abs_path.into(), (exclude, false));
4884 }
4885 state
4886 .insert_git_repository_for_path(
4887 work_directory,
4888 ancestor_dot_git.into(),
4889 self.fs.as_ref(),
4890 self.watcher.as_ref(),
4891 )
4892 .await
4893 .log_err();
4894 }
4895 }
4896 }
4897 Ok(None) => {
4898 self.remove_repo_path(path.clone(), &mut state.snapshot);
4899 }
4900 Err(err) => {
4901 log::error!("error reading file {abs_path:?} on event: {err:#}");
4902 }
4903 }
4904 }
4905
4906 util::extend_sorted(
4907 &mut state.changed_paths,
4908 relative_paths.iter().cloned(),
4909 usize::MAX,
4910 Ord::cmp,
4911 );
4912 }
4913
4914 fn remove_repo_path(&self, path: Arc<RelPath>, snapshot: &mut LocalSnapshot) -> Option<()> {
4915 if !path.components().any(|component| component == DOT_GIT)
4916 && let Some(local_repo) = snapshot.local_repo_for_work_directory_path(&path)
4917 {
4918 let id = local_repo.work_directory_id;
4919 log::debug!("remove repo path: {:?}", path);
4920 snapshot.git_repositories.remove(&id);
4921 return Some(());
4922 }
4923
4924 Some(())
4925 }
4926
4927 async fn update_ignore_statuses_for_paths(
4928 &self,
4929 scan_job_tx: Sender<ScanJob>,
4930 prev_snapshot: LocalSnapshot,
4931 ignores_to_update: Vec<(Arc<Path>, IgnoreStack)>,
4932 ) {
4933 let (ignore_queue_tx, ignore_queue_rx) = channel::unbounded();
4934 {
4935 for (parent_abs_path, ignore_stack) in ignores_to_update {
4936 ignore_queue_tx
4937 .send_blocking(UpdateIgnoreStatusJob {
4938 abs_path: parent_abs_path,
4939 ignore_stack,
4940 ignore_queue: ignore_queue_tx.clone(),
4941 scan_queue: scan_job_tx.clone(),
4942 })
4943 .unwrap();
4944 }
4945 }
4946 drop(ignore_queue_tx);
4947
4948 self.executor
4949 .scoped(|scope| {
4950 for _ in 0..self.executor.num_cpus() {
4951 scope.spawn(async {
4952 loop {
4953 select_biased! {
4954 // Process any path refresh requests before moving on to process
4955 // the queue of ignore statuses.
4956 request = self.next_scan_request().fuse() => {
4957 let Ok(request) = request else { break };
4958 if !self.process_scan_request(request, true).await {
4959 return;
4960 }
4961 }
4962
4963 // Recursively process directories whose ignores have changed.
4964 job = ignore_queue_rx.recv().fuse() => {
4965 let Ok(job) = job else { break };
4966 self.update_ignore_status(job, &prev_snapshot).await;
4967 }
4968 }
4969 }
4970 });
4971 }
4972 })
4973 .await;
4974 }
4975
4976 async fn ignores_needing_update(&self) -> Vec<Arc<Path>> {
4977 let mut ignores_to_update = Vec::new();
4978 let mut excludes_to_load: Vec<(Arc<Path>, PathBuf)> = Vec::new();
4979
4980 // First pass: collect updates and drop stale entries without awaiting.
4981 {
4982 let snapshot = &mut self.state.lock().await.snapshot;
4983 let abs_path = snapshot.abs_path.clone();
4984 let mut repo_exclude_keys_to_remove: Vec<Arc<Path>> = Vec::new();
4985
4986 for (work_dir_abs_path, (_, needs_update)) in
4987 snapshot.repo_exclude_by_work_dir_abs_path.iter_mut()
4988 {
4989 let repository = snapshot
4990 .git_repositories
4991 .iter()
4992 .find(|(_, repo)| &repo.work_directory_abs_path == work_dir_abs_path);
4993
4994 if *needs_update {
4995 *needs_update = false;
4996 ignores_to_update.push(work_dir_abs_path.clone());
4997
4998 if let Some((_, repository)) = repository {
4999 let exclude_abs_path = repository.common_dir_abs_path.join(REPO_EXCLUDE);
5000 excludes_to_load.push((work_dir_abs_path.clone(), exclude_abs_path));
5001 }
5002 }
5003
5004 if repository.is_none() {
5005 repo_exclude_keys_to_remove.push(work_dir_abs_path.clone());
5006 }
5007 }
5008
5009 for key in repo_exclude_keys_to_remove {
5010 snapshot.repo_exclude_by_work_dir_abs_path.remove(&key);
5011 }
5012
5013 snapshot
5014 .ignores_by_parent_abs_path
5015 .retain(|parent_abs_path, (_, needs_update)| {
5016 if let Ok(parent_path) = parent_abs_path.strip_prefix(abs_path.as_path())
5017 && let Some(parent_path) =
5018 RelPath::new(&parent_path, PathStyle::local()).log_err()
5019 {
5020 if *needs_update {
5021 *needs_update = false;
5022 if snapshot.snapshot.entry_for_path(&parent_path).is_some() {
5023 ignores_to_update.push(parent_abs_path.clone());
5024 }
5025 }
5026
5027 let ignore_path = parent_path.join(RelPath::unix(GITIGNORE).unwrap());
5028 if snapshot.snapshot.entry_for_path(&ignore_path).is_none() {
5029 return false;
5030 }
5031 }
5032 true
5033 });
5034 }
5035
5036 // Load gitignores asynchronously (outside the lock)
5037 let mut loaded_excludes: Vec<(Arc<Path>, Arc<Gitignore>)> = Vec::new();
5038 for (work_dir_abs_path, exclude_abs_path) in excludes_to_load {
5039 if let Ok(current_exclude) = build_gitignore(&exclude_abs_path, self.fs.as_ref()).await
5040 {
5041 loaded_excludes.push((work_dir_abs_path, Arc::new(current_exclude)));
5042 }
5043 }
5044
5045 // Second pass: apply updates.
5046 if !loaded_excludes.is_empty() {
5047 let snapshot = &mut self.state.lock().await.snapshot;
5048
5049 for (work_dir_abs_path, exclude) in loaded_excludes {
5050 if let Some((existing_exclude, _)) = snapshot
5051 .repo_exclude_by_work_dir_abs_path
5052 .get_mut(&work_dir_abs_path)
5053 {
5054 *existing_exclude = exclude;
5055 }
5056 }
5057 }
5058
5059 ignores_to_update
5060 }
5061
5062 async fn order_ignores(&self, mut ignores: Vec<Arc<Path>>) -> Vec<(Arc<Path>, IgnoreStack)> {
5063 let fs = self.fs.clone();
5064 let snapshot = self.state.lock().await.snapshot.clone();
5065 ignores.sort_unstable();
5066 let mut ignores_to_update = ignores.into_iter().peekable();
5067
5068 let mut result = vec![];
5069 while let Some(parent_abs_path) = ignores_to_update.next() {
5070 while ignores_to_update
5071 .peek()
5072 .map_or(false, |p| p.starts_with(&parent_abs_path))
5073 {
5074 ignores_to_update.next().unwrap();
5075 }
5076 let ignore_stack = snapshot
5077 .ignore_stack_for_abs_path(&parent_abs_path, true, fs.as_ref())
5078 .await;
5079 result.push((parent_abs_path, ignore_stack));
5080 }
5081
5082 result
5083 }
5084
5085 async fn update_ignore_status(&self, job: UpdateIgnoreStatusJob, snapshot: &LocalSnapshot) {
5086 log::trace!("update ignore status {:?}", job.abs_path);
5087
5088 let mut ignore_stack = job.ignore_stack;
5089 if let Some((ignore, _)) = snapshot.ignores_by_parent_abs_path.get(&job.abs_path) {
5090 ignore_stack =
5091 ignore_stack.append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
5092 }
5093
5094 let mut entries_by_id_edits = Vec::new();
5095 let mut entries_by_path_edits = Vec::new();
5096 let Some(path) = job
5097 .abs_path
5098 .strip_prefix(snapshot.abs_path.as_path())
5099 .map_err(|_| {
5100 anyhow::anyhow!(
5101 "Failed to strip prefix '{}' from path '{}'",
5102 snapshot.abs_path.as_path().display(),
5103 job.abs_path.display()
5104 )
5105 })
5106 .log_err()
5107 else {
5108 return;
5109 };
5110
5111 let Some(path) = RelPath::new(&path, PathStyle::local()).log_err() else {
5112 return;
5113 };
5114
5115 if let Ok(Some(metadata)) = self.fs.metadata(&job.abs_path.join(DOT_GIT)).await
5116 && metadata.is_dir
5117 {
5118 ignore_stack.repo_root = Some(job.abs_path.clone());
5119 }
5120
5121 for mut entry in snapshot.child_entries(&path).cloned() {
5122 let was_ignored = entry.is_ignored;
5123 let abs_path: Arc<Path> = snapshot.absolutize(&entry.path).into();
5124 entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, entry.is_dir());
5125
5126 if entry.is_dir() {
5127 let child_ignore_stack = if entry.is_ignored {
5128 IgnoreStack::all()
5129 } else {
5130 ignore_stack.clone()
5131 };
5132
5133 // Scan any directories that were previously ignored and weren't previously scanned.
5134 if was_ignored && !entry.is_ignored && entry.kind.is_unloaded() {
5135 let state = self.state.lock().await;
5136 if state.should_scan_directory(&entry) {
5137 state
5138 .enqueue_scan_dir(
5139 abs_path.clone(),
5140 &entry,
5141 &job.scan_queue,
5142 self.fs.as_ref(),
5143 )
5144 .await;
5145 }
5146 }
5147
5148 job.ignore_queue
5149 .send(UpdateIgnoreStatusJob {
5150 abs_path: abs_path.clone(),
5151 ignore_stack: child_ignore_stack,
5152 ignore_queue: job.ignore_queue.clone(),
5153 scan_queue: job.scan_queue.clone(),
5154 })
5155 .await
5156 .unwrap();
5157 }
5158
5159 if entry.is_ignored != was_ignored {
5160 let mut path_entry = snapshot.entries_by_id.get(&entry.id, ()).unwrap().clone();
5161 path_entry.scan_id = snapshot.scan_id;
5162 path_entry.is_ignored = entry.is_ignored;
5163 entries_by_id_edits.push(Edit::Insert(path_entry));
5164 entries_by_path_edits.push(Edit::Insert(entry));
5165 }
5166 }
5167
5168 let state = &mut self.state.lock().await;
5169 for edit in &entries_by_path_edits {
5170 if let Edit::Insert(entry) = edit
5171 && let Err(ix) = state.changed_paths.binary_search(&entry.path)
5172 {
5173 state.changed_paths.insert(ix, entry.path.clone());
5174 }
5175 }
5176
5177 state
5178 .snapshot
5179 .entries_by_path
5180 .edit(entries_by_path_edits, ());
5181 state.snapshot.entries_by_id.edit(entries_by_id_edits, ());
5182 }
5183
5184 async fn update_git_repositories(&self, dot_git_paths: Vec<PathBuf>) -> Vec<Arc<Path>> {
5185 log::trace!("reloading repositories: {dot_git_paths:?}");
5186 let mut state = self.state.lock().await;
5187 let scan_id = state.snapshot.scan_id;
5188 let mut affected_repo_roots = Vec::new();
5189 for dot_git_dir in dot_git_paths {
5190 let existing_repository_entry =
5191 state
5192 .snapshot
5193 .git_repositories
5194 .iter()
5195 .find_map(|(_, repo)| {
5196 let dot_git_dir = SanitizedPath::new(&dot_git_dir);
5197 if SanitizedPath::new(repo.common_dir_abs_path.as_ref()) == dot_git_dir
5198 || SanitizedPath::new(repo.repository_dir_abs_path.as_ref())
5199 == dot_git_dir
5200 {
5201 Some(repo.clone())
5202 } else {
5203 None
5204 }
5205 });
5206
5207 match existing_repository_entry {
5208 None => {
5209 let Ok(relative) = dot_git_dir.strip_prefix(state.snapshot.abs_path()) else {
5210 debug_panic!(
5211 "update_git_repositories called with .git directory outside the worktree root"
5212 );
5213 return Vec::new();
5214 };
5215 affected_repo_roots.push(dot_git_dir.parent().unwrap().into());
5216 state
5217 .insert_git_repository(
5218 RelPath::new(relative, PathStyle::local())
5219 .unwrap()
5220 .into_arc(),
5221 self.fs.as_ref(),
5222 self.watcher.as_ref(),
5223 )
5224 .await;
5225 }
5226 Some(local_repository) => {
5227 state.snapshot.git_repositories.update(
5228 &local_repository.work_directory_id,
5229 |entry| {
5230 entry.git_dir_scan_id = scan_id;
5231 },
5232 );
5233 }
5234 };
5235 }
5236
5237 // Remove any git repositories whose .git entry no longer exists.
5238 let snapshot = &mut state.snapshot;
5239 let mut ids_to_preserve = HashSet::default();
5240 for (&work_directory_id, entry) in snapshot.git_repositories.iter() {
5241 let exists_in_snapshot =
5242 snapshot
5243 .entry_for_id(work_directory_id)
5244 .is_some_and(|entry| {
5245 snapshot
5246 .entry_for_path(&entry.path.join(RelPath::unix(DOT_GIT).unwrap()))
5247 .is_some()
5248 });
5249
5250 if exists_in_snapshot
5251 || matches!(
5252 self.fs.metadata(&entry.common_dir_abs_path).await,
5253 Ok(Some(_))
5254 )
5255 {
5256 ids_to_preserve.insert(work_directory_id);
5257 }
5258 }
5259
5260 snapshot
5261 .git_repositories
5262 .retain(|work_directory_id, entry| {
5263 let preserve = ids_to_preserve.contains(work_directory_id);
5264 if !preserve {
5265 affected_repo_roots.push(entry.dot_git_abs_path.parent().unwrap().into());
5266 snapshot
5267 .repo_exclude_by_work_dir_abs_path
5268 .remove(&entry.work_directory_abs_path);
5269 }
5270 preserve
5271 });
5272
5273 affected_repo_roots
5274 }
5275
5276 async fn progress_timer(&self, running: bool) {
5277 if !running {
5278 return futures::future::pending().await;
5279 }
5280
5281 #[cfg(feature = "test-support")]
5282 if self.fs.is_fake() {
5283 return self.executor.simulate_random_delay().await;
5284 }
5285
5286 self.executor.timer(FS_WATCH_LATENCY).await
5287 }
5288
5289 fn is_path_private(&self, path: &RelPath) -> bool {
5290 !self.share_private_files && self.settings.is_path_private(path)
5291 }
5292
5293 async fn next_scan_request(&self) -> Result<ScanRequest> {
5294 let mut request = self.scan_requests_rx.recv().await?;
5295 while let Ok(next_request) = self.scan_requests_rx.try_recv() {
5296 request.relative_paths.extend(next_request.relative_paths);
5297 request.done.extend(next_request.done);
5298 }
5299 Ok(request)
5300 }
5301}
5302
5303async fn discover_ancestor_git_repo(
5304 fs: Arc<dyn Fs>,
5305 root_abs_path: &SanitizedPath,
5306) -> (
5307 HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
5308 Option<Arc<Gitignore>>,
5309 Option<(PathBuf, WorkDirectory)>,
5310) {
5311 let mut exclude = None;
5312 let mut ignores = HashMap::default();
5313 for (index, ancestor) in root_abs_path.as_path().ancestors().enumerate() {
5314 if index != 0 {
5315 if ancestor == paths::home_dir() {
5316 // Unless $HOME is itself the worktree root, don't consider it as a
5317 // containing git repository---expensive and likely unwanted.
5318 break;
5319 } else if let Ok(ignore) = build_gitignore(&ancestor.join(GITIGNORE), fs.as_ref()).await
5320 {
5321 ignores.insert(ancestor.into(), (ignore.into(), false));
5322 }
5323 }
5324
5325 let ancestor_dot_git = ancestor.join(DOT_GIT);
5326 log::trace!("considering ancestor: {ancestor_dot_git:?}");
5327 // Check whether the directory or file called `.git` exists (in the
5328 // case of worktrees it's a file.)
5329 if fs
5330 .metadata(&ancestor_dot_git)
5331 .await
5332 .is_ok_and(|metadata| metadata.is_some())
5333 {
5334 if index != 0 {
5335 // We canonicalize, since the FS events use the canonicalized path.
5336 if let Some(ancestor_dot_git) = fs.canonicalize(&ancestor_dot_git).await.log_err() {
5337 let location_in_repo = root_abs_path
5338 .as_path()
5339 .strip_prefix(ancestor)
5340 .unwrap()
5341 .into();
5342 log::info!("inserting parent git repo for this worktree: {location_in_repo:?}");
5343 // We associate the external git repo with our root folder and
5344 // also mark where in the git repo the root folder is located.
5345 return (
5346 ignores,
5347 exclude,
5348 Some((
5349 ancestor_dot_git,
5350 WorkDirectory::AboveProject {
5351 absolute_path: ancestor.into(),
5352 location_in_repo,
5353 },
5354 )),
5355 );
5356 };
5357 }
5358
5359 let repo_exclude_abs_path = ancestor_dot_git.join(REPO_EXCLUDE);
5360 if let Ok(repo_exclude) = build_gitignore(&repo_exclude_abs_path, fs.as_ref()).await {
5361 exclude = Some(Arc::new(repo_exclude));
5362 }
5363
5364 // Reached root of git repository.
5365 break;
5366 }
5367 }
5368
5369 (ignores, exclude, None)
5370}
5371
5372fn merge_event_roots(changed_paths: &[Arc<RelPath>], event_roots: &[EventRoot]) -> Vec<EventRoot> {
5373 let mut merged_event_roots = Vec::with_capacity(changed_paths.len() + event_roots.len());
5374 let mut changed_paths = changed_paths.iter().peekable();
5375 let mut event_roots = event_roots.iter().peekable();
5376 while let (Some(path), Some(event_root)) = (changed_paths.peek(), event_roots.peek()) {
5377 match path.cmp(&&event_root.path) {
5378 Ordering::Less => {
5379 merged_event_roots.push(EventRoot {
5380 path: (*changed_paths.next().expect("peeked changed path")).clone(),
5381 was_rescanned: false,
5382 });
5383 }
5384 Ordering::Equal => {
5385 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5386 changed_paths.next();
5387 }
5388 Ordering::Greater => {
5389 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5390 }
5391 }
5392 }
5393 merged_event_roots.extend(changed_paths.map(|path| EventRoot {
5394 path: path.clone(),
5395 was_rescanned: false,
5396 }));
5397 merged_event_roots.extend(event_roots.cloned());
5398 merged_event_roots
5399}
5400
5401fn build_diff(
5402 phase: BackgroundScannerPhase,
5403 old_snapshot: &Snapshot,
5404 new_snapshot: &Snapshot,
5405 event_roots: &[EventRoot],
5406) -> UpdatedEntriesSet {
5407 use BackgroundScannerPhase::*;
5408 use PathChange::{Added, AddedOrUpdated, Loaded, Removed, Updated};
5409
5410 // Identify which paths have changed. Use the known set of changed
5411 // parent paths to optimize the search.
5412 let mut changes = Vec::new();
5413
5414 let mut old_paths = old_snapshot.entries_by_path.cursor::<PathKey>(());
5415 let mut new_paths = new_snapshot.entries_by_path.cursor::<PathKey>(());
5416 let mut last_newly_loaded_dir_path = None;
5417 old_paths.next();
5418 new_paths.next();
5419 for event_root in event_roots {
5420 let path = PathKey(event_root.path.clone());
5421 if old_paths.item().is_some_and(|e| e.path < path.0) {
5422 old_paths.seek_forward(&path, Bias::Left);
5423 }
5424 if new_paths.item().is_some_and(|e| e.path < path.0) {
5425 new_paths.seek_forward(&path, Bias::Left);
5426 }
5427 loop {
5428 match (old_paths.item(), new_paths.item()) {
5429 (Some(old_entry), Some(new_entry)) => {
5430 if old_entry.path > path.0
5431 && new_entry.path > path.0
5432 && !old_entry.path.starts_with(&path.0)
5433 && !new_entry.path.starts_with(&path.0)
5434 {
5435 break;
5436 }
5437
5438 match Ord::cmp(&old_entry.path, &new_entry.path) {
5439 Ordering::Less => {
5440 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5441 old_paths.next();
5442 }
5443 Ordering::Equal => {
5444 if phase == EventsReceivedDuringInitialScan {
5445 if old_entry.id != new_entry.id {
5446 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5447 }
5448 // If the worktree was not fully initialized when this event was generated,
5449 // we can't know whether this entry was added during the scan or whether
5450 // it was merely updated.
5451 changes.push((
5452 new_entry.path.clone(),
5453 new_entry.id,
5454 AddedOrUpdated,
5455 ));
5456 } else if old_entry.id != new_entry.id {
5457 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5458 changes.push((new_entry.path.clone(), new_entry.id, Added));
5459 } else if old_entry != new_entry {
5460 if old_entry.kind.is_unloaded() {
5461 last_newly_loaded_dir_path = Some(&new_entry.path);
5462 changes.push((new_entry.path.clone(), new_entry.id, Loaded));
5463 } else {
5464 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5465 }
5466 } else if event_root.was_rescanned {
5467 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5468 }
5469 old_paths.next();
5470 new_paths.next();
5471 }
5472 Ordering::Greater => {
5473 let is_newly_loaded = phase == InitialScan
5474 || last_newly_loaded_dir_path
5475 .as_ref()
5476 .is_some_and(|dir| new_entry.path.starts_with(dir));
5477 changes.push((
5478 new_entry.path.clone(),
5479 new_entry.id,
5480 if is_newly_loaded { Loaded } else { Added },
5481 ));
5482 new_paths.next();
5483 }
5484 }
5485 }
5486 (Some(old_entry), None) => {
5487 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5488 old_paths.next();
5489 }
5490 (None, Some(new_entry)) => {
5491 let is_newly_loaded = phase == InitialScan
5492 || last_newly_loaded_dir_path
5493 .as_ref()
5494 .is_some_and(|dir| new_entry.path.starts_with(dir));
5495 changes.push((
5496 new_entry.path.clone(),
5497 new_entry.id,
5498 if is_newly_loaded { Loaded } else { Added },
5499 ));
5500 new_paths.next();
5501 }
5502 (None, None) => break,
5503 }
5504 }
5505 }
5506
5507 changes.into()
5508}
5509
5510fn swap_to_front(child_paths: &mut Vec<PathBuf>, file: &str) {
5511 let position = child_paths
5512 .iter()
5513 .position(|path| path.file_name().unwrap() == file);
5514 if let Some(position) = position {
5515 let temp = child_paths.remove(position);
5516 child_paths.insert(0, temp);
5517 }
5518}
5519
5520fn char_bag_for_path(root_char_bag: CharBag, path: &RelPath) -> CharBag {
5521 let mut result = root_char_bag;
5522 result.extend(path.as_unix_str().chars().map(|c| c.to_ascii_lowercase()));
5523 result
5524}
5525
5526#[derive(Debug)]
5527struct ScanJob {
5528 abs_path: Arc<Path>,
5529 path: Arc<RelPath>,
5530 ignore_stack: IgnoreStack,
5531 scan_queue: Sender<ScanJob>,
5532 ancestor_inodes: TreeSet<u64>,
5533 is_external: bool,
5534}
5535
5536struct UpdateIgnoreStatusJob {
5537 abs_path: Arc<Path>,
5538 ignore_stack: IgnoreStack,
5539 ignore_queue: Sender<UpdateIgnoreStatusJob>,
5540 scan_queue: Sender<ScanJob>,
5541}
5542
5543pub trait WorktreeModelHandle {
5544 #[cfg(feature = "test-support")]
5545 fn flush_fs_events<'a>(
5546 &self,
5547 cx: &'a mut gpui::TestAppContext,
5548 ) -> futures::future::LocalBoxFuture<'a, ()>;
5549
5550 #[cfg(feature = "test-support")]
5551 fn flush_fs_events_in_root_git_repository<'a>(
5552 &self,
5553 cx: &'a mut gpui::TestAppContext,
5554 ) -> futures::future::LocalBoxFuture<'a, ()>;
5555}
5556
5557impl WorktreeModelHandle for Entity<Worktree> {
5558 // When the worktree's FS event stream sometimes delivers "redundant" events for FS changes that
5559 // occurred before the worktree was constructed. These events can cause the worktree to perform
5560 // extra directory scans, and emit extra scan-state notifications.
5561 //
5562 // This function mutates the worktree's directory and waits for those mutations to be picked up,
5563 // to ensure that all redundant FS events have already been processed.
5564 #[cfg(feature = "test-support")]
5565 fn flush_fs_events<'a>(
5566 &self,
5567 cx: &'a mut gpui::TestAppContext,
5568 ) -> futures::future::LocalBoxFuture<'a, ()> {
5569 let file_name = "fs-event-sentinel";
5570
5571 let tree = self.clone();
5572 let (fs, root_path) = self.read_with(cx, |tree, _| {
5573 let tree = tree.as_local().unwrap();
5574 (tree.fs.clone(), tree.abs_path.clone())
5575 });
5576
5577 async move {
5578 // Subscribe to events BEFORE creating the file to avoid race condition
5579 // where events fire before subscription is set up
5580 let mut events = cx.events(&tree);
5581
5582 fs.create_file(&root_path.join(file_name), Default::default())
5583 .await
5584 .unwrap();
5585
5586 // Check if condition is already met before waiting for events
5587 let file_exists = || {
5588 tree.read_with(cx, |tree, _| {
5589 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5590 .is_some()
5591 })
5592 };
5593
5594 // Use select to avoid blocking indefinitely if events are delayed
5595 while !file_exists() {
5596 futures::select_biased! {
5597 _ = events.next() => {}
5598 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5599 }
5600 }
5601
5602 fs.remove_file(&root_path.join(file_name), Default::default())
5603 .await
5604 .unwrap();
5605
5606 // Check if condition is already met before waiting for events
5607 let file_gone = || {
5608 tree.read_with(cx, |tree, _| {
5609 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5610 .is_none()
5611 })
5612 };
5613
5614 // Use select to avoid blocking indefinitely if events are delayed
5615 while !file_gone() {
5616 futures::select_biased! {
5617 _ = events.next() => {}
5618 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5619 }
5620 }
5621
5622 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5623 .await;
5624 }
5625 .boxed_local()
5626 }
5627
5628 // This function is similar to flush_fs_events, except that it waits for events to be flushed in
5629 // the .git folder of the root repository.
5630 // The reason for its existence is that a repository's .git folder might live *outside* of the
5631 // worktree and thus its FS events might go through a different path.
5632 // In order to flush those, we need to create artificial events in the .git folder and wait
5633 // for the repository to be reloaded.
5634 #[cfg(feature = "test-support")]
5635 fn flush_fs_events_in_root_git_repository<'a>(
5636 &self,
5637 cx: &'a mut gpui::TestAppContext,
5638 ) -> futures::future::LocalBoxFuture<'a, ()> {
5639 let file_name = "fs-event-sentinel";
5640
5641 let tree = self.clone();
5642 let (fs, root_path, mut git_dir_scan_id) = self.read_with(cx, |tree, _| {
5643 let tree = tree.as_local().unwrap();
5644 let local_repo_entry = tree
5645 .git_repositories
5646 .values()
5647 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5648 .unwrap();
5649 (
5650 tree.fs.clone(),
5651 local_repo_entry.common_dir_abs_path.clone(),
5652 local_repo_entry.git_dir_scan_id,
5653 )
5654 });
5655
5656 let scan_id_increased = |tree: &mut Worktree, git_dir_scan_id: &mut usize| {
5657 let tree = tree.as_local().unwrap();
5658 // let repository = tree.repositories.first().unwrap();
5659 let local_repo_entry = tree
5660 .git_repositories
5661 .values()
5662 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5663 .unwrap();
5664
5665 if local_repo_entry.git_dir_scan_id > *git_dir_scan_id {
5666 *git_dir_scan_id = local_repo_entry.git_dir_scan_id;
5667 true
5668 } else {
5669 false
5670 }
5671 };
5672
5673 async move {
5674 // Subscribe to events BEFORE creating the file to avoid race condition
5675 // where events fire before subscription is set up
5676 let mut events = cx.events(&tree);
5677
5678 fs.create_file(&root_path.join(file_name), Default::default())
5679 .await
5680 .unwrap();
5681
5682 // Use select to avoid blocking indefinitely if events are delayed
5683 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5684 futures::select_biased! {
5685 _ = events.next() => {}
5686 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5687 }
5688 }
5689
5690 fs.remove_file(&root_path.join(file_name), Default::default())
5691 .await
5692 .unwrap();
5693
5694 // Use select to avoid blocking indefinitely if events are delayed
5695 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5696 futures::select_biased! {
5697 _ = events.next() => {}
5698 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5699 }
5700 }
5701
5702 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5703 .await;
5704 }
5705 .boxed_local()
5706 }
5707}
5708
5709#[derive(Clone, Debug)]
5710struct TraversalProgress<'a> {
5711 max_path: &'a RelPath,
5712 count: usize,
5713 non_ignored_count: usize,
5714 file_count: usize,
5715 non_ignored_file_count: usize,
5716}
5717
5718impl TraversalProgress<'_> {
5719 fn count(&self, include_files: bool, include_dirs: bool, include_ignored: bool) -> usize {
5720 match (include_files, include_dirs, include_ignored) {
5721 (true, true, true) => self.count,
5722 (true, true, false) => self.non_ignored_count,
5723 (true, false, true) => self.file_count,
5724 (true, false, false) => self.non_ignored_file_count,
5725 (false, true, true) => self.count - self.file_count,
5726 (false, true, false) => self.non_ignored_count - self.non_ignored_file_count,
5727 (false, false, _) => 0,
5728 }
5729 }
5730}
5731
5732impl<'a> sum_tree::Dimension<'a, EntrySummary> for TraversalProgress<'a> {
5733 fn zero(_cx: ()) -> Self {
5734 Default::default()
5735 }
5736
5737 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
5738 self.max_path = summary.max_path.as_ref();
5739 self.count += summary.count;
5740 self.non_ignored_count += summary.non_ignored_count;
5741 self.file_count += summary.file_count;
5742 self.non_ignored_file_count += summary.non_ignored_file_count;
5743 }
5744}
5745
5746impl Default for TraversalProgress<'_> {
5747 fn default() -> Self {
5748 Self {
5749 max_path: RelPath::empty(),
5750 count: 0,
5751 non_ignored_count: 0,
5752 file_count: 0,
5753 non_ignored_file_count: 0,
5754 }
5755 }
5756}
5757
5758#[derive(Debug)]
5759pub struct Traversal<'a> {
5760 snapshot: &'a Snapshot,
5761 cursor: sum_tree::Cursor<'a, 'static, Entry, TraversalProgress<'a>>,
5762 include_ignored: bool,
5763 include_files: bool,
5764 include_dirs: bool,
5765}
5766
5767impl<'a> Traversal<'a> {
5768 fn new(
5769 snapshot: &'a Snapshot,
5770 include_files: bool,
5771 include_dirs: bool,
5772 include_ignored: bool,
5773 start_path: &RelPath,
5774 ) -> Self {
5775 let mut cursor = snapshot.entries_by_path.cursor(());
5776 cursor.seek(&TraversalTarget::path(start_path), Bias::Left);
5777 let mut traversal = Self {
5778 snapshot,
5779 cursor,
5780 include_files,
5781 include_dirs,
5782 include_ignored,
5783 };
5784 if traversal.end_offset() == traversal.start_offset() {
5785 traversal.next();
5786 }
5787 traversal
5788 }
5789
5790 pub fn advance(&mut self) -> bool {
5791 self.advance_by(1)
5792 }
5793
5794 pub fn advance_by(&mut self, count: usize) -> bool {
5795 self.cursor.seek_forward(
5796 &TraversalTarget::Count {
5797 count: self.end_offset() + count,
5798 include_dirs: self.include_dirs,
5799 include_files: self.include_files,
5800 include_ignored: self.include_ignored,
5801 },
5802 Bias::Left,
5803 )
5804 }
5805
5806 pub fn advance_to_sibling(&mut self) -> bool {
5807 while let Some(entry) = self.cursor.item() {
5808 self.cursor
5809 .seek_forward(&TraversalTarget::successor(&entry.path), Bias::Left);
5810 if let Some(entry) = self.cursor.item()
5811 && (self.include_files || !entry.is_file())
5812 && (self.include_dirs || !entry.is_dir())
5813 && (self.include_ignored || !entry.is_ignored || entry.is_always_included)
5814 {
5815 return true;
5816 }
5817 }
5818 false
5819 }
5820
5821 pub fn back_to_parent(&mut self) -> bool {
5822 let Some(parent_path) = self.cursor.item().and_then(|entry| entry.path.parent()) else {
5823 return false;
5824 };
5825 self.cursor
5826 .seek(&TraversalTarget::path(parent_path), Bias::Left)
5827 }
5828
5829 pub fn entry(&self) -> Option<&'a Entry> {
5830 self.cursor.item()
5831 }
5832
5833 pub fn snapshot(&self) -> &'a Snapshot {
5834 self.snapshot
5835 }
5836
5837 pub fn start_offset(&self) -> usize {
5838 self.cursor
5839 .start()
5840 .count(self.include_files, self.include_dirs, self.include_ignored)
5841 }
5842
5843 pub fn end_offset(&self) -> usize {
5844 self.cursor
5845 .end()
5846 .count(self.include_files, self.include_dirs, self.include_ignored)
5847 }
5848}
5849
5850impl<'a> Iterator for Traversal<'a> {
5851 type Item = &'a Entry;
5852
5853 fn next(&mut self) -> Option<Self::Item> {
5854 if let Some(item) = self.entry() {
5855 self.advance();
5856 Some(item)
5857 } else {
5858 None
5859 }
5860 }
5861}
5862
5863#[derive(Debug, Clone, Copy)]
5864pub enum PathTarget<'a> {
5865 Path(&'a RelPath),
5866 Successor(&'a RelPath),
5867}
5868
5869impl PathTarget<'_> {
5870 fn cmp_path(&self, other: &RelPath) -> Ordering {
5871 match self {
5872 PathTarget::Path(path) => path.cmp(&other),
5873 PathTarget::Successor(path) => {
5874 if other.starts_with(path) {
5875 Ordering::Greater
5876 } else {
5877 Ordering::Equal
5878 }
5879 }
5880 }
5881 }
5882}
5883
5884impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, PathProgress<'a>> for PathTarget<'_> {
5885 fn cmp(&self, cursor_location: &PathProgress<'a>, _: S::Context<'_>) -> Ordering {
5886 self.cmp_path(cursor_location.max_path)
5887 }
5888}
5889
5890impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, TraversalProgress<'a>> for PathTarget<'_> {
5891 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: S::Context<'_>) -> Ordering {
5892 self.cmp_path(cursor_location.max_path)
5893 }
5894}
5895
5896#[derive(Debug)]
5897enum TraversalTarget<'a> {
5898 Path(PathTarget<'a>),
5899 Count {
5900 count: usize,
5901 include_files: bool,
5902 include_ignored: bool,
5903 include_dirs: bool,
5904 },
5905}
5906
5907impl<'a> TraversalTarget<'a> {
5908 fn path(path: &'a RelPath) -> Self {
5909 Self::Path(PathTarget::Path(path))
5910 }
5911
5912 fn successor(path: &'a RelPath) -> Self {
5913 Self::Path(PathTarget::Successor(path))
5914 }
5915
5916 fn cmp_progress(&self, progress: &TraversalProgress) -> Ordering {
5917 match self {
5918 TraversalTarget::Path(path) => path.cmp_path(progress.max_path),
5919 TraversalTarget::Count {
5920 count,
5921 include_files,
5922 include_dirs,
5923 include_ignored,
5924 } => Ord::cmp(
5925 count,
5926 &progress.count(*include_files, *include_dirs, *include_ignored),
5927 ),
5928 }
5929 }
5930}
5931
5932impl<'a> SeekTarget<'a, EntrySummary, TraversalProgress<'a>> for TraversalTarget<'_> {
5933 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5934 self.cmp_progress(cursor_location)
5935 }
5936}
5937
5938impl<'a> SeekTarget<'a, PathSummary<sum_tree::NoSummary>, TraversalProgress<'a>>
5939 for TraversalTarget<'_>
5940{
5941 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5942 self.cmp_progress(cursor_location)
5943 }
5944}
5945
5946pub struct ChildEntriesOptions {
5947 pub include_files: bool,
5948 pub include_dirs: bool,
5949 pub include_ignored: bool,
5950}
5951
5952pub struct ChildEntriesIter<'a> {
5953 parent_path: &'a RelPath,
5954 traversal: Traversal<'a>,
5955}
5956
5957impl<'a> Iterator for ChildEntriesIter<'a> {
5958 type Item = &'a Entry;
5959
5960 fn next(&mut self) -> Option<Self::Item> {
5961 if let Some(item) = self.traversal.entry()
5962 && item.path.starts_with(self.parent_path)
5963 {
5964 self.traversal.advance_to_sibling();
5965 return Some(item);
5966 }
5967 None
5968 }
5969}
5970
5971impl<'a> From<&'a Entry> for proto::Entry {
5972 fn from(entry: &'a Entry) -> Self {
5973 Self {
5974 id: entry.id.to_proto(),
5975 is_dir: entry.is_dir(),
5976 path: entry.path.as_ref().to_proto(),
5977 inode: entry.inode,
5978 mtime: entry.mtime.map(|time| time.into()),
5979 is_ignored: entry.is_ignored,
5980 is_hidden: entry.is_hidden,
5981 is_external: entry.is_external,
5982 is_fifo: entry.is_fifo,
5983 size: Some(entry.size),
5984 canonical_path: entry
5985 .canonical_path
5986 .as_ref()
5987 .map(|path| path.to_string_lossy().into_owned()),
5988 }
5989 }
5990}
5991
5992impl TryFrom<(&CharBag, &PathMatcher, proto::Entry)> for Entry {
5993 type Error = anyhow::Error;
5994
5995 fn try_from(
5996 (root_char_bag, always_included, entry): (&CharBag, &PathMatcher, proto::Entry),
5997 ) -> Result<Self> {
5998 let kind = if entry.is_dir {
5999 EntryKind::Dir
6000 } else {
6001 EntryKind::File
6002 };
6003
6004 let path =
6005 RelPath::from_proto(&entry.path).context("invalid relative path in proto message")?;
6006 let char_bag = char_bag_for_path(*root_char_bag, &path);
6007 let is_always_included = always_included.is_match(&path);
6008 Ok(Entry {
6009 id: ProjectEntryId::from_proto(entry.id),
6010 kind,
6011 path,
6012 inode: entry.inode,
6013 mtime: entry.mtime.map(|time| time.into()),
6014 size: entry.size.unwrap_or(0),
6015 canonical_path: entry
6016 .canonical_path
6017 .map(|path_string| Arc::from(PathBuf::from(path_string))),
6018 is_ignored: entry.is_ignored,
6019 is_hidden: entry.is_hidden,
6020 is_always_included,
6021 is_external: entry.is_external,
6022 is_private: false,
6023 char_bag,
6024 is_fifo: entry.is_fifo,
6025 })
6026 }
6027}
6028
6029#[derive(Clone, Copy, Debug, Default, Hash, PartialEq, Eq, PartialOrd, Ord)]
6030pub struct ProjectEntryId(usize);
6031
6032impl ProjectEntryId {
6033 pub const MAX: Self = Self(usize::MAX);
6034 pub const MIN: Self = Self(usize::MIN);
6035
6036 pub fn new(counter: &AtomicUsize) -> Self {
6037 Self(counter.fetch_add(1, SeqCst))
6038 }
6039
6040 pub fn from_proto(id: u64) -> Self {
6041 Self(id as usize)
6042 }
6043
6044 pub fn to_proto(self) -> u64 {
6045 self.0 as u64
6046 }
6047
6048 pub fn from_usize(id: usize) -> Self {
6049 ProjectEntryId(id)
6050 }
6051
6052 pub fn to_usize(self) -> usize {
6053 self.0
6054 }
6055}
6056
6057#[cfg(feature = "test-support")]
6058impl CreatedEntry {
6059 pub fn into_included(self) -> Option<Entry> {
6060 match self {
6061 CreatedEntry::Included(entry) => Some(entry),
6062 CreatedEntry::Excluded { .. } => None,
6063 }
6064 }
6065}
6066
6067fn parse_gitfile(content: &str) -> anyhow::Result<&Path> {
6068 let path = content
6069 .strip_prefix("gitdir:")
6070 .with_context(|| format!("parsing gitfile content {content:?}"))?;
6071 Ok(Path::new(path.trim()))
6072}
6073
6074async fn discover_git_paths(dot_git_abs_path: &Arc<Path>, fs: &dyn Fs) -> (Arc<Path>, Arc<Path>) {
6075 let mut repository_dir_abs_path = dot_git_abs_path.clone();
6076 let mut common_dir_abs_path = dot_git_abs_path.clone();
6077
6078 if let Some(path) = fs
6079 .load(dot_git_abs_path)
6080 .await
6081 .ok()
6082 .as_ref()
6083 .and_then(|contents| parse_gitfile(contents).log_err())
6084 {
6085 let path = dot_git_abs_path
6086 .parent()
6087 .unwrap_or(Path::new(""))
6088 .join(path);
6089 if let Some(path) = fs.canonicalize(&path).await.log_err() {
6090 repository_dir_abs_path = Path::new(&path).into();
6091 common_dir_abs_path = repository_dir_abs_path.clone();
6092
6093 if let Some(commondir_contents) = fs.load(&path.join("commondir")).await.ok()
6094 && let Some(commondir_path) = fs
6095 .canonicalize(&path.join(commondir_contents.trim()))
6096 .await
6097 .log_err()
6098 {
6099 common_dir_abs_path = commondir_path.as_path().into();
6100 }
6101 }
6102 };
6103 (repository_dir_abs_path, common_dir_abs_path)
6104}
6105
6106struct NullWatcher;
6107
6108impl fs::Watcher for NullWatcher {
6109 fn add(&self, _path: &Path) -> Result<()> {
6110 Ok(())
6111 }
6112
6113 fn remove(&self, _path: &Path) -> Result<()> {
6114 Ok(())
6115 }
6116}
6117
6118const FILE_ANALYSIS_BYTES: usize = 1024;
6119
6120async fn decode_file_text(
6121 fs: &dyn Fs,
6122 abs_path: &Path,
6123) -> Result<(String, &'static Encoding, bool)> {
6124 let mut file = fs
6125 .open_sync(&abs_path)
6126 .await
6127 .with_context(|| format!("opening file {abs_path:?}"))?;
6128
6129 // First, read the beginning of the file to determine its kind and encoding.
6130 // We do not want to load an entire large blob into memory only to discard it.
6131 let mut file_first_bytes = Vec::with_capacity(FILE_ANALYSIS_BYTES);
6132 let mut buf = [0u8; FILE_ANALYSIS_BYTES];
6133 let mut reached_eof = false;
6134 loop {
6135 if file_first_bytes.len() >= FILE_ANALYSIS_BYTES {
6136 break;
6137 }
6138 let n = file
6139 .read(&mut buf)
6140 .with_context(|| format!("reading bytes of the file {abs_path:?}"))?;
6141 if n == 0 {
6142 reached_eof = true;
6143 break;
6144 }
6145 file_first_bytes.extend_from_slice(&buf[..n]);
6146 }
6147 let (bom_encoding, byte_content) = decode_byte_header(&file_first_bytes);
6148 anyhow::ensure!(
6149 byte_content != ByteContent::Binary,
6150 "Binary files are not supported"
6151 );
6152
6153 // If the file is eligible for opening, read the rest of the file.
6154 let mut content = file_first_bytes;
6155 if !reached_eof {
6156 let mut buf = [0u8; 8 * 1024];
6157 loop {
6158 let n = file
6159 .read(&mut buf)
6160 .with_context(|| format!("reading remaining bytes of the file {abs_path:?}"))?;
6161 if n == 0 {
6162 break;
6163 }
6164 content.extend_from_slice(&buf[..n]);
6165 }
6166 }
6167 decode_byte_full(content, bom_encoding, byte_content)
6168}
6169
6170fn decode_byte_header(prefix: &[u8]) -> (Option<&'static Encoding>, ByteContent) {
6171 if let Some((encoding, _bom_len)) = Encoding::for_bom(prefix) {
6172 return (Some(encoding), ByteContent::Unknown);
6173 }
6174 (None, analyze_byte_content(prefix))
6175}
6176
6177fn decode_byte_full(
6178 bytes: Vec<u8>,
6179 bom_encoding: Option<&'static Encoding>,
6180 byte_content: ByteContent,
6181) -> Result<(String, &'static Encoding, bool)> {
6182 if let Some(encoding) = bom_encoding {
6183 let (cow, _) = encoding.decode_with_bom_removal(&bytes);
6184 return Ok((cow.into_owned(), encoding, true));
6185 }
6186
6187 match byte_content {
6188 ByteContent::Utf16Le => {
6189 let encoding = encoding_rs::UTF_16LE;
6190 let (cow, _, _) = encoding.decode(&bytes);
6191 return Ok((cow.into_owned(), encoding, false));
6192 }
6193 ByteContent::Utf16Be => {
6194 let encoding = encoding_rs::UTF_16BE;
6195 let (cow, _, _) = encoding.decode(&bytes);
6196 return Ok((cow.into_owned(), encoding, false));
6197 }
6198 ByteContent::Binary => {
6199 anyhow::bail!("Binary files are not supported");
6200 }
6201 ByteContent::Unknown => {}
6202 }
6203
6204 fn detect_encoding(bytes: Vec<u8>) -> (String, &'static Encoding) {
6205 let mut detector = EncodingDetector::new();
6206 detector.feed(&bytes, true);
6207
6208 let encoding = detector.guess(None, true); // Use None for TLD hint to ensure neutral detection logic.
6209
6210 let (cow, _, _) = encoding.decode(&bytes);
6211 (cow.into_owned(), encoding)
6212 }
6213
6214 match String::from_utf8(bytes) {
6215 Ok(text) => {
6216 // ISO-2022-JP (and other ISO-2022 variants) consists entirely of 7-bit ASCII bytes,
6217 // so it is valid UTF-8. However, it contains escape sequences starting with '\x1b'.
6218 // If we find an escape character, we double-check the encoding to prevent
6219 // displaying raw escape sequences instead of the correct characters.
6220 if text.contains('\x1b') {
6221 let (s, enc) = detect_encoding(text.into_bytes());
6222 Ok((s, enc, false))
6223 } else {
6224 Ok((text, encoding_rs::UTF_8, false))
6225 }
6226 }
6227 Err(e) => {
6228 let (s, enc) = detect_encoding(e.into_bytes());
6229 Ok((s, enc, false))
6230 }
6231 }
6232}
6233
6234#[derive(Debug, PartialEq)]
6235enum ByteContent {
6236 Utf16Le,
6237 Utf16Be,
6238 Binary,
6239 Unknown,
6240}
6241
6242// Heuristic check using null byte distribution plus a generic text-likeness
6243// heuristic. This prefers UTF-16 when many bytes are NUL and otherwise
6244// distinguishes between text-like and binary-like content.
6245fn analyze_byte_content(bytes: &[u8]) -> ByteContent {
6246 if bytes.len() < 2 {
6247 return ByteContent::Unknown;
6248 }
6249
6250 if is_known_binary_header(bytes) {
6251 return ByteContent::Binary;
6252 }
6253
6254 let limit = bytes.len().min(FILE_ANALYSIS_BYTES);
6255 let mut even_null_count = 0usize;
6256 let mut odd_null_count = 0usize;
6257 let mut non_text_like_count = 0usize;
6258
6259 for (i, &byte) in bytes[..limit].iter().enumerate() {
6260 if byte == 0 {
6261 if i % 2 == 0 {
6262 even_null_count += 1;
6263 } else {
6264 odd_null_count += 1;
6265 }
6266 non_text_like_count += 1;
6267 continue;
6268 }
6269
6270 let is_text_like = match byte {
6271 b'\t' | b'\n' | b'\r' | 0x0C => true,
6272 0x20..=0x7E => true,
6273 // Treat bytes that are likely part of UTF-8 or single-byte encodings as text-like.
6274 0x80..=0xBF | 0xC2..=0xF4 => true,
6275 _ => false,
6276 };
6277
6278 if !is_text_like {
6279 non_text_like_count += 1;
6280 }
6281 }
6282
6283 let total_null_count = even_null_count + odd_null_count;
6284
6285 // If there are no NUL bytes at all, this is overwhelmingly likely to be text.
6286 if total_null_count == 0 {
6287 return ByteContent::Unknown;
6288 }
6289
6290 let has_significant_nulls = total_null_count >= limit / 16;
6291 let nulls_skew_to_even = even_null_count > odd_null_count * 4;
6292 let nulls_skew_to_odd = odd_null_count > even_null_count * 4;
6293
6294 if has_significant_nulls {
6295 let sample = &bytes[..limit];
6296
6297 // UTF-16BE ASCII: [0x00, char] — nulls at even positions (high byte first)
6298 // UTF-16LE ASCII: [char, 0x00] — nulls at odd positions (low byte first)
6299
6300 if nulls_skew_to_even && is_plausible_utf16_text(sample, false) {
6301 return ByteContent::Utf16Be;
6302 }
6303
6304 if nulls_skew_to_odd && is_plausible_utf16_text(sample, true) {
6305 return ByteContent::Utf16Le;
6306 }
6307
6308 return ByteContent::Binary;
6309 }
6310
6311 if non_text_like_count * 100 < limit * 8 {
6312 ByteContent::Unknown
6313 } else {
6314 ByteContent::Binary
6315 }
6316}
6317
6318fn is_known_binary_header(bytes: &[u8]) -> bool {
6319 bytes.starts_with(b"%PDF-") // PDF
6320 || bytes.starts_with(b"PK\x03\x04") // ZIP local header
6321 || bytes.starts_with(b"PK\x05\x06") // ZIP end of central directory
6322 || bytes.starts_with(b"PK\x07\x08") // ZIP spanning/splitting
6323 || bytes.starts_with(b"\x89PNG\r\n\x1a\n") // PNG
6324 || bytes.starts_with(b"\xFF\xD8\xFF") // JPEG
6325 || bytes.starts_with(b"GIF87a") // GIF87a
6326 || bytes.starts_with(b"GIF89a") // GIF89a
6327 || bytes.starts_with(b"IWAD") // Doom IWAD archive
6328 || bytes.starts_with(b"PWAD") // Doom PWAD archive
6329 || bytes.starts_with(b"RIFF") // WAV, AVI, WebP
6330 || bytes.starts_with(b"OggS") // OGG (Vorbis, Opus, FLAC)
6331 || bytes.starts_with(b"fLaC") // FLAC
6332 || bytes.starts_with(b"ID3") // MP3 with ID3v2 tag
6333 || bytes.starts_with(b"\xFF\xFB") // MP3 frame sync (MPEG1 Layer3)
6334 || bytes.starts_with(b"\xFF\xFA") // MP3 frame sync (MPEG1 Layer3)
6335 || bytes.starts_with(b"\xFF\xF3") // MP3 frame sync (MPEG2 Layer3)
6336 || bytes.starts_with(b"\xFF\xF2") // MP3 frame sync (MPEG2 Layer3)
6337}
6338
6339// Null byte skew alone is not enough to identify UTF-16 -- binary formats with
6340// small 16-bit values (like PCM audio) produce the same pattern. Decode the
6341// bytes as UTF-16 and reject if too many code units land in control character
6342// ranges or form unpaired surrogates, which real text almost never contains.
6343fn is_plausible_utf16_text(bytes: &[u8], little_endian: bool) -> bool {
6344 let mut suspicious_count = 0usize;
6345 let mut total = 0usize;
6346
6347 let mut i = 0;
6348 while let Some(code_unit) = read_u16(bytes, i, little_endian) {
6349 total += 1;
6350
6351 match code_unit {
6352 0x0009 | 0x000A | 0x000C | 0x000D => {}
6353 // C0/C1 control characters and non-characters
6354 0x0000..=0x001F | 0x007F..=0x009F | 0xFFFE | 0xFFFF => suspicious_count += 1,
6355 0xD800..=0xDBFF => {
6356 let next_offset = i + 2;
6357 let has_low_surrogate = read_u16(bytes, next_offset, little_endian)
6358 .is_some_and(|next| (0xDC00..=0xDFFF).contains(&next));
6359 if has_low_surrogate {
6360 total += 1;
6361 i += 2;
6362 } else {
6363 suspicious_count += 1;
6364 }
6365 }
6366 // Lone low surrogate without a preceding high surrogate
6367 0xDC00..=0xDFFF => suspicious_count += 1,
6368 _ => {}
6369 }
6370
6371 i += 2;
6372 }
6373
6374 if total == 0 {
6375 return false;
6376 }
6377
6378 // Real UTF-16 text has near-zero control characters; binary data with
6379 // small 16-bit values typically exceeds 5%. 2% provides a safe margin.
6380 suspicious_count * 100 < total * 2
6381}
6382
6383fn read_u16(bytes: &[u8], offset: usize, little_endian: bool) -> Option<u16> {
6384 let pair = [*bytes.get(offset)?, *bytes.get(offset + 1)?];
6385 if little_endian {
6386 return Some(u16::from_le_bytes(pair));
6387 }
6388 Some(u16::from_be_bytes(pair))
6389}
6390
6391#[cfg(test)]
6392mod tests {
6393 use super::*;
6394
6395 /// reproduction of issue #50785
6396 fn build_pcm16_wav_bytes() -> Vec<u8> {
6397 let header: Vec<u8> = vec![
6398 /* RIFF header */
6399 0x52, 0x49, 0x46, 0x46, // "RIFF"
6400 0xc6, 0xcf, 0x00, 0x00, // file size: 8
6401 0x57, 0x41, 0x56, 0x45, // "WAVE"
6402 /* fmt chunk */
6403 0x66, 0x6d, 0x74, 0x20, // "fmt "
6404 0x10, 0x00, 0x00, 0x00, // chunk size: 16
6405 0x01, 0x00, // format: PCM (1)
6406 0x01, 0x00, // channels: 1 (mono)
6407 0x80, 0x3e, 0x00, 0x00, // sample rate: 16000
6408 0x00, 0x7d, 0x00, 0x00, // byte rate: 32000
6409 0x02, 0x00, // block align: 2
6410 0x10, 0x00, // bits per sample: 16
6411 /* LIST chunk */
6412 0x4c, 0x49, 0x53, 0x54, // "LIST"
6413 0x1a, 0x00, 0x00, 0x00, // chunk size: 26
6414 0x49, 0x4e, 0x46, 0x4f, // "INFO"
6415 0x49, 0x53, 0x46, 0x54, // "ISFT"
6416 0x0d, 0x00, 0x00, 0x00, // sub-chunk size: 13
6417 0x4c, 0x61, 0x76, 0x66, 0x36, 0x32, 0x2e, 0x33, // "Lavf62.3"
6418 0x2e, 0x31, 0x30, 0x30, 0x00, // ".100\0"
6419 /* padding byte for word alignment */
6420 0x00, // data chunk header
6421 0x64, 0x61, 0x74, 0x61, // "data"
6422 0x80, 0xcf, 0x00, 0x00, // chunk size
6423 ];
6424
6425 let mut bytes = header;
6426
6427 // fill remaining space up to `FILE_ANALYSIS_BYTES` with synthetic PCM
6428 let audio_bytes_needed = FILE_ANALYSIS_BYTES - bytes.len();
6429 for i in 0..(audio_bytes_needed / 2) {
6430 let sample = (i & 0xFF) as u8;
6431 bytes.push(sample); // low byte: varies
6432 bytes.push(0x00); // high byte: zero for small values
6433 }
6434
6435 bytes
6436 }
6437
6438 #[test]
6439 fn test_pcm16_wav_detected_as_binary() {
6440 let wav_bytes = build_pcm16_wav_bytes();
6441 assert_eq!(wav_bytes.len(), FILE_ANALYSIS_BYTES);
6442
6443 let result = analyze_byte_content(&wav_bytes);
6444 assert_eq!(
6445 result,
6446 ByteContent::Binary,
6447 "PCM 16-bit WAV should be detected as Binary via RIFF header"
6448 );
6449 }
6450
6451 #[test]
6452 fn test_le16_binary_not_misdetected_as_utf16le() {
6453 let mut bytes = b"FAKE".to_vec();
6454 while bytes.len() < FILE_ANALYSIS_BYTES {
6455 let sample = (bytes.len() & 0xFF) as u8;
6456 bytes.push(sample);
6457 bytes.push(0x00);
6458 }
6459 bytes.truncate(FILE_ANALYSIS_BYTES);
6460
6461 let result = analyze_byte_content(&bytes);
6462 assert_eq!(
6463 result,
6464 ByteContent::Binary,
6465 "LE 16-bit binary with control characters should be detected as Binary"
6466 );
6467 }
6468
6469 #[test]
6470 fn test_be16_binary_not_misdetected_as_utf16be() {
6471 let mut bytes = b"FAKE".to_vec();
6472 while bytes.len() < FILE_ANALYSIS_BYTES {
6473 bytes.push(0x00);
6474 let sample = (bytes.len() & 0xFF) as u8;
6475 bytes.push(sample);
6476 }
6477 bytes.truncate(FILE_ANALYSIS_BYTES);
6478
6479 let result = analyze_byte_content(&bytes);
6480 assert_eq!(
6481 result,
6482 ByteContent::Binary,
6483 "BE 16-bit binary with control characters should be detected as Binary"
6484 );
6485 }
6486
6487 #[test]
6488 fn test_utf16le_text_detected_as_utf16le() {
6489 let text = "Hello, world! This is a UTF-16 test string. ";
6490 let mut bytes = Vec::new();
6491 while bytes.len() < FILE_ANALYSIS_BYTES {
6492 bytes.extend(text.encode_utf16().flat_map(|u| u.to_le_bytes()));
6493 }
6494 bytes.truncate(FILE_ANALYSIS_BYTES);
6495
6496 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Le);
6497 }
6498
6499 #[test]
6500 fn test_utf16be_text_detected_as_utf16be() {
6501 let text = "Hello, world! This is a UTF-16 test string. ";
6502 let mut bytes = Vec::new();
6503 while bytes.len() < FILE_ANALYSIS_BYTES {
6504 bytes.extend(text.encode_utf16().flat_map(|u| u.to_be_bytes()));
6505 }
6506 bytes.truncate(FILE_ANALYSIS_BYTES);
6507
6508 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Be);
6509 }
6510
6511 #[test]
6512 fn test_known_binary_headers() {
6513 let cases: &[(&[u8], &str)] = &[
6514 (b"RIFF\x00\x00\x00\x00WAVE", "WAV"),
6515 (b"RIFF\x00\x00\x00\x00AVI ", "AVI"),
6516 (b"OggS\x00\x02", "OGG"),
6517 (b"fLaC\x00\x00", "FLAC"),
6518 (b"ID3\x03\x00", "MP3 ID3v2"),
6519 (b"\xFF\xFB\x90\x00", "MP3 MPEG1 Layer3"),
6520 (b"\xFF\xF3\x90\x00", "MP3 MPEG2 Layer3"),
6521 ];
6522
6523 for (header, label) in cases {
6524 let mut bytes = header.to_vec();
6525 bytes.resize(FILE_ANALYSIS_BYTES, 0x41); // pad with 'A'
6526 assert_eq!(
6527 analyze_byte_content(&bytes),
6528 ByteContent::Binary,
6529 "{label} should be detected as Binary"
6530 );
6531 }
6532 }
6533}