1mod ignore;
2mod worktree_settings;
3
4use ::ignore::gitignore::{Gitignore, GitignoreBuilder};
5use anyhow::{Context as _, Result, anyhow};
6use chardetng::EncodingDetector;
7use clock::ReplicaId;
8use collections::{HashMap, HashSet, VecDeque};
9use encoding_rs::Encoding;
10use fs::{Fs, MTime, PathEvent, RemoveOptions, Watcher, copy_recursive, read_dir_items};
11use futures::{
12 FutureExt as _, Stream, StreamExt,
13 channel::{
14 mpsc::{self, UnboundedSender},
15 oneshot,
16 },
17 select_biased, stream,
18 task::Poll,
19};
20use fuzzy::CharBag;
21use git::{
22 COMMIT_MESSAGE, DOT_GIT, FSMONITOR_DAEMON, GITIGNORE, INDEX_LOCK, LFS_DIR, REPO_EXCLUDE,
23 status::GitSummary,
24};
25use gpui::{
26 App, AppContext as _, AsyncApp, BackgroundExecutor, Context, Entity, EventEmitter, Priority,
27 Task,
28};
29use ignore::IgnoreStack;
30use language::DiskState;
31
32use parking_lot::Mutex;
33use paths::{local_settings_folder_name, local_vscode_folder_name};
34use postage::{
35 barrier,
36 prelude::{Sink as _, Stream as _},
37 watch,
38};
39use rpc::{
40 AnyProtoClient,
41 proto::{self, split_worktree_update},
42};
43pub use settings::WorktreeId;
44use settings::{Settings, SettingsLocation, SettingsStore};
45use smallvec::{SmallVec, smallvec};
46use smol::channel::{self, Sender};
47use std::{
48 any::Any,
49 borrow::Borrow as _,
50 cmp::Ordering,
51 collections::hash_map,
52 convert::TryFrom,
53 ffi::OsStr,
54 fmt,
55 future::Future,
56 mem::{self},
57 ops::{Deref, DerefMut, Range},
58 path::{Path, PathBuf},
59 pin::Pin,
60 sync::{
61 Arc,
62 atomic::{AtomicUsize, Ordering::SeqCst},
63 },
64 time::{Duration, Instant},
65};
66use sum_tree::{Bias, Dimensions, Edit, KeyedItem, SeekTarget, SumTree, Summary, TreeMap, TreeSet};
67use text::{LineEnding, Rope};
68use util::{
69 ResultExt, debug_panic, maybe,
70 paths::{PathMatcher, PathStyle, SanitizedPath, home_dir},
71 rel_path::RelPath,
72};
73pub use worktree_settings::WorktreeSettings;
74
75use crate::ignore::IgnoreKind;
76
77pub const FS_WATCH_LATENCY: Duration = Duration::from_millis(100);
78
79/// A set of local or remote files that are being opened as part of a project.
80/// Responsible for tracking related FS (for local)/collab (for remote) events and corresponding updates.
81/// Stores git repositories data and the diagnostics for the file(s).
82///
83/// Has an absolute path, and may be set to be visible in Zed UI or not.
84/// May correspond to a directory or a single file.
85/// Possible examples:
86/// * a drag and dropped file — may be added as an invisible, "ephemeral" entry to the current worktree
87/// * a directory opened in Zed — may be added as a visible entry to the current worktree
88///
89/// Uses [`Entry`] to track the state of each file/directory, can look up absolute paths for entries.
90pub enum Worktree {
91 Local(LocalWorktree),
92 Remote(RemoteWorktree),
93}
94
95/// An entry, created in the worktree.
96#[derive(Debug)]
97pub enum CreatedEntry {
98 /// Got created and indexed by the worktree, receiving a corresponding entry.
99 Included(Entry),
100 /// Got created, but not indexed due to falling under exclusion filters.
101 Excluded { abs_path: PathBuf },
102}
103
104#[derive(Debug)]
105pub struct LoadedFile {
106 pub file: Arc<File>,
107 pub text: String,
108 pub encoding: &'static Encoding,
109 pub has_bom: bool,
110}
111
112pub struct LoadedBinaryFile {
113 pub file: Arc<File>,
114 pub content: Vec<u8>,
115}
116
117impl fmt::Debug for LoadedBinaryFile {
118 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
119 f.debug_struct("LoadedBinaryFile")
120 .field("file", &self.file)
121 .field("content_bytes", &self.content.len())
122 .finish()
123 }
124}
125
126pub struct LocalWorktree {
127 snapshot: LocalSnapshot,
128 scan_requests_tx: channel::Sender<ScanRequest>,
129 path_prefixes_to_scan_tx: channel::Sender<PathPrefixScanRequest>,
130 is_scanning: (watch::Sender<bool>, watch::Receiver<bool>),
131 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
132 _background_scanner_tasks: Vec<Task<()>>,
133 update_observer: Option<UpdateObservationState>,
134 fs: Arc<dyn Fs>,
135 fs_case_sensitive: bool,
136 visible: bool,
137 next_entry_id: Arc<AtomicUsize>,
138 settings: WorktreeSettings,
139 share_private_files: bool,
140 scanning_enabled: bool,
141}
142
143pub struct PathPrefixScanRequest {
144 path: Arc<RelPath>,
145 done: SmallVec<[barrier::Sender; 1]>,
146}
147
148struct ScanRequest {
149 relative_paths: Vec<Arc<RelPath>>,
150 done: SmallVec<[barrier::Sender; 1]>,
151}
152
153pub struct RemoteWorktree {
154 snapshot: Snapshot,
155 background_snapshot: Arc<Mutex<(Snapshot, Vec<proto::UpdateWorktree>)>>,
156 project_id: u64,
157 client: AnyProtoClient,
158 file_scan_inclusions: PathMatcher,
159 updates_tx: Option<UnboundedSender<proto::UpdateWorktree>>,
160 update_observer: Option<mpsc::UnboundedSender<proto::UpdateWorktree>>,
161 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
162 replica_id: ReplicaId,
163 visible: bool,
164 disconnected: bool,
165}
166
167#[derive(Clone)]
168pub struct Snapshot {
169 id: WorktreeId,
170 /// The absolute path of the worktree root.
171 abs_path: Arc<SanitizedPath>,
172 path_style: PathStyle,
173 root_name: Arc<RelPath>,
174 root_char_bag: CharBag,
175 entries_by_path: SumTree<Entry>,
176 entries_by_id: SumTree<PathEntry>,
177 always_included_entries: Vec<Arc<RelPath>>,
178
179 /// A number that increases every time the worktree begins scanning
180 /// a set of paths from the filesystem. This scanning could be caused
181 /// by some operation performed on the worktree, such as reading or
182 /// writing a file, or by an event reported by the filesystem.
183 scan_id: usize,
184
185 /// The latest scan id that has completed, and whose preceding scans
186 /// have all completed. The current `scan_id` could be more than one
187 /// greater than the `completed_scan_id` if operations are performed
188 /// on the worktree while it is processing a file-system event.
189 completed_scan_id: usize,
190}
191
192/// This path corresponds to the 'content path' of a repository in relation
193/// to Zed's project root.
194/// In the majority of the cases, this is the folder that contains the .git folder.
195/// But if a sub-folder of a git repository is opened, this corresponds to the
196/// project root and the .git folder is located in a parent directory.
197#[derive(Clone, Debug, Ord, PartialOrd, Eq, PartialEq, Hash)]
198pub enum WorkDirectory {
199 InProject {
200 relative_path: Arc<RelPath>,
201 },
202 AboveProject {
203 absolute_path: Arc<Path>,
204 location_in_repo: Arc<Path>,
205 },
206}
207
208impl WorkDirectory {
209 fn path_key(&self) -> PathKey {
210 match self {
211 WorkDirectory::InProject { relative_path } => PathKey(relative_path.clone()),
212 WorkDirectory::AboveProject { .. } => PathKey(RelPath::empty().into()),
213 }
214 }
215
216 /// Returns true if the given path is a child of the work directory.
217 ///
218 /// Note that the path may not be a member of this repository, if there
219 /// is a repository in a directory between these two paths
220 /// external .git folder in a parent folder of the project root.
221 #[track_caller]
222 pub fn directory_contains(&self, path: &RelPath) -> bool {
223 match self {
224 WorkDirectory::InProject { relative_path } => path.starts_with(relative_path),
225 WorkDirectory::AboveProject { .. } => true,
226 }
227 }
228}
229
230impl Default for WorkDirectory {
231 fn default() -> Self {
232 Self::InProject {
233 relative_path: Arc::from(RelPath::empty()),
234 }
235 }
236}
237
238#[derive(Clone)]
239pub struct LocalSnapshot {
240 snapshot: Snapshot,
241 global_gitignore: Option<Arc<Gitignore>>,
242 /// Exclude files for all git repositories in the worktree, indexed by their absolute path.
243 /// The boolean indicates whether the gitignore needs to be updated.
244 repo_exclude_by_work_dir_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
245 /// All of the gitignore files in the worktree, indexed by their absolute path.
246 /// The boolean indicates whether the gitignore needs to be updated.
247 ignores_by_parent_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
248 /// All of the git repositories in the worktree, indexed by the project entry
249 /// id of their parent directory.
250 git_repositories: TreeMap<ProjectEntryId, LocalRepositoryEntry>,
251 /// The file handle of the worktree root
252 /// (so we can find it after it's been moved)
253 root_file_handle: Option<Arc<dyn fs::FileHandle>>,
254}
255
256struct BackgroundScannerState {
257 snapshot: LocalSnapshot,
258 scanned_dirs: HashSet<ProjectEntryId>,
259 path_prefixes_to_scan: HashSet<Arc<RelPath>>,
260 paths_to_scan: HashSet<Arc<RelPath>>,
261 /// The ids of all of the entries that were removed from the snapshot
262 /// as part of the current update. These entry ids may be re-used
263 /// if the same inode is discovered at a new path, or if the given
264 /// path is re-created after being deleted.
265 removed_entries: HashMap<u64, Entry>,
266 changed_paths: Vec<Arc<RelPath>>,
267 prev_snapshot: Snapshot,
268 scanning_enabled: bool,
269}
270
271#[derive(Clone, Debug, Eq, PartialEq)]
272struct EventRoot {
273 path: Arc<RelPath>,
274 was_rescanned: bool,
275}
276
277#[derive(Debug, Clone)]
278struct LocalRepositoryEntry {
279 work_directory_id: ProjectEntryId,
280 work_directory: WorkDirectory,
281 work_directory_abs_path: Arc<Path>,
282 git_dir_scan_id: usize,
283 /// Absolute path to the original .git entry that caused us to create this repository.
284 ///
285 /// This is normally a directory, but may be a "gitfile" that points to a directory elsewhere
286 /// (whose path we then store in `repository_dir_abs_path`).
287 dot_git_abs_path: Arc<Path>,
288 /// Absolute path to the "commondir" for this repository.
289 ///
290 /// This is always a directory. For a normal repository, this is the same as dot_git_abs_path,
291 /// but in the case of a submodule or a worktree it is the path to the "parent" .git directory
292 /// from which the submodule/worktree was derived.
293 common_dir_abs_path: Arc<Path>,
294 /// Absolute path to the directory holding the repository's state.
295 ///
296 /// For a normal repository, this is a directory and coincides with `dot_git_abs_path` and
297 /// `common_dir_abs_path`. For a submodule or worktree, this is some subdirectory of the
298 /// commondir like `/project/.git/modules/foo`.
299 repository_dir_abs_path: Arc<Path>,
300}
301
302impl sum_tree::Item for LocalRepositoryEntry {
303 type Summary = PathSummary<sum_tree::NoSummary>;
304
305 fn summary(&self, _: <Self::Summary as Summary>::Context<'_>) -> Self::Summary {
306 PathSummary {
307 max_path: self.work_directory.path_key().0,
308 item_summary: sum_tree::NoSummary,
309 }
310 }
311}
312
313impl KeyedItem for LocalRepositoryEntry {
314 type Key = PathKey;
315
316 fn key(&self) -> Self::Key {
317 self.work_directory.path_key()
318 }
319}
320
321impl Deref for LocalRepositoryEntry {
322 type Target = WorkDirectory;
323
324 fn deref(&self) -> &Self::Target {
325 &self.work_directory
326 }
327}
328
329impl Deref for LocalSnapshot {
330 type Target = Snapshot;
331
332 fn deref(&self) -> &Self::Target {
333 &self.snapshot
334 }
335}
336
337impl DerefMut for LocalSnapshot {
338 fn deref_mut(&mut self) -> &mut Self::Target {
339 &mut self.snapshot
340 }
341}
342
343enum ScanState {
344 Started,
345 Updated {
346 snapshot: LocalSnapshot,
347 changes: UpdatedEntriesSet,
348 barrier: SmallVec<[barrier::Sender; 1]>,
349 scanning: bool,
350 },
351 RootUpdated {
352 new_path: Arc<SanitizedPath>,
353 },
354}
355
356struct UpdateObservationState {
357 snapshots_tx: mpsc::UnboundedSender<(LocalSnapshot, UpdatedEntriesSet)>,
358 resume_updates: watch::Sender<()>,
359 _maintain_remote_snapshot: Task<Option<()>>,
360}
361
362#[derive(Debug, Clone)]
363pub enum Event {
364 UpdatedEntries(UpdatedEntriesSet),
365 UpdatedGitRepositories(UpdatedGitRepositoriesSet),
366 DeletedEntry(ProjectEntryId),
367}
368
369impl EventEmitter<Event> for Worktree {}
370
371impl Worktree {
372 pub async fn local(
373 path: impl Into<Arc<Path>>,
374 visible: bool,
375 fs: Arc<dyn Fs>,
376 next_entry_id: Arc<AtomicUsize>,
377 scanning_enabled: bool,
378 worktree_id: WorktreeId,
379 cx: &mut AsyncApp,
380 ) -> Result<Entity<Self>> {
381 let abs_path = path.into();
382 let metadata = fs
383 .metadata(&abs_path)
384 .await
385 .context("failed to stat worktree path")?;
386
387 let fs_case_sensitive = fs.is_case_sensitive().await;
388
389 let root_file_handle = if metadata.as_ref().is_some() {
390 fs.open_handle(&abs_path)
391 .await
392 .with_context(|| {
393 format!(
394 "failed to open local worktree root at {}",
395 abs_path.display()
396 )
397 })
398 .log_err()
399 } else {
400 None
401 };
402
403 Ok(cx.new(move |cx: &mut Context<Worktree>| {
404 let mut snapshot = LocalSnapshot {
405 ignores_by_parent_abs_path: Default::default(),
406 global_gitignore: Default::default(),
407 repo_exclude_by_work_dir_abs_path: Default::default(),
408 git_repositories: Default::default(),
409 snapshot: Snapshot::new(
410 worktree_id,
411 abs_path
412 .file_name()
413 .and_then(|f| f.to_str())
414 .map_or(RelPath::empty().into(), |f| {
415 RelPath::unix(f).unwrap().into()
416 }),
417 abs_path.clone(),
418 PathStyle::local(),
419 ),
420 root_file_handle,
421 };
422
423 let worktree_id = snapshot.id();
424 let settings_location = Some(SettingsLocation {
425 worktree_id,
426 path: RelPath::empty(),
427 });
428
429 let settings = WorktreeSettings::get(settings_location, cx).clone();
430 cx.observe_global::<SettingsStore>(move |this, cx| {
431 if let Self::Local(this) = this {
432 let settings = WorktreeSettings::get(settings_location, cx).clone();
433 if this.settings != settings {
434 this.settings = settings;
435 this.restart_background_scanners(cx);
436 }
437 }
438 })
439 .detach();
440
441 let share_private_files = false;
442 if let Some(metadata) = metadata {
443 let mut entry = Entry::new(
444 RelPath::empty().into(),
445 &metadata,
446 ProjectEntryId::new(&next_entry_id),
447 snapshot.root_char_bag,
448 None,
449 );
450 if metadata.is_dir {
451 if !scanning_enabled {
452 entry.kind = EntryKind::UnloadedDir;
453 }
454 } else {
455 if let Some(file_name) = abs_path.file_name()
456 && let Some(file_name) = file_name.to_str()
457 && let Ok(path) = RelPath::unix(file_name)
458 {
459 entry.is_private = !share_private_files && settings.is_path_private(path);
460 entry.is_hidden = settings.is_path_hidden(path);
461 }
462 }
463 cx.foreground_executor()
464 .block_on(snapshot.insert_entry(entry, fs.as_ref()));
465 }
466
467 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
468 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
469 let mut worktree = LocalWorktree {
470 share_private_files,
471 next_entry_id,
472 snapshot,
473 is_scanning: watch::channel_with(true),
474 snapshot_subscriptions: Default::default(),
475 update_observer: None,
476 scan_requests_tx,
477 path_prefixes_to_scan_tx,
478 _background_scanner_tasks: Vec::new(),
479 fs,
480 fs_case_sensitive,
481 visible,
482 settings,
483 scanning_enabled,
484 };
485 worktree.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
486 Worktree::Local(worktree)
487 }))
488 }
489
490 pub fn remote(
491 project_id: u64,
492 replica_id: ReplicaId,
493 worktree: proto::WorktreeMetadata,
494 client: AnyProtoClient,
495 path_style: PathStyle,
496 cx: &mut App,
497 ) -> Entity<Self> {
498 cx.new(|cx: &mut Context<Self>| {
499 let snapshot = Snapshot::new(
500 WorktreeId::from_proto(worktree.id),
501 RelPath::from_proto(&worktree.root_name)
502 .unwrap_or_else(|_| RelPath::empty().into()),
503 Path::new(&worktree.abs_path).into(),
504 path_style,
505 );
506
507 let background_snapshot = Arc::new(Mutex::new((
508 snapshot.clone(),
509 Vec::<proto::UpdateWorktree>::new(),
510 )));
511 let (background_updates_tx, mut background_updates_rx) =
512 mpsc::unbounded::<proto::UpdateWorktree>();
513 let (mut snapshot_updated_tx, mut snapshot_updated_rx) = watch::channel();
514
515 let worktree_id = snapshot.id();
516 let settings_location = Some(SettingsLocation {
517 worktree_id,
518 path: RelPath::empty(),
519 });
520
521 let settings = WorktreeSettings::get(settings_location, cx).clone();
522 let worktree = RemoteWorktree {
523 client,
524 project_id,
525 replica_id,
526 snapshot,
527 file_scan_inclusions: settings.parent_dir_scan_inclusions.clone(),
528 background_snapshot: background_snapshot.clone(),
529 updates_tx: Some(background_updates_tx),
530 update_observer: None,
531 snapshot_subscriptions: Default::default(),
532 visible: worktree.visible,
533 disconnected: false,
534 };
535
536 // Apply updates to a separate snapshot in a background task, then
537 // send them to a foreground task which updates the model.
538 cx.background_spawn(async move {
539 while let Some(update) = background_updates_rx.next().await {
540 {
541 let mut lock = background_snapshot.lock();
542 lock.0.apply_remote_update(
543 update.clone(),
544 &settings.parent_dir_scan_inclusions,
545 );
546 lock.1.push(update);
547 }
548 snapshot_updated_tx.send(()).await.ok();
549 }
550 })
551 .detach();
552
553 // On the foreground task, update to the latest snapshot and notify
554 // any update observer of all updates that led to that snapshot.
555 cx.spawn(async move |this, cx| {
556 while (snapshot_updated_rx.recv().await).is_some() {
557 this.update(cx, |this, cx| {
558 let mut entries_changed = false;
559 let this = this.as_remote_mut().unwrap();
560 {
561 let mut lock = this.background_snapshot.lock();
562 this.snapshot = lock.0.clone();
563 for update in lock.1.drain(..) {
564 entries_changed |= !update.updated_entries.is_empty()
565 || !update.removed_entries.is_empty();
566 if let Some(tx) = &this.update_observer {
567 tx.unbounded_send(update).ok();
568 }
569 }
570 };
571
572 if entries_changed {
573 cx.emit(Event::UpdatedEntries(Arc::default()));
574 }
575 cx.notify();
576 while let Some((scan_id, _)) = this.snapshot_subscriptions.front() {
577 if this.observed_snapshot(*scan_id) {
578 let (_, tx) = this.snapshot_subscriptions.pop_front().unwrap();
579 let _ = tx.send(());
580 } else {
581 break;
582 }
583 }
584 })?;
585 }
586 anyhow::Ok(())
587 })
588 .detach();
589
590 Worktree::Remote(worktree)
591 })
592 }
593
594 pub fn as_local(&self) -> Option<&LocalWorktree> {
595 if let Worktree::Local(worktree) = self {
596 Some(worktree)
597 } else {
598 None
599 }
600 }
601
602 pub fn as_remote(&self) -> Option<&RemoteWorktree> {
603 if let Worktree::Remote(worktree) = self {
604 Some(worktree)
605 } else {
606 None
607 }
608 }
609
610 pub fn as_local_mut(&mut self) -> Option<&mut LocalWorktree> {
611 if let Worktree::Local(worktree) = self {
612 Some(worktree)
613 } else {
614 None
615 }
616 }
617
618 pub fn as_remote_mut(&mut self) -> Option<&mut RemoteWorktree> {
619 if let Worktree::Remote(worktree) = self {
620 Some(worktree)
621 } else {
622 None
623 }
624 }
625
626 pub fn is_local(&self) -> bool {
627 matches!(self, Worktree::Local(_))
628 }
629
630 pub fn is_remote(&self) -> bool {
631 !self.is_local()
632 }
633
634 pub fn settings_location(&self, _: &Context<Self>) -> SettingsLocation<'static> {
635 SettingsLocation {
636 worktree_id: self.id(),
637 path: RelPath::empty(),
638 }
639 }
640
641 pub fn snapshot(&self) -> Snapshot {
642 match self {
643 Worktree::Local(worktree) => worktree.snapshot.snapshot.clone(),
644 Worktree::Remote(worktree) => worktree.snapshot.clone(),
645 }
646 }
647
648 pub fn scan_id(&self) -> usize {
649 match self {
650 Worktree::Local(worktree) => worktree.snapshot.scan_id,
651 Worktree::Remote(worktree) => worktree.snapshot.scan_id,
652 }
653 }
654
655 pub fn metadata_proto(&self) -> proto::WorktreeMetadata {
656 proto::WorktreeMetadata {
657 id: self.id().to_proto(),
658 root_name: self.root_name().to_proto(),
659 visible: self.is_visible(),
660 abs_path: self.abs_path().to_string_lossy().into_owned(),
661 }
662 }
663
664 pub fn completed_scan_id(&self) -> usize {
665 match self {
666 Worktree::Local(worktree) => worktree.snapshot.completed_scan_id,
667 Worktree::Remote(worktree) => worktree.snapshot.completed_scan_id,
668 }
669 }
670
671 pub fn is_visible(&self) -> bool {
672 match self {
673 Worktree::Local(worktree) => worktree.visible,
674 Worktree::Remote(worktree) => worktree.visible,
675 }
676 }
677
678 pub fn replica_id(&self) -> ReplicaId {
679 match self {
680 Worktree::Local(_) => ReplicaId::LOCAL,
681 Worktree::Remote(worktree) => worktree.replica_id,
682 }
683 }
684
685 pub fn abs_path(&self) -> Arc<Path> {
686 match self {
687 Worktree::Local(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
688 Worktree::Remote(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
689 }
690 }
691
692 pub fn root_file(&self, cx: &Context<Self>) -> Option<Arc<File>> {
693 let entry = self.root_entry()?;
694 Some(File::for_entry(entry.clone(), cx.entity()))
695 }
696
697 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
698 where
699 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
700 Fut: 'static + Send + Future<Output = bool>,
701 {
702 match self {
703 Worktree::Local(this) => this.observe_updates(project_id, cx, callback),
704 Worktree::Remote(this) => this.observe_updates(project_id, cx, callback),
705 }
706 }
707
708 pub fn stop_observing_updates(&mut self) {
709 match self {
710 Worktree::Local(this) => {
711 this.update_observer.take();
712 }
713 Worktree::Remote(this) => {
714 this.update_observer.take();
715 }
716 }
717 }
718
719 pub fn wait_for_snapshot(
720 &mut self,
721 scan_id: usize,
722 ) -> impl Future<Output = Result<()>> + use<> {
723 match self {
724 Worktree::Local(this) => this.wait_for_snapshot(scan_id).boxed(),
725 Worktree::Remote(this) => this.wait_for_snapshot(scan_id).boxed(),
726 }
727 }
728
729 #[cfg(feature = "test-support")]
730 pub fn has_update_observer(&self) -> bool {
731 match self {
732 Worktree::Local(this) => this.update_observer.is_some(),
733 Worktree::Remote(this) => this.update_observer.is_some(),
734 }
735 }
736
737 pub fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
738 match self {
739 Worktree::Local(this) => this.load_file(path, cx),
740 Worktree::Remote(_) => {
741 Task::ready(Err(anyhow!("remote worktrees can't yet load files")))
742 }
743 }
744 }
745
746 pub fn load_binary_file(
747 &self,
748 path: &RelPath,
749 cx: &Context<Worktree>,
750 ) -> Task<Result<LoadedBinaryFile>> {
751 match self {
752 Worktree::Local(this) => this.load_binary_file(path, cx),
753 Worktree::Remote(_) => {
754 Task::ready(Err(anyhow!("remote worktrees can't yet load binary files")))
755 }
756 }
757 }
758
759 pub fn write_file(
760 &self,
761 path: Arc<RelPath>,
762 text: Rope,
763 line_ending: LineEnding,
764 encoding: &'static Encoding,
765 has_bom: bool,
766 cx: &Context<Worktree>,
767 ) -> Task<Result<Arc<File>>> {
768 match self {
769 Worktree::Local(this) => {
770 this.write_file(path, text, line_ending, encoding, has_bom, cx)
771 }
772 Worktree::Remote(_) => {
773 Task::ready(Err(anyhow!("remote worktree can't yet write files")))
774 }
775 }
776 }
777
778 pub fn create_entry(
779 &mut self,
780 path: Arc<RelPath>,
781 is_directory: bool,
782 content: Option<Vec<u8>>,
783 cx: &Context<Worktree>,
784 ) -> Task<Result<CreatedEntry>> {
785 let worktree_id = self.id();
786 match self {
787 Worktree::Local(this) => this.create_entry(path, is_directory, content, cx),
788 Worktree::Remote(this) => {
789 let project_id = this.project_id;
790 let request = this.client.request(proto::CreateProjectEntry {
791 worktree_id: worktree_id.to_proto(),
792 project_id,
793 path: path.as_ref().to_proto(),
794 content,
795 is_directory,
796 });
797 cx.spawn(async move |this, cx| {
798 let response = request.await?;
799 match response.entry {
800 Some(entry) => this
801 .update(cx, |worktree, cx| {
802 worktree.as_remote_mut().unwrap().insert_entry(
803 entry,
804 response.worktree_scan_id as usize,
805 cx,
806 )
807 })?
808 .await
809 .map(CreatedEntry::Included),
810 None => {
811 let abs_path =
812 this.read_with(cx, |worktree, _| worktree.absolutize(&path))?;
813 Ok(CreatedEntry::Excluded { abs_path })
814 }
815 }
816 })
817 }
818 }
819 }
820
821 pub fn delete_entry(
822 &mut self,
823 entry_id: ProjectEntryId,
824 trash: bool,
825 cx: &mut Context<Worktree>,
826 ) -> Option<Task<Result<()>>> {
827 let task = match self {
828 Worktree::Local(this) => this.delete_entry(entry_id, trash, cx),
829 Worktree::Remote(this) => this.delete_entry(entry_id, trash, cx),
830 }?;
831
832 let entry = match &*self {
833 Worktree::Local(this) => this.entry_for_id(entry_id),
834 Worktree::Remote(this) => this.entry_for_id(entry_id),
835 }?;
836
837 let mut ids = vec![entry_id];
838 let path = &*entry.path;
839
840 self.get_children_ids_recursive(path, &mut ids);
841
842 for id in ids {
843 cx.emit(Event::DeletedEntry(id));
844 }
845 Some(task)
846 }
847
848 fn get_children_ids_recursive(&self, path: &RelPath, ids: &mut Vec<ProjectEntryId>) {
849 let children_iter = self.child_entries(path);
850 for child in children_iter {
851 ids.push(child.id);
852 self.get_children_ids_recursive(&child.path, ids);
853 }
854 }
855
856 // pub fn rename_entry(
857 // &mut self,
858 // entry_id: ProjectEntryId,
859 // new_path: Arc<RelPath>,
860 // cx: &Context<Self>,
861 // ) -> Task<Result<CreatedEntry>> {
862 // match self {
863 // Worktree::Local(this) => this.rename_entry(entry_id, new_path, cx),
864 // Worktree::Remote(this) => this.rename_entry(entry_id, new_path, cx),
865 // }
866 // }
867
868 pub fn copy_external_entries(
869 &mut self,
870 target_directory: Arc<RelPath>,
871 paths: Vec<Arc<Path>>,
872 fs: Arc<dyn Fs>,
873 cx: &Context<Worktree>,
874 ) -> Task<Result<Vec<ProjectEntryId>>> {
875 match self {
876 Worktree::Local(this) => this.copy_external_entries(target_directory, paths, cx),
877 Worktree::Remote(this) => this.copy_external_entries(target_directory, paths, fs, cx),
878 }
879 }
880
881 pub fn expand_entry(
882 &mut self,
883 entry_id: ProjectEntryId,
884 cx: &Context<Worktree>,
885 ) -> Option<Task<Result<()>>> {
886 match self {
887 Worktree::Local(this) => this.expand_entry(entry_id, cx),
888 Worktree::Remote(this) => {
889 let response = this.client.request(proto::ExpandProjectEntry {
890 project_id: this.project_id,
891 entry_id: entry_id.to_proto(),
892 });
893 Some(cx.spawn(async move |this, cx| {
894 let response = response.await?;
895 this.update(cx, |this, _| {
896 this.as_remote_mut()
897 .unwrap()
898 .wait_for_snapshot(response.worktree_scan_id as usize)
899 })?
900 .await?;
901 Ok(())
902 }))
903 }
904 }
905 }
906
907 pub fn expand_all_for_entry(
908 &mut self,
909 entry_id: ProjectEntryId,
910 cx: &Context<Worktree>,
911 ) -> Option<Task<Result<()>>> {
912 match self {
913 Worktree::Local(this) => this.expand_all_for_entry(entry_id, cx),
914 Worktree::Remote(this) => {
915 let response = this.client.request(proto::ExpandAllForProjectEntry {
916 project_id: this.project_id,
917 entry_id: entry_id.to_proto(),
918 });
919 Some(cx.spawn(async move |this, cx| {
920 let response = response.await?;
921 this.update(cx, |this, _| {
922 this.as_remote_mut()
923 .unwrap()
924 .wait_for_snapshot(response.worktree_scan_id as usize)
925 })?
926 .await?;
927 Ok(())
928 }))
929 }
930 }
931 }
932
933 pub async fn handle_create_entry(
934 this: Entity<Self>,
935 request: proto::CreateProjectEntry,
936 mut cx: AsyncApp,
937 ) -> Result<proto::ProjectEntryResponse> {
938 let (scan_id, entry) = this.update(&mut cx, |this, cx| {
939 anyhow::Ok((
940 this.scan_id(),
941 this.create_entry(
942 RelPath::from_proto(&request.path).with_context(|| {
943 format!("received invalid relative path {:?}", request.path)
944 })?,
945 request.is_directory,
946 request.content,
947 cx,
948 ),
949 ))
950 })?;
951 Ok(proto::ProjectEntryResponse {
952 entry: match &entry.await? {
953 CreatedEntry::Included(entry) => Some(entry.into()),
954 CreatedEntry::Excluded { .. } => None,
955 },
956 worktree_scan_id: scan_id as u64,
957 })
958 }
959
960 pub async fn handle_delete_entry(
961 this: Entity<Self>,
962 request: proto::DeleteProjectEntry,
963 mut cx: AsyncApp,
964 ) -> Result<proto::ProjectEntryResponse> {
965 let (scan_id, task) = this.update(&mut cx, |this, cx| {
966 (
967 this.scan_id(),
968 this.delete_entry(
969 ProjectEntryId::from_proto(request.entry_id),
970 request.use_trash,
971 cx,
972 ),
973 )
974 });
975 task.ok_or_else(|| anyhow::anyhow!("invalid entry"))?
976 .await?;
977 Ok(proto::ProjectEntryResponse {
978 entry: None,
979 worktree_scan_id: scan_id as u64,
980 })
981 }
982
983 pub async fn handle_expand_entry(
984 this: Entity<Self>,
985 request: proto::ExpandProjectEntry,
986 mut cx: AsyncApp,
987 ) -> Result<proto::ExpandProjectEntryResponse> {
988 let task = this.update(&mut cx, |this, cx| {
989 this.expand_entry(ProjectEntryId::from_proto(request.entry_id), cx)
990 });
991 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
992 .await?;
993 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
994 Ok(proto::ExpandProjectEntryResponse {
995 worktree_scan_id: scan_id as u64,
996 })
997 }
998
999 pub async fn handle_expand_all_for_entry(
1000 this: Entity<Self>,
1001 request: proto::ExpandAllForProjectEntry,
1002 mut cx: AsyncApp,
1003 ) -> Result<proto::ExpandAllForProjectEntryResponse> {
1004 let task = this.update(&mut cx, |this, cx| {
1005 this.expand_all_for_entry(ProjectEntryId::from_proto(request.entry_id), cx)
1006 });
1007 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
1008 .await?;
1009 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
1010 Ok(proto::ExpandAllForProjectEntryResponse {
1011 worktree_scan_id: scan_id as u64,
1012 })
1013 }
1014
1015 pub fn is_single_file(&self) -> bool {
1016 self.root_dir().is_none()
1017 }
1018
1019 /// For visible worktrees, returns the path with the worktree name as the first component.
1020 /// Otherwise, returns an absolute path.
1021 pub fn full_path(&self, worktree_relative_path: &RelPath) -> PathBuf {
1022 if self.is_visible() {
1023 self.root_name()
1024 .join(worktree_relative_path)
1025 .display(self.path_style)
1026 .to_string()
1027 .into()
1028 } else {
1029 let full_path = self.abs_path();
1030 let mut full_path_string = if self.is_local()
1031 && let Ok(stripped) = full_path.strip_prefix(home_dir())
1032 {
1033 self.path_style
1034 .join("~", &*stripped.to_string_lossy())
1035 .unwrap()
1036 } else {
1037 full_path.to_string_lossy().into_owned()
1038 };
1039
1040 if worktree_relative_path.components().next().is_some() {
1041 full_path_string.push_str(self.path_style.primary_separator());
1042 full_path_string.push_str(&worktree_relative_path.display(self.path_style));
1043 }
1044
1045 full_path_string.into()
1046 }
1047 }
1048}
1049
1050impl LocalWorktree {
1051 pub fn fs(&self) -> &Arc<dyn Fs> {
1052 &self.fs
1053 }
1054
1055 pub fn is_path_private(&self, path: &RelPath) -> bool {
1056 !self.share_private_files && self.settings.is_path_private(path)
1057 }
1058
1059 pub fn fs_is_case_sensitive(&self) -> bool {
1060 self.fs_case_sensitive
1061 }
1062
1063 fn restart_background_scanners(&mut self, cx: &Context<Worktree>) {
1064 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
1065 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
1066 self.scan_requests_tx = scan_requests_tx;
1067 self.path_prefixes_to_scan_tx = path_prefixes_to_scan_tx;
1068
1069 self.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
1070 let always_included_entries = mem::take(&mut self.snapshot.always_included_entries);
1071 log::debug!(
1072 "refreshing entries for the following always included paths: {:?}",
1073 always_included_entries
1074 );
1075
1076 // Cleans up old always included entries to ensure they get updated properly. Otherwise,
1077 // nested always included entries may not get updated and will result in out-of-date info.
1078 self.refresh_entries_for_paths(always_included_entries);
1079 }
1080
1081 fn start_background_scanner(
1082 &mut self,
1083 scan_requests_rx: channel::Receiver<ScanRequest>,
1084 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
1085 cx: &Context<Worktree>,
1086 ) {
1087 let snapshot = self.snapshot();
1088 let share_private_files = self.share_private_files;
1089 let next_entry_id = self.next_entry_id.clone();
1090 let fs = self.fs.clone();
1091 let scanning_enabled = self.scanning_enabled;
1092 let settings = self.settings.clone();
1093 let (scan_states_tx, mut scan_states_rx) = mpsc::unbounded();
1094 let background_scanner = cx.background_spawn({
1095 let abs_path = snapshot.abs_path.as_path().to_path_buf();
1096 let background = cx.background_executor().clone();
1097 async move {
1098 let (events, watcher) = if scanning_enabled {
1099 fs.watch(&abs_path, FS_WATCH_LATENCY).await
1100 } else {
1101 (Box::pin(stream::pending()) as _, Arc::new(NullWatcher) as _)
1102 };
1103 let fs_case_sensitive = fs.is_case_sensitive().await;
1104
1105 let mut scanner = BackgroundScanner {
1106 fs,
1107 fs_case_sensitive,
1108 status_updates_tx: scan_states_tx,
1109 executor: background,
1110 scan_requests_rx,
1111 path_prefixes_to_scan_rx,
1112 next_entry_id,
1113 state: async_lock::Mutex::new(BackgroundScannerState {
1114 prev_snapshot: snapshot.snapshot.clone(),
1115 snapshot,
1116 scanned_dirs: Default::default(),
1117 scanning_enabled,
1118 path_prefixes_to_scan: Default::default(),
1119 paths_to_scan: Default::default(),
1120 removed_entries: Default::default(),
1121 changed_paths: Default::default(),
1122 }),
1123 phase: BackgroundScannerPhase::InitialScan,
1124 share_private_files,
1125 settings,
1126 watcher,
1127 };
1128
1129 scanner
1130 .run(Box::pin(events.map(|events| events.into_iter().collect())))
1131 .await;
1132 }
1133 });
1134 let scan_state_updater = cx.spawn(async move |this, cx| {
1135 while let Some((state, this)) = scan_states_rx.next().await.zip(this.upgrade()) {
1136 this.update(cx, |this, cx| {
1137 let this = this.as_local_mut().unwrap();
1138 match state {
1139 ScanState::Started => {
1140 *this.is_scanning.0.borrow_mut() = true;
1141 }
1142 ScanState::Updated {
1143 snapshot,
1144 changes,
1145 barrier,
1146 scanning,
1147 } => {
1148 *this.is_scanning.0.borrow_mut() = scanning;
1149 this.set_snapshot(snapshot, changes, cx);
1150 drop(barrier);
1151 }
1152 ScanState::RootUpdated { new_path } => {
1153 this.update_abs_path_and_refresh(new_path, cx);
1154 }
1155 }
1156 });
1157 }
1158 });
1159 self._background_scanner_tasks = vec![background_scanner, scan_state_updater];
1160 *self.is_scanning.0.borrow_mut() = true;
1161 }
1162
1163 fn set_snapshot(
1164 &mut self,
1165 mut new_snapshot: LocalSnapshot,
1166 entry_changes: UpdatedEntriesSet,
1167 cx: &mut Context<Worktree>,
1168 ) {
1169 let repo_changes = self.changed_repos(&self.snapshot, &mut new_snapshot);
1170 self.snapshot = new_snapshot;
1171
1172 if let Some(share) = self.update_observer.as_mut() {
1173 share
1174 .snapshots_tx
1175 .unbounded_send((self.snapshot.clone(), entry_changes.clone()))
1176 .ok();
1177 }
1178
1179 if !entry_changes.is_empty() {
1180 cx.emit(Event::UpdatedEntries(entry_changes));
1181 }
1182 if !repo_changes.is_empty() {
1183 cx.emit(Event::UpdatedGitRepositories(repo_changes));
1184 }
1185
1186 while let Some((scan_id, _)) = self.snapshot_subscriptions.front() {
1187 if self.snapshot.completed_scan_id >= *scan_id {
1188 let (_, tx) = self.snapshot_subscriptions.pop_front().unwrap();
1189 tx.send(()).ok();
1190 } else {
1191 break;
1192 }
1193 }
1194 }
1195
1196 fn changed_repos(
1197 &self,
1198 old_snapshot: &LocalSnapshot,
1199 new_snapshot: &mut LocalSnapshot,
1200 ) -> UpdatedGitRepositoriesSet {
1201 let mut changes = Vec::new();
1202 let mut old_repos = old_snapshot.git_repositories.iter().peekable();
1203 let new_repos = new_snapshot.git_repositories.clone();
1204 let mut new_repos = new_repos.iter().peekable();
1205
1206 loop {
1207 match (new_repos.peek().map(clone), old_repos.peek().map(clone)) {
1208 (Some((new_entry_id, new_repo)), Some((old_entry_id, old_repo))) => {
1209 match Ord::cmp(&new_entry_id, &old_entry_id) {
1210 Ordering::Less => {
1211 changes.push(UpdatedGitRepository {
1212 work_directory_id: new_entry_id,
1213 old_work_directory_abs_path: None,
1214 new_work_directory_abs_path: Some(
1215 new_repo.work_directory_abs_path.clone(),
1216 ),
1217 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1218 repository_dir_abs_path: Some(
1219 new_repo.repository_dir_abs_path.clone(),
1220 ),
1221 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1222 });
1223 new_repos.next();
1224 }
1225 Ordering::Equal => {
1226 if new_repo.git_dir_scan_id != old_repo.git_dir_scan_id
1227 || new_repo.work_directory_abs_path
1228 != old_repo.work_directory_abs_path
1229 {
1230 changes.push(UpdatedGitRepository {
1231 work_directory_id: new_entry_id,
1232 old_work_directory_abs_path: Some(
1233 old_repo.work_directory_abs_path.clone(),
1234 ),
1235 new_work_directory_abs_path: Some(
1236 new_repo.work_directory_abs_path.clone(),
1237 ),
1238 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1239 repository_dir_abs_path: Some(
1240 new_repo.repository_dir_abs_path.clone(),
1241 ),
1242 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1243 });
1244 }
1245 new_repos.next();
1246 old_repos.next();
1247 }
1248 Ordering::Greater => {
1249 changes.push(UpdatedGitRepository {
1250 work_directory_id: old_entry_id,
1251 old_work_directory_abs_path: Some(
1252 old_repo.work_directory_abs_path.clone(),
1253 ),
1254 new_work_directory_abs_path: None,
1255 dot_git_abs_path: None,
1256 repository_dir_abs_path: None,
1257 common_dir_abs_path: None,
1258 });
1259 old_repos.next();
1260 }
1261 }
1262 }
1263 (Some((entry_id, repo)), None) => {
1264 changes.push(UpdatedGitRepository {
1265 work_directory_id: entry_id,
1266 old_work_directory_abs_path: None,
1267 new_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1268 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1269 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1270 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1271 });
1272 new_repos.next();
1273 }
1274 (None, Some((entry_id, repo))) => {
1275 changes.push(UpdatedGitRepository {
1276 work_directory_id: entry_id,
1277 old_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1278 new_work_directory_abs_path: None,
1279 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1280 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1281 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1282 });
1283 old_repos.next();
1284 }
1285 (None, None) => break,
1286 }
1287 }
1288
1289 fn clone<T: Clone, U: Clone>(value: &(&T, &U)) -> (T, U) {
1290 (value.0.clone(), value.1.clone())
1291 }
1292
1293 changes.into()
1294 }
1295
1296 pub fn scan_complete(&self) -> impl Future<Output = ()> + use<> {
1297 let mut is_scanning_rx = self.is_scanning.1.clone();
1298 async move {
1299 let mut is_scanning = *is_scanning_rx.borrow();
1300 while is_scanning {
1301 if let Some(value) = is_scanning_rx.recv().await {
1302 is_scanning = value;
1303 } else {
1304 break;
1305 }
1306 }
1307 }
1308 }
1309
1310 pub fn wait_for_snapshot(
1311 &mut self,
1312 scan_id: usize,
1313 ) -> impl Future<Output = Result<()>> + use<> {
1314 let (tx, rx) = oneshot::channel();
1315 if self.snapshot.completed_scan_id >= scan_id {
1316 tx.send(()).ok();
1317 } else {
1318 match self
1319 .snapshot_subscriptions
1320 .binary_search_by_key(&scan_id, |probe| probe.0)
1321 {
1322 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
1323 }
1324 }
1325
1326 async move {
1327 rx.await?;
1328 Ok(())
1329 }
1330 }
1331
1332 pub fn snapshot(&self) -> LocalSnapshot {
1333 self.snapshot.clone()
1334 }
1335
1336 pub fn settings(&self) -> WorktreeSettings {
1337 self.settings.clone()
1338 }
1339
1340 fn load_binary_file(
1341 &self,
1342 path: &RelPath,
1343 cx: &Context<Worktree>,
1344 ) -> Task<Result<LoadedBinaryFile>> {
1345 let path = Arc::from(path);
1346 let abs_path = self.absolutize(&path);
1347 let fs = self.fs.clone();
1348 let entry = self.refresh_entry(path.clone(), None, cx);
1349 let is_private = self.is_path_private(&path);
1350
1351 let worktree = cx.weak_entity();
1352 cx.background_spawn(async move {
1353 let content = fs.load_bytes(&abs_path).await?;
1354
1355 let worktree = worktree.upgrade().context("worktree was dropped")?;
1356 let file = match entry.await? {
1357 Some(entry) => File::for_entry(entry, worktree),
1358 None => {
1359 let metadata = fs
1360 .metadata(&abs_path)
1361 .await
1362 .with_context(|| {
1363 format!("Loading metadata for excluded file {abs_path:?}")
1364 })?
1365 .with_context(|| {
1366 format!("Excluded file {abs_path:?} got removed during loading")
1367 })?;
1368 Arc::new(File {
1369 entry_id: None,
1370 worktree,
1371 path,
1372 disk_state: DiskState::Present {
1373 mtime: metadata.mtime,
1374 size: metadata.len,
1375 },
1376 is_local: true,
1377 is_private,
1378 })
1379 }
1380 };
1381
1382 Ok(LoadedBinaryFile { file, content })
1383 })
1384 }
1385
1386 #[ztracing::instrument(skip_all)]
1387 fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
1388 let path = Arc::from(path);
1389 let abs_path = self.absolutize(&path);
1390 let fs = self.fs.clone();
1391 let entry = self.refresh_entry(path.clone(), None, cx);
1392 let is_private = self.is_path_private(path.as_ref());
1393
1394 let this = cx.weak_entity();
1395 cx.background_spawn(async move {
1396 // WARN: Temporary workaround for #27283.
1397 // We are not efficient with our memory usage per file, and use in excess of 64GB for a 10GB file
1398 // Therefore, as a temporary workaround to prevent system freezes, we just bail before opening a file
1399 // if it is too large
1400 // 5GB seems to be more reasonable, peaking at ~16GB, while 6GB jumps up to >24GB which seems like a
1401 // reasonable limit
1402 {
1403 const FILE_SIZE_MAX: u64 = 6 * 1024 * 1024 * 1024; // 6GB
1404 if let Ok(Some(metadata)) = fs.metadata(&abs_path).await
1405 && metadata.len >= FILE_SIZE_MAX
1406 {
1407 anyhow::bail!("File is too large to load");
1408 }
1409 }
1410 let (text, encoding, has_bom) = decode_file_text(fs.as_ref(), &abs_path).await?;
1411
1412 let worktree = this.upgrade().context("worktree was dropped")?;
1413 let file = match entry.await? {
1414 Some(entry) => File::for_entry(entry, worktree),
1415 None => {
1416 let metadata = fs
1417 .metadata(&abs_path)
1418 .await
1419 .with_context(|| {
1420 format!("Loading metadata for excluded file {abs_path:?}")
1421 })?
1422 .with_context(|| {
1423 format!("Excluded file {abs_path:?} got removed during loading")
1424 })?;
1425 Arc::new(File {
1426 entry_id: None,
1427 worktree,
1428 path,
1429 disk_state: DiskState::Present {
1430 mtime: metadata.mtime,
1431 size: metadata.len,
1432 },
1433 is_local: true,
1434 is_private,
1435 })
1436 }
1437 };
1438
1439 Ok(LoadedFile {
1440 file,
1441 text,
1442 encoding,
1443 has_bom,
1444 })
1445 })
1446 }
1447
1448 /// Find the lowest path in the worktree's datastructures that is an ancestor
1449 fn lowest_ancestor(&self, path: &RelPath) -> Arc<RelPath> {
1450 let mut lowest_ancestor = None;
1451 for path in path.ancestors() {
1452 if self.entry_for_path(path).is_some() {
1453 lowest_ancestor = Some(path.into());
1454 break;
1455 }
1456 }
1457
1458 lowest_ancestor.unwrap_or_else(|| RelPath::empty().into())
1459 }
1460
1461 pub fn create_entry(
1462 &self,
1463 path: Arc<RelPath>,
1464 is_dir: bool,
1465 content: Option<Vec<u8>>,
1466 cx: &Context<Worktree>,
1467 ) -> Task<Result<CreatedEntry>> {
1468 let abs_path = self.absolutize(&path);
1469 let path_excluded = self.settings.is_path_excluded(&path);
1470 let fs = self.fs.clone();
1471 let task_abs_path = abs_path.clone();
1472 let write = cx.background_spawn(async move {
1473 if is_dir {
1474 fs.create_dir(&task_abs_path)
1475 .await
1476 .with_context(|| format!("creating directory {task_abs_path:?}"))
1477 } else {
1478 fs.write(&task_abs_path, content.as_deref().unwrap_or(&[]))
1479 .await
1480 .with_context(|| format!("creating file {task_abs_path:?}"))
1481 }
1482 });
1483
1484 let lowest_ancestor = self.lowest_ancestor(&path);
1485 cx.spawn(async move |this, cx| {
1486 write.await?;
1487 if path_excluded {
1488 return Ok(CreatedEntry::Excluded { abs_path });
1489 }
1490
1491 let (result, refreshes) = this.update(cx, |this, cx| {
1492 let mut refreshes = Vec::new();
1493 let refresh_paths = path.strip_prefix(&lowest_ancestor).unwrap();
1494 for refresh_path in refresh_paths.ancestors() {
1495 if refresh_path == RelPath::empty() {
1496 continue;
1497 }
1498 let refresh_full_path = lowest_ancestor.join(refresh_path);
1499
1500 refreshes.push(this.as_local_mut().unwrap().refresh_entry(
1501 refresh_full_path,
1502 None,
1503 cx,
1504 ));
1505 }
1506 (
1507 this.as_local_mut().unwrap().refresh_entry(path, None, cx),
1508 refreshes,
1509 )
1510 })?;
1511 for refresh in refreshes {
1512 refresh.await.log_err();
1513 }
1514
1515 Ok(result
1516 .await?
1517 .map(CreatedEntry::Included)
1518 .unwrap_or_else(|| CreatedEntry::Excluded { abs_path }))
1519 })
1520 }
1521
1522 pub fn write_file(
1523 &self,
1524 path: Arc<RelPath>,
1525 text: Rope,
1526 line_ending: LineEnding,
1527 encoding: &'static Encoding,
1528 has_bom: bool,
1529 cx: &Context<Worktree>,
1530 ) -> Task<Result<Arc<File>>> {
1531 let fs = self.fs.clone();
1532 let is_private = self.is_path_private(&path);
1533 let abs_path = self.absolutize(&path);
1534
1535 let write = cx.background_spawn({
1536 let fs = fs.clone();
1537 let abs_path = abs_path.clone();
1538 async move {
1539 // For UTF-8, use the optimized `fs.save` which writes Rope chunks directly to disk
1540 // without allocating a contiguous string.
1541 if encoding == encoding_rs::UTF_8 && !has_bom {
1542 return fs.save(&abs_path, &text, line_ending).await;
1543 }
1544
1545 // For legacy encodings (e.g. Shift-JIS), we fall back to converting the entire Rope
1546 // to a String/Bytes in memory before writing.
1547 //
1548 // Note: This is inefficient for very large files compared to the streaming approach above,
1549 // but supporting streaming writes for arbitrary encodings would require a significant
1550 // refactor of the `fs` crate to expose a Writer interface.
1551 let text_string = text.to_string();
1552 let normalized_text = match line_ending {
1553 LineEnding::Unix => text_string,
1554 LineEnding::Windows => text_string.replace('\n', "\r\n"),
1555 };
1556
1557 // Create the byte vector manually for UTF-16 encodings because encoding_rs encodes to UTF-8 by default (per WHATWG standards),
1558 // which is not what we want for saving files.
1559 let bytes = if encoding == encoding_rs::UTF_16BE {
1560 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1561 if has_bom {
1562 data.extend_from_slice(&[0xFE, 0xFF]); // BOM
1563 }
1564 let utf16be_bytes =
1565 normalized_text.encode_utf16().flat_map(|u| u.to_be_bytes());
1566 data.extend(utf16be_bytes);
1567 data.into()
1568 } else if encoding == encoding_rs::UTF_16LE {
1569 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1570 if has_bom {
1571 data.extend_from_slice(&[0xFF, 0xFE]); // BOM
1572 }
1573 let utf16le_bytes =
1574 normalized_text.encode_utf16().flat_map(|u| u.to_le_bytes());
1575 data.extend(utf16le_bytes);
1576 data.into()
1577 } else {
1578 // For other encodings (Shift-JIS, UTF-8 with BOM, etc.), delegate to encoding_rs.
1579 let bom_bytes = if has_bom {
1580 if encoding == encoding_rs::UTF_8 {
1581 vec![0xEF, 0xBB, 0xBF]
1582 } else {
1583 vec![]
1584 }
1585 } else {
1586 vec![]
1587 };
1588 let (cow, _, _) = encoding.encode(&normalized_text);
1589 if !bom_bytes.is_empty() {
1590 let mut bytes = bom_bytes;
1591 bytes.extend_from_slice(&cow);
1592 bytes.into()
1593 } else {
1594 cow
1595 }
1596 };
1597
1598 fs.write(&abs_path, &bytes).await
1599 }
1600 });
1601
1602 cx.spawn(async move |this, cx| {
1603 write.await?;
1604 let entry = this
1605 .update(cx, |this, cx| {
1606 this.as_local_mut()
1607 .unwrap()
1608 .refresh_entry(path.clone(), None, cx)
1609 })?
1610 .await?;
1611 let worktree = this.upgrade().context("worktree dropped")?;
1612 if let Some(entry) = entry {
1613 Ok(File::for_entry(entry, worktree))
1614 } else {
1615 let metadata = fs
1616 .metadata(&abs_path)
1617 .await
1618 .with_context(|| {
1619 format!("Fetching metadata after saving the excluded buffer {abs_path:?}")
1620 })?
1621 .with_context(|| {
1622 format!("Excluded buffer {path:?} got removed during saving")
1623 })?;
1624 Ok(Arc::new(File {
1625 worktree,
1626 path,
1627 disk_state: DiskState::Present {
1628 mtime: metadata.mtime,
1629 size: metadata.len,
1630 },
1631 entry_id: None,
1632 is_local: true,
1633 is_private,
1634 }))
1635 }
1636 })
1637 }
1638
1639 pub fn delete_entry(
1640 &self,
1641 entry_id: ProjectEntryId,
1642 trash: bool,
1643 cx: &Context<Worktree>,
1644 ) -> Option<Task<Result<()>>> {
1645 let entry = self.entry_for_id(entry_id)?.clone();
1646 let abs_path = self.absolutize(&entry.path);
1647 let fs = self.fs.clone();
1648
1649 let delete = cx.background_spawn(async move {
1650 if entry.is_file() {
1651 if trash {
1652 fs.trash_file(&abs_path, Default::default()).await?;
1653 } else {
1654 fs.remove_file(&abs_path, Default::default()).await?;
1655 }
1656 } else if trash {
1657 fs.trash_dir(
1658 &abs_path,
1659 RemoveOptions {
1660 recursive: true,
1661 ignore_if_not_exists: false,
1662 },
1663 )
1664 .await?;
1665 } else {
1666 fs.remove_dir(
1667 &abs_path,
1668 RemoveOptions {
1669 recursive: true,
1670 ignore_if_not_exists: false,
1671 },
1672 )
1673 .await?;
1674 }
1675 anyhow::Ok(entry.path)
1676 });
1677
1678 Some(cx.spawn(async move |this, cx| {
1679 let path = delete.await?;
1680 this.update(cx, |this, _| {
1681 this.as_local_mut()
1682 .unwrap()
1683 .refresh_entries_for_paths(vec![path])
1684 })?
1685 .recv()
1686 .await;
1687 Ok(())
1688 }))
1689 }
1690
1691 pub fn copy_external_entries(
1692 &self,
1693 target_directory: Arc<RelPath>,
1694 paths: Vec<Arc<Path>>,
1695 cx: &Context<Worktree>,
1696 ) -> Task<Result<Vec<ProjectEntryId>>> {
1697 let target_directory = self.absolutize(&target_directory);
1698 let worktree_path = self.abs_path().clone();
1699 let fs = self.fs.clone();
1700 let paths = paths
1701 .into_iter()
1702 .filter_map(|source| {
1703 let file_name = source.file_name()?;
1704 let mut target = target_directory.clone();
1705 target.push(file_name);
1706
1707 // Do not allow copying the same file to itself.
1708 if source.as_ref() != target.as_path() {
1709 Some((source, target))
1710 } else {
1711 None
1712 }
1713 })
1714 .collect::<Vec<_>>();
1715
1716 let paths_to_refresh = paths
1717 .iter()
1718 .filter_map(|(_, target)| {
1719 RelPath::new(
1720 target.strip_prefix(&worktree_path).ok()?,
1721 PathStyle::local(),
1722 )
1723 .ok()
1724 .map(|path| path.into_arc())
1725 })
1726 .collect::<Vec<_>>();
1727
1728 cx.spawn(async move |this, cx| {
1729 cx.background_spawn(async move {
1730 for (source, target) in paths {
1731 copy_recursive(
1732 fs.as_ref(),
1733 &source,
1734 &target,
1735 fs::CopyOptions {
1736 overwrite: true,
1737 ..Default::default()
1738 },
1739 )
1740 .await
1741 .with_context(|| {
1742 format!("Failed to copy file from {source:?} to {target:?}")
1743 })?;
1744 }
1745 anyhow::Ok(())
1746 })
1747 .await
1748 .log_err();
1749 let mut refresh = cx.read_entity(
1750 &this.upgrade().with_context(|| "Dropped worktree")?,
1751 |this, _| {
1752 anyhow::Ok::<postage::barrier::Receiver>(
1753 this.as_local()
1754 .with_context(|| "Worktree is not local")?
1755 .refresh_entries_for_paths(paths_to_refresh.clone()),
1756 )
1757 },
1758 )?;
1759
1760 cx.background_spawn(async move {
1761 refresh.next().await;
1762 anyhow::Ok(())
1763 })
1764 .await
1765 .log_err();
1766
1767 let this = this.upgrade().with_context(|| "Dropped worktree")?;
1768 Ok(cx.read_entity(&this, |this, _| {
1769 paths_to_refresh
1770 .iter()
1771 .filter_map(|path| Some(this.entry_for_path(path)?.id))
1772 .collect()
1773 }))
1774 })
1775 }
1776
1777 fn expand_entry(
1778 &self,
1779 entry_id: ProjectEntryId,
1780 cx: &Context<Worktree>,
1781 ) -> Option<Task<Result<()>>> {
1782 let path = self.entry_for_id(entry_id)?.path.clone();
1783 let mut refresh = self.refresh_entries_for_paths(vec![path]);
1784 Some(cx.background_spawn(async move {
1785 refresh.next().await;
1786 Ok(())
1787 }))
1788 }
1789
1790 fn expand_all_for_entry(
1791 &self,
1792 entry_id: ProjectEntryId,
1793 cx: &Context<Worktree>,
1794 ) -> Option<Task<Result<()>>> {
1795 let path = self.entry_for_id(entry_id).unwrap().path.clone();
1796 let mut rx = self.add_path_prefix_to_scan(path);
1797 Some(cx.background_spawn(async move {
1798 rx.next().await;
1799 Ok(())
1800 }))
1801 }
1802
1803 pub fn refresh_entries_for_paths(&self, paths: Vec<Arc<RelPath>>) -> barrier::Receiver {
1804 let (tx, rx) = barrier::channel();
1805 self.scan_requests_tx
1806 .try_send(ScanRequest {
1807 relative_paths: paths,
1808 done: smallvec![tx],
1809 })
1810 .ok();
1811 rx
1812 }
1813
1814 #[cfg(feature = "test-support")]
1815 pub fn manually_refresh_entries_for_paths(
1816 &self,
1817 paths: Vec<Arc<RelPath>>,
1818 ) -> barrier::Receiver {
1819 self.refresh_entries_for_paths(paths)
1820 }
1821
1822 pub fn add_path_prefix_to_scan(&self, path_prefix: Arc<RelPath>) -> barrier::Receiver {
1823 let (tx, rx) = barrier::channel();
1824 self.path_prefixes_to_scan_tx
1825 .try_send(PathPrefixScanRequest {
1826 path: path_prefix,
1827 done: smallvec![tx],
1828 })
1829 .ok();
1830 rx
1831 }
1832
1833 pub fn refresh_entry(
1834 &self,
1835 path: Arc<RelPath>,
1836 old_path: Option<Arc<RelPath>>,
1837 cx: &Context<Worktree>,
1838 ) -> Task<Result<Option<Entry>>> {
1839 if self.settings.is_path_excluded(&path) {
1840 return Task::ready(Ok(None));
1841 }
1842 let paths = if let Some(old_path) = old_path.as_ref() {
1843 vec![old_path.clone(), path.clone()]
1844 } else {
1845 vec![path.clone()]
1846 };
1847 let t0 = Instant::now();
1848 let mut refresh = self.refresh_entries_for_paths(paths);
1849 // todo(lw): Hot foreground spawn
1850 cx.spawn(async move |this, cx| {
1851 refresh.recv().await;
1852 log::trace!("refreshed entry {path:?} in {:?}", t0.elapsed());
1853 let new_entry = this.read_with(cx, |this, _| {
1854 this.entry_for_path(&path).cloned().with_context(|| {
1855 format!("Could not find entry in worktree for {path:?} after refresh")
1856 })
1857 })??;
1858 Ok(Some(new_entry))
1859 })
1860 }
1861
1862 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1863 where
1864 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1865 Fut: 'static + Send + Future<Output = bool>,
1866 {
1867 if let Some(observer) = self.update_observer.as_mut() {
1868 *observer.resume_updates.borrow_mut() = ();
1869 return;
1870 }
1871
1872 let (resume_updates_tx, mut resume_updates_rx) = watch::channel::<()>();
1873 let (snapshots_tx, mut snapshots_rx) =
1874 mpsc::unbounded::<(LocalSnapshot, UpdatedEntriesSet)>();
1875 snapshots_tx
1876 .unbounded_send((self.snapshot(), Arc::default()))
1877 .ok();
1878
1879 let worktree_id = self.id.to_proto();
1880 let _maintain_remote_snapshot = cx.background_spawn(async move {
1881 let mut is_first = true;
1882 while let Some((snapshot, entry_changes)) = snapshots_rx.next().await {
1883 let update = if is_first {
1884 is_first = false;
1885 snapshot.build_initial_update(project_id, worktree_id)
1886 } else {
1887 snapshot.build_update(project_id, worktree_id, entry_changes)
1888 };
1889
1890 for update in proto::split_worktree_update(update) {
1891 let _ = resume_updates_rx.try_recv();
1892 loop {
1893 let result = callback(update.clone());
1894 if result.await {
1895 break;
1896 } else {
1897 log::info!("waiting to resume updates");
1898 if resume_updates_rx.next().await.is_none() {
1899 return Some(());
1900 }
1901 }
1902 }
1903 }
1904 }
1905 Some(())
1906 });
1907
1908 self.update_observer = Some(UpdateObservationState {
1909 snapshots_tx,
1910 resume_updates: resume_updates_tx,
1911 _maintain_remote_snapshot,
1912 });
1913 }
1914
1915 pub fn share_private_files(&mut self, cx: &Context<Worktree>) {
1916 self.share_private_files = true;
1917 self.restart_background_scanners(cx);
1918 }
1919
1920 pub fn update_abs_path_and_refresh(
1921 &mut self,
1922 new_path: Arc<SanitizedPath>,
1923 cx: &Context<Worktree>,
1924 ) {
1925 self.snapshot.git_repositories = Default::default();
1926 self.snapshot.ignores_by_parent_abs_path = Default::default();
1927 let root_name = new_path
1928 .as_path()
1929 .file_name()
1930 .and_then(|f| f.to_str())
1931 .map_or(RelPath::empty().into(), |f| {
1932 RelPath::unix(f).unwrap().into()
1933 });
1934 self.snapshot.update_abs_path(new_path, root_name);
1935 self.restart_background_scanners(cx);
1936 }
1937 #[cfg(feature = "test-support")]
1938 pub fn repositories(&self) -> Vec<Arc<Path>> {
1939 self.git_repositories
1940 .values()
1941 .map(|entry| entry.work_directory_abs_path.clone())
1942 .collect::<Vec<_>>()
1943 }
1944}
1945
1946impl RemoteWorktree {
1947 pub fn project_id(&self) -> u64 {
1948 self.project_id
1949 }
1950
1951 pub fn client(&self) -> AnyProtoClient {
1952 self.client.clone()
1953 }
1954
1955 pub fn disconnected_from_host(&mut self) {
1956 self.updates_tx.take();
1957 self.snapshot_subscriptions.clear();
1958 self.disconnected = true;
1959 }
1960
1961 pub fn update_from_remote(&self, update: proto::UpdateWorktree) {
1962 if let Some(updates_tx) = &self.updates_tx {
1963 updates_tx
1964 .unbounded_send(update)
1965 .expect("consumer runs to completion");
1966 }
1967 }
1968
1969 fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1970 where
1971 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1972 Fut: 'static + Send + Future<Output = bool>,
1973 {
1974 let (tx, mut rx) = mpsc::unbounded();
1975 let initial_update = self
1976 .snapshot
1977 .build_initial_update(project_id, self.id().to_proto());
1978 self.update_observer = Some(tx);
1979 cx.spawn(async move |this, cx| {
1980 let mut update = initial_update;
1981 'outer: loop {
1982 // SSH projects use a special project ID of 0, and we need to
1983 // remap it to the correct one here.
1984 update.project_id = project_id;
1985
1986 for chunk in split_worktree_update(update) {
1987 if !callback(chunk).await {
1988 break 'outer;
1989 }
1990 }
1991
1992 if let Some(next_update) = rx.next().await {
1993 update = next_update;
1994 } else {
1995 break;
1996 }
1997 }
1998 this.update(cx, |this, _| {
1999 let this = this.as_remote_mut().unwrap();
2000 this.update_observer.take();
2001 })
2002 })
2003 .detach();
2004 }
2005
2006 fn observed_snapshot(&self, scan_id: usize) -> bool {
2007 self.completed_scan_id >= scan_id
2008 }
2009
2010 pub fn wait_for_snapshot(
2011 &mut self,
2012 scan_id: usize,
2013 ) -> impl Future<Output = Result<()>> + use<> {
2014 let (tx, rx) = oneshot::channel();
2015 if self.observed_snapshot(scan_id) {
2016 let _ = tx.send(());
2017 } else if self.disconnected {
2018 drop(tx);
2019 } else {
2020 match self
2021 .snapshot_subscriptions
2022 .binary_search_by_key(&scan_id, |probe| probe.0)
2023 {
2024 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
2025 }
2026 }
2027
2028 async move {
2029 rx.await?;
2030 Ok(())
2031 }
2032 }
2033
2034 pub fn insert_entry(
2035 &mut self,
2036 entry: proto::Entry,
2037 scan_id: usize,
2038 cx: &Context<Worktree>,
2039 ) -> Task<Result<Entry>> {
2040 let wait_for_snapshot = self.wait_for_snapshot(scan_id);
2041 cx.spawn(async move |this, cx| {
2042 wait_for_snapshot.await?;
2043 this.update(cx, |worktree, _| {
2044 let worktree = worktree.as_remote_mut().unwrap();
2045 let snapshot = &mut worktree.background_snapshot.lock().0;
2046 let entry = snapshot.insert_entry(entry, &worktree.file_scan_inclusions);
2047 worktree.snapshot = snapshot.clone();
2048 entry
2049 })?
2050 })
2051 }
2052
2053 fn delete_entry(
2054 &self,
2055 entry_id: ProjectEntryId,
2056 trash: bool,
2057 cx: &Context<Worktree>,
2058 ) -> Option<Task<Result<()>>> {
2059 let response = self.client.request(proto::DeleteProjectEntry {
2060 project_id: self.project_id,
2061 entry_id: entry_id.to_proto(),
2062 use_trash: trash,
2063 });
2064 Some(cx.spawn(async move |this, cx| {
2065 let response = response.await?;
2066 let scan_id = response.worktree_scan_id as usize;
2067
2068 this.update(cx, move |this, _| {
2069 this.as_remote_mut().unwrap().wait_for_snapshot(scan_id)
2070 })?
2071 .await?;
2072
2073 this.update(cx, |this, _| {
2074 let this = this.as_remote_mut().unwrap();
2075 let snapshot = &mut this.background_snapshot.lock().0;
2076 snapshot.delete_entry(entry_id);
2077 this.snapshot = snapshot.clone();
2078 })
2079 }))
2080 }
2081
2082 // fn rename_entry(
2083 // &self,
2084 // entry_id: ProjectEntryId,
2085 // new_path: impl Into<Arc<RelPath>>,
2086 // cx: &Context<Worktree>,
2087 // ) -> Task<Result<CreatedEntry>> {
2088 // let new_path: Arc<RelPath> = new_path.into();
2089 // let response = self.client.request(proto::RenameProjectEntry {
2090 // project_id: self.project_id,
2091 // entry_id: entry_id.to_proto(),
2092 // new_worktree_id: new_path.worktree_id,
2093 // new_path: new_path.as_ref().to_proto(),
2094 // });
2095 // cx.spawn(async move |this, cx| {
2096 // let response = response.await?;
2097 // match response.entry {
2098 // Some(entry) => this
2099 // .update(cx, |this, cx| {
2100 // this.as_remote_mut().unwrap().insert_entry(
2101 // entry,
2102 // response.worktree_scan_id as usize,
2103 // cx,
2104 // )
2105 // })?
2106 // .await
2107 // .map(CreatedEntry::Included),
2108 // None => {
2109 // let abs_path =
2110 // this.read_with(cx, |worktree, _| worktree.absolutize(&new_path))?;
2111 // Ok(CreatedEntry::Excluded { abs_path })
2112 // }
2113 // }
2114 // })
2115 // }
2116
2117 fn copy_external_entries(
2118 &self,
2119 target_directory: Arc<RelPath>,
2120 paths_to_copy: Vec<Arc<Path>>,
2121 local_fs: Arc<dyn Fs>,
2122 cx: &Context<Worktree>,
2123 ) -> Task<anyhow::Result<Vec<ProjectEntryId>>> {
2124 let client = self.client.clone();
2125 let worktree_id = self.id().to_proto();
2126 let project_id = self.project_id;
2127
2128 cx.background_spawn(async move {
2129 let mut requests = Vec::new();
2130 for root_path_to_copy in paths_to_copy {
2131 let Some(filename) = root_path_to_copy
2132 .file_name()
2133 .and_then(|name| name.to_str())
2134 .and_then(|filename| RelPath::unix(filename).ok())
2135 else {
2136 continue;
2137 };
2138 for (abs_path, is_directory) in
2139 read_dir_items(local_fs.as_ref(), &root_path_to_copy).await?
2140 {
2141 let Some(relative_path) = abs_path
2142 .strip_prefix(&root_path_to_copy)
2143 .map_err(|e| anyhow::Error::from(e))
2144 .and_then(|relative_path| RelPath::new(relative_path, PathStyle::local()))
2145 .log_err()
2146 else {
2147 continue;
2148 };
2149 let content = if is_directory {
2150 None
2151 } else {
2152 Some(local_fs.load_bytes(&abs_path).await?)
2153 };
2154
2155 let mut target_path = target_directory.join(filename);
2156 if relative_path.file_name().is_some() {
2157 target_path = target_path.join(&relative_path);
2158 }
2159
2160 requests.push(proto::CreateProjectEntry {
2161 project_id,
2162 worktree_id,
2163 path: target_path.to_proto(),
2164 is_directory,
2165 content,
2166 });
2167 }
2168 }
2169 requests.sort_unstable_by(|a, b| a.path.cmp(&b.path));
2170 requests.dedup();
2171
2172 let mut copied_entry_ids = Vec::new();
2173 for request in requests {
2174 let response = client.request(request).await?;
2175 copied_entry_ids.extend(response.entry.map(|e| ProjectEntryId::from_proto(e.id)));
2176 }
2177
2178 Ok(copied_entry_ids)
2179 })
2180 }
2181}
2182
2183impl Snapshot {
2184 pub fn new(
2185 id: WorktreeId,
2186 root_name: Arc<RelPath>,
2187 abs_path: Arc<Path>,
2188 path_style: PathStyle,
2189 ) -> Self {
2190 Snapshot {
2191 id,
2192 abs_path: SanitizedPath::from_arc(abs_path),
2193 path_style,
2194 root_char_bag: root_name
2195 .as_unix_str()
2196 .chars()
2197 .map(|c| c.to_ascii_lowercase())
2198 .collect(),
2199 root_name,
2200 always_included_entries: Default::default(),
2201 entries_by_path: Default::default(),
2202 entries_by_id: Default::default(),
2203 scan_id: 1,
2204 completed_scan_id: 0,
2205 }
2206 }
2207
2208 pub fn id(&self) -> WorktreeId {
2209 self.id
2210 }
2211
2212 // TODO:
2213 // Consider the following:
2214 //
2215 // ```rust
2216 // let abs_path: Arc<Path> = snapshot.abs_path(); // e.g. "C:\Users\user\Desktop\project"
2217 // let some_non_trimmed_path = Path::new("\\\\?\\C:\\Users\\user\\Desktop\\project\\main.rs");
2218 // // The caller perform some actions here:
2219 // some_non_trimmed_path.strip_prefix(abs_path); // This fails
2220 // some_non_trimmed_path.starts_with(abs_path); // This fails too
2221 // ```
2222 //
2223 // This is definitely a bug, but it's not clear if we should handle it here or not.
2224 pub fn abs_path(&self) -> &Arc<Path> {
2225 SanitizedPath::cast_arc_ref(&self.abs_path)
2226 }
2227
2228 fn build_initial_update(&self, project_id: u64, worktree_id: u64) -> proto::UpdateWorktree {
2229 let mut updated_entries = self
2230 .entries_by_path
2231 .iter()
2232 .map(proto::Entry::from)
2233 .collect::<Vec<_>>();
2234 updated_entries.sort_unstable_by_key(|e| e.id);
2235
2236 proto::UpdateWorktree {
2237 project_id,
2238 worktree_id,
2239 abs_path: self.abs_path().to_string_lossy().into_owned(),
2240 root_name: self.root_name().to_proto(),
2241 updated_entries,
2242 removed_entries: Vec::new(),
2243 scan_id: self.scan_id as u64,
2244 is_last_update: self.completed_scan_id == self.scan_id,
2245 // Sent in separate messages.
2246 updated_repositories: Vec::new(),
2247 removed_repositories: Vec::new(),
2248 }
2249 }
2250
2251 pub fn work_directory_abs_path(&self, work_directory: &WorkDirectory) -> PathBuf {
2252 match work_directory {
2253 WorkDirectory::InProject { relative_path } => self.absolutize(relative_path),
2254 WorkDirectory::AboveProject { absolute_path, .. } => absolute_path.as_ref().to_owned(),
2255 }
2256 }
2257
2258 pub fn absolutize(&self, path: &RelPath) -> PathBuf {
2259 if path.file_name().is_some() {
2260 let mut abs_path = self.abs_path.to_string();
2261 for component in path.components() {
2262 if !abs_path.ends_with(self.path_style.primary_separator()) {
2263 abs_path.push_str(self.path_style.primary_separator());
2264 }
2265 abs_path.push_str(component);
2266 }
2267 PathBuf::from(abs_path)
2268 } else {
2269 self.abs_path.as_path().to_path_buf()
2270 }
2271 }
2272
2273 pub fn contains_entry(&self, entry_id: ProjectEntryId) -> bool {
2274 self.entries_by_id.get(&entry_id, ()).is_some()
2275 }
2276
2277 fn insert_entry(
2278 &mut self,
2279 entry: proto::Entry,
2280 always_included_paths: &PathMatcher,
2281 ) -> Result<Entry> {
2282 let entry = Entry::try_from((&self.root_char_bag, always_included_paths, entry))?;
2283 let old_entry = self.entries_by_id.insert_or_replace(
2284 PathEntry {
2285 id: entry.id,
2286 path: entry.path.clone(),
2287 is_ignored: entry.is_ignored,
2288 scan_id: 0,
2289 },
2290 (),
2291 );
2292 if let Some(old_entry) = old_entry {
2293 self.entries_by_path.remove(&PathKey(old_entry.path), ());
2294 }
2295 self.entries_by_path.insert_or_replace(entry.clone(), ());
2296 Ok(entry)
2297 }
2298
2299 fn delete_entry(&mut self, entry_id: ProjectEntryId) -> Option<Arc<RelPath>> {
2300 let removed_entry = self.entries_by_id.remove(&entry_id, ())?;
2301 self.entries_by_path = {
2302 let mut cursor = self.entries_by_path.cursor::<TraversalProgress>(());
2303 let mut new_entries_by_path =
2304 cursor.slice(&TraversalTarget::path(&removed_entry.path), Bias::Left);
2305 while let Some(entry) = cursor.item() {
2306 if entry.path.starts_with(&removed_entry.path) {
2307 self.entries_by_id.remove(&entry.id, ());
2308 cursor.next();
2309 } else {
2310 break;
2311 }
2312 }
2313 new_entries_by_path.append(cursor.suffix(), ());
2314 new_entries_by_path
2315 };
2316
2317 Some(removed_entry.path)
2318 }
2319
2320 fn update_abs_path(&mut self, abs_path: Arc<SanitizedPath>, root_name: Arc<RelPath>) {
2321 self.abs_path = abs_path;
2322 if root_name != self.root_name {
2323 self.root_char_bag = root_name
2324 .as_unix_str()
2325 .chars()
2326 .map(|c| c.to_ascii_lowercase())
2327 .collect();
2328 self.root_name = root_name;
2329 }
2330 }
2331
2332 pub fn apply_remote_update(
2333 &mut self,
2334 update: proto::UpdateWorktree,
2335 always_included_paths: &PathMatcher,
2336 ) {
2337 log::debug!(
2338 "applying remote worktree update. {} entries updated, {} removed",
2339 update.updated_entries.len(),
2340 update.removed_entries.len()
2341 );
2342 if let Some(root_name) = RelPath::from_proto(&update.root_name).log_err() {
2343 self.update_abs_path(
2344 SanitizedPath::new_arc(&Path::new(&update.abs_path)),
2345 root_name,
2346 );
2347 }
2348
2349 let mut entries_by_path_edits = Vec::new();
2350 let mut entries_by_id_edits = Vec::new();
2351
2352 for entry_id in update.removed_entries {
2353 let entry_id = ProjectEntryId::from_proto(entry_id);
2354 entries_by_id_edits.push(Edit::Remove(entry_id));
2355 if let Some(entry) = self.entry_for_id(entry_id) {
2356 entries_by_path_edits.push(Edit::Remove(PathKey(entry.path.clone())));
2357 }
2358 }
2359
2360 for entry in update.updated_entries {
2361 let Some(entry) =
2362 Entry::try_from((&self.root_char_bag, always_included_paths, entry)).log_err()
2363 else {
2364 continue;
2365 };
2366 if let Some(PathEntry { path, .. }) = self.entries_by_id.get(&entry.id, ()) {
2367 entries_by_path_edits.push(Edit::Remove(PathKey(path.clone())));
2368 }
2369 if let Some(old_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2370 && old_entry.id != entry.id
2371 {
2372 entries_by_id_edits.push(Edit::Remove(old_entry.id));
2373 }
2374 entries_by_id_edits.push(Edit::Insert(PathEntry {
2375 id: entry.id,
2376 path: entry.path.clone(),
2377 is_ignored: entry.is_ignored,
2378 scan_id: 0,
2379 }));
2380 entries_by_path_edits.push(Edit::Insert(entry));
2381 }
2382
2383 self.entries_by_path.edit(entries_by_path_edits, ());
2384 self.entries_by_id.edit(entries_by_id_edits, ());
2385
2386 self.scan_id = update.scan_id as usize;
2387 if update.is_last_update {
2388 self.completed_scan_id = update.scan_id as usize;
2389 }
2390 }
2391
2392 pub fn entry_count(&self) -> usize {
2393 self.entries_by_path.summary().count
2394 }
2395
2396 pub fn visible_entry_count(&self) -> usize {
2397 self.entries_by_path.summary().non_ignored_count
2398 }
2399
2400 pub fn dir_count(&self) -> usize {
2401 let summary = self.entries_by_path.summary();
2402 summary.count - summary.file_count
2403 }
2404
2405 pub fn visible_dir_count(&self) -> usize {
2406 let summary = self.entries_by_path.summary();
2407 summary.non_ignored_count - summary.non_ignored_file_count
2408 }
2409
2410 pub fn file_count(&self) -> usize {
2411 self.entries_by_path.summary().file_count
2412 }
2413
2414 pub fn visible_file_count(&self) -> usize {
2415 self.entries_by_path.summary().non_ignored_file_count
2416 }
2417
2418 fn traverse_from_offset(
2419 &self,
2420 include_files: bool,
2421 include_dirs: bool,
2422 include_ignored: bool,
2423 start_offset: usize,
2424 ) -> Traversal<'_> {
2425 let mut cursor = self.entries_by_path.cursor(());
2426 cursor.seek(
2427 &TraversalTarget::Count {
2428 count: start_offset,
2429 include_files,
2430 include_dirs,
2431 include_ignored,
2432 },
2433 Bias::Right,
2434 );
2435 Traversal {
2436 snapshot: self,
2437 cursor,
2438 include_files,
2439 include_dirs,
2440 include_ignored,
2441 }
2442 }
2443
2444 pub fn traverse_from_path(
2445 &self,
2446 include_files: bool,
2447 include_dirs: bool,
2448 include_ignored: bool,
2449 path: &RelPath,
2450 ) -> Traversal<'_> {
2451 Traversal::new(self, include_files, include_dirs, include_ignored, path)
2452 }
2453
2454 pub fn files(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2455 self.traverse_from_offset(true, false, include_ignored, start)
2456 }
2457
2458 pub fn directories(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2459 self.traverse_from_offset(false, true, include_ignored, start)
2460 }
2461
2462 pub fn entries(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2463 self.traverse_from_offset(true, true, include_ignored, start)
2464 }
2465
2466 pub fn paths(&self) -> impl Iterator<Item = &RelPath> {
2467 self.entries_by_path
2468 .cursor::<()>(())
2469 .filter(move |entry| !entry.path.is_empty())
2470 .map(|entry| entry.path.as_ref())
2471 }
2472
2473 pub fn child_entries<'a>(&'a self, parent_path: &'a RelPath) -> ChildEntriesIter<'a> {
2474 let options = ChildEntriesOptions {
2475 include_files: true,
2476 include_dirs: true,
2477 include_ignored: true,
2478 };
2479 self.child_entries_with_options(parent_path, options)
2480 }
2481
2482 pub fn child_entries_with_options<'a>(
2483 &'a self,
2484 parent_path: &'a RelPath,
2485 options: ChildEntriesOptions,
2486 ) -> ChildEntriesIter<'a> {
2487 let mut cursor = self.entries_by_path.cursor(());
2488 cursor.seek(&TraversalTarget::path(parent_path), Bias::Right);
2489 let traversal = Traversal {
2490 snapshot: self,
2491 cursor,
2492 include_files: options.include_files,
2493 include_dirs: options.include_dirs,
2494 include_ignored: options.include_ignored,
2495 };
2496 ChildEntriesIter {
2497 traversal,
2498 parent_path,
2499 }
2500 }
2501
2502 pub fn root_entry(&self) -> Option<&Entry> {
2503 self.entries_by_path.first()
2504 }
2505
2506 /// Returns `None` for a single file worktree, or `Some(self.abs_path())` if
2507 /// it is a directory.
2508 pub fn root_dir(&self) -> Option<Arc<Path>> {
2509 self.root_entry()
2510 .filter(|entry| entry.is_dir())
2511 .map(|_| self.abs_path().clone())
2512 }
2513
2514 pub fn root_name(&self) -> &RelPath {
2515 &self.root_name
2516 }
2517
2518 pub fn root_name_str(&self) -> &str {
2519 self.root_name.as_unix_str()
2520 }
2521
2522 pub fn scan_id(&self) -> usize {
2523 self.scan_id
2524 }
2525
2526 pub fn entry_for_path(&self, path: &RelPath) -> Option<&Entry> {
2527 self.traverse_from_path(true, true, true, path)
2528 .entry()
2529 .and_then(|entry| {
2530 if entry.path.as_ref() == path {
2531 Some(entry)
2532 } else {
2533 None
2534 }
2535 })
2536 }
2537
2538 /// Resolves a path to an executable using the following heuristics:
2539 ///
2540 /// 1. If the path starts with `~`, it is expanded to the user's home directory.
2541 /// 2. If the path is relative and contains more than one component,
2542 /// it is joined to the worktree root path.
2543 /// 3. If the path is relative and exists in the worktree
2544 /// (even if falls under an exclusion filter),
2545 /// it is joined to the worktree root path.
2546 /// 4. Otherwise the path is returned unmodified.
2547 ///
2548 /// Relative paths that do not exist in the worktree may
2549 /// still be found using the `PATH` environment variable.
2550 pub fn resolve_relative_path(&self, path: PathBuf) -> PathBuf {
2551 if let Some(path_str) = path.to_str() {
2552 if let Some(remaining_path) = path_str.strip_prefix("~/") {
2553 return home_dir().join(remaining_path);
2554 } else if path_str == "~" {
2555 return home_dir().to_path_buf();
2556 }
2557 }
2558
2559 if let Ok(rel_path) = RelPath::new(&path, self.path_style)
2560 && (path.components().count() > 1 || self.entry_for_path(&rel_path).is_some())
2561 {
2562 self.abs_path().join(path)
2563 } else {
2564 path
2565 }
2566 }
2567
2568 pub fn entry_for_id(&self, id: ProjectEntryId) -> Option<&Entry> {
2569 let entry = self.entries_by_id.get(&id, ())?;
2570 self.entry_for_path(&entry.path)
2571 }
2572
2573 pub fn path_style(&self) -> PathStyle {
2574 self.path_style
2575 }
2576}
2577
2578impl LocalSnapshot {
2579 fn local_repo_for_work_directory_path(&self, path: &RelPath) -> Option<&LocalRepositoryEntry> {
2580 self.git_repositories
2581 .iter()
2582 .map(|(_, entry)| entry)
2583 .find(|entry| entry.work_directory.path_key() == PathKey(path.into()))
2584 }
2585
2586 fn build_update(
2587 &self,
2588 project_id: u64,
2589 worktree_id: u64,
2590 entry_changes: UpdatedEntriesSet,
2591 ) -> proto::UpdateWorktree {
2592 let mut updated_entries = Vec::new();
2593 let mut removed_entries = Vec::new();
2594
2595 for (_, entry_id, path_change) in entry_changes.iter() {
2596 if let PathChange::Removed = path_change {
2597 removed_entries.push(entry_id.0 as u64);
2598 } else if let Some(entry) = self.entry_for_id(*entry_id) {
2599 updated_entries.push(proto::Entry::from(entry));
2600 }
2601 }
2602
2603 removed_entries.sort_unstable();
2604 updated_entries.sort_unstable_by_key(|e| e.id);
2605
2606 // TODO - optimize, knowing that removed_entries are sorted.
2607 removed_entries.retain(|id| updated_entries.binary_search_by_key(id, |e| e.id).is_err());
2608
2609 proto::UpdateWorktree {
2610 project_id,
2611 worktree_id,
2612 abs_path: self.abs_path().to_string_lossy().into_owned(),
2613 root_name: self.root_name().to_proto(),
2614 updated_entries,
2615 removed_entries,
2616 scan_id: self.scan_id as u64,
2617 is_last_update: self.completed_scan_id == self.scan_id,
2618 // Sent in separate messages.
2619 updated_repositories: Vec::new(),
2620 removed_repositories: Vec::new(),
2621 }
2622 }
2623
2624 async fn insert_entry(&mut self, mut entry: Entry, fs: &dyn Fs) -> Entry {
2625 log::trace!("insert entry {:?}", entry.path);
2626 if entry.is_file() && entry.path.file_name() == Some(&GITIGNORE) {
2627 let abs_path = self.absolutize(&entry.path);
2628 match build_gitignore(&abs_path, fs).await {
2629 Ok(ignore) => {
2630 self.ignores_by_parent_abs_path
2631 .insert(abs_path.parent().unwrap().into(), (Arc::new(ignore), true));
2632 }
2633 Err(error) => {
2634 log::error!(
2635 "error loading .gitignore file {:?} - {:?}",
2636 &entry.path,
2637 error
2638 );
2639 }
2640 }
2641 }
2642
2643 if entry.kind == EntryKind::PendingDir
2644 && let Some(existing_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2645 {
2646 entry.kind = existing_entry.kind;
2647 }
2648
2649 let scan_id = self.scan_id;
2650 let removed = self.entries_by_path.insert_or_replace(entry.clone(), ());
2651 if let Some(removed) = removed
2652 && removed.id != entry.id
2653 {
2654 self.entries_by_id.remove(&removed.id, ());
2655 }
2656 self.entries_by_id.insert_or_replace(
2657 PathEntry {
2658 id: entry.id,
2659 path: entry.path.clone(),
2660 is_ignored: entry.is_ignored,
2661 scan_id,
2662 },
2663 (),
2664 );
2665
2666 entry
2667 }
2668
2669 fn ancestor_inodes_for_path(&self, path: &RelPath) -> TreeSet<u64> {
2670 let mut inodes = TreeSet::default();
2671 for ancestor in path.ancestors().skip(1) {
2672 if let Some(entry) = self.entry_for_path(ancestor) {
2673 inodes.insert(entry.inode);
2674 }
2675 }
2676 inodes
2677 }
2678
2679 async fn ignore_stack_for_abs_path(
2680 &self,
2681 abs_path: &Path,
2682 is_dir: bool,
2683 fs: &dyn Fs,
2684 ) -> IgnoreStack {
2685 let mut new_ignores = Vec::new();
2686 let mut repo_root = None;
2687 for (index, ancestor) in abs_path.ancestors().enumerate() {
2688 if index > 0 {
2689 if let Some((ignore, _)) = self.ignores_by_parent_abs_path.get(ancestor) {
2690 new_ignores.push((ancestor, Some(ignore.clone())));
2691 } else {
2692 new_ignores.push((ancestor, None));
2693 }
2694 }
2695
2696 let metadata = fs.metadata(&ancestor.join(DOT_GIT)).await.ok().flatten();
2697 if metadata.is_some() {
2698 repo_root = Some(Arc::from(ancestor));
2699 break;
2700 }
2701 }
2702
2703 let mut ignore_stack = if let Some(global_gitignore) = self.global_gitignore.clone() {
2704 IgnoreStack::global(global_gitignore)
2705 } else {
2706 IgnoreStack::none()
2707 };
2708
2709 if let Some((repo_exclude, _)) = repo_root
2710 .as_ref()
2711 .and_then(|abs_path| self.repo_exclude_by_work_dir_abs_path.get(abs_path))
2712 {
2713 ignore_stack = ignore_stack.append(IgnoreKind::RepoExclude, repo_exclude.clone());
2714 }
2715 ignore_stack.repo_root = repo_root;
2716 for (parent_abs_path, ignore) in new_ignores.into_iter().rev() {
2717 if ignore_stack.is_abs_path_ignored(parent_abs_path, true) {
2718 ignore_stack = IgnoreStack::all();
2719 break;
2720 } else if let Some(ignore) = ignore {
2721 ignore_stack =
2722 ignore_stack.append(IgnoreKind::Gitignore(parent_abs_path.into()), ignore);
2723 }
2724 }
2725
2726 if ignore_stack.is_abs_path_ignored(abs_path, is_dir) {
2727 ignore_stack = IgnoreStack::all();
2728 }
2729
2730 ignore_stack
2731 }
2732
2733 #[cfg(feature = "test-support")]
2734 pub fn expanded_entries(&self) -> impl Iterator<Item = &Entry> {
2735 self.entries_by_path
2736 .cursor::<()>(())
2737 .filter(|entry| entry.kind == EntryKind::Dir && (entry.is_external || entry.is_ignored))
2738 }
2739
2740 #[cfg(feature = "test-support")]
2741 pub fn check_invariants(&self, git_state: bool) {
2742 use pretty_assertions::assert_eq;
2743
2744 assert_eq!(
2745 self.entries_by_path
2746 .cursor::<()>(())
2747 .map(|e| (&e.path, e.id))
2748 .collect::<Vec<_>>(),
2749 self.entries_by_id
2750 .cursor::<()>(())
2751 .map(|e| (&e.path, e.id))
2752 .collect::<collections::BTreeSet<_>>()
2753 .into_iter()
2754 .collect::<Vec<_>>(),
2755 "entries_by_path and entries_by_id are inconsistent"
2756 );
2757
2758 let mut files = self.files(true, 0);
2759 let mut visible_files = self.files(false, 0);
2760 for entry in self.entries_by_path.cursor::<()>(()) {
2761 if entry.is_file() {
2762 assert_eq!(files.next().unwrap().inode, entry.inode);
2763 if (!entry.is_ignored && !entry.is_external) || entry.is_always_included {
2764 assert_eq!(visible_files.next().unwrap().inode, entry.inode);
2765 }
2766 }
2767 }
2768
2769 assert!(files.next().is_none());
2770 assert!(visible_files.next().is_none());
2771
2772 let mut bfs_paths = Vec::new();
2773 let mut stack = self
2774 .root_entry()
2775 .map(|e| e.path.as_ref())
2776 .into_iter()
2777 .collect::<Vec<_>>();
2778 while let Some(path) = stack.pop() {
2779 bfs_paths.push(path);
2780 let ix = stack.len();
2781 for child_entry in self.child_entries(path) {
2782 stack.insert(ix, &child_entry.path);
2783 }
2784 }
2785
2786 let dfs_paths_via_iter = self
2787 .entries_by_path
2788 .cursor::<()>(())
2789 .map(|e| e.path.as_ref())
2790 .collect::<Vec<_>>();
2791 assert_eq!(bfs_paths, dfs_paths_via_iter);
2792
2793 let dfs_paths_via_traversal = self
2794 .entries(true, 0)
2795 .map(|e| e.path.as_ref())
2796 .collect::<Vec<_>>();
2797
2798 assert_eq!(dfs_paths_via_traversal, dfs_paths_via_iter);
2799
2800 if git_state {
2801 for ignore_parent_abs_path in self.ignores_by_parent_abs_path.keys() {
2802 let ignore_parent_path = &RelPath::new(
2803 ignore_parent_abs_path
2804 .strip_prefix(self.abs_path.as_path())
2805 .unwrap(),
2806 PathStyle::local(),
2807 )
2808 .unwrap();
2809 assert!(self.entry_for_path(ignore_parent_path).is_some());
2810 assert!(
2811 self.entry_for_path(
2812 &ignore_parent_path.join(RelPath::unix(GITIGNORE).unwrap())
2813 )
2814 .is_some()
2815 );
2816 }
2817 }
2818 }
2819
2820 #[cfg(feature = "test-support")]
2821 pub fn entries_without_ids(&self, include_ignored: bool) -> Vec<(&RelPath, u64, bool)> {
2822 let mut paths = Vec::new();
2823 for entry in self.entries_by_path.cursor::<()>(()) {
2824 if include_ignored || !entry.is_ignored {
2825 paths.push((entry.path.as_ref(), entry.inode, entry.is_ignored));
2826 }
2827 }
2828 paths.sort_by(|a, b| a.0.cmp(b.0));
2829 paths
2830 }
2831}
2832
2833impl BackgroundScannerState {
2834 fn should_scan_directory(&self, entry: &Entry) -> bool {
2835 (self.scanning_enabled && !entry.is_external && (!entry.is_ignored || entry.is_always_included))
2836 || entry.path.file_name() == Some(DOT_GIT)
2837 || entry.path.file_name() == Some(local_settings_folder_name())
2838 || entry.path.file_name() == Some(local_vscode_folder_name())
2839 || self.scanned_dirs.contains(&entry.id) // If we've ever scanned it, keep scanning
2840 || self
2841 .paths_to_scan
2842 .iter()
2843 .any(|p| p.starts_with(&entry.path))
2844 || self
2845 .path_prefixes_to_scan
2846 .iter()
2847 .any(|p| entry.path.starts_with(p))
2848 }
2849
2850 async fn enqueue_scan_dir(
2851 &self,
2852 abs_path: Arc<Path>,
2853 entry: &Entry,
2854 scan_job_tx: &Sender<ScanJob>,
2855 fs: &dyn Fs,
2856 ) {
2857 let path = entry.path.clone();
2858 let ignore_stack = self
2859 .snapshot
2860 .ignore_stack_for_abs_path(&abs_path, true, fs)
2861 .await;
2862 let mut ancestor_inodes = self.snapshot.ancestor_inodes_for_path(&path);
2863
2864 if !ancestor_inodes.contains(&entry.inode) {
2865 ancestor_inodes.insert(entry.inode);
2866 scan_job_tx
2867 .try_send(ScanJob {
2868 abs_path,
2869 path,
2870 ignore_stack,
2871 scan_queue: scan_job_tx.clone(),
2872 ancestor_inodes,
2873 is_external: entry.is_external,
2874 })
2875 .unwrap();
2876 }
2877 }
2878
2879 fn reuse_entry_id(&mut self, entry: &mut Entry) {
2880 if let Some(mtime) = entry.mtime {
2881 // If an entry with the same inode was removed from the worktree during this scan,
2882 // then it *might* represent the same file or directory. But the OS might also have
2883 // re-used the inode for a completely different file or directory.
2884 //
2885 // Conditionally reuse the old entry's id:
2886 // * if the mtime is the same, the file was probably been renamed.
2887 // * if the path is the same, the file may just have been updated
2888 if let Some(removed_entry) = self.removed_entries.remove(&entry.inode) {
2889 if removed_entry.mtime == Some(mtime) || removed_entry.path == entry.path {
2890 entry.id = removed_entry.id;
2891 }
2892 } else if let Some(existing_entry) = self.snapshot.entry_for_path(&entry.path) {
2893 entry.id = existing_entry.id;
2894 }
2895 }
2896 }
2897
2898 fn entry_id_for(
2899 &mut self,
2900 next_entry_id: &AtomicUsize,
2901 path: &RelPath,
2902 metadata: &fs::Metadata,
2903 ) -> ProjectEntryId {
2904 // If an entry with the same inode was removed from the worktree during this scan,
2905 // then it *might* represent the same file or directory. But the OS might also have
2906 // re-used the inode for a completely different file or directory.
2907 //
2908 // Conditionally reuse the old entry's id:
2909 // * if the mtime is the same, the file was probably been renamed.
2910 // * if the path is the same, the file may just have been updated
2911 if let Some(removed_entry) = self.removed_entries.remove(&metadata.inode) {
2912 if removed_entry.mtime == Some(metadata.mtime) || *removed_entry.path == *path {
2913 return removed_entry.id;
2914 }
2915 } else if let Some(existing_entry) = self.snapshot.entry_for_path(path) {
2916 return existing_entry.id;
2917 }
2918 ProjectEntryId::new(next_entry_id)
2919 }
2920
2921 async fn insert_entry(&mut self, entry: Entry, fs: &dyn Fs, watcher: &dyn Watcher) -> Entry {
2922 let entry = self.snapshot.insert_entry(entry, fs).await;
2923 if entry.path.file_name() == Some(&DOT_GIT) {
2924 self.insert_git_repository(entry.path.clone(), fs, watcher)
2925 .await;
2926 }
2927
2928 #[cfg(feature = "test-support")]
2929 self.snapshot.check_invariants(false);
2930
2931 entry
2932 }
2933
2934 fn populate_dir(
2935 &mut self,
2936 parent_path: Arc<RelPath>,
2937 entries: impl IntoIterator<Item = Entry>,
2938 ignore: Option<Arc<Gitignore>>,
2939 ) {
2940 let mut parent_entry = if let Some(parent_entry) = self
2941 .snapshot
2942 .entries_by_path
2943 .get(&PathKey(parent_path.clone()), ())
2944 {
2945 parent_entry.clone()
2946 } else {
2947 log::warn!(
2948 "populating a directory {:?} that has been removed",
2949 parent_path
2950 );
2951 return;
2952 };
2953
2954 match parent_entry.kind {
2955 EntryKind::PendingDir | EntryKind::UnloadedDir => parent_entry.kind = EntryKind::Dir,
2956 EntryKind::Dir => {}
2957 _ => return,
2958 }
2959
2960 if let Some(ignore) = ignore {
2961 let abs_parent_path = self
2962 .snapshot
2963 .abs_path
2964 .as_path()
2965 .join(parent_path.as_std_path())
2966 .into();
2967 self.snapshot
2968 .ignores_by_parent_abs_path
2969 .insert(abs_parent_path, (ignore, false));
2970 }
2971
2972 let parent_entry_id = parent_entry.id;
2973 self.scanned_dirs.insert(parent_entry_id);
2974 let mut entries_by_path_edits = vec![Edit::Insert(parent_entry)];
2975 let mut entries_by_id_edits = Vec::new();
2976
2977 for entry in entries {
2978 entries_by_id_edits.push(Edit::Insert(PathEntry {
2979 id: entry.id,
2980 path: entry.path.clone(),
2981 is_ignored: entry.is_ignored,
2982 scan_id: self.snapshot.scan_id,
2983 }));
2984 entries_by_path_edits.push(Edit::Insert(entry));
2985 }
2986
2987 self.snapshot
2988 .entries_by_path
2989 .edit(entries_by_path_edits, ());
2990 self.snapshot.entries_by_id.edit(entries_by_id_edits, ());
2991
2992 if let Err(ix) = self.changed_paths.binary_search(&parent_path) {
2993 self.changed_paths.insert(ix, parent_path.clone());
2994 }
2995
2996 #[cfg(feature = "test-support")]
2997 self.snapshot.check_invariants(false);
2998 }
2999
3000 fn remove_path(&mut self, path: &RelPath, watcher: &dyn Watcher) {
3001 log::trace!("background scanner removing path {path:?}");
3002 let mut new_entries;
3003 let removed_entries;
3004 {
3005 let mut cursor = self
3006 .snapshot
3007 .entries_by_path
3008 .cursor::<TraversalProgress>(());
3009 new_entries = cursor.slice(&TraversalTarget::path(path), Bias::Left);
3010 removed_entries = cursor.slice(&TraversalTarget::successor(path), Bias::Left);
3011 new_entries.append(cursor.suffix(), ());
3012 }
3013 self.snapshot.entries_by_path = new_entries;
3014
3015 let mut removed_ids = Vec::with_capacity(removed_entries.summary().count);
3016 let mut removed_dir_abs_paths = Vec::new();
3017 for entry in removed_entries.cursor::<()>(()) {
3018 if entry.is_dir() {
3019 removed_dir_abs_paths.push(self.snapshot.absolutize(&entry.path));
3020 }
3021
3022 match self.removed_entries.entry(entry.inode) {
3023 hash_map::Entry::Occupied(mut e) => {
3024 let prev_removed_entry = e.get_mut();
3025 if entry.id > prev_removed_entry.id {
3026 *prev_removed_entry = entry.clone();
3027 }
3028 }
3029 hash_map::Entry::Vacant(e) => {
3030 e.insert(entry.clone());
3031 }
3032 }
3033
3034 if entry.path.file_name() == Some(GITIGNORE) {
3035 let abs_parent_path = self.snapshot.absolutize(&entry.path.parent().unwrap());
3036 if let Some((_, needs_update)) = self
3037 .snapshot
3038 .ignores_by_parent_abs_path
3039 .get_mut(abs_parent_path.as_path())
3040 {
3041 *needs_update = true;
3042 }
3043 }
3044
3045 if let Err(ix) = removed_ids.binary_search(&entry.id) {
3046 removed_ids.insert(ix, entry.id);
3047 }
3048 }
3049
3050 self.snapshot
3051 .entries_by_id
3052 .edit(removed_ids.iter().map(|&id| Edit::Remove(id)).collect(), ());
3053 self.snapshot
3054 .git_repositories
3055 .retain(|id, _| removed_ids.binary_search(id).is_err());
3056
3057 for removed_dir_abs_path in removed_dir_abs_paths {
3058 watcher.remove(&removed_dir_abs_path).log_err();
3059 }
3060
3061 #[cfg(feature = "test-support")]
3062 self.snapshot.check_invariants(false);
3063 }
3064
3065 async fn insert_git_repository(
3066 &mut self,
3067 dot_git_path: Arc<RelPath>,
3068 fs: &dyn Fs,
3069 watcher: &dyn Watcher,
3070 ) {
3071 let work_dir_path: Arc<RelPath> = match dot_git_path.parent() {
3072 Some(parent_dir) => {
3073 // Guard against repositories inside the repository metadata
3074 if parent_dir
3075 .components()
3076 .any(|component| component == DOT_GIT)
3077 {
3078 log::debug!(
3079 "not building git repository for nested `.git` directory, `.git` path in the worktree: {dot_git_path:?}"
3080 );
3081 return;
3082 };
3083
3084 parent_dir.into()
3085 }
3086 None => {
3087 // `dot_git_path.parent().is_none()` means `.git` directory is the opened worktree itself,
3088 // no files inside that directory are tracked by git, so no need to build the repo around it
3089 log::debug!(
3090 "not building git repository for the worktree itself, `.git` path in the worktree: {dot_git_path:?}"
3091 );
3092 return;
3093 }
3094 };
3095
3096 let dot_git_abs_path = Arc::from(self.snapshot.absolutize(&dot_git_path).as_ref());
3097
3098 self.insert_git_repository_for_path(
3099 WorkDirectory::InProject {
3100 relative_path: work_dir_path,
3101 },
3102 dot_git_abs_path,
3103 fs,
3104 watcher,
3105 )
3106 .await
3107 .log_err();
3108 }
3109
3110 async fn insert_git_repository_for_path(
3111 &mut self,
3112 work_directory: WorkDirectory,
3113 dot_git_abs_path: Arc<Path>,
3114 fs: &dyn Fs,
3115 watcher: &dyn Watcher,
3116 ) -> Result<LocalRepositoryEntry> {
3117 let work_dir_entry = self
3118 .snapshot
3119 .entry_for_path(&work_directory.path_key().0)
3120 .with_context(|| {
3121 format!(
3122 "working directory `{}` not indexed",
3123 work_directory
3124 .path_key()
3125 .0
3126 .display(self.snapshot.path_style)
3127 )
3128 })?;
3129 let work_directory_abs_path = self.snapshot.work_directory_abs_path(&work_directory);
3130
3131 let (repository_dir_abs_path, common_dir_abs_path) =
3132 discover_git_paths(&dot_git_abs_path, fs).await;
3133 watcher
3134 .add(&common_dir_abs_path)
3135 .context("failed to add common directory to watcher")
3136 .log_err();
3137 watcher
3138 .add(&repository_dir_abs_path)
3139 .context("failed to add repository directory to watcher")
3140 .log_err();
3141
3142 let work_directory_id = work_dir_entry.id;
3143
3144 let local_repository = LocalRepositoryEntry {
3145 work_directory_id,
3146 work_directory,
3147 work_directory_abs_path: work_directory_abs_path.as_path().into(),
3148 git_dir_scan_id: 0,
3149 dot_git_abs_path,
3150 common_dir_abs_path,
3151 repository_dir_abs_path,
3152 };
3153
3154 self.snapshot
3155 .git_repositories
3156 .insert(work_directory_id, local_repository.clone());
3157
3158 log::trace!("inserting new local git repository");
3159 Ok(local_repository)
3160 }
3161}
3162
3163async fn is_git_dir(path: &Path, fs: &dyn Fs) -> bool {
3164 if let Some(file_name) = path.file_name()
3165 && file_name == DOT_GIT
3166 {
3167 return true;
3168 }
3169
3170 // If we're in a bare repository, we are not inside a `.git` folder. In a
3171 // bare repository, the root folder contains what would normally be in the
3172 // `.git` folder.
3173 let head_metadata = fs.metadata(&path.join("HEAD")).await;
3174 if !matches!(head_metadata, Ok(Some(_))) {
3175 return false;
3176 }
3177 let config_metadata = fs.metadata(&path.join("config")).await;
3178 matches!(config_metadata, Ok(Some(_)))
3179}
3180
3181async fn build_gitignore(abs_path: &Path, fs: &dyn Fs) -> Result<Gitignore> {
3182 let contents = fs
3183 .load(abs_path)
3184 .await
3185 .with_context(|| format!("failed to load gitignore file at {}", abs_path.display()))?;
3186 let parent = abs_path.parent().unwrap_or_else(|| Path::new("/"));
3187 let mut builder = GitignoreBuilder::new(parent);
3188 for line in contents.lines() {
3189 builder.add_line(Some(abs_path.into()), line)?;
3190 }
3191 Ok(builder.build()?)
3192}
3193
3194impl Deref for Worktree {
3195 type Target = Snapshot;
3196
3197 fn deref(&self) -> &Self::Target {
3198 match self {
3199 Worktree::Local(worktree) => &worktree.snapshot,
3200 Worktree::Remote(worktree) => &worktree.snapshot,
3201 }
3202 }
3203}
3204
3205impl Deref for LocalWorktree {
3206 type Target = LocalSnapshot;
3207
3208 fn deref(&self) -> &Self::Target {
3209 &self.snapshot
3210 }
3211}
3212
3213impl Deref for RemoteWorktree {
3214 type Target = Snapshot;
3215
3216 fn deref(&self) -> &Self::Target {
3217 &self.snapshot
3218 }
3219}
3220
3221impl fmt::Debug for LocalWorktree {
3222 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3223 self.snapshot.fmt(f)
3224 }
3225}
3226
3227impl fmt::Debug for Snapshot {
3228 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3229 struct EntriesById<'a>(&'a SumTree<PathEntry>);
3230 struct EntriesByPath<'a>(&'a SumTree<Entry>);
3231
3232 impl fmt::Debug for EntriesByPath<'_> {
3233 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3234 f.debug_map()
3235 .entries(self.0.iter().map(|entry| (&entry.path, entry.id)))
3236 .finish()
3237 }
3238 }
3239
3240 impl fmt::Debug for EntriesById<'_> {
3241 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3242 f.debug_list().entries(self.0.iter()).finish()
3243 }
3244 }
3245
3246 f.debug_struct("Snapshot")
3247 .field("id", &self.id)
3248 .field("root_name", &self.root_name)
3249 .field("entries_by_path", &EntriesByPath(&self.entries_by_path))
3250 .field("entries_by_id", &EntriesById(&self.entries_by_id))
3251 .finish()
3252 }
3253}
3254
3255#[derive(Debug, Clone, PartialEq)]
3256pub struct File {
3257 pub worktree: Entity<Worktree>,
3258 pub path: Arc<RelPath>,
3259 pub disk_state: DiskState,
3260 pub entry_id: Option<ProjectEntryId>,
3261 pub is_local: bool,
3262 pub is_private: bool,
3263}
3264
3265impl language::File for File {
3266 fn as_local(&self) -> Option<&dyn language::LocalFile> {
3267 if self.is_local { Some(self) } else { None }
3268 }
3269
3270 fn disk_state(&self) -> DiskState {
3271 self.disk_state
3272 }
3273
3274 fn path(&self) -> &Arc<RelPath> {
3275 &self.path
3276 }
3277
3278 fn full_path(&self, cx: &App) -> PathBuf {
3279 self.worktree.read(cx).full_path(&self.path)
3280 }
3281
3282 /// Returns the last component of this handle's absolute path. If this handle refers to the root
3283 /// of its worktree, then this method will return the name of the worktree itself.
3284 fn file_name<'a>(&'a self, cx: &'a App) -> &'a str {
3285 self.path
3286 .file_name()
3287 .unwrap_or_else(|| self.worktree.read(cx).root_name_str())
3288 }
3289
3290 fn worktree_id(&self, cx: &App) -> WorktreeId {
3291 self.worktree.read(cx).id()
3292 }
3293
3294 fn to_proto(&self, cx: &App) -> rpc::proto::File {
3295 rpc::proto::File {
3296 worktree_id: self.worktree.read(cx).id().to_proto(),
3297 entry_id: self.entry_id.map(|id| id.to_proto()),
3298 path: self.path.as_ref().to_proto(),
3299 mtime: self.disk_state.mtime().map(|time| time.into()),
3300 is_deleted: self.disk_state.is_deleted(),
3301 is_historic: matches!(self.disk_state, DiskState::Historic { .. }),
3302 }
3303 }
3304
3305 fn is_private(&self) -> bool {
3306 self.is_private
3307 }
3308
3309 fn path_style(&self, cx: &App) -> PathStyle {
3310 self.worktree.read(cx).path_style()
3311 }
3312
3313 fn can_open(&self) -> bool {
3314 true
3315 }
3316}
3317
3318impl language::LocalFile for File {
3319 fn abs_path(&self, cx: &App) -> PathBuf {
3320 self.worktree.read(cx).absolutize(&self.path)
3321 }
3322
3323 fn load(&self, cx: &App) -> Task<Result<String>> {
3324 let worktree = self.worktree.read(cx).as_local().unwrap();
3325 let abs_path = worktree.absolutize(&self.path);
3326 let fs = worktree.fs.clone();
3327 cx.background_spawn(async move { fs.load(&abs_path).await })
3328 }
3329
3330 fn load_bytes(&self, cx: &App) -> Task<Result<Vec<u8>>> {
3331 let worktree = self.worktree.read(cx).as_local().unwrap();
3332 let abs_path = worktree.absolutize(&self.path);
3333 let fs = worktree.fs.clone();
3334 cx.background_spawn(async move { fs.load_bytes(&abs_path).await })
3335 }
3336}
3337
3338impl File {
3339 pub fn for_entry(entry: Entry, worktree: Entity<Worktree>) -> Arc<Self> {
3340 Arc::new(Self {
3341 worktree,
3342 path: entry.path.clone(),
3343 disk_state: if let Some(mtime) = entry.mtime {
3344 DiskState::Present {
3345 mtime,
3346 size: entry.size,
3347 }
3348 } else {
3349 DiskState::New
3350 },
3351 entry_id: Some(entry.id),
3352 is_local: true,
3353 is_private: entry.is_private,
3354 })
3355 }
3356
3357 pub fn from_proto(
3358 proto: rpc::proto::File,
3359 worktree: Entity<Worktree>,
3360 cx: &App,
3361 ) -> Result<Self> {
3362 let worktree_id = worktree.read(cx).as_remote().context("not remote")?.id();
3363
3364 anyhow::ensure!(
3365 worktree_id.to_proto() == proto.worktree_id,
3366 "worktree id does not match file"
3367 );
3368
3369 let disk_state = if proto.is_historic {
3370 DiskState::Historic {
3371 was_deleted: proto.is_deleted,
3372 }
3373 } else if proto.is_deleted {
3374 DiskState::Deleted
3375 } else if let Some(mtime) = proto.mtime.map(&Into::into) {
3376 DiskState::Present { mtime, size: 0 }
3377 } else {
3378 DiskState::New
3379 };
3380
3381 Ok(Self {
3382 worktree,
3383 path: RelPath::from_proto(&proto.path).context("invalid path in file protobuf")?,
3384 disk_state,
3385 entry_id: proto.entry_id.map(ProjectEntryId::from_proto),
3386 is_local: false,
3387 is_private: false,
3388 })
3389 }
3390
3391 pub fn from_dyn(file: Option<&Arc<dyn language::File>>) -> Option<&Self> {
3392 file.and_then(|f| {
3393 let f: &dyn language::File = f.borrow();
3394 let f: &dyn Any = f;
3395 f.downcast_ref()
3396 })
3397 }
3398
3399 pub fn worktree_id(&self, cx: &App) -> WorktreeId {
3400 self.worktree.read(cx).id()
3401 }
3402
3403 pub fn project_entry_id(&self) -> Option<ProjectEntryId> {
3404 match self.disk_state {
3405 DiskState::Deleted => None,
3406 _ => self.entry_id,
3407 }
3408 }
3409}
3410
3411#[derive(Clone, Debug, PartialEq, Eq)]
3412pub struct Entry {
3413 pub id: ProjectEntryId,
3414 pub kind: EntryKind,
3415 pub path: Arc<RelPath>,
3416 pub inode: u64,
3417 pub mtime: Option<MTime>,
3418
3419 pub canonical_path: Option<Arc<Path>>,
3420 /// Whether this entry is ignored by Git.
3421 ///
3422 /// We only scan ignored entries once the directory is expanded and
3423 /// exclude them from searches.
3424 pub is_ignored: bool,
3425
3426 /// Whether this entry is hidden or inside hidden directory.
3427 ///
3428 /// We only scan hidden entries once the directory is expanded.
3429 pub is_hidden: bool,
3430
3431 /// Whether this entry is always included in searches.
3432 ///
3433 /// This is used for entries that are always included in searches, even
3434 /// if they are ignored by git. Overridden by file_scan_exclusions.
3435 pub is_always_included: bool,
3436
3437 /// Whether this entry's canonical path is outside of the worktree.
3438 /// This means the entry is only accessible from the worktree root via a
3439 /// symlink.
3440 ///
3441 /// We only scan entries outside of the worktree once the symlinked
3442 /// directory is expanded. External entries are treated like gitignored
3443 /// entries in that they are not included in searches.
3444 pub is_external: bool,
3445
3446 /// Whether this entry is considered to be a `.env` file.
3447 pub is_private: bool,
3448 /// The entry's size on disk, in bytes.
3449 pub size: u64,
3450 pub char_bag: CharBag,
3451 pub is_fifo: bool,
3452}
3453
3454#[derive(Clone, Copy, Debug, PartialEq, Eq)]
3455pub enum EntryKind {
3456 UnloadedDir,
3457 PendingDir,
3458 Dir,
3459 File,
3460}
3461
3462#[derive(Clone, Copy, Debug, PartialEq)]
3463pub enum PathChange {
3464 /// A filesystem entry was was created.
3465 Added,
3466 /// A filesystem entry was removed.
3467 Removed,
3468 /// A filesystem entry was updated.
3469 Updated,
3470 /// A filesystem entry was either updated or added. We don't know
3471 /// whether or not it already existed, because the path had not
3472 /// been loaded before the event.
3473 AddedOrUpdated,
3474 /// A filesystem entry was found during the initial scan of the worktree.
3475 Loaded,
3476}
3477
3478#[derive(Clone, Debug, PartialEq, Eq)]
3479pub struct UpdatedGitRepository {
3480 /// ID of the repository's working directory.
3481 ///
3482 /// For a repo that's above the worktree root, this is the ID of the worktree root, and hence not unique.
3483 /// It's included here to aid the GitStore in detecting when a repository's working directory is renamed.
3484 pub work_directory_id: ProjectEntryId,
3485 pub old_work_directory_abs_path: Option<Arc<Path>>,
3486 pub new_work_directory_abs_path: Option<Arc<Path>>,
3487 /// For a normal git repository checkout, the absolute path to the .git directory.
3488 /// For a worktree, the absolute path to the worktree's subdirectory inside the .git directory.
3489 pub dot_git_abs_path: Option<Arc<Path>>,
3490 pub repository_dir_abs_path: Option<Arc<Path>>,
3491 pub common_dir_abs_path: Option<Arc<Path>>,
3492}
3493
3494pub type UpdatedEntriesSet = Arc<[(Arc<RelPath>, ProjectEntryId, PathChange)]>;
3495pub type UpdatedGitRepositoriesSet = Arc<[UpdatedGitRepository]>;
3496
3497#[derive(Clone, Debug)]
3498pub struct PathProgress<'a> {
3499 pub max_path: &'a RelPath,
3500}
3501
3502#[derive(Clone, Debug)]
3503pub struct PathSummary<S> {
3504 pub max_path: Arc<RelPath>,
3505 pub item_summary: S,
3506}
3507
3508impl<S: Summary> Summary for PathSummary<S> {
3509 type Context<'a> = S::Context<'a>;
3510
3511 fn zero(cx: Self::Context<'_>) -> Self {
3512 Self {
3513 max_path: RelPath::empty().into(),
3514 item_summary: S::zero(cx),
3515 }
3516 }
3517
3518 fn add_summary(&mut self, rhs: &Self, cx: Self::Context<'_>) {
3519 self.max_path = rhs.max_path.clone();
3520 self.item_summary.add_summary(&rhs.item_summary, cx);
3521 }
3522}
3523
3524impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathProgress<'a> {
3525 fn zero(_: <PathSummary<S> as Summary>::Context<'_>) -> Self {
3526 Self {
3527 max_path: RelPath::empty(),
3528 }
3529 }
3530
3531 fn add_summary(
3532 &mut self,
3533 summary: &'a PathSummary<S>,
3534 _: <PathSummary<S> as Summary>::Context<'_>,
3535 ) {
3536 self.max_path = summary.max_path.as_ref()
3537 }
3538}
3539
3540impl<'a> sum_tree::Dimension<'a, PathSummary<GitSummary>> for GitSummary {
3541 fn zero(_cx: ()) -> Self {
3542 Default::default()
3543 }
3544
3545 fn add_summary(&mut self, summary: &'a PathSummary<GitSummary>, _: ()) {
3546 *self += summary.item_summary
3547 }
3548}
3549
3550impl<'a>
3551 sum_tree::SeekTarget<'a, PathSummary<GitSummary>, Dimensions<TraversalProgress<'a>, GitSummary>>
3552 for PathTarget<'_>
3553{
3554 fn cmp(
3555 &self,
3556 cursor_location: &Dimensions<TraversalProgress<'a>, GitSummary>,
3557 _: (),
3558 ) -> Ordering {
3559 self.cmp_path(cursor_location.0.max_path)
3560 }
3561}
3562
3563impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathKey {
3564 fn zero(_: S::Context<'_>) -> Self {
3565 Default::default()
3566 }
3567
3568 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3569 self.0 = summary.max_path.clone();
3570 }
3571}
3572
3573impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for TraversalProgress<'a> {
3574 fn zero(_cx: S::Context<'_>) -> Self {
3575 Default::default()
3576 }
3577
3578 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3579 self.max_path = summary.max_path.as_ref();
3580 }
3581}
3582
3583impl Entry {
3584 fn new(
3585 path: Arc<RelPath>,
3586 metadata: &fs::Metadata,
3587 id: ProjectEntryId,
3588 root_char_bag: CharBag,
3589 canonical_path: Option<Arc<Path>>,
3590 ) -> Self {
3591 let char_bag = char_bag_for_path(root_char_bag, &path);
3592 Self {
3593 id,
3594 kind: if metadata.is_dir {
3595 EntryKind::PendingDir
3596 } else {
3597 EntryKind::File
3598 },
3599 path,
3600 inode: metadata.inode,
3601 mtime: Some(metadata.mtime),
3602 size: metadata.len,
3603 canonical_path,
3604 is_ignored: false,
3605 is_hidden: false,
3606 is_always_included: false,
3607 is_external: false,
3608 is_private: false,
3609 char_bag,
3610 is_fifo: metadata.is_fifo,
3611 }
3612 }
3613
3614 pub fn is_created(&self) -> bool {
3615 self.mtime.is_some()
3616 }
3617
3618 pub fn is_dir(&self) -> bool {
3619 self.kind.is_dir()
3620 }
3621
3622 pub fn is_file(&self) -> bool {
3623 self.kind.is_file()
3624 }
3625}
3626
3627impl EntryKind {
3628 pub fn is_dir(&self) -> bool {
3629 matches!(
3630 self,
3631 EntryKind::Dir | EntryKind::PendingDir | EntryKind::UnloadedDir
3632 )
3633 }
3634
3635 pub fn is_unloaded(&self) -> bool {
3636 matches!(self, EntryKind::UnloadedDir)
3637 }
3638
3639 pub fn is_file(&self) -> bool {
3640 matches!(self, EntryKind::File)
3641 }
3642}
3643
3644impl sum_tree::Item for Entry {
3645 type Summary = EntrySummary;
3646
3647 fn summary(&self, _cx: ()) -> Self::Summary {
3648 let non_ignored_count = if (self.is_ignored || self.is_external) && !self.is_always_included
3649 {
3650 0
3651 } else {
3652 1
3653 };
3654 let file_count;
3655 let non_ignored_file_count;
3656 if self.is_file() {
3657 file_count = 1;
3658 non_ignored_file_count = non_ignored_count;
3659 } else {
3660 file_count = 0;
3661 non_ignored_file_count = 0;
3662 }
3663
3664 EntrySummary {
3665 max_path: self.path.clone(),
3666 count: 1,
3667 non_ignored_count,
3668 file_count,
3669 non_ignored_file_count,
3670 }
3671 }
3672}
3673
3674impl sum_tree::KeyedItem for Entry {
3675 type Key = PathKey;
3676
3677 fn key(&self) -> Self::Key {
3678 PathKey(self.path.clone())
3679 }
3680}
3681
3682#[derive(Clone, Debug)]
3683pub struct EntrySummary {
3684 max_path: Arc<RelPath>,
3685 count: usize,
3686 non_ignored_count: usize,
3687 file_count: usize,
3688 non_ignored_file_count: usize,
3689}
3690
3691impl Default for EntrySummary {
3692 fn default() -> Self {
3693 Self {
3694 max_path: Arc::from(RelPath::empty()),
3695 count: 0,
3696 non_ignored_count: 0,
3697 file_count: 0,
3698 non_ignored_file_count: 0,
3699 }
3700 }
3701}
3702
3703impl sum_tree::ContextLessSummary for EntrySummary {
3704 fn zero() -> Self {
3705 Default::default()
3706 }
3707
3708 fn add_summary(&mut self, rhs: &Self) {
3709 self.max_path = rhs.max_path.clone();
3710 self.count += rhs.count;
3711 self.non_ignored_count += rhs.non_ignored_count;
3712 self.file_count += rhs.file_count;
3713 self.non_ignored_file_count += rhs.non_ignored_file_count;
3714 }
3715}
3716
3717#[derive(Clone, Debug)]
3718struct PathEntry {
3719 id: ProjectEntryId,
3720 path: Arc<RelPath>,
3721 is_ignored: bool,
3722 scan_id: usize,
3723}
3724
3725impl sum_tree::Item for PathEntry {
3726 type Summary = PathEntrySummary;
3727
3728 fn summary(&self, _cx: ()) -> Self::Summary {
3729 PathEntrySummary { max_id: self.id }
3730 }
3731}
3732
3733impl sum_tree::KeyedItem for PathEntry {
3734 type Key = ProjectEntryId;
3735
3736 fn key(&self) -> Self::Key {
3737 self.id
3738 }
3739}
3740
3741#[derive(Clone, Debug, Default)]
3742struct PathEntrySummary {
3743 max_id: ProjectEntryId,
3744}
3745
3746impl sum_tree::ContextLessSummary for PathEntrySummary {
3747 fn zero() -> Self {
3748 Default::default()
3749 }
3750
3751 fn add_summary(&mut self, summary: &Self) {
3752 self.max_id = summary.max_id;
3753 }
3754}
3755
3756impl<'a> sum_tree::Dimension<'a, PathEntrySummary> for ProjectEntryId {
3757 fn zero(_cx: ()) -> Self {
3758 Default::default()
3759 }
3760
3761 fn add_summary(&mut self, summary: &'a PathEntrySummary, _: ()) {
3762 *self = summary.max_id;
3763 }
3764}
3765
3766#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
3767pub struct PathKey(pub Arc<RelPath>);
3768
3769impl Default for PathKey {
3770 fn default() -> Self {
3771 Self(RelPath::empty().into())
3772 }
3773}
3774
3775impl<'a> sum_tree::Dimension<'a, EntrySummary> for PathKey {
3776 fn zero(_cx: ()) -> Self {
3777 Default::default()
3778 }
3779
3780 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
3781 self.0 = summary.max_path.clone();
3782 }
3783}
3784
3785struct BackgroundScanner {
3786 state: async_lock::Mutex<BackgroundScannerState>,
3787 fs: Arc<dyn Fs>,
3788 fs_case_sensitive: bool,
3789 status_updates_tx: UnboundedSender<ScanState>,
3790 executor: BackgroundExecutor,
3791 scan_requests_rx: channel::Receiver<ScanRequest>,
3792 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
3793 next_entry_id: Arc<AtomicUsize>,
3794 phase: BackgroundScannerPhase,
3795 watcher: Arc<dyn Watcher>,
3796 settings: WorktreeSettings,
3797 share_private_files: bool,
3798}
3799
3800#[derive(Copy, Clone, PartialEq)]
3801enum BackgroundScannerPhase {
3802 InitialScan,
3803 EventsReceivedDuringInitialScan,
3804 Events,
3805}
3806
3807impl BackgroundScanner {
3808 async fn run(&mut self, mut fs_events_rx: Pin<Box<dyn Send + Stream<Item = Vec<PathEvent>>>>) {
3809 let root_abs_path;
3810 let scanning_enabled;
3811 {
3812 let state = self.state.lock().await;
3813 root_abs_path = state.snapshot.abs_path.clone();
3814 scanning_enabled = state.scanning_enabled;
3815 }
3816
3817 // If the worktree root does not contain a git repository, then find
3818 // the git repository in an ancestor directory. Find any gitignore files
3819 // in ancestor directories.
3820 let repo = if scanning_enabled {
3821 let (ignores, exclude, repo) =
3822 discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await;
3823 self.state
3824 .lock()
3825 .await
3826 .snapshot
3827 .ignores_by_parent_abs_path
3828 .extend(ignores);
3829 if let Some(exclude) = exclude {
3830 self.state
3831 .lock()
3832 .await
3833 .snapshot
3834 .repo_exclude_by_work_dir_abs_path
3835 .insert(root_abs_path.as_path().into(), (exclude, false));
3836 }
3837
3838 repo
3839 } else {
3840 None
3841 };
3842
3843 let containing_git_repository = if let Some((ancestor_dot_git, work_directory)) = repo
3844 && scanning_enabled
3845 {
3846 maybe!(async {
3847 self.state
3848 .lock()
3849 .await
3850 .insert_git_repository_for_path(
3851 work_directory,
3852 ancestor_dot_git.clone().into(),
3853 self.fs.as_ref(),
3854 self.watcher.as_ref(),
3855 )
3856 .await
3857 .log_err()?;
3858 Some(ancestor_dot_git)
3859 })
3860 .await
3861 } else {
3862 None
3863 };
3864
3865 log::trace!("containing git repository: {containing_git_repository:?}");
3866
3867 let global_gitignore_file = paths::global_gitignore_path();
3868 let mut global_gitignore_events = if let Some(global_gitignore_path) =
3869 &global_gitignore_file
3870 && scanning_enabled
3871 {
3872 let is_file = self.fs.is_file(&global_gitignore_path).await;
3873 self.state.lock().await.snapshot.global_gitignore = if is_file {
3874 build_gitignore(global_gitignore_path, self.fs.as_ref())
3875 .await
3876 .ok()
3877 .map(Arc::new)
3878 } else {
3879 None
3880 };
3881 if is_file {
3882 self.fs
3883 .watch(global_gitignore_path, FS_WATCH_LATENCY)
3884 .await
3885 .0
3886 } else {
3887 Box::pin(futures::stream::pending())
3888 }
3889 } else {
3890 self.state.lock().await.snapshot.global_gitignore = None;
3891 Box::pin(futures::stream::pending())
3892 };
3893
3894 let (scan_job_tx, scan_job_rx) = channel::unbounded();
3895 {
3896 let mut state = self.state.lock().await;
3897 state.snapshot.scan_id += 1;
3898 if let Some(mut root_entry) = state.snapshot.root_entry().cloned() {
3899 let ignore_stack = state
3900 .snapshot
3901 .ignore_stack_for_abs_path(root_abs_path.as_path(), true, self.fs.as_ref())
3902 .await;
3903 if ignore_stack.is_abs_path_ignored(root_abs_path.as_path(), true) {
3904 root_entry.is_ignored = true;
3905 let mut root_entry = root_entry.clone();
3906 state.reuse_entry_id(&mut root_entry);
3907 state
3908 .insert_entry(root_entry, self.fs.as_ref(), self.watcher.as_ref())
3909 .await;
3910 }
3911 if root_entry.is_dir() && state.scanning_enabled {
3912 state
3913 .enqueue_scan_dir(
3914 root_abs_path.as_path().into(),
3915 &root_entry,
3916 &scan_job_tx,
3917 self.fs.as_ref(),
3918 )
3919 .await;
3920 }
3921 }
3922 };
3923
3924 // Perform an initial scan of the directory.
3925 drop(scan_job_tx);
3926 self.scan_dirs(true, scan_job_rx).await;
3927 {
3928 let mut state = self.state.lock().await;
3929 state.snapshot.completed_scan_id = state.snapshot.scan_id;
3930 }
3931
3932 self.send_status_update(false, SmallVec::new(), &[]).await;
3933
3934 // Process any any FS events that occurred while performing the initial scan.
3935 // For these events, update events cannot be as precise, because we didn't
3936 // have the previous state loaded yet.
3937 self.phase = BackgroundScannerPhase::EventsReceivedDuringInitialScan;
3938 if let Poll::Ready(Some(mut paths)) = futures::poll!(fs_events_rx.next()) {
3939 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
3940 paths.extend(more_paths);
3941 }
3942 self.process_events(
3943 paths
3944 .into_iter()
3945 .filter(|event| event.kind.is_some())
3946 .collect(),
3947 )
3948 .await;
3949 }
3950 if let Some(abs_path) = containing_git_repository {
3951 self.process_events(vec![PathEvent {
3952 path: abs_path,
3953 kind: Some(fs::PathEventKind::Changed),
3954 }])
3955 .await;
3956 }
3957
3958 // Continue processing events until the worktree is dropped.
3959 self.phase = BackgroundScannerPhase::Events;
3960
3961 loop {
3962 select_biased! {
3963 // Process any path refresh requests from the worktree. Prioritize
3964 // these before handling changes reported by the filesystem.
3965 request = self.next_scan_request().fuse() => {
3966 let Ok(request) = request else { break };
3967 if !self.process_scan_request(request, false).await {
3968 return;
3969 }
3970 }
3971
3972 path_prefix_request = self.path_prefixes_to_scan_rx.recv().fuse() => {
3973 let Ok(request) = path_prefix_request else { break };
3974 log::trace!("adding path prefix {:?}", request.path);
3975
3976 let did_scan = self.forcibly_load_paths(std::slice::from_ref(&request.path)).await;
3977 if did_scan {
3978 let abs_path =
3979 {
3980 let mut state = self.state.lock().await;
3981 state.path_prefixes_to_scan.insert(request.path.clone());
3982 state.snapshot.absolutize(&request.path)
3983 };
3984
3985 if let Some(abs_path) = self.fs.canonicalize(&abs_path).await.log_err() {
3986 self.process_events(vec![PathEvent {
3987 path: abs_path,
3988 kind: Some(fs::PathEventKind::Changed),
3989 }])
3990 .await;
3991 }
3992 }
3993 self.send_status_update(false, request.done, &[]).await;
3994 }
3995
3996 paths = fs_events_rx.next().fuse() => {
3997 let Some(mut paths) = paths else { break };
3998 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
3999 paths.extend(more_paths);
4000 }
4001 self.process_events(paths.into_iter().filter(|event| event.kind.is_some()).collect()).await;
4002 }
4003
4004 _ = global_gitignore_events.next().fuse() => {
4005 if let Some(path) = &global_gitignore_file {
4006 self.update_global_gitignore(&path).await;
4007 }
4008 }
4009 }
4010 }
4011 }
4012
4013 async fn process_scan_request(&self, mut request: ScanRequest, scanning: bool) -> bool {
4014 log::debug!("rescanning paths {:?}", request.relative_paths);
4015
4016 request.relative_paths.sort_unstable();
4017 self.forcibly_load_paths(&request.relative_paths).await;
4018
4019 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4020 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4021 let root_canonical_path = match &root_canonical_path {
4022 Ok(path) => SanitizedPath::new(path),
4023 Err(err) => {
4024 log::error!("failed to canonicalize root path {root_path:?}: {err:#}");
4025 return true;
4026 }
4027 };
4028 let abs_paths = request
4029 .relative_paths
4030 .iter()
4031 .map(|path| {
4032 if path.file_name().is_some() {
4033 root_canonical_path.as_path().join(path.as_std_path())
4034 } else {
4035 root_canonical_path.as_path().to_path_buf()
4036 }
4037 })
4038 .collect::<Vec<_>>();
4039
4040 {
4041 let mut state = self.state.lock().await;
4042 let is_idle = state.snapshot.completed_scan_id == state.snapshot.scan_id;
4043 state.snapshot.scan_id += 1;
4044 if is_idle {
4045 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4046 }
4047 }
4048
4049 self.reload_entries_for_paths(
4050 &root_path,
4051 &root_canonical_path,
4052 &request.relative_paths,
4053 abs_paths,
4054 None,
4055 )
4056 .await;
4057
4058 self.send_status_update(scanning, request.done, &[]).await
4059 }
4060
4061 async fn process_events(&self, mut events: Vec<PathEvent>) {
4062 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4063 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4064 let root_canonical_path = match &root_canonical_path {
4065 Ok(path) => SanitizedPath::new(path),
4066 Err(err) => {
4067 let new_path = self
4068 .state
4069 .lock()
4070 .await
4071 .snapshot
4072 .root_file_handle
4073 .clone()
4074 .and_then(|handle| match handle.current_path(&self.fs) {
4075 Ok(new_path) => Some(new_path),
4076 Err(e) => {
4077 log::error!("Failed to refresh worktree root path: {e:#}");
4078 None
4079 }
4080 })
4081 .map(|path| SanitizedPath::new_arc(&path))
4082 .filter(|new_path| *new_path != root_path);
4083
4084 if let Some(new_path) = new_path {
4085 log::info!(
4086 "root renamed from {:?} to {:?}",
4087 root_path.as_path(),
4088 new_path.as_path(),
4089 );
4090 self.status_updates_tx
4091 .unbounded_send(ScanState::RootUpdated { new_path })
4092 .ok();
4093 } else {
4094 log::error!("root path could not be canonicalized: {err:#}");
4095 }
4096 return;
4097 }
4098 };
4099
4100 // Certain directories may have FS changes, but do not lead to git data changes that Zed cares about.
4101 // Ignore these, to avoid Zed unnecessarily rescanning git metadata.
4102 let skipped_files_in_dot_git = [COMMIT_MESSAGE, INDEX_LOCK];
4103 let skipped_dirs_in_dot_git = [FSMONITOR_DAEMON, LFS_DIR];
4104
4105 let mut relative_paths = Vec::with_capacity(events.len());
4106 let mut dot_git_abs_paths = Vec::new();
4107 let mut work_dirs_needing_exclude_update = Vec::new();
4108 events.sort_unstable_by(|left, right| left.path.cmp(&right.path));
4109 events.dedup_by(|left, right| {
4110 if left.path == right.path {
4111 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4112 right.kind = left.kind;
4113 }
4114 true
4115 } else if left.path.starts_with(&right.path) {
4116 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4117 right.kind = left.kind;
4118 }
4119 true
4120 } else {
4121 false
4122 }
4123 });
4124 {
4125 let snapshot = &self.state.lock().await.snapshot;
4126
4127 let mut ranges_to_drop = SmallVec::<[Range<usize>; 4]>::new();
4128
4129 fn skip_ix(ranges: &mut SmallVec<[Range<usize>; 4]>, ix: usize) {
4130 if let Some(last_range) = ranges.last_mut()
4131 && last_range.end == ix
4132 {
4133 last_range.end += 1;
4134 } else {
4135 ranges.push(ix..ix + 1);
4136 }
4137 }
4138
4139 for (ix, event) in events.iter().enumerate() {
4140 let abs_path = &SanitizedPath::new(&event.path);
4141
4142 let mut is_git_related = false;
4143 let mut dot_git_paths = None;
4144
4145 for ancestor in abs_path.as_path().ancestors() {
4146 if is_git_dir(ancestor, self.fs.as_ref()).await {
4147 let path_in_git_dir = abs_path
4148 .as_path()
4149 .strip_prefix(ancestor)
4150 .expect("stripping off the ancestor");
4151 dot_git_paths = Some((ancestor.to_owned(), path_in_git_dir.to_owned()));
4152 break;
4153 }
4154 }
4155
4156 if let Some((dot_git_abs_path, path_in_git_dir)) = dot_git_paths {
4157 if skipped_files_in_dot_git
4158 .iter()
4159 .any(|skipped| OsStr::new(skipped) == path_in_git_dir.as_path().as_os_str())
4160 || skipped_dirs_in_dot_git.iter().any(|skipped_git_subdir| {
4161 path_in_git_dir.starts_with(skipped_git_subdir)
4162 })
4163 {
4164 log::debug!(
4165 "ignoring event {abs_path:?} as it's in the .git directory among skipped files or directories"
4166 );
4167 skip_ix(&mut ranges_to_drop, ix);
4168 continue;
4169 }
4170
4171 is_git_related = true;
4172 if !dot_git_abs_paths.contains(&dot_git_abs_path) {
4173 dot_git_abs_paths.push(dot_git_abs_path);
4174 }
4175 }
4176
4177 let relative_path = if let Ok(path) = abs_path.strip_prefix(&root_canonical_path)
4178 && let Ok(path) = RelPath::new(path, PathStyle::local())
4179 {
4180 path
4181 } else {
4182 if is_git_related {
4183 log::debug!(
4184 "ignoring event {abs_path:?}, since it's in git dir outside of root path {root_canonical_path:?}",
4185 );
4186 } else {
4187 log::error!(
4188 "ignoring event {abs_path:?} outside of root path {root_canonical_path:?}",
4189 );
4190 }
4191 skip_ix(&mut ranges_to_drop, ix);
4192 continue;
4193 };
4194
4195 let absolute_path = abs_path.to_path_buf();
4196 if absolute_path.ends_with(Path::new(DOT_GIT).join(REPO_EXCLUDE)) {
4197 if let Some(repository) = snapshot
4198 .git_repositories
4199 .values()
4200 .find(|repo| repo.common_dir_abs_path.join(REPO_EXCLUDE) == absolute_path)
4201 {
4202 work_dirs_needing_exclude_update
4203 .push(repository.work_directory_abs_path.clone());
4204 }
4205 }
4206
4207 if abs_path.file_name() == Some(OsStr::new(GITIGNORE)) {
4208 for (_, repo) in snapshot
4209 .git_repositories
4210 .iter()
4211 .filter(|(_, repo)| repo.directory_contains(&relative_path))
4212 {
4213 if !dot_git_abs_paths.iter().any(|dot_git_abs_path| {
4214 dot_git_abs_path == repo.common_dir_abs_path.as_ref()
4215 }) {
4216 dot_git_abs_paths.push(repo.common_dir_abs_path.to_path_buf());
4217 }
4218 }
4219 }
4220
4221 let parent_dir_is_loaded = relative_path.parent().is_none_or(|parent| {
4222 snapshot
4223 .entry_for_path(parent)
4224 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4225 });
4226 if !parent_dir_is_loaded {
4227 log::debug!("ignoring event {relative_path:?} within unloaded directory");
4228 skip_ix(&mut ranges_to_drop, ix);
4229 continue;
4230 }
4231
4232 if self.settings.is_path_excluded(&relative_path) {
4233 if !is_git_related {
4234 log::debug!("ignoring FS event for excluded path {relative_path:?}");
4235 }
4236 skip_ix(&mut ranges_to_drop, ix);
4237 continue;
4238 }
4239
4240 relative_paths.push(EventRoot {
4241 path: relative_path.into_arc(),
4242 was_rescanned: matches!(event.kind, Some(fs::PathEventKind::Rescan)),
4243 });
4244 }
4245
4246 for range_to_drop in ranges_to_drop.into_iter().rev() {
4247 events.drain(range_to_drop);
4248 }
4249 }
4250
4251 if relative_paths.is_empty() && dot_git_abs_paths.is_empty() {
4252 return;
4253 }
4254
4255 if !work_dirs_needing_exclude_update.is_empty() {
4256 let mut state = self.state.lock().await;
4257 for work_dir_abs_path in work_dirs_needing_exclude_update {
4258 if let Some((_, needs_update)) = state
4259 .snapshot
4260 .repo_exclude_by_work_dir_abs_path
4261 .get_mut(&work_dir_abs_path)
4262 {
4263 *needs_update = true;
4264 }
4265 }
4266 }
4267
4268 self.state.lock().await.snapshot.scan_id += 1;
4269
4270 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4271 log::debug!(
4272 "received fs events {:?}",
4273 relative_paths
4274 .iter()
4275 .map(|event_root| &event_root.path)
4276 .collect::<Vec<_>>()
4277 );
4278 self.reload_entries_for_paths(
4279 &root_path,
4280 &root_canonical_path,
4281 &relative_paths
4282 .iter()
4283 .map(|event_root| event_root.path.clone())
4284 .collect::<Vec<_>>(),
4285 events
4286 .into_iter()
4287 .map(|event| event.path)
4288 .collect::<Vec<_>>(),
4289 Some(scan_job_tx.clone()),
4290 )
4291 .await;
4292
4293 let affected_repo_roots = if !dot_git_abs_paths.is_empty() {
4294 self.update_git_repositories(dot_git_abs_paths).await
4295 } else {
4296 Vec::new()
4297 };
4298
4299 {
4300 let mut ignores_to_update = self.ignores_needing_update().await;
4301 ignores_to_update.extend(affected_repo_roots);
4302 let ignores_to_update = self.order_ignores(ignores_to_update).await;
4303 let snapshot = self.state.lock().await.snapshot.clone();
4304 self.update_ignore_statuses_for_paths(scan_job_tx, snapshot, ignores_to_update)
4305 .await;
4306 self.scan_dirs(false, scan_job_rx).await;
4307 }
4308
4309 {
4310 let mut state = self.state.lock().await;
4311 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4312 for (_, entry) in mem::take(&mut state.removed_entries) {
4313 state.scanned_dirs.remove(&entry.id);
4314 }
4315 }
4316 self.send_status_update(false, SmallVec::new(), &relative_paths)
4317 .await;
4318 }
4319
4320 async fn update_global_gitignore(&self, abs_path: &Path) {
4321 let ignore = build_gitignore(abs_path, self.fs.as_ref())
4322 .await
4323 .log_err()
4324 .map(Arc::new);
4325 let (prev_snapshot, ignore_stack, abs_path) = {
4326 let mut state = self.state.lock().await;
4327 state.snapshot.global_gitignore = ignore;
4328 let abs_path = state.snapshot.abs_path().clone();
4329 let ignore_stack = state
4330 .snapshot
4331 .ignore_stack_for_abs_path(&abs_path, true, self.fs.as_ref())
4332 .await;
4333 (state.snapshot.clone(), ignore_stack, abs_path)
4334 };
4335 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4336 self.update_ignore_statuses_for_paths(
4337 scan_job_tx,
4338 prev_snapshot,
4339 vec![(abs_path, ignore_stack)],
4340 )
4341 .await;
4342 self.scan_dirs(false, scan_job_rx).await;
4343 self.send_status_update(false, SmallVec::new(), &[]).await;
4344 }
4345
4346 async fn forcibly_load_paths(&self, paths: &[Arc<RelPath>]) -> bool {
4347 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4348 {
4349 let mut state = self.state.lock().await;
4350 let root_path = state.snapshot.abs_path.clone();
4351 for path in paths {
4352 for ancestor in path.ancestors() {
4353 if let Some(entry) = state.snapshot.entry_for_path(ancestor)
4354 && entry.kind == EntryKind::UnloadedDir
4355 {
4356 let abs_path = root_path.join(ancestor.as_std_path());
4357 state
4358 .enqueue_scan_dir(
4359 abs_path.into(),
4360 entry,
4361 &scan_job_tx,
4362 self.fs.as_ref(),
4363 )
4364 .await;
4365 state.paths_to_scan.insert(path.clone());
4366 break;
4367 }
4368 }
4369 }
4370 drop(scan_job_tx);
4371 }
4372 while let Ok(job) = scan_job_rx.recv().await {
4373 self.scan_dir(&job).await.log_err();
4374 }
4375
4376 !mem::take(&mut self.state.lock().await.paths_to_scan).is_empty()
4377 }
4378
4379 async fn scan_dirs(
4380 &self,
4381 enable_progress_updates: bool,
4382 scan_jobs_rx: channel::Receiver<ScanJob>,
4383 ) {
4384 if self
4385 .status_updates_tx
4386 .unbounded_send(ScanState::Started)
4387 .is_err()
4388 {
4389 return;
4390 }
4391
4392 let progress_update_count = AtomicUsize::new(0);
4393 self.executor
4394 .scoped_priority(Priority::Low, |scope| {
4395 for _ in 0..self.executor.num_cpus() {
4396 scope.spawn(async {
4397 let mut last_progress_update_count = 0;
4398 let progress_update_timer = self.progress_timer(enable_progress_updates).fuse();
4399 futures::pin_mut!(progress_update_timer);
4400
4401 loop {
4402 select_biased! {
4403 // Process any path refresh requests before moving on to process
4404 // the scan queue, so that user operations are prioritized.
4405 request = self.next_scan_request().fuse() => {
4406 let Ok(request) = request else { break };
4407 if !self.process_scan_request(request, true).await {
4408 return;
4409 }
4410 }
4411
4412 // Send periodic progress updates to the worktree. Use an atomic counter
4413 // to ensure that only one of the workers sends a progress update after
4414 // the update interval elapses.
4415 _ = progress_update_timer => {
4416 match progress_update_count.compare_exchange(
4417 last_progress_update_count,
4418 last_progress_update_count + 1,
4419 SeqCst,
4420 SeqCst
4421 ) {
4422 Ok(_) => {
4423 last_progress_update_count += 1;
4424 self.send_status_update(true, SmallVec::new(), &[])
4425 .await;
4426 }
4427 Err(count) => {
4428 last_progress_update_count = count;
4429 }
4430 }
4431 progress_update_timer.set(self.progress_timer(enable_progress_updates).fuse());
4432 }
4433
4434 // Recursively load directories from the file system.
4435 job = scan_jobs_rx.recv().fuse() => {
4436 let Ok(job) = job else { break };
4437 if let Err(err) = self.scan_dir(&job).await
4438 && job.path.is_empty() {
4439 log::error!("error scanning directory {:?}: {}", job.abs_path, err);
4440 }
4441 }
4442 }
4443 }
4444 });
4445 }
4446 })
4447 .await;
4448 }
4449
4450 async fn send_status_update(
4451 &self,
4452 scanning: bool,
4453 barrier: SmallVec<[barrier::Sender; 1]>,
4454 event_roots: &[EventRoot],
4455 ) -> bool {
4456 let mut state = self.state.lock().await;
4457 if state.changed_paths.is_empty() && event_roots.is_empty() && scanning {
4458 return true;
4459 }
4460
4461 let merged_event_roots = merge_event_roots(&state.changed_paths, event_roots);
4462
4463 let new_snapshot = state.snapshot.clone();
4464 let old_snapshot = mem::replace(&mut state.prev_snapshot, new_snapshot.snapshot.clone());
4465 let changes = build_diff(
4466 self.phase,
4467 &old_snapshot,
4468 &new_snapshot,
4469 &merged_event_roots,
4470 );
4471 state.changed_paths.clear();
4472
4473 self.status_updates_tx
4474 .unbounded_send(ScanState::Updated {
4475 snapshot: new_snapshot,
4476 changes,
4477 scanning,
4478 barrier,
4479 })
4480 .is_ok()
4481 }
4482
4483 async fn scan_dir(&self, job: &ScanJob) -> Result<()> {
4484 let root_abs_path;
4485 let root_char_bag;
4486 {
4487 let snapshot = &self.state.lock().await.snapshot;
4488 if self.settings.is_path_excluded(&job.path) {
4489 log::error!("skipping excluded directory {:?}", job.path);
4490 return Ok(());
4491 }
4492 log::trace!("scanning directory {:?}", job.path);
4493 root_abs_path = snapshot.abs_path().clone();
4494 root_char_bag = snapshot.root_char_bag;
4495 }
4496
4497 let next_entry_id = self.next_entry_id.clone();
4498 let mut ignore_stack = job.ignore_stack.clone();
4499 let mut new_ignore = None;
4500 let mut root_canonical_path = None;
4501 let mut new_entries: Vec<Entry> = Vec::new();
4502 let mut new_jobs: Vec<Option<ScanJob>> = Vec::new();
4503 let mut child_paths = self
4504 .fs
4505 .read_dir(&job.abs_path)
4506 .await?
4507 .filter_map(|entry| async {
4508 match entry {
4509 Ok(entry) => Some(entry),
4510 Err(error) => {
4511 log::error!("error processing entry {:?}", error);
4512 None
4513 }
4514 }
4515 })
4516 .collect::<Vec<_>>()
4517 .await;
4518
4519 // Ensure that .git and .gitignore are processed first.
4520 swap_to_front(&mut child_paths, GITIGNORE);
4521 swap_to_front(&mut child_paths, DOT_GIT);
4522
4523 if let Some(path) = child_paths.first()
4524 && path.ends_with(DOT_GIT)
4525 {
4526 ignore_stack.repo_root = Some(job.abs_path.clone());
4527 }
4528
4529 for child_abs_path in child_paths {
4530 let child_abs_path: Arc<Path> = child_abs_path.into();
4531 let child_name = child_abs_path.file_name().unwrap();
4532 let Some(child_path) = child_name
4533 .to_str()
4534 .and_then(|name| Some(job.path.join(RelPath::unix(name).ok()?)))
4535 else {
4536 continue;
4537 };
4538
4539 if child_name == DOT_GIT {
4540 let mut state = self.state.lock().await;
4541 state
4542 .insert_git_repository(
4543 child_path.clone(),
4544 self.fs.as_ref(),
4545 self.watcher.as_ref(),
4546 )
4547 .await;
4548 } else if child_name == GITIGNORE {
4549 match build_gitignore(&child_abs_path, self.fs.as_ref()).await {
4550 Ok(ignore) => {
4551 let ignore = Arc::new(ignore);
4552 ignore_stack = ignore_stack
4553 .append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4554 new_ignore = Some(ignore);
4555 }
4556 Err(error) => {
4557 log::error!(
4558 "error loading .gitignore file {:?} - {:?}",
4559 child_name,
4560 error
4561 );
4562 }
4563 }
4564 }
4565
4566 if self.settings.is_path_excluded(&child_path) {
4567 log::debug!("skipping excluded child entry {child_path:?}");
4568 self.state
4569 .lock()
4570 .await
4571 .remove_path(&child_path, self.watcher.as_ref());
4572 continue;
4573 }
4574
4575 let child_metadata = match self.fs.metadata(&child_abs_path).await {
4576 Ok(Some(metadata)) => metadata,
4577 Ok(None) => continue,
4578 Err(err) => {
4579 log::error!("error processing {:?}: {err:#}", child_abs_path.display());
4580 continue;
4581 }
4582 };
4583
4584 let mut child_entry = Entry::new(
4585 child_path.clone(),
4586 &child_metadata,
4587 ProjectEntryId::new(&next_entry_id),
4588 root_char_bag,
4589 None,
4590 );
4591
4592 if job.is_external {
4593 child_entry.is_external = true;
4594 } else if child_metadata.is_symlink {
4595 let canonical_path = match self.fs.canonicalize(&child_abs_path).await {
4596 Ok(path) => path,
4597 Err(err) => {
4598 log::error!("error reading target of symlink {child_abs_path:?}: {err:#}",);
4599 continue;
4600 }
4601 };
4602
4603 // lazily canonicalize the root path in order to determine if
4604 // symlinks point outside of the worktree.
4605 let root_canonical_path = match &root_canonical_path {
4606 Some(path) => path,
4607 None => match self.fs.canonicalize(&root_abs_path).await {
4608 Ok(path) => root_canonical_path.insert(path),
4609 Err(err) => {
4610 log::error!("error canonicalizing root {:?}: {:?}", root_abs_path, err);
4611 continue;
4612 }
4613 },
4614 };
4615
4616 if !canonical_path.starts_with(root_canonical_path) {
4617 child_entry.is_external = true;
4618 }
4619
4620 child_entry.canonical_path = Some(canonical_path.into());
4621 }
4622
4623 if child_entry.is_dir() {
4624 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, true);
4625 child_entry.is_always_included =
4626 self.settings.is_path_always_included(&child_path, true);
4627
4628 // Avoid recursing until crash in the case of a recursive symlink
4629 if job.ancestor_inodes.contains(&child_entry.inode) {
4630 new_jobs.push(None);
4631 } else {
4632 let mut ancestor_inodes = job.ancestor_inodes.clone();
4633 ancestor_inodes.insert(child_entry.inode);
4634
4635 new_jobs.push(Some(ScanJob {
4636 abs_path: child_abs_path.clone(),
4637 path: child_path,
4638 is_external: child_entry.is_external,
4639 ignore_stack: if child_entry.is_ignored {
4640 IgnoreStack::all()
4641 } else {
4642 ignore_stack.clone()
4643 },
4644 ancestor_inodes,
4645 scan_queue: job.scan_queue.clone(),
4646 }));
4647 }
4648 } else {
4649 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, false);
4650 child_entry.is_always_included =
4651 self.settings.is_path_always_included(&child_path, false);
4652 }
4653
4654 {
4655 let relative_path = job
4656 .path
4657 .join(RelPath::unix(child_name.to_str().unwrap()).unwrap());
4658 if self.is_path_private(&relative_path) {
4659 log::debug!("detected private file: {relative_path:?}");
4660 child_entry.is_private = true;
4661 }
4662 if self.settings.is_path_hidden(&relative_path) {
4663 log::debug!("detected hidden file: {relative_path:?}");
4664 child_entry.is_hidden = true;
4665 }
4666 }
4667
4668 new_entries.push(child_entry);
4669 }
4670
4671 let mut state = self.state.lock().await;
4672
4673 // Identify any subdirectories that should not be scanned.
4674 let mut job_ix = 0;
4675 for entry in &mut new_entries {
4676 state.reuse_entry_id(entry);
4677 if entry.is_dir() {
4678 if state.should_scan_directory(entry) {
4679 job_ix += 1;
4680 } else {
4681 log::debug!("defer scanning directory {:?}", entry.path);
4682 entry.kind = EntryKind::UnloadedDir;
4683 new_jobs.remove(job_ix);
4684 }
4685 }
4686 if entry.is_always_included {
4687 state
4688 .snapshot
4689 .always_included_entries
4690 .push(entry.path.clone());
4691 }
4692 }
4693
4694 state.populate_dir(job.path.clone(), new_entries, new_ignore);
4695 self.watcher.add(job.abs_path.as_ref()).log_err();
4696
4697 for new_job in new_jobs.into_iter().flatten() {
4698 job.scan_queue
4699 .try_send(new_job)
4700 .expect("channel is unbounded");
4701 }
4702
4703 Ok(())
4704 }
4705
4706 /// All list arguments should be sorted before calling this function
4707 async fn reload_entries_for_paths(
4708 &self,
4709 root_abs_path: &SanitizedPath,
4710 root_canonical_path: &SanitizedPath,
4711 relative_paths: &[Arc<RelPath>],
4712 abs_paths: Vec<PathBuf>,
4713 scan_queue_tx: Option<Sender<ScanJob>>,
4714 ) {
4715 // grab metadata for all requested paths
4716 let metadata = futures::future::join_all(
4717 abs_paths
4718 .iter()
4719 .map(|abs_path| async move {
4720 let metadata = self.fs.metadata(abs_path).await?;
4721 if let Some(metadata) = metadata {
4722 let canonical_path = self.fs.canonicalize(abs_path).await?;
4723
4724 // If we're on a case-insensitive filesystem (default on macOS), we want
4725 // to only ignore metadata for non-symlink files if their absolute-path matches
4726 // the canonical-path.
4727 // Because if not, this might be a case-only-renaming (`mv test.txt TEST.TXT`)
4728 // and we want to ignore the metadata for the old path (`test.txt`) so it's
4729 // treated as removed.
4730 if !self.fs_case_sensitive && !metadata.is_symlink {
4731 let canonical_file_name = canonical_path.file_name();
4732 let file_name = abs_path.file_name();
4733 if canonical_file_name != file_name {
4734 return Ok(None);
4735 }
4736 }
4737
4738 anyhow::Ok(Some((metadata, SanitizedPath::new_arc(&canonical_path))))
4739 } else {
4740 Ok(None)
4741 }
4742 })
4743 .collect::<Vec<_>>(),
4744 )
4745 .await;
4746
4747 let mut new_ancestor_repo = if relative_paths.iter().any(|path| path.is_empty()) {
4748 Some(discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await)
4749 } else {
4750 None
4751 };
4752
4753 let mut state = self.state.lock().await;
4754 let doing_recursive_update = scan_queue_tx.is_some();
4755
4756 // Remove any entries for paths that no longer exist or are being recursively
4757 // refreshed. Do this before adding any new entries, so that renames can be
4758 // detected regardless of the order of the paths.
4759 for (path, metadata) in relative_paths.iter().zip(metadata.iter()) {
4760 if matches!(metadata, Ok(None)) || doing_recursive_update {
4761 state.remove_path(path, self.watcher.as_ref());
4762 }
4763 }
4764
4765 for (path, metadata) in relative_paths.iter().zip(metadata.into_iter()) {
4766 let abs_path: Arc<Path> = root_abs_path.join(path.as_std_path()).into();
4767 match metadata {
4768 Ok(Some((metadata, canonical_path))) => {
4769 let ignore_stack = state
4770 .snapshot
4771 .ignore_stack_for_abs_path(&abs_path, metadata.is_dir, self.fs.as_ref())
4772 .await;
4773 let is_external = !canonical_path.starts_with(&root_canonical_path);
4774 let entry_id = state.entry_id_for(self.next_entry_id.as_ref(), path, &metadata);
4775 let mut fs_entry = Entry::new(
4776 path.clone(),
4777 &metadata,
4778 entry_id,
4779 state.snapshot.root_char_bag,
4780 if metadata.is_symlink {
4781 Some(canonical_path.as_path().to_path_buf().into())
4782 } else {
4783 None
4784 },
4785 );
4786
4787 let is_dir = fs_entry.is_dir();
4788 fs_entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, is_dir);
4789 fs_entry.is_external = is_external;
4790 fs_entry.is_private = self.is_path_private(path);
4791 fs_entry.is_always_included =
4792 self.settings.is_path_always_included(path, is_dir);
4793 fs_entry.is_hidden = self.settings.is_path_hidden(path);
4794
4795 if let (Some(scan_queue_tx), true) = (&scan_queue_tx, is_dir) {
4796 if state.should_scan_directory(&fs_entry)
4797 || (fs_entry.path.is_empty()
4798 && abs_path.file_name() == Some(OsStr::new(DOT_GIT)))
4799 {
4800 state
4801 .enqueue_scan_dir(
4802 abs_path,
4803 &fs_entry,
4804 scan_queue_tx,
4805 self.fs.as_ref(),
4806 )
4807 .await;
4808 } else {
4809 fs_entry.kind = EntryKind::UnloadedDir;
4810 }
4811 }
4812
4813 state
4814 .insert_entry(fs_entry.clone(), self.fs.as_ref(), self.watcher.as_ref())
4815 .await;
4816
4817 if path.is_empty()
4818 && let Some((ignores, exclude, repo)) = new_ancestor_repo.take()
4819 {
4820 log::trace!("updating ancestor git repository");
4821 state.snapshot.ignores_by_parent_abs_path.extend(ignores);
4822 if let Some((ancestor_dot_git, work_directory)) = repo {
4823 if let Some(exclude) = exclude {
4824 let work_directory_abs_path = self
4825 .state
4826 .lock()
4827 .await
4828 .snapshot
4829 .work_directory_abs_path(&work_directory);
4830
4831 state
4832 .snapshot
4833 .repo_exclude_by_work_dir_abs_path
4834 .insert(work_directory_abs_path.into(), (exclude, false));
4835 }
4836 state
4837 .insert_git_repository_for_path(
4838 work_directory,
4839 ancestor_dot_git.into(),
4840 self.fs.as_ref(),
4841 self.watcher.as_ref(),
4842 )
4843 .await
4844 .log_err();
4845 }
4846 }
4847 }
4848 Ok(None) => {
4849 self.remove_repo_path(path.clone(), &mut state.snapshot);
4850 }
4851 Err(err) => {
4852 log::error!("error reading file {abs_path:?} on event: {err:#}");
4853 }
4854 }
4855 }
4856
4857 util::extend_sorted(
4858 &mut state.changed_paths,
4859 relative_paths.iter().cloned(),
4860 usize::MAX,
4861 Ord::cmp,
4862 );
4863 }
4864
4865 fn remove_repo_path(&self, path: Arc<RelPath>, snapshot: &mut LocalSnapshot) -> Option<()> {
4866 if !path.components().any(|component| component == DOT_GIT)
4867 && let Some(local_repo) = snapshot.local_repo_for_work_directory_path(&path)
4868 {
4869 let id = local_repo.work_directory_id;
4870 log::debug!("remove repo path: {:?}", path);
4871 snapshot.git_repositories.remove(&id);
4872 return Some(());
4873 }
4874
4875 Some(())
4876 }
4877
4878 async fn update_ignore_statuses_for_paths(
4879 &self,
4880 scan_job_tx: Sender<ScanJob>,
4881 prev_snapshot: LocalSnapshot,
4882 ignores_to_update: Vec<(Arc<Path>, IgnoreStack)>,
4883 ) {
4884 let (ignore_queue_tx, ignore_queue_rx) = channel::unbounded();
4885 {
4886 for (parent_abs_path, ignore_stack) in ignores_to_update {
4887 ignore_queue_tx
4888 .send_blocking(UpdateIgnoreStatusJob {
4889 abs_path: parent_abs_path,
4890 ignore_stack,
4891 ignore_queue: ignore_queue_tx.clone(),
4892 scan_queue: scan_job_tx.clone(),
4893 })
4894 .unwrap();
4895 }
4896 }
4897 drop(ignore_queue_tx);
4898
4899 self.executor
4900 .scoped(|scope| {
4901 for _ in 0..self.executor.num_cpus() {
4902 scope.spawn(async {
4903 loop {
4904 select_biased! {
4905 // Process any path refresh requests before moving on to process
4906 // the queue of ignore statuses.
4907 request = self.next_scan_request().fuse() => {
4908 let Ok(request) = request else { break };
4909 if !self.process_scan_request(request, true).await {
4910 return;
4911 }
4912 }
4913
4914 // Recursively process directories whose ignores have changed.
4915 job = ignore_queue_rx.recv().fuse() => {
4916 let Ok(job) = job else { break };
4917 self.update_ignore_status(job, &prev_snapshot).await;
4918 }
4919 }
4920 }
4921 });
4922 }
4923 })
4924 .await;
4925 }
4926
4927 async fn ignores_needing_update(&self) -> Vec<Arc<Path>> {
4928 let mut ignores_to_update = Vec::new();
4929 let mut excludes_to_load: Vec<(Arc<Path>, PathBuf)> = Vec::new();
4930
4931 // First pass: collect updates and drop stale entries without awaiting.
4932 {
4933 let snapshot = &mut self.state.lock().await.snapshot;
4934 let abs_path = snapshot.abs_path.clone();
4935 let mut repo_exclude_keys_to_remove: Vec<Arc<Path>> = Vec::new();
4936
4937 for (work_dir_abs_path, (_, needs_update)) in
4938 snapshot.repo_exclude_by_work_dir_abs_path.iter_mut()
4939 {
4940 let repository = snapshot
4941 .git_repositories
4942 .iter()
4943 .find(|(_, repo)| &repo.work_directory_abs_path == work_dir_abs_path);
4944
4945 if *needs_update {
4946 *needs_update = false;
4947 ignores_to_update.push(work_dir_abs_path.clone());
4948
4949 if let Some((_, repository)) = repository {
4950 let exclude_abs_path = repository.common_dir_abs_path.join(REPO_EXCLUDE);
4951 excludes_to_load.push((work_dir_abs_path.clone(), exclude_abs_path));
4952 }
4953 }
4954
4955 if repository.is_none() {
4956 repo_exclude_keys_to_remove.push(work_dir_abs_path.clone());
4957 }
4958 }
4959
4960 for key in repo_exclude_keys_to_remove {
4961 snapshot.repo_exclude_by_work_dir_abs_path.remove(&key);
4962 }
4963
4964 snapshot
4965 .ignores_by_parent_abs_path
4966 .retain(|parent_abs_path, (_, needs_update)| {
4967 if let Ok(parent_path) = parent_abs_path.strip_prefix(abs_path.as_path())
4968 && let Some(parent_path) =
4969 RelPath::new(&parent_path, PathStyle::local()).log_err()
4970 {
4971 if *needs_update {
4972 *needs_update = false;
4973 if snapshot.snapshot.entry_for_path(&parent_path).is_some() {
4974 ignores_to_update.push(parent_abs_path.clone());
4975 }
4976 }
4977
4978 let ignore_path = parent_path.join(RelPath::unix(GITIGNORE).unwrap());
4979 if snapshot.snapshot.entry_for_path(&ignore_path).is_none() {
4980 return false;
4981 }
4982 }
4983 true
4984 });
4985 }
4986
4987 // Load gitignores asynchronously (outside the lock)
4988 let mut loaded_excludes: Vec<(Arc<Path>, Arc<Gitignore>)> = Vec::new();
4989 for (work_dir_abs_path, exclude_abs_path) in excludes_to_load {
4990 if let Ok(current_exclude) = build_gitignore(&exclude_abs_path, self.fs.as_ref()).await
4991 {
4992 loaded_excludes.push((work_dir_abs_path, Arc::new(current_exclude)));
4993 }
4994 }
4995
4996 // Second pass: apply updates.
4997 if !loaded_excludes.is_empty() {
4998 let snapshot = &mut self.state.lock().await.snapshot;
4999
5000 for (work_dir_abs_path, exclude) in loaded_excludes {
5001 if let Some((existing_exclude, _)) = snapshot
5002 .repo_exclude_by_work_dir_abs_path
5003 .get_mut(&work_dir_abs_path)
5004 {
5005 *existing_exclude = exclude;
5006 }
5007 }
5008 }
5009
5010 ignores_to_update
5011 }
5012
5013 async fn order_ignores(&self, mut ignores: Vec<Arc<Path>>) -> Vec<(Arc<Path>, IgnoreStack)> {
5014 let fs = self.fs.clone();
5015 let snapshot = self.state.lock().await.snapshot.clone();
5016 ignores.sort_unstable();
5017 let mut ignores_to_update = ignores.into_iter().peekable();
5018
5019 let mut result = vec![];
5020 while let Some(parent_abs_path) = ignores_to_update.next() {
5021 while ignores_to_update
5022 .peek()
5023 .map_or(false, |p| p.starts_with(&parent_abs_path))
5024 {
5025 ignores_to_update.next().unwrap();
5026 }
5027 let ignore_stack = snapshot
5028 .ignore_stack_for_abs_path(&parent_abs_path, true, fs.as_ref())
5029 .await;
5030 result.push((parent_abs_path, ignore_stack));
5031 }
5032
5033 result
5034 }
5035
5036 async fn update_ignore_status(&self, job: UpdateIgnoreStatusJob, snapshot: &LocalSnapshot) {
5037 log::trace!("update ignore status {:?}", job.abs_path);
5038
5039 let mut ignore_stack = job.ignore_stack;
5040 if let Some((ignore, _)) = snapshot.ignores_by_parent_abs_path.get(&job.abs_path) {
5041 ignore_stack =
5042 ignore_stack.append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
5043 }
5044
5045 let mut entries_by_id_edits = Vec::new();
5046 let mut entries_by_path_edits = Vec::new();
5047 let Some(path) = job
5048 .abs_path
5049 .strip_prefix(snapshot.abs_path.as_path())
5050 .map_err(|_| {
5051 anyhow::anyhow!(
5052 "Failed to strip prefix '{}' from path '{}'",
5053 snapshot.abs_path.as_path().display(),
5054 job.abs_path.display()
5055 )
5056 })
5057 .log_err()
5058 else {
5059 return;
5060 };
5061
5062 let Some(path) = RelPath::new(&path, PathStyle::local()).log_err() else {
5063 return;
5064 };
5065
5066 if let Ok(Some(metadata)) = self.fs.metadata(&job.abs_path.join(DOT_GIT)).await
5067 && metadata.is_dir
5068 {
5069 ignore_stack.repo_root = Some(job.abs_path.clone());
5070 }
5071
5072 for mut entry in snapshot.child_entries(&path).cloned() {
5073 let was_ignored = entry.is_ignored;
5074 let abs_path: Arc<Path> = snapshot.absolutize(&entry.path).into();
5075 entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, entry.is_dir());
5076
5077 if entry.is_dir() {
5078 let child_ignore_stack = if entry.is_ignored {
5079 IgnoreStack::all()
5080 } else {
5081 ignore_stack.clone()
5082 };
5083
5084 // Scan any directories that were previously ignored and weren't previously scanned.
5085 if was_ignored && !entry.is_ignored && entry.kind.is_unloaded() {
5086 let state = self.state.lock().await;
5087 if state.should_scan_directory(&entry) {
5088 state
5089 .enqueue_scan_dir(
5090 abs_path.clone(),
5091 &entry,
5092 &job.scan_queue,
5093 self.fs.as_ref(),
5094 )
5095 .await;
5096 }
5097 }
5098
5099 job.ignore_queue
5100 .send(UpdateIgnoreStatusJob {
5101 abs_path: abs_path.clone(),
5102 ignore_stack: child_ignore_stack,
5103 ignore_queue: job.ignore_queue.clone(),
5104 scan_queue: job.scan_queue.clone(),
5105 })
5106 .await
5107 .unwrap();
5108 }
5109
5110 if entry.is_ignored != was_ignored {
5111 let mut path_entry = snapshot.entries_by_id.get(&entry.id, ()).unwrap().clone();
5112 path_entry.scan_id = snapshot.scan_id;
5113 path_entry.is_ignored = entry.is_ignored;
5114 entries_by_id_edits.push(Edit::Insert(path_entry));
5115 entries_by_path_edits.push(Edit::Insert(entry));
5116 }
5117 }
5118
5119 let state = &mut self.state.lock().await;
5120 for edit in &entries_by_path_edits {
5121 if let Edit::Insert(entry) = edit
5122 && let Err(ix) = state.changed_paths.binary_search(&entry.path)
5123 {
5124 state.changed_paths.insert(ix, entry.path.clone());
5125 }
5126 }
5127
5128 state
5129 .snapshot
5130 .entries_by_path
5131 .edit(entries_by_path_edits, ());
5132 state.snapshot.entries_by_id.edit(entries_by_id_edits, ());
5133 }
5134
5135 async fn update_git_repositories(&self, dot_git_paths: Vec<PathBuf>) -> Vec<Arc<Path>> {
5136 log::trace!("reloading repositories: {dot_git_paths:?}");
5137 let mut state = self.state.lock().await;
5138 let scan_id = state.snapshot.scan_id;
5139 let mut affected_repo_roots = Vec::new();
5140 for dot_git_dir in dot_git_paths {
5141 let existing_repository_entry =
5142 state
5143 .snapshot
5144 .git_repositories
5145 .iter()
5146 .find_map(|(_, repo)| {
5147 let dot_git_dir = SanitizedPath::new(&dot_git_dir);
5148 if SanitizedPath::new(repo.common_dir_abs_path.as_ref()) == dot_git_dir
5149 || SanitizedPath::new(repo.repository_dir_abs_path.as_ref())
5150 == dot_git_dir
5151 {
5152 Some(repo.clone())
5153 } else {
5154 None
5155 }
5156 });
5157
5158 match existing_repository_entry {
5159 None => {
5160 let Ok(relative) = dot_git_dir.strip_prefix(state.snapshot.abs_path()) else {
5161 debug_panic!(
5162 "update_git_repositories called with .git directory outside the worktree root"
5163 );
5164 return Vec::new();
5165 };
5166 affected_repo_roots.push(dot_git_dir.parent().unwrap().into());
5167 state
5168 .insert_git_repository(
5169 RelPath::new(relative, PathStyle::local())
5170 .unwrap()
5171 .into_arc(),
5172 self.fs.as_ref(),
5173 self.watcher.as_ref(),
5174 )
5175 .await;
5176 }
5177 Some(local_repository) => {
5178 state.snapshot.git_repositories.update(
5179 &local_repository.work_directory_id,
5180 |entry| {
5181 entry.git_dir_scan_id = scan_id;
5182 },
5183 );
5184 }
5185 };
5186 }
5187
5188 // Remove any git repositories whose .git entry no longer exists.
5189 let snapshot = &mut state.snapshot;
5190 let mut ids_to_preserve = HashSet::default();
5191 for (&work_directory_id, entry) in snapshot.git_repositories.iter() {
5192 let exists_in_snapshot =
5193 snapshot
5194 .entry_for_id(work_directory_id)
5195 .is_some_and(|entry| {
5196 snapshot
5197 .entry_for_path(&entry.path.join(RelPath::unix(DOT_GIT).unwrap()))
5198 .is_some()
5199 });
5200
5201 if exists_in_snapshot
5202 || matches!(
5203 self.fs.metadata(&entry.common_dir_abs_path).await,
5204 Ok(Some(_))
5205 )
5206 {
5207 ids_to_preserve.insert(work_directory_id);
5208 }
5209 }
5210
5211 snapshot
5212 .git_repositories
5213 .retain(|work_directory_id, entry| {
5214 let preserve = ids_to_preserve.contains(work_directory_id);
5215 if !preserve {
5216 affected_repo_roots.push(entry.dot_git_abs_path.parent().unwrap().into());
5217 snapshot
5218 .repo_exclude_by_work_dir_abs_path
5219 .remove(&entry.work_directory_abs_path);
5220 }
5221 preserve
5222 });
5223
5224 affected_repo_roots
5225 }
5226
5227 async fn progress_timer(&self, running: bool) {
5228 if !running {
5229 return futures::future::pending().await;
5230 }
5231
5232 #[cfg(feature = "test-support")]
5233 if self.fs.is_fake() {
5234 return self.executor.simulate_random_delay().await;
5235 }
5236
5237 self.executor.timer(FS_WATCH_LATENCY).await
5238 }
5239
5240 fn is_path_private(&self, path: &RelPath) -> bool {
5241 !self.share_private_files && self.settings.is_path_private(path)
5242 }
5243
5244 async fn next_scan_request(&self) -> Result<ScanRequest> {
5245 let mut request = self.scan_requests_rx.recv().await?;
5246 while let Ok(next_request) = self.scan_requests_rx.try_recv() {
5247 request.relative_paths.extend(next_request.relative_paths);
5248 request.done.extend(next_request.done);
5249 }
5250 Ok(request)
5251 }
5252}
5253
5254async fn discover_ancestor_git_repo(
5255 fs: Arc<dyn Fs>,
5256 root_abs_path: &SanitizedPath,
5257) -> (
5258 HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
5259 Option<Arc<Gitignore>>,
5260 Option<(PathBuf, WorkDirectory)>,
5261) {
5262 let mut exclude = None;
5263 let mut ignores = HashMap::default();
5264 for (index, ancestor) in root_abs_path.as_path().ancestors().enumerate() {
5265 if index != 0 {
5266 if ancestor == paths::home_dir() {
5267 // Unless $HOME is itself the worktree root, don't consider it as a
5268 // containing git repository---expensive and likely unwanted.
5269 break;
5270 } else if let Ok(ignore) = build_gitignore(&ancestor.join(GITIGNORE), fs.as_ref()).await
5271 {
5272 ignores.insert(ancestor.into(), (ignore.into(), false));
5273 }
5274 }
5275
5276 let ancestor_dot_git = ancestor.join(DOT_GIT);
5277 log::trace!("considering ancestor: {ancestor_dot_git:?}");
5278 // Check whether the directory or file called `.git` exists (in the
5279 // case of worktrees it's a file.)
5280 if fs
5281 .metadata(&ancestor_dot_git)
5282 .await
5283 .is_ok_and(|metadata| metadata.is_some())
5284 {
5285 if index != 0 {
5286 // We canonicalize, since the FS events use the canonicalized path.
5287 if let Some(ancestor_dot_git) = fs.canonicalize(&ancestor_dot_git).await.log_err() {
5288 let location_in_repo = root_abs_path
5289 .as_path()
5290 .strip_prefix(ancestor)
5291 .unwrap()
5292 .into();
5293 log::info!("inserting parent git repo for this worktree: {location_in_repo:?}");
5294 // We associate the external git repo with our root folder and
5295 // also mark where in the git repo the root folder is located.
5296 return (
5297 ignores,
5298 exclude,
5299 Some((
5300 ancestor_dot_git,
5301 WorkDirectory::AboveProject {
5302 absolute_path: ancestor.into(),
5303 location_in_repo,
5304 },
5305 )),
5306 );
5307 };
5308 }
5309
5310 let repo_exclude_abs_path = ancestor_dot_git.join(REPO_EXCLUDE);
5311 if let Ok(repo_exclude) = build_gitignore(&repo_exclude_abs_path, fs.as_ref()).await {
5312 exclude = Some(Arc::new(repo_exclude));
5313 }
5314
5315 // Reached root of git repository.
5316 break;
5317 }
5318 }
5319
5320 (ignores, exclude, None)
5321}
5322
5323fn merge_event_roots(changed_paths: &[Arc<RelPath>], event_roots: &[EventRoot]) -> Vec<EventRoot> {
5324 let mut merged_event_roots = Vec::with_capacity(changed_paths.len() + event_roots.len());
5325 let mut changed_paths = changed_paths.iter().peekable();
5326 let mut event_roots = event_roots.iter().peekable();
5327 while let (Some(path), Some(event_root)) = (changed_paths.peek(), event_roots.peek()) {
5328 match path.cmp(&&event_root.path) {
5329 Ordering::Less => {
5330 merged_event_roots.push(EventRoot {
5331 path: (*changed_paths.next().expect("peeked changed path")).clone(),
5332 was_rescanned: false,
5333 });
5334 }
5335 Ordering::Equal => {
5336 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5337 changed_paths.next();
5338 }
5339 Ordering::Greater => {
5340 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5341 }
5342 }
5343 }
5344 merged_event_roots.extend(changed_paths.map(|path| EventRoot {
5345 path: path.clone(),
5346 was_rescanned: false,
5347 }));
5348 merged_event_roots.extend(event_roots.cloned());
5349 merged_event_roots
5350}
5351
5352fn build_diff(
5353 phase: BackgroundScannerPhase,
5354 old_snapshot: &Snapshot,
5355 new_snapshot: &Snapshot,
5356 event_roots: &[EventRoot],
5357) -> UpdatedEntriesSet {
5358 use BackgroundScannerPhase::*;
5359 use PathChange::{Added, AddedOrUpdated, Loaded, Removed, Updated};
5360
5361 // Identify which paths have changed. Use the known set of changed
5362 // parent paths to optimize the search.
5363 let mut changes = Vec::new();
5364
5365 let mut old_paths = old_snapshot.entries_by_path.cursor::<PathKey>(());
5366 let mut new_paths = new_snapshot.entries_by_path.cursor::<PathKey>(());
5367 let mut last_newly_loaded_dir_path = None;
5368 old_paths.next();
5369 new_paths.next();
5370 for event_root in event_roots {
5371 let path = PathKey(event_root.path.clone());
5372 if old_paths.item().is_some_and(|e| e.path < path.0) {
5373 old_paths.seek_forward(&path, Bias::Left);
5374 }
5375 if new_paths.item().is_some_and(|e| e.path < path.0) {
5376 new_paths.seek_forward(&path, Bias::Left);
5377 }
5378 loop {
5379 match (old_paths.item(), new_paths.item()) {
5380 (Some(old_entry), Some(new_entry)) => {
5381 if old_entry.path > path.0
5382 && new_entry.path > path.0
5383 && !old_entry.path.starts_with(&path.0)
5384 && !new_entry.path.starts_with(&path.0)
5385 {
5386 break;
5387 }
5388
5389 match Ord::cmp(&old_entry.path, &new_entry.path) {
5390 Ordering::Less => {
5391 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5392 old_paths.next();
5393 }
5394 Ordering::Equal => {
5395 if phase == EventsReceivedDuringInitialScan {
5396 if old_entry.id != new_entry.id {
5397 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5398 }
5399 // If the worktree was not fully initialized when this event was generated,
5400 // we can't know whether this entry was added during the scan or whether
5401 // it was merely updated.
5402 changes.push((
5403 new_entry.path.clone(),
5404 new_entry.id,
5405 AddedOrUpdated,
5406 ));
5407 } else if old_entry.id != new_entry.id {
5408 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5409 changes.push((new_entry.path.clone(), new_entry.id, Added));
5410 } else if old_entry != new_entry {
5411 if old_entry.kind.is_unloaded() {
5412 last_newly_loaded_dir_path = Some(&new_entry.path);
5413 changes.push((new_entry.path.clone(), new_entry.id, Loaded));
5414 } else {
5415 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5416 }
5417 } else if event_root.was_rescanned {
5418 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5419 }
5420 old_paths.next();
5421 new_paths.next();
5422 }
5423 Ordering::Greater => {
5424 let is_newly_loaded = phase == InitialScan
5425 || last_newly_loaded_dir_path
5426 .as_ref()
5427 .is_some_and(|dir| new_entry.path.starts_with(dir));
5428 changes.push((
5429 new_entry.path.clone(),
5430 new_entry.id,
5431 if is_newly_loaded { Loaded } else { Added },
5432 ));
5433 new_paths.next();
5434 }
5435 }
5436 }
5437 (Some(old_entry), None) => {
5438 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5439 old_paths.next();
5440 }
5441 (None, Some(new_entry)) => {
5442 let is_newly_loaded = phase == InitialScan
5443 || last_newly_loaded_dir_path
5444 .as_ref()
5445 .is_some_and(|dir| new_entry.path.starts_with(dir));
5446 changes.push((
5447 new_entry.path.clone(),
5448 new_entry.id,
5449 if is_newly_loaded { Loaded } else { Added },
5450 ));
5451 new_paths.next();
5452 }
5453 (None, None) => break,
5454 }
5455 }
5456 }
5457
5458 changes.into()
5459}
5460
5461fn swap_to_front(child_paths: &mut Vec<PathBuf>, file: &str) {
5462 let position = child_paths
5463 .iter()
5464 .position(|path| path.file_name().unwrap() == file);
5465 if let Some(position) = position {
5466 let temp = child_paths.remove(position);
5467 child_paths.insert(0, temp);
5468 }
5469}
5470
5471fn char_bag_for_path(root_char_bag: CharBag, path: &RelPath) -> CharBag {
5472 let mut result = root_char_bag;
5473 result.extend(path.as_unix_str().chars().map(|c| c.to_ascii_lowercase()));
5474 result
5475}
5476
5477#[derive(Debug)]
5478struct ScanJob {
5479 abs_path: Arc<Path>,
5480 path: Arc<RelPath>,
5481 ignore_stack: IgnoreStack,
5482 scan_queue: Sender<ScanJob>,
5483 ancestor_inodes: TreeSet<u64>,
5484 is_external: bool,
5485}
5486
5487struct UpdateIgnoreStatusJob {
5488 abs_path: Arc<Path>,
5489 ignore_stack: IgnoreStack,
5490 ignore_queue: Sender<UpdateIgnoreStatusJob>,
5491 scan_queue: Sender<ScanJob>,
5492}
5493
5494pub trait WorktreeModelHandle {
5495 #[cfg(feature = "test-support")]
5496 fn flush_fs_events<'a>(
5497 &self,
5498 cx: &'a mut gpui::TestAppContext,
5499 ) -> futures::future::LocalBoxFuture<'a, ()>;
5500
5501 #[cfg(feature = "test-support")]
5502 fn flush_fs_events_in_root_git_repository<'a>(
5503 &self,
5504 cx: &'a mut gpui::TestAppContext,
5505 ) -> futures::future::LocalBoxFuture<'a, ()>;
5506}
5507
5508impl WorktreeModelHandle for Entity<Worktree> {
5509 // When the worktree's FS event stream sometimes delivers "redundant" events for FS changes that
5510 // occurred before the worktree was constructed. These events can cause the worktree to perform
5511 // extra directory scans, and emit extra scan-state notifications.
5512 //
5513 // This function mutates the worktree's directory and waits for those mutations to be picked up,
5514 // to ensure that all redundant FS events have already been processed.
5515 #[cfg(feature = "test-support")]
5516 fn flush_fs_events<'a>(
5517 &self,
5518 cx: &'a mut gpui::TestAppContext,
5519 ) -> futures::future::LocalBoxFuture<'a, ()> {
5520 let file_name = "fs-event-sentinel";
5521
5522 let tree = self.clone();
5523 let (fs, root_path) = self.read_with(cx, |tree, _| {
5524 let tree = tree.as_local().unwrap();
5525 (tree.fs.clone(), tree.abs_path.clone())
5526 });
5527
5528 async move {
5529 // Subscribe to events BEFORE creating the file to avoid race condition
5530 // where events fire before subscription is set up
5531 let mut events = cx.events(&tree);
5532
5533 fs.create_file(&root_path.join(file_name), Default::default())
5534 .await
5535 .unwrap();
5536
5537 // Check if condition is already met before waiting for events
5538 let file_exists = || {
5539 tree.read_with(cx, |tree, _| {
5540 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5541 .is_some()
5542 })
5543 };
5544
5545 // Use select to avoid blocking indefinitely if events are delayed
5546 while !file_exists() {
5547 futures::select_biased! {
5548 _ = events.next() => {}
5549 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5550 }
5551 }
5552
5553 fs.remove_file(&root_path.join(file_name), Default::default())
5554 .await
5555 .unwrap();
5556
5557 // Check if condition is already met before waiting for events
5558 let file_gone = || {
5559 tree.read_with(cx, |tree, _| {
5560 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5561 .is_none()
5562 })
5563 };
5564
5565 // Use select to avoid blocking indefinitely if events are delayed
5566 while !file_gone() {
5567 futures::select_biased! {
5568 _ = events.next() => {}
5569 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5570 }
5571 }
5572
5573 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5574 .await;
5575 }
5576 .boxed_local()
5577 }
5578
5579 // This function is similar to flush_fs_events, except that it waits for events to be flushed in
5580 // the .git folder of the root repository.
5581 // The reason for its existence is that a repository's .git folder might live *outside* of the
5582 // worktree and thus its FS events might go through a different path.
5583 // In order to flush those, we need to create artificial events in the .git folder and wait
5584 // for the repository to be reloaded.
5585 #[cfg(feature = "test-support")]
5586 fn flush_fs_events_in_root_git_repository<'a>(
5587 &self,
5588 cx: &'a mut gpui::TestAppContext,
5589 ) -> futures::future::LocalBoxFuture<'a, ()> {
5590 let file_name = "fs-event-sentinel";
5591
5592 let tree = self.clone();
5593 let (fs, root_path, mut git_dir_scan_id) = self.read_with(cx, |tree, _| {
5594 let tree = tree.as_local().unwrap();
5595 let local_repo_entry = tree
5596 .git_repositories
5597 .values()
5598 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5599 .unwrap();
5600 (
5601 tree.fs.clone(),
5602 local_repo_entry.common_dir_abs_path.clone(),
5603 local_repo_entry.git_dir_scan_id,
5604 )
5605 });
5606
5607 let scan_id_increased = |tree: &mut Worktree, git_dir_scan_id: &mut usize| {
5608 let tree = tree.as_local().unwrap();
5609 // let repository = tree.repositories.first().unwrap();
5610 let local_repo_entry = tree
5611 .git_repositories
5612 .values()
5613 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5614 .unwrap();
5615
5616 if local_repo_entry.git_dir_scan_id > *git_dir_scan_id {
5617 *git_dir_scan_id = local_repo_entry.git_dir_scan_id;
5618 true
5619 } else {
5620 false
5621 }
5622 };
5623
5624 async move {
5625 // Subscribe to events BEFORE creating the file to avoid race condition
5626 // where events fire before subscription is set up
5627 let mut events = cx.events(&tree);
5628
5629 fs.create_file(&root_path.join(file_name), Default::default())
5630 .await
5631 .unwrap();
5632
5633 // Use select to avoid blocking indefinitely if events are delayed
5634 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5635 futures::select_biased! {
5636 _ = events.next() => {}
5637 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5638 }
5639 }
5640
5641 fs.remove_file(&root_path.join(file_name), Default::default())
5642 .await
5643 .unwrap();
5644
5645 // Use select to avoid blocking indefinitely if events are delayed
5646 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5647 futures::select_biased! {
5648 _ = events.next() => {}
5649 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5650 }
5651 }
5652
5653 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5654 .await;
5655 }
5656 .boxed_local()
5657 }
5658}
5659
5660#[derive(Clone, Debug)]
5661struct TraversalProgress<'a> {
5662 max_path: &'a RelPath,
5663 count: usize,
5664 non_ignored_count: usize,
5665 file_count: usize,
5666 non_ignored_file_count: usize,
5667}
5668
5669impl TraversalProgress<'_> {
5670 fn count(&self, include_files: bool, include_dirs: bool, include_ignored: bool) -> usize {
5671 match (include_files, include_dirs, include_ignored) {
5672 (true, true, true) => self.count,
5673 (true, true, false) => self.non_ignored_count,
5674 (true, false, true) => self.file_count,
5675 (true, false, false) => self.non_ignored_file_count,
5676 (false, true, true) => self.count - self.file_count,
5677 (false, true, false) => self.non_ignored_count - self.non_ignored_file_count,
5678 (false, false, _) => 0,
5679 }
5680 }
5681}
5682
5683impl<'a> sum_tree::Dimension<'a, EntrySummary> for TraversalProgress<'a> {
5684 fn zero(_cx: ()) -> Self {
5685 Default::default()
5686 }
5687
5688 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
5689 self.max_path = summary.max_path.as_ref();
5690 self.count += summary.count;
5691 self.non_ignored_count += summary.non_ignored_count;
5692 self.file_count += summary.file_count;
5693 self.non_ignored_file_count += summary.non_ignored_file_count;
5694 }
5695}
5696
5697impl Default for TraversalProgress<'_> {
5698 fn default() -> Self {
5699 Self {
5700 max_path: RelPath::empty(),
5701 count: 0,
5702 non_ignored_count: 0,
5703 file_count: 0,
5704 non_ignored_file_count: 0,
5705 }
5706 }
5707}
5708
5709#[derive(Debug)]
5710pub struct Traversal<'a> {
5711 snapshot: &'a Snapshot,
5712 cursor: sum_tree::Cursor<'a, 'static, Entry, TraversalProgress<'a>>,
5713 include_ignored: bool,
5714 include_files: bool,
5715 include_dirs: bool,
5716}
5717
5718impl<'a> Traversal<'a> {
5719 fn new(
5720 snapshot: &'a Snapshot,
5721 include_files: bool,
5722 include_dirs: bool,
5723 include_ignored: bool,
5724 start_path: &RelPath,
5725 ) -> Self {
5726 let mut cursor = snapshot.entries_by_path.cursor(());
5727 cursor.seek(&TraversalTarget::path(start_path), Bias::Left);
5728 let mut traversal = Self {
5729 snapshot,
5730 cursor,
5731 include_files,
5732 include_dirs,
5733 include_ignored,
5734 };
5735 if traversal.end_offset() == traversal.start_offset() {
5736 traversal.next();
5737 }
5738 traversal
5739 }
5740
5741 pub fn advance(&mut self) -> bool {
5742 self.advance_by(1)
5743 }
5744
5745 pub fn advance_by(&mut self, count: usize) -> bool {
5746 self.cursor.seek_forward(
5747 &TraversalTarget::Count {
5748 count: self.end_offset() + count,
5749 include_dirs: self.include_dirs,
5750 include_files: self.include_files,
5751 include_ignored: self.include_ignored,
5752 },
5753 Bias::Left,
5754 )
5755 }
5756
5757 pub fn advance_to_sibling(&mut self) -> bool {
5758 while let Some(entry) = self.cursor.item() {
5759 self.cursor
5760 .seek_forward(&TraversalTarget::successor(&entry.path), Bias::Left);
5761 if let Some(entry) = self.cursor.item()
5762 && (self.include_files || !entry.is_file())
5763 && (self.include_dirs || !entry.is_dir())
5764 && (self.include_ignored || !entry.is_ignored || entry.is_always_included)
5765 {
5766 return true;
5767 }
5768 }
5769 false
5770 }
5771
5772 pub fn back_to_parent(&mut self) -> bool {
5773 let Some(parent_path) = self.cursor.item().and_then(|entry| entry.path.parent()) else {
5774 return false;
5775 };
5776 self.cursor
5777 .seek(&TraversalTarget::path(parent_path), Bias::Left)
5778 }
5779
5780 pub fn entry(&self) -> Option<&'a Entry> {
5781 self.cursor.item()
5782 }
5783
5784 pub fn snapshot(&self) -> &'a Snapshot {
5785 self.snapshot
5786 }
5787
5788 pub fn start_offset(&self) -> usize {
5789 self.cursor
5790 .start()
5791 .count(self.include_files, self.include_dirs, self.include_ignored)
5792 }
5793
5794 pub fn end_offset(&self) -> usize {
5795 self.cursor
5796 .end()
5797 .count(self.include_files, self.include_dirs, self.include_ignored)
5798 }
5799}
5800
5801impl<'a> Iterator for Traversal<'a> {
5802 type Item = &'a Entry;
5803
5804 fn next(&mut self) -> Option<Self::Item> {
5805 if let Some(item) = self.entry() {
5806 self.advance();
5807 Some(item)
5808 } else {
5809 None
5810 }
5811 }
5812}
5813
5814#[derive(Debug, Clone, Copy)]
5815pub enum PathTarget<'a> {
5816 Path(&'a RelPath),
5817 Successor(&'a RelPath),
5818}
5819
5820impl PathTarget<'_> {
5821 fn cmp_path(&self, other: &RelPath) -> Ordering {
5822 match self {
5823 PathTarget::Path(path) => path.cmp(&other),
5824 PathTarget::Successor(path) => {
5825 if other.starts_with(path) {
5826 Ordering::Greater
5827 } else {
5828 Ordering::Equal
5829 }
5830 }
5831 }
5832 }
5833}
5834
5835impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, PathProgress<'a>> for PathTarget<'_> {
5836 fn cmp(&self, cursor_location: &PathProgress<'a>, _: S::Context<'_>) -> Ordering {
5837 self.cmp_path(cursor_location.max_path)
5838 }
5839}
5840
5841impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, TraversalProgress<'a>> for PathTarget<'_> {
5842 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: S::Context<'_>) -> Ordering {
5843 self.cmp_path(cursor_location.max_path)
5844 }
5845}
5846
5847#[derive(Debug)]
5848enum TraversalTarget<'a> {
5849 Path(PathTarget<'a>),
5850 Count {
5851 count: usize,
5852 include_files: bool,
5853 include_ignored: bool,
5854 include_dirs: bool,
5855 },
5856}
5857
5858impl<'a> TraversalTarget<'a> {
5859 fn path(path: &'a RelPath) -> Self {
5860 Self::Path(PathTarget::Path(path))
5861 }
5862
5863 fn successor(path: &'a RelPath) -> Self {
5864 Self::Path(PathTarget::Successor(path))
5865 }
5866
5867 fn cmp_progress(&self, progress: &TraversalProgress) -> Ordering {
5868 match self {
5869 TraversalTarget::Path(path) => path.cmp_path(progress.max_path),
5870 TraversalTarget::Count {
5871 count,
5872 include_files,
5873 include_dirs,
5874 include_ignored,
5875 } => Ord::cmp(
5876 count,
5877 &progress.count(*include_files, *include_dirs, *include_ignored),
5878 ),
5879 }
5880 }
5881}
5882
5883impl<'a> SeekTarget<'a, EntrySummary, TraversalProgress<'a>> for TraversalTarget<'_> {
5884 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5885 self.cmp_progress(cursor_location)
5886 }
5887}
5888
5889impl<'a> SeekTarget<'a, PathSummary<sum_tree::NoSummary>, TraversalProgress<'a>>
5890 for TraversalTarget<'_>
5891{
5892 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5893 self.cmp_progress(cursor_location)
5894 }
5895}
5896
5897pub struct ChildEntriesOptions {
5898 pub include_files: bool,
5899 pub include_dirs: bool,
5900 pub include_ignored: bool,
5901}
5902
5903pub struct ChildEntriesIter<'a> {
5904 parent_path: &'a RelPath,
5905 traversal: Traversal<'a>,
5906}
5907
5908impl<'a> Iterator for ChildEntriesIter<'a> {
5909 type Item = &'a Entry;
5910
5911 fn next(&mut self) -> Option<Self::Item> {
5912 if let Some(item) = self.traversal.entry()
5913 && item.path.starts_with(self.parent_path)
5914 {
5915 self.traversal.advance_to_sibling();
5916 return Some(item);
5917 }
5918 None
5919 }
5920}
5921
5922impl<'a> From<&'a Entry> for proto::Entry {
5923 fn from(entry: &'a Entry) -> Self {
5924 Self {
5925 id: entry.id.to_proto(),
5926 is_dir: entry.is_dir(),
5927 path: entry.path.as_ref().to_proto(),
5928 inode: entry.inode,
5929 mtime: entry.mtime.map(|time| time.into()),
5930 is_ignored: entry.is_ignored,
5931 is_hidden: entry.is_hidden,
5932 is_external: entry.is_external,
5933 is_fifo: entry.is_fifo,
5934 size: Some(entry.size),
5935 canonical_path: entry
5936 .canonical_path
5937 .as_ref()
5938 .map(|path| path.to_string_lossy().into_owned()),
5939 }
5940 }
5941}
5942
5943impl TryFrom<(&CharBag, &PathMatcher, proto::Entry)> for Entry {
5944 type Error = anyhow::Error;
5945
5946 fn try_from(
5947 (root_char_bag, always_included, entry): (&CharBag, &PathMatcher, proto::Entry),
5948 ) -> Result<Self> {
5949 let kind = if entry.is_dir {
5950 EntryKind::Dir
5951 } else {
5952 EntryKind::File
5953 };
5954
5955 let path =
5956 RelPath::from_proto(&entry.path).context("invalid relative path in proto message")?;
5957 let char_bag = char_bag_for_path(*root_char_bag, &path);
5958 let is_always_included = always_included.is_match(&path);
5959 Ok(Entry {
5960 id: ProjectEntryId::from_proto(entry.id),
5961 kind,
5962 path,
5963 inode: entry.inode,
5964 mtime: entry.mtime.map(|time| time.into()),
5965 size: entry.size.unwrap_or(0),
5966 canonical_path: entry
5967 .canonical_path
5968 .map(|path_string| Arc::from(PathBuf::from(path_string))),
5969 is_ignored: entry.is_ignored,
5970 is_hidden: entry.is_hidden,
5971 is_always_included,
5972 is_external: entry.is_external,
5973 is_private: false,
5974 char_bag,
5975 is_fifo: entry.is_fifo,
5976 })
5977 }
5978}
5979
5980#[derive(Clone, Copy, Debug, Default, Hash, PartialEq, Eq, PartialOrd, Ord)]
5981pub struct ProjectEntryId(usize);
5982
5983impl ProjectEntryId {
5984 pub const MAX: Self = Self(usize::MAX);
5985 pub const MIN: Self = Self(usize::MIN);
5986
5987 pub fn new(counter: &AtomicUsize) -> Self {
5988 Self(counter.fetch_add(1, SeqCst))
5989 }
5990
5991 pub fn from_proto(id: u64) -> Self {
5992 Self(id as usize)
5993 }
5994
5995 pub fn to_proto(self) -> u64 {
5996 self.0 as u64
5997 }
5998
5999 pub fn from_usize(id: usize) -> Self {
6000 ProjectEntryId(id)
6001 }
6002
6003 pub fn to_usize(self) -> usize {
6004 self.0
6005 }
6006}
6007
6008#[cfg(feature = "test-support")]
6009impl CreatedEntry {
6010 pub fn into_included(self) -> Option<Entry> {
6011 match self {
6012 CreatedEntry::Included(entry) => Some(entry),
6013 CreatedEntry::Excluded { .. } => None,
6014 }
6015 }
6016}
6017
6018fn parse_gitfile(content: &str) -> anyhow::Result<&Path> {
6019 let path = content
6020 .strip_prefix("gitdir:")
6021 .with_context(|| format!("parsing gitfile content {content:?}"))?;
6022 Ok(Path::new(path.trim()))
6023}
6024
6025async fn discover_git_paths(dot_git_abs_path: &Arc<Path>, fs: &dyn Fs) -> (Arc<Path>, Arc<Path>) {
6026 let mut repository_dir_abs_path = dot_git_abs_path.clone();
6027 let mut common_dir_abs_path = dot_git_abs_path.clone();
6028
6029 if let Some(path) = fs
6030 .load(dot_git_abs_path)
6031 .await
6032 .ok()
6033 .as_ref()
6034 .and_then(|contents| parse_gitfile(contents).log_err())
6035 {
6036 let path = dot_git_abs_path
6037 .parent()
6038 .unwrap_or(Path::new(""))
6039 .join(path);
6040 if let Some(path) = fs.canonicalize(&path).await.log_err() {
6041 repository_dir_abs_path = Path::new(&path).into();
6042 common_dir_abs_path = repository_dir_abs_path.clone();
6043
6044 if let Some(commondir_contents) = fs.load(&path.join("commondir")).await.ok()
6045 && let Some(commondir_path) = fs
6046 .canonicalize(&path.join(commondir_contents.trim()))
6047 .await
6048 .log_err()
6049 {
6050 common_dir_abs_path = commondir_path.as_path().into();
6051 }
6052 }
6053 };
6054 (repository_dir_abs_path, common_dir_abs_path)
6055}
6056
6057struct NullWatcher;
6058
6059impl fs::Watcher for NullWatcher {
6060 fn add(&self, _path: &Path) -> Result<()> {
6061 Ok(())
6062 }
6063
6064 fn remove(&self, _path: &Path) -> Result<()> {
6065 Ok(())
6066 }
6067}
6068
6069const FILE_ANALYSIS_BYTES: usize = 1024;
6070
6071async fn decode_file_text(
6072 fs: &dyn Fs,
6073 abs_path: &Path,
6074) -> Result<(String, &'static Encoding, bool)> {
6075 let mut file = fs
6076 .open_sync(&abs_path)
6077 .await
6078 .with_context(|| format!("opening file {abs_path:?}"))?;
6079
6080 // First, read the beginning of the file to determine its kind and encoding.
6081 // We do not want to load an entire large blob into memory only to discard it.
6082 let mut file_first_bytes = Vec::with_capacity(FILE_ANALYSIS_BYTES);
6083 let mut buf = [0u8; FILE_ANALYSIS_BYTES];
6084 let mut reached_eof = false;
6085 loop {
6086 if file_first_bytes.len() >= FILE_ANALYSIS_BYTES {
6087 break;
6088 }
6089 let n = file
6090 .read(&mut buf)
6091 .with_context(|| format!("reading bytes of the file {abs_path:?}"))?;
6092 if n == 0 {
6093 reached_eof = true;
6094 break;
6095 }
6096 file_first_bytes.extend_from_slice(&buf[..n]);
6097 }
6098 let (bom_encoding, byte_content) = decode_byte_header(&file_first_bytes);
6099 anyhow::ensure!(
6100 byte_content != ByteContent::Binary,
6101 "Binary files are not supported"
6102 );
6103
6104 // If the file is eligible for opening, read the rest of the file.
6105 let mut content = file_first_bytes;
6106 if !reached_eof {
6107 let mut buf = [0u8; 8 * 1024];
6108 loop {
6109 let n = file
6110 .read(&mut buf)
6111 .with_context(|| format!("reading remaining bytes of the file {abs_path:?}"))?;
6112 if n == 0 {
6113 break;
6114 }
6115 content.extend_from_slice(&buf[..n]);
6116 }
6117 }
6118 decode_byte_full(content, bom_encoding, byte_content)
6119}
6120
6121fn decode_byte_header(prefix: &[u8]) -> (Option<&'static Encoding>, ByteContent) {
6122 if let Some((encoding, _bom_len)) = Encoding::for_bom(prefix) {
6123 return (Some(encoding), ByteContent::Unknown);
6124 }
6125 (None, analyze_byte_content(prefix))
6126}
6127
6128fn decode_byte_full(
6129 bytes: Vec<u8>,
6130 bom_encoding: Option<&'static Encoding>,
6131 byte_content: ByteContent,
6132) -> Result<(String, &'static Encoding, bool)> {
6133 if let Some(encoding) = bom_encoding {
6134 let (cow, _) = encoding.decode_with_bom_removal(&bytes);
6135 return Ok((cow.into_owned(), encoding, true));
6136 }
6137
6138 match byte_content {
6139 ByteContent::Utf16Le => {
6140 let encoding = encoding_rs::UTF_16LE;
6141 let (cow, _, _) = encoding.decode(&bytes);
6142 return Ok((cow.into_owned(), encoding, false));
6143 }
6144 ByteContent::Utf16Be => {
6145 let encoding = encoding_rs::UTF_16BE;
6146 let (cow, _, _) = encoding.decode(&bytes);
6147 return Ok((cow.into_owned(), encoding, false));
6148 }
6149 ByteContent::Binary => {
6150 anyhow::bail!("Binary files are not supported");
6151 }
6152 ByteContent::Unknown => {}
6153 }
6154
6155 fn detect_encoding(bytes: Vec<u8>) -> (String, &'static Encoding) {
6156 let mut detector = EncodingDetector::new();
6157 detector.feed(&bytes, true);
6158
6159 let encoding = detector.guess(None, true); // Use None for TLD hint to ensure neutral detection logic.
6160
6161 let (cow, _, _) = encoding.decode(&bytes);
6162 (cow.into_owned(), encoding)
6163 }
6164
6165 match String::from_utf8(bytes) {
6166 Ok(text) => {
6167 // ISO-2022-JP (and other ISO-2022 variants) consists entirely of 7-bit ASCII bytes,
6168 // so it is valid UTF-8. However, it contains escape sequences starting with '\x1b'.
6169 // If we find an escape character, we double-check the encoding to prevent
6170 // displaying raw escape sequences instead of the correct characters.
6171 if text.contains('\x1b') {
6172 let (s, enc) = detect_encoding(text.into_bytes());
6173 Ok((s, enc, false))
6174 } else {
6175 Ok((text, encoding_rs::UTF_8, false))
6176 }
6177 }
6178 Err(e) => {
6179 let (s, enc) = detect_encoding(e.into_bytes());
6180 Ok((s, enc, false))
6181 }
6182 }
6183}
6184
6185#[derive(Debug, PartialEq)]
6186enum ByteContent {
6187 Utf16Le,
6188 Utf16Be,
6189 Binary,
6190 Unknown,
6191}
6192
6193// Heuristic check using null byte distribution plus a generic text-likeness
6194// heuristic. This prefers UTF-16 when many bytes are NUL and otherwise
6195// distinguishes between text-like and binary-like content.
6196fn analyze_byte_content(bytes: &[u8]) -> ByteContent {
6197 if bytes.len() < 2 {
6198 return ByteContent::Unknown;
6199 }
6200
6201 if is_known_binary_header(bytes) {
6202 return ByteContent::Binary;
6203 }
6204
6205 let limit = bytes.len().min(FILE_ANALYSIS_BYTES);
6206 let mut even_null_count = 0usize;
6207 let mut odd_null_count = 0usize;
6208 let mut non_text_like_count = 0usize;
6209
6210 for (i, &byte) in bytes[..limit].iter().enumerate() {
6211 if byte == 0 {
6212 if i % 2 == 0 {
6213 even_null_count += 1;
6214 } else {
6215 odd_null_count += 1;
6216 }
6217 non_text_like_count += 1;
6218 continue;
6219 }
6220
6221 let is_text_like = match byte {
6222 b'\t' | b'\n' | b'\r' | 0x0C => true,
6223 0x20..=0x7E => true,
6224 // Treat bytes that are likely part of UTF-8 or single-byte encodings as text-like.
6225 0x80..=0xBF | 0xC2..=0xF4 => true,
6226 _ => false,
6227 };
6228
6229 if !is_text_like {
6230 non_text_like_count += 1;
6231 }
6232 }
6233
6234 let total_null_count = even_null_count + odd_null_count;
6235
6236 // If there are no NUL bytes at all, this is overwhelmingly likely to be text.
6237 if total_null_count == 0 {
6238 return ByteContent::Unknown;
6239 }
6240
6241 let has_significant_nulls = total_null_count >= limit / 16;
6242 let nulls_skew_to_even = even_null_count > odd_null_count * 4;
6243 let nulls_skew_to_odd = odd_null_count > even_null_count * 4;
6244
6245 if has_significant_nulls {
6246 let sample = &bytes[..limit];
6247
6248 // UTF-16BE ASCII: [0x00, char] — nulls at even positions (high byte first)
6249 // UTF-16LE ASCII: [char, 0x00] — nulls at odd positions (low byte first)
6250
6251 if nulls_skew_to_even && is_plausible_utf16_text(sample, false) {
6252 return ByteContent::Utf16Be;
6253 }
6254
6255 if nulls_skew_to_odd && is_plausible_utf16_text(sample, true) {
6256 return ByteContent::Utf16Le;
6257 }
6258
6259 return ByteContent::Binary;
6260 }
6261
6262 if non_text_like_count * 100 < limit * 8 {
6263 ByteContent::Unknown
6264 } else {
6265 ByteContent::Binary
6266 }
6267}
6268
6269fn is_known_binary_header(bytes: &[u8]) -> bool {
6270 bytes.starts_with(b"%PDF-") // PDF
6271 || bytes.starts_with(b"PK\x03\x04") // ZIP local header
6272 || bytes.starts_with(b"PK\x05\x06") // ZIP end of central directory
6273 || bytes.starts_with(b"PK\x07\x08") // ZIP spanning/splitting
6274 || bytes.starts_with(b"\x89PNG\r\n\x1a\n") // PNG
6275 || bytes.starts_with(b"\xFF\xD8\xFF") // JPEG
6276 || bytes.starts_with(b"GIF87a") // GIF87a
6277 || bytes.starts_with(b"GIF89a") // GIF89a
6278 || bytes.starts_with(b"IWAD") // Doom IWAD archive
6279 || bytes.starts_with(b"PWAD") // Doom PWAD archive
6280 || bytes.starts_with(b"RIFF") // WAV, AVI, WebP
6281 || bytes.starts_with(b"OggS") // OGG (Vorbis, Opus, FLAC)
6282 || bytes.starts_with(b"fLaC") // FLAC
6283 || bytes.starts_with(b"ID3") // MP3 with ID3v2 tag
6284 || bytes.starts_with(b"\xFF\xFB") // MP3 frame sync (MPEG1 Layer3)
6285 || bytes.starts_with(b"\xFF\xFA") // MP3 frame sync (MPEG1 Layer3)
6286 || bytes.starts_with(b"\xFF\xF3") // MP3 frame sync (MPEG2 Layer3)
6287 || bytes.starts_with(b"\xFF\xF2") // MP3 frame sync (MPEG2 Layer3)
6288}
6289
6290// Null byte skew alone is not enough to identify UTF-16 -- binary formats with
6291// small 16-bit values (like PCM audio) produce the same pattern. Decode the
6292// bytes as UTF-16 and reject if too many code units land in control character
6293// ranges or form unpaired surrogates, which real text almost never contains.
6294fn is_plausible_utf16_text(bytes: &[u8], little_endian: bool) -> bool {
6295 let mut suspicious_count = 0usize;
6296 let mut total = 0usize;
6297
6298 let mut i = 0;
6299 while let Some(code_unit) = read_u16(bytes, i, little_endian) {
6300 total += 1;
6301
6302 match code_unit {
6303 0x0009 | 0x000A | 0x000C | 0x000D => {}
6304 // C0/C1 control characters and non-characters
6305 0x0000..=0x001F | 0x007F..=0x009F | 0xFFFE | 0xFFFF => suspicious_count += 1,
6306 0xD800..=0xDBFF => {
6307 let next_offset = i + 2;
6308 let has_low_surrogate = read_u16(bytes, next_offset, little_endian)
6309 .is_some_and(|next| (0xDC00..=0xDFFF).contains(&next));
6310 if has_low_surrogate {
6311 total += 1;
6312 i += 2;
6313 } else {
6314 suspicious_count += 1;
6315 }
6316 }
6317 // Lone low surrogate without a preceding high surrogate
6318 0xDC00..=0xDFFF => suspicious_count += 1,
6319 _ => {}
6320 }
6321
6322 i += 2;
6323 }
6324
6325 if total == 0 {
6326 return false;
6327 }
6328
6329 // Real UTF-16 text has near-zero control characters; binary data with
6330 // small 16-bit values typically exceeds 5%. 2% provides a safe margin.
6331 suspicious_count * 100 < total * 2
6332}
6333
6334fn read_u16(bytes: &[u8], offset: usize, little_endian: bool) -> Option<u16> {
6335 let pair = [*bytes.get(offset)?, *bytes.get(offset + 1)?];
6336 if little_endian {
6337 return Some(u16::from_le_bytes(pair));
6338 }
6339 Some(u16::from_be_bytes(pair))
6340}
6341
6342#[cfg(test)]
6343mod tests {
6344 use super::*;
6345
6346 /// reproduction of issue #50785
6347 fn build_pcm16_wav_bytes() -> Vec<u8> {
6348 let header: Vec<u8> = vec![
6349 /* RIFF header */
6350 0x52, 0x49, 0x46, 0x46, // "RIFF"
6351 0xc6, 0xcf, 0x00, 0x00, // file size: 8
6352 0x57, 0x41, 0x56, 0x45, // "WAVE"
6353 /* fmt chunk */
6354 0x66, 0x6d, 0x74, 0x20, // "fmt "
6355 0x10, 0x00, 0x00, 0x00, // chunk size: 16
6356 0x01, 0x00, // format: PCM (1)
6357 0x01, 0x00, // channels: 1 (mono)
6358 0x80, 0x3e, 0x00, 0x00, // sample rate: 16000
6359 0x00, 0x7d, 0x00, 0x00, // byte rate: 32000
6360 0x02, 0x00, // block align: 2
6361 0x10, 0x00, // bits per sample: 16
6362 /* LIST chunk */
6363 0x4c, 0x49, 0x53, 0x54, // "LIST"
6364 0x1a, 0x00, 0x00, 0x00, // chunk size: 26
6365 0x49, 0x4e, 0x46, 0x4f, // "INFO"
6366 0x49, 0x53, 0x46, 0x54, // "ISFT"
6367 0x0d, 0x00, 0x00, 0x00, // sub-chunk size: 13
6368 0x4c, 0x61, 0x76, 0x66, 0x36, 0x32, 0x2e, 0x33, // "Lavf62.3"
6369 0x2e, 0x31, 0x30, 0x30, 0x00, // ".100\0"
6370 /* padding byte for word alignment */
6371 0x00, // data chunk header
6372 0x64, 0x61, 0x74, 0x61, // "data"
6373 0x80, 0xcf, 0x00, 0x00, // chunk size
6374 ];
6375
6376 let mut bytes = header;
6377
6378 // fill remaining space up to `FILE_ANALYSIS_BYTES` with synthetic PCM
6379 let audio_bytes_needed = FILE_ANALYSIS_BYTES - bytes.len();
6380 for i in 0..(audio_bytes_needed / 2) {
6381 let sample = (i & 0xFF) as u8;
6382 bytes.push(sample); // low byte: varies
6383 bytes.push(0x00); // high byte: zero for small values
6384 }
6385
6386 bytes
6387 }
6388
6389 #[test]
6390 fn test_pcm16_wav_detected_as_binary() {
6391 let wav_bytes = build_pcm16_wav_bytes();
6392 assert_eq!(wav_bytes.len(), FILE_ANALYSIS_BYTES);
6393
6394 let result = analyze_byte_content(&wav_bytes);
6395 assert_eq!(
6396 result,
6397 ByteContent::Binary,
6398 "PCM 16-bit WAV should be detected as Binary via RIFF header"
6399 );
6400 }
6401
6402 #[test]
6403 fn test_le16_binary_not_misdetected_as_utf16le() {
6404 let mut bytes = b"FAKE".to_vec();
6405 while bytes.len() < FILE_ANALYSIS_BYTES {
6406 let sample = (bytes.len() & 0xFF) as u8;
6407 bytes.push(sample);
6408 bytes.push(0x00);
6409 }
6410 bytes.truncate(FILE_ANALYSIS_BYTES);
6411
6412 let result = analyze_byte_content(&bytes);
6413 assert_eq!(
6414 result,
6415 ByteContent::Binary,
6416 "LE 16-bit binary with control characters should be detected as Binary"
6417 );
6418 }
6419
6420 #[test]
6421 fn test_be16_binary_not_misdetected_as_utf16be() {
6422 let mut bytes = b"FAKE".to_vec();
6423 while bytes.len() < FILE_ANALYSIS_BYTES {
6424 bytes.push(0x00);
6425 let sample = (bytes.len() & 0xFF) as u8;
6426 bytes.push(sample);
6427 }
6428 bytes.truncate(FILE_ANALYSIS_BYTES);
6429
6430 let result = analyze_byte_content(&bytes);
6431 assert_eq!(
6432 result,
6433 ByteContent::Binary,
6434 "BE 16-bit binary with control characters should be detected as Binary"
6435 );
6436 }
6437
6438 #[test]
6439 fn test_utf16le_text_detected_as_utf16le() {
6440 let text = "Hello, world! This is a UTF-16 test string. ";
6441 let mut bytes = Vec::new();
6442 while bytes.len() < FILE_ANALYSIS_BYTES {
6443 bytes.extend(text.encode_utf16().flat_map(|u| u.to_le_bytes()));
6444 }
6445 bytes.truncate(FILE_ANALYSIS_BYTES);
6446
6447 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Le);
6448 }
6449
6450 #[test]
6451 fn test_utf16be_text_detected_as_utf16be() {
6452 let text = "Hello, world! This is a UTF-16 test string. ";
6453 let mut bytes = Vec::new();
6454 while bytes.len() < FILE_ANALYSIS_BYTES {
6455 bytes.extend(text.encode_utf16().flat_map(|u| u.to_be_bytes()));
6456 }
6457 bytes.truncate(FILE_ANALYSIS_BYTES);
6458
6459 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Be);
6460 }
6461
6462 #[test]
6463 fn test_known_binary_headers() {
6464 let cases: &[(&[u8], &str)] = &[
6465 (b"RIFF\x00\x00\x00\x00WAVE", "WAV"),
6466 (b"RIFF\x00\x00\x00\x00AVI ", "AVI"),
6467 (b"OggS\x00\x02", "OGG"),
6468 (b"fLaC\x00\x00", "FLAC"),
6469 (b"ID3\x03\x00", "MP3 ID3v2"),
6470 (b"\xFF\xFB\x90\x00", "MP3 MPEG1 Layer3"),
6471 (b"\xFF\xF3\x90\x00", "MP3 MPEG2 Layer3"),
6472 ];
6473
6474 for (header, label) in cases {
6475 let mut bytes = header.to_vec();
6476 bytes.resize(FILE_ANALYSIS_BYTES, 0x41); // pad with 'A'
6477 assert_eq!(
6478 analyze_byte_content(&bytes),
6479 ByteContent::Binary,
6480 "{label} should be detected as Binary"
6481 );
6482 }
6483 }
6484}