1mod ignore;
2mod worktree_settings;
3
4use ::ignore::gitignore::{Gitignore, GitignoreBuilder};
5use anyhow::{Context as _, Result, anyhow};
6use chardetng::EncodingDetector;
7use clock::ReplicaId;
8use collections::{HashMap, HashSet, VecDeque};
9use encoding_rs::Encoding;
10use fs::{
11 Fs, MTime, PathEvent, PathEventKind, RemoveOptions, Watcher, copy_recursive, read_dir_items,
12};
13use futures::{
14 FutureExt as _, Stream, StreamExt,
15 channel::{
16 mpsc::{self, UnboundedSender},
17 oneshot,
18 },
19 select_biased, stream,
20 task::Poll,
21};
22use fuzzy::CharBag;
23use git::{
24 COMMIT_MESSAGE, DOT_GIT, FSMONITOR_DAEMON, GITIGNORE, INDEX_LOCK, LFS_DIR, REPO_EXCLUDE,
25 status::GitSummary,
26};
27use gpui::{
28 App, AppContext as _, AsyncApp, BackgroundExecutor, Context, Entity, EventEmitter, Priority,
29 Task,
30};
31use ignore::IgnoreStack;
32use language::DiskState;
33
34use parking_lot::Mutex;
35use paths::{local_settings_folder_name, local_vscode_folder_name};
36use postage::{
37 barrier,
38 prelude::{Sink as _, Stream as _},
39 watch,
40};
41use rpc::{
42 AnyProtoClient,
43 proto::{self, split_worktree_update},
44};
45pub use settings::WorktreeId;
46use settings::{Settings, SettingsLocation, SettingsStore};
47use smallvec::{SmallVec, smallvec};
48use smol::channel::{self, Sender};
49use std::{
50 any::Any,
51 borrow::Borrow as _,
52 cmp::Ordering,
53 collections::hash_map,
54 convert::TryFrom,
55 ffi::OsStr,
56 fmt,
57 future::Future,
58 mem::{self},
59 ops::{Deref, DerefMut, Range},
60 path::{Path, PathBuf},
61 pin::Pin,
62 sync::{
63 Arc,
64 atomic::{AtomicUsize, Ordering::SeqCst},
65 },
66 time::{Duration, Instant},
67};
68use sum_tree::{Bias, Dimensions, Edit, KeyedItem, SeekTarget, SumTree, Summary, TreeMap, TreeSet};
69use text::{LineEnding, Rope};
70use util::{
71 ResultExt, debug_panic, maybe,
72 paths::{PathMatcher, PathStyle, SanitizedPath, home_dir},
73 rel_path::RelPath,
74};
75pub use worktree_settings::WorktreeSettings;
76
77use crate::ignore::IgnoreKind;
78
79pub const FS_WATCH_LATENCY: Duration = Duration::from_millis(100);
80
81/// A set of local or remote files that are being opened as part of a project.
82/// Responsible for tracking related FS (for local)/collab (for remote) events and corresponding updates.
83/// Stores git repositories data and the diagnostics for the file(s).
84///
85/// Has an absolute path, and may be set to be visible in Zed UI or not.
86/// May correspond to a directory or a single file.
87/// Possible examples:
88/// * a drag and dropped file — may be added as an invisible, "ephemeral" entry to the current worktree
89/// * a directory opened in Zed — may be added as a visible entry to the current worktree
90///
91/// Uses [`Entry`] to track the state of each file/directory, can look up absolute paths for entries.
92pub enum Worktree {
93 Local(LocalWorktree),
94 Remote(RemoteWorktree),
95}
96
97/// An entry, created in the worktree.
98#[derive(Debug)]
99pub enum CreatedEntry {
100 /// Got created and indexed by the worktree, receiving a corresponding entry.
101 Included(Entry),
102 /// Got created, but not indexed due to falling under exclusion filters.
103 Excluded { abs_path: PathBuf },
104}
105
106#[derive(Debug)]
107pub struct LoadedFile {
108 pub file: Arc<File>,
109 pub text: String,
110 pub encoding: &'static Encoding,
111 pub has_bom: bool,
112}
113
114pub struct LoadedBinaryFile {
115 pub file: Arc<File>,
116 pub content: Vec<u8>,
117}
118
119impl fmt::Debug for LoadedBinaryFile {
120 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
121 f.debug_struct("LoadedBinaryFile")
122 .field("file", &self.file)
123 .field("content_bytes", &self.content.len())
124 .finish()
125 }
126}
127
128pub struct LocalWorktree {
129 snapshot: LocalSnapshot,
130 scan_requests_tx: channel::Sender<ScanRequest>,
131 path_prefixes_to_scan_tx: channel::Sender<PathPrefixScanRequest>,
132 is_scanning: (watch::Sender<bool>, watch::Receiver<bool>),
133 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
134 _background_scanner_tasks: Vec<Task<()>>,
135 update_observer: Option<UpdateObservationState>,
136 fs: Arc<dyn Fs>,
137 fs_case_sensitive: bool,
138 visible: bool,
139 next_entry_id: Arc<AtomicUsize>,
140 settings: WorktreeSettings,
141 share_private_files: bool,
142 scanning_enabled: bool,
143}
144
145pub struct PathPrefixScanRequest {
146 path: Arc<RelPath>,
147 done: SmallVec<[barrier::Sender; 1]>,
148}
149
150struct ScanRequest {
151 relative_paths: Vec<Arc<RelPath>>,
152 done: SmallVec<[barrier::Sender; 1]>,
153}
154
155pub struct RemoteWorktree {
156 snapshot: Snapshot,
157 background_snapshot: Arc<Mutex<(Snapshot, Vec<proto::UpdateWorktree>)>>,
158 project_id: u64,
159 client: AnyProtoClient,
160 file_scan_inclusions: PathMatcher,
161 updates_tx: Option<UnboundedSender<proto::UpdateWorktree>>,
162 update_observer: Option<mpsc::UnboundedSender<proto::UpdateWorktree>>,
163 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
164 replica_id: ReplicaId,
165 visible: bool,
166 disconnected: bool,
167}
168
169#[derive(Clone)]
170pub struct Snapshot {
171 id: WorktreeId,
172 /// The absolute path of the worktree root.
173 abs_path: Arc<SanitizedPath>,
174 path_style: PathStyle,
175 root_name: Arc<RelPath>,
176 root_char_bag: CharBag,
177 entries_by_path: SumTree<Entry>,
178 entries_by_id: SumTree<PathEntry>,
179 always_included_entries: Vec<Arc<RelPath>>,
180
181 /// A number that increases every time the worktree begins scanning
182 /// a set of paths from the filesystem. This scanning could be caused
183 /// by some operation performed on the worktree, such as reading or
184 /// writing a file, or by an event reported by the filesystem.
185 scan_id: usize,
186
187 /// The latest scan id that has completed, and whose preceding scans
188 /// have all completed. The current `scan_id` could be more than one
189 /// greater than the `completed_scan_id` if operations are performed
190 /// on the worktree while it is processing a file-system event.
191 completed_scan_id: usize,
192}
193
194/// This path corresponds to the 'content path' of a repository in relation
195/// to Zed's project root.
196/// In the majority of the cases, this is the folder that contains the .git folder.
197/// But if a sub-folder of a git repository is opened, this corresponds to the
198/// project root and the .git folder is located in a parent directory.
199#[derive(Clone, Debug, Ord, PartialOrd, Eq, PartialEq, Hash)]
200pub enum WorkDirectory {
201 InProject {
202 relative_path: Arc<RelPath>,
203 },
204 AboveProject {
205 absolute_path: Arc<Path>,
206 location_in_repo: Arc<Path>,
207 },
208}
209
210impl WorkDirectory {
211 fn path_key(&self) -> PathKey {
212 match self {
213 WorkDirectory::InProject { relative_path } => PathKey(relative_path.clone()),
214 WorkDirectory::AboveProject { .. } => PathKey(RelPath::empty().into()),
215 }
216 }
217
218 /// Returns true if the given path is a child of the work directory.
219 ///
220 /// Note that the path may not be a member of this repository, if there
221 /// is a repository in a directory between these two paths
222 /// external .git folder in a parent folder of the project root.
223 #[track_caller]
224 pub fn directory_contains(&self, path: &RelPath) -> bool {
225 match self {
226 WorkDirectory::InProject { relative_path } => path.starts_with(relative_path),
227 WorkDirectory::AboveProject { .. } => true,
228 }
229 }
230}
231
232impl Default for WorkDirectory {
233 fn default() -> Self {
234 Self::InProject {
235 relative_path: Arc::from(RelPath::empty()),
236 }
237 }
238}
239
240#[derive(Clone)]
241pub struct LocalSnapshot {
242 snapshot: Snapshot,
243 global_gitignore: Option<Arc<Gitignore>>,
244 /// Exclude files for all git repositories in the worktree, indexed by their absolute path.
245 /// The boolean indicates whether the gitignore needs to be updated.
246 repo_exclude_by_work_dir_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
247 /// All of the gitignore files in the worktree, indexed by their absolute path.
248 /// The boolean indicates whether the gitignore needs to be updated.
249 ignores_by_parent_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
250 /// All of the git repositories in the worktree, indexed by the project entry
251 /// id of their parent directory.
252 git_repositories: TreeMap<ProjectEntryId, LocalRepositoryEntry>,
253 /// The file handle of the worktree root
254 /// (so we can find it after it's been moved)
255 root_file_handle: Option<Arc<dyn fs::FileHandle>>,
256}
257
258struct BackgroundScannerState {
259 snapshot: LocalSnapshot,
260 scanned_dirs: HashSet<ProjectEntryId>,
261 path_prefixes_to_scan: HashSet<Arc<RelPath>>,
262 paths_to_scan: HashSet<Arc<RelPath>>,
263 /// The ids of all of the entries that were removed from the snapshot
264 /// as part of the current update. These entry ids may be re-used
265 /// if the same inode is discovered at a new path, or if the given
266 /// path is re-created after being deleted.
267 removed_entries: HashMap<u64, Entry>,
268 changed_paths: Vec<Arc<RelPath>>,
269 prev_snapshot: Snapshot,
270 scanning_enabled: bool,
271}
272
273#[derive(Clone, Debug, Eq, PartialEq)]
274struct EventRoot {
275 path: Arc<RelPath>,
276 was_rescanned: bool,
277}
278
279#[derive(Debug, Clone)]
280struct LocalRepositoryEntry {
281 work_directory_id: ProjectEntryId,
282 work_directory: WorkDirectory,
283 work_directory_abs_path: Arc<Path>,
284 git_dir_scan_id: usize,
285 /// Absolute path to the original .git entry that caused us to create this repository.
286 ///
287 /// This is normally a directory, but may be a "gitfile" that points to a directory elsewhere
288 /// (whose path we then store in `repository_dir_abs_path`).
289 dot_git_abs_path: Arc<Path>,
290 /// Absolute path to the "commondir" for this repository.
291 ///
292 /// This is always a directory. For a normal repository, this is the same as dot_git_abs_path,
293 /// but in the case of a submodule or a worktree it is the path to the "parent" .git directory
294 /// from which the submodule/worktree was derived.
295 common_dir_abs_path: Arc<Path>,
296 /// Absolute path to the directory holding the repository's state.
297 ///
298 /// For a normal repository, this is a directory and coincides with `dot_git_abs_path` and
299 /// `common_dir_abs_path`. For a submodule or worktree, this is some subdirectory of the
300 /// commondir like `/project/.git/modules/foo`.
301 repository_dir_abs_path: Arc<Path>,
302}
303
304impl sum_tree::Item for LocalRepositoryEntry {
305 type Summary = PathSummary<sum_tree::NoSummary>;
306
307 fn summary(&self, _: <Self::Summary as Summary>::Context<'_>) -> Self::Summary {
308 PathSummary {
309 max_path: self.work_directory.path_key().0,
310 item_summary: sum_tree::NoSummary,
311 }
312 }
313}
314
315impl KeyedItem for LocalRepositoryEntry {
316 type Key = PathKey;
317
318 fn key(&self) -> Self::Key {
319 self.work_directory.path_key()
320 }
321}
322
323impl Deref for LocalRepositoryEntry {
324 type Target = WorkDirectory;
325
326 fn deref(&self) -> &Self::Target {
327 &self.work_directory
328 }
329}
330
331impl Deref for LocalSnapshot {
332 type Target = Snapshot;
333
334 fn deref(&self) -> &Self::Target {
335 &self.snapshot
336 }
337}
338
339impl DerefMut for LocalSnapshot {
340 fn deref_mut(&mut self) -> &mut Self::Target {
341 &mut self.snapshot
342 }
343}
344
345enum ScanState {
346 Started,
347 Updated {
348 snapshot: LocalSnapshot,
349 changes: UpdatedEntriesSet,
350 barrier: SmallVec<[barrier::Sender; 1]>,
351 scanning: bool,
352 },
353 RootUpdated {
354 new_path: Arc<SanitizedPath>,
355 },
356}
357
358struct UpdateObservationState {
359 snapshots_tx: mpsc::UnboundedSender<(LocalSnapshot, UpdatedEntriesSet)>,
360 resume_updates: watch::Sender<()>,
361 _maintain_remote_snapshot: Task<Option<()>>,
362}
363
364#[derive(Debug, Clone)]
365pub enum Event {
366 UpdatedEntries(UpdatedEntriesSet),
367 UpdatedGitRepositories(UpdatedGitRepositoriesSet),
368 DeletedEntry(ProjectEntryId),
369}
370
371impl EventEmitter<Event> for Worktree {}
372
373impl Worktree {
374 pub async fn local(
375 path: impl Into<Arc<Path>>,
376 visible: bool,
377 fs: Arc<dyn Fs>,
378 next_entry_id: Arc<AtomicUsize>,
379 scanning_enabled: bool,
380 worktree_id: WorktreeId,
381 cx: &mut AsyncApp,
382 ) -> Result<Entity<Self>> {
383 let abs_path = path.into();
384 let metadata = fs
385 .metadata(&abs_path)
386 .await
387 .context("failed to stat worktree path")?;
388
389 let fs_case_sensitive = fs.is_case_sensitive().await;
390
391 let root_file_handle = if metadata.as_ref().is_some() {
392 fs.open_handle(&abs_path)
393 .await
394 .with_context(|| {
395 format!(
396 "failed to open local worktree root at {}",
397 abs_path.display()
398 )
399 })
400 .log_err()
401 } else {
402 None
403 };
404
405 Ok(cx.new(move |cx: &mut Context<Worktree>| {
406 let mut snapshot = LocalSnapshot {
407 ignores_by_parent_abs_path: Default::default(),
408 global_gitignore: Default::default(),
409 repo_exclude_by_work_dir_abs_path: Default::default(),
410 git_repositories: Default::default(),
411 snapshot: Snapshot::new(
412 worktree_id,
413 abs_path
414 .file_name()
415 .and_then(|f| f.to_str())
416 .map_or(RelPath::empty().into(), |f| {
417 RelPath::unix(f).unwrap().into()
418 }),
419 abs_path.clone(),
420 PathStyle::local(),
421 ),
422 root_file_handle,
423 };
424
425 let worktree_id = snapshot.id();
426 let settings_location = Some(SettingsLocation {
427 worktree_id,
428 path: RelPath::empty(),
429 });
430
431 let settings = WorktreeSettings::get(settings_location, cx).clone();
432 cx.observe_global::<SettingsStore>(move |this, cx| {
433 if let Self::Local(this) = this {
434 let settings = WorktreeSettings::get(settings_location, cx).clone();
435 if this.settings != settings {
436 this.settings = settings;
437 this.restart_background_scanners(cx);
438 }
439 }
440 })
441 .detach();
442
443 let share_private_files = false;
444 if let Some(metadata) = metadata {
445 let mut entry = Entry::new(
446 RelPath::empty().into(),
447 &metadata,
448 ProjectEntryId::new(&next_entry_id),
449 snapshot.root_char_bag,
450 None,
451 );
452 if metadata.is_dir {
453 if !scanning_enabled {
454 entry.kind = EntryKind::UnloadedDir;
455 }
456 } else {
457 if let Some(file_name) = abs_path.file_name()
458 && let Some(file_name) = file_name.to_str()
459 && let Ok(path) = RelPath::unix(file_name)
460 {
461 entry.is_private = !share_private_files && settings.is_path_private(path);
462 entry.is_hidden = settings.is_path_hidden(path);
463 }
464 }
465 cx.foreground_executor()
466 .block_on(snapshot.insert_entry(entry, fs.as_ref()));
467 }
468
469 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
470 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
471 let mut worktree = LocalWorktree {
472 share_private_files,
473 next_entry_id,
474 snapshot,
475 is_scanning: watch::channel_with(true),
476 snapshot_subscriptions: Default::default(),
477 update_observer: None,
478 scan_requests_tx,
479 path_prefixes_to_scan_tx,
480 _background_scanner_tasks: Vec::new(),
481 fs,
482 fs_case_sensitive,
483 visible,
484 settings,
485 scanning_enabled,
486 };
487 worktree.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
488 Worktree::Local(worktree)
489 }))
490 }
491
492 pub fn remote(
493 project_id: u64,
494 replica_id: ReplicaId,
495 worktree: proto::WorktreeMetadata,
496 client: AnyProtoClient,
497 path_style: PathStyle,
498 cx: &mut App,
499 ) -> Entity<Self> {
500 cx.new(|cx: &mut Context<Self>| {
501 let snapshot = Snapshot::new(
502 WorktreeId::from_proto(worktree.id),
503 RelPath::from_proto(&worktree.root_name)
504 .unwrap_or_else(|_| RelPath::empty().into()),
505 Path::new(&worktree.abs_path).into(),
506 path_style,
507 );
508
509 let background_snapshot = Arc::new(Mutex::new((
510 snapshot.clone(),
511 Vec::<proto::UpdateWorktree>::new(),
512 )));
513 let (background_updates_tx, mut background_updates_rx) =
514 mpsc::unbounded::<proto::UpdateWorktree>();
515 let (mut snapshot_updated_tx, mut snapshot_updated_rx) = watch::channel();
516
517 let worktree_id = snapshot.id();
518 let settings_location = Some(SettingsLocation {
519 worktree_id,
520 path: RelPath::empty(),
521 });
522
523 let settings = WorktreeSettings::get(settings_location, cx).clone();
524 let worktree = RemoteWorktree {
525 client,
526 project_id,
527 replica_id,
528 snapshot,
529 file_scan_inclusions: settings.parent_dir_scan_inclusions.clone(),
530 background_snapshot: background_snapshot.clone(),
531 updates_tx: Some(background_updates_tx),
532 update_observer: None,
533 snapshot_subscriptions: Default::default(),
534 visible: worktree.visible,
535 disconnected: false,
536 };
537
538 // Apply updates to a separate snapshot in a background task, then
539 // send them to a foreground task which updates the model.
540 cx.background_spawn(async move {
541 while let Some(update) = background_updates_rx.next().await {
542 {
543 let mut lock = background_snapshot.lock();
544 lock.0.apply_remote_update(
545 update.clone(),
546 &settings.parent_dir_scan_inclusions,
547 );
548 lock.1.push(update);
549 }
550 snapshot_updated_tx.send(()).await.ok();
551 }
552 })
553 .detach();
554
555 // On the foreground task, update to the latest snapshot and notify
556 // any update observer of all updates that led to that snapshot.
557 cx.spawn(async move |this, cx| {
558 while (snapshot_updated_rx.recv().await).is_some() {
559 this.update(cx, |this, cx| {
560 let mut entries_changed = false;
561 let this = this.as_remote_mut().unwrap();
562 {
563 let mut lock = this.background_snapshot.lock();
564 this.snapshot = lock.0.clone();
565 for update in lock.1.drain(..) {
566 entries_changed |= !update.updated_entries.is_empty()
567 || !update.removed_entries.is_empty();
568 if let Some(tx) = &this.update_observer {
569 tx.unbounded_send(update).ok();
570 }
571 }
572 };
573
574 if entries_changed {
575 cx.emit(Event::UpdatedEntries(Arc::default()));
576 }
577 cx.notify();
578 while let Some((scan_id, _)) = this.snapshot_subscriptions.front() {
579 if this.observed_snapshot(*scan_id) {
580 let (_, tx) = this.snapshot_subscriptions.pop_front().unwrap();
581 let _ = tx.send(());
582 } else {
583 break;
584 }
585 }
586 })?;
587 }
588 anyhow::Ok(())
589 })
590 .detach();
591
592 Worktree::Remote(worktree)
593 })
594 }
595
596 pub fn as_local(&self) -> Option<&LocalWorktree> {
597 if let Worktree::Local(worktree) = self {
598 Some(worktree)
599 } else {
600 None
601 }
602 }
603
604 pub fn as_remote(&self) -> Option<&RemoteWorktree> {
605 if let Worktree::Remote(worktree) = self {
606 Some(worktree)
607 } else {
608 None
609 }
610 }
611
612 pub fn as_local_mut(&mut self) -> Option<&mut LocalWorktree> {
613 if let Worktree::Local(worktree) = self {
614 Some(worktree)
615 } else {
616 None
617 }
618 }
619
620 pub fn as_remote_mut(&mut self) -> Option<&mut RemoteWorktree> {
621 if let Worktree::Remote(worktree) = self {
622 Some(worktree)
623 } else {
624 None
625 }
626 }
627
628 pub fn is_local(&self) -> bool {
629 matches!(self, Worktree::Local(_))
630 }
631
632 pub fn is_remote(&self) -> bool {
633 !self.is_local()
634 }
635
636 pub fn settings_location(&self, _: &Context<Self>) -> SettingsLocation<'static> {
637 SettingsLocation {
638 worktree_id: self.id(),
639 path: RelPath::empty(),
640 }
641 }
642
643 pub fn snapshot(&self) -> Snapshot {
644 match self {
645 Worktree::Local(worktree) => worktree.snapshot.snapshot.clone(),
646 Worktree::Remote(worktree) => worktree.snapshot.clone(),
647 }
648 }
649
650 pub fn scan_id(&self) -> usize {
651 match self {
652 Worktree::Local(worktree) => worktree.snapshot.scan_id,
653 Worktree::Remote(worktree) => worktree.snapshot.scan_id,
654 }
655 }
656
657 pub fn metadata_proto(&self) -> proto::WorktreeMetadata {
658 proto::WorktreeMetadata {
659 id: self.id().to_proto(),
660 root_name: self.root_name().to_proto(),
661 visible: self.is_visible(),
662 abs_path: self.abs_path().to_string_lossy().into_owned(),
663 }
664 }
665
666 pub fn completed_scan_id(&self) -> usize {
667 match self {
668 Worktree::Local(worktree) => worktree.snapshot.completed_scan_id,
669 Worktree::Remote(worktree) => worktree.snapshot.completed_scan_id,
670 }
671 }
672
673 pub fn is_visible(&self) -> bool {
674 match self {
675 Worktree::Local(worktree) => worktree.visible,
676 Worktree::Remote(worktree) => worktree.visible,
677 }
678 }
679
680 pub fn replica_id(&self) -> ReplicaId {
681 match self {
682 Worktree::Local(_) => ReplicaId::LOCAL,
683 Worktree::Remote(worktree) => worktree.replica_id,
684 }
685 }
686
687 pub fn abs_path(&self) -> Arc<Path> {
688 match self {
689 Worktree::Local(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
690 Worktree::Remote(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
691 }
692 }
693
694 pub fn root_file(&self, cx: &Context<Self>) -> Option<Arc<File>> {
695 let entry = self.root_entry()?;
696 Some(File::for_entry(entry.clone(), cx.entity()))
697 }
698
699 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
700 where
701 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
702 Fut: 'static + Send + Future<Output = bool>,
703 {
704 match self {
705 Worktree::Local(this) => this.observe_updates(project_id, cx, callback),
706 Worktree::Remote(this) => this.observe_updates(project_id, cx, callback),
707 }
708 }
709
710 pub fn stop_observing_updates(&mut self) {
711 match self {
712 Worktree::Local(this) => {
713 this.update_observer.take();
714 }
715 Worktree::Remote(this) => {
716 this.update_observer.take();
717 }
718 }
719 }
720
721 pub fn wait_for_snapshot(
722 &mut self,
723 scan_id: usize,
724 ) -> impl Future<Output = Result<()>> + use<> {
725 match self {
726 Worktree::Local(this) => this.wait_for_snapshot(scan_id).boxed(),
727 Worktree::Remote(this) => this.wait_for_snapshot(scan_id).boxed(),
728 }
729 }
730
731 #[cfg(feature = "test-support")]
732 pub fn has_update_observer(&self) -> bool {
733 match self {
734 Worktree::Local(this) => this.update_observer.is_some(),
735 Worktree::Remote(this) => this.update_observer.is_some(),
736 }
737 }
738
739 pub fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
740 match self {
741 Worktree::Local(this) => this.load_file(path, cx),
742 Worktree::Remote(_) => {
743 Task::ready(Err(anyhow!("remote worktrees can't yet load files")))
744 }
745 }
746 }
747
748 pub fn load_binary_file(
749 &self,
750 path: &RelPath,
751 cx: &Context<Worktree>,
752 ) -> Task<Result<LoadedBinaryFile>> {
753 match self {
754 Worktree::Local(this) => this.load_binary_file(path, cx),
755 Worktree::Remote(_) => {
756 Task::ready(Err(anyhow!("remote worktrees can't yet load binary files")))
757 }
758 }
759 }
760
761 pub fn write_file(
762 &self,
763 path: Arc<RelPath>,
764 text: Rope,
765 line_ending: LineEnding,
766 encoding: &'static Encoding,
767 has_bom: bool,
768 cx: &Context<Worktree>,
769 ) -> Task<Result<Arc<File>>> {
770 match self {
771 Worktree::Local(this) => {
772 this.write_file(path, text, line_ending, encoding, has_bom, cx)
773 }
774 Worktree::Remote(_) => {
775 Task::ready(Err(anyhow!("remote worktree can't yet write files")))
776 }
777 }
778 }
779
780 pub fn create_entry(
781 &mut self,
782 path: Arc<RelPath>,
783 is_directory: bool,
784 content: Option<Vec<u8>>,
785 cx: &Context<Worktree>,
786 ) -> Task<Result<CreatedEntry>> {
787 let worktree_id = self.id();
788 match self {
789 Worktree::Local(this) => this.create_entry(path, is_directory, content, cx),
790 Worktree::Remote(this) => {
791 let project_id = this.project_id;
792 let request = this.client.request(proto::CreateProjectEntry {
793 worktree_id: worktree_id.to_proto(),
794 project_id,
795 path: path.as_ref().to_proto(),
796 content,
797 is_directory,
798 });
799 cx.spawn(async move |this, cx| {
800 let response = request.await?;
801 match response.entry {
802 Some(entry) => this
803 .update(cx, |worktree, cx| {
804 worktree.as_remote_mut().unwrap().insert_entry(
805 entry,
806 response.worktree_scan_id as usize,
807 cx,
808 )
809 })?
810 .await
811 .map(CreatedEntry::Included),
812 None => {
813 let abs_path =
814 this.read_with(cx, |worktree, _| worktree.absolutize(&path))?;
815 Ok(CreatedEntry::Excluded { abs_path })
816 }
817 }
818 })
819 }
820 }
821 }
822
823 pub fn delete_entry(
824 &mut self,
825 entry_id: ProjectEntryId,
826 trash: bool,
827 cx: &mut Context<Worktree>,
828 ) -> Option<Task<Result<()>>> {
829 let task = match self {
830 Worktree::Local(this) => this.delete_entry(entry_id, trash, cx),
831 Worktree::Remote(this) => this.delete_entry(entry_id, trash, cx),
832 }?;
833
834 let entry = match &*self {
835 Worktree::Local(this) => this.entry_for_id(entry_id),
836 Worktree::Remote(this) => this.entry_for_id(entry_id),
837 }?;
838
839 let mut ids = vec![entry_id];
840 let path = &*entry.path;
841
842 self.get_children_ids_recursive(path, &mut ids);
843
844 for id in ids {
845 cx.emit(Event::DeletedEntry(id));
846 }
847 Some(task)
848 }
849
850 fn get_children_ids_recursive(&self, path: &RelPath, ids: &mut Vec<ProjectEntryId>) {
851 let children_iter = self.child_entries(path);
852 for child in children_iter {
853 ids.push(child.id);
854 self.get_children_ids_recursive(&child.path, ids);
855 }
856 }
857
858 // pub fn rename_entry(
859 // &mut self,
860 // entry_id: ProjectEntryId,
861 // new_path: Arc<RelPath>,
862 // cx: &Context<Self>,
863 // ) -> Task<Result<CreatedEntry>> {
864 // match self {
865 // Worktree::Local(this) => this.rename_entry(entry_id, new_path, cx),
866 // Worktree::Remote(this) => this.rename_entry(entry_id, new_path, cx),
867 // }
868 // }
869
870 pub fn copy_external_entries(
871 &mut self,
872 target_directory: Arc<RelPath>,
873 paths: Vec<Arc<Path>>,
874 fs: Arc<dyn Fs>,
875 cx: &Context<Worktree>,
876 ) -> Task<Result<Vec<ProjectEntryId>>> {
877 match self {
878 Worktree::Local(this) => this.copy_external_entries(target_directory, paths, cx),
879 Worktree::Remote(this) => this.copy_external_entries(target_directory, paths, fs, cx),
880 }
881 }
882
883 pub fn expand_entry(
884 &mut self,
885 entry_id: ProjectEntryId,
886 cx: &Context<Worktree>,
887 ) -> Option<Task<Result<()>>> {
888 match self {
889 Worktree::Local(this) => this.expand_entry(entry_id, cx),
890 Worktree::Remote(this) => {
891 let response = this.client.request(proto::ExpandProjectEntry {
892 project_id: this.project_id,
893 entry_id: entry_id.to_proto(),
894 });
895 Some(cx.spawn(async move |this, cx| {
896 let response = response.await?;
897 this.update(cx, |this, _| {
898 this.as_remote_mut()
899 .unwrap()
900 .wait_for_snapshot(response.worktree_scan_id as usize)
901 })?
902 .await?;
903 Ok(())
904 }))
905 }
906 }
907 }
908
909 pub fn expand_all_for_entry(
910 &mut self,
911 entry_id: ProjectEntryId,
912 cx: &Context<Worktree>,
913 ) -> Option<Task<Result<()>>> {
914 match self {
915 Worktree::Local(this) => this.expand_all_for_entry(entry_id, cx),
916 Worktree::Remote(this) => {
917 let response = this.client.request(proto::ExpandAllForProjectEntry {
918 project_id: this.project_id,
919 entry_id: entry_id.to_proto(),
920 });
921 Some(cx.spawn(async move |this, cx| {
922 let response = response.await?;
923 this.update(cx, |this, _| {
924 this.as_remote_mut()
925 .unwrap()
926 .wait_for_snapshot(response.worktree_scan_id as usize)
927 })?
928 .await?;
929 Ok(())
930 }))
931 }
932 }
933 }
934
935 pub async fn handle_create_entry(
936 this: Entity<Self>,
937 request: proto::CreateProjectEntry,
938 mut cx: AsyncApp,
939 ) -> Result<proto::ProjectEntryResponse> {
940 let (scan_id, entry) = this.update(&mut cx, |this, cx| {
941 anyhow::Ok((
942 this.scan_id(),
943 this.create_entry(
944 RelPath::from_proto(&request.path).with_context(|| {
945 format!("received invalid relative path {:?}", request.path)
946 })?,
947 request.is_directory,
948 request.content,
949 cx,
950 ),
951 ))
952 })?;
953 Ok(proto::ProjectEntryResponse {
954 entry: match &entry.await? {
955 CreatedEntry::Included(entry) => Some(entry.into()),
956 CreatedEntry::Excluded { .. } => None,
957 },
958 worktree_scan_id: scan_id as u64,
959 })
960 }
961
962 pub async fn handle_delete_entry(
963 this: Entity<Self>,
964 request: proto::DeleteProjectEntry,
965 mut cx: AsyncApp,
966 ) -> Result<proto::ProjectEntryResponse> {
967 let (scan_id, task) = this.update(&mut cx, |this, cx| {
968 (
969 this.scan_id(),
970 this.delete_entry(
971 ProjectEntryId::from_proto(request.entry_id),
972 request.use_trash,
973 cx,
974 ),
975 )
976 });
977 task.ok_or_else(|| anyhow::anyhow!("invalid entry"))?
978 .await?;
979 Ok(proto::ProjectEntryResponse {
980 entry: None,
981 worktree_scan_id: scan_id as u64,
982 })
983 }
984
985 pub async fn handle_expand_entry(
986 this: Entity<Self>,
987 request: proto::ExpandProjectEntry,
988 mut cx: AsyncApp,
989 ) -> Result<proto::ExpandProjectEntryResponse> {
990 let task = this.update(&mut cx, |this, cx| {
991 this.expand_entry(ProjectEntryId::from_proto(request.entry_id), cx)
992 });
993 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
994 .await?;
995 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
996 Ok(proto::ExpandProjectEntryResponse {
997 worktree_scan_id: scan_id as u64,
998 })
999 }
1000
1001 pub async fn handle_expand_all_for_entry(
1002 this: Entity<Self>,
1003 request: proto::ExpandAllForProjectEntry,
1004 mut cx: AsyncApp,
1005 ) -> Result<proto::ExpandAllForProjectEntryResponse> {
1006 let task = this.update(&mut cx, |this, cx| {
1007 this.expand_all_for_entry(ProjectEntryId::from_proto(request.entry_id), cx)
1008 });
1009 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
1010 .await?;
1011 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
1012 Ok(proto::ExpandAllForProjectEntryResponse {
1013 worktree_scan_id: scan_id as u64,
1014 })
1015 }
1016
1017 pub fn is_single_file(&self) -> bool {
1018 self.root_dir().is_none()
1019 }
1020
1021 /// For visible worktrees, returns the path with the worktree name as the first component.
1022 /// Otherwise, returns an absolute path.
1023 pub fn full_path(&self, worktree_relative_path: &RelPath) -> PathBuf {
1024 if self.is_visible() {
1025 self.root_name()
1026 .join(worktree_relative_path)
1027 .display(self.path_style)
1028 .to_string()
1029 .into()
1030 } else {
1031 let full_path = self.abs_path();
1032 let mut full_path_string = if self.is_local()
1033 && let Ok(stripped) = full_path.strip_prefix(home_dir())
1034 {
1035 self.path_style
1036 .join("~", &*stripped.to_string_lossy())
1037 .unwrap()
1038 } else {
1039 full_path.to_string_lossy().into_owned()
1040 };
1041
1042 if worktree_relative_path.components().next().is_some() {
1043 full_path_string.push_str(self.path_style.primary_separator());
1044 full_path_string.push_str(&worktree_relative_path.display(self.path_style));
1045 }
1046
1047 full_path_string.into()
1048 }
1049 }
1050}
1051
1052impl LocalWorktree {
1053 pub fn fs(&self) -> &Arc<dyn Fs> {
1054 &self.fs
1055 }
1056
1057 pub fn is_path_private(&self, path: &RelPath) -> bool {
1058 !self.share_private_files && self.settings.is_path_private(path)
1059 }
1060
1061 pub fn fs_is_case_sensitive(&self) -> bool {
1062 self.fs_case_sensitive
1063 }
1064
1065 fn restart_background_scanners(&mut self, cx: &Context<Worktree>) {
1066 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
1067 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
1068 self.scan_requests_tx = scan_requests_tx;
1069 self.path_prefixes_to_scan_tx = path_prefixes_to_scan_tx;
1070
1071 self.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
1072 let always_included_entries = mem::take(&mut self.snapshot.always_included_entries);
1073 log::debug!(
1074 "refreshing entries for the following always included paths: {:?}",
1075 always_included_entries
1076 );
1077
1078 // Cleans up old always included entries to ensure they get updated properly. Otherwise,
1079 // nested always included entries may not get updated and will result in out-of-date info.
1080 self.refresh_entries_for_paths(always_included_entries);
1081 }
1082
1083 fn start_background_scanner(
1084 &mut self,
1085 scan_requests_rx: channel::Receiver<ScanRequest>,
1086 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
1087 cx: &Context<Worktree>,
1088 ) {
1089 let snapshot = self.snapshot();
1090 let share_private_files = self.share_private_files;
1091 let next_entry_id = self.next_entry_id.clone();
1092 let fs = self.fs.clone();
1093 let scanning_enabled = self.scanning_enabled;
1094 let settings = self.settings.clone();
1095 let (scan_states_tx, mut scan_states_rx) = mpsc::unbounded();
1096 let background_scanner = cx.background_spawn({
1097 let abs_path = snapshot.abs_path.as_path().to_path_buf();
1098 let background = cx.background_executor().clone();
1099 async move {
1100 let (events, watcher) = if scanning_enabled {
1101 fs.watch(&abs_path, FS_WATCH_LATENCY).await
1102 } else {
1103 (Box::pin(stream::pending()) as _, Arc::new(NullWatcher) as _)
1104 };
1105 let fs_case_sensitive = fs.is_case_sensitive().await;
1106
1107 let mut scanner = BackgroundScanner {
1108 fs,
1109 fs_case_sensitive,
1110 status_updates_tx: scan_states_tx,
1111 executor: background,
1112 scan_requests_rx,
1113 path_prefixes_to_scan_rx,
1114 next_entry_id,
1115 state: async_lock::Mutex::new(BackgroundScannerState {
1116 prev_snapshot: snapshot.snapshot.clone(),
1117 snapshot,
1118 scanned_dirs: Default::default(),
1119 scanning_enabled,
1120 path_prefixes_to_scan: Default::default(),
1121 paths_to_scan: Default::default(),
1122 removed_entries: Default::default(),
1123 changed_paths: Default::default(),
1124 }),
1125 phase: BackgroundScannerPhase::InitialScan,
1126 share_private_files,
1127 settings,
1128 watcher,
1129 };
1130
1131 scanner
1132 .run(Box::pin(events.map(|events| events.into_iter().collect())))
1133 .await;
1134 }
1135 });
1136 let scan_state_updater = cx.spawn(async move |this, cx| {
1137 while let Some((state, this)) = scan_states_rx.next().await.zip(this.upgrade()) {
1138 this.update(cx, |this, cx| {
1139 let this = this.as_local_mut().unwrap();
1140 match state {
1141 ScanState::Started => {
1142 *this.is_scanning.0.borrow_mut() = true;
1143 }
1144 ScanState::Updated {
1145 snapshot,
1146 changes,
1147 barrier,
1148 scanning,
1149 } => {
1150 *this.is_scanning.0.borrow_mut() = scanning;
1151 this.set_snapshot(snapshot, changes, cx);
1152 drop(barrier);
1153 }
1154 ScanState::RootUpdated { new_path } => {
1155 this.update_abs_path_and_refresh(new_path, cx);
1156 }
1157 }
1158 });
1159 }
1160 });
1161 self._background_scanner_tasks = vec![background_scanner, scan_state_updater];
1162 *self.is_scanning.0.borrow_mut() = true;
1163 }
1164
1165 fn set_snapshot(
1166 &mut self,
1167 mut new_snapshot: LocalSnapshot,
1168 entry_changes: UpdatedEntriesSet,
1169 cx: &mut Context<Worktree>,
1170 ) {
1171 let repo_changes = self.changed_repos(&self.snapshot, &mut new_snapshot);
1172 self.snapshot = new_snapshot;
1173
1174 if let Some(share) = self.update_observer.as_mut() {
1175 share
1176 .snapshots_tx
1177 .unbounded_send((self.snapshot.clone(), entry_changes.clone()))
1178 .ok();
1179 }
1180
1181 if !entry_changes.is_empty() {
1182 cx.emit(Event::UpdatedEntries(entry_changes));
1183 }
1184 if !repo_changes.is_empty() {
1185 cx.emit(Event::UpdatedGitRepositories(repo_changes));
1186 }
1187
1188 while let Some((scan_id, _)) = self.snapshot_subscriptions.front() {
1189 if self.snapshot.completed_scan_id >= *scan_id {
1190 let (_, tx) = self.snapshot_subscriptions.pop_front().unwrap();
1191 tx.send(()).ok();
1192 } else {
1193 break;
1194 }
1195 }
1196 }
1197
1198 fn changed_repos(
1199 &self,
1200 old_snapshot: &LocalSnapshot,
1201 new_snapshot: &mut LocalSnapshot,
1202 ) -> UpdatedGitRepositoriesSet {
1203 let mut changes = Vec::new();
1204 let mut old_repos = old_snapshot.git_repositories.iter().peekable();
1205 let new_repos = new_snapshot.git_repositories.clone();
1206 let mut new_repos = new_repos.iter().peekable();
1207
1208 loop {
1209 match (new_repos.peek().map(clone), old_repos.peek().map(clone)) {
1210 (Some((new_entry_id, new_repo)), Some((old_entry_id, old_repo))) => {
1211 match Ord::cmp(&new_entry_id, &old_entry_id) {
1212 Ordering::Less => {
1213 changes.push(UpdatedGitRepository {
1214 work_directory_id: new_entry_id,
1215 old_work_directory_abs_path: None,
1216 new_work_directory_abs_path: Some(
1217 new_repo.work_directory_abs_path.clone(),
1218 ),
1219 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1220 repository_dir_abs_path: Some(
1221 new_repo.repository_dir_abs_path.clone(),
1222 ),
1223 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1224 });
1225 new_repos.next();
1226 }
1227 Ordering::Equal => {
1228 if new_repo.git_dir_scan_id != old_repo.git_dir_scan_id
1229 || new_repo.work_directory_abs_path
1230 != old_repo.work_directory_abs_path
1231 {
1232 changes.push(UpdatedGitRepository {
1233 work_directory_id: new_entry_id,
1234 old_work_directory_abs_path: Some(
1235 old_repo.work_directory_abs_path.clone(),
1236 ),
1237 new_work_directory_abs_path: Some(
1238 new_repo.work_directory_abs_path.clone(),
1239 ),
1240 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1241 repository_dir_abs_path: Some(
1242 new_repo.repository_dir_abs_path.clone(),
1243 ),
1244 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1245 });
1246 }
1247 new_repos.next();
1248 old_repos.next();
1249 }
1250 Ordering::Greater => {
1251 changes.push(UpdatedGitRepository {
1252 work_directory_id: old_entry_id,
1253 old_work_directory_abs_path: Some(
1254 old_repo.work_directory_abs_path.clone(),
1255 ),
1256 new_work_directory_abs_path: None,
1257 dot_git_abs_path: None,
1258 repository_dir_abs_path: None,
1259 common_dir_abs_path: None,
1260 });
1261 old_repos.next();
1262 }
1263 }
1264 }
1265 (Some((entry_id, repo)), None) => {
1266 changes.push(UpdatedGitRepository {
1267 work_directory_id: entry_id,
1268 old_work_directory_abs_path: None,
1269 new_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1270 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1271 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1272 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1273 });
1274 new_repos.next();
1275 }
1276 (None, Some((entry_id, repo))) => {
1277 changes.push(UpdatedGitRepository {
1278 work_directory_id: entry_id,
1279 old_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1280 new_work_directory_abs_path: None,
1281 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1282 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1283 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1284 });
1285 old_repos.next();
1286 }
1287 (None, None) => break,
1288 }
1289 }
1290
1291 fn clone<T: Clone, U: Clone>(value: &(&T, &U)) -> (T, U) {
1292 (value.0.clone(), value.1.clone())
1293 }
1294
1295 changes.into()
1296 }
1297
1298 pub fn scan_complete(&self) -> impl Future<Output = ()> + use<> {
1299 let mut is_scanning_rx = self.is_scanning.1.clone();
1300 async move {
1301 let mut is_scanning = *is_scanning_rx.borrow();
1302 while is_scanning {
1303 if let Some(value) = is_scanning_rx.recv().await {
1304 is_scanning = value;
1305 } else {
1306 break;
1307 }
1308 }
1309 }
1310 }
1311
1312 pub fn wait_for_snapshot(
1313 &mut self,
1314 scan_id: usize,
1315 ) -> impl Future<Output = Result<()>> + use<> {
1316 let (tx, rx) = oneshot::channel();
1317 if self.snapshot.completed_scan_id >= scan_id {
1318 tx.send(()).ok();
1319 } else {
1320 match self
1321 .snapshot_subscriptions
1322 .binary_search_by_key(&scan_id, |probe| probe.0)
1323 {
1324 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
1325 }
1326 }
1327
1328 async move {
1329 rx.await?;
1330 Ok(())
1331 }
1332 }
1333
1334 pub fn snapshot(&self) -> LocalSnapshot {
1335 self.snapshot.clone()
1336 }
1337
1338 pub fn settings(&self) -> WorktreeSettings {
1339 self.settings.clone()
1340 }
1341
1342 fn load_binary_file(
1343 &self,
1344 path: &RelPath,
1345 cx: &Context<Worktree>,
1346 ) -> Task<Result<LoadedBinaryFile>> {
1347 let path = Arc::from(path);
1348 let abs_path = self.absolutize(&path);
1349 let fs = self.fs.clone();
1350 let entry = self.refresh_entry(path.clone(), None, cx);
1351 let is_private = self.is_path_private(&path);
1352
1353 let worktree = cx.weak_entity();
1354 cx.background_spawn(async move {
1355 let content = fs.load_bytes(&abs_path).await?;
1356
1357 let worktree = worktree.upgrade().context("worktree was dropped")?;
1358 let file = match entry.await? {
1359 Some(entry) => File::for_entry(entry, worktree),
1360 None => {
1361 let metadata = fs
1362 .metadata(&abs_path)
1363 .await
1364 .with_context(|| {
1365 format!("Loading metadata for excluded file {abs_path:?}")
1366 })?
1367 .with_context(|| {
1368 format!("Excluded file {abs_path:?} got removed during loading")
1369 })?;
1370 Arc::new(File {
1371 entry_id: None,
1372 worktree,
1373 path,
1374 disk_state: DiskState::Present {
1375 mtime: metadata.mtime,
1376 size: metadata.len,
1377 },
1378 is_local: true,
1379 is_private,
1380 })
1381 }
1382 };
1383
1384 Ok(LoadedBinaryFile { file, content })
1385 })
1386 }
1387
1388 #[ztracing::instrument(skip_all)]
1389 fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
1390 let path = Arc::from(path);
1391 let abs_path = self.absolutize(&path);
1392 let fs = self.fs.clone();
1393 let entry = self.refresh_entry(path.clone(), None, cx);
1394 let is_private = self.is_path_private(path.as_ref());
1395
1396 let this = cx.weak_entity();
1397 cx.background_spawn(async move {
1398 // WARN: Temporary workaround for #27283.
1399 // We are not efficient with our memory usage per file, and use in excess of 64GB for a 10GB file
1400 // Therefore, as a temporary workaround to prevent system freezes, we just bail before opening a file
1401 // if it is too large
1402 // 5GB seems to be more reasonable, peaking at ~16GB, while 6GB jumps up to >24GB which seems like a
1403 // reasonable limit
1404 {
1405 const FILE_SIZE_MAX: u64 = 6 * 1024 * 1024 * 1024; // 6GB
1406 if let Ok(Some(metadata)) = fs.metadata(&abs_path).await
1407 && metadata.len >= FILE_SIZE_MAX
1408 {
1409 anyhow::bail!("File is too large to load");
1410 }
1411 }
1412 let (text, encoding, has_bom) = decode_file_text(fs.as_ref(), &abs_path).await?;
1413
1414 let worktree = this.upgrade().context("worktree was dropped")?;
1415 let file = match entry.await? {
1416 Some(entry) => File::for_entry(entry, worktree),
1417 None => {
1418 let metadata = fs
1419 .metadata(&abs_path)
1420 .await
1421 .with_context(|| {
1422 format!("Loading metadata for excluded file {abs_path:?}")
1423 })?
1424 .with_context(|| {
1425 format!("Excluded file {abs_path:?} got removed during loading")
1426 })?;
1427 Arc::new(File {
1428 entry_id: None,
1429 worktree,
1430 path,
1431 disk_state: DiskState::Present {
1432 mtime: metadata.mtime,
1433 size: metadata.len,
1434 },
1435 is_local: true,
1436 is_private,
1437 })
1438 }
1439 };
1440
1441 Ok(LoadedFile {
1442 file,
1443 text,
1444 encoding,
1445 has_bom,
1446 })
1447 })
1448 }
1449
1450 /// Find the lowest path in the worktree's datastructures that is an ancestor
1451 fn lowest_ancestor(&self, path: &RelPath) -> Arc<RelPath> {
1452 let mut lowest_ancestor = None;
1453 for path in path.ancestors() {
1454 if self.entry_for_path(path).is_some() {
1455 lowest_ancestor = Some(path.into());
1456 break;
1457 }
1458 }
1459
1460 lowest_ancestor.unwrap_or_else(|| RelPath::empty().into())
1461 }
1462
1463 pub fn create_entry(
1464 &self,
1465 path: Arc<RelPath>,
1466 is_dir: bool,
1467 content: Option<Vec<u8>>,
1468 cx: &Context<Worktree>,
1469 ) -> Task<Result<CreatedEntry>> {
1470 let abs_path = self.absolutize(&path);
1471 let path_excluded = self.settings.is_path_excluded(&path);
1472 let fs = self.fs.clone();
1473 let task_abs_path = abs_path.clone();
1474 let write = cx.background_spawn(async move {
1475 if is_dir {
1476 fs.create_dir(&task_abs_path)
1477 .await
1478 .with_context(|| format!("creating directory {task_abs_path:?}"))
1479 } else {
1480 fs.write(&task_abs_path, content.as_deref().unwrap_or(&[]))
1481 .await
1482 .with_context(|| format!("creating file {task_abs_path:?}"))
1483 }
1484 });
1485
1486 let lowest_ancestor = self.lowest_ancestor(&path);
1487 cx.spawn(async move |this, cx| {
1488 write.await?;
1489 if path_excluded {
1490 return Ok(CreatedEntry::Excluded { abs_path });
1491 }
1492
1493 let (result, refreshes) = this.update(cx, |this, cx| {
1494 let mut refreshes = Vec::new();
1495 let refresh_paths = path.strip_prefix(&lowest_ancestor).unwrap();
1496 for refresh_path in refresh_paths.ancestors() {
1497 if refresh_path == RelPath::empty() {
1498 continue;
1499 }
1500 let refresh_full_path = lowest_ancestor.join(refresh_path);
1501
1502 refreshes.push(this.as_local_mut().unwrap().refresh_entry(
1503 refresh_full_path,
1504 None,
1505 cx,
1506 ));
1507 }
1508 (
1509 this.as_local_mut().unwrap().refresh_entry(path, None, cx),
1510 refreshes,
1511 )
1512 })?;
1513 for refresh in refreshes {
1514 refresh.await.log_err();
1515 }
1516
1517 Ok(result
1518 .await?
1519 .map(CreatedEntry::Included)
1520 .unwrap_or_else(|| CreatedEntry::Excluded { abs_path }))
1521 })
1522 }
1523
1524 pub fn write_file(
1525 &self,
1526 path: Arc<RelPath>,
1527 text: Rope,
1528 line_ending: LineEnding,
1529 encoding: &'static Encoding,
1530 has_bom: bool,
1531 cx: &Context<Worktree>,
1532 ) -> Task<Result<Arc<File>>> {
1533 let fs = self.fs.clone();
1534 let is_private = self.is_path_private(&path);
1535 let abs_path = self.absolutize(&path);
1536
1537 let write = cx.background_spawn({
1538 let fs = fs.clone();
1539 let abs_path = abs_path.clone();
1540 async move {
1541 // For UTF-8, use the optimized `fs.save` which writes Rope chunks directly to disk
1542 // without allocating a contiguous string.
1543 if encoding == encoding_rs::UTF_8 && !has_bom {
1544 return fs.save(&abs_path, &text, line_ending).await;
1545 }
1546
1547 // For legacy encodings (e.g. Shift-JIS), we fall back to converting the entire Rope
1548 // to a String/Bytes in memory before writing.
1549 //
1550 // Note: This is inefficient for very large files compared to the streaming approach above,
1551 // but supporting streaming writes for arbitrary encodings would require a significant
1552 // refactor of the `fs` crate to expose a Writer interface.
1553 let text_string = text.to_string();
1554 let normalized_text = match line_ending {
1555 LineEnding::Unix => text_string,
1556 LineEnding::Windows => text_string.replace('\n', "\r\n"),
1557 };
1558
1559 // Create the byte vector manually for UTF-16 encodings because encoding_rs encodes to UTF-8 by default (per WHATWG standards),
1560 // which is not what we want for saving files.
1561 let bytes = if encoding == encoding_rs::UTF_16BE {
1562 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1563 if has_bom {
1564 data.extend_from_slice(&[0xFE, 0xFF]); // BOM
1565 }
1566 let utf16be_bytes =
1567 normalized_text.encode_utf16().flat_map(|u| u.to_be_bytes());
1568 data.extend(utf16be_bytes);
1569 data.into()
1570 } else if encoding == encoding_rs::UTF_16LE {
1571 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1572 if has_bom {
1573 data.extend_from_slice(&[0xFF, 0xFE]); // BOM
1574 }
1575 let utf16le_bytes =
1576 normalized_text.encode_utf16().flat_map(|u| u.to_le_bytes());
1577 data.extend(utf16le_bytes);
1578 data.into()
1579 } else {
1580 // For other encodings (Shift-JIS, UTF-8 with BOM, etc.), delegate to encoding_rs.
1581 let bom_bytes = if has_bom {
1582 if encoding == encoding_rs::UTF_8 {
1583 vec![0xEF, 0xBB, 0xBF]
1584 } else {
1585 vec![]
1586 }
1587 } else {
1588 vec![]
1589 };
1590 let (cow, _, _) = encoding.encode(&normalized_text);
1591 if !bom_bytes.is_empty() {
1592 let mut bytes = bom_bytes;
1593 bytes.extend_from_slice(&cow);
1594 bytes.into()
1595 } else {
1596 cow
1597 }
1598 };
1599
1600 fs.write(&abs_path, &bytes).await
1601 }
1602 });
1603
1604 cx.spawn(async move |this, cx| {
1605 write.await?;
1606 let entry = this
1607 .update(cx, |this, cx| {
1608 this.as_local_mut()
1609 .unwrap()
1610 .refresh_entry(path.clone(), None, cx)
1611 })?
1612 .await?;
1613 let worktree = this.upgrade().context("worktree dropped")?;
1614 if let Some(entry) = entry {
1615 Ok(File::for_entry(entry, worktree))
1616 } else {
1617 let metadata = fs
1618 .metadata(&abs_path)
1619 .await
1620 .with_context(|| {
1621 format!("Fetching metadata after saving the excluded buffer {abs_path:?}")
1622 })?
1623 .with_context(|| {
1624 format!("Excluded buffer {path:?} got removed during saving")
1625 })?;
1626 Ok(Arc::new(File {
1627 worktree,
1628 path,
1629 disk_state: DiskState::Present {
1630 mtime: metadata.mtime,
1631 size: metadata.len,
1632 },
1633 entry_id: None,
1634 is_local: true,
1635 is_private,
1636 }))
1637 }
1638 })
1639 }
1640
1641 pub fn delete_entry(
1642 &self,
1643 entry_id: ProjectEntryId,
1644 trash: bool,
1645 cx: &Context<Worktree>,
1646 ) -> Option<Task<Result<()>>> {
1647 let entry = self.entry_for_id(entry_id)?.clone();
1648 let abs_path = self.absolutize(&entry.path);
1649 let fs = self.fs.clone();
1650
1651 let delete = cx.background_spawn(async move {
1652 if entry.is_file() {
1653 if trash {
1654 fs.trash_file(&abs_path, Default::default()).await?;
1655 } else {
1656 fs.remove_file(&abs_path, Default::default()).await?;
1657 }
1658 } else if trash {
1659 fs.trash_dir(
1660 &abs_path,
1661 RemoveOptions {
1662 recursive: true,
1663 ignore_if_not_exists: false,
1664 },
1665 )
1666 .await?;
1667 } else {
1668 fs.remove_dir(
1669 &abs_path,
1670 RemoveOptions {
1671 recursive: true,
1672 ignore_if_not_exists: false,
1673 },
1674 )
1675 .await?;
1676 }
1677 anyhow::Ok(entry.path)
1678 });
1679
1680 Some(cx.spawn(async move |this, cx| {
1681 let path = delete.await?;
1682 this.update(cx, |this, _| {
1683 this.as_local_mut()
1684 .unwrap()
1685 .refresh_entries_for_paths(vec![path])
1686 })?
1687 .recv()
1688 .await;
1689 Ok(())
1690 }))
1691 }
1692
1693 pub fn copy_external_entries(
1694 &self,
1695 target_directory: Arc<RelPath>,
1696 paths: Vec<Arc<Path>>,
1697 cx: &Context<Worktree>,
1698 ) -> Task<Result<Vec<ProjectEntryId>>> {
1699 let target_directory = self.absolutize(&target_directory);
1700 let worktree_path = self.abs_path().clone();
1701 let fs = self.fs.clone();
1702 let paths = paths
1703 .into_iter()
1704 .filter_map(|source| {
1705 let file_name = source.file_name()?;
1706 let mut target = target_directory.clone();
1707 target.push(file_name);
1708
1709 // Do not allow copying the same file to itself.
1710 if source.as_ref() != target.as_path() {
1711 Some((source, target))
1712 } else {
1713 None
1714 }
1715 })
1716 .collect::<Vec<_>>();
1717
1718 let paths_to_refresh = paths
1719 .iter()
1720 .filter_map(|(_, target)| {
1721 RelPath::new(
1722 target.strip_prefix(&worktree_path).ok()?,
1723 PathStyle::local(),
1724 )
1725 .ok()
1726 .map(|path| path.into_arc())
1727 })
1728 .collect::<Vec<_>>();
1729
1730 cx.spawn(async move |this, cx| {
1731 cx.background_spawn(async move {
1732 for (source, target) in paths {
1733 copy_recursive(
1734 fs.as_ref(),
1735 &source,
1736 &target,
1737 fs::CopyOptions {
1738 overwrite: true,
1739 ..Default::default()
1740 },
1741 )
1742 .await
1743 .with_context(|| {
1744 format!("Failed to copy file from {source:?} to {target:?}")
1745 })?;
1746 }
1747 anyhow::Ok(())
1748 })
1749 .await
1750 .log_err();
1751 let mut refresh = cx.read_entity(
1752 &this.upgrade().with_context(|| "Dropped worktree")?,
1753 |this, _| {
1754 anyhow::Ok::<postage::barrier::Receiver>(
1755 this.as_local()
1756 .with_context(|| "Worktree is not local")?
1757 .refresh_entries_for_paths(paths_to_refresh.clone()),
1758 )
1759 },
1760 )?;
1761
1762 cx.background_spawn(async move {
1763 refresh.next().await;
1764 anyhow::Ok(())
1765 })
1766 .await
1767 .log_err();
1768
1769 let this = this.upgrade().with_context(|| "Dropped worktree")?;
1770 Ok(cx.read_entity(&this, |this, _| {
1771 paths_to_refresh
1772 .iter()
1773 .filter_map(|path| Some(this.entry_for_path(path)?.id))
1774 .collect()
1775 }))
1776 })
1777 }
1778
1779 fn expand_entry(
1780 &self,
1781 entry_id: ProjectEntryId,
1782 cx: &Context<Worktree>,
1783 ) -> Option<Task<Result<()>>> {
1784 let path = self.entry_for_id(entry_id)?.path.clone();
1785 let mut refresh = self.refresh_entries_for_paths(vec![path]);
1786 Some(cx.background_spawn(async move {
1787 refresh.next().await;
1788 Ok(())
1789 }))
1790 }
1791
1792 fn expand_all_for_entry(
1793 &self,
1794 entry_id: ProjectEntryId,
1795 cx: &Context<Worktree>,
1796 ) -> Option<Task<Result<()>>> {
1797 let path = self.entry_for_id(entry_id).unwrap().path.clone();
1798 let mut rx = self.add_path_prefix_to_scan(path);
1799 Some(cx.background_spawn(async move {
1800 rx.next().await;
1801 Ok(())
1802 }))
1803 }
1804
1805 pub fn refresh_entries_for_paths(&self, paths: Vec<Arc<RelPath>>) -> barrier::Receiver {
1806 let (tx, rx) = barrier::channel();
1807 self.scan_requests_tx
1808 .try_send(ScanRequest {
1809 relative_paths: paths,
1810 done: smallvec![tx],
1811 })
1812 .ok();
1813 rx
1814 }
1815
1816 #[cfg(feature = "test-support")]
1817 pub fn manually_refresh_entries_for_paths(
1818 &self,
1819 paths: Vec<Arc<RelPath>>,
1820 ) -> barrier::Receiver {
1821 self.refresh_entries_for_paths(paths)
1822 }
1823
1824 pub fn add_path_prefix_to_scan(&self, path_prefix: Arc<RelPath>) -> barrier::Receiver {
1825 let (tx, rx) = barrier::channel();
1826 self.path_prefixes_to_scan_tx
1827 .try_send(PathPrefixScanRequest {
1828 path: path_prefix,
1829 done: smallvec![tx],
1830 })
1831 .ok();
1832 rx
1833 }
1834
1835 pub fn refresh_entry(
1836 &self,
1837 path: Arc<RelPath>,
1838 old_path: Option<Arc<RelPath>>,
1839 cx: &Context<Worktree>,
1840 ) -> Task<Result<Option<Entry>>> {
1841 if self.settings.is_path_excluded(&path) {
1842 return Task::ready(Ok(None));
1843 }
1844 let paths = if let Some(old_path) = old_path.as_ref() {
1845 vec![old_path.clone(), path.clone()]
1846 } else {
1847 vec![path.clone()]
1848 };
1849 let t0 = Instant::now();
1850 let mut refresh = self.refresh_entries_for_paths(paths);
1851 // todo(lw): Hot foreground spawn
1852 cx.spawn(async move |this, cx| {
1853 refresh.recv().await;
1854 log::trace!("refreshed entry {path:?} in {:?}", t0.elapsed());
1855 let new_entry = this.read_with(cx, |this, _| {
1856 this.entry_for_path(&path).cloned().with_context(|| {
1857 format!("Could not find entry in worktree for {path:?} after refresh")
1858 })
1859 })??;
1860 Ok(Some(new_entry))
1861 })
1862 }
1863
1864 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1865 where
1866 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1867 Fut: 'static + Send + Future<Output = bool>,
1868 {
1869 if let Some(observer) = self.update_observer.as_mut() {
1870 *observer.resume_updates.borrow_mut() = ();
1871 return;
1872 }
1873
1874 let (resume_updates_tx, mut resume_updates_rx) = watch::channel::<()>();
1875 let (snapshots_tx, mut snapshots_rx) =
1876 mpsc::unbounded::<(LocalSnapshot, UpdatedEntriesSet)>();
1877 snapshots_tx
1878 .unbounded_send((self.snapshot(), Arc::default()))
1879 .ok();
1880
1881 let worktree_id = self.id.to_proto();
1882 let _maintain_remote_snapshot = cx.background_spawn(async move {
1883 let mut is_first = true;
1884 while let Some((snapshot, entry_changes)) = snapshots_rx.next().await {
1885 let update = if is_first {
1886 is_first = false;
1887 snapshot.build_initial_update(project_id, worktree_id)
1888 } else {
1889 snapshot.build_update(project_id, worktree_id, entry_changes)
1890 };
1891
1892 for update in proto::split_worktree_update(update) {
1893 let _ = resume_updates_rx.try_recv();
1894 loop {
1895 let result = callback(update.clone());
1896 if result.await {
1897 break;
1898 } else {
1899 log::info!("waiting to resume updates");
1900 if resume_updates_rx.next().await.is_none() {
1901 return Some(());
1902 }
1903 }
1904 }
1905 }
1906 }
1907 Some(())
1908 });
1909
1910 self.update_observer = Some(UpdateObservationState {
1911 snapshots_tx,
1912 resume_updates: resume_updates_tx,
1913 _maintain_remote_snapshot,
1914 });
1915 }
1916
1917 pub fn share_private_files(&mut self, cx: &Context<Worktree>) {
1918 self.share_private_files = true;
1919 self.restart_background_scanners(cx);
1920 }
1921
1922 pub fn update_abs_path_and_refresh(
1923 &mut self,
1924 new_path: Arc<SanitizedPath>,
1925 cx: &Context<Worktree>,
1926 ) {
1927 self.snapshot.git_repositories = Default::default();
1928 self.snapshot.ignores_by_parent_abs_path = Default::default();
1929 let root_name = new_path
1930 .as_path()
1931 .file_name()
1932 .and_then(|f| f.to_str())
1933 .map_or(RelPath::empty().into(), |f| {
1934 RelPath::unix(f).unwrap().into()
1935 });
1936 self.snapshot.update_abs_path(new_path, root_name);
1937 self.restart_background_scanners(cx);
1938 }
1939 #[cfg(feature = "test-support")]
1940 pub fn repositories(&self) -> Vec<Arc<Path>> {
1941 self.git_repositories
1942 .values()
1943 .map(|entry| entry.work_directory_abs_path.clone())
1944 .collect::<Vec<_>>()
1945 }
1946}
1947
1948impl RemoteWorktree {
1949 pub fn project_id(&self) -> u64 {
1950 self.project_id
1951 }
1952
1953 pub fn client(&self) -> AnyProtoClient {
1954 self.client.clone()
1955 }
1956
1957 pub fn disconnected_from_host(&mut self) {
1958 self.updates_tx.take();
1959 self.snapshot_subscriptions.clear();
1960 self.disconnected = true;
1961 }
1962
1963 pub fn update_from_remote(&self, update: proto::UpdateWorktree) {
1964 if let Some(updates_tx) = &self.updates_tx {
1965 updates_tx
1966 .unbounded_send(update)
1967 .expect("consumer runs to completion");
1968 }
1969 }
1970
1971 fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1972 where
1973 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1974 Fut: 'static + Send + Future<Output = bool>,
1975 {
1976 let (tx, mut rx) = mpsc::unbounded();
1977 let initial_update = self
1978 .snapshot
1979 .build_initial_update(project_id, self.id().to_proto());
1980 self.update_observer = Some(tx);
1981 cx.spawn(async move |this, cx| {
1982 let mut update = initial_update;
1983 'outer: loop {
1984 // SSH projects use a special project ID of 0, and we need to
1985 // remap it to the correct one here.
1986 update.project_id = project_id;
1987
1988 for chunk in split_worktree_update(update) {
1989 if !callback(chunk).await {
1990 break 'outer;
1991 }
1992 }
1993
1994 if let Some(next_update) = rx.next().await {
1995 update = next_update;
1996 } else {
1997 break;
1998 }
1999 }
2000 this.update(cx, |this, _| {
2001 let this = this.as_remote_mut().unwrap();
2002 this.update_observer.take();
2003 })
2004 })
2005 .detach();
2006 }
2007
2008 fn observed_snapshot(&self, scan_id: usize) -> bool {
2009 self.completed_scan_id >= scan_id
2010 }
2011
2012 pub fn wait_for_snapshot(
2013 &mut self,
2014 scan_id: usize,
2015 ) -> impl Future<Output = Result<()>> + use<> {
2016 let (tx, rx) = oneshot::channel();
2017 if self.observed_snapshot(scan_id) {
2018 let _ = tx.send(());
2019 } else if self.disconnected {
2020 drop(tx);
2021 } else {
2022 match self
2023 .snapshot_subscriptions
2024 .binary_search_by_key(&scan_id, |probe| probe.0)
2025 {
2026 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
2027 }
2028 }
2029
2030 async move {
2031 rx.await?;
2032 Ok(())
2033 }
2034 }
2035
2036 pub fn insert_entry(
2037 &mut self,
2038 entry: proto::Entry,
2039 scan_id: usize,
2040 cx: &Context<Worktree>,
2041 ) -> Task<Result<Entry>> {
2042 let wait_for_snapshot = self.wait_for_snapshot(scan_id);
2043 cx.spawn(async move |this, cx| {
2044 wait_for_snapshot.await?;
2045 this.update(cx, |worktree, _| {
2046 let worktree = worktree.as_remote_mut().unwrap();
2047 let snapshot = &mut worktree.background_snapshot.lock().0;
2048 let entry = snapshot.insert_entry(entry, &worktree.file_scan_inclusions);
2049 worktree.snapshot = snapshot.clone();
2050 entry
2051 })?
2052 })
2053 }
2054
2055 fn delete_entry(
2056 &self,
2057 entry_id: ProjectEntryId,
2058 trash: bool,
2059 cx: &Context<Worktree>,
2060 ) -> Option<Task<Result<()>>> {
2061 let response = self.client.request(proto::DeleteProjectEntry {
2062 project_id: self.project_id,
2063 entry_id: entry_id.to_proto(),
2064 use_trash: trash,
2065 });
2066 Some(cx.spawn(async move |this, cx| {
2067 let response = response.await?;
2068 let scan_id = response.worktree_scan_id as usize;
2069
2070 this.update(cx, move |this, _| {
2071 this.as_remote_mut().unwrap().wait_for_snapshot(scan_id)
2072 })?
2073 .await?;
2074
2075 this.update(cx, |this, _| {
2076 let this = this.as_remote_mut().unwrap();
2077 let snapshot = &mut this.background_snapshot.lock().0;
2078 snapshot.delete_entry(entry_id);
2079 this.snapshot = snapshot.clone();
2080 })
2081 }))
2082 }
2083
2084 // fn rename_entry(
2085 // &self,
2086 // entry_id: ProjectEntryId,
2087 // new_path: impl Into<Arc<RelPath>>,
2088 // cx: &Context<Worktree>,
2089 // ) -> Task<Result<CreatedEntry>> {
2090 // let new_path: Arc<RelPath> = new_path.into();
2091 // let response = self.client.request(proto::RenameProjectEntry {
2092 // project_id: self.project_id,
2093 // entry_id: entry_id.to_proto(),
2094 // new_worktree_id: new_path.worktree_id,
2095 // new_path: new_path.as_ref().to_proto(),
2096 // });
2097 // cx.spawn(async move |this, cx| {
2098 // let response = response.await?;
2099 // match response.entry {
2100 // Some(entry) => this
2101 // .update(cx, |this, cx| {
2102 // this.as_remote_mut().unwrap().insert_entry(
2103 // entry,
2104 // response.worktree_scan_id as usize,
2105 // cx,
2106 // )
2107 // })?
2108 // .await
2109 // .map(CreatedEntry::Included),
2110 // None => {
2111 // let abs_path =
2112 // this.read_with(cx, |worktree, _| worktree.absolutize(&new_path))?;
2113 // Ok(CreatedEntry::Excluded { abs_path })
2114 // }
2115 // }
2116 // })
2117 // }
2118
2119 fn copy_external_entries(
2120 &self,
2121 target_directory: Arc<RelPath>,
2122 paths_to_copy: Vec<Arc<Path>>,
2123 local_fs: Arc<dyn Fs>,
2124 cx: &Context<Worktree>,
2125 ) -> Task<anyhow::Result<Vec<ProjectEntryId>>> {
2126 let client = self.client.clone();
2127 let worktree_id = self.id().to_proto();
2128 let project_id = self.project_id;
2129
2130 cx.background_spawn(async move {
2131 let mut requests = Vec::new();
2132 for root_path_to_copy in paths_to_copy {
2133 let Some(filename) = root_path_to_copy
2134 .file_name()
2135 .and_then(|name| name.to_str())
2136 .and_then(|filename| RelPath::unix(filename).ok())
2137 else {
2138 continue;
2139 };
2140 for (abs_path, is_directory) in
2141 read_dir_items(local_fs.as_ref(), &root_path_to_copy).await?
2142 {
2143 let Some(relative_path) = abs_path
2144 .strip_prefix(&root_path_to_copy)
2145 .map_err(|e| anyhow::Error::from(e))
2146 .and_then(|relative_path| RelPath::new(relative_path, PathStyle::local()))
2147 .log_err()
2148 else {
2149 continue;
2150 };
2151 let content = if is_directory {
2152 None
2153 } else {
2154 Some(local_fs.load_bytes(&abs_path).await?)
2155 };
2156
2157 let mut target_path = target_directory.join(filename);
2158 if relative_path.file_name().is_some() {
2159 target_path = target_path.join(&relative_path);
2160 }
2161
2162 requests.push(proto::CreateProjectEntry {
2163 project_id,
2164 worktree_id,
2165 path: target_path.to_proto(),
2166 is_directory,
2167 content,
2168 });
2169 }
2170 }
2171 requests.sort_unstable_by(|a, b| a.path.cmp(&b.path));
2172 requests.dedup();
2173
2174 let mut copied_entry_ids = Vec::new();
2175 for request in requests {
2176 let response = client.request(request).await?;
2177 copied_entry_ids.extend(response.entry.map(|e| ProjectEntryId::from_proto(e.id)));
2178 }
2179
2180 Ok(copied_entry_ids)
2181 })
2182 }
2183}
2184
2185impl Snapshot {
2186 pub fn new(
2187 id: WorktreeId,
2188 root_name: Arc<RelPath>,
2189 abs_path: Arc<Path>,
2190 path_style: PathStyle,
2191 ) -> Self {
2192 Snapshot {
2193 id,
2194 abs_path: SanitizedPath::from_arc(abs_path),
2195 path_style,
2196 root_char_bag: root_name
2197 .as_unix_str()
2198 .chars()
2199 .map(|c| c.to_ascii_lowercase())
2200 .collect(),
2201 root_name,
2202 always_included_entries: Default::default(),
2203 entries_by_path: Default::default(),
2204 entries_by_id: Default::default(),
2205 scan_id: 1,
2206 completed_scan_id: 0,
2207 }
2208 }
2209
2210 pub fn id(&self) -> WorktreeId {
2211 self.id
2212 }
2213
2214 // TODO:
2215 // Consider the following:
2216 //
2217 // ```rust
2218 // let abs_path: Arc<Path> = snapshot.abs_path(); // e.g. "C:\Users\user\Desktop\project"
2219 // let some_non_trimmed_path = Path::new("\\\\?\\C:\\Users\\user\\Desktop\\project\\main.rs");
2220 // // The caller perform some actions here:
2221 // some_non_trimmed_path.strip_prefix(abs_path); // This fails
2222 // some_non_trimmed_path.starts_with(abs_path); // This fails too
2223 // ```
2224 //
2225 // This is definitely a bug, but it's not clear if we should handle it here or not.
2226 pub fn abs_path(&self) -> &Arc<Path> {
2227 SanitizedPath::cast_arc_ref(&self.abs_path)
2228 }
2229
2230 fn build_initial_update(&self, project_id: u64, worktree_id: u64) -> proto::UpdateWorktree {
2231 let mut updated_entries = self
2232 .entries_by_path
2233 .iter()
2234 .map(proto::Entry::from)
2235 .collect::<Vec<_>>();
2236 updated_entries.sort_unstable_by_key(|e| e.id);
2237
2238 proto::UpdateWorktree {
2239 project_id,
2240 worktree_id,
2241 abs_path: self.abs_path().to_string_lossy().into_owned(),
2242 root_name: self.root_name().to_proto(),
2243 updated_entries,
2244 removed_entries: Vec::new(),
2245 scan_id: self.scan_id as u64,
2246 is_last_update: self.completed_scan_id == self.scan_id,
2247 // Sent in separate messages.
2248 updated_repositories: Vec::new(),
2249 removed_repositories: Vec::new(),
2250 }
2251 }
2252
2253 pub fn work_directory_abs_path(&self, work_directory: &WorkDirectory) -> PathBuf {
2254 match work_directory {
2255 WorkDirectory::InProject { relative_path } => self.absolutize(relative_path),
2256 WorkDirectory::AboveProject { absolute_path, .. } => absolute_path.as_ref().to_owned(),
2257 }
2258 }
2259
2260 pub fn absolutize(&self, path: &RelPath) -> PathBuf {
2261 if path.file_name().is_some() {
2262 let mut abs_path = self.abs_path.to_string();
2263 for component in path.components() {
2264 if !abs_path.ends_with(self.path_style.primary_separator()) {
2265 abs_path.push_str(self.path_style.primary_separator());
2266 }
2267 abs_path.push_str(component);
2268 }
2269 PathBuf::from(abs_path)
2270 } else {
2271 self.abs_path.as_path().to_path_buf()
2272 }
2273 }
2274
2275 pub fn contains_entry(&self, entry_id: ProjectEntryId) -> bool {
2276 self.entries_by_id.get(&entry_id, ()).is_some()
2277 }
2278
2279 fn insert_entry(
2280 &mut self,
2281 entry: proto::Entry,
2282 always_included_paths: &PathMatcher,
2283 ) -> Result<Entry> {
2284 let entry = Entry::try_from((&self.root_char_bag, always_included_paths, entry))?;
2285 let old_entry = self.entries_by_id.insert_or_replace(
2286 PathEntry {
2287 id: entry.id,
2288 path: entry.path.clone(),
2289 is_ignored: entry.is_ignored,
2290 scan_id: 0,
2291 },
2292 (),
2293 );
2294 if let Some(old_entry) = old_entry {
2295 self.entries_by_path.remove(&PathKey(old_entry.path), ());
2296 }
2297 self.entries_by_path.insert_or_replace(entry.clone(), ());
2298 Ok(entry)
2299 }
2300
2301 fn delete_entry(&mut self, entry_id: ProjectEntryId) -> Option<Arc<RelPath>> {
2302 let removed_entry = self.entries_by_id.remove(&entry_id, ())?;
2303 self.entries_by_path = {
2304 let mut cursor = self.entries_by_path.cursor::<TraversalProgress>(());
2305 let mut new_entries_by_path =
2306 cursor.slice(&TraversalTarget::path(&removed_entry.path), Bias::Left);
2307 while let Some(entry) = cursor.item() {
2308 if entry.path.starts_with(&removed_entry.path) {
2309 self.entries_by_id.remove(&entry.id, ());
2310 cursor.next();
2311 } else {
2312 break;
2313 }
2314 }
2315 new_entries_by_path.append(cursor.suffix(), ());
2316 new_entries_by_path
2317 };
2318
2319 Some(removed_entry.path)
2320 }
2321
2322 fn update_abs_path(&mut self, abs_path: Arc<SanitizedPath>, root_name: Arc<RelPath>) {
2323 self.abs_path = abs_path;
2324 if root_name != self.root_name {
2325 self.root_char_bag = root_name
2326 .as_unix_str()
2327 .chars()
2328 .map(|c| c.to_ascii_lowercase())
2329 .collect();
2330 self.root_name = root_name;
2331 }
2332 }
2333
2334 pub fn apply_remote_update(
2335 &mut self,
2336 update: proto::UpdateWorktree,
2337 always_included_paths: &PathMatcher,
2338 ) {
2339 log::debug!(
2340 "applying remote worktree update. {} entries updated, {} removed",
2341 update.updated_entries.len(),
2342 update.removed_entries.len()
2343 );
2344 if let Some(root_name) = RelPath::from_proto(&update.root_name).log_err() {
2345 self.update_abs_path(
2346 SanitizedPath::new_arc(&Path::new(&update.abs_path)),
2347 root_name,
2348 );
2349 }
2350
2351 let mut entries_by_path_edits = Vec::new();
2352 let mut entries_by_id_edits = Vec::new();
2353
2354 for entry_id in update.removed_entries {
2355 let entry_id = ProjectEntryId::from_proto(entry_id);
2356 entries_by_id_edits.push(Edit::Remove(entry_id));
2357 if let Some(entry) = self.entry_for_id(entry_id) {
2358 entries_by_path_edits.push(Edit::Remove(PathKey(entry.path.clone())));
2359 }
2360 }
2361
2362 for entry in update.updated_entries {
2363 let Some(entry) =
2364 Entry::try_from((&self.root_char_bag, always_included_paths, entry)).log_err()
2365 else {
2366 continue;
2367 };
2368 if let Some(PathEntry { path, .. }) = self.entries_by_id.get(&entry.id, ()) {
2369 entries_by_path_edits.push(Edit::Remove(PathKey(path.clone())));
2370 }
2371 if let Some(old_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2372 && old_entry.id != entry.id
2373 {
2374 entries_by_id_edits.push(Edit::Remove(old_entry.id));
2375 }
2376 entries_by_id_edits.push(Edit::Insert(PathEntry {
2377 id: entry.id,
2378 path: entry.path.clone(),
2379 is_ignored: entry.is_ignored,
2380 scan_id: 0,
2381 }));
2382 entries_by_path_edits.push(Edit::Insert(entry));
2383 }
2384
2385 self.entries_by_path.edit(entries_by_path_edits, ());
2386 self.entries_by_id.edit(entries_by_id_edits, ());
2387
2388 self.scan_id = update.scan_id as usize;
2389 if update.is_last_update {
2390 self.completed_scan_id = update.scan_id as usize;
2391 }
2392 }
2393
2394 pub fn entry_count(&self) -> usize {
2395 self.entries_by_path.summary().count
2396 }
2397
2398 pub fn visible_entry_count(&self) -> usize {
2399 self.entries_by_path.summary().non_ignored_count
2400 }
2401
2402 pub fn dir_count(&self) -> usize {
2403 let summary = self.entries_by_path.summary();
2404 summary.count - summary.file_count
2405 }
2406
2407 pub fn visible_dir_count(&self) -> usize {
2408 let summary = self.entries_by_path.summary();
2409 summary.non_ignored_count - summary.non_ignored_file_count
2410 }
2411
2412 pub fn file_count(&self) -> usize {
2413 self.entries_by_path.summary().file_count
2414 }
2415
2416 pub fn visible_file_count(&self) -> usize {
2417 self.entries_by_path.summary().non_ignored_file_count
2418 }
2419
2420 fn traverse_from_offset(
2421 &self,
2422 include_files: bool,
2423 include_dirs: bool,
2424 include_ignored: bool,
2425 start_offset: usize,
2426 ) -> Traversal<'_> {
2427 let mut cursor = self.entries_by_path.cursor(());
2428 cursor.seek(
2429 &TraversalTarget::Count {
2430 count: start_offset,
2431 include_files,
2432 include_dirs,
2433 include_ignored,
2434 },
2435 Bias::Right,
2436 );
2437 Traversal {
2438 snapshot: self,
2439 cursor,
2440 include_files,
2441 include_dirs,
2442 include_ignored,
2443 }
2444 }
2445
2446 pub fn traverse_from_path(
2447 &self,
2448 include_files: bool,
2449 include_dirs: bool,
2450 include_ignored: bool,
2451 path: &RelPath,
2452 ) -> Traversal<'_> {
2453 Traversal::new(self, include_files, include_dirs, include_ignored, path)
2454 }
2455
2456 pub fn files(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2457 self.traverse_from_offset(true, false, include_ignored, start)
2458 }
2459
2460 pub fn directories(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2461 self.traverse_from_offset(false, true, include_ignored, start)
2462 }
2463
2464 pub fn entries(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2465 self.traverse_from_offset(true, true, include_ignored, start)
2466 }
2467
2468 pub fn paths(&self) -> impl Iterator<Item = &RelPath> {
2469 self.entries_by_path
2470 .cursor::<()>(())
2471 .filter(move |entry| !entry.path.is_empty())
2472 .map(|entry| entry.path.as_ref())
2473 }
2474
2475 pub fn child_entries<'a>(&'a self, parent_path: &'a RelPath) -> ChildEntriesIter<'a> {
2476 let options = ChildEntriesOptions {
2477 include_files: true,
2478 include_dirs: true,
2479 include_ignored: true,
2480 };
2481 self.child_entries_with_options(parent_path, options)
2482 }
2483
2484 pub fn child_entries_with_options<'a>(
2485 &'a self,
2486 parent_path: &'a RelPath,
2487 options: ChildEntriesOptions,
2488 ) -> ChildEntriesIter<'a> {
2489 let mut cursor = self.entries_by_path.cursor(());
2490 cursor.seek(&TraversalTarget::path(parent_path), Bias::Right);
2491 let traversal = Traversal {
2492 snapshot: self,
2493 cursor,
2494 include_files: options.include_files,
2495 include_dirs: options.include_dirs,
2496 include_ignored: options.include_ignored,
2497 };
2498 ChildEntriesIter {
2499 traversal,
2500 parent_path,
2501 }
2502 }
2503
2504 pub fn root_entry(&self) -> Option<&Entry> {
2505 self.entries_by_path.first()
2506 }
2507
2508 /// Returns `None` for a single file worktree, or `Some(self.abs_path())` if
2509 /// it is a directory.
2510 pub fn root_dir(&self) -> Option<Arc<Path>> {
2511 self.root_entry()
2512 .filter(|entry| entry.is_dir())
2513 .map(|_| self.abs_path().clone())
2514 }
2515
2516 pub fn root_name(&self) -> &RelPath {
2517 &self.root_name
2518 }
2519
2520 pub fn root_name_str(&self) -> &str {
2521 self.root_name.as_unix_str()
2522 }
2523
2524 pub fn scan_id(&self) -> usize {
2525 self.scan_id
2526 }
2527
2528 pub fn entry_for_path(&self, path: &RelPath) -> Option<&Entry> {
2529 self.traverse_from_path(true, true, true, path)
2530 .entry()
2531 .and_then(|entry| {
2532 if entry.path.as_ref() == path {
2533 Some(entry)
2534 } else {
2535 None
2536 }
2537 })
2538 }
2539
2540 /// Resolves a path to an executable using the following heuristics:
2541 ///
2542 /// 1. If the path starts with `~`, it is expanded to the user's home directory.
2543 /// 2. If the path is relative and contains more than one component,
2544 /// it is joined to the worktree root path.
2545 /// 3. If the path is relative and exists in the worktree
2546 /// (even if falls under an exclusion filter),
2547 /// it is joined to the worktree root path.
2548 /// 4. Otherwise the path is returned unmodified.
2549 ///
2550 /// Relative paths that do not exist in the worktree may
2551 /// still be found using the `PATH` environment variable.
2552 pub fn resolve_relative_path(&self, path: PathBuf) -> PathBuf {
2553 if let Some(path_str) = path.to_str() {
2554 if let Some(remaining_path) = path_str.strip_prefix("~/") {
2555 return home_dir().join(remaining_path);
2556 } else if path_str == "~" {
2557 return home_dir().to_path_buf();
2558 }
2559 }
2560
2561 if let Ok(rel_path) = RelPath::new(&path, self.path_style)
2562 && (path.components().count() > 1 || self.entry_for_path(&rel_path).is_some())
2563 {
2564 self.abs_path().join(path)
2565 } else {
2566 path
2567 }
2568 }
2569
2570 pub fn entry_for_id(&self, id: ProjectEntryId) -> Option<&Entry> {
2571 let entry = self.entries_by_id.get(&id, ())?;
2572 self.entry_for_path(&entry.path)
2573 }
2574
2575 pub fn path_style(&self) -> PathStyle {
2576 self.path_style
2577 }
2578}
2579
2580impl LocalSnapshot {
2581 fn local_repo_for_work_directory_path(&self, path: &RelPath) -> Option<&LocalRepositoryEntry> {
2582 self.git_repositories
2583 .iter()
2584 .map(|(_, entry)| entry)
2585 .find(|entry| entry.work_directory.path_key() == PathKey(path.into()))
2586 }
2587
2588 fn build_update(
2589 &self,
2590 project_id: u64,
2591 worktree_id: u64,
2592 entry_changes: UpdatedEntriesSet,
2593 ) -> proto::UpdateWorktree {
2594 let mut updated_entries = Vec::new();
2595 let mut removed_entries = Vec::new();
2596
2597 for (_, entry_id, path_change) in entry_changes.iter() {
2598 if let PathChange::Removed = path_change {
2599 removed_entries.push(entry_id.0 as u64);
2600 } else if let Some(entry) = self.entry_for_id(*entry_id) {
2601 updated_entries.push(proto::Entry::from(entry));
2602 }
2603 }
2604
2605 removed_entries.sort_unstable();
2606 updated_entries.sort_unstable_by_key(|e| e.id);
2607
2608 // TODO - optimize, knowing that removed_entries are sorted.
2609 removed_entries.retain(|id| updated_entries.binary_search_by_key(id, |e| e.id).is_err());
2610
2611 proto::UpdateWorktree {
2612 project_id,
2613 worktree_id,
2614 abs_path: self.abs_path().to_string_lossy().into_owned(),
2615 root_name: self.root_name().to_proto(),
2616 updated_entries,
2617 removed_entries,
2618 scan_id: self.scan_id as u64,
2619 is_last_update: self.completed_scan_id == self.scan_id,
2620 // Sent in separate messages.
2621 updated_repositories: Vec::new(),
2622 removed_repositories: Vec::new(),
2623 }
2624 }
2625
2626 async fn insert_entry(&mut self, mut entry: Entry, fs: &dyn Fs) -> Entry {
2627 log::trace!("insert entry {:?}", entry.path);
2628 if entry.is_file() && entry.path.file_name() == Some(&GITIGNORE) {
2629 let abs_path = self.absolutize(&entry.path);
2630 match build_gitignore(&abs_path, fs).await {
2631 Ok(ignore) => {
2632 self.ignores_by_parent_abs_path
2633 .insert(abs_path.parent().unwrap().into(), (Arc::new(ignore), true));
2634 }
2635 Err(error) => {
2636 log::error!(
2637 "error loading .gitignore file {:?} - {:?}",
2638 &entry.path,
2639 error
2640 );
2641 }
2642 }
2643 }
2644
2645 if entry.kind == EntryKind::PendingDir
2646 && let Some(existing_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2647 {
2648 entry.kind = existing_entry.kind;
2649 }
2650
2651 let scan_id = self.scan_id;
2652 let removed = self.entries_by_path.insert_or_replace(entry.clone(), ());
2653 if let Some(removed) = removed
2654 && removed.id != entry.id
2655 {
2656 self.entries_by_id.remove(&removed.id, ());
2657 }
2658 self.entries_by_id.insert_or_replace(
2659 PathEntry {
2660 id: entry.id,
2661 path: entry.path.clone(),
2662 is_ignored: entry.is_ignored,
2663 scan_id,
2664 },
2665 (),
2666 );
2667
2668 entry
2669 }
2670
2671 fn ancestor_inodes_for_path(&self, path: &RelPath) -> TreeSet<u64> {
2672 let mut inodes = TreeSet::default();
2673 for ancestor in path.ancestors().skip(1) {
2674 if let Some(entry) = self.entry_for_path(ancestor) {
2675 inodes.insert(entry.inode);
2676 }
2677 }
2678 inodes
2679 }
2680
2681 async fn ignore_stack_for_abs_path(
2682 &self,
2683 abs_path: &Path,
2684 is_dir: bool,
2685 fs: &dyn Fs,
2686 ) -> IgnoreStack {
2687 let mut new_ignores = Vec::new();
2688 let mut repo_root = None;
2689 for (index, ancestor) in abs_path.ancestors().enumerate() {
2690 if index > 0 {
2691 if let Some((ignore, _)) = self.ignores_by_parent_abs_path.get(ancestor) {
2692 new_ignores.push((ancestor, Some(ignore.clone())));
2693 } else {
2694 new_ignores.push((ancestor, None));
2695 }
2696 }
2697
2698 let metadata = fs.metadata(&ancestor.join(DOT_GIT)).await.ok().flatten();
2699 if metadata.is_some() {
2700 repo_root = Some(Arc::from(ancestor));
2701 break;
2702 }
2703 }
2704
2705 let mut ignore_stack = if let Some(global_gitignore) = self.global_gitignore.clone() {
2706 IgnoreStack::global(global_gitignore)
2707 } else {
2708 IgnoreStack::none()
2709 };
2710
2711 if let Some((repo_exclude, _)) = repo_root
2712 .as_ref()
2713 .and_then(|abs_path| self.repo_exclude_by_work_dir_abs_path.get(abs_path))
2714 {
2715 ignore_stack = ignore_stack.append(IgnoreKind::RepoExclude, repo_exclude.clone());
2716 }
2717 ignore_stack.repo_root = repo_root;
2718 for (parent_abs_path, ignore) in new_ignores.into_iter().rev() {
2719 if ignore_stack.is_abs_path_ignored(parent_abs_path, true) {
2720 ignore_stack = IgnoreStack::all();
2721 break;
2722 } else if let Some(ignore) = ignore {
2723 ignore_stack =
2724 ignore_stack.append(IgnoreKind::Gitignore(parent_abs_path.into()), ignore);
2725 }
2726 }
2727
2728 if ignore_stack.is_abs_path_ignored(abs_path, is_dir) {
2729 ignore_stack = IgnoreStack::all();
2730 }
2731
2732 ignore_stack
2733 }
2734
2735 #[cfg(feature = "test-support")]
2736 pub fn expanded_entries(&self) -> impl Iterator<Item = &Entry> {
2737 self.entries_by_path
2738 .cursor::<()>(())
2739 .filter(|entry| entry.kind == EntryKind::Dir && (entry.is_external || entry.is_ignored))
2740 }
2741
2742 #[cfg(feature = "test-support")]
2743 pub fn check_invariants(&self, git_state: bool) {
2744 use pretty_assertions::assert_eq;
2745
2746 assert_eq!(
2747 self.entries_by_path
2748 .cursor::<()>(())
2749 .map(|e| (&e.path, e.id))
2750 .collect::<Vec<_>>(),
2751 self.entries_by_id
2752 .cursor::<()>(())
2753 .map(|e| (&e.path, e.id))
2754 .collect::<collections::BTreeSet<_>>()
2755 .into_iter()
2756 .collect::<Vec<_>>(),
2757 "entries_by_path and entries_by_id are inconsistent"
2758 );
2759
2760 let mut files = self.files(true, 0);
2761 let mut visible_files = self.files(false, 0);
2762 for entry in self.entries_by_path.cursor::<()>(()) {
2763 if entry.is_file() {
2764 assert_eq!(files.next().unwrap().inode, entry.inode);
2765 if (!entry.is_ignored && !entry.is_external) || entry.is_always_included {
2766 assert_eq!(visible_files.next().unwrap().inode, entry.inode);
2767 }
2768 }
2769 }
2770
2771 assert!(files.next().is_none());
2772 assert!(visible_files.next().is_none());
2773
2774 let mut bfs_paths = Vec::new();
2775 let mut stack = self
2776 .root_entry()
2777 .map(|e| e.path.as_ref())
2778 .into_iter()
2779 .collect::<Vec<_>>();
2780 while let Some(path) = stack.pop() {
2781 bfs_paths.push(path);
2782 let ix = stack.len();
2783 for child_entry in self.child_entries(path) {
2784 stack.insert(ix, &child_entry.path);
2785 }
2786 }
2787
2788 let dfs_paths_via_iter = self
2789 .entries_by_path
2790 .cursor::<()>(())
2791 .map(|e| e.path.as_ref())
2792 .collect::<Vec<_>>();
2793 assert_eq!(bfs_paths, dfs_paths_via_iter);
2794
2795 let dfs_paths_via_traversal = self
2796 .entries(true, 0)
2797 .map(|e| e.path.as_ref())
2798 .collect::<Vec<_>>();
2799
2800 assert_eq!(dfs_paths_via_traversal, dfs_paths_via_iter);
2801
2802 if git_state {
2803 for ignore_parent_abs_path in self.ignores_by_parent_abs_path.keys() {
2804 let ignore_parent_path = &RelPath::new(
2805 ignore_parent_abs_path
2806 .strip_prefix(self.abs_path.as_path())
2807 .unwrap(),
2808 PathStyle::local(),
2809 )
2810 .unwrap();
2811 assert!(self.entry_for_path(ignore_parent_path).is_some());
2812 assert!(
2813 self.entry_for_path(
2814 &ignore_parent_path.join(RelPath::unix(GITIGNORE).unwrap())
2815 )
2816 .is_some()
2817 );
2818 }
2819 }
2820 }
2821
2822 #[cfg(feature = "test-support")]
2823 pub fn entries_without_ids(&self, include_ignored: bool) -> Vec<(&RelPath, u64, bool)> {
2824 let mut paths = Vec::new();
2825 for entry in self.entries_by_path.cursor::<()>(()) {
2826 if include_ignored || !entry.is_ignored {
2827 paths.push((entry.path.as_ref(), entry.inode, entry.is_ignored));
2828 }
2829 }
2830 paths.sort_by(|a, b| a.0.cmp(b.0));
2831 paths
2832 }
2833}
2834
2835impl BackgroundScannerState {
2836 fn should_scan_directory(&self, entry: &Entry) -> bool {
2837 (self.scanning_enabled && !entry.is_external && (!entry.is_ignored || entry.is_always_included))
2838 || entry.path.file_name() == Some(DOT_GIT)
2839 || entry.path.file_name() == Some(local_settings_folder_name())
2840 || entry.path.file_name() == Some(local_vscode_folder_name())
2841 || self.scanned_dirs.contains(&entry.id) // If we've ever scanned it, keep scanning
2842 || self
2843 .paths_to_scan
2844 .iter()
2845 .any(|p| p.starts_with(&entry.path))
2846 || self
2847 .path_prefixes_to_scan
2848 .iter()
2849 .any(|p| entry.path.starts_with(p))
2850 }
2851
2852 async fn enqueue_scan_dir(
2853 &self,
2854 abs_path: Arc<Path>,
2855 entry: &Entry,
2856 scan_job_tx: &Sender<ScanJob>,
2857 fs: &dyn Fs,
2858 ) {
2859 let path = entry.path.clone();
2860 let ignore_stack = self
2861 .snapshot
2862 .ignore_stack_for_abs_path(&abs_path, true, fs)
2863 .await;
2864 let mut ancestor_inodes = self.snapshot.ancestor_inodes_for_path(&path);
2865
2866 if !ancestor_inodes.contains(&entry.inode) {
2867 ancestor_inodes.insert(entry.inode);
2868 scan_job_tx
2869 .try_send(ScanJob {
2870 abs_path,
2871 path,
2872 ignore_stack,
2873 scan_queue: scan_job_tx.clone(),
2874 ancestor_inodes,
2875 is_external: entry.is_external,
2876 })
2877 .unwrap();
2878 }
2879 }
2880
2881 fn reuse_entry_id(&mut self, entry: &mut Entry) {
2882 if let Some(mtime) = entry.mtime {
2883 // If an entry with the same inode was removed from the worktree during this scan,
2884 // then it *might* represent the same file or directory. But the OS might also have
2885 // re-used the inode for a completely different file or directory.
2886 //
2887 // Conditionally reuse the old entry's id:
2888 // * if the mtime is the same, the file was probably been renamed.
2889 // * if the path is the same, the file may just have been updated
2890 if let Some(removed_entry) = self.removed_entries.remove(&entry.inode) {
2891 if removed_entry.mtime == Some(mtime) || removed_entry.path == entry.path {
2892 entry.id = removed_entry.id;
2893 }
2894 } else if let Some(existing_entry) = self.snapshot.entry_for_path(&entry.path) {
2895 entry.id = existing_entry.id;
2896 }
2897 }
2898 }
2899
2900 fn entry_id_for(
2901 &mut self,
2902 next_entry_id: &AtomicUsize,
2903 path: &RelPath,
2904 metadata: &fs::Metadata,
2905 ) -> ProjectEntryId {
2906 // If an entry with the same inode was removed from the worktree during this scan,
2907 // then it *might* represent the same file or directory. But the OS might also have
2908 // re-used the inode for a completely different file or directory.
2909 //
2910 // Conditionally reuse the old entry's id:
2911 // * if the mtime is the same, the file was probably been renamed.
2912 // * if the path is the same, the file may just have been updated
2913 if let Some(removed_entry) = self.removed_entries.remove(&metadata.inode) {
2914 if removed_entry.mtime == Some(metadata.mtime) || *removed_entry.path == *path {
2915 return removed_entry.id;
2916 }
2917 } else if let Some(existing_entry) = self.snapshot.entry_for_path(path) {
2918 return existing_entry.id;
2919 }
2920 ProjectEntryId::new(next_entry_id)
2921 }
2922
2923 async fn insert_entry(&mut self, entry: Entry, fs: &dyn Fs, watcher: &dyn Watcher) -> Entry {
2924 let entry = self.snapshot.insert_entry(entry, fs).await;
2925 if entry.path.file_name() == Some(&DOT_GIT) {
2926 self.insert_git_repository(entry.path.clone(), fs, watcher)
2927 .await;
2928 }
2929
2930 #[cfg(feature = "test-support")]
2931 self.snapshot.check_invariants(false);
2932
2933 entry
2934 }
2935
2936 fn populate_dir(
2937 &mut self,
2938 parent_path: Arc<RelPath>,
2939 entries: impl IntoIterator<Item = Entry>,
2940 ignore: Option<Arc<Gitignore>>,
2941 ) {
2942 let mut parent_entry = if let Some(parent_entry) = self
2943 .snapshot
2944 .entries_by_path
2945 .get(&PathKey(parent_path.clone()), ())
2946 {
2947 parent_entry.clone()
2948 } else {
2949 log::warn!(
2950 "populating a directory {:?} that has been removed",
2951 parent_path
2952 );
2953 return;
2954 };
2955
2956 match parent_entry.kind {
2957 EntryKind::PendingDir | EntryKind::UnloadedDir => parent_entry.kind = EntryKind::Dir,
2958 EntryKind::Dir => {}
2959 _ => return,
2960 }
2961
2962 if let Some(ignore) = ignore {
2963 let abs_parent_path = self
2964 .snapshot
2965 .abs_path
2966 .as_path()
2967 .join(parent_path.as_std_path())
2968 .into();
2969 self.snapshot
2970 .ignores_by_parent_abs_path
2971 .insert(abs_parent_path, (ignore, false));
2972 }
2973
2974 let parent_entry_id = parent_entry.id;
2975 self.scanned_dirs.insert(parent_entry_id);
2976 let mut entries_by_path_edits = vec![Edit::Insert(parent_entry)];
2977 let mut entries_by_id_edits = Vec::new();
2978
2979 for entry in entries {
2980 entries_by_id_edits.push(Edit::Insert(PathEntry {
2981 id: entry.id,
2982 path: entry.path.clone(),
2983 is_ignored: entry.is_ignored,
2984 scan_id: self.snapshot.scan_id,
2985 }));
2986 entries_by_path_edits.push(Edit::Insert(entry));
2987 }
2988
2989 self.snapshot
2990 .entries_by_path
2991 .edit(entries_by_path_edits, ());
2992 self.snapshot.entries_by_id.edit(entries_by_id_edits, ());
2993
2994 if let Err(ix) = self.changed_paths.binary_search(&parent_path) {
2995 self.changed_paths.insert(ix, parent_path.clone());
2996 }
2997
2998 #[cfg(feature = "test-support")]
2999 self.snapshot.check_invariants(false);
3000 }
3001
3002 fn remove_path(&mut self, path: &RelPath, watcher: &dyn Watcher) {
3003 log::trace!("background scanner removing path {path:?}");
3004 let mut new_entries;
3005 let removed_entries;
3006 {
3007 let mut cursor = self
3008 .snapshot
3009 .entries_by_path
3010 .cursor::<TraversalProgress>(());
3011 new_entries = cursor.slice(&TraversalTarget::path(path), Bias::Left);
3012 removed_entries = cursor.slice(&TraversalTarget::successor(path), Bias::Left);
3013 new_entries.append(cursor.suffix(), ());
3014 }
3015 self.snapshot.entries_by_path = new_entries;
3016
3017 let mut removed_ids = Vec::with_capacity(removed_entries.summary().count);
3018 let mut removed_dir_abs_paths = Vec::new();
3019 for entry in removed_entries.cursor::<()>(()) {
3020 if entry.is_dir() {
3021 removed_dir_abs_paths.push(self.snapshot.absolutize(&entry.path));
3022 }
3023
3024 match self.removed_entries.entry(entry.inode) {
3025 hash_map::Entry::Occupied(mut e) => {
3026 let prev_removed_entry = e.get_mut();
3027 if entry.id > prev_removed_entry.id {
3028 *prev_removed_entry = entry.clone();
3029 }
3030 }
3031 hash_map::Entry::Vacant(e) => {
3032 e.insert(entry.clone());
3033 }
3034 }
3035
3036 if entry.path.file_name() == Some(GITIGNORE) {
3037 let abs_parent_path = self.snapshot.absolutize(&entry.path.parent().unwrap());
3038 if let Some((_, needs_update)) = self
3039 .snapshot
3040 .ignores_by_parent_abs_path
3041 .get_mut(abs_parent_path.as_path())
3042 {
3043 *needs_update = true;
3044 }
3045 }
3046
3047 if let Err(ix) = removed_ids.binary_search(&entry.id) {
3048 removed_ids.insert(ix, entry.id);
3049 }
3050 }
3051
3052 self.snapshot
3053 .entries_by_id
3054 .edit(removed_ids.iter().map(|&id| Edit::Remove(id)).collect(), ());
3055 self.snapshot
3056 .git_repositories
3057 .retain(|id, _| removed_ids.binary_search(id).is_err());
3058
3059 for removed_dir_abs_path in removed_dir_abs_paths {
3060 watcher.remove(&removed_dir_abs_path).log_err();
3061 }
3062
3063 #[cfg(feature = "test-support")]
3064 self.snapshot.check_invariants(false);
3065 }
3066
3067 async fn insert_git_repository(
3068 &mut self,
3069 dot_git_path: Arc<RelPath>,
3070 fs: &dyn Fs,
3071 watcher: &dyn Watcher,
3072 ) {
3073 let work_dir_path: Arc<RelPath> = match dot_git_path.parent() {
3074 Some(parent_dir) => {
3075 // Guard against repositories inside the repository metadata
3076 if parent_dir
3077 .components()
3078 .any(|component| component == DOT_GIT)
3079 {
3080 log::debug!(
3081 "not building git repository for nested `.git` directory, `.git` path in the worktree: {dot_git_path:?}"
3082 );
3083 return;
3084 };
3085
3086 parent_dir.into()
3087 }
3088 None => {
3089 // `dot_git_path.parent().is_none()` means `.git` directory is the opened worktree itself,
3090 // no files inside that directory are tracked by git, so no need to build the repo around it
3091 log::debug!(
3092 "not building git repository for the worktree itself, `.git` path in the worktree: {dot_git_path:?}"
3093 );
3094 return;
3095 }
3096 };
3097
3098 let dot_git_abs_path = Arc::from(self.snapshot.absolutize(&dot_git_path).as_ref());
3099
3100 self.insert_git_repository_for_path(
3101 WorkDirectory::InProject {
3102 relative_path: work_dir_path,
3103 },
3104 dot_git_abs_path,
3105 fs,
3106 watcher,
3107 )
3108 .await
3109 .log_err();
3110 }
3111
3112 async fn insert_git_repository_for_path(
3113 &mut self,
3114 work_directory: WorkDirectory,
3115 dot_git_abs_path: Arc<Path>,
3116 fs: &dyn Fs,
3117 watcher: &dyn Watcher,
3118 ) -> Result<LocalRepositoryEntry> {
3119 let work_dir_entry = self
3120 .snapshot
3121 .entry_for_path(&work_directory.path_key().0)
3122 .with_context(|| {
3123 format!(
3124 "working directory `{}` not indexed",
3125 work_directory
3126 .path_key()
3127 .0
3128 .display(self.snapshot.path_style)
3129 )
3130 })?;
3131 let work_directory_abs_path = self.snapshot.work_directory_abs_path(&work_directory);
3132
3133 let (repository_dir_abs_path, common_dir_abs_path) =
3134 discover_git_paths(&dot_git_abs_path, fs).await;
3135 watcher
3136 .add(&common_dir_abs_path)
3137 .context("failed to add common directory to watcher")
3138 .log_err();
3139 watcher
3140 .add(&repository_dir_abs_path)
3141 .context("failed to add repository directory to watcher")
3142 .log_err();
3143
3144 let work_directory_id = work_dir_entry.id;
3145
3146 let local_repository = LocalRepositoryEntry {
3147 work_directory_id,
3148 work_directory,
3149 work_directory_abs_path: work_directory_abs_path.as_path().into(),
3150 git_dir_scan_id: 0,
3151 dot_git_abs_path,
3152 common_dir_abs_path,
3153 repository_dir_abs_path,
3154 };
3155
3156 self.snapshot
3157 .git_repositories
3158 .insert(work_directory_id, local_repository.clone());
3159
3160 log::trace!("inserting new local git repository");
3161 Ok(local_repository)
3162 }
3163}
3164
3165async fn is_git_dir(path: &Path, fs: &dyn Fs) -> bool {
3166 if let Some(file_name) = path.file_name()
3167 && file_name == DOT_GIT
3168 {
3169 return true;
3170 }
3171
3172 // If we're in a bare repository, we are not inside a `.git` folder. In a
3173 // bare repository, the root folder contains what would normally be in the
3174 // `.git` folder.
3175 let head_metadata = fs.metadata(&path.join("HEAD")).await;
3176 if !matches!(head_metadata, Ok(Some(_))) {
3177 return false;
3178 }
3179 let config_metadata = fs.metadata(&path.join("config")).await;
3180 matches!(config_metadata, Ok(Some(_)))
3181}
3182
3183async fn build_gitignore(abs_path: &Path, fs: &dyn Fs) -> Result<Gitignore> {
3184 let contents = fs
3185 .load(abs_path)
3186 .await
3187 .with_context(|| format!("failed to load gitignore file at {}", abs_path.display()))?;
3188 let parent = abs_path.parent().unwrap_or_else(|| Path::new("/"));
3189 let mut builder = GitignoreBuilder::new(parent);
3190 for line in contents.lines() {
3191 builder.add_line(Some(abs_path.into()), line)?;
3192 }
3193 Ok(builder.build()?)
3194}
3195
3196impl Deref for Worktree {
3197 type Target = Snapshot;
3198
3199 fn deref(&self) -> &Self::Target {
3200 match self {
3201 Worktree::Local(worktree) => &worktree.snapshot,
3202 Worktree::Remote(worktree) => &worktree.snapshot,
3203 }
3204 }
3205}
3206
3207impl Deref for LocalWorktree {
3208 type Target = LocalSnapshot;
3209
3210 fn deref(&self) -> &Self::Target {
3211 &self.snapshot
3212 }
3213}
3214
3215impl Deref for RemoteWorktree {
3216 type Target = Snapshot;
3217
3218 fn deref(&self) -> &Self::Target {
3219 &self.snapshot
3220 }
3221}
3222
3223impl fmt::Debug for LocalWorktree {
3224 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3225 self.snapshot.fmt(f)
3226 }
3227}
3228
3229impl fmt::Debug for Snapshot {
3230 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3231 struct EntriesById<'a>(&'a SumTree<PathEntry>);
3232 struct EntriesByPath<'a>(&'a SumTree<Entry>);
3233
3234 impl fmt::Debug for EntriesByPath<'_> {
3235 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3236 f.debug_map()
3237 .entries(self.0.iter().map(|entry| (&entry.path, entry.id)))
3238 .finish()
3239 }
3240 }
3241
3242 impl fmt::Debug for EntriesById<'_> {
3243 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3244 f.debug_list().entries(self.0.iter()).finish()
3245 }
3246 }
3247
3248 f.debug_struct("Snapshot")
3249 .field("id", &self.id)
3250 .field("root_name", &self.root_name)
3251 .field("entries_by_path", &EntriesByPath(&self.entries_by_path))
3252 .field("entries_by_id", &EntriesById(&self.entries_by_id))
3253 .finish()
3254 }
3255}
3256
3257#[derive(Debug, Clone, PartialEq)]
3258pub struct File {
3259 pub worktree: Entity<Worktree>,
3260 pub path: Arc<RelPath>,
3261 pub disk_state: DiskState,
3262 pub entry_id: Option<ProjectEntryId>,
3263 pub is_local: bool,
3264 pub is_private: bool,
3265}
3266
3267impl language::File for File {
3268 fn as_local(&self) -> Option<&dyn language::LocalFile> {
3269 if self.is_local { Some(self) } else { None }
3270 }
3271
3272 fn disk_state(&self) -> DiskState {
3273 self.disk_state
3274 }
3275
3276 fn path(&self) -> &Arc<RelPath> {
3277 &self.path
3278 }
3279
3280 fn full_path(&self, cx: &App) -> PathBuf {
3281 self.worktree.read(cx).full_path(&self.path)
3282 }
3283
3284 /// Returns the last component of this handle's absolute path. If this handle refers to the root
3285 /// of its worktree, then this method will return the name of the worktree itself.
3286 fn file_name<'a>(&'a self, cx: &'a App) -> &'a str {
3287 self.path
3288 .file_name()
3289 .unwrap_or_else(|| self.worktree.read(cx).root_name_str())
3290 }
3291
3292 fn worktree_id(&self, cx: &App) -> WorktreeId {
3293 self.worktree.read(cx).id()
3294 }
3295
3296 fn to_proto(&self, cx: &App) -> rpc::proto::File {
3297 rpc::proto::File {
3298 worktree_id: self.worktree.read(cx).id().to_proto(),
3299 entry_id: self.entry_id.map(|id| id.to_proto()),
3300 path: self.path.as_ref().to_proto(),
3301 mtime: self.disk_state.mtime().map(|time| time.into()),
3302 is_deleted: self.disk_state.is_deleted(),
3303 is_historic: matches!(self.disk_state, DiskState::Historic { .. }),
3304 }
3305 }
3306
3307 fn is_private(&self) -> bool {
3308 self.is_private
3309 }
3310
3311 fn path_style(&self, cx: &App) -> PathStyle {
3312 self.worktree.read(cx).path_style()
3313 }
3314
3315 fn can_open(&self) -> bool {
3316 true
3317 }
3318}
3319
3320impl language::LocalFile for File {
3321 fn abs_path(&self, cx: &App) -> PathBuf {
3322 self.worktree.read(cx).absolutize(&self.path)
3323 }
3324
3325 fn load(&self, cx: &App) -> Task<Result<String>> {
3326 let worktree = self.worktree.read(cx).as_local().unwrap();
3327 let abs_path = worktree.absolutize(&self.path);
3328 let fs = worktree.fs.clone();
3329 cx.background_spawn(async move { fs.load(&abs_path).await })
3330 }
3331
3332 fn load_bytes(&self, cx: &App) -> Task<Result<Vec<u8>>> {
3333 let worktree = self.worktree.read(cx).as_local().unwrap();
3334 let abs_path = worktree.absolutize(&self.path);
3335 let fs = worktree.fs.clone();
3336 cx.background_spawn(async move { fs.load_bytes(&abs_path).await })
3337 }
3338}
3339
3340impl File {
3341 pub fn for_entry(entry: Entry, worktree: Entity<Worktree>) -> Arc<Self> {
3342 Arc::new(Self {
3343 worktree,
3344 path: entry.path.clone(),
3345 disk_state: if let Some(mtime) = entry.mtime {
3346 DiskState::Present {
3347 mtime,
3348 size: entry.size,
3349 }
3350 } else {
3351 DiskState::New
3352 },
3353 entry_id: Some(entry.id),
3354 is_local: true,
3355 is_private: entry.is_private,
3356 })
3357 }
3358
3359 pub fn from_proto(
3360 proto: rpc::proto::File,
3361 worktree: Entity<Worktree>,
3362 cx: &App,
3363 ) -> Result<Self> {
3364 let worktree_id = worktree.read(cx).as_remote().context("not remote")?.id();
3365
3366 anyhow::ensure!(
3367 worktree_id.to_proto() == proto.worktree_id,
3368 "worktree id does not match file"
3369 );
3370
3371 let disk_state = if proto.is_historic {
3372 DiskState::Historic {
3373 was_deleted: proto.is_deleted,
3374 }
3375 } else if proto.is_deleted {
3376 DiskState::Deleted
3377 } else if let Some(mtime) = proto.mtime.map(&Into::into) {
3378 DiskState::Present { mtime, size: 0 }
3379 } else {
3380 DiskState::New
3381 };
3382
3383 Ok(Self {
3384 worktree,
3385 path: RelPath::from_proto(&proto.path).context("invalid path in file protobuf")?,
3386 disk_state,
3387 entry_id: proto.entry_id.map(ProjectEntryId::from_proto),
3388 is_local: false,
3389 is_private: false,
3390 })
3391 }
3392
3393 pub fn from_dyn(file: Option<&Arc<dyn language::File>>) -> Option<&Self> {
3394 file.and_then(|f| {
3395 let f: &dyn language::File = f.borrow();
3396 let f: &dyn Any = f;
3397 f.downcast_ref()
3398 })
3399 }
3400
3401 pub fn worktree_id(&self, cx: &App) -> WorktreeId {
3402 self.worktree.read(cx).id()
3403 }
3404
3405 pub fn project_entry_id(&self) -> Option<ProjectEntryId> {
3406 match self.disk_state {
3407 DiskState::Deleted => None,
3408 _ => self.entry_id,
3409 }
3410 }
3411}
3412
3413#[derive(Clone, Debug, PartialEq, Eq)]
3414pub struct Entry {
3415 pub id: ProjectEntryId,
3416 pub kind: EntryKind,
3417 pub path: Arc<RelPath>,
3418 pub inode: u64,
3419 pub mtime: Option<MTime>,
3420
3421 pub canonical_path: Option<Arc<Path>>,
3422 /// Whether this entry is ignored by Git.
3423 ///
3424 /// We only scan ignored entries once the directory is expanded and
3425 /// exclude them from searches.
3426 pub is_ignored: bool,
3427
3428 /// Whether this entry is hidden or inside hidden directory.
3429 ///
3430 /// We only scan hidden entries once the directory is expanded.
3431 pub is_hidden: bool,
3432
3433 /// Whether this entry is always included in searches.
3434 ///
3435 /// This is used for entries that are always included in searches, even
3436 /// if they are ignored by git. Overridden by file_scan_exclusions.
3437 pub is_always_included: bool,
3438
3439 /// Whether this entry's canonical path is outside of the worktree.
3440 /// This means the entry is only accessible from the worktree root via a
3441 /// symlink.
3442 ///
3443 /// We only scan entries outside of the worktree once the symlinked
3444 /// directory is expanded. External entries are treated like gitignored
3445 /// entries in that they are not included in searches.
3446 pub is_external: bool,
3447
3448 /// Whether this entry is considered to be a `.env` file.
3449 pub is_private: bool,
3450 /// The entry's size on disk, in bytes.
3451 pub size: u64,
3452 pub char_bag: CharBag,
3453 pub is_fifo: bool,
3454}
3455
3456#[derive(Clone, Copy, Debug, PartialEq, Eq)]
3457pub enum EntryKind {
3458 UnloadedDir,
3459 PendingDir,
3460 Dir,
3461 File,
3462}
3463
3464#[derive(Clone, Copy, Debug, PartialEq)]
3465pub enum PathChange {
3466 /// A filesystem entry was was created.
3467 Added,
3468 /// A filesystem entry was removed.
3469 Removed,
3470 /// A filesystem entry was updated.
3471 Updated,
3472 /// A filesystem entry was either updated or added. We don't know
3473 /// whether or not it already existed, because the path had not
3474 /// been loaded before the event.
3475 AddedOrUpdated,
3476 /// A filesystem entry was found during the initial scan of the worktree.
3477 Loaded,
3478}
3479
3480#[derive(Clone, Debug, PartialEq, Eq)]
3481pub struct UpdatedGitRepository {
3482 /// ID of the repository's working directory.
3483 ///
3484 /// For a repo that's above the worktree root, this is the ID of the worktree root, and hence not unique.
3485 /// It's included here to aid the GitStore in detecting when a repository's working directory is renamed.
3486 pub work_directory_id: ProjectEntryId,
3487 pub old_work_directory_abs_path: Option<Arc<Path>>,
3488 pub new_work_directory_abs_path: Option<Arc<Path>>,
3489 /// For a normal git repository checkout, the absolute path to the .git directory.
3490 /// For a worktree, the absolute path to the worktree's subdirectory inside the .git directory.
3491 pub dot_git_abs_path: Option<Arc<Path>>,
3492 pub repository_dir_abs_path: Option<Arc<Path>>,
3493 pub common_dir_abs_path: Option<Arc<Path>>,
3494}
3495
3496pub type UpdatedEntriesSet = Arc<[(Arc<RelPath>, ProjectEntryId, PathChange)]>;
3497pub type UpdatedGitRepositoriesSet = Arc<[UpdatedGitRepository]>;
3498
3499#[derive(Clone, Debug)]
3500pub struct PathProgress<'a> {
3501 pub max_path: &'a RelPath,
3502}
3503
3504#[derive(Clone, Debug)]
3505pub struct PathSummary<S> {
3506 pub max_path: Arc<RelPath>,
3507 pub item_summary: S,
3508}
3509
3510impl<S: Summary> Summary for PathSummary<S> {
3511 type Context<'a> = S::Context<'a>;
3512
3513 fn zero(cx: Self::Context<'_>) -> Self {
3514 Self {
3515 max_path: RelPath::empty().into(),
3516 item_summary: S::zero(cx),
3517 }
3518 }
3519
3520 fn add_summary(&mut self, rhs: &Self, cx: Self::Context<'_>) {
3521 self.max_path = rhs.max_path.clone();
3522 self.item_summary.add_summary(&rhs.item_summary, cx);
3523 }
3524}
3525
3526impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathProgress<'a> {
3527 fn zero(_: <PathSummary<S> as Summary>::Context<'_>) -> Self {
3528 Self {
3529 max_path: RelPath::empty(),
3530 }
3531 }
3532
3533 fn add_summary(
3534 &mut self,
3535 summary: &'a PathSummary<S>,
3536 _: <PathSummary<S> as Summary>::Context<'_>,
3537 ) {
3538 self.max_path = summary.max_path.as_ref()
3539 }
3540}
3541
3542impl<'a> sum_tree::Dimension<'a, PathSummary<GitSummary>> for GitSummary {
3543 fn zero(_cx: ()) -> Self {
3544 Default::default()
3545 }
3546
3547 fn add_summary(&mut self, summary: &'a PathSummary<GitSummary>, _: ()) {
3548 *self += summary.item_summary
3549 }
3550}
3551
3552impl<'a>
3553 sum_tree::SeekTarget<'a, PathSummary<GitSummary>, Dimensions<TraversalProgress<'a>, GitSummary>>
3554 for PathTarget<'_>
3555{
3556 fn cmp(
3557 &self,
3558 cursor_location: &Dimensions<TraversalProgress<'a>, GitSummary>,
3559 _: (),
3560 ) -> Ordering {
3561 self.cmp_path(cursor_location.0.max_path)
3562 }
3563}
3564
3565impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathKey {
3566 fn zero(_: S::Context<'_>) -> Self {
3567 Default::default()
3568 }
3569
3570 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3571 self.0 = summary.max_path.clone();
3572 }
3573}
3574
3575impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for TraversalProgress<'a> {
3576 fn zero(_cx: S::Context<'_>) -> Self {
3577 Default::default()
3578 }
3579
3580 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3581 self.max_path = summary.max_path.as_ref();
3582 }
3583}
3584
3585impl Entry {
3586 fn new(
3587 path: Arc<RelPath>,
3588 metadata: &fs::Metadata,
3589 id: ProjectEntryId,
3590 root_char_bag: CharBag,
3591 canonical_path: Option<Arc<Path>>,
3592 ) -> Self {
3593 let char_bag = char_bag_for_path(root_char_bag, &path);
3594 Self {
3595 id,
3596 kind: if metadata.is_dir {
3597 EntryKind::PendingDir
3598 } else {
3599 EntryKind::File
3600 },
3601 path,
3602 inode: metadata.inode,
3603 mtime: Some(metadata.mtime),
3604 size: metadata.len,
3605 canonical_path,
3606 is_ignored: false,
3607 is_hidden: false,
3608 is_always_included: false,
3609 is_external: false,
3610 is_private: false,
3611 char_bag,
3612 is_fifo: metadata.is_fifo,
3613 }
3614 }
3615
3616 pub fn is_created(&self) -> bool {
3617 self.mtime.is_some()
3618 }
3619
3620 pub fn is_dir(&self) -> bool {
3621 self.kind.is_dir()
3622 }
3623
3624 pub fn is_file(&self) -> bool {
3625 self.kind.is_file()
3626 }
3627}
3628
3629impl EntryKind {
3630 pub fn is_dir(&self) -> bool {
3631 matches!(
3632 self,
3633 EntryKind::Dir | EntryKind::PendingDir | EntryKind::UnloadedDir
3634 )
3635 }
3636
3637 pub fn is_unloaded(&self) -> bool {
3638 matches!(self, EntryKind::UnloadedDir)
3639 }
3640
3641 pub fn is_file(&self) -> bool {
3642 matches!(self, EntryKind::File)
3643 }
3644}
3645
3646impl sum_tree::Item for Entry {
3647 type Summary = EntrySummary;
3648
3649 fn summary(&self, _cx: ()) -> Self::Summary {
3650 let non_ignored_count = if (self.is_ignored || self.is_external) && !self.is_always_included
3651 {
3652 0
3653 } else {
3654 1
3655 };
3656 let file_count;
3657 let non_ignored_file_count;
3658 if self.is_file() {
3659 file_count = 1;
3660 non_ignored_file_count = non_ignored_count;
3661 } else {
3662 file_count = 0;
3663 non_ignored_file_count = 0;
3664 }
3665
3666 EntrySummary {
3667 max_path: self.path.clone(),
3668 count: 1,
3669 non_ignored_count,
3670 file_count,
3671 non_ignored_file_count,
3672 }
3673 }
3674}
3675
3676impl sum_tree::KeyedItem for Entry {
3677 type Key = PathKey;
3678
3679 fn key(&self) -> Self::Key {
3680 PathKey(self.path.clone())
3681 }
3682}
3683
3684#[derive(Clone, Debug)]
3685pub struct EntrySummary {
3686 max_path: Arc<RelPath>,
3687 count: usize,
3688 non_ignored_count: usize,
3689 file_count: usize,
3690 non_ignored_file_count: usize,
3691}
3692
3693impl Default for EntrySummary {
3694 fn default() -> Self {
3695 Self {
3696 max_path: Arc::from(RelPath::empty()),
3697 count: 0,
3698 non_ignored_count: 0,
3699 file_count: 0,
3700 non_ignored_file_count: 0,
3701 }
3702 }
3703}
3704
3705impl sum_tree::ContextLessSummary for EntrySummary {
3706 fn zero() -> Self {
3707 Default::default()
3708 }
3709
3710 fn add_summary(&mut self, rhs: &Self) {
3711 self.max_path = rhs.max_path.clone();
3712 self.count += rhs.count;
3713 self.non_ignored_count += rhs.non_ignored_count;
3714 self.file_count += rhs.file_count;
3715 self.non_ignored_file_count += rhs.non_ignored_file_count;
3716 }
3717}
3718
3719#[derive(Clone, Debug)]
3720struct PathEntry {
3721 id: ProjectEntryId,
3722 path: Arc<RelPath>,
3723 is_ignored: bool,
3724 scan_id: usize,
3725}
3726
3727impl sum_tree::Item for PathEntry {
3728 type Summary = PathEntrySummary;
3729
3730 fn summary(&self, _cx: ()) -> Self::Summary {
3731 PathEntrySummary { max_id: self.id }
3732 }
3733}
3734
3735impl sum_tree::KeyedItem for PathEntry {
3736 type Key = ProjectEntryId;
3737
3738 fn key(&self) -> Self::Key {
3739 self.id
3740 }
3741}
3742
3743#[derive(Clone, Debug, Default)]
3744struct PathEntrySummary {
3745 max_id: ProjectEntryId,
3746}
3747
3748impl sum_tree::ContextLessSummary for PathEntrySummary {
3749 fn zero() -> Self {
3750 Default::default()
3751 }
3752
3753 fn add_summary(&mut self, summary: &Self) {
3754 self.max_id = summary.max_id;
3755 }
3756}
3757
3758impl<'a> sum_tree::Dimension<'a, PathEntrySummary> for ProjectEntryId {
3759 fn zero(_cx: ()) -> Self {
3760 Default::default()
3761 }
3762
3763 fn add_summary(&mut self, summary: &'a PathEntrySummary, _: ()) {
3764 *self = summary.max_id;
3765 }
3766}
3767
3768#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
3769pub struct PathKey(pub Arc<RelPath>);
3770
3771impl Default for PathKey {
3772 fn default() -> Self {
3773 Self(RelPath::empty().into())
3774 }
3775}
3776
3777impl<'a> sum_tree::Dimension<'a, EntrySummary> for PathKey {
3778 fn zero(_cx: ()) -> Self {
3779 Default::default()
3780 }
3781
3782 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
3783 self.0 = summary.max_path.clone();
3784 }
3785}
3786
3787struct BackgroundScanner {
3788 state: async_lock::Mutex<BackgroundScannerState>,
3789 fs: Arc<dyn Fs>,
3790 fs_case_sensitive: bool,
3791 status_updates_tx: UnboundedSender<ScanState>,
3792 executor: BackgroundExecutor,
3793 scan_requests_rx: channel::Receiver<ScanRequest>,
3794 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
3795 next_entry_id: Arc<AtomicUsize>,
3796 phase: BackgroundScannerPhase,
3797 watcher: Arc<dyn Watcher>,
3798 settings: WorktreeSettings,
3799 share_private_files: bool,
3800}
3801
3802#[derive(Copy, Clone, PartialEq)]
3803enum BackgroundScannerPhase {
3804 InitialScan,
3805 EventsReceivedDuringInitialScan,
3806 Events,
3807}
3808
3809impl BackgroundScanner {
3810 async fn run(&mut self, mut fs_events_rx: Pin<Box<dyn Send + Stream<Item = Vec<PathEvent>>>>) {
3811 let root_abs_path;
3812 let scanning_enabled;
3813 {
3814 let state = self.state.lock().await;
3815 root_abs_path = state.snapshot.abs_path.clone();
3816 scanning_enabled = state.scanning_enabled;
3817 }
3818
3819 // If the worktree root does not contain a git repository, then find
3820 // the git repository in an ancestor directory. Find any gitignore files
3821 // in ancestor directories.
3822 let repo = if scanning_enabled {
3823 let (ignores, exclude, repo) =
3824 discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await;
3825 self.state
3826 .lock()
3827 .await
3828 .snapshot
3829 .ignores_by_parent_abs_path
3830 .extend(ignores);
3831 if let Some(exclude) = exclude {
3832 self.state
3833 .lock()
3834 .await
3835 .snapshot
3836 .repo_exclude_by_work_dir_abs_path
3837 .insert(root_abs_path.as_path().into(), (exclude, false));
3838 }
3839
3840 repo
3841 } else {
3842 None
3843 };
3844
3845 let containing_git_repository = if let Some((ancestor_dot_git, work_directory)) = repo
3846 && scanning_enabled
3847 {
3848 maybe!(async {
3849 self.state
3850 .lock()
3851 .await
3852 .insert_git_repository_for_path(
3853 work_directory,
3854 ancestor_dot_git.clone().into(),
3855 self.fs.as_ref(),
3856 self.watcher.as_ref(),
3857 )
3858 .await
3859 .log_err()?;
3860 Some(ancestor_dot_git)
3861 })
3862 .await
3863 } else {
3864 None
3865 };
3866
3867 log::trace!("containing git repository: {containing_git_repository:?}");
3868
3869 let global_gitignore_file = paths::global_gitignore_path();
3870 let mut global_gitignore_events = if let Some(global_gitignore_path) =
3871 &global_gitignore_file
3872 && scanning_enabled
3873 {
3874 let is_file = self.fs.is_file(&global_gitignore_path).await;
3875 self.state.lock().await.snapshot.global_gitignore = if is_file {
3876 build_gitignore(global_gitignore_path, self.fs.as_ref())
3877 .await
3878 .ok()
3879 .map(Arc::new)
3880 } else {
3881 None
3882 };
3883 if is_file {
3884 self.fs
3885 .watch(global_gitignore_path, FS_WATCH_LATENCY)
3886 .await
3887 .0
3888 } else {
3889 Box::pin(futures::stream::pending())
3890 }
3891 } else {
3892 self.state.lock().await.snapshot.global_gitignore = None;
3893 Box::pin(futures::stream::pending())
3894 };
3895
3896 let (scan_job_tx, scan_job_rx) = channel::unbounded();
3897 {
3898 let mut state = self.state.lock().await;
3899 state.snapshot.scan_id += 1;
3900 if let Some(mut root_entry) = state.snapshot.root_entry().cloned() {
3901 let ignore_stack = state
3902 .snapshot
3903 .ignore_stack_for_abs_path(root_abs_path.as_path(), true, self.fs.as_ref())
3904 .await;
3905 if ignore_stack.is_abs_path_ignored(root_abs_path.as_path(), true) {
3906 root_entry.is_ignored = true;
3907 let mut root_entry = root_entry.clone();
3908 state.reuse_entry_id(&mut root_entry);
3909 state
3910 .insert_entry(root_entry, self.fs.as_ref(), self.watcher.as_ref())
3911 .await;
3912 }
3913 if root_entry.is_dir() && state.scanning_enabled {
3914 state
3915 .enqueue_scan_dir(
3916 root_abs_path.as_path().into(),
3917 &root_entry,
3918 &scan_job_tx,
3919 self.fs.as_ref(),
3920 )
3921 .await;
3922 }
3923 }
3924 };
3925
3926 // Perform an initial scan of the directory.
3927 drop(scan_job_tx);
3928 self.scan_dirs(true, scan_job_rx).await;
3929 {
3930 let mut state = self.state.lock().await;
3931 state.snapshot.completed_scan_id = state.snapshot.scan_id;
3932 }
3933
3934 self.send_status_update(false, SmallVec::new(), &[]).await;
3935
3936 // Process any any FS events that occurred while performing the initial scan.
3937 // For these events, update events cannot be as precise, because we didn't
3938 // have the previous state loaded yet.
3939 self.phase = BackgroundScannerPhase::EventsReceivedDuringInitialScan;
3940 if let Poll::Ready(Some(mut paths)) = futures::poll!(fs_events_rx.next()) {
3941 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
3942 paths.extend(more_paths);
3943 }
3944 self.process_events(
3945 paths
3946 .into_iter()
3947 .filter(|event| event.kind.is_some())
3948 .collect(),
3949 )
3950 .await;
3951 }
3952 if let Some(abs_path) = containing_git_repository {
3953 self.process_events(vec![PathEvent {
3954 path: abs_path,
3955 kind: Some(fs::PathEventKind::Changed),
3956 }])
3957 .await;
3958 }
3959
3960 // Continue processing events until the worktree is dropped.
3961 self.phase = BackgroundScannerPhase::Events;
3962
3963 loop {
3964 select_biased! {
3965 // Process any path refresh requests from the worktree. Prioritize
3966 // these before handling changes reported by the filesystem.
3967 request = self.next_scan_request().fuse() => {
3968 let Ok(request) = request else { break };
3969 if !self.process_scan_request(request, false).await {
3970 return;
3971 }
3972 }
3973
3974 path_prefix_request = self.path_prefixes_to_scan_rx.recv().fuse() => {
3975 let Ok(request) = path_prefix_request else { break };
3976 log::trace!("adding path prefix {:?}", request.path);
3977
3978 let did_scan = self.forcibly_load_paths(std::slice::from_ref(&request.path)).await;
3979 if did_scan {
3980 let abs_path =
3981 {
3982 let mut state = self.state.lock().await;
3983 state.path_prefixes_to_scan.insert(request.path.clone());
3984 state.snapshot.absolutize(&request.path)
3985 };
3986
3987 if let Some(abs_path) = self.fs.canonicalize(&abs_path).await.log_err() {
3988 self.process_events(vec![PathEvent {
3989 path: abs_path,
3990 kind: Some(fs::PathEventKind::Changed),
3991 }])
3992 .await;
3993 }
3994 }
3995 self.send_status_update(false, request.done, &[]).await;
3996 }
3997
3998 paths = fs_events_rx.next().fuse() => {
3999 let Some(mut paths) = paths else { break };
4000 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
4001 paths.extend(more_paths);
4002 }
4003 self.process_events(paths.into_iter().filter(|event| event.kind.is_some()).collect()).await;
4004 }
4005
4006 _ = global_gitignore_events.next().fuse() => {
4007 if let Some(path) = &global_gitignore_file {
4008 self.update_global_gitignore(&path).await;
4009 }
4010 }
4011 }
4012 }
4013 }
4014
4015 async fn process_scan_request(&self, mut request: ScanRequest, scanning: bool) -> bool {
4016 log::debug!("rescanning paths {:?}", request.relative_paths);
4017
4018 request.relative_paths.sort_unstable();
4019 self.forcibly_load_paths(&request.relative_paths).await;
4020
4021 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4022 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4023 let root_canonical_path = match &root_canonical_path {
4024 Ok(path) => SanitizedPath::new(path),
4025 Err(err) => {
4026 log::error!("failed to canonicalize root path {root_path:?}: {err:#}");
4027 return true;
4028 }
4029 };
4030 let abs_paths = request
4031 .relative_paths
4032 .iter()
4033 .map(|path| {
4034 if path.file_name().is_some() {
4035 root_canonical_path.as_path().join(path.as_std_path())
4036 } else {
4037 root_canonical_path.as_path().to_path_buf()
4038 }
4039 })
4040 .collect::<Vec<_>>();
4041
4042 {
4043 let mut state = self.state.lock().await;
4044 let is_idle = state.snapshot.completed_scan_id == state.snapshot.scan_id;
4045 state.snapshot.scan_id += 1;
4046 if is_idle {
4047 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4048 }
4049 }
4050
4051 self.reload_entries_for_paths(
4052 &root_path,
4053 &root_canonical_path,
4054 &request.relative_paths,
4055 abs_paths,
4056 None,
4057 )
4058 .await;
4059
4060 self.send_status_update(scanning, request.done, &[]).await
4061 }
4062
4063 async fn process_events(&self, mut events: Vec<PathEvent>) {
4064 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4065 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4066 let root_canonical_path = match &root_canonical_path {
4067 Ok(path) => SanitizedPath::new(path),
4068 Err(err) => {
4069 let new_path = self
4070 .state
4071 .lock()
4072 .await
4073 .snapshot
4074 .root_file_handle
4075 .clone()
4076 .and_then(|handle| match handle.current_path(&self.fs) {
4077 Ok(new_path) => Some(new_path),
4078 Err(e) => {
4079 log::error!("Failed to refresh worktree root path: {e:#}");
4080 None
4081 }
4082 })
4083 .map(|path| SanitizedPath::new_arc(&path))
4084 .filter(|new_path| *new_path != root_path);
4085
4086 if let Some(new_path) = new_path {
4087 log::info!(
4088 "root renamed from {:?} to {:?}",
4089 root_path.as_path(),
4090 new_path.as_path(),
4091 );
4092 self.status_updates_tx
4093 .unbounded_send(ScanState::RootUpdated { new_path })
4094 .ok();
4095 } else {
4096 log::error!("root path could not be canonicalized: {err:#}");
4097 }
4098 return;
4099 }
4100 };
4101
4102 // Certain directories may have FS changes, but do not lead to git data changes that Zed cares about.
4103 // Ignore these, to avoid Zed unnecessarily rescanning git metadata.
4104 let skipped_files_in_dot_git = [COMMIT_MESSAGE, INDEX_LOCK];
4105 let skipped_dirs_in_dot_git = [FSMONITOR_DAEMON, LFS_DIR];
4106
4107 let mut relative_paths = Vec::with_capacity(events.len());
4108 let mut dot_git_abs_paths = Vec::new();
4109 let mut work_dirs_needing_exclude_update = Vec::new();
4110 events.sort_unstable_by(|left, right| left.path.cmp(&right.path));
4111 events.dedup_by(|left, right| {
4112 if left.path == right.path {
4113 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4114 right.kind = left.kind;
4115 }
4116 true
4117 } else if left.path.starts_with(&right.path) {
4118 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4119 right.kind = left.kind;
4120 }
4121 true
4122 } else {
4123 false
4124 }
4125 });
4126 {
4127 let snapshot = &self.state.lock().await.snapshot;
4128
4129 let mut ranges_to_drop = SmallVec::<[Range<usize>; 4]>::new();
4130
4131 fn skip_ix(ranges: &mut SmallVec<[Range<usize>; 4]>, ix: usize) {
4132 if let Some(last_range) = ranges.last_mut()
4133 && last_range.end == ix
4134 {
4135 last_range.end += 1;
4136 } else {
4137 ranges.push(ix..ix + 1);
4138 }
4139 }
4140
4141 for (ix, event) in events.iter().enumerate() {
4142 let abs_path = SanitizedPath::new(&event.path);
4143
4144 let mut is_git_related = false;
4145 let mut dot_git_paths = None;
4146
4147 for ancestor in abs_path.as_path().ancestors() {
4148 if is_git_dir(ancestor, self.fs.as_ref()).await {
4149 let path_in_git_dir = abs_path
4150 .as_path()
4151 .strip_prefix(ancestor)
4152 .expect("stripping off the ancestor");
4153 dot_git_paths = Some((ancestor.to_owned(), path_in_git_dir.to_owned()));
4154 break;
4155 }
4156 }
4157
4158 if let Some((dot_git_abs_path, path_in_git_dir)) = dot_git_paths {
4159 // We ignore `""` as well, as that is going to be the
4160 // `.git` folder itself. WE do not care about it, if
4161 // there are changes within we will see them, we need
4162 // this ignore to prevent us from accidentally observing
4163 // the ignored created file due to the events not being
4164 // empty after filtering.
4165
4166 let is_dot_git_changed = {
4167 path_in_git_dir == Path::new("")
4168 && event.kind == Some(PathEventKind::Changed)
4169 && abs_path
4170 .strip_prefix(root_canonical_path)
4171 .ok()
4172 .and_then(|it| RelPath::new(it, PathStyle::local()).ok())
4173 .is_some_and(|it| {
4174 snapshot
4175 .entry_for_path(&it)
4176 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4177 })
4178 };
4179 let condition = skipped_files_in_dot_git.iter().any(|skipped| {
4180 OsStr::new(skipped) == path_in_git_dir.as_path().as_os_str()
4181 }) || skipped_dirs_in_dot_git
4182 .iter()
4183 .any(|skipped_git_subdir| path_in_git_dir.starts_with(skipped_git_subdir))
4184 || is_dot_git_changed;
4185 if condition {
4186 log::debug!(
4187 "ignoring event {abs_path:?} as it's in the .git directory among skipped files or directories"
4188 );
4189 skip_ix(&mut ranges_to_drop, ix);
4190 continue;
4191 }
4192
4193 is_git_related = true;
4194 if !dot_git_abs_paths.contains(&dot_git_abs_path) {
4195 dot_git_abs_paths.push(dot_git_abs_path);
4196 }
4197 }
4198
4199 let relative_path = if let Ok(path) = abs_path.strip_prefix(&root_canonical_path)
4200 && let Ok(path) = RelPath::new(path, PathStyle::local())
4201 {
4202 path
4203 } else {
4204 if is_git_related {
4205 log::debug!(
4206 "ignoring event {abs_path:?}, since it's in git dir outside of root path {root_canonical_path:?}",
4207 );
4208 } else {
4209 log::error!(
4210 "ignoring event {abs_path:?} outside of root path {root_canonical_path:?}",
4211 );
4212 }
4213 skip_ix(&mut ranges_to_drop, ix);
4214 continue;
4215 };
4216
4217 let absolute_path = abs_path.to_path_buf();
4218 if absolute_path.ends_with(Path::new(DOT_GIT).join(REPO_EXCLUDE)) {
4219 if let Some(repository) = snapshot
4220 .git_repositories
4221 .values()
4222 .find(|repo| repo.common_dir_abs_path.join(REPO_EXCLUDE) == absolute_path)
4223 {
4224 work_dirs_needing_exclude_update
4225 .push(repository.work_directory_abs_path.clone());
4226 }
4227 }
4228
4229 if abs_path.file_name() == Some(OsStr::new(GITIGNORE)) {
4230 for (_, repo) in snapshot
4231 .git_repositories
4232 .iter()
4233 .filter(|(_, repo)| repo.directory_contains(&relative_path))
4234 {
4235 if !dot_git_abs_paths.iter().any(|dot_git_abs_path| {
4236 dot_git_abs_path == repo.common_dir_abs_path.as_ref()
4237 }) {
4238 dot_git_abs_paths.push(repo.common_dir_abs_path.to_path_buf());
4239 }
4240 }
4241 }
4242
4243 let parent_dir_is_loaded = relative_path.parent().is_none_or(|parent| {
4244 snapshot
4245 .entry_for_path(parent)
4246 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4247 });
4248 if !parent_dir_is_loaded {
4249 log::debug!("ignoring event {relative_path:?} within unloaded directory");
4250 skip_ix(&mut ranges_to_drop, ix);
4251 continue;
4252 }
4253
4254 if self.settings.is_path_excluded(&relative_path) {
4255 if !is_git_related {
4256 log::debug!("ignoring FS event for excluded path {relative_path:?}");
4257 }
4258 skip_ix(&mut ranges_to_drop, ix);
4259 continue;
4260 }
4261
4262 relative_paths.push(EventRoot {
4263 path: relative_path.into_arc(),
4264 was_rescanned: matches!(event.kind, Some(fs::PathEventKind::Rescan)),
4265 });
4266 }
4267
4268 for range_to_drop in ranges_to_drop.into_iter().rev() {
4269 events.drain(range_to_drop);
4270 }
4271 }
4272
4273 if relative_paths.is_empty() && dot_git_abs_paths.is_empty() {
4274 return;
4275 }
4276
4277 if !work_dirs_needing_exclude_update.is_empty() {
4278 let mut state = self.state.lock().await;
4279 for work_dir_abs_path in work_dirs_needing_exclude_update {
4280 if let Some((_, needs_update)) = state
4281 .snapshot
4282 .repo_exclude_by_work_dir_abs_path
4283 .get_mut(&work_dir_abs_path)
4284 {
4285 *needs_update = true;
4286 }
4287 }
4288 }
4289
4290 self.state.lock().await.snapshot.scan_id += 1;
4291
4292 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4293 log::debug!(
4294 "received fs events {:?}",
4295 relative_paths
4296 .iter()
4297 .map(|event_root| &event_root.path)
4298 .collect::<Vec<_>>()
4299 );
4300 self.reload_entries_for_paths(
4301 &root_path,
4302 &root_canonical_path,
4303 &relative_paths
4304 .iter()
4305 .map(|event_root| event_root.path.clone())
4306 .collect::<Vec<_>>(),
4307 events
4308 .into_iter()
4309 .map(|event| event.path)
4310 .collect::<Vec<_>>(),
4311 Some(scan_job_tx.clone()),
4312 )
4313 .await;
4314
4315 let affected_repo_roots = if !dot_git_abs_paths.is_empty() {
4316 self.update_git_repositories(dot_git_abs_paths).await
4317 } else {
4318 Vec::new()
4319 };
4320
4321 {
4322 let mut ignores_to_update = self.ignores_needing_update().await;
4323 ignores_to_update.extend(affected_repo_roots);
4324 let ignores_to_update = self.order_ignores(ignores_to_update).await;
4325 let snapshot = self.state.lock().await.snapshot.clone();
4326 self.update_ignore_statuses_for_paths(scan_job_tx, snapshot, ignores_to_update)
4327 .await;
4328 self.scan_dirs(false, scan_job_rx).await;
4329 }
4330
4331 {
4332 let mut state = self.state.lock().await;
4333 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4334 for (_, entry) in mem::take(&mut state.removed_entries) {
4335 state.scanned_dirs.remove(&entry.id);
4336 }
4337 }
4338 self.send_status_update(false, SmallVec::new(), &relative_paths)
4339 .await;
4340 }
4341
4342 async fn update_global_gitignore(&self, abs_path: &Path) {
4343 let ignore = build_gitignore(abs_path, self.fs.as_ref())
4344 .await
4345 .log_err()
4346 .map(Arc::new);
4347 let (prev_snapshot, ignore_stack, abs_path) = {
4348 let mut state = self.state.lock().await;
4349 state.snapshot.global_gitignore = ignore;
4350 let abs_path = state.snapshot.abs_path().clone();
4351 let ignore_stack = state
4352 .snapshot
4353 .ignore_stack_for_abs_path(&abs_path, true, self.fs.as_ref())
4354 .await;
4355 (state.snapshot.clone(), ignore_stack, abs_path)
4356 };
4357 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4358 self.update_ignore_statuses_for_paths(
4359 scan_job_tx,
4360 prev_snapshot,
4361 vec![(abs_path, ignore_stack)],
4362 )
4363 .await;
4364 self.scan_dirs(false, scan_job_rx).await;
4365 self.send_status_update(false, SmallVec::new(), &[]).await;
4366 }
4367
4368 async fn forcibly_load_paths(&self, paths: &[Arc<RelPath>]) -> bool {
4369 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4370 {
4371 let mut state = self.state.lock().await;
4372 let root_path = state.snapshot.abs_path.clone();
4373 for path in paths {
4374 for ancestor in path.ancestors() {
4375 if let Some(entry) = state.snapshot.entry_for_path(ancestor)
4376 && entry.kind == EntryKind::UnloadedDir
4377 {
4378 let abs_path = root_path.join(ancestor.as_std_path());
4379 state
4380 .enqueue_scan_dir(
4381 abs_path.into(),
4382 entry,
4383 &scan_job_tx,
4384 self.fs.as_ref(),
4385 )
4386 .await;
4387 state.paths_to_scan.insert(path.clone());
4388 break;
4389 }
4390 }
4391 }
4392 drop(scan_job_tx);
4393 }
4394 while let Ok(job) = scan_job_rx.recv().await {
4395 self.scan_dir(&job).await.log_err();
4396 }
4397
4398 !mem::take(&mut self.state.lock().await.paths_to_scan).is_empty()
4399 }
4400
4401 async fn scan_dirs(
4402 &self,
4403 enable_progress_updates: bool,
4404 scan_jobs_rx: channel::Receiver<ScanJob>,
4405 ) {
4406 if self
4407 .status_updates_tx
4408 .unbounded_send(ScanState::Started)
4409 .is_err()
4410 {
4411 return;
4412 }
4413
4414 let progress_update_count = AtomicUsize::new(0);
4415 self.executor
4416 .scoped_priority(Priority::Low, |scope| {
4417 for _ in 0..self.executor.num_cpus() {
4418 scope.spawn(async {
4419 let mut last_progress_update_count = 0;
4420 let progress_update_timer = self.progress_timer(enable_progress_updates).fuse();
4421 futures::pin_mut!(progress_update_timer);
4422
4423 loop {
4424 select_biased! {
4425 // Process any path refresh requests before moving on to process
4426 // the scan queue, so that user operations are prioritized.
4427 request = self.next_scan_request().fuse() => {
4428 let Ok(request) = request else { break };
4429 if !self.process_scan_request(request, true).await {
4430 return;
4431 }
4432 }
4433
4434 // Send periodic progress updates to the worktree. Use an atomic counter
4435 // to ensure that only one of the workers sends a progress update after
4436 // the update interval elapses.
4437 _ = progress_update_timer => {
4438 match progress_update_count.compare_exchange(
4439 last_progress_update_count,
4440 last_progress_update_count + 1,
4441 SeqCst,
4442 SeqCst
4443 ) {
4444 Ok(_) => {
4445 last_progress_update_count += 1;
4446 self.send_status_update(true, SmallVec::new(), &[])
4447 .await;
4448 }
4449 Err(count) => {
4450 last_progress_update_count = count;
4451 }
4452 }
4453 progress_update_timer.set(self.progress_timer(enable_progress_updates).fuse());
4454 }
4455
4456 // Recursively load directories from the file system.
4457 job = scan_jobs_rx.recv().fuse() => {
4458 let Ok(job) = job else { break };
4459 if let Err(err) = self.scan_dir(&job).await
4460 && job.path.is_empty() {
4461 log::error!("error scanning directory {:?}: {}", job.abs_path, err);
4462 }
4463 }
4464 }
4465 }
4466 });
4467 }
4468 })
4469 .await;
4470 }
4471
4472 async fn send_status_update(
4473 &self,
4474 scanning: bool,
4475 barrier: SmallVec<[barrier::Sender; 1]>,
4476 event_roots: &[EventRoot],
4477 ) -> bool {
4478 let mut state = self.state.lock().await;
4479 if state.changed_paths.is_empty() && event_roots.is_empty() && scanning {
4480 return true;
4481 }
4482
4483 let merged_event_roots = merge_event_roots(&state.changed_paths, event_roots);
4484
4485 let new_snapshot = state.snapshot.clone();
4486 let old_snapshot = mem::replace(&mut state.prev_snapshot, new_snapshot.snapshot.clone());
4487 let changes = build_diff(
4488 self.phase,
4489 &old_snapshot,
4490 &new_snapshot,
4491 &merged_event_roots,
4492 );
4493 state.changed_paths.clear();
4494
4495 self.status_updates_tx
4496 .unbounded_send(ScanState::Updated {
4497 snapshot: new_snapshot,
4498 changes,
4499 scanning,
4500 barrier,
4501 })
4502 .is_ok()
4503 }
4504
4505 async fn scan_dir(&self, job: &ScanJob) -> Result<()> {
4506 let root_abs_path;
4507 let root_char_bag;
4508 {
4509 let snapshot = &self.state.lock().await.snapshot;
4510 if self.settings.is_path_excluded(&job.path) {
4511 log::error!("skipping excluded directory {:?}", job.path);
4512 return Ok(());
4513 }
4514 log::trace!("scanning directory {:?}", job.path);
4515 root_abs_path = snapshot.abs_path().clone();
4516 root_char_bag = snapshot.root_char_bag;
4517 }
4518
4519 let next_entry_id = self.next_entry_id.clone();
4520 let mut ignore_stack = job.ignore_stack.clone();
4521 let mut new_ignore = None;
4522 let mut root_canonical_path = None;
4523 let mut new_entries: Vec<Entry> = Vec::new();
4524 let mut new_jobs: Vec<Option<ScanJob>> = Vec::new();
4525 let mut child_paths = self
4526 .fs
4527 .read_dir(&job.abs_path)
4528 .await?
4529 .filter_map(|entry| async {
4530 match entry {
4531 Ok(entry) => Some(entry),
4532 Err(error) => {
4533 log::error!("error processing entry {:?}", error);
4534 None
4535 }
4536 }
4537 })
4538 .collect::<Vec<_>>()
4539 .await;
4540
4541 // Ensure that .git and .gitignore are processed first.
4542 swap_to_front(&mut child_paths, GITIGNORE);
4543 swap_to_front(&mut child_paths, DOT_GIT);
4544
4545 if let Some(path) = child_paths.first()
4546 && path.ends_with(DOT_GIT)
4547 {
4548 ignore_stack.repo_root = Some(job.abs_path.clone());
4549 }
4550
4551 for child_abs_path in child_paths {
4552 let child_abs_path: Arc<Path> = child_abs_path.into();
4553 let child_name = child_abs_path.file_name().unwrap();
4554 let Some(child_path) = child_name
4555 .to_str()
4556 .and_then(|name| Some(job.path.join(RelPath::unix(name).ok()?)))
4557 else {
4558 continue;
4559 };
4560
4561 if child_name == DOT_GIT {
4562 let mut state = self.state.lock().await;
4563 state
4564 .insert_git_repository(
4565 child_path.clone(),
4566 self.fs.as_ref(),
4567 self.watcher.as_ref(),
4568 )
4569 .await;
4570 } else if child_name == GITIGNORE {
4571 match build_gitignore(&child_abs_path, self.fs.as_ref()).await {
4572 Ok(ignore) => {
4573 let ignore = Arc::new(ignore);
4574 ignore_stack = ignore_stack
4575 .append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4576 new_ignore = Some(ignore);
4577 }
4578 Err(error) => {
4579 log::error!(
4580 "error loading .gitignore file {:?} - {:?}",
4581 child_name,
4582 error
4583 );
4584 }
4585 }
4586 }
4587
4588 if self.settings.is_path_excluded(&child_path) {
4589 log::debug!("skipping excluded child entry {child_path:?}");
4590 self.state
4591 .lock()
4592 .await
4593 .remove_path(&child_path, self.watcher.as_ref());
4594 continue;
4595 }
4596
4597 let child_metadata = match self.fs.metadata(&child_abs_path).await {
4598 Ok(Some(metadata)) => metadata,
4599 Ok(None) => continue,
4600 Err(err) => {
4601 log::error!("error processing {:?}: {err:#}", child_abs_path.display());
4602 continue;
4603 }
4604 };
4605
4606 let mut child_entry = Entry::new(
4607 child_path.clone(),
4608 &child_metadata,
4609 ProjectEntryId::new(&next_entry_id),
4610 root_char_bag,
4611 None,
4612 );
4613
4614 if job.is_external {
4615 child_entry.is_external = true;
4616 } else if child_metadata.is_symlink {
4617 let canonical_path = match self.fs.canonicalize(&child_abs_path).await {
4618 Ok(path) => path,
4619 Err(err) => {
4620 log::error!("error reading target of symlink {child_abs_path:?}: {err:#}",);
4621 continue;
4622 }
4623 };
4624
4625 // lazily canonicalize the root path in order to determine if
4626 // symlinks point outside of the worktree.
4627 let root_canonical_path = match &root_canonical_path {
4628 Some(path) => path,
4629 None => match self.fs.canonicalize(&root_abs_path).await {
4630 Ok(path) => root_canonical_path.insert(path),
4631 Err(err) => {
4632 log::error!("error canonicalizing root {:?}: {:?}", root_abs_path, err);
4633 continue;
4634 }
4635 },
4636 };
4637
4638 if !canonical_path.starts_with(root_canonical_path) {
4639 child_entry.is_external = true;
4640 }
4641
4642 child_entry.canonical_path = Some(canonical_path.into());
4643 }
4644
4645 if child_entry.is_dir() {
4646 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, true);
4647 child_entry.is_always_included =
4648 self.settings.is_path_always_included(&child_path, true);
4649
4650 // Avoid recursing until crash in the case of a recursive symlink
4651 if job.ancestor_inodes.contains(&child_entry.inode) {
4652 new_jobs.push(None);
4653 } else {
4654 let mut ancestor_inodes = job.ancestor_inodes.clone();
4655 ancestor_inodes.insert(child_entry.inode);
4656
4657 new_jobs.push(Some(ScanJob {
4658 abs_path: child_abs_path.clone(),
4659 path: child_path,
4660 is_external: child_entry.is_external,
4661 ignore_stack: if child_entry.is_ignored {
4662 IgnoreStack::all()
4663 } else {
4664 ignore_stack.clone()
4665 },
4666 ancestor_inodes,
4667 scan_queue: job.scan_queue.clone(),
4668 }));
4669 }
4670 } else {
4671 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, false);
4672 child_entry.is_always_included =
4673 self.settings.is_path_always_included(&child_path, false);
4674 }
4675
4676 {
4677 let relative_path = job
4678 .path
4679 .join(RelPath::unix(child_name.to_str().unwrap()).unwrap());
4680 if self.is_path_private(&relative_path) {
4681 log::debug!("detected private file: {relative_path:?}");
4682 child_entry.is_private = true;
4683 }
4684 if self.settings.is_path_hidden(&relative_path) {
4685 log::debug!("detected hidden file: {relative_path:?}");
4686 child_entry.is_hidden = true;
4687 }
4688 }
4689
4690 new_entries.push(child_entry);
4691 }
4692
4693 let mut state = self.state.lock().await;
4694
4695 // Identify any subdirectories that should not be scanned.
4696 let mut job_ix = 0;
4697 for entry in &mut new_entries {
4698 state.reuse_entry_id(entry);
4699 if entry.is_dir() {
4700 if state.should_scan_directory(entry) {
4701 job_ix += 1;
4702 } else {
4703 log::debug!("defer scanning directory {:?}", entry.path);
4704 entry.kind = EntryKind::UnloadedDir;
4705 new_jobs.remove(job_ix);
4706 }
4707 }
4708 if entry.is_always_included {
4709 state
4710 .snapshot
4711 .always_included_entries
4712 .push(entry.path.clone());
4713 }
4714 }
4715
4716 state.populate_dir(job.path.clone(), new_entries, new_ignore);
4717 self.watcher.add(job.abs_path.as_ref()).log_err();
4718
4719 for new_job in new_jobs.into_iter().flatten() {
4720 job.scan_queue
4721 .try_send(new_job)
4722 .expect("channel is unbounded");
4723 }
4724
4725 Ok(())
4726 }
4727
4728 /// All list arguments should be sorted before calling this function
4729 async fn reload_entries_for_paths(
4730 &self,
4731 root_abs_path: &SanitizedPath,
4732 root_canonical_path: &SanitizedPath,
4733 relative_paths: &[Arc<RelPath>],
4734 abs_paths: Vec<PathBuf>,
4735 scan_queue_tx: Option<Sender<ScanJob>>,
4736 ) {
4737 // grab metadata for all requested paths
4738 let metadata = futures::future::join_all(
4739 abs_paths
4740 .iter()
4741 .map(|abs_path| async move {
4742 let metadata = self.fs.metadata(abs_path).await?;
4743 if let Some(metadata) = metadata {
4744 let canonical_path = self.fs.canonicalize(abs_path).await?;
4745
4746 // If we're on a case-insensitive filesystem (default on macOS), we want
4747 // to only ignore metadata for non-symlink files if their absolute-path matches
4748 // the canonical-path.
4749 // Because if not, this might be a case-only-renaming (`mv test.txt TEST.TXT`)
4750 // and we want to ignore the metadata for the old path (`test.txt`) so it's
4751 // treated as removed.
4752 if !self.fs_case_sensitive && !metadata.is_symlink {
4753 let canonical_file_name = canonical_path.file_name();
4754 let file_name = abs_path.file_name();
4755 if canonical_file_name != file_name {
4756 return Ok(None);
4757 }
4758 }
4759
4760 anyhow::Ok(Some((metadata, SanitizedPath::new_arc(&canonical_path))))
4761 } else {
4762 Ok(None)
4763 }
4764 })
4765 .collect::<Vec<_>>(),
4766 )
4767 .await;
4768
4769 let mut new_ancestor_repo = if relative_paths.iter().any(|path| path.is_empty()) {
4770 Some(discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await)
4771 } else {
4772 None
4773 };
4774
4775 let mut state = self.state.lock().await;
4776 let doing_recursive_update = scan_queue_tx.is_some();
4777
4778 // Remove any entries for paths that no longer exist or are being recursively
4779 // refreshed. Do this before adding any new entries, so that renames can be
4780 // detected regardless of the order of the paths.
4781 for (path, metadata) in relative_paths.iter().zip(metadata.iter()) {
4782 if matches!(metadata, Ok(None)) || doing_recursive_update {
4783 state.remove_path(path, self.watcher.as_ref());
4784 }
4785 }
4786
4787 for (path, metadata) in relative_paths.iter().zip(metadata.into_iter()) {
4788 let abs_path: Arc<Path> = root_abs_path.join(path.as_std_path()).into();
4789 match metadata {
4790 Ok(Some((metadata, canonical_path))) => {
4791 let ignore_stack = state
4792 .snapshot
4793 .ignore_stack_for_abs_path(&abs_path, metadata.is_dir, self.fs.as_ref())
4794 .await;
4795 let is_external = !canonical_path.starts_with(&root_canonical_path);
4796 let entry_id = state.entry_id_for(self.next_entry_id.as_ref(), path, &metadata);
4797 let mut fs_entry = Entry::new(
4798 path.clone(),
4799 &metadata,
4800 entry_id,
4801 state.snapshot.root_char_bag,
4802 if metadata.is_symlink {
4803 Some(canonical_path.as_path().to_path_buf().into())
4804 } else {
4805 None
4806 },
4807 );
4808
4809 let is_dir = fs_entry.is_dir();
4810 fs_entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, is_dir);
4811 fs_entry.is_external = is_external;
4812 fs_entry.is_private = self.is_path_private(path);
4813 fs_entry.is_always_included =
4814 self.settings.is_path_always_included(path, is_dir);
4815 fs_entry.is_hidden = self.settings.is_path_hidden(path);
4816
4817 if let (Some(scan_queue_tx), true) = (&scan_queue_tx, is_dir) {
4818 if state.should_scan_directory(&fs_entry)
4819 || (fs_entry.path.is_empty()
4820 && abs_path.file_name() == Some(OsStr::new(DOT_GIT)))
4821 {
4822 state
4823 .enqueue_scan_dir(
4824 abs_path,
4825 &fs_entry,
4826 scan_queue_tx,
4827 self.fs.as_ref(),
4828 )
4829 .await;
4830 } else {
4831 fs_entry.kind = EntryKind::UnloadedDir;
4832 }
4833 }
4834
4835 state
4836 .insert_entry(fs_entry.clone(), self.fs.as_ref(), self.watcher.as_ref())
4837 .await;
4838
4839 if path.is_empty()
4840 && let Some((ignores, exclude, repo)) = new_ancestor_repo.take()
4841 {
4842 log::trace!("updating ancestor git repository");
4843 state.snapshot.ignores_by_parent_abs_path.extend(ignores);
4844 if let Some((ancestor_dot_git, work_directory)) = repo {
4845 if let Some(exclude) = exclude {
4846 let work_directory_abs_path = self
4847 .state
4848 .lock()
4849 .await
4850 .snapshot
4851 .work_directory_abs_path(&work_directory);
4852
4853 state
4854 .snapshot
4855 .repo_exclude_by_work_dir_abs_path
4856 .insert(work_directory_abs_path.into(), (exclude, false));
4857 }
4858 state
4859 .insert_git_repository_for_path(
4860 work_directory,
4861 ancestor_dot_git.into(),
4862 self.fs.as_ref(),
4863 self.watcher.as_ref(),
4864 )
4865 .await
4866 .log_err();
4867 }
4868 }
4869 }
4870 Ok(None) => {
4871 self.remove_repo_path(path.clone(), &mut state.snapshot);
4872 }
4873 Err(err) => {
4874 log::error!("error reading file {abs_path:?} on event: {err:#}");
4875 }
4876 }
4877 }
4878
4879 util::extend_sorted(
4880 &mut state.changed_paths,
4881 relative_paths.iter().cloned(),
4882 usize::MAX,
4883 Ord::cmp,
4884 );
4885 }
4886
4887 fn remove_repo_path(&self, path: Arc<RelPath>, snapshot: &mut LocalSnapshot) -> Option<()> {
4888 if !path.components().any(|component| component == DOT_GIT)
4889 && let Some(local_repo) = snapshot.local_repo_for_work_directory_path(&path)
4890 {
4891 let id = local_repo.work_directory_id;
4892 log::debug!("remove repo path: {:?}", path);
4893 snapshot.git_repositories.remove(&id);
4894 return Some(());
4895 }
4896
4897 Some(())
4898 }
4899
4900 async fn update_ignore_statuses_for_paths(
4901 &self,
4902 scan_job_tx: Sender<ScanJob>,
4903 prev_snapshot: LocalSnapshot,
4904 ignores_to_update: Vec<(Arc<Path>, IgnoreStack)>,
4905 ) {
4906 let (ignore_queue_tx, ignore_queue_rx) = channel::unbounded();
4907 {
4908 for (parent_abs_path, ignore_stack) in ignores_to_update {
4909 ignore_queue_tx
4910 .send_blocking(UpdateIgnoreStatusJob {
4911 abs_path: parent_abs_path,
4912 ignore_stack,
4913 ignore_queue: ignore_queue_tx.clone(),
4914 scan_queue: scan_job_tx.clone(),
4915 })
4916 .unwrap();
4917 }
4918 }
4919 drop(ignore_queue_tx);
4920
4921 self.executor
4922 .scoped(|scope| {
4923 for _ in 0..self.executor.num_cpus() {
4924 scope.spawn(async {
4925 loop {
4926 select_biased! {
4927 // Process any path refresh requests before moving on to process
4928 // the queue of ignore statuses.
4929 request = self.next_scan_request().fuse() => {
4930 let Ok(request) = request else { break };
4931 if !self.process_scan_request(request, true).await {
4932 return;
4933 }
4934 }
4935
4936 // Recursively process directories whose ignores have changed.
4937 job = ignore_queue_rx.recv().fuse() => {
4938 let Ok(job) = job else { break };
4939 self.update_ignore_status(job, &prev_snapshot).await;
4940 }
4941 }
4942 }
4943 });
4944 }
4945 })
4946 .await;
4947 }
4948
4949 async fn ignores_needing_update(&self) -> Vec<Arc<Path>> {
4950 let mut ignores_to_update = Vec::new();
4951 let mut excludes_to_load: Vec<(Arc<Path>, PathBuf)> = Vec::new();
4952
4953 // First pass: collect updates and drop stale entries without awaiting.
4954 {
4955 let snapshot = &mut self.state.lock().await.snapshot;
4956 let abs_path = snapshot.abs_path.clone();
4957 let mut repo_exclude_keys_to_remove: Vec<Arc<Path>> = Vec::new();
4958
4959 for (work_dir_abs_path, (_, needs_update)) in
4960 snapshot.repo_exclude_by_work_dir_abs_path.iter_mut()
4961 {
4962 let repository = snapshot
4963 .git_repositories
4964 .iter()
4965 .find(|(_, repo)| &repo.work_directory_abs_path == work_dir_abs_path);
4966
4967 if *needs_update {
4968 *needs_update = false;
4969 ignores_to_update.push(work_dir_abs_path.clone());
4970
4971 if let Some((_, repository)) = repository {
4972 let exclude_abs_path = repository.common_dir_abs_path.join(REPO_EXCLUDE);
4973 excludes_to_load.push((work_dir_abs_path.clone(), exclude_abs_path));
4974 }
4975 }
4976
4977 if repository.is_none() {
4978 repo_exclude_keys_to_remove.push(work_dir_abs_path.clone());
4979 }
4980 }
4981
4982 for key in repo_exclude_keys_to_remove {
4983 snapshot.repo_exclude_by_work_dir_abs_path.remove(&key);
4984 }
4985
4986 snapshot
4987 .ignores_by_parent_abs_path
4988 .retain(|parent_abs_path, (_, needs_update)| {
4989 if let Ok(parent_path) = parent_abs_path.strip_prefix(abs_path.as_path())
4990 && let Some(parent_path) =
4991 RelPath::new(&parent_path, PathStyle::local()).log_err()
4992 {
4993 if *needs_update {
4994 *needs_update = false;
4995 if snapshot.snapshot.entry_for_path(&parent_path).is_some() {
4996 ignores_to_update.push(parent_abs_path.clone());
4997 }
4998 }
4999
5000 let ignore_path = parent_path.join(RelPath::unix(GITIGNORE).unwrap());
5001 if snapshot.snapshot.entry_for_path(&ignore_path).is_none() {
5002 return false;
5003 }
5004 }
5005 true
5006 });
5007 }
5008
5009 // Load gitignores asynchronously (outside the lock)
5010 let mut loaded_excludes: Vec<(Arc<Path>, Arc<Gitignore>)> = Vec::new();
5011 for (work_dir_abs_path, exclude_abs_path) in excludes_to_load {
5012 if let Ok(current_exclude) = build_gitignore(&exclude_abs_path, self.fs.as_ref()).await
5013 {
5014 loaded_excludes.push((work_dir_abs_path, Arc::new(current_exclude)));
5015 }
5016 }
5017
5018 // Second pass: apply updates.
5019 if !loaded_excludes.is_empty() {
5020 let snapshot = &mut self.state.lock().await.snapshot;
5021
5022 for (work_dir_abs_path, exclude) in loaded_excludes {
5023 if let Some((existing_exclude, _)) = snapshot
5024 .repo_exclude_by_work_dir_abs_path
5025 .get_mut(&work_dir_abs_path)
5026 {
5027 *existing_exclude = exclude;
5028 }
5029 }
5030 }
5031
5032 ignores_to_update
5033 }
5034
5035 async fn order_ignores(&self, mut ignores: Vec<Arc<Path>>) -> Vec<(Arc<Path>, IgnoreStack)> {
5036 let fs = self.fs.clone();
5037 let snapshot = self.state.lock().await.snapshot.clone();
5038 ignores.sort_unstable();
5039 let mut ignores_to_update = ignores.into_iter().peekable();
5040
5041 let mut result = vec![];
5042 while let Some(parent_abs_path) = ignores_to_update.next() {
5043 while ignores_to_update
5044 .peek()
5045 .map_or(false, |p| p.starts_with(&parent_abs_path))
5046 {
5047 ignores_to_update.next().unwrap();
5048 }
5049 let ignore_stack = snapshot
5050 .ignore_stack_for_abs_path(&parent_abs_path, true, fs.as_ref())
5051 .await;
5052 result.push((parent_abs_path, ignore_stack));
5053 }
5054
5055 result
5056 }
5057
5058 async fn update_ignore_status(&self, job: UpdateIgnoreStatusJob, snapshot: &LocalSnapshot) {
5059 log::trace!("update ignore status {:?}", job.abs_path);
5060
5061 let mut ignore_stack = job.ignore_stack;
5062 if let Some((ignore, _)) = snapshot.ignores_by_parent_abs_path.get(&job.abs_path) {
5063 ignore_stack =
5064 ignore_stack.append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
5065 }
5066
5067 let mut entries_by_id_edits = Vec::new();
5068 let mut entries_by_path_edits = Vec::new();
5069 let Some(path) = job
5070 .abs_path
5071 .strip_prefix(snapshot.abs_path.as_path())
5072 .map_err(|_| {
5073 anyhow::anyhow!(
5074 "Failed to strip prefix '{}' from path '{}'",
5075 snapshot.abs_path.as_path().display(),
5076 job.abs_path.display()
5077 )
5078 })
5079 .log_err()
5080 else {
5081 return;
5082 };
5083
5084 let Some(path) = RelPath::new(&path, PathStyle::local()).log_err() else {
5085 return;
5086 };
5087
5088 if let Ok(Some(metadata)) = self.fs.metadata(&job.abs_path.join(DOT_GIT)).await
5089 && metadata.is_dir
5090 {
5091 ignore_stack.repo_root = Some(job.abs_path.clone());
5092 }
5093
5094 for mut entry in snapshot.child_entries(&path).cloned() {
5095 let was_ignored = entry.is_ignored;
5096 let abs_path: Arc<Path> = snapshot.absolutize(&entry.path).into();
5097 entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, entry.is_dir());
5098
5099 if entry.is_dir() {
5100 let child_ignore_stack = if entry.is_ignored {
5101 IgnoreStack::all()
5102 } else {
5103 ignore_stack.clone()
5104 };
5105
5106 // Scan any directories that were previously ignored and weren't previously scanned.
5107 if was_ignored && !entry.is_ignored && entry.kind.is_unloaded() {
5108 let state = self.state.lock().await;
5109 if state.should_scan_directory(&entry) {
5110 state
5111 .enqueue_scan_dir(
5112 abs_path.clone(),
5113 &entry,
5114 &job.scan_queue,
5115 self.fs.as_ref(),
5116 )
5117 .await;
5118 }
5119 }
5120
5121 job.ignore_queue
5122 .send(UpdateIgnoreStatusJob {
5123 abs_path: abs_path.clone(),
5124 ignore_stack: child_ignore_stack,
5125 ignore_queue: job.ignore_queue.clone(),
5126 scan_queue: job.scan_queue.clone(),
5127 })
5128 .await
5129 .unwrap();
5130 }
5131
5132 if entry.is_ignored != was_ignored {
5133 let mut path_entry = snapshot.entries_by_id.get(&entry.id, ()).unwrap().clone();
5134 path_entry.scan_id = snapshot.scan_id;
5135 path_entry.is_ignored = entry.is_ignored;
5136 entries_by_id_edits.push(Edit::Insert(path_entry));
5137 entries_by_path_edits.push(Edit::Insert(entry));
5138 }
5139 }
5140
5141 let state = &mut self.state.lock().await;
5142 for edit in &entries_by_path_edits {
5143 if let Edit::Insert(entry) = edit
5144 && let Err(ix) = state.changed_paths.binary_search(&entry.path)
5145 {
5146 state.changed_paths.insert(ix, entry.path.clone());
5147 }
5148 }
5149
5150 state
5151 .snapshot
5152 .entries_by_path
5153 .edit(entries_by_path_edits, ());
5154 state.snapshot.entries_by_id.edit(entries_by_id_edits, ());
5155 }
5156
5157 async fn update_git_repositories(&self, dot_git_paths: Vec<PathBuf>) -> Vec<Arc<Path>> {
5158 log::trace!("reloading repositories: {dot_git_paths:?}");
5159 let mut state = self.state.lock().await;
5160 let scan_id = state.snapshot.scan_id;
5161 let mut affected_repo_roots = Vec::new();
5162 for dot_git_dir in dot_git_paths {
5163 let existing_repository_entry =
5164 state
5165 .snapshot
5166 .git_repositories
5167 .iter()
5168 .find_map(|(_, repo)| {
5169 let dot_git_dir = SanitizedPath::new(&dot_git_dir);
5170 if SanitizedPath::new(repo.common_dir_abs_path.as_ref()) == dot_git_dir
5171 || SanitizedPath::new(repo.repository_dir_abs_path.as_ref())
5172 == dot_git_dir
5173 {
5174 Some(repo.clone())
5175 } else {
5176 None
5177 }
5178 });
5179
5180 match existing_repository_entry {
5181 None => {
5182 let Ok(relative) = dot_git_dir.strip_prefix(state.snapshot.abs_path()) else {
5183 debug_panic!(
5184 "update_git_repositories called with .git directory outside the worktree root"
5185 );
5186 return Vec::new();
5187 };
5188 affected_repo_roots.push(dot_git_dir.parent().unwrap().into());
5189 state
5190 .insert_git_repository(
5191 RelPath::new(relative, PathStyle::local())
5192 .unwrap()
5193 .into_arc(),
5194 self.fs.as_ref(),
5195 self.watcher.as_ref(),
5196 )
5197 .await;
5198 }
5199 Some(local_repository) => {
5200 state.snapshot.git_repositories.update(
5201 &local_repository.work_directory_id,
5202 |entry| {
5203 entry.git_dir_scan_id = scan_id;
5204 },
5205 );
5206 }
5207 };
5208 }
5209
5210 // Remove any git repositories whose .git entry no longer exists.
5211 let snapshot = &mut state.snapshot;
5212 let mut ids_to_preserve = HashSet::default();
5213 for (&work_directory_id, entry) in snapshot.git_repositories.iter() {
5214 let exists_in_snapshot =
5215 snapshot
5216 .entry_for_id(work_directory_id)
5217 .is_some_and(|entry| {
5218 snapshot
5219 .entry_for_path(&entry.path.join(RelPath::unix(DOT_GIT).unwrap()))
5220 .is_some()
5221 });
5222
5223 if exists_in_snapshot
5224 || matches!(
5225 self.fs.metadata(&entry.common_dir_abs_path).await,
5226 Ok(Some(_))
5227 )
5228 {
5229 ids_to_preserve.insert(work_directory_id);
5230 }
5231 }
5232
5233 snapshot
5234 .git_repositories
5235 .retain(|work_directory_id, entry| {
5236 let preserve = ids_to_preserve.contains(work_directory_id);
5237 if !preserve {
5238 affected_repo_roots.push(entry.dot_git_abs_path.parent().unwrap().into());
5239 snapshot
5240 .repo_exclude_by_work_dir_abs_path
5241 .remove(&entry.work_directory_abs_path);
5242 }
5243 preserve
5244 });
5245
5246 affected_repo_roots
5247 }
5248
5249 async fn progress_timer(&self, running: bool) {
5250 if !running {
5251 return futures::future::pending().await;
5252 }
5253
5254 #[cfg(feature = "test-support")]
5255 if self.fs.is_fake() {
5256 return self.executor.simulate_random_delay().await;
5257 }
5258
5259 self.executor.timer(FS_WATCH_LATENCY).await
5260 }
5261
5262 fn is_path_private(&self, path: &RelPath) -> bool {
5263 !self.share_private_files && self.settings.is_path_private(path)
5264 }
5265
5266 async fn next_scan_request(&self) -> Result<ScanRequest> {
5267 let mut request = self.scan_requests_rx.recv().await?;
5268 while let Ok(next_request) = self.scan_requests_rx.try_recv() {
5269 request.relative_paths.extend(next_request.relative_paths);
5270 request.done.extend(next_request.done);
5271 }
5272 Ok(request)
5273 }
5274}
5275
5276async fn discover_ancestor_git_repo(
5277 fs: Arc<dyn Fs>,
5278 root_abs_path: &SanitizedPath,
5279) -> (
5280 HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
5281 Option<Arc<Gitignore>>,
5282 Option<(PathBuf, WorkDirectory)>,
5283) {
5284 let mut exclude = None;
5285 let mut ignores = HashMap::default();
5286 for (index, ancestor) in root_abs_path.as_path().ancestors().enumerate() {
5287 if index != 0 {
5288 if ancestor == paths::home_dir() {
5289 // Unless $HOME is itself the worktree root, don't consider it as a
5290 // containing git repository---expensive and likely unwanted.
5291 break;
5292 } else if let Ok(ignore) = build_gitignore(&ancestor.join(GITIGNORE), fs.as_ref()).await
5293 {
5294 ignores.insert(ancestor.into(), (ignore.into(), false));
5295 }
5296 }
5297
5298 let ancestor_dot_git = ancestor.join(DOT_GIT);
5299 log::trace!("considering ancestor: {ancestor_dot_git:?}");
5300 // Check whether the directory or file called `.git` exists (in the
5301 // case of worktrees it's a file.)
5302 if fs
5303 .metadata(&ancestor_dot_git)
5304 .await
5305 .is_ok_and(|metadata| metadata.is_some())
5306 {
5307 if index != 0 {
5308 // We canonicalize, since the FS events use the canonicalized path.
5309 if let Some(ancestor_dot_git) = fs.canonicalize(&ancestor_dot_git).await.log_err() {
5310 let location_in_repo = root_abs_path
5311 .as_path()
5312 .strip_prefix(ancestor)
5313 .unwrap()
5314 .into();
5315 log::info!("inserting parent git repo for this worktree: {location_in_repo:?}");
5316 // We associate the external git repo with our root folder and
5317 // also mark where in the git repo the root folder is located.
5318 return (
5319 ignores,
5320 exclude,
5321 Some((
5322 ancestor_dot_git,
5323 WorkDirectory::AboveProject {
5324 absolute_path: ancestor.into(),
5325 location_in_repo,
5326 },
5327 )),
5328 );
5329 };
5330 }
5331
5332 let repo_exclude_abs_path = ancestor_dot_git.join(REPO_EXCLUDE);
5333 if let Ok(repo_exclude) = build_gitignore(&repo_exclude_abs_path, fs.as_ref()).await {
5334 exclude = Some(Arc::new(repo_exclude));
5335 }
5336
5337 // Reached root of git repository.
5338 break;
5339 }
5340 }
5341
5342 (ignores, exclude, None)
5343}
5344
5345fn merge_event_roots(changed_paths: &[Arc<RelPath>], event_roots: &[EventRoot]) -> Vec<EventRoot> {
5346 let mut merged_event_roots = Vec::with_capacity(changed_paths.len() + event_roots.len());
5347 let mut changed_paths = changed_paths.iter().peekable();
5348 let mut event_roots = event_roots.iter().peekable();
5349 while let (Some(path), Some(event_root)) = (changed_paths.peek(), event_roots.peek()) {
5350 match path.cmp(&&event_root.path) {
5351 Ordering::Less => {
5352 merged_event_roots.push(EventRoot {
5353 path: (*changed_paths.next().expect("peeked changed path")).clone(),
5354 was_rescanned: false,
5355 });
5356 }
5357 Ordering::Equal => {
5358 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5359 changed_paths.next();
5360 }
5361 Ordering::Greater => {
5362 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5363 }
5364 }
5365 }
5366 merged_event_roots.extend(changed_paths.map(|path| EventRoot {
5367 path: path.clone(),
5368 was_rescanned: false,
5369 }));
5370 merged_event_roots.extend(event_roots.cloned());
5371 merged_event_roots
5372}
5373
5374fn build_diff(
5375 phase: BackgroundScannerPhase,
5376 old_snapshot: &Snapshot,
5377 new_snapshot: &Snapshot,
5378 event_roots: &[EventRoot],
5379) -> UpdatedEntriesSet {
5380 use BackgroundScannerPhase::*;
5381 use PathChange::{Added, AddedOrUpdated, Loaded, Removed, Updated};
5382
5383 // Identify which paths have changed. Use the known set of changed
5384 // parent paths to optimize the search.
5385 let mut changes = Vec::new();
5386
5387 let mut old_paths = old_snapshot.entries_by_path.cursor::<PathKey>(());
5388 let mut new_paths = new_snapshot.entries_by_path.cursor::<PathKey>(());
5389 let mut last_newly_loaded_dir_path = None;
5390 old_paths.next();
5391 new_paths.next();
5392 for event_root in event_roots {
5393 let path = PathKey(event_root.path.clone());
5394 if old_paths.item().is_some_and(|e| e.path < path.0) {
5395 old_paths.seek_forward(&path, Bias::Left);
5396 }
5397 if new_paths.item().is_some_and(|e| e.path < path.0) {
5398 new_paths.seek_forward(&path, Bias::Left);
5399 }
5400 loop {
5401 match (old_paths.item(), new_paths.item()) {
5402 (Some(old_entry), Some(new_entry)) => {
5403 if old_entry.path > path.0
5404 && new_entry.path > path.0
5405 && !old_entry.path.starts_with(&path.0)
5406 && !new_entry.path.starts_with(&path.0)
5407 {
5408 break;
5409 }
5410
5411 match Ord::cmp(&old_entry.path, &new_entry.path) {
5412 Ordering::Less => {
5413 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5414 old_paths.next();
5415 }
5416 Ordering::Equal => {
5417 if phase == EventsReceivedDuringInitialScan {
5418 if old_entry.id != new_entry.id {
5419 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5420 }
5421 // If the worktree was not fully initialized when this event was generated,
5422 // we can't know whether this entry was added during the scan or whether
5423 // it was merely updated.
5424 changes.push((
5425 new_entry.path.clone(),
5426 new_entry.id,
5427 AddedOrUpdated,
5428 ));
5429 } else if old_entry.id != new_entry.id {
5430 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5431 changes.push((new_entry.path.clone(), new_entry.id, Added));
5432 } else if old_entry != new_entry {
5433 if old_entry.kind.is_unloaded() {
5434 last_newly_loaded_dir_path = Some(&new_entry.path);
5435 changes.push((new_entry.path.clone(), new_entry.id, Loaded));
5436 } else {
5437 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5438 }
5439 } else if event_root.was_rescanned {
5440 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5441 }
5442 old_paths.next();
5443 new_paths.next();
5444 }
5445 Ordering::Greater => {
5446 let is_newly_loaded = phase == InitialScan
5447 || last_newly_loaded_dir_path
5448 .as_ref()
5449 .is_some_and(|dir| new_entry.path.starts_with(dir));
5450 changes.push((
5451 new_entry.path.clone(),
5452 new_entry.id,
5453 if is_newly_loaded { Loaded } else { Added },
5454 ));
5455 new_paths.next();
5456 }
5457 }
5458 }
5459 (Some(old_entry), None) => {
5460 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5461 old_paths.next();
5462 }
5463 (None, Some(new_entry)) => {
5464 let is_newly_loaded = phase == InitialScan
5465 || last_newly_loaded_dir_path
5466 .as_ref()
5467 .is_some_and(|dir| new_entry.path.starts_with(dir));
5468 changes.push((
5469 new_entry.path.clone(),
5470 new_entry.id,
5471 if is_newly_loaded { Loaded } else { Added },
5472 ));
5473 new_paths.next();
5474 }
5475 (None, None) => break,
5476 }
5477 }
5478 }
5479
5480 changes.into()
5481}
5482
5483fn swap_to_front(child_paths: &mut Vec<PathBuf>, file: &str) {
5484 let position = child_paths
5485 .iter()
5486 .position(|path| path.file_name().unwrap() == file);
5487 if let Some(position) = position {
5488 let temp = child_paths.remove(position);
5489 child_paths.insert(0, temp);
5490 }
5491}
5492
5493fn char_bag_for_path(root_char_bag: CharBag, path: &RelPath) -> CharBag {
5494 let mut result = root_char_bag;
5495 result.extend(path.as_unix_str().chars().map(|c| c.to_ascii_lowercase()));
5496 result
5497}
5498
5499#[derive(Debug)]
5500struct ScanJob {
5501 abs_path: Arc<Path>,
5502 path: Arc<RelPath>,
5503 ignore_stack: IgnoreStack,
5504 scan_queue: Sender<ScanJob>,
5505 ancestor_inodes: TreeSet<u64>,
5506 is_external: bool,
5507}
5508
5509struct UpdateIgnoreStatusJob {
5510 abs_path: Arc<Path>,
5511 ignore_stack: IgnoreStack,
5512 ignore_queue: Sender<UpdateIgnoreStatusJob>,
5513 scan_queue: Sender<ScanJob>,
5514}
5515
5516pub trait WorktreeModelHandle {
5517 #[cfg(feature = "test-support")]
5518 fn flush_fs_events<'a>(
5519 &self,
5520 cx: &'a mut gpui::TestAppContext,
5521 ) -> futures::future::LocalBoxFuture<'a, ()>;
5522
5523 #[cfg(feature = "test-support")]
5524 fn flush_fs_events_in_root_git_repository<'a>(
5525 &self,
5526 cx: &'a mut gpui::TestAppContext,
5527 ) -> futures::future::LocalBoxFuture<'a, ()>;
5528}
5529
5530impl WorktreeModelHandle for Entity<Worktree> {
5531 // When the worktree's FS event stream sometimes delivers "redundant" events for FS changes that
5532 // occurred before the worktree was constructed. These events can cause the worktree to perform
5533 // extra directory scans, and emit extra scan-state notifications.
5534 //
5535 // This function mutates the worktree's directory and waits for those mutations to be picked up,
5536 // to ensure that all redundant FS events have already been processed.
5537 #[cfg(feature = "test-support")]
5538 fn flush_fs_events<'a>(
5539 &self,
5540 cx: &'a mut gpui::TestAppContext,
5541 ) -> futures::future::LocalBoxFuture<'a, ()> {
5542 let file_name = "fs-event-sentinel";
5543
5544 let tree = self.clone();
5545 let (fs, root_path) = self.read_with(cx, |tree, _| {
5546 let tree = tree.as_local().unwrap();
5547 (tree.fs.clone(), tree.abs_path.clone())
5548 });
5549
5550 async move {
5551 // Subscribe to events BEFORE creating the file to avoid race condition
5552 // where events fire before subscription is set up
5553 let mut events = cx.events(&tree);
5554
5555 fs.create_file(&root_path.join(file_name), Default::default())
5556 .await
5557 .unwrap();
5558
5559 // Check if condition is already met before waiting for events
5560 let file_exists = || {
5561 tree.read_with(cx, |tree, _| {
5562 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5563 .is_some()
5564 })
5565 };
5566
5567 // Use select to avoid blocking indefinitely if events are delayed
5568 while !file_exists() {
5569 futures::select_biased! {
5570 _ = events.next() => {}
5571 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5572 }
5573 }
5574
5575 fs.remove_file(&root_path.join(file_name), Default::default())
5576 .await
5577 .unwrap();
5578
5579 // Check if condition is already met before waiting for events
5580 let file_gone = || {
5581 tree.read_with(cx, |tree, _| {
5582 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5583 .is_none()
5584 })
5585 };
5586
5587 // Use select to avoid blocking indefinitely if events are delayed
5588 while !file_gone() {
5589 futures::select_biased! {
5590 _ = events.next() => {}
5591 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5592 }
5593 }
5594
5595 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5596 .await;
5597 }
5598 .boxed_local()
5599 }
5600
5601 // This function is similar to flush_fs_events, except that it waits for events to be flushed in
5602 // the .git folder of the root repository.
5603 // The reason for its existence is that a repository's .git folder might live *outside* of the
5604 // worktree and thus its FS events might go through a different path.
5605 // In order to flush those, we need to create artificial events in the .git folder and wait
5606 // for the repository to be reloaded.
5607 #[cfg(feature = "test-support")]
5608 fn flush_fs_events_in_root_git_repository<'a>(
5609 &self,
5610 cx: &'a mut gpui::TestAppContext,
5611 ) -> futures::future::LocalBoxFuture<'a, ()> {
5612 let file_name = "fs-event-sentinel";
5613
5614 let tree = self.clone();
5615 let (fs, root_path, mut git_dir_scan_id) = self.read_with(cx, |tree, _| {
5616 let tree = tree.as_local().unwrap();
5617 let local_repo_entry = tree
5618 .git_repositories
5619 .values()
5620 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5621 .unwrap();
5622 (
5623 tree.fs.clone(),
5624 local_repo_entry.common_dir_abs_path.clone(),
5625 local_repo_entry.git_dir_scan_id,
5626 )
5627 });
5628
5629 let scan_id_increased = |tree: &mut Worktree, git_dir_scan_id: &mut usize| {
5630 let tree = tree.as_local().unwrap();
5631 // let repository = tree.repositories.first().unwrap();
5632 let local_repo_entry = tree
5633 .git_repositories
5634 .values()
5635 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5636 .unwrap();
5637
5638 if local_repo_entry.git_dir_scan_id > *git_dir_scan_id {
5639 *git_dir_scan_id = local_repo_entry.git_dir_scan_id;
5640 true
5641 } else {
5642 false
5643 }
5644 };
5645
5646 async move {
5647 // Subscribe to events BEFORE creating the file to avoid race condition
5648 // where events fire before subscription is set up
5649 let mut events = cx.events(&tree);
5650
5651 fs.create_file(&root_path.join(file_name), Default::default())
5652 .await
5653 .unwrap();
5654
5655 // Use select to avoid blocking indefinitely if events are delayed
5656 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5657 futures::select_biased! {
5658 _ = events.next() => {}
5659 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5660 }
5661 }
5662
5663 fs.remove_file(&root_path.join(file_name), Default::default())
5664 .await
5665 .unwrap();
5666
5667 // Use select to avoid blocking indefinitely if events are delayed
5668 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5669 futures::select_biased! {
5670 _ = events.next() => {}
5671 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5672 }
5673 }
5674
5675 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5676 .await;
5677 }
5678 .boxed_local()
5679 }
5680}
5681
5682#[derive(Clone, Debug)]
5683struct TraversalProgress<'a> {
5684 max_path: &'a RelPath,
5685 count: usize,
5686 non_ignored_count: usize,
5687 file_count: usize,
5688 non_ignored_file_count: usize,
5689}
5690
5691impl TraversalProgress<'_> {
5692 fn count(&self, include_files: bool, include_dirs: bool, include_ignored: bool) -> usize {
5693 match (include_files, include_dirs, include_ignored) {
5694 (true, true, true) => self.count,
5695 (true, true, false) => self.non_ignored_count,
5696 (true, false, true) => self.file_count,
5697 (true, false, false) => self.non_ignored_file_count,
5698 (false, true, true) => self.count - self.file_count,
5699 (false, true, false) => self.non_ignored_count - self.non_ignored_file_count,
5700 (false, false, _) => 0,
5701 }
5702 }
5703}
5704
5705impl<'a> sum_tree::Dimension<'a, EntrySummary> for TraversalProgress<'a> {
5706 fn zero(_cx: ()) -> Self {
5707 Default::default()
5708 }
5709
5710 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
5711 self.max_path = summary.max_path.as_ref();
5712 self.count += summary.count;
5713 self.non_ignored_count += summary.non_ignored_count;
5714 self.file_count += summary.file_count;
5715 self.non_ignored_file_count += summary.non_ignored_file_count;
5716 }
5717}
5718
5719impl Default for TraversalProgress<'_> {
5720 fn default() -> Self {
5721 Self {
5722 max_path: RelPath::empty(),
5723 count: 0,
5724 non_ignored_count: 0,
5725 file_count: 0,
5726 non_ignored_file_count: 0,
5727 }
5728 }
5729}
5730
5731#[derive(Debug)]
5732pub struct Traversal<'a> {
5733 snapshot: &'a Snapshot,
5734 cursor: sum_tree::Cursor<'a, 'static, Entry, TraversalProgress<'a>>,
5735 include_ignored: bool,
5736 include_files: bool,
5737 include_dirs: bool,
5738}
5739
5740impl<'a> Traversal<'a> {
5741 fn new(
5742 snapshot: &'a Snapshot,
5743 include_files: bool,
5744 include_dirs: bool,
5745 include_ignored: bool,
5746 start_path: &RelPath,
5747 ) -> Self {
5748 let mut cursor = snapshot.entries_by_path.cursor(());
5749 cursor.seek(&TraversalTarget::path(start_path), Bias::Left);
5750 let mut traversal = Self {
5751 snapshot,
5752 cursor,
5753 include_files,
5754 include_dirs,
5755 include_ignored,
5756 };
5757 if traversal.end_offset() == traversal.start_offset() {
5758 traversal.next();
5759 }
5760 traversal
5761 }
5762
5763 pub fn advance(&mut self) -> bool {
5764 self.advance_by(1)
5765 }
5766
5767 pub fn advance_by(&mut self, count: usize) -> bool {
5768 self.cursor.seek_forward(
5769 &TraversalTarget::Count {
5770 count: self.end_offset() + count,
5771 include_dirs: self.include_dirs,
5772 include_files: self.include_files,
5773 include_ignored: self.include_ignored,
5774 },
5775 Bias::Left,
5776 )
5777 }
5778
5779 pub fn advance_to_sibling(&mut self) -> bool {
5780 while let Some(entry) = self.cursor.item() {
5781 self.cursor
5782 .seek_forward(&TraversalTarget::successor(&entry.path), Bias::Left);
5783 if let Some(entry) = self.cursor.item()
5784 && (self.include_files || !entry.is_file())
5785 && (self.include_dirs || !entry.is_dir())
5786 && (self.include_ignored || !entry.is_ignored || entry.is_always_included)
5787 {
5788 return true;
5789 }
5790 }
5791 false
5792 }
5793
5794 pub fn back_to_parent(&mut self) -> bool {
5795 let Some(parent_path) = self.cursor.item().and_then(|entry| entry.path.parent()) else {
5796 return false;
5797 };
5798 self.cursor
5799 .seek(&TraversalTarget::path(parent_path), Bias::Left)
5800 }
5801
5802 pub fn entry(&self) -> Option<&'a Entry> {
5803 self.cursor.item()
5804 }
5805
5806 pub fn snapshot(&self) -> &'a Snapshot {
5807 self.snapshot
5808 }
5809
5810 pub fn start_offset(&self) -> usize {
5811 self.cursor
5812 .start()
5813 .count(self.include_files, self.include_dirs, self.include_ignored)
5814 }
5815
5816 pub fn end_offset(&self) -> usize {
5817 self.cursor
5818 .end()
5819 .count(self.include_files, self.include_dirs, self.include_ignored)
5820 }
5821}
5822
5823impl<'a> Iterator for Traversal<'a> {
5824 type Item = &'a Entry;
5825
5826 fn next(&mut self) -> Option<Self::Item> {
5827 if let Some(item) = self.entry() {
5828 self.advance();
5829 Some(item)
5830 } else {
5831 None
5832 }
5833 }
5834}
5835
5836#[derive(Debug, Clone, Copy)]
5837pub enum PathTarget<'a> {
5838 Path(&'a RelPath),
5839 Successor(&'a RelPath),
5840}
5841
5842impl PathTarget<'_> {
5843 fn cmp_path(&self, other: &RelPath) -> Ordering {
5844 match self {
5845 PathTarget::Path(path) => path.cmp(&other),
5846 PathTarget::Successor(path) => {
5847 if other.starts_with(path) {
5848 Ordering::Greater
5849 } else {
5850 Ordering::Equal
5851 }
5852 }
5853 }
5854 }
5855}
5856
5857impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, PathProgress<'a>> for PathTarget<'_> {
5858 fn cmp(&self, cursor_location: &PathProgress<'a>, _: S::Context<'_>) -> Ordering {
5859 self.cmp_path(cursor_location.max_path)
5860 }
5861}
5862
5863impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, TraversalProgress<'a>> for PathTarget<'_> {
5864 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: S::Context<'_>) -> Ordering {
5865 self.cmp_path(cursor_location.max_path)
5866 }
5867}
5868
5869#[derive(Debug)]
5870enum TraversalTarget<'a> {
5871 Path(PathTarget<'a>),
5872 Count {
5873 count: usize,
5874 include_files: bool,
5875 include_ignored: bool,
5876 include_dirs: bool,
5877 },
5878}
5879
5880impl<'a> TraversalTarget<'a> {
5881 fn path(path: &'a RelPath) -> Self {
5882 Self::Path(PathTarget::Path(path))
5883 }
5884
5885 fn successor(path: &'a RelPath) -> Self {
5886 Self::Path(PathTarget::Successor(path))
5887 }
5888
5889 fn cmp_progress(&self, progress: &TraversalProgress) -> Ordering {
5890 match self {
5891 TraversalTarget::Path(path) => path.cmp_path(progress.max_path),
5892 TraversalTarget::Count {
5893 count,
5894 include_files,
5895 include_dirs,
5896 include_ignored,
5897 } => Ord::cmp(
5898 count,
5899 &progress.count(*include_files, *include_dirs, *include_ignored),
5900 ),
5901 }
5902 }
5903}
5904
5905impl<'a> SeekTarget<'a, EntrySummary, TraversalProgress<'a>> for TraversalTarget<'_> {
5906 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5907 self.cmp_progress(cursor_location)
5908 }
5909}
5910
5911impl<'a> SeekTarget<'a, PathSummary<sum_tree::NoSummary>, TraversalProgress<'a>>
5912 for TraversalTarget<'_>
5913{
5914 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
5915 self.cmp_progress(cursor_location)
5916 }
5917}
5918
5919pub struct ChildEntriesOptions {
5920 pub include_files: bool,
5921 pub include_dirs: bool,
5922 pub include_ignored: bool,
5923}
5924
5925pub struct ChildEntriesIter<'a> {
5926 parent_path: &'a RelPath,
5927 traversal: Traversal<'a>,
5928}
5929
5930impl<'a> Iterator for ChildEntriesIter<'a> {
5931 type Item = &'a Entry;
5932
5933 fn next(&mut self) -> Option<Self::Item> {
5934 if let Some(item) = self.traversal.entry()
5935 && item.path.starts_with(self.parent_path)
5936 {
5937 self.traversal.advance_to_sibling();
5938 return Some(item);
5939 }
5940 None
5941 }
5942}
5943
5944impl<'a> From<&'a Entry> for proto::Entry {
5945 fn from(entry: &'a Entry) -> Self {
5946 Self {
5947 id: entry.id.to_proto(),
5948 is_dir: entry.is_dir(),
5949 path: entry.path.as_ref().to_proto(),
5950 inode: entry.inode,
5951 mtime: entry.mtime.map(|time| time.into()),
5952 is_ignored: entry.is_ignored,
5953 is_hidden: entry.is_hidden,
5954 is_external: entry.is_external,
5955 is_fifo: entry.is_fifo,
5956 size: Some(entry.size),
5957 canonical_path: entry
5958 .canonical_path
5959 .as_ref()
5960 .map(|path| path.to_string_lossy().into_owned()),
5961 }
5962 }
5963}
5964
5965impl TryFrom<(&CharBag, &PathMatcher, proto::Entry)> for Entry {
5966 type Error = anyhow::Error;
5967
5968 fn try_from(
5969 (root_char_bag, always_included, entry): (&CharBag, &PathMatcher, proto::Entry),
5970 ) -> Result<Self> {
5971 let kind = if entry.is_dir {
5972 EntryKind::Dir
5973 } else {
5974 EntryKind::File
5975 };
5976
5977 let path =
5978 RelPath::from_proto(&entry.path).context("invalid relative path in proto message")?;
5979 let char_bag = char_bag_for_path(*root_char_bag, &path);
5980 let is_always_included = always_included.is_match(&path);
5981 Ok(Entry {
5982 id: ProjectEntryId::from_proto(entry.id),
5983 kind,
5984 path,
5985 inode: entry.inode,
5986 mtime: entry.mtime.map(|time| time.into()),
5987 size: entry.size.unwrap_or(0),
5988 canonical_path: entry
5989 .canonical_path
5990 .map(|path_string| Arc::from(PathBuf::from(path_string))),
5991 is_ignored: entry.is_ignored,
5992 is_hidden: entry.is_hidden,
5993 is_always_included,
5994 is_external: entry.is_external,
5995 is_private: false,
5996 char_bag,
5997 is_fifo: entry.is_fifo,
5998 })
5999 }
6000}
6001
6002#[derive(Clone, Copy, Debug, Default, Hash, PartialEq, Eq, PartialOrd, Ord)]
6003pub struct ProjectEntryId(usize);
6004
6005impl ProjectEntryId {
6006 pub const MAX: Self = Self(usize::MAX);
6007 pub const MIN: Self = Self(usize::MIN);
6008
6009 pub fn new(counter: &AtomicUsize) -> Self {
6010 Self(counter.fetch_add(1, SeqCst))
6011 }
6012
6013 pub fn from_proto(id: u64) -> Self {
6014 Self(id as usize)
6015 }
6016
6017 pub fn to_proto(self) -> u64 {
6018 self.0 as u64
6019 }
6020
6021 pub fn from_usize(id: usize) -> Self {
6022 ProjectEntryId(id)
6023 }
6024
6025 pub fn to_usize(self) -> usize {
6026 self.0
6027 }
6028}
6029
6030#[cfg(feature = "test-support")]
6031impl CreatedEntry {
6032 pub fn into_included(self) -> Option<Entry> {
6033 match self {
6034 CreatedEntry::Included(entry) => Some(entry),
6035 CreatedEntry::Excluded { .. } => None,
6036 }
6037 }
6038}
6039
6040fn parse_gitfile(content: &str) -> anyhow::Result<&Path> {
6041 let path = content
6042 .strip_prefix("gitdir:")
6043 .with_context(|| format!("parsing gitfile content {content:?}"))?;
6044 Ok(Path::new(path.trim()))
6045}
6046
6047async fn discover_git_paths(dot_git_abs_path: &Arc<Path>, fs: &dyn Fs) -> (Arc<Path>, Arc<Path>) {
6048 let mut repository_dir_abs_path = dot_git_abs_path.clone();
6049 let mut common_dir_abs_path = dot_git_abs_path.clone();
6050
6051 if let Some(path) = fs
6052 .load(dot_git_abs_path)
6053 .await
6054 .ok()
6055 .as_ref()
6056 .and_then(|contents| parse_gitfile(contents).log_err())
6057 {
6058 let path = dot_git_abs_path
6059 .parent()
6060 .unwrap_or(Path::new(""))
6061 .join(path);
6062 if let Some(path) = fs.canonicalize(&path).await.log_err() {
6063 repository_dir_abs_path = Path::new(&path).into();
6064 common_dir_abs_path = repository_dir_abs_path.clone();
6065
6066 if let Some(commondir_contents) = fs.load(&path.join("commondir")).await.ok()
6067 && let Some(commondir_path) = fs
6068 .canonicalize(&path.join(commondir_contents.trim()))
6069 .await
6070 .log_err()
6071 {
6072 common_dir_abs_path = commondir_path.as_path().into();
6073 }
6074 }
6075 };
6076 (repository_dir_abs_path, common_dir_abs_path)
6077}
6078
6079struct NullWatcher;
6080
6081impl fs::Watcher for NullWatcher {
6082 fn add(&self, _path: &Path) -> Result<()> {
6083 Ok(())
6084 }
6085
6086 fn remove(&self, _path: &Path) -> Result<()> {
6087 Ok(())
6088 }
6089}
6090
6091const FILE_ANALYSIS_BYTES: usize = 1024;
6092
6093async fn decode_file_text(
6094 fs: &dyn Fs,
6095 abs_path: &Path,
6096) -> Result<(String, &'static Encoding, bool)> {
6097 let mut file = fs
6098 .open_sync(&abs_path)
6099 .await
6100 .with_context(|| format!("opening file {abs_path:?}"))?;
6101
6102 // First, read the beginning of the file to determine its kind and encoding.
6103 // We do not want to load an entire large blob into memory only to discard it.
6104 let mut file_first_bytes = Vec::with_capacity(FILE_ANALYSIS_BYTES);
6105 let mut buf = [0u8; FILE_ANALYSIS_BYTES];
6106 let mut reached_eof = false;
6107 loop {
6108 if file_first_bytes.len() >= FILE_ANALYSIS_BYTES {
6109 break;
6110 }
6111 let n = file
6112 .read(&mut buf)
6113 .with_context(|| format!("reading bytes of the file {abs_path:?}"))?;
6114 if n == 0 {
6115 reached_eof = true;
6116 break;
6117 }
6118 file_first_bytes.extend_from_slice(&buf[..n]);
6119 }
6120 let (bom_encoding, byte_content) = decode_byte_header(&file_first_bytes);
6121 anyhow::ensure!(
6122 byte_content != ByteContent::Binary,
6123 "Binary files are not supported"
6124 );
6125
6126 // If the file is eligible for opening, read the rest of the file.
6127 let mut content = file_first_bytes;
6128 if !reached_eof {
6129 let mut buf = [0u8; 8 * 1024];
6130 loop {
6131 let n = file
6132 .read(&mut buf)
6133 .with_context(|| format!("reading remaining bytes of the file {abs_path:?}"))?;
6134 if n == 0 {
6135 break;
6136 }
6137 content.extend_from_slice(&buf[..n]);
6138 }
6139 }
6140 decode_byte_full(content, bom_encoding, byte_content)
6141}
6142
6143fn decode_byte_header(prefix: &[u8]) -> (Option<&'static Encoding>, ByteContent) {
6144 if let Some((encoding, _bom_len)) = Encoding::for_bom(prefix) {
6145 return (Some(encoding), ByteContent::Unknown);
6146 }
6147 (None, analyze_byte_content(prefix))
6148}
6149
6150fn decode_byte_full(
6151 bytes: Vec<u8>,
6152 bom_encoding: Option<&'static Encoding>,
6153 byte_content: ByteContent,
6154) -> Result<(String, &'static Encoding, bool)> {
6155 if let Some(encoding) = bom_encoding {
6156 let (cow, _) = encoding.decode_with_bom_removal(&bytes);
6157 return Ok((cow.into_owned(), encoding, true));
6158 }
6159
6160 match byte_content {
6161 ByteContent::Utf16Le => {
6162 let encoding = encoding_rs::UTF_16LE;
6163 let (cow, _, _) = encoding.decode(&bytes);
6164 return Ok((cow.into_owned(), encoding, false));
6165 }
6166 ByteContent::Utf16Be => {
6167 let encoding = encoding_rs::UTF_16BE;
6168 let (cow, _, _) = encoding.decode(&bytes);
6169 return Ok((cow.into_owned(), encoding, false));
6170 }
6171 ByteContent::Binary => {
6172 anyhow::bail!("Binary files are not supported");
6173 }
6174 ByteContent::Unknown => {}
6175 }
6176
6177 fn detect_encoding(bytes: Vec<u8>) -> (String, &'static Encoding) {
6178 let mut detector = EncodingDetector::new();
6179 detector.feed(&bytes, true);
6180
6181 let encoding = detector.guess(None, true); // Use None for TLD hint to ensure neutral detection logic.
6182
6183 let (cow, _, _) = encoding.decode(&bytes);
6184 (cow.into_owned(), encoding)
6185 }
6186
6187 match String::from_utf8(bytes) {
6188 Ok(text) => {
6189 // ISO-2022-JP (and other ISO-2022 variants) consists entirely of 7-bit ASCII bytes,
6190 // so it is valid UTF-8. However, it contains escape sequences starting with '\x1b'.
6191 // If we find an escape character, we double-check the encoding to prevent
6192 // displaying raw escape sequences instead of the correct characters.
6193 if text.contains('\x1b') {
6194 let (s, enc) = detect_encoding(text.into_bytes());
6195 Ok((s, enc, false))
6196 } else {
6197 Ok((text, encoding_rs::UTF_8, false))
6198 }
6199 }
6200 Err(e) => {
6201 let (s, enc) = detect_encoding(e.into_bytes());
6202 Ok((s, enc, false))
6203 }
6204 }
6205}
6206
6207#[derive(Debug, PartialEq)]
6208enum ByteContent {
6209 Utf16Le,
6210 Utf16Be,
6211 Binary,
6212 Unknown,
6213}
6214
6215// Heuristic check using null byte distribution plus a generic text-likeness
6216// heuristic. This prefers UTF-16 when many bytes are NUL and otherwise
6217// distinguishes between text-like and binary-like content.
6218fn analyze_byte_content(bytes: &[u8]) -> ByteContent {
6219 if bytes.len() < 2 {
6220 return ByteContent::Unknown;
6221 }
6222
6223 if is_known_binary_header(bytes) {
6224 return ByteContent::Binary;
6225 }
6226
6227 let limit = bytes.len().min(FILE_ANALYSIS_BYTES);
6228 let mut even_null_count = 0usize;
6229 let mut odd_null_count = 0usize;
6230 let mut non_text_like_count = 0usize;
6231
6232 for (i, &byte) in bytes[..limit].iter().enumerate() {
6233 if byte == 0 {
6234 if i % 2 == 0 {
6235 even_null_count += 1;
6236 } else {
6237 odd_null_count += 1;
6238 }
6239 non_text_like_count += 1;
6240 continue;
6241 }
6242
6243 let is_text_like = match byte {
6244 b'\t' | b'\n' | b'\r' | 0x0C => true,
6245 0x20..=0x7E => true,
6246 // Treat bytes that are likely part of UTF-8 or single-byte encodings as text-like.
6247 0x80..=0xBF | 0xC2..=0xF4 => true,
6248 _ => false,
6249 };
6250
6251 if !is_text_like {
6252 non_text_like_count += 1;
6253 }
6254 }
6255
6256 let total_null_count = even_null_count + odd_null_count;
6257
6258 // If there are no NUL bytes at all, this is overwhelmingly likely to be text.
6259 if total_null_count == 0 {
6260 return ByteContent::Unknown;
6261 }
6262
6263 let has_significant_nulls = total_null_count >= limit / 16;
6264 let nulls_skew_to_even = even_null_count > odd_null_count * 4;
6265 let nulls_skew_to_odd = odd_null_count > even_null_count * 4;
6266
6267 if has_significant_nulls {
6268 let sample = &bytes[..limit];
6269
6270 // UTF-16BE ASCII: [0x00, char] — nulls at even positions (high byte first)
6271 // UTF-16LE ASCII: [char, 0x00] — nulls at odd positions (low byte first)
6272
6273 if nulls_skew_to_even && is_plausible_utf16_text(sample, false) {
6274 return ByteContent::Utf16Be;
6275 }
6276
6277 if nulls_skew_to_odd && is_plausible_utf16_text(sample, true) {
6278 return ByteContent::Utf16Le;
6279 }
6280
6281 return ByteContent::Binary;
6282 }
6283
6284 if non_text_like_count * 100 < limit * 8 {
6285 ByteContent::Unknown
6286 } else {
6287 ByteContent::Binary
6288 }
6289}
6290
6291fn is_known_binary_header(bytes: &[u8]) -> bool {
6292 bytes.starts_with(b"%PDF-") // PDF
6293 || bytes.starts_with(b"PK\x03\x04") // ZIP local header
6294 || bytes.starts_with(b"PK\x05\x06") // ZIP end of central directory
6295 || bytes.starts_with(b"PK\x07\x08") // ZIP spanning/splitting
6296 || bytes.starts_with(b"\x89PNG\r\n\x1a\n") // PNG
6297 || bytes.starts_with(b"\xFF\xD8\xFF") // JPEG
6298 || bytes.starts_with(b"GIF87a") // GIF87a
6299 || bytes.starts_with(b"GIF89a") // GIF89a
6300 || bytes.starts_with(b"IWAD") // Doom IWAD archive
6301 || bytes.starts_with(b"PWAD") // Doom PWAD archive
6302 || bytes.starts_with(b"RIFF") // WAV, AVI, WebP
6303 || bytes.starts_with(b"OggS") // OGG (Vorbis, Opus, FLAC)
6304 || bytes.starts_with(b"fLaC") // FLAC
6305 || bytes.starts_with(b"ID3") // MP3 with ID3v2 tag
6306 || bytes.starts_with(b"\xFF\xFB") // MP3 frame sync (MPEG1 Layer3)
6307 || bytes.starts_with(b"\xFF\xFA") // MP3 frame sync (MPEG1 Layer3)
6308 || bytes.starts_with(b"\xFF\xF3") // MP3 frame sync (MPEG2 Layer3)
6309 || bytes.starts_with(b"\xFF\xF2") // MP3 frame sync (MPEG2 Layer3)
6310}
6311
6312// Null byte skew alone is not enough to identify UTF-16 -- binary formats with
6313// small 16-bit values (like PCM audio) produce the same pattern. Decode the
6314// bytes as UTF-16 and reject if too many code units land in control character
6315// ranges or form unpaired surrogates, which real text almost never contains.
6316fn is_plausible_utf16_text(bytes: &[u8], little_endian: bool) -> bool {
6317 let mut suspicious_count = 0usize;
6318 let mut total = 0usize;
6319
6320 let mut i = 0;
6321 while let Some(code_unit) = read_u16(bytes, i, little_endian) {
6322 total += 1;
6323
6324 match code_unit {
6325 0x0009 | 0x000A | 0x000C | 0x000D => {}
6326 // C0/C1 control characters and non-characters
6327 0x0000..=0x001F | 0x007F..=0x009F | 0xFFFE | 0xFFFF => suspicious_count += 1,
6328 0xD800..=0xDBFF => {
6329 let next_offset = i + 2;
6330 let has_low_surrogate = read_u16(bytes, next_offset, little_endian)
6331 .is_some_and(|next| (0xDC00..=0xDFFF).contains(&next));
6332 if has_low_surrogate {
6333 total += 1;
6334 i += 2;
6335 } else {
6336 suspicious_count += 1;
6337 }
6338 }
6339 // Lone low surrogate without a preceding high surrogate
6340 0xDC00..=0xDFFF => suspicious_count += 1,
6341 _ => {}
6342 }
6343
6344 i += 2;
6345 }
6346
6347 if total == 0 {
6348 return false;
6349 }
6350
6351 // Real UTF-16 text has near-zero control characters; binary data with
6352 // small 16-bit values typically exceeds 5%. 2% provides a safe margin.
6353 suspicious_count * 100 < total * 2
6354}
6355
6356fn read_u16(bytes: &[u8], offset: usize, little_endian: bool) -> Option<u16> {
6357 let pair = [*bytes.get(offset)?, *bytes.get(offset + 1)?];
6358 if little_endian {
6359 return Some(u16::from_le_bytes(pair));
6360 }
6361 Some(u16::from_be_bytes(pair))
6362}
6363
6364#[cfg(test)]
6365mod tests {
6366 use super::*;
6367
6368 /// reproduction of issue #50785
6369 fn build_pcm16_wav_bytes() -> Vec<u8> {
6370 let header: Vec<u8> = vec![
6371 /* RIFF header */
6372 0x52, 0x49, 0x46, 0x46, // "RIFF"
6373 0xc6, 0xcf, 0x00, 0x00, // file size: 8
6374 0x57, 0x41, 0x56, 0x45, // "WAVE"
6375 /* fmt chunk */
6376 0x66, 0x6d, 0x74, 0x20, // "fmt "
6377 0x10, 0x00, 0x00, 0x00, // chunk size: 16
6378 0x01, 0x00, // format: PCM (1)
6379 0x01, 0x00, // channels: 1 (mono)
6380 0x80, 0x3e, 0x00, 0x00, // sample rate: 16000
6381 0x00, 0x7d, 0x00, 0x00, // byte rate: 32000
6382 0x02, 0x00, // block align: 2
6383 0x10, 0x00, // bits per sample: 16
6384 /* LIST chunk */
6385 0x4c, 0x49, 0x53, 0x54, // "LIST"
6386 0x1a, 0x00, 0x00, 0x00, // chunk size: 26
6387 0x49, 0x4e, 0x46, 0x4f, // "INFO"
6388 0x49, 0x53, 0x46, 0x54, // "ISFT"
6389 0x0d, 0x00, 0x00, 0x00, // sub-chunk size: 13
6390 0x4c, 0x61, 0x76, 0x66, 0x36, 0x32, 0x2e, 0x33, // "Lavf62.3"
6391 0x2e, 0x31, 0x30, 0x30, 0x00, // ".100\0"
6392 /* padding byte for word alignment */
6393 0x00, // data chunk header
6394 0x64, 0x61, 0x74, 0x61, // "data"
6395 0x80, 0xcf, 0x00, 0x00, // chunk size
6396 ];
6397
6398 let mut bytes = header;
6399
6400 // fill remaining space up to `FILE_ANALYSIS_BYTES` with synthetic PCM
6401 let audio_bytes_needed = FILE_ANALYSIS_BYTES - bytes.len();
6402 for i in 0..(audio_bytes_needed / 2) {
6403 let sample = (i & 0xFF) as u8;
6404 bytes.push(sample); // low byte: varies
6405 bytes.push(0x00); // high byte: zero for small values
6406 }
6407
6408 bytes
6409 }
6410
6411 #[test]
6412 fn test_pcm16_wav_detected_as_binary() {
6413 let wav_bytes = build_pcm16_wav_bytes();
6414 assert_eq!(wav_bytes.len(), FILE_ANALYSIS_BYTES);
6415
6416 let result = analyze_byte_content(&wav_bytes);
6417 assert_eq!(
6418 result,
6419 ByteContent::Binary,
6420 "PCM 16-bit WAV should be detected as Binary via RIFF header"
6421 );
6422 }
6423
6424 #[test]
6425 fn test_le16_binary_not_misdetected_as_utf16le() {
6426 let mut bytes = b"FAKE".to_vec();
6427 while bytes.len() < FILE_ANALYSIS_BYTES {
6428 let sample = (bytes.len() & 0xFF) as u8;
6429 bytes.push(sample);
6430 bytes.push(0x00);
6431 }
6432 bytes.truncate(FILE_ANALYSIS_BYTES);
6433
6434 let result = analyze_byte_content(&bytes);
6435 assert_eq!(
6436 result,
6437 ByteContent::Binary,
6438 "LE 16-bit binary with control characters should be detected as Binary"
6439 );
6440 }
6441
6442 #[test]
6443 fn test_be16_binary_not_misdetected_as_utf16be() {
6444 let mut bytes = b"FAKE".to_vec();
6445 while bytes.len() < FILE_ANALYSIS_BYTES {
6446 bytes.push(0x00);
6447 let sample = (bytes.len() & 0xFF) as u8;
6448 bytes.push(sample);
6449 }
6450 bytes.truncate(FILE_ANALYSIS_BYTES);
6451
6452 let result = analyze_byte_content(&bytes);
6453 assert_eq!(
6454 result,
6455 ByteContent::Binary,
6456 "BE 16-bit binary with control characters should be detected as Binary"
6457 );
6458 }
6459
6460 #[test]
6461 fn test_utf16le_text_detected_as_utf16le() {
6462 let text = "Hello, world! This is a UTF-16 test string. ";
6463 let mut bytes = Vec::new();
6464 while bytes.len() < FILE_ANALYSIS_BYTES {
6465 bytes.extend(text.encode_utf16().flat_map(|u| u.to_le_bytes()));
6466 }
6467 bytes.truncate(FILE_ANALYSIS_BYTES);
6468
6469 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Le);
6470 }
6471
6472 #[test]
6473 fn test_utf16be_text_detected_as_utf16be() {
6474 let text = "Hello, world! This is a UTF-16 test string. ";
6475 let mut bytes = Vec::new();
6476 while bytes.len() < FILE_ANALYSIS_BYTES {
6477 bytes.extend(text.encode_utf16().flat_map(|u| u.to_be_bytes()));
6478 }
6479 bytes.truncate(FILE_ANALYSIS_BYTES);
6480
6481 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Be);
6482 }
6483
6484 #[test]
6485 fn test_known_binary_headers() {
6486 let cases: &[(&[u8], &str)] = &[
6487 (b"RIFF\x00\x00\x00\x00WAVE", "WAV"),
6488 (b"RIFF\x00\x00\x00\x00AVI ", "AVI"),
6489 (b"OggS\x00\x02", "OGG"),
6490 (b"fLaC\x00\x00", "FLAC"),
6491 (b"ID3\x03\x00", "MP3 ID3v2"),
6492 (b"\xFF\xFB\x90\x00", "MP3 MPEG1 Layer3"),
6493 (b"\xFF\xF3\x90\x00", "MP3 MPEG2 Layer3"),
6494 ];
6495
6496 for (header, label) in cases {
6497 let mut bytes = header.to_vec();
6498 bytes.resize(FILE_ANALYSIS_BYTES, 0x41); // pad with 'A'
6499 assert_eq!(
6500 analyze_byte_content(&bytes),
6501 ByteContent::Binary,
6502 "{label} should be detected as Binary"
6503 );
6504 }
6505 }
6506}