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, TrashId, Watcher, copy_recursive,
12 read_dir_items,
13};
14use futures::{
15 FutureExt as _, Stream, StreamExt,
16 channel::{
17 mpsc::{self, UnboundedSender},
18 oneshot,
19 },
20 select_biased, stream,
21 task::Poll,
22};
23use fuzzy::CharBag;
24use git::{
25 COMMIT_MESSAGE, DOT_GIT, FSMONITOR_DAEMON, GITIGNORE, INDEX_LOCK, LFS_DIR, REPO_EXCLUDE,
26 status::GitSummary,
27};
28use gpui::{
29 App, AppContext as _, AsyncApp, BackgroundExecutor, Context, Entity, EventEmitter, Priority,
30 Task,
31};
32use ignore::IgnoreStack;
33use language::DiskState;
34
35use parking_lot::Mutex;
36use paths::{local_settings_folder_name, local_vscode_folder_name};
37use postage::{
38 barrier,
39 prelude::{Sink as _, Stream as _},
40 watch,
41};
42use rpc::{
43 AnyProtoClient,
44 proto::{self, split_worktree_update},
45};
46pub use settings::WorktreeId;
47use settings::{Settings, SettingsLocation, SettingsStore};
48use smallvec::{SmallVec, smallvec};
49use smol::channel::{self, Sender};
50use std::{
51 any::Any,
52 borrow::Borrow as _,
53 cmp::Ordering,
54 collections::hash_map,
55 convert::TryFrom,
56 ffi::OsStr,
57 fmt,
58 future::Future,
59 mem::{self},
60 ops::{Deref, DerefMut, Range},
61 path::{Path, PathBuf},
62 pin::Pin,
63 sync::{
64 Arc,
65 atomic::{AtomicUsize, Ordering::SeqCst},
66 },
67 time::{Duration, Instant},
68};
69use sum_tree::{Bias, Dimensions, Edit, KeyedItem, SeekTarget, SumTree, Summary, TreeMap, TreeSet};
70use text::{LineEnding, Rope};
71use util::{
72 ResultExt, maybe,
73 paths::{PathMatcher, PathStyle, SanitizedPath, home_dir},
74 rel_path::{RelPath, RelPathBuf},
75};
76pub use worktree_settings::WorktreeSettings;
77
78use crate::ignore::IgnoreKind;
79
80pub const FS_WATCH_LATENCY: Duration = Duration::from_millis(100);
81
82/// A set of local or remote files that are being opened as part of a project.
83/// Responsible for tracking related FS (for local)/collab (for remote) events and corresponding updates.
84/// Stores git repositories data and the diagnostics for the file(s).
85///
86/// Has an absolute path, and may be set to be visible in Zed UI or not.
87/// May correspond to a directory or a single file.
88/// Possible examples:
89/// * a drag and dropped file — may be added as an invisible, "ephemeral" entry to the current worktree
90/// * a directory opened in Zed — may be added as a visible entry to the current worktree
91///
92/// Uses [`Entry`] to track the state of each file/directory, can look up absolute paths for entries.
93pub enum Worktree {
94 Local(LocalWorktree),
95 Remote(RemoteWorktree),
96}
97
98/// An entry, created in the worktree.
99#[derive(Debug)]
100pub enum CreatedEntry {
101 /// Got created and indexed by the worktree, receiving a corresponding entry.
102 Included(Entry),
103 /// Got created, but not indexed due to falling under exclusion filters.
104 Excluded { abs_path: PathBuf },
105}
106
107#[derive(Debug)]
108pub struct LoadedFile {
109 pub file: Arc<File>,
110 pub text: String,
111 pub encoding: &'static Encoding,
112 pub has_bom: bool,
113}
114
115pub struct LoadedBinaryFile {
116 pub file: Arc<File>,
117 pub content: Vec<u8>,
118}
119
120impl fmt::Debug for LoadedBinaryFile {
121 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
122 f.debug_struct("LoadedBinaryFile")
123 .field("file", &self.file)
124 .field("content_bytes", &self.content.len())
125 .finish()
126 }
127}
128
129pub struct LocalWorktree {
130 snapshot: LocalSnapshot,
131 scan_requests_tx: channel::Sender<ScanRequest>,
132 path_prefixes_to_scan_tx: channel::Sender<PathPrefixScanRequest>,
133 is_scanning: (watch::Sender<bool>, watch::Receiver<bool>),
134 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
135 _background_scanner_tasks: Vec<Task<()>>,
136 update_observer: Option<UpdateObservationState>,
137 fs: Arc<dyn Fs>,
138 fs_case_sensitive: bool,
139 visible: bool,
140 next_entry_id: Arc<AtomicUsize>,
141 settings: WorktreeSettings,
142 share_private_files: bool,
143 scanning_enabled: bool,
144}
145
146pub struct PathPrefixScanRequest {
147 path: Arc<RelPath>,
148 done: SmallVec<[barrier::Sender; 1]>,
149}
150
151struct ScanRequest {
152 relative_paths: Vec<Arc<RelPath>>,
153 done: SmallVec<[barrier::Sender; 1]>,
154}
155
156pub struct RemoteWorktree {
157 snapshot: Snapshot,
158 background_snapshot: Arc<Mutex<(Snapshot, Vec<proto::UpdateWorktree>)>>,
159 project_id: u64,
160 client: AnyProtoClient,
161 file_scan_inclusions: PathMatcher,
162 updates_tx: Option<UnboundedSender<proto::UpdateWorktree>>,
163 update_observer: Option<mpsc::UnboundedSender<proto::UpdateWorktree>>,
164 snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>,
165 replica_id: ReplicaId,
166 visible: bool,
167 disconnected: bool,
168}
169
170#[derive(Clone)]
171pub struct Snapshot {
172 id: WorktreeId,
173 /// The absolute path of the worktree root.
174 abs_path: Arc<SanitizedPath>,
175 path_style: PathStyle,
176 root_name: Arc<RelPath>,
177 root_char_bag: CharBag,
178 entries_by_path: SumTree<Entry>,
179 entries_by_id: SumTree<PathEntry>,
180 root_repo_common_dir: Option<Arc<SanitizedPath>>,
181 always_included_entries: Vec<Arc<RelPath>>,
182
183 /// A number that increases every time the worktree begins scanning
184 /// a set of paths from the filesystem. This scanning could be caused
185 /// by some operation performed on the worktree, such as reading or
186 /// writing a file, or by an event reported by the filesystem.
187 scan_id: usize,
188
189 /// The latest scan id that has completed, and whose preceding scans
190 /// have all completed. The current `scan_id` could be more than one
191 /// greater than the `completed_scan_id` if operations are performed
192 /// on the worktree while it is processing a file-system event.
193 completed_scan_id: usize,
194}
195
196/// This path corresponds to the 'content path' of a repository in relation
197/// to Zed's project root.
198/// In the majority of the cases, this is the folder that contains the .git folder.
199/// But if a sub-folder of a git repository is opened, this corresponds to the
200/// project root and the .git folder is located in a parent directory.
201#[derive(Clone, Debug, Ord, PartialOrd, Eq, PartialEq, Hash)]
202pub enum WorkDirectory {
203 InProject {
204 relative_path: Arc<RelPath>,
205 },
206 AboveProject {
207 absolute_path: Arc<Path>,
208 location_in_repo: Arc<Path>,
209 },
210}
211
212impl WorkDirectory {
213 fn path_key(&self) -> PathKey {
214 match self {
215 WorkDirectory::InProject { relative_path } => PathKey(relative_path.clone()),
216 WorkDirectory::AboveProject { .. } => PathKey(RelPath::empty().into()),
217 }
218 }
219
220 /// Returns true if the given path is a child of the work directory.
221 ///
222 /// Note that the path may not be a member of this repository, if there
223 /// is a repository in a directory between these two paths
224 /// external .git folder in a parent folder of the project root.
225 #[track_caller]
226 pub fn directory_contains(&self, path: &RelPath) -> bool {
227 match self {
228 WorkDirectory::InProject { relative_path } => path.starts_with(relative_path),
229 WorkDirectory::AboveProject { .. } => true,
230 }
231 }
232}
233
234impl Default for WorkDirectory {
235 fn default() -> Self {
236 Self::InProject {
237 relative_path: Arc::from(RelPath::empty()),
238 }
239 }
240}
241
242#[derive(Clone)]
243pub struct LocalSnapshot {
244 snapshot: Snapshot,
245 global_gitignore: Option<Arc<Gitignore>>,
246 /// Exclude files for all git repositories in the worktree, indexed by their absolute path.
247 /// The boolean indicates whether the gitignore needs to be updated.
248 repo_exclude_by_work_dir_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
249 /// All of the gitignore files in the worktree, indexed by their absolute path.
250 /// The boolean indicates whether the gitignore needs to be updated.
251 ignores_by_parent_abs_path: HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
252 /// All of the git repositories in the worktree, indexed by the project entry
253 /// id of their parent directory.
254 git_repositories: TreeMap<ProjectEntryId, LocalRepositoryEntry>,
255 /// The file handle of the worktree root
256 /// (so we can find it after it's been moved)
257 root_file_handle: Option<Arc<dyn fs::FileHandle>>,
258}
259
260struct BackgroundScannerState {
261 snapshot: LocalSnapshot,
262 scanned_dirs: HashSet<ProjectEntryId>,
263 path_prefixes_to_scan: HashSet<Arc<RelPath>>,
264 paths_to_scan: HashSet<Arc<RelPath>>,
265 /// The ids of all of the entries that were removed from the snapshot
266 /// as part of the current update. These entry ids may be re-used
267 /// if the same inode is discovered at a new path, or if the given
268 /// path is re-created after being deleted.
269 removed_entries: HashMap<u64, Entry>,
270 changed_paths: Vec<Arc<RelPath>>,
271 prev_snapshot: Snapshot,
272 scanning_enabled: bool,
273}
274
275#[derive(Clone, Debug, Eq, PartialEq)]
276struct EventRoot {
277 path: Arc<RelPath>,
278 was_rescanned: bool,
279}
280
281#[derive(Debug, Clone)]
282struct LocalRepositoryEntry {
283 work_directory_id: ProjectEntryId,
284 work_directory: WorkDirectory,
285 work_directory_abs_path: Arc<Path>,
286 git_dir_scan_id: usize,
287 /// Absolute path to the original .git entry that caused us to create this repository.
288 ///
289 /// This is normally a directory, but may be a "gitfile" that points to a directory elsewhere
290 /// (whose path we then store in `repository_dir_abs_path`).
291 dot_git_abs_path: Arc<Path>,
292 /// Absolute path to the "commondir" for this repository.
293 ///
294 /// This is always a directory. For a normal repository, this is the same as
295 /// `dot_git_abs_path`. For a linked worktree, this is the main repo's `.git`
296 /// directory (resolved from the worktree's `commondir` file). For a submodule,
297 /// this equals `repository_dir_abs_path` (submodules don't have a `commondir`
298 /// file).
299 common_dir_abs_path: Arc<Path>,
300 /// Absolute path to the directory holding the repository's state.
301 ///
302 /// For a normal repository, this is a directory and coincides with `dot_git_abs_path` and
303 /// `common_dir_abs_path`. For a submodule or worktree, this is some subdirectory of the
304 /// commondir like `/project/.git/modules/foo`.
305 repository_dir_abs_path: Arc<Path>,
306}
307
308impl sum_tree::Item for LocalRepositoryEntry {
309 type Summary = PathSummary<sum_tree::NoSummary>;
310
311 fn summary(&self, _: <Self::Summary as Summary>::Context<'_>) -> Self::Summary {
312 PathSummary {
313 max_path: self.work_directory.path_key().0,
314 item_summary: sum_tree::NoSummary,
315 }
316 }
317}
318
319impl KeyedItem for LocalRepositoryEntry {
320 type Key = PathKey;
321
322 fn key(&self) -> Self::Key {
323 self.work_directory.path_key()
324 }
325}
326
327impl Deref for LocalRepositoryEntry {
328 type Target = WorkDirectory;
329
330 fn deref(&self) -> &Self::Target {
331 &self.work_directory
332 }
333}
334
335impl Deref for LocalSnapshot {
336 type Target = Snapshot;
337
338 fn deref(&self) -> &Self::Target {
339 &self.snapshot
340 }
341}
342
343impl DerefMut for LocalSnapshot {
344 fn deref_mut(&mut self) -> &mut Self::Target {
345 &mut self.snapshot
346 }
347}
348
349enum ScanState {
350 Started,
351 Updated {
352 snapshot: LocalSnapshot,
353 changes: UpdatedEntriesSet,
354 barrier: SmallVec<[barrier::Sender; 1]>,
355 scanning: bool,
356 },
357 RootUpdated {
358 new_path: Arc<SanitizedPath>,
359 },
360 RootDeleted,
361}
362
363struct UpdateObservationState {
364 snapshots_tx: mpsc::UnboundedSender<(LocalSnapshot, UpdatedEntriesSet)>,
365 resume_updates: watch::Sender<()>,
366 _maintain_remote_snapshot: Task<Option<()>>,
367}
368
369#[derive(Debug, Clone)]
370pub enum Event {
371 UpdatedEntries(UpdatedEntriesSet),
372 UpdatedGitRepositories(UpdatedGitRepositoriesSet),
373 UpdatedRootRepoCommonDir,
374 DeletedEntry(ProjectEntryId),
375 /// The worktree root itself has been deleted (for single-file worktrees)
376 Deleted,
377}
378
379impl EventEmitter<Event> for Worktree {}
380
381impl Worktree {
382 pub async fn local(
383 path: impl Into<Arc<Path>>,
384 visible: bool,
385 fs: Arc<dyn Fs>,
386 next_entry_id: Arc<AtomicUsize>,
387 scanning_enabled: bool,
388 worktree_id: WorktreeId,
389 cx: &mut AsyncApp,
390 ) -> Result<Entity<Self>> {
391 let abs_path = path.into();
392 let metadata = fs
393 .metadata(&abs_path)
394 .await
395 .context("failed to stat worktree path")?;
396
397 let fs_case_sensitive = fs.is_case_sensitive().await;
398
399 let root_file_handle = if metadata.as_ref().is_some() {
400 fs.open_handle(&abs_path)
401 .await
402 .with_context(|| {
403 format!(
404 "failed to open local worktree root at {}",
405 abs_path.display()
406 )
407 })
408 .log_err()
409 } else {
410 None
411 };
412
413 let root_repo_common_dir = discover_root_repo_common_dir(&abs_path, fs.as_ref())
414 .await
415 .map(SanitizedPath::from_arc);
416
417 Ok(cx.new(move |cx: &mut Context<Worktree>| {
418 let mut snapshot = LocalSnapshot {
419 ignores_by_parent_abs_path: Default::default(),
420 global_gitignore: Default::default(),
421 repo_exclude_by_work_dir_abs_path: Default::default(),
422 git_repositories: Default::default(),
423 snapshot: Snapshot::new(
424 worktree_id,
425 abs_path
426 .file_name()
427 .and_then(|f| f.to_str())
428 .map_or(RelPath::empty().into(), |f| {
429 RelPath::unix(f).unwrap().into()
430 }),
431 abs_path.clone(),
432 PathStyle::local(),
433 ),
434 root_file_handle,
435 };
436 snapshot.root_repo_common_dir = root_repo_common_dir;
437
438 let worktree_id = snapshot.id();
439 let settings_location = Some(SettingsLocation {
440 worktree_id,
441 path: RelPath::empty(),
442 });
443
444 let settings = WorktreeSettings::get(settings_location, cx).clone();
445 cx.observe_global::<SettingsStore>(move |this, cx| {
446 if let Self::Local(this) = this {
447 let settings = WorktreeSettings::get(settings_location, cx).clone();
448 if this.settings != settings {
449 this.settings = settings;
450 this.restart_background_scanners(cx);
451 }
452 }
453 })
454 .detach();
455
456 let share_private_files = false;
457 if let Some(metadata) = metadata {
458 let mut entry = Entry::new(
459 RelPath::empty().into(),
460 &metadata,
461 ProjectEntryId::new(&next_entry_id),
462 snapshot.root_char_bag,
463 None,
464 );
465 if metadata.is_dir {
466 if !scanning_enabled {
467 entry.kind = EntryKind::UnloadedDir;
468 }
469 } else {
470 if let Some(file_name) = abs_path.file_name()
471 && let Some(file_name) = file_name.to_str()
472 && let Ok(path) = RelPath::unix(file_name)
473 {
474 entry.is_private = !share_private_files && settings.is_path_private(path);
475 entry.is_hidden = settings.is_path_hidden(path);
476 }
477 }
478 cx.foreground_executor()
479 .block_on(snapshot.insert_entry(entry, fs.as_ref()));
480 }
481
482 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
483 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
484 let mut worktree = LocalWorktree {
485 share_private_files,
486 next_entry_id,
487 snapshot,
488 is_scanning: watch::channel_with(true),
489 snapshot_subscriptions: Default::default(),
490 update_observer: None,
491 scan_requests_tx,
492 path_prefixes_to_scan_tx,
493 _background_scanner_tasks: Vec::new(),
494 fs,
495 fs_case_sensitive,
496 visible,
497 settings,
498 scanning_enabled,
499 };
500 worktree.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
501 Worktree::Local(worktree)
502 }))
503 }
504
505 pub fn remote(
506 project_id: u64,
507 replica_id: ReplicaId,
508 worktree: proto::WorktreeMetadata,
509 client: AnyProtoClient,
510 path_style: PathStyle,
511 cx: &mut App,
512 ) -> Entity<Self> {
513 cx.new(|cx: &mut Context<Self>| {
514 let mut snapshot = Snapshot::new(
515 WorktreeId::from_proto(worktree.id),
516 RelPath::from_proto(&worktree.root_name)
517 .unwrap_or_else(|_| RelPath::empty().into()),
518 Path::new(&worktree.abs_path).into(),
519 path_style,
520 );
521
522 snapshot.root_repo_common_dir = worktree
523 .root_repo_common_dir
524 .map(|p| SanitizedPath::new_arc(Path::new(&p)));
525
526 let background_snapshot = Arc::new(Mutex::new((
527 snapshot.clone(),
528 Vec::<proto::UpdateWorktree>::new(),
529 )));
530 let (background_updates_tx, mut background_updates_rx) =
531 mpsc::unbounded::<proto::UpdateWorktree>();
532 let (mut snapshot_updated_tx, mut snapshot_updated_rx) = watch::channel();
533
534 let worktree_id = snapshot.id();
535 let settings_location = Some(SettingsLocation {
536 worktree_id,
537 path: RelPath::empty(),
538 });
539
540 let settings = WorktreeSettings::get(settings_location, cx).clone();
541 let worktree = RemoteWorktree {
542 client,
543 project_id,
544 replica_id,
545 snapshot,
546 file_scan_inclusions: settings.parent_dir_scan_inclusions.clone(),
547 background_snapshot: background_snapshot.clone(),
548 updates_tx: Some(background_updates_tx),
549 update_observer: None,
550 snapshot_subscriptions: Default::default(),
551 visible: worktree.visible,
552 disconnected: false,
553 };
554
555 // Apply updates to a separate snapshot in a background task, then
556 // send them to a foreground task which updates the model.
557 cx.background_spawn(async move {
558 while let Some(update) = background_updates_rx.next().await {
559 {
560 let mut lock = background_snapshot.lock();
561 lock.0.apply_remote_update(
562 update.clone(),
563 &settings.parent_dir_scan_inclusions,
564 );
565 lock.1.push(update);
566 }
567 snapshot_updated_tx.send(()).await.ok();
568 }
569 })
570 .detach();
571
572 // On the foreground task, update to the latest snapshot and notify
573 // any update observer of all updates that led to that snapshot.
574 cx.spawn(async move |this, cx| {
575 while (snapshot_updated_rx.recv().await).is_some() {
576 this.update(cx, |this, cx| {
577 let mut entries_changed = false;
578 let this = this.as_remote_mut().unwrap();
579 let old_root_repo_common_dir = this.snapshot.root_repo_common_dir.clone();
580 {
581 let mut lock = this.background_snapshot.lock();
582 this.snapshot = lock.0.clone();
583 for update in lock.1.drain(..) {
584 entries_changed |= !update.updated_entries.is_empty()
585 || !update.removed_entries.is_empty();
586 if let Some(tx) = &this.update_observer {
587 tx.unbounded_send(update).ok();
588 }
589 }
590 };
591
592 if entries_changed {
593 cx.emit(Event::UpdatedEntries(Arc::default()));
594 }
595 if this.snapshot.root_repo_common_dir != old_root_repo_common_dir {
596 cx.emit(Event::UpdatedRootRepoCommonDir);
597 }
598 cx.notify();
599 while let Some((scan_id, _)) = this.snapshot_subscriptions.front() {
600 if this.observed_snapshot(*scan_id) {
601 let (_, tx) = this.snapshot_subscriptions.pop_front().unwrap();
602 let _ = tx.send(());
603 } else {
604 break;
605 }
606 }
607 })?;
608 }
609 anyhow::Ok(())
610 })
611 .detach();
612
613 Worktree::Remote(worktree)
614 })
615 }
616
617 pub fn as_local(&self) -> Option<&LocalWorktree> {
618 if let Worktree::Local(worktree) = self {
619 Some(worktree)
620 } else {
621 None
622 }
623 }
624
625 pub fn as_remote(&self) -> Option<&RemoteWorktree> {
626 if let Worktree::Remote(worktree) = self {
627 Some(worktree)
628 } else {
629 None
630 }
631 }
632
633 pub fn as_local_mut(&mut self) -> Option<&mut LocalWorktree> {
634 if let Worktree::Local(worktree) = self {
635 Some(worktree)
636 } else {
637 None
638 }
639 }
640
641 pub fn as_remote_mut(&mut self) -> Option<&mut RemoteWorktree> {
642 if let Worktree::Remote(worktree) = self {
643 Some(worktree)
644 } else {
645 None
646 }
647 }
648
649 pub fn is_local(&self) -> bool {
650 matches!(self, Worktree::Local(_))
651 }
652
653 pub fn is_remote(&self) -> bool {
654 !self.is_local()
655 }
656
657 pub fn settings_location(&self, _: &Context<Self>) -> SettingsLocation<'static> {
658 SettingsLocation {
659 worktree_id: self.id(),
660 path: RelPath::empty(),
661 }
662 }
663
664 pub fn snapshot(&self) -> Snapshot {
665 match self {
666 Worktree::Local(worktree) => worktree.snapshot.snapshot.clone(),
667 Worktree::Remote(worktree) => worktree.snapshot.clone(),
668 }
669 }
670
671 pub fn scan_id(&self) -> usize {
672 match self {
673 Worktree::Local(worktree) => worktree.snapshot.scan_id,
674 Worktree::Remote(worktree) => worktree.snapshot.scan_id,
675 }
676 }
677
678 pub fn metadata_proto(&self) -> proto::WorktreeMetadata {
679 proto::WorktreeMetadata {
680 id: self.id().to_proto(),
681 root_name: self.root_name().to_proto(),
682 visible: self.is_visible(),
683 abs_path: self.abs_path().to_string_lossy().into_owned(),
684 root_repo_common_dir: self
685 .root_repo_common_dir()
686 .map(|p| p.to_string_lossy().into_owned()),
687 }
688 }
689
690 pub fn completed_scan_id(&self) -> usize {
691 match self {
692 Worktree::Local(worktree) => worktree.snapshot.completed_scan_id,
693 Worktree::Remote(worktree) => worktree.snapshot.completed_scan_id,
694 }
695 }
696
697 pub fn is_visible(&self) -> bool {
698 match self {
699 Worktree::Local(worktree) => worktree.visible,
700 Worktree::Remote(worktree) => worktree.visible,
701 }
702 }
703
704 pub fn replica_id(&self) -> ReplicaId {
705 match self {
706 Worktree::Local(_) => ReplicaId::LOCAL,
707 Worktree::Remote(worktree) => worktree.replica_id,
708 }
709 }
710
711 pub fn abs_path(&self) -> Arc<Path> {
712 match self {
713 Worktree::Local(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
714 Worktree::Remote(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()),
715 }
716 }
717
718 pub fn root_file(&self, cx: &Context<Self>) -> Option<Arc<File>> {
719 let entry = self.root_entry()?;
720 Some(File::for_entry(entry.clone(), cx.entity()))
721 }
722
723 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
724 where
725 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
726 Fut: 'static + Send + Future<Output = bool>,
727 {
728 match self {
729 Worktree::Local(this) => this.observe_updates(project_id, cx, callback),
730 Worktree::Remote(this) => this.observe_updates(project_id, cx, callback),
731 }
732 }
733
734 pub fn stop_observing_updates(&mut self) {
735 match self {
736 Worktree::Local(this) => {
737 this.update_observer.take();
738 }
739 Worktree::Remote(this) => {
740 this.update_observer.take();
741 }
742 }
743 }
744
745 pub fn wait_for_snapshot(
746 &mut self,
747 scan_id: usize,
748 ) -> impl Future<Output = Result<()>> + use<> {
749 match self {
750 Worktree::Local(this) => this.wait_for_snapshot(scan_id).boxed(),
751 Worktree::Remote(this) => this.wait_for_snapshot(scan_id).boxed(),
752 }
753 }
754
755 #[cfg(feature = "test-support")]
756 pub fn has_update_observer(&self) -> bool {
757 match self {
758 Worktree::Local(this) => this.update_observer.is_some(),
759 Worktree::Remote(this) => this.update_observer.is_some(),
760 }
761 }
762
763 pub fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
764 match self {
765 Worktree::Local(this) => this.load_file(path, cx),
766 Worktree::Remote(_) => {
767 Task::ready(Err(anyhow!("remote worktrees can't yet load files")))
768 }
769 }
770 }
771
772 pub fn load_binary_file(
773 &self,
774 path: &RelPath,
775 cx: &Context<Worktree>,
776 ) -> Task<Result<LoadedBinaryFile>> {
777 match self {
778 Worktree::Local(this) => this.load_binary_file(path, cx),
779 Worktree::Remote(_) => {
780 Task::ready(Err(anyhow!("remote worktrees can't yet load binary files")))
781 }
782 }
783 }
784
785 pub fn write_file(
786 &self,
787 path: Arc<RelPath>,
788 text: Rope,
789 line_ending: LineEnding,
790 encoding: &'static Encoding,
791 has_bom: bool,
792 cx: &Context<Worktree>,
793 ) -> Task<Result<Arc<File>>> {
794 match self {
795 Worktree::Local(this) => {
796 this.write_file(path, text, line_ending, encoding, has_bom, cx)
797 }
798 Worktree::Remote(_) => {
799 Task::ready(Err(anyhow!("remote worktree can't yet write files")))
800 }
801 }
802 }
803
804 pub fn create_entry(
805 &mut self,
806 path: Arc<RelPath>,
807 is_directory: bool,
808 content: Option<Vec<u8>>,
809 cx: &Context<Worktree>,
810 ) -> Task<Result<CreatedEntry>> {
811 let worktree_id = self.id();
812 match self {
813 Worktree::Local(this) => this.create_entry(path, is_directory, content, cx),
814 Worktree::Remote(this) => {
815 let project_id = this.project_id;
816 let request = this.client.request(proto::CreateProjectEntry {
817 worktree_id: worktree_id.to_proto(),
818 project_id,
819 path: path.as_ref().to_proto(),
820 content,
821 is_directory,
822 });
823 cx.spawn(async move |this, cx| {
824 let response = request.await?; // TODO!(yara) HERE
825 match response.entry {
826 Some(entry) => this
827 .update(cx, |worktree, cx| {
828 worktree.as_remote_mut().unwrap().insert_entry(
829 entry,
830 response.worktree_scan_id as usize,
831 cx,
832 )
833 })?
834 .await
835 .map(CreatedEntry::Included),
836 None => {
837 let abs_path =
838 this.read_with(cx, |worktree, _| worktree.absolutize(&path))?;
839 Ok(CreatedEntry::Excluded { abs_path })
840 }
841 }
842 })
843 }
844 }
845 }
846
847 pub fn delete_entry(
848 &mut self,
849 entry_id: ProjectEntryId,
850 trash: bool,
851 cx: &mut Context<Worktree>,
852 ) -> Option<Task<Result<Option<TrashId>>>> {
853 let task = match self {
854 Worktree::Local(this) => this.delete_entry(entry_id, trash, cx),
855 Worktree::Remote(this) => this.delete_entry(entry_id, trash, cx),
856 }?;
857
858 let entry = match &*self {
859 Worktree::Local(this) => this.entry_for_id(entry_id),
860 Worktree::Remote(this) => this.entry_for_id(entry_id),
861 }?;
862
863 let mut ids = vec![entry_id];
864 let path = &*entry.path;
865
866 self.get_children_ids_recursive(path, &mut ids);
867
868 for id in ids {
869 cx.emit(Event::DeletedEntry(id));
870 }
871 Some(task)
872 }
873
874 pub async fn restore_entry(
875 trash_id: TrashId,
876 worktree: Entity<Self>,
877 cx: &mut AsyncApp,
878 ) -> Result<RelPathBuf> {
879 let is_local = worktree.read_with(cx, |this, _| this.is_local());
880 if is_local {
881 LocalWorktree::restore_entry(trash_id, worktree, cx).await
882 } else {
883 // TODO(dino): Add support for restoring entries in remote worktrees.
884 Err(anyhow!("Unsupported"))
885 }
886 }
887
888 fn get_children_ids_recursive(&self, path: &RelPath, ids: &mut Vec<ProjectEntryId>) {
889 let children_iter = self.child_entries(path);
890 for child in children_iter {
891 ids.push(child.id);
892 self.get_children_ids_recursive(&child.path, ids);
893 }
894 }
895
896 pub fn copy_external_entries(
897 &mut self,
898 target_directory: Arc<RelPath>,
899 paths: Vec<Arc<Path>>,
900 fs: Arc<dyn Fs>,
901 cx: &Context<Worktree>,
902 ) -> Task<Result<Vec<ProjectEntryId>>> {
903 match self {
904 Worktree::Local(this) => this.copy_external_entries(target_directory, paths, cx),
905 Worktree::Remote(this) => this.copy_external_entries(target_directory, paths, fs, cx),
906 }
907 }
908
909 pub fn expand_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_entry(entry_id, cx),
916 Worktree::Remote(this) => {
917 let response = this.client.request(proto::ExpandProjectEntry {
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 fn expand_all_for_entry(
936 &mut self,
937 entry_id: ProjectEntryId,
938 cx: &Context<Worktree>,
939 ) -> Option<Task<Result<()>>> {
940 match self {
941 Worktree::Local(this) => this.expand_all_for_entry(entry_id, cx),
942 Worktree::Remote(this) => {
943 let response = this.client.request(proto::ExpandAllForProjectEntry {
944 project_id: this.project_id,
945 entry_id: entry_id.to_proto(),
946 });
947 Some(cx.spawn(async move |this, cx| {
948 let response = response.await?;
949 this.update(cx, |this, _| {
950 this.as_remote_mut()
951 .unwrap()
952 .wait_for_snapshot(response.worktree_scan_id as usize)
953 })?
954 .await?;
955 Ok(())
956 }))
957 }
958 }
959 }
960
961 pub async fn handle_create_entry(
962 this: Entity<Self>,
963 request: proto::CreateProjectEntry,
964 mut cx: AsyncApp,
965 ) -> Result<proto::ProjectEntryResponse> {
966 let (scan_id, entry) = this.update(&mut cx, |this, cx| {
967 anyhow::Ok((
968 this.scan_id(),
969 this.create_entry(
970 RelPath::from_proto(&request.path).with_context(|| {
971 format!("received invalid relative path {:?}", request.path)
972 })?,
973 request.is_directory,
974 request.content,
975 cx,
976 ),
977 ))
978 })?;
979 Ok(proto::ProjectEntryResponse {
980 entry: match &entry.await? {
981 CreatedEntry::Included(entry) => Some(entry.into()),
982 CreatedEntry::Excluded { .. } => None,
983 },
984 worktree_scan_id: scan_id as u64,
985 })
986 }
987
988 pub async fn handle_delete_entry(
989 this: Entity<Self>,
990 request: proto::DeleteProjectEntry,
991 mut cx: AsyncApp,
992 ) -> Result<proto::ProjectEntryResponse> {
993 let (scan_id, task) = this.update(&mut cx, |this, cx| {
994 (
995 this.scan_id(),
996 this.delete_entry(
997 ProjectEntryId::from_proto(request.entry_id),
998 request.use_trash,
999 cx,
1000 ),
1001 )
1002 });
1003 let trash_id = task
1004 .ok_or_else(|| anyhow::anyhow!("invalid entry"))?
1005 .await?;
1006 Ok(proto::ProjectEntryResponse {
1007 entry: None,
1008 worktree_scan_id: scan_id as u64,
1009 })
1010 }
1011
1012 pub async fn handle_expand_entry(
1013 this: Entity<Self>,
1014 request: proto::ExpandProjectEntry,
1015 mut cx: AsyncApp,
1016 ) -> Result<proto::ExpandProjectEntryResponse> {
1017 let task = this.update(&mut cx, |this, cx| {
1018 this.expand_entry(ProjectEntryId::from_proto(request.entry_id), cx)
1019 });
1020 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
1021 .await?;
1022 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
1023 Ok(proto::ExpandProjectEntryResponse {
1024 worktree_scan_id: scan_id as u64,
1025 })
1026 }
1027
1028 pub async fn handle_expand_all_for_entry(
1029 this: Entity<Self>,
1030 request: proto::ExpandAllForProjectEntry,
1031 mut cx: AsyncApp,
1032 ) -> Result<proto::ExpandAllForProjectEntryResponse> {
1033 let task = this.update(&mut cx, |this, cx| {
1034 this.expand_all_for_entry(ProjectEntryId::from_proto(request.entry_id), cx)
1035 });
1036 task.ok_or_else(|| anyhow::anyhow!("no such entry"))?
1037 .await?;
1038 let scan_id = this.read_with(&cx, |this, _| this.scan_id());
1039 Ok(proto::ExpandAllForProjectEntryResponse {
1040 worktree_scan_id: scan_id as u64,
1041 })
1042 }
1043
1044 pub fn is_single_file(&self) -> bool {
1045 self.root_dir().is_none()
1046 }
1047
1048 /// For visible worktrees, returns the path with the worktree name as the first component.
1049 /// Otherwise, returns an absolute path.
1050 pub fn full_path(&self, worktree_relative_path: &RelPath) -> PathBuf {
1051 if self.is_visible() {
1052 self.root_name()
1053 .join(worktree_relative_path)
1054 .display(self.path_style)
1055 .to_string()
1056 .into()
1057 } else {
1058 let full_path = self.abs_path();
1059 let mut full_path_string = if self.is_local()
1060 && let Ok(stripped) = full_path.strip_prefix(home_dir())
1061 {
1062 self.path_style
1063 .join("~", &*stripped.to_string_lossy())
1064 .unwrap()
1065 } else {
1066 full_path.to_string_lossy().into_owned()
1067 };
1068
1069 if worktree_relative_path.components().next().is_some() {
1070 full_path_string.push_str(self.path_style.primary_separator());
1071 full_path_string.push_str(&worktree_relative_path.display(self.path_style));
1072 }
1073
1074 full_path_string.into()
1075 }
1076 }
1077}
1078
1079impl LocalWorktree {
1080 pub fn fs(&self) -> &Arc<dyn Fs> {
1081 &self.fs
1082 }
1083
1084 pub fn is_path_private(&self, path: &RelPath) -> bool {
1085 !self.share_private_files && self.settings.is_path_private(path)
1086 }
1087
1088 pub fn fs_is_case_sensitive(&self) -> bool {
1089 self.fs_case_sensitive
1090 }
1091
1092 fn restart_background_scanners(&mut self, cx: &Context<Worktree>) {
1093 let (scan_requests_tx, scan_requests_rx) = channel::unbounded();
1094 let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = channel::unbounded();
1095 self.scan_requests_tx = scan_requests_tx;
1096 self.path_prefixes_to_scan_tx = path_prefixes_to_scan_tx;
1097
1098 self.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx);
1099 let always_included_entries = mem::take(&mut self.snapshot.always_included_entries);
1100 log::debug!(
1101 "refreshing entries for the following always included paths: {:?}",
1102 always_included_entries
1103 );
1104
1105 // Cleans up old always included entries to ensure they get updated properly. Otherwise,
1106 // nested always included entries may not get updated and will result in out-of-date info.
1107 self.refresh_entries_for_paths(always_included_entries);
1108 }
1109
1110 fn start_background_scanner(
1111 &mut self,
1112 scan_requests_rx: channel::Receiver<ScanRequest>,
1113 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
1114 cx: &Context<Worktree>,
1115 ) {
1116 let snapshot = self.snapshot();
1117 let share_private_files = self.share_private_files;
1118 let next_entry_id = self.next_entry_id.clone();
1119 let fs = self.fs.clone();
1120 let scanning_enabled = self.scanning_enabled;
1121 let settings = self.settings.clone();
1122 let (scan_states_tx, mut scan_states_rx) = mpsc::unbounded();
1123 let background_scanner = cx.background_spawn({
1124 let abs_path = snapshot.abs_path.as_path().to_path_buf();
1125 let background = cx.background_executor().clone();
1126 async move {
1127 let (events, watcher) = if scanning_enabled {
1128 fs.watch(&abs_path, FS_WATCH_LATENCY).await
1129 } else {
1130 (Box::pin(stream::pending()) as _, Arc::new(NullWatcher) as _)
1131 };
1132 let fs_case_sensitive = fs.is_case_sensitive().await;
1133
1134 let is_single_file = snapshot.snapshot.root_dir().is_none();
1135 let mut scanner = BackgroundScanner {
1136 fs,
1137 fs_case_sensitive,
1138 status_updates_tx: scan_states_tx,
1139 executor: background,
1140 scan_requests_rx,
1141 path_prefixes_to_scan_rx,
1142 next_entry_id,
1143 state: async_lock::Mutex::new(BackgroundScannerState {
1144 prev_snapshot: snapshot.snapshot.clone(),
1145 snapshot,
1146 scanned_dirs: Default::default(),
1147 scanning_enabled,
1148 path_prefixes_to_scan: Default::default(),
1149 paths_to_scan: Default::default(),
1150 removed_entries: Default::default(),
1151 changed_paths: Default::default(),
1152 }),
1153 phase: BackgroundScannerPhase::InitialScan,
1154 share_private_files,
1155 settings,
1156 watcher,
1157 is_single_file,
1158 };
1159
1160 scanner
1161 .run(Box::pin(events.map(|events| events.into_iter().collect())))
1162 .await;
1163 }
1164 });
1165 let scan_state_updater = cx.spawn(async move |this, cx| {
1166 while let Some((state, this)) = scan_states_rx.next().await.zip(this.upgrade()) {
1167 this.update(cx, |this, cx| {
1168 let this = this.as_local_mut().unwrap();
1169 match state {
1170 ScanState::Started => {
1171 *this.is_scanning.0.borrow_mut() = true;
1172 }
1173 ScanState::Updated {
1174 snapshot,
1175 changes,
1176 barrier,
1177 scanning,
1178 } => {
1179 *this.is_scanning.0.borrow_mut() = scanning;
1180 this.set_snapshot(snapshot, changes, cx);
1181 drop(barrier);
1182 }
1183 ScanState::RootUpdated { new_path } => {
1184 this.update_abs_path_and_refresh(new_path, cx);
1185 }
1186 ScanState::RootDeleted => {
1187 log::info!(
1188 "worktree root {} no longer exists, closing worktree",
1189 this.abs_path().display()
1190 );
1191 cx.emit(Event::Deleted);
1192 }
1193 }
1194 });
1195 }
1196 });
1197 self._background_scanner_tasks = vec![background_scanner, scan_state_updater];
1198 *self.is_scanning.0.borrow_mut() = true;
1199 }
1200
1201 fn set_snapshot(
1202 &mut self,
1203 mut new_snapshot: LocalSnapshot,
1204 entry_changes: UpdatedEntriesSet,
1205 cx: &mut Context<Worktree>,
1206 ) {
1207 let repo_changes = self.changed_repos(&self.snapshot, &mut new_snapshot);
1208
1209 new_snapshot.root_repo_common_dir = new_snapshot
1210 .local_repo_for_work_directory_path(RelPath::empty())
1211 .map(|repo| SanitizedPath::from_arc(repo.common_dir_abs_path.clone()));
1212
1213 let root_repo_common_dir_changed =
1214 self.snapshot.root_repo_common_dir != new_snapshot.root_repo_common_dir;
1215 self.snapshot = new_snapshot;
1216
1217 if let Some(share) = self.update_observer.as_mut() {
1218 share
1219 .snapshots_tx
1220 .unbounded_send((self.snapshot.clone(), entry_changes.clone()))
1221 .ok();
1222 }
1223
1224 if !entry_changes.is_empty() {
1225 cx.emit(Event::UpdatedEntries(entry_changes));
1226 }
1227 if !repo_changes.is_empty() {
1228 cx.emit(Event::UpdatedGitRepositories(repo_changes));
1229 }
1230 if root_repo_common_dir_changed {
1231 cx.emit(Event::UpdatedRootRepoCommonDir);
1232 }
1233
1234 while let Some((scan_id, _)) = self.snapshot_subscriptions.front() {
1235 if self.snapshot.completed_scan_id >= *scan_id {
1236 let (_, tx) = self.snapshot_subscriptions.pop_front().unwrap();
1237 tx.send(()).ok();
1238 } else {
1239 break;
1240 }
1241 }
1242 }
1243
1244 fn changed_repos(
1245 &self,
1246 old_snapshot: &LocalSnapshot,
1247 new_snapshot: &mut LocalSnapshot,
1248 ) -> UpdatedGitRepositoriesSet {
1249 let mut changes = Vec::new();
1250 let mut old_repos = old_snapshot.git_repositories.iter().peekable();
1251 let new_repos = new_snapshot.git_repositories.clone();
1252 let mut new_repos = new_repos.iter().peekable();
1253
1254 loop {
1255 match (new_repos.peek().map(clone), old_repos.peek().map(clone)) {
1256 (Some((new_entry_id, new_repo)), Some((old_entry_id, old_repo))) => {
1257 match Ord::cmp(&new_entry_id, &old_entry_id) {
1258 Ordering::Less => {
1259 changes.push(UpdatedGitRepository {
1260 work_directory_id: new_entry_id,
1261 old_work_directory_abs_path: None,
1262 new_work_directory_abs_path: Some(
1263 new_repo.work_directory_abs_path.clone(),
1264 ),
1265 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1266 repository_dir_abs_path: Some(
1267 new_repo.repository_dir_abs_path.clone(),
1268 ),
1269 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1270 });
1271 new_repos.next();
1272 }
1273 Ordering::Equal => {
1274 if new_repo.git_dir_scan_id != old_repo.git_dir_scan_id
1275 || new_repo.work_directory_abs_path
1276 != old_repo.work_directory_abs_path
1277 {
1278 changes.push(UpdatedGitRepository {
1279 work_directory_id: new_entry_id,
1280 old_work_directory_abs_path: Some(
1281 old_repo.work_directory_abs_path.clone(),
1282 ),
1283 new_work_directory_abs_path: Some(
1284 new_repo.work_directory_abs_path.clone(),
1285 ),
1286 dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()),
1287 repository_dir_abs_path: Some(
1288 new_repo.repository_dir_abs_path.clone(),
1289 ),
1290 common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()),
1291 });
1292 }
1293 new_repos.next();
1294 old_repos.next();
1295 }
1296 Ordering::Greater => {
1297 changes.push(UpdatedGitRepository {
1298 work_directory_id: old_entry_id,
1299 old_work_directory_abs_path: Some(
1300 old_repo.work_directory_abs_path.clone(),
1301 ),
1302 new_work_directory_abs_path: None,
1303 dot_git_abs_path: None,
1304 repository_dir_abs_path: None,
1305 common_dir_abs_path: None,
1306 });
1307 old_repos.next();
1308 }
1309 }
1310 }
1311 (Some((entry_id, repo)), None) => {
1312 changes.push(UpdatedGitRepository {
1313 work_directory_id: entry_id,
1314 old_work_directory_abs_path: None,
1315 new_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1316 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1317 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1318 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1319 });
1320 new_repos.next();
1321 }
1322 (None, Some((entry_id, repo))) => {
1323 changes.push(UpdatedGitRepository {
1324 work_directory_id: entry_id,
1325 old_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()),
1326 new_work_directory_abs_path: None,
1327 dot_git_abs_path: Some(repo.dot_git_abs_path.clone()),
1328 repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()),
1329 common_dir_abs_path: Some(repo.common_dir_abs_path.clone()),
1330 });
1331 old_repos.next();
1332 }
1333 (None, None) => break,
1334 }
1335 }
1336
1337 fn clone<T: Clone, U: Clone>(value: &(&T, &U)) -> (T, U) {
1338 (value.0.clone(), value.1.clone())
1339 }
1340
1341 changes.into()
1342 }
1343
1344 pub fn scan_complete(&self) -> impl Future<Output = ()> + use<> {
1345 let mut is_scanning_rx = self.is_scanning.1.clone();
1346 async move {
1347 let mut is_scanning = *is_scanning_rx.borrow();
1348 while is_scanning {
1349 if let Some(value) = is_scanning_rx.recv().await {
1350 is_scanning = value;
1351 } else {
1352 break;
1353 }
1354 }
1355 }
1356 }
1357
1358 pub fn wait_for_snapshot(
1359 &mut self,
1360 scan_id: usize,
1361 ) -> impl Future<Output = Result<()>> + use<> {
1362 let (tx, rx) = oneshot::channel();
1363 if self.snapshot.completed_scan_id >= scan_id {
1364 tx.send(()).ok();
1365 } else {
1366 match self
1367 .snapshot_subscriptions
1368 .binary_search_by_key(&scan_id, |probe| probe.0)
1369 {
1370 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
1371 }
1372 }
1373
1374 async move {
1375 rx.await?;
1376 Ok(())
1377 }
1378 }
1379
1380 pub fn snapshot(&self) -> LocalSnapshot {
1381 self.snapshot.clone()
1382 }
1383
1384 pub fn settings(&self) -> WorktreeSettings {
1385 self.settings.clone()
1386 }
1387
1388 fn load_binary_file(
1389 &self,
1390 path: &RelPath,
1391 cx: &Context<Worktree>,
1392 ) -> Task<Result<LoadedBinaryFile>> {
1393 let path = Arc::from(path);
1394 let abs_path = self.absolutize(&path);
1395 let fs = self.fs.clone();
1396 let entry = self.refresh_entry(path.clone(), None, cx);
1397 let is_private = self.is_path_private(&path);
1398
1399 let worktree = cx.weak_entity();
1400 cx.background_spawn(async move {
1401 let content = fs.load_bytes(&abs_path).await?;
1402
1403 let worktree = worktree.upgrade().context("worktree was dropped")?;
1404 let file = match entry.await? {
1405 Some(entry) => File::for_entry(entry, worktree),
1406 None => {
1407 let metadata = fs
1408 .metadata(&abs_path)
1409 .await
1410 .with_context(|| {
1411 format!("Loading metadata for excluded file {abs_path:?}")
1412 })?
1413 .with_context(|| {
1414 format!("Excluded file {abs_path:?} got removed during loading")
1415 })?;
1416 Arc::new(File {
1417 entry_id: None,
1418 worktree,
1419 path,
1420 disk_state: DiskState::Present {
1421 mtime: metadata.mtime,
1422 size: metadata.len,
1423 },
1424 is_local: true,
1425 is_private,
1426 })
1427 }
1428 };
1429
1430 Ok(LoadedBinaryFile { file, content })
1431 })
1432 }
1433
1434 #[ztracing::instrument(skip_all)]
1435 fn load_file(&self, path: &RelPath, cx: &Context<Worktree>) -> Task<Result<LoadedFile>> {
1436 let path = Arc::from(path);
1437 let abs_path = self.absolutize(&path);
1438 let fs = self.fs.clone();
1439 let entry = self.refresh_entry(path.clone(), None, cx);
1440 let is_private = self.is_path_private(path.as_ref());
1441
1442 let this = cx.weak_entity();
1443 cx.background_spawn(async move {
1444 // WARN: Temporary workaround for #27283.
1445 // We are not efficient with our memory usage per file, and use in excess of 64GB for a 10GB file
1446 // Therefore, as a temporary workaround to prevent system freezes, we just bail before opening a file
1447 // if it is too large
1448 // 5GB seems to be more reasonable, peaking at ~16GB, while 6GB jumps up to >24GB which seems like a
1449 // reasonable limit
1450 {
1451 const FILE_SIZE_MAX: u64 = 6 * 1024 * 1024 * 1024; // 6GB
1452 if let Ok(Some(metadata)) = fs.metadata(&abs_path).await
1453 && metadata.len >= FILE_SIZE_MAX
1454 {
1455 anyhow::bail!("File is too large to load");
1456 }
1457 }
1458 let (text, encoding, has_bom) = decode_file_text(fs.as_ref(), &abs_path).await?;
1459
1460 let worktree = this.upgrade().context("worktree was dropped")?;
1461 let file = match entry.await? {
1462 Some(entry) => File::for_entry(entry, worktree),
1463 None => {
1464 let metadata = fs
1465 .metadata(&abs_path)
1466 .await
1467 .with_context(|| {
1468 format!("Loading metadata for excluded file {abs_path:?}")
1469 })?
1470 .with_context(|| {
1471 format!("Excluded file {abs_path:?} got removed during loading")
1472 })?;
1473 Arc::new(File {
1474 entry_id: None,
1475 worktree,
1476 path,
1477 disk_state: DiskState::Present {
1478 mtime: metadata.mtime,
1479 size: metadata.len,
1480 },
1481 is_local: true,
1482 is_private,
1483 })
1484 }
1485 };
1486
1487 Ok(LoadedFile {
1488 file,
1489 text,
1490 encoding,
1491 has_bom,
1492 })
1493 })
1494 }
1495
1496 /// Find the lowest path in the worktree's datastructures that is an ancestor
1497 fn lowest_ancestor(&self, path: &RelPath) -> Arc<RelPath> {
1498 let mut lowest_ancestor = None;
1499 for path in path.ancestors() {
1500 if self.entry_for_path(path).is_some() {
1501 lowest_ancestor = Some(path.into());
1502 break;
1503 }
1504 }
1505
1506 lowest_ancestor.unwrap_or_else(|| RelPath::empty().into())
1507 }
1508
1509 pub fn create_entry(
1510 &self,
1511 path: Arc<RelPath>,
1512 is_dir: bool,
1513 content: Option<Vec<u8>>,
1514 cx: &Context<Worktree>,
1515 ) -> Task<Result<CreatedEntry>> {
1516 let abs_path = self.absolutize(&path);
1517 let path_excluded = self.settings.is_path_excluded(&path);
1518 let fs = self.fs.clone();
1519 let task_abs_path = abs_path.clone();
1520 let write = cx.background_spawn(async move {
1521 if is_dir {
1522 fs.create_dir(&task_abs_path)
1523 .await
1524 .with_context(|| format!("creating directory {task_abs_path:?}"))
1525 } else {
1526 fs.write(&task_abs_path, content.as_deref().unwrap_or(&[]))
1527 .await
1528 .with_context(|| format!("creating file {task_abs_path:?}"))
1529 }
1530 });
1531
1532 let lowest_ancestor = self.lowest_ancestor(&path);
1533 cx.spawn(async move |this, cx| {
1534 write.await?;
1535 if path_excluded {
1536 return Ok(CreatedEntry::Excluded { abs_path });
1537 }
1538
1539 let (result, refreshes) = this.update(cx, |this, cx| {
1540 let mut refreshes = Vec::new();
1541 let refresh_paths = path.strip_prefix(&lowest_ancestor).unwrap();
1542 for refresh_path in refresh_paths.ancestors() {
1543 if refresh_path == RelPath::empty() {
1544 continue;
1545 }
1546 let refresh_full_path = lowest_ancestor.join(refresh_path);
1547
1548 refreshes.push(this.as_local_mut().unwrap().refresh_entry(
1549 refresh_full_path,
1550 None,
1551 cx,
1552 ));
1553 }
1554 (
1555 this.as_local_mut().unwrap().refresh_entry(path, None, cx),
1556 refreshes,
1557 )
1558 })?;
1559 for refresh in refreshes {
1560 refresh.await.log_err();
1561 }
1562
1563 Ok(result
1564 .await?
1565 .map(CreatedEntry::Included)
1566 .unwrap_or_else(|| CreatedEntry::Excluded { abs_path }))
1567 })
1568 }
1569
1570 pub fn write_file(
1571 &self,
1572 path: Arc<RelPath>,
1573 text: Rope,
1574 line_ending: LineEnding,
1575 encoding: &'static Encoding,
1576 has_bom: bool,
1577 cx: &Context<Worktree>,
1578 ) -> Task<Result<Arc<File>>> {
1579 let fs = self.fs.clone();
1580 let is_private = self.is_path_private(&path);
1581 let abs_path = self.absolutize(&path);
1582
1583 let write = cx.background_spawn({
1584 let fs = fs.clone();
1585 let abs_path = abs_path.clone();
1586 async move {
1587 // For UTF-8, use the optimized `fs.save` which writes Rope chunks directly to disk
1588 // without allocating a contiguous string.
1589 if encoding == encoding_rs::UTF_8 && !has_bom {
1590 return fs.save(&abs_path, &text, line_ending).await;
1591 }
1592
1593 // For legacy encodings (e.g. Shift-JIS), we fall back to converting the entire Rope
1594 // to a String/Bytes in memory before writing.
1595 //
1596 // Note: This is inefficient for very large files compared to the streaming approach above,
1597 // but supporting streaming writes for arbitrary encodings would require a significant
1598 // refactor of the `fs` crate to expose a Writer interface.
1599 let text_string = text.to_string();
1600 let normalized_text = match line_ending {
1601 LineEnding::Unix => text_string,
1602 LineEnding::Windows => text_string.replace('\n', "\r\n"),
1603 };
1604
1605 // Create the byte vector manually for UTF-16 encodings because encoding_rs encodes to UTF-8 by default (per WHATWG standards),
1606 // which is not what we want for saving files.
1607 let bytes = if encoding == encoding_rs::UTF_16BE {
1608 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1609 if has_bom {
1610 data.extend_from_slice(&[0xFE, 0xFF]); // BOM
1611 }
1612 let utf16be_bytes =
1613 normalized_text.encode_utf16().flat_map(|u| u.to_be_bytes());
1614 data.extend(utf16be_bytes);
1615 data.into()
1616 } else if encoding == encoding_rs::UTF_16LE {
1617 let mut data = Vec::with_capacity(normalized_text.len() * 2 + 2);
1618 if has_bom {
1619 data.extend_from_slice(&[0xFF, 0xFE]); // BOM
1620 }
1621 let utf16le_bytes =
1622 normalized_text.encode_utf16().flat_map(|u| u.to_le_bytes());
1623 data.extend(utf16le_bytes);
1624 data.into()
1625 } else {
1626 // For other encodings (Shift-JIS, UTF-8 with BOM, etc.), delegate to encoding_rs.
1627 let bom_bytes = if has_bom {
1628 if encoding == encoding_rs::UTF_8 {
1629 vec![0xEF, 0xBB, 0xBF]
1630 } else {
1631 vec![]
1632 }
1633 } else {
1634 vec![]
1635 };
1636 let (cow, _, _) = encoding.encode(&normalized_text);
1637 if !bom_bytes.is_empty() {
1638 let mut bytes = bom_bytes;
1639 bytes.extend_from_slice(&cow);
1640 bytes.into()
1641 } else {
1642 cow
1643 }
1644 };
1645
1646 fs.write(&abs_path, &bytes).await
1647 }
1648 });
1649
1650 cx.spawn(async move |this, cx| {
1651 write.await?;
1652 let entry = this
1653 .update(cx, |this, cx| {
1654 this.as_local_mut()
1655 .unwrap()
1656 .refresh_entry(path.clone(), None, cx)
1657 })?
1658 .await?;
1659 let worktree = this.upgrade().context("worktree dropped")?;
1660 if let Some(entry) = entry {
1661 Ok(File::for_entry(entry, worktree))
1662 } else {
1663 let metadata = fs
1664 .metadata(&abs_path)
1665 .await
1666 .with_context(|| {
1667 format!("Fetching metadata after saving the excluded buffer {abs_path:?}")
1668 })?
1669 .with_context(|| {
1670 format!("Excluded buffer {path:?} got removed during saving")
1671 })?;
1672 Ok(Arc::new(File {
1673 worktree,
1674 path,
1675 disk_state: DiskState::Present {
1676 mtime: metadata.mtime,
1677 size: metadata.len,
1678 },
1679 entry_id: None,
1680 is_local: true,
1681 is_private,
1682 }))
1683 }
1684 })
1685 }
1686
1687 pub fn delete_entry(
1688 &self,
1689 entry_id: ProjectEntryId,
1690 trash: bool,
1691 cx: &Context<Worktree>,
1692 ) -> Option<Task<Result<Option<TrashId>>>> {
1693 let entry = self.entry_for_id(entry_id)?.clone();
1694 let abs_path = self.absolutize(&entry.path);
1695 let fs = self.fs.clone();
1696
1697 let delete = cx.background_spawn(async move {
1698 let trash_id = match (entry.is_file(), trash) {
1699 (true, true) => Some(fs.trash(&abs_path, Default::default()).await?),
1700 (false, true) => Some(
1701 fs.trash(
1702 &abs_path,
1703 RemoveOptions {
1704 recursive: true,
1705 ignore_if_not_exists: false,
1706 },
1707 )
1708 .await?,
1709 ),
1710 (true, false) => {
1711 fs.remove_file(&abs_path, Default::default()).await?;
1712 None
1713 }
1714 (false, false) => {
1715 fs.remove_dir(
1716 &abs_path,
1717 RemoveOptions {
1718 recursive: true,
1719 ignore_if_not_exists: false,
1720 },
1721 )
1722 .await?;
1723 None
1724 }
1725 };
1726
1727 anyhow::Ok((trash_id, entry.path))
1728 });
1729
1730 Some(cx.spawn(async move |this, cx| {
1731 let (trashed_entry, path) = delete.await?;
1732 this.update(cx, |this, _| {
1733 this.as_local_mut()
1734 .unwrap()
1735 .refresh_entries_for_paths(vec![path])
1736 })?
1737 .recv()
1738 .await;
1739
1740 Ok(trashed_entry)
1741 }))
1742 }
1743
1744 pub async fn restore_entry(
1745 trash_entry: TrashId,
1746 this: Entity<Worktree>,
1747 cx: &mut AsyncApp,
1748 ) -> Result<RelPathBuf> {
1749 let Some((fs, worktree_abs_path, path_style)) = this.read_with(cx, |this, _cx| {
1750 let local_worktree = match this {
1751 Worktree::Local(local_worktree) => local_worktree,
1752 Worktree::Remote(_) => return None,
1753 };
1754
1755 let fs = local_worktree.fs.clone();
1756 let path_style = local_worktree.path_style();
1757 Some((fs, Arc::clone(local_worktree.abs_path()), path_style))
1758 }) else {
1759 return Err(anyhow!("Localworktree should not change into a remote one"));
1760 };
1761
1762 let path_buf = fs.restore(trash_entry).await?;
1763 let path = path_buf
1764 .strip_prefix(worktree_abs_path)
1765 .context("Could not strip prefix")?;
1766 let path = RelPath::new(&path, path_style)?;
1767 let path = path.into_owned();
1768
1769 Ok(path)
1770 }
1771
1772 pub fn copy_external_entries(
1773 &self,
1774 target_directory: Arc<RelPath>,
1775 paths: Vec<Arc<Path>>,
1776 cx: &Context<Worktree>,
1777 ) -> Task<Result<Vec<ProjectEntryId>>> {
1778 let target_directory = self.absolutize(&target_directory);
1779 let worktree_path = self.abs_path().clone();
1780 let fs = self.fs.clone();
1781 let paths = paths
1782 .into_iter()
1783 .filter_map(|source| {
1784 let file_name = source.file_name()?;
1785 let mut target = target_directory.clone();
1786 target.push(file_name);
1787
1788 // Do not allow copying the same file to itself.
1789 if source.as_ref() != target.as_path() {
1790 Some((source, target))
1791 } else {
1792 None
1793 }
1794 })
1795 .collect::<Vec<_>>();
1796
1797 let paths_to_refresh = paths
1798 .iter()
1799 .filter_map(|(_, target)| {
1800 RelPath::new(
1801 target.strip_prefix(&worktree_path).ok()?,
1802 PathStyle::local(),
1803 )
1804 .ok()
1805 .map(|path| path.into_arc())
1806 })
1807 .collect::<Vec<_>>();
1808
1809 cx.spawn(async move |this, cx| {
1810 cx.background_spawn(async move {
1811 for (source, target) in paths {
1812 copy_recursive(
1813 fs.as_ref(),
1814 &source,
1815 &target,
1816 fs::CopyOptions {
1817 overwrite: true,
1818 ..Default::default()
1819 },
1820 )
1821 .await
1822 .with_context(|| {
1823 format!("Failed to copy file from {source:?} to {target:?}")
1824 })?;
1825 }
1826 anyhow::Ok(())
1827 })
1828 .await
1829 .log_err();
1830 let mut refresh = cx.read_entity(
1831 &this.upgrade().with_context(|| "Dropped worktree")?,
1832 |this, _| {
1833 anyhow::Ok::<postage::barrier::Receiver>(
1834 this.as_local()
1835 .with_context(|| "Worktree is not local")?
1836 .refresh_entries_for_paths(paths_to_refresh.clone()),
1837 )
1838 },
1839 )?;
1840
1841 cx.background_spawn(async move {
1842 refresh.next().await;
1843 anyhow::Ok(())
1844 })
1845 .await
1846 .log_err();
1847
1848 let this = this.upgrade().with_context(|| "Dropped worktree")?;
1849 Ok(cx.read_entity(&this, |this, _| {
1850 paths_to_refresh
1851 .iter()
1852 .filter_map(|path| Some(this.entry_for_path(path)?.id))
1853 .collect()
1854 }))
1855 })
1856 }
1857
1858 fn expand_entry(
1859 &self,
1860 entry_id: ProjectEntryId,
1861 cx: &Context<Worktree>,
1862 ) -> Option<Task<Result<()>>> {
1863 let path = self.entry_for_id(entry_id)?.path.clone();
1864 let mut refresh = self.refresh_entries_for_paths(vec![path]);
1865 Some(cx.background_spawn(async move {
1866 refresh.next().await;
1867 Ok(())
1868 }))
1869 }
1870
1871 fn expand_all_for_entry(
1872 &self,
1873 entry_id: ProjectEntryId,
1874 cx: &Context<Worktree>,
1875 ) -> Option<Task<Result<()>>> {
1876 let path = self.entry_for_id(entry_id).unwrap().path.clone();
1877 let mut rx = self.add_path_prefix_to_scan(path);
1878 Some(cx.background_spawn(async move {
1879 rx.next().await;
1880 Ok(())
1881 }))
1882 }
1883
1884 pub fn refresh_entries_for_paths(&self, paths: Vec<Arc<RelPath>>) -> barrier::Receiver {
1885 let (tx, rx) = barrier::channel();
1886 self.scan_requests_tx
1887 .try_send(ScanRequest {
1888 relative_paths: paths,
1889 done: smallvec![tx],
1890 })
1891 .ok();
1892 rx
1893 }
1894
1895 #[cfg(feature = "test-support")]
1896 pub fn manually_refresh_entries_for_paths(
1897 &self,
1898 paths: Vec<Arc<RelPath>>,
1899 ) -> barrier::Receiver {
1900 self.refresh_entries_for_paths(paths)
1901 }
1902
1903 pub fn add_path_prefix_to_scan(&self, path_prefix: Arc<RelPath>) -> barrier::Receiver {
1904 let (tx, rx) = barrier::channel();
1905 self.path_prefixes_to_scan_tx
1906 .try_send(PathPrefixScanRequest {
1907 path: path_prefix,
1908 done: smallvec![tx],
1909 })
1910 .ok();
1911 rx
1912 }
1913
1914 pub fn refresh_entry(
1915 &self,
1916 path: Arc<RelPath>,
1917 old_path: Option<Arc<RelPath>>,
1918 cx: &Context<Worktree>,
1919 ) -> Task<Result<Option<Entry>>> {
1920 if self.settings.is_path_excluded(&path) {
1921 return Task::ready(Ok(None));
1922 }
1923 let paths = if let Some(old_path) = old_path.as_ref() {
1924 vec![old_path.clone(), path.clone()]
1925 } else {
1926 vec![path.clone()]
1927 };
1928 let t0 = Instant::now();
1929 let mut refresh = self.refresh_entries_for_paths(paths);
1930 // todo(lw): Hot foreground spawn
1931 cx.spawn(async move |this, cx| {
1932 refresh.recv().await;
1933 log::trace!("refreshed entry {path:?} in {:?}", t0.elapsed());
1934 let new_entry = this.read_with(cx, |this, _| {
1935 this.entry_for_path(&path).cloned().with_context(|| {
1936 format!("Could not find entry in worktree for {path:?} after refresh")
1937 })
1938 })??;
1939 Ok(Some(new_entry))
1940 })
1941 }
1942
1943 pub fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
1944 where
1945 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
1946 Fut: 'static + Send + Future<Output = bool>,
1947 {
1948 if let Some(observer) = self.update_observer.as_mut() {
1949 *observer.resume_updates.borrow_mut() = ();
1950 return;
1951 }
1952
1953 let (resume_updates_tx, mut resume_updates_rx) = watch::channel::<()>();
1954 let (snapshots_tx, mut snapshots_rx) =
1955 mpsc::unbounded::<(LocalSnapshot, UpdatedEntriesSet)>();
1956 snapshots_tx
1957 .unbounded_send((self.snapshot(), Arc::default()))
1958 .ok();
1959
1960 let worktree_id = self.id.to_proto();
1961 let _maintain_remote_snapshot = cx.background_spawn(async move {
1962 let mut is_first = true;
1963 while let Some((snapshot, entry_changes)) = snapshots_rx.next().await {
1964 let update = if is_first {
1965 is_first = false;
1966 snapshot.build_initial_update(project_id, worktree_id)
1967 } else {
1968 snapshot.build_update(project_id, worktree_id, entry_changes)
1969 };
1970
1971 for update in proto::split_worktree_update(update) {
1972 let _ = resume_updates_rx.try_recv();
1973 loop {
1974 let result = callback(update.clone());
1975 if result.await {
1976 break;
1977 } else {
1978 log::info!("waiting to resume updates");
1979 if resume_updates_rx.next().await.is_none() {
1980 return Some(());
1981 }
1982 }
1983 }
1984 }
1985 }
1986 Some(())
1987 });
1988
1989 self.update_observer = Some(UpdateObservationState {
1990 snapshots_tx,
1991 resume_updates: resume_updates_tx,
1992 _maintain_remote_snapshot,
1993 });
1994 }
1995
1996 pub fn share_private_files(&mut self, cx: &Context<Worktree>) {
1997 self.share_private_files = true;
1998 self.restart_background_scanners(cx);
1999 }
2000
2001 pub fn update_abs_path_and_refresh(
2002 &mut self,
2003 new_path: Arc<SanitizedPath>,
2004 cx: &Context<Worktree>,
2005 ) {
2006 self.snapshot.git_repositories = Default::default();
2007 self.snapshot.ignores_by_parent_abs_path = Default::default();
2008 let root_name = new_path
2009 .as_path()
2010 .file_name()
2011 .and_then(|f| f.to_str())
2012 .map_or(RelPath::empty().into(), |f| {
2013 RelPath::unix(f).unwrap().into()
2014 });
2015 self.snapshot.update_abs_path(new_path, root_name);
2016 self.restart_background_scanners(cx);
2017 }
2018 #[cfg(feature = "test-support")]
2019 pub fn repositories(&self) -> Vec<Arc<Path>> {
2020 self.git_repositories
2021 .values()
2022 .map(|entry| entry.work_directory_abs_path.clone())
2023 .collect::<Vec<_>>()
2024 }
2025}
2026
2027impl RemoteWorktree {
2028 pub fn project_id(&self) -> u64 {
2029 self.project_id
2030 }
2031
2032 pub fn client(&self) -> AnyProtoClient {
2033 self.client.clone()
2034 }
2035
2036 pub fn disconnected_from_host(&mut self) {
2037 self.updates_tx.take();
2038 self.snapshot_subscriptions.clear();
2039 self.disconnected = true;
2040 }
2041
2042 pub fn update_from_remote(&self, update: proto::UpdateWorktree) {
2043 if let Some(updates_tx) = &self.updates_tx {
2044 updates_tx
2045 .unbounded_send(update)
2046 .expect("consumer runs to completion");
2047 }
2048 }
2049
2050 fn observe_updates<F, Fut>(&mut self, project_id: u64, cx: &Context<Worktree>, callback: F)
2051 where
2052 F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut,
2053 Fut: 'static + Send + Future<Output = bool>,
2054 {
2055 let (tx, mut rx) = mpsc::unbounded();
2056 let initial_update = self
2057 .snapshot
2058 .build_initial_update(project_id, self.id().to_proto());
2059 self.update_observer = Some(tx);
2060 cx.spawn(async move |this, cx| {
2061 let mut update = initial_update;
2062 'outer: loop {
2063 // SSH projects use a special project ID of 0, and we need to
2064 // remap it to the correct one here.
2065 update.project_id = project_id;
2066
2067 for chunk in split_worktree_update(update) {
2068 if !callback(chunk).await {
2069 break 'outer;
2070 }
2071 }
2072
2073 if let Some(next_update) = rx.next().await {
2074 update = next_update;
2075 } else {
2076 break;
2077 }
2078 }
2079 this.update(cx, |this, _| {
2080 let this = this.as_remote_mut().unwrap();
2081 this.update_observer.take();
2082 })
2083 })
2084 .detach();
2085 }
2086
2087 fn observed_snapshot(&self, scan_id: usize) -> bool {
2088 self.completed_scan_id >= scan_id
2089 }
2090
2091 pub fn wait_for_snapshot(
2092 &mut self,
2093 scan_id: usize,
2094 ) -> impl Future<Output = Result<()>> + use<> {
2095 let (tx, rx) = oneshot::channel();
2096 if self.observed_snapshot(scan_id) {
2097 let _ = tx.send(());
2098 } else if self.disconnected {
2099 drop(tx);
2100 } else {
2101 match self
2102 .snapshot_subscriptions
2103 .binary_search_by_key(&scan_id, |probe| probe.0)
2104 {
2105 Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)),
2106 }
2107 }
2108
2109 async move {
2110 rx.await?;
2111 Ok(())
2112 }
2113 }
2114
2115 pub fn insert_entry(
2116 &mut self,
2117 entry: proto::Entry,
2118 scan_id: usize,
2119 cx: &Context<Worktree>,
2120 ) -> Task<Result<Entry>> {
2121 let wait_for_snapshot = self.wait_for_snapshot(scan_id);
2122 cx.spawn(async move |this, cx| {
2123 wait_for_snapshot.await?;
2124 this.update(cx, |worktree, _| {
2125 let worktree = worktree.as_remote_mut().unwrap();
2126 let snapshot = &mut worktree.background_snapshot.lock().0;
2127 let entry = snapshot.insert_entry(entry, &worktree.file_scan_inclusions);
2128 worktree.snapshot = snapshot.clone();
2129 entry
2130 })?
2131 })
2132 }
2133
2134 fn trash_entry(
2135 &self,
2136 entry_id: ProjectEntryId,
2137 cx: &Context<Worktree>,
2138 ) -> Task<Result<TrashId>> {
2139 let response = self.client.request(proto::TrashProjectEntry {
2140 project_id: self.project_id,
2141 entry_id: entry_id.to_proto(),
2142 });
2143
2144 cx.spawn(async move |this, cx| {
2145 let response = response.await?;
2146 let scan_id = response.worktree_scan_id as usize;
2147 let trash_id = response.trash_id;
2148
2149 this.update(cx, move |this, _| {
2150 this.as_remote_mut().unwrap().wait_for_snapshot(scan_id)
2151 })?
2152 .await?;
2153
2154 this.update(cx, |this, _| {
2155 let this = this.as_remote_mut().unwrap();
2156 let snapshot = &mut this.background_snapshot.lock().0;
2157 snapshot.delete_entry(entry_id);
2158 this.snapshot = snapshot.clone();
2159 })?;
2160
2161 Ok(TrashId::from_u64(trash_id))
2162 })
2163 }
2164
2165 fn delete_entry(&self, entry_id: ProjectEntryId, cx: &Context<Worktree>) -> Task<Result<()>> {
2166 let response = self.client.request(proto::DeleteProjectEntry {
2167 project_id: self.project_id,
2168 entry_id: entry_id.to_proto(),
2169 use_trash: false, // deprecated
2170 });
2171
2172 cx.spawn(async move |this, cx| {
2173 let response = response.await?;
2174 let scan_id = response.worktree_scan_id as usize;
2175
2176 this.update(cx, move |this, _| {
2177 this.as_remote_mut().unwrap().wait_for_snapshot(scan_id)
2178 })?
2179 .await?;
2180
2181 this.update(cx, |this, _| {
2182 let this = this.as_remote_mut().unwrap();
2183 let snapshot = &mut this.background_snapshot.lock().0;
2184 snapshot.delete_entry(entry_id);
2185 this.snapshot = snapshot.clone();
2186 })
2187 })
2188 }
2189
2190 fn copy_external_entries(
2191 &self,
2192 target_directory: Arc<RelPath>,
2193 paths_to_copy: Vec<Arc<Path>>,
2194 local_fs: Arc<dyn Fs>,
2195 cx: &Context<Worktree>,
2196 ) -> Task<anyhow::Result<Vec<ProjectEntryId>>> {
2197 let client = self.client.clone();
2198 let worktree_id = self.id().to_proto();
2199 let project_id = self.project_id;
2200
2201 cx.background_spawn(async move {
2202 let mut requests = Vec::new();
2203 for root_path_to_copy in paths_to_copy {
2204 let Some(filename) = root_path_to_copy
2205 .file_name()
2206 .and_then(|name| name.to_str())
2207 .and_then(|filename| RelPath::unix(filename).ok())
2208 else {
2209 continue;
2210 };
2211 for (abs_path, is_directory) in
2212 read_dir_items(local_fs.as_ref(), &root_path_to_copy).await?
2213 {
2214 let Some(relative_path) = abs_path
2215 .strip_prefix(&root_path_to_copy)
2216 .map_err(|e| anyhow::Error::from(e))
2217 .and_then(|relative_path| RelPath::new(relative_path, PathStyle::local()))
2218 .log_err()
2219 else {
2220 continue;
2221 };
2222 let content = if is_directory {
2223 None
2224 } else {
2225 Some(local_fs.load_bytes(&abs_path).await?)
2226 };
2227
2228 let mut target_path = target_directory.join(filename);
2229 if relative_path.file_name().is_some() {
2230 target_path = target_path.join(&relative_path);
2231 }
2232
2233 requests.push(proto::CreateProjectEntry {
2234 project_id,
2235 worktree_id,
2236 path: target_path.to_proto(),
2237 is_directory,
2238 content,
2239 });
2240 }
2241 }
2242 requests.sort_unstable_by(|a, b| a.path.cmp(&b.path));
2243 requests.dedup();
2244
2245 let mut copied_entry_ids = Vec::new();
2246 for request in requests {
2247 let response = client.request(request).await?;
2248 copied_entry_ids.extend(response.entry.map(|e| ProjectEntryId::from_proto(e.id)));
2249 }
2250
2251 Ok(copied_entry_ids)
2252 })
2253 }
2254}
2255
2256impl Snapshot {
2257 pub fn new(
2258 id: WorktreeId,
2259 root_name: Arc<RelPath>,
2260 abs_path: Arc<Path>,
2261 path_style: PathStyle,
2262 ) -> Self {
2263 Snapshot {
2264 id,
2265 abs_path: SanitizedPath::from_arc(abs_path),
2266 path_style,
2267 root_char_bag: root_name
2268 .as_unix_str()
2269 .chars()
2270 .map(|c| c.to_ascii_lowercase())
2271 .collect(),
2272 root_name,
2273 always_included_entries: Default::default(),
2274 entries_by_path: Default::default(),
2275 entries_by_id: Default::default(),
2276 root_repo_common_dir: None,
2277 scan_id: 1,
2278 completed_scan_id: 0,
2279 }
2280 }
2281
2282 pub fn id(&self) -> WorktreeId {
2283 self.id
2284 }
2285
2286 // TODO:
2287 // Consider the following:
2288 //
2289 // ```rust
2290 // let abs_path: Arc<Path> = snapshot.abs_path(); // e.g. "C:\Users\user\Desktop\project"
2291 // let some_non_trimmed_path = Path::new("\\\\?\\C:\\Users\\user\\Desktop\\project\\main.rs");
2292 // // The caller perform some actions here:
2293 // some_non_trimmed_path.strip_prefix(abs_path); // This fails
2294 // some_non_trimmed_path.starts_with(abs_path); // This fails too
2295 // ```
2296 //
2297 // This is definitely a bug, but it's not clear if we should handle it here or not.
2298 pub fn abs_path(&self) -> &Arc<Path> {
2299 SanitizedPath::cast_arc_ref(&self.abs_path)
2300 }
2301
2302 pub fn root_repo_common_dir(&self) -> Option<&Arc<Path>> {
2303 self.root_repo_common_dir
2304 .as_ref()
2305 .map(SanitizedPath::cast_arc_ref)
2306 }
2307
2308 fn build_initial_update(&self, project_id: u64, worktree_id: u64) -> proto::UpdateWorktree {
2309 let mut updated_entries = self
2310 .entries_by_path
2311 .iter()
2312 .map(proto::Entry::from)
2313 .collect::<Vec<_>>();
2314 updated_entries.sort_unstable_by_key(|e| e.id);
2315
2316 proto::UpdateWorktree {
2317 project_id,
2318 worktree_id,
2319 abs_path: self.abs_path().to_string_lossy().into_owned(),
2320 root_name: self.root_name().to_proto(),
2321 root_repo_common_dir: self
2322 .root_repo_common_dir()
2323 .map(|p| p.to_string_lossy().into_owned()),
2324 updated_entries,
2325 removed_entries: Vec::new(),
2326 scan_id: self.scan_id as u64,
2327 is_last_update: self.completed_scan_id == self.scan_id,
2328 // Sent in separate messages.
2329 updated_repositories: Vec::new(),
2330 removed_repositories: Vec::new(),
2331 }
2332 }
2333
2334 pub fn work_directory_abs_path(&self, work_directory: &WorkDirectory) -> PathBuf {
2335 match work_directory {
2336 WorkDirectory::InProject { relative_path } => self.absolutize(relative_path),
2337 WorkDirectory::AboveProject { absolute_path, .. } => absolute_path.as_ref().to_owned(),
2338 }
2339 }
2340
2341 pub fn absolutize(&self, path: &RelPath) -> PathBuf {
2342 if path.file_name().is_some() {
2343 let mut abs_path = self.abs_path.to_string();
2344 for component in path.components() {
2345 if !abs_path.ends_with(self.path_style.primary_separator()) {
2346 abs_path.push_str(self.path_style.primary_separator());
2347 }
2348 abs_path.push_str(component);
2349 }
2350 PathBuf::from(abs_path)
2351 } else {
2352 self.abs_path.as_path().to_path_buf()
2353 }
2354 }
2355
2356 pub fn contains_entry(&self, entry_id: ProjectEntryId) -> bool {
2357 self.entries_by_id.get(&entry_id, ()).is_some()
2358 }
2359
2360 fn insert_entry(
2361 &mut self,
2362 entry: proto::Entry,
2363 always_included_paths: &PathMatcher,
2364 ) -> Result<Entry> {
2365 let entry = Entry::try_from((&self.root_char_bag, always_included_paths, entry))?;
2366 let old_entry = self.entries_by_id.insert_or_replace(
2367 PathEntry {
2368 id: entry.id,
2369 path: entry.path.clone(),
2370 is_ignored: entry.is_ignored,
2371 scan_id: 0,
2372 },
2373 (),
2374 );
2375 if let Some(old_entry) = old_entry {
2376 self.entries_by_path.remove(&PathKey(old_entry.path), ());
2377 }
2378 self.entries_by_path.insert_or_replace(entry.clone(), ());
2379 Ok(entry)
2380 }
2381
2382 fn delete_entry(&mut self, entry_id: ProjectEntryId) -> Option<Arc<RelPath>> {
2383 let removed_entry = self.entries_by_id.remove(&entry_id, ())?;
2384 self.entries_by_path = {
2385 let mut cursor = self.entries_by_path.cursor::<TraversalProgress>(());
2386 let mut new_entries_by_path =
2387 cursor.slice(&TraversalTarget::path(&removed_entry.path), Bias::Left);
2388 while let Some(entry) = cursor.item() {
2389 if entry.path.starts_with(&removed_entry.path) {
2390 self.entries_by_id.remove(&entry.id, ());
2391 cursor.next();
2392 } else {
2393 break;
2394 }
2395 }
2396 new_entries_by_path.append(cursor.suffix(), ());
2397 new_entries_by_path
2398 };
2399
2400 Some(removed_entry.path)
2401 }
2402
2403 fn update_abs_path(&mut self, abs_path: Arc<SanitizedPath>, root_name: Arc<RelPath>) {
2404 self.abs_path = abs_path;
2405 if root_name != self.root_name {
2406 self.root_char_bag = root_name
2407 .as_unix_str()
2408 .chars()
2409 .map(|c| c.to_ascii_lowercase())
2410 .collect();
2411 self.root_name = root_name;
2412 }
2413 }
2414
2415 pub fn apply_remote_update(
2416 &mut self,
2417 update: proto::UpdateWorktree,
2418 always_included_paths: &PathMatcher,
2419 ) {
2420 log::debug!(
2421 "applying remote worktree update. {} entries updated, {} removed",
2422 update.updated_entries.len(),
2423 update.removed_entries.len()
2424 );
2425 if let Some(root_name) = RelPath::from_proto(&update.root_name).log_err() {
2426 self.update_abs_path(
2427 SanitizedPath::new_arc(&Path::new(&update.abs_path)),
2428 root_name,
2429 );
2430 }
2431
2432 let mut entries_by_path_edits = Vec::new();
2433 let mut entries_by_id_edits = Vec::new();
2434
2435 for entry_id in update.removed_entries {
2436 let entry_id = ProjectEntryId::from_proto(entry_id);
2437 entries_by_id_edits.push(Edit::Remove(entry_id));
2438 if let Some(entry) = self.entry_for_id(entry_id) {
2439 entries_by_path_edits.push(Edit::Remove(PathKey(entry.path.clone())));
2440 }
2441 }
2442
2443 for entry in update.updated_entries {
2444 let Some(entry) =
2445 Entry::try_from((&self.root_char_bag, always_included_paths, entry)).log_err()
2446 else {
2447 continue;
2448 };
2449 if let Some(PathEntry { path, .. }) = self.entries_by_id.get(&entry.id, ()) {
2450 entries_by_path_edits.push(Edit::Remove(PathKey(path.clone())));
2451 }
2452 if let Some(old_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2453 && old_entry.id != entry.id
2454 {
2455 entries_by_id_edits.push(Edit::Remove(old_entry.id));
2456 }
2457 entries_by_id_edits.push(Edit::Insert(PathEntry {
2458 id: entry.id,
2459 path: entry.path.clone(),
2460 is_ignored: entry.is_ignored,
2461 scan_id: 0,
2462 }));
2463 entries_by_path_edits.push(Edit::Insert(entry));
2464 }
2465
2466 self.entries_by_path.edit(entries_by_path_edits, ());
2467 self.entries_by_id.edit(entries_by_id_edits, ());
2468
2469 if let Some(dir) = update
2470 .root_repo_common_dir
2471 .map(|p| SanitizedPath::new_arc(Path::new(&p)))
2472 {
2473 self.root_repo_common_dir = Some(dir);
2474 }
2475
2476 self.scan_id = update.scan_id as usize;
2477 /// TODO!(yara) batch the scans
2478 if update.is_last_update {
2479 self.completed_scan_id = update.scan_id as usize;
2480 }
2481 }
2482
2483 pub fn entry_count(&self) -> usize {
2484 self.entries_by_path.summary().count
2485 }
2486
2487 pub fn visible_entry_count(&self) -> usize {
2488 self.entries_by_path.summary().non_ignored_count
2489 }
2490
2491 pub fn dir_count(&self) -> usize {
2492 let summary = self.entries_by_path.summary();
2493 summary.count - summary.file_count
2494 }
2495
2496 pub fn visible_dir_count(&self) -> usize {
2497 let summary = self.entries_by_path.summary();
2498 summary.non_ignored_count - summary.non_ignored_file_count
2499 }
2500
2501 pub fn file_count(&self) -> usize {
2502 self.entries_by_path.summary().file_count
2503 }
2504
2505 pub fn visible_file_count(&self) -> usize {
2506 self.entries_by_path.summary().non_ignored_file_count
2507 }
2508
2509 fn traverse_from_offset(
2510 &self,
2511 include_files: bool,
2512 include_dirs: bool,
2513 include_ignored: bool,
2514 start_offset: usize,
2515 ) -> Traversal<'_> {
2516 let mut cursor = self.entries_by_path.cursor(());
2517 cursor.seek(
2518 &TraversalTarget::Count {
2519 count: start_offset,
2520 include_files,
2521 include_dirs,
2522 include_ignored,
2523 },
2524 Bias::Right,
2525 );
2526 Traversal {
2527 snapshot: self,
2528 cursor,
2529 include_files,
2530 include_dirs,
2531 include_ignored,
2532 }
2533 }
2534
2535 pub fn traverse_from_path(
2536 &self,
2537 include_files: bool,
2538 include_dirs: bool,
2539 include_ignored: bool,
2540 path: &RelPath,
2541 ) -> Traversal<'_> {
2542 Traversal::new(self, include_files, include_dirs, include_ignored, path)
2543 }
2544
2545 pub fn files(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2546 self.traverse_from_offset(true, false, include_ignored, start)
2547 }
2548
2549 pub fn directories(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2550 self.traverse_from_offset(false, true, include_ignored, start)
2551 }
2552
2553 pub fn entries(&self, include_ignored: bool, start: usize) -> Traversal<'_> {
2554 self.traverse_from_offset(true, true, include_ignored, start)
2555 }
2556
2557 pub fn paths(&self) -> impl Iterator<Item = &RelPath> {
2558 self.entries_by_path
2559 .cursor::<()>(())
2560 .filter(move |entry| !entry.path.is_empty())
2561 .map(|entry| entry.path.as_ref())
2562 }
2563
2564 pub fn child_entries<'a>(&'a self, parent_path: &'a RelPath) -> ChildEntriesIter<'a> {
2565 let options = ChildEntriesOptions {
2566 include_files: true,
2567 include_dirs: true,
2568 include_ignored: true,
2569 };
2570 self.child_entries_with_options(parent_path, options)
2571 }
2572
2573 pub fn child_entries_with_options<'a>(
2574 &'a self,
2575 parent_path: &'a RelPath,
2576 options: ChildEntriesOptions,
2577 ) -> ChildEntriesIter<'a> {
2578 let mut cursor = self.entries_by_path.cursor(());
2579 cursor.seek(&TraversalTarget::path(parent_path), Bias::Right);
2580 let traversal = Traversal {
2581 snapshot: self,
2582 cursor,
2583 include_files: options.include_files,
2584 include_dirs: options.include_dirs,
2585 include_ignored: options.include_ignored,
2586 };
2587 ChildEntriesIter {
2588 traversal,
2589 parent_path,
2590 }
2591 }
2592
2593 pub fn root_entry(&self) -> Option<&Entry> {
2594 self.entries_by_path.first()
2595 }
2596
2597 /// Returns `None` for a single file worktree, or `Some(self.abs_path())` if
2598 /// it is a directory.
2599 pub fn root_dir(&self) -> Option<Arc<Path>> {
2600 self.root_entry()
2601 .filter(|entry| entry.is_dir())
2602 .map(|_| self.abs_path().clone())
2603 }
2604
2605 pub fn root_name(&self) -> &RelPath {
2606 &self.root_name
2607 }
2608
2609 pub fn root_name_str(&self) -> &str {
2610 self.root_name.as_unix_str()
2611 }
2612
2613 pub fn scan_id(&self) -> usize {
2614 self.scan_id
2615 }
2616
2617 pub fn entry_for_path(&self, path: &RelPath) -> Option<&Entry> {
2618 let entry = self.traverse_from_path(true, true, true, path).entry();
2619 entry.and_then(|entry| {
2620 if entry.path.as_ref() == path {
2621 Some(entry)
2622 } else {
2623 None
2624 }
2625 })
2626 }
2627
2628 /// Resolves a path to an executable using the following heuristics:
2629 ///
2630 /// 1. If the path starts with `~`, it is expanded to the user's home directory.
2631 /// 2. If the path is relative and contains more than one component,
2632 /// it is joined to the worktree root path.
2633 /// 3. If the path is relative and exists in the worktree
2634 /// (even if falls under an exclusion filter),
2635 /// it is joined to the worktree root path.
2636 /// 4. Otherwise the path is returned unmodified.
2637 ///
2638 /// Relative paths that do not exist in the worktree may
2639 /// still be found using the `PATH` environment variable.
2640 pub fn resolve_relative_path(&self, path: PathBuf) -> PathBuf {
2641 if let Some(path_str) = path.to_str() {
2642 if let Some(remaining_path) = path_str.strip_prefix("~/") {
2643 return home_dir().join(remaining_path);
2644 } else if path_str == "~" {
2645 return home_dir().to_path_buf();
2646 }
2647 }
2648
2649 if let Ok(rel_path) = RelPath::new(&path, self.path_style)
2650 && (path.components().count() > 1 || self.entry_for_path(&rel_path).is_some())
2651 {
2652 self.abs_path().join(path)
2653 } else {
2654 path
2655 }
2656 }
2657
2658 pub fn entry_for_id(&self, id: ProjectEntryId) -> Option<&Entry> {
2659 let entry = self.entries_by_id.get(&id, ())?;
2660 self.entry_for_path(&entry.path)
2661 }
2662
2663 pub fn path_style(&self) -> PathStyle {
2664 self.path_style
2665 }
2666}
2667
2668impl LocalSnapshot {
2669 fn local_repo_for_work_directory_path(&self, path: &RelPath) -> Option<&LocalRepositoryEntry> {
2670 self.git_repositories
2671 .iter()
2672 .map(|(_, entry)| entry)
2673 .find(|entry| entry.work_directory.path_key() == PathKey(path.into()))
2674 }
2675
2676 fn build_update(
2677 &self,
2678 project_id: u64,
2679 worktree_id: u64,
2680 entry_changes: UpdatedEntriesSet,
2681 ) -> proto::UpdateWorktree {
2682 let mut updated_entries = Vec::new();
2683 let mut removed_entries = Vec::new();
2684
2685 for (_, entry_id, path_change) in entry_changes.iter() {
2686 if let PathChange::Removed = path_change {
2687 removed_entries.push(entry_id.0 as u64);
2688 } else if let Some(entry) = self.entry_for_id(*entry_id) {
2689 updated_entries.push(proto::Entry::from(entry));
2690 }
2691 }
2692
2693 removed_entries.sort_unstable();
2694 updated_entries.sort_unstable_by_key(|e| e.id);
2695
2696 // TODO - optimize, knowing that removed_entries are sorted.
2697 removed_entries.retain(|id| updated_entries.binary_search_by_key(id, |e| e.id).is_err());
2698
2699 proto::UpdateWorktree {
2700 project_id,
2701 worktree_id,
2702 abs_path: self.abs_path().to_string_lossy().into_owned(),
2703 root_name: self.root_name().to_proto(),
2704 root_repo_common_dir: self
2705 .root_repo_common_dir()
2706 .map(|p| p.to_string_lossy().into_owned()),
2707 updated_entries,
2708 removed_entries,
2709 scan_id: self.scan_id as u64,
2710 is_last_update: self.completed_scan_id == self.scan_id,
2711 // Sent in separate messages.
2712 updated_repositories: Vec::new(),
2713 removed_repositories: Vec::new(),
2714 }
2715 }
2716
2717 async fn insert_entry(&mut self, mut entry: Entry, fs: &dyn Fs) -> Entry {
2718 log::trace!("insert entry {:?}", entry.path);
2719 if entry.is_file() && entry.path.file_name() == Some(&GITIGNORE) {
2720 let abs_path = self.absolutize(&entry.path);
2721 match build_gitignore(&abs_path, fs).await {
2722 Ok(ignore) => {
2723 self.ignores_by_parent_abs_path
2724 .insert(abs_path.parent().unwrap().into(), (Arc::new(ignore), true));
2725 }
2726 Err(error) => {
2727 log::error!(
2728 "error loading .gitignore file {:?} - {:?}",
2729 &entry.path,
2730 error
2731 );
2732 }
2733 }
2734 }
2735
2736 if entry.kind == EntryKind::PendingDir
2737 && let Some(existing_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ())
2738 {
2739 entry.kind = existing_entry.kind;
2740 }
2741
2742 let scan_id = self.scan_id;
2743 let removed = self.entries_by_path.insert_or_replace(entry.clone(), ());
2744 if let Some(removed) = removed
2745 && removed.id != entry.id
2746 {
2747 self.entries_by_id.remove(&removed.id, ());
2748 }
2749 self.entries_by_id.insert_or_replace(
2750 PathEntry {
2751 id: entry.id,
2752 path: entry.path.clone(),
2753 is_ignored: entry.is_ignored,
2754 scan_id,
2755 },
2756 (),
2757 );
2758
2759 entry
2760 }
2761
2762 fn ancestor_inodes_for_path(&self, path: &RelPath) -> TreeSet<u64> {
2763 let mut inodes = TreeSet::default();
2764 for ancestor in path.ancestors().skip(1) {
2765 if let Some(entry) = self.entry_for_path(ancestor) {
2766 inodes.insert(entry.inode);
2767 }
2768 }
2769 inodes
2770 }
2771
2772 async fn ignore_stack_for_abs_path(
2773 &self,
2774 abs_path: &Path,
2775 is_dir: bool,
2776 fs: &dyn Fs,
2777 ) -> IgnoreStack {
2778 let mut new_ignores = Vec::new();
2779 let mut repo_root = None;
2780 for (index, ancestor) in abs_path.ancestors().enumerate() {
2781 if index > 0 {
2782 if let Some((ignore, _)) = self.ignores_by_parent_abs_path.get(ancestor) {
2783 new_ignores.push((ancestor, Some(ignore.clone())));
2784 } else {
2785 new_ignores.push((ancestor, None));
2786 }
2787 }
2788
2789 let metadata = fs.metadata(&ancestor.join(DOT_GIT)).await.ok().flatten();
2790 if metadata.is_some() {
2791 repo_root = Some(Arc::from(ancestor));
2792 break;
2793 }
2794 }
2795
2796 let mut ignore_stack = if let Some(global_gitignore) = self.global_gitignore.clone() {
2797 IgnoreStack::global(global_gitignore)
2798 } else {
2799 IgnoreStack::none()
2800 };
2801
2802 if let Some((repo_exclude, _)) = repo_root
2803 .as_ref()
2804 .and_then(|abs_path| self.repo_exclude_by_work_dir_abs_path.get(abs_path))
2805 {
2806 ignore_stack = ignore_stack.append(IgnoreKind::RepoExclude, repo_exclude.clone());
2807 }
2808 ignore_stack.repo_root = repo_root;
2809 for (parent_abs_path, ignore) in new_ignores.into_iter().rev() {
2810 if ignore_stack.is_abs_path_ignored(parent_abs_path, true) {
2811 ignore_stack = IgnoreStack::all();
2812 break;
2813 } else if let Some(ignore) = ignore {
2814 ignore_stack =
2815 ignore_stack.append(IgnoreKind::Gitignore(parent_abs_path.into()), ignore);
2816 }
2817 }
2818
2819 if ignore_stack.is_abs_path_ignored(abs_path, is_dir) {
2820 ignore_stack = IgnoreStack::all();
2821 }
2822
2823 ignore_stack
2824 }
2825
2826 #[cfg(feature = "test-support")]
2827 pub fn expanded_entries(&self) -> impl Iterator<Item = &Entry> {
2828 self.entries_by_path
2829 .cursor::<()>(())
2830 .filter(|entry| entry.kind == EntryKind::Dir && (entry.is_external || entry.is_ignored))
2831 }
2832
2833 #[cfg(feature = "test-support")]
2834 pub fn check_invariants(&self, git_state: bool) {
2835 use pretty_assertions::assert_eq;
2836
2837 assert_eq!(
2838 self.entries_by_path
2839 .cursor::<()>(())
2840 .map(|e| (&e.path, e.id))
2841 .collect::<Vec<_>>(),
2842 self.entries_by_id
2843 .cursor::<()>(())
2844 .map(|e| (&e.path, e.id))
2845 .collect::<collections::BTreeSet<_>>()
2846 .into_iter()
2847 .collect::<Vec<_>>(),
2848 "entries_by_path and entries_by_id are inconsistent"
2849 );
2850
2851 let mut files = self.files(true, 0);
2852 let mut visible_files = self.files(false, 0);
2853 for entry in self.entries_by_path.cursor::<()>(()) {
2854 if entry.is_file() {
2855 assert_eq!(files.next().unwrap().inode, entry.inode);
2856 if !entry.is_ignored || entry.is_always_included {
2857 assert_eq!(visible_files.next().unwrap().inode, entry.inode);
2858 }
2859 }
2860 }
2861
2862 assert!(files.next().is_none());
2863 assert!(visible_files.next().is_none());
2864
2865 let mut bfs_paths = Vec::new();
2866 let mut stack = self
2867 .root_entry()
2868 .map(|e| e.path.as_ref())
2869 .into_iter()
2870 .collect::<Vec<_>>();
2871 while let Some(path) = stack.pop() {
2872 bfs_paths.push(path);
2873 let ix = stack.len();
2874 for child_entry in self.child_entries(path) {
2875 stack.insert(ix, &child_entry.path);
2876 }
2877 }
2878
2879 let dfs_paths_via_iter = self
2880 .entries_by_path
2881 .cursor::<()>(())
2882 .map(|e| e.path.as_ref())
2883 .collect::<Vec<_>>();
2884 assert_eq!(bfs_paths, dfs_paths_via_iter);
2885
2886 let dfs_paths_via_traversal = self
2887 .entries(true, 0)
2888 .map(|e| e.path.as_ref())
2889 .collect::<Vec<_>>();
2890
2891 assert_eq!(dfs_paths_via_traversal, dfs_paths_via_iter);
2892
2893 if git_state {
2894 for ignore_parent_abs_path in self.ignores_by_parent_abs_path.keys() {
2895 let ignore_parent_path = &RelPath::new(
2896 ignore_parent_abs_path
2897 .strip_prefix(self.abs_path.as_path())
2898 .unwrap(),
2899 PathStyle::local(),
2900 )
2901 .unwrap();
2902 assert!(self.entry_for_path(ignore_parent_path).is_some());
2903 assert!(
2904 self.entry_for_path(
2905 &ignore_parent_path.join(RelPath::unix(GITIGNORE).unwrap())
2906 )
2907 .is_some()
2908 );
2909 }
2910 }
2911 }
2912
2913 #[cfg(feature = "test-support")]
2914 pub fn entries_without_ids(&self, include_ignored: bool) -> Vec<(&RelPath, u64, bool)> {
2915 let mut paths = Vec::new();
2916 for entry in self.entries_by_path.cursor::<()>(()) {
2917 if include_ignored || !entry.is_ignored {
2918 paths.push((entry.path.as_ref(), entry.inode, entry.is_ignored));
2919 }
2920 }
2921 paths.sort_by(|a, b| a.0.cmp(b.0));
2922 paths
2923 }
2924}
2925
2926impl BackgroundScannerState {
2927 fn should_scan_directory(&self, entry: &Entry) -> bool {
2928 (self.scanning_enabled && !entry.is_external && (!entry.is_ignored || entry.is_always_included))
2929 || entry.path.file_name() == Some(DOT_GIT)
2930 || entry.path.file_name() == Some(local_settings_folder_name())
2931 || entry.path.file_name() == Some(local_vscode_folder_name())
2932 || self.scanned_dirs.contains(&entry.id) // If we've ever scanned it, keep scanning
2933 || self
2934 .paths_to_scan
2935 .iter()
2936 .any(|p| p.starts_with(&entry.path))
2937 || self
2938 .path_prefixes_to_scan
2939 .iter()
2940 .any(|p| entry.path.starts_with(p))
2941 }
2942
2943 async fn enqueue_scan_dir(
2944 &self,
2945 abs_path: Arc<Path>,
2946 entry: &Entry,
2947 scan_job_tx: &Sender<ScanJob>,
2948 fs: &dyn Fs,
2949 ) {
2950 let path = entry.path.clone();
2951 let ignore_stack = self
2952 .snapshot
2953 .ignore_stack_for_abs_path(&abs_path, true, fs)
2954 .await;
2955 let mut ancestor_inodes = self.snapshot.ancestor_inodes_for_path(&path);
2956
2957 if !ancestor_inodes.contains(&entry.inode) {
2958 ancestor_inodes.insert(entry.inode);
2959 scan_job_tx
2960 .try_send(ScanJob {
2961 abs_path,
2962 path,
2963 ignore_stack,
2964 scan_queue: scan_job_tx.clone(),
2965 ancestor_inodes,
2966 is_external: entry.is_external,
2967 })
2968 .unwrap();
2969 }
2970 }
2971
2972 fn reuse_entry_id(&mut self, entry: &mut Entry) {
2973 if let Some(mtime) = entry.mtime {
2974 // If an entry with the same inode was removed from the worktree during this scan,
2975 // then it *might* represent the same file or directory. But the OS might also have
2976 // re-used the inode for a completely different file or directory.
2977 //
2978 // Conditionally reuse the old entry's id:
2979 // * if the mtime is the same, the file was probably been renamed.
2980 // * if the path is the same, the file may just have been updated
2981 if let Some(removed_entry) = self.removed_entries.remove(&entry.inode) {
2982 if removed_entry.mtime == Some(mtime) || removed_entry.path == entry.path {
2983 entry.id = removed_entry.id;
2984 }
2985 } else if let Some(existing_entry) = self.snapshot.entry_for_path(&entry.path) {
2986 entry.id = existing_entry.id;
2987 }
2988 }
2989 }
2990
2991 fn entry_id_for(
2992 &mut self,
2993 next_entry_id: &AtomicUsize,
2994 path: &RelPath,
2995 metadata: &fs::Metadata,
2996 ) -> ProjectEntryId {
2997 // If an entry with the same inode was removed from the worktree during this scan,
2998 // then it *might* represent the same file or directory. But the OS might also have
2999 // re-used the inode for a completely different file or directory.
3000 //
3001 // Conditionally reuse the old entry's id:
3002 // * if the mtime is the same, the file was probably been renamed.
3003 // * if the path is the same, the file may just have been updated
3004 if let Some(removed_entry) = self.removed_entries.remove(&metadata.inode) {
3005 if removed_entry.mtime == Some(metadata.mtime) || *removed_entry.path == *path {
3006 return removed_entry.id;
3007 }
3008 } else if let Some(existing_entry) = self.snapshot.entry_for_path(path) {
3009 return existing_entry.id;
3010 }
3011 ProjectEntryId::new(next_entry_id)
3012 }
3013
3014 async fn insert_entry(&mut self, entry: Entry, fs: &dyn Fs, watcher: &dyn Watcher) -> Entry {
3015 let entry = self.snapshot.insert_entry(entry, fs).await;
3016 if entry.path.file_name() == Some(&DOT_GIT) {
3017 self.insert_git_repository(entry.path.clone(), fs, watcher)
3018 .await;
3019 }
3020
3021 #[cfg(feature = "test-support")]
3022 self.snapshot.check_invariants(false);
3023
3024 entry
3025 }
3026
3027 fn populate_dir(
3028 &mut self,
3029 parent_path: Arc<RelPath>,
3030 entries: impl IntoIterator<Item = Entry>,
3031 ignore: Option<Arc<Gitignore>>,
3032 ) {
3033 let mut parent_entry = if let Some(parent_entry) = self
3034 .snapshot
3035 .entries_by_path
3036 .get(&PathKey(parent_path.clone()), ())
3037 {
3038 parent_entry.clone()
3039 } else {
3040 log::warn!(
3041 "populating a directory {:?} that has been removed",
3042 parent_path
3043 );
3044 return;
3045 };
3046
3047 match parent_entry.kind {
3048 EntryKind::PendingDir | EntryKind::UnloadedDir => parent_entry.kind = EntryKind::Dir,
3049 EntryKind::Dir => {}
3050 _ => return,
3051 }
3052
3053 if let Some(ignore) = ignore {
3054 let abs_parent_path = self
3055 .snapshot
3056 .abs_path
3057 .as_path()
3058 .join(parent_path.as_std_path())
3059 .into();
3060 self.snapshot
3061 .ignores_by_parent_abs_path
3062 .insert(abs_parent_path, (ignore, false));
3063 }
3064
3065 let parent_entry_id = parent_entry.id;
3066 self.scanned_dirs.insert(parent_entry_id);
3067 let mut entries_by_path_edits = vec![Edit::Insert(parent_entry)];
3068 let mut entries_by_id_edits = Vec::new();
3069
3070 for entry in entries {
3071 entries_by_id_edits.push(Edit::Insert(PathEntry {
3072 id: entry.id,
3073 path: entry.path.clone(),
3074 is_ignored: entry.is_ignored,
3075 scan_id: self.snapshot.scan_id,
3076 }));
3077 entries_by_path_edits.push(Edit::Insert(entry));
3078 }
3079
3080 self.snapshot
3081 .entries_by_path
3082 .edit(entries_by_path_edits, ());
3083 self.snapshot.entries_by_id.edit(entries_by_id_edits, ());
3084
3085 if let Err(ix) = self.changed_paths.binary_search(&parent_path) {
3086 self.changed_paths.insert(ix, parent_path.clone());
3087 }
3088
3089 #[cfg(feature = "test-support")]
3090 self.snapshot.check_invariants(false);
3091 }
3092
3093 fn remove_path(&mut self, path: &RelPath, watcher: &dyn Watcher) {
3094 log::trace!("background scanner removing path {path:?}");
3095 let mut new_entries;
3096 let removed_entries;
3097 {
3098 let mut cursor = self
3099 .snapshot
3100 .entries_by_path
3101 .cursor::<TraversalProgress>(());
3102 new_entries = cursor.slice(&TraversalTarget::path(path), Bias::Left);
3103 removed_entries = cursor.slice(&TraversalTarget::successor(path), Bias::Left);
3104 new_entries.append(cursor.suffix(), ());
3105 }
3106 self.snapshot.entries_by_path = new_entries;
3107
3108 let mut removed_ids = Vec::with_capacity(removed_entries.summary().count);
3109 let mut removed_dir_abs_paths = Vec::new();
3110 for entry in removed_entries.cursor::<()>(()) {
3111 if entry.is_dir() {
3112 removed_dir_abs_paths.push(self.snapshot.absolutize(&entry.path));
3113 }
3114
3115 match self.removed_entries.entry(entry.inode) {
3116 hash_map::Entry::Occupied(mut e) => {
3117 let prev_removed_entry = e.get_mut();
3118 if entry.id > prev_removed_entry.id {
3119 *prev_removed_entry = entry.clone();
3120 }
3121 }
3122 hash_map::Entry::Vacant(e) => {
3123 e.insert(entry.clone());
3124 }
3125 }
3126
3127 if entry.path.file_name() == Some(GITIGNORE) {
3128 let abs_parent_path = self.snapshot.absolutize(&entry.path.parent().unwrap());
3129 if let Some((_, needs_update)) = self
3130 .snapshot
3131 .ignores_by_parent_abs_path
3132 .get_mut(abs_parent_path.as_path())
3133 {
3134 *needs_update = true;
3135 }
3136 }
3137
3138 if let Err(ix) = removed_ids.binary_search(&entry.id) {
3139 removed_ids.insert(ix, entry.id);
3140 }
3141 }
3142
3143 self.snapshot
3144 .entries_by_id
3145 .edit(removed_ids.iter().map(|&id| Edit::Remove(id)).collect(), ());
3146 self.snapshot
3147 .git_repositories
3148 .retain(|id, _| removed_ids.binary_search(id).is_err());
3149
3150 for removed_dir_abs_path in removed_dir_abs_paths {
3151 watcher.remove(&removed_dir_abs_path).log_err();
3152 }
3153
3154 #[cfg(feature = "test-support")]
3155 self.snapshot.check_invariants(false);
3156 }
3157
3158 async fn insert_git_repository(
3159 &mut self,
3160 dot_git_path: Arc<RelPath>,
3161 fs: &dyn Fs,
3162 watcher: &dyn Watcher,
3163 ) {
3164 let work_dir_path: Arc<RelPath> = match dot_git_path.parent() {
3165 Some(parent_dir) => {
3166 // Guard against repositories inside the repository metadata
3167 if parent_dir
3168 .components()
3169 .any(|component| component == DOT_GIT)
3170 {
3171 log::debug!(
3172 "not building git repository for nested `.git` directory, `.git` path in the worktree: {dot_git_path:?}"
3173 );
3174 return;
3175 };
3176
3177 parent_dir.into()
3178 }
3179 None => {
3180 // `dot_git_path.parent().is_none()` means `.git` directory is the opened worktree itself,
3181 // no files inside that directory are tracked by git, so no need to build the repo around it
3182 log::debug!(
3183 "not building git repository for the worktree itself, `.git` path in the worktree: {dot_git_path:?}"
3184 );
3185 return;
3186 }
3187 };
3188
3189 let dot_git_abs_path = Arc::from(self.snapshot.absolutize(&dot_git_path).as_ref());
3190
3191 self.insert_git_repository_for_path(
3192 WorkDirectory::InProject {
3193 relative_path: work_dir_path,
3194 },
3195 dot_git_abs_path,
3196 fs,
3197 watcher,
3198 )
3199 .await
3200 .log_err();
3201 }
3202
3203 async fn insert_git_repository_for_path(
3204 &mut self,
3205 work_directory: WorkDirectory,
3206 dot_git_abs_path: Arc<Path>,
3207 fs: &dyn Fs,
3208 watcher: &dyn Watcher,
3209 ) -> Result<LocalRepositoryEntry> {
3210 let work_dir_entry = self
3211 .snapshot
3212 .entry_for_path(&work_directory.path_key().0)
3213 .with_context(|| {
3214 format!(
3215 "working directory `{}` not indexed",
3216 work_directory
3217 .path_key()
3218 .0
3219 .display(self.snapshot.path_style)
3220 )
3221 })?;
3222 let work_directory_abs_path = self.snapshot.work_directory_abs_path(&work_directory);
3223
3224 let (repository_dir_abs_path, common_dir_abs_path) =
3225 discover_git_paths(&dot_git_abs_path, fs).await;
3226 watcher
3227 .add(&common_dir_abs_path)
3228 .context("failed to add common directory to watcher")
3229 .log_err();
3230 watcher
3231 .add(&repository_dir_abs_path)
3232 .context("failed to add repository directory to watcher")
3233 .log_err();
3234
3235 let work_directory_id = work_dir_entry.id;
3236
3237 let local_repository = LocalRepositoryEntry {
3238 work_directory_id,
3239 work_directory,
3240 work_directory_abs_path: work_directory_abs_path.as_path().into(),
3241 git_dir_scan_id: 0,
3242 dot_git_abs_path,
3243 common_dir_abs_path,
3244 repository_dir_abs_path,
3245 };
3246
3247 self.snapshot
3248 .git_repositories
3249 .insert(work_directory_id, local_repository.clone());
3250
3251 log::trace!("inserting new local git repository");
3252 Ok(local_repository)
3253 }
3254}
3255
3256async fn is_git_dir(path: &Path, fs: &dyn Fs) -> bool {
3257 if let Some(file_name) = path.file_name()
3258 && file_name == DOT_GIT
3259 {
3260 return true;
3261 }
3262
3263 // If we're in a bare repository, we are not inside a `.git` folder. In a
3264 // bare repository, the root folder contains what would normally be in the
3265 // `.git` folder.
3266 let head_metadata = fs.metadata(&path.join("HEAD")).await;
3267 if !matches!(head_metadata, Ok(Some(_))) {
3268 return false;
3269 }
3270 let config_metadata = fs.metadata(&path.join("config")).await;
3271 matches!(config_metadata, Ok(Some(_)))
3272}
3273
3274async fn build_gitignore(abs_path: &Path, fs: &dyn Fs) -> Result<Gitignore> {
3275 let contents = fs
3276 .load(abs_path)
3277 .await
3278 .with_context(|| format!("failed to load gitignore file at {}", abs_path.display()))?;
3279 let parent = abs_path.parent().unwrap_or_else(|| Path::new("/"));
3280 let mut builder = GitignoreBuilder::new(parent);
3281 for line in contents.lines() {
3282 builder.add_line(Some(abs_path.into()), line)?;
3283 }
3284 Ok(builder.build()?)
3285}
3286
3287impl Deref for Worktree {
3288 type Target = Snapshot;
3289
3290 fn deref(&self) -> &Self::Target {
3291 match self {
3292 Worktree::Local(worktree) => &worktree.snapshot,
3293 Worktree::Remote(worktree) => &worktree.snapshot,
3294 }
3295 }
3296}
3297
3298impl Deref for LocalWorktree {
3299 type Target = LocalSnapshot;
3300
3301 fn deref(&self) -> &Self::Target {
3302 &self.snapshot
3303 }
3304}
3305
3306impl Deref for RemoteWorktree {
3307 type Target = Snapshot;
3308
3309 fn deref(&self) -> &Self::Target {
3310 &self.snapshot
3311 }
3312}
3313
3314impl fmt::Debug for LocalWorktree {
3315 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3316 self.snapshot.fmt(f)
3317 }
3318}
3319
3320impl fmt::Debug for Snapshot {
3321 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3322 struct EntriesById<'a>(&'a SumTree<PathEntry>);
3323 struct EntriesByPath<'a>(&'a SumTree<Entry>);
3324
3325 impl fmt::Debug for EntriesByPath<'_> {
3326 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3327 f.debug_map()
3328 .entries(self.0.iter().map(|entry| (&entry.path, entry.id)))
3329 .finish()
3330 }
3331 }
3332
3333 impl fmt::Debug for EntriesById<'_> {
3334 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
3335 f.debug_list().entries(self.0.iter()).finish()
3336 }
3337 }
3338
3339 f.debug_struct("Snapshot")
3340 .field("id", &self.id)
3341 .field("root_name", &self.root_name)
3342 .field("entries_by_path", &EntriesByPath(&self.entries_by_path))
3343 .field("entries_by_id", &EntriesById(&self.entries_by_id))
3344 .finish()
3345 }
3346}
3347
3348#[derive(Debug, Clone, PartialEq)]
3349pub struct File {
3350 pub worktree: Entity<Worktree>,
3351 pub path: Arc<RelPath>,
3352 pub disk_state: DiskState,
3353 pub entry_id: Option<ProjectEntryId>,
3354 pub is_local: bool,
3355 pub is_private: bool,
3356}
3357
3358impl language::File for File {
3359 fn as_local(&self) -> Option<&dyn language::LocalFile> {
3360 if self.is_local { Some(self) } else { None }
3361 }
3362
3363 fn disk_state(&self) -> DiskState {
3364 self.disk_state
3365 }
3366
3367 fn path(&self) -> &Arc<RelPath> {
3368 &self.path
3369 }
3370
3371 fn full_path(&self, cx: &App) -> PathBuf {
3372 self.worktree.read(cx).full_path(&self.path)
3373 }
3374
3375 /// Returns the last component of this handle's absolute path. If this handle refers to the root
3376 /// of its worktree, then this method will return the name of the worktree itself.
3377 fn file_name<'a>(&'a self, cx: &'a App) -> &'a str {
3378 self.path
3379 .file_name()
3380 .unwrap_or_else(|| self.worktree.read(cx).root_name_str())
3381 }
3382
3383 fn worktree_id(&self, cx: &App) -> WorktreeId {
3384 self.worktree.read(cx).id()
3385 }
3386
3387 fn to_proto(&self, cx: &App) -> rpc::proto::File {
3388 rpc::proto::File {
3389 worktree_id: self.worktree.read(cx).id().to_proto(),
3390 entry_id: self.entry_id.map(|id| id.to_proto()),
3391 path: self.path.as_ref().to_proto(),
3392 mtime: self.disk_state.mtime().map(|time| time.into()),
3393 is_deleted: self.disk_state.is_deleted(),
3394 is_historic: matches!(self.disk_state, DiskState::Historic { .. }),
3395 }
3396 }
3397
3398 fn is_private(&self) -> bool {
3399 self.is_private
3400 }
3401
3402 fn path_style(&self, cx: &App) -> PathStyle {
3403 self.worktree.read(cx).path_style()
3404 }
3405
3406 fn can_open(&self) -> bool {
3407 true
3408 }
3409}
3410
3411impl language::LocalFile for File {
3412 fn abs_path(&self, cx: &App) -> PathBuf {
3413 self.worktree.read(cx).absolutize(&self.path)
3414 }
3415
3416 fn load(&self, cx: &App) -> Task<Result<String>> {
3417 let worktree = self.worktree.read(cx).as_local().unwrap();
3418 let abs_path = worktree.absolutize(&self.path);
3419 let fs = worktree.fs.clone();
3420 cx.background_spawn(async move { fs.load(&abs_path).await })
3421 }
3422
3423 fn load_bytes(&self, cx: &App) -> Task<Result<Vec<u8>>> {
3424 let worktree = self.worktree.read(cx).as_local().unwrap();
3425 let abs_path = worktree.absolutize(&self.path);
3426 let fs = worktree.fs.clone();
3427 cx.background_spawn(async move { fs.load_bytes(&abs_path).await })
3428 }
3429}
3430
3431impl File {
3432 pub fn for_entry(entry: Entry, worktree: Entity<Worktree>) -> Arc<Self> {
3433 Arc::new(Self {
3434 worktree,
3435 path: entry.path.clone(),
3436 disk_state: if let Some(mtime) = entry.mtime {
3437 DiskState::Present {
3438 mtime,
3439 size: entry.size,
3440 }
3441 } else {
3442 DiskState::New
3443 },
3444 entry_id: Some(entry.id),
3445 is_local: true,
3446 is_private: entry.is_private,
3447 })
3448 }
3449
3450 pub fn from_proto(
3451 proto: rpc::proto::File,
3452 worktree: Entity<Worktree>,
3453 cx: &App,
3454 ) -> Result<Self> {
3455 let worktree_id = worktree.read(cx).as_remote().context("not remote")?.id();
3456
3457 anyhow::ensure!(
3458 worktree_id.to_proto() == proto.worktree_id,
3459 "worktree id does not match file"
3460 );
3461
3462 let disk_state = if proto.is_historic {
3463 DiskState::Historic {
3464 was_deleted: proto.is_deleted,
3465 }
3466 } else if proto.is_deleted {
3467 DiskState::Deleted
3468 } else if let Some(mtime) = proto.mtime.map(&Into::into) {
3469 DiskState::Present { mtime, size: 0 }
3470 } else {
3471 DiskState::New
3472 };
3473
3474 Ok(Self {
3475 worktree,
3476 path: RelPath::from_proto(&proto.path).context("invalid path in file protobuf")?,
3477 disk_state,
3478 entry_id: proto.entry_id.map(ProjectEntryId::from_proto),
3479 is_local: false,
3480 is_private: false,
3481 })
3482 }
3483
3484 pub fn from_dyn(file: Option<&Arc<dyn language::File>>) -> Option<&Self> {
3485 file.and_then(|f| {
3486 let f: &dyn language::File = f.borrow();
3487 let f: &dyn Any = f;
3488 f.downcast_ref()
3489 })
3490 }
3491
3492 pub fn worktree_id(&self, cx: &App) -> WorktreeId {
3493 self.worktree.read(cx).id()
3494 }
3495
3496 pub fn project_entry_id(&self) -> Option<ProjectEntryId> {
3497 match self.disk_state {
3498 DiskState::Deleted => None,
3499 _ => self.entry_id,
3500 }
3501 }
3502}
3503
3504#[derive(Clone, Debug, PartialEq, Eq)]
3505pub struct Entry {
3506 pub id: ProjectEntryId,
3507 pub kind: EntryKind,
3508 pub path: Arc<RelPath>,
3509 pub inode: u64,
3510 pub mtime: Option<MTime>,
3511
3512 pub canonical_path: Option<Arc<Path>>,
3513 /// Whether this entry is ignored by Git.
3514 ///
3515 /// We only scan ignored entries once the directory is expanded and
3516 /// exclude them from searches.
3517 pub is_ignored: bool,
3518
3519 /// Whether this entry is hidden or inside hidden directory.
3520 ///
3521 /// We only scan hidden entries once the directory is expanded.
3522 pub is_hidden: bool,
3523
3524 /// Whether this entry is always included in searches.
3525 ///
3526 /// This is used for entries that are always included in searches, even
3527 /// if they are ignored by git. Overridden by file_scan_exclusions.
3528 pub is_always_included: bool,
3529
3530 /// Whether this entry's canonical path is outside of the worktree.
3531 /// This means the entry is only accessible from the worktree root via a
3532 /// symlink.
3533 ///
3534 /// We only scan entries outside of the worktree once the symlinked
3535 /// directory is expanded.
3536 pub is_external: bool,
3537
3538 /// Whether this entry is considered to be a `.env` file.
3539 pub is_private: bool,
3540 /// The entry's size on disk, in bytes.
3541 pub size: u64,
3542 pub char_bag: CharBag,
3543 pub is_fifo: bool,
3544}
3545
3546#[derive(Clone, Copy, Debug, PartialEq, Eq)]
3547pub enum EntryKind {
3548 UnloadedDir,
3549 PendingDir,
3550 Dir,
3551 File,
3552}
3553
3554#[derive(Clone, Copy, Debug, PartialEq)]
3555pub enum PathChange {
3556 /// A filesystem entry was was created.
3557 Added,
3558 /// A filesystem entry was removed.
3559 Removed,
3560 /// A filesystem entry was updated.
3561 Updated,
3562 /// A filesystem entry was either updated or added. We don't know
3563 /// whether or not it already existed, because the path had not
3564 /// been loaded before the event.
3565 AddedOrUpdated,
3566 /// A filesystem entry was found during the initial scan of the worktree.
3567 Loaded,
3568}
3569
3570#[derive(Clone, Debug, PartialEq, Eq)]
3571pub struct UpdatedGitRepository {
3572 /// ID of the repository's working directory.
3573 ///
3574 /// For a repo that's above the worktree root, this is the ID of the worktree root, and hence not unique.
3575 /// It's included here to aid the GitStore in detecting when a repository's working directory is renamed.
3576 pub work_directory_id: ProjectEntryId,
3577 pub old_work_directory_abs_path: Option<Arc<Path>>,
3578 pub new_work_directory_abs_path: Option<Arc<Path>>,
3579 /// For a normal git repository checkout, the absolute path to the .git directory.
3580 /// For a worktree, the absolute path to the worktree's subdirectory inside the .git directory.
3581 pub dot_git_abs_path: Option<Arc<Path>>,
3582 pub repository_dir_abs_path: Option<Arc<Path>>,
3583 pub common_dir_abs_path: Option<Arc<Path>>,
3584}
3585
3586pub type UpdatedEntriesSet = Arc<[(Arc<RelPath>, ProjectEntryId, PathChange)]>;
3587pub type UpdatedGitRepositoriesSet = Arc<[UpdatedGitRepository]>;
3588
3589#[derive(Clone, Debug)]
3590pub struct PathProgress<'a> {
3591 pub max_path: &'a RelPath,
3592}
3593
3594#[derive(Clone, Debug)]
3595pub struct PathSummary<S> {
3596 pub max_path: Arc<RelPath>,
3597 pub item_summary: S,
3598}
3599
3600impl<S: Summary> Summary for PathSummary<S> {
3601 type Context<'a> = S::Context<'a>;
3602
3603 fn zero(cx: Self::Context<'_>) -> Self {
3604 Self {
3605 max_path: RelPath::empty().into(),
3606 item_summary: S::zero(cx),
3607 }
3608 }
3609
3610 fn add_summary(&mut self, rhs: &Self, cx: Self::Context<'_>) {
3611 self.max_path = rhs.max_path.clone();
3612 self.item_summary.add_summary(&rhs.item_summary, cx);
3613 }
3614}
3615
3616impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathProgress<'a> {
3617 fn zero(_: <PathSummary<S> as Summary>::Context<'_>) -> Self {
3618 Self {
3619 max_path: RelPath::empty(),
3620 }
3621 }
3622
3623 fn add_summary(
3624 &mut self,
3625 summary: &'a PathSummary<S>,
3626 _: <PathSummary<S> as Summary>::Context<'_>,
3627 ) {
3628 self.max_path = summary.max_path.as_ref()
3629 }
3630}
3631
3632impl<'a> sum_tree::Dimension<'a, PathSummary<GitSummary>> for GitSummary {
3633 fn zero(_cx: ()) -> Self {
3634 Default::default()
3635 }
3636
3637 fn add_summary(&mut self, summary: &'a PathSummary<GitSummary>, _: ()) {
3638 *self += summary.item_summary
3639 }
3640}
3641
3642impl<'a>
3643 sum_tree::SeekTarget<'a, PathSummary<GitSummary>, Dimensions<TraversalProgress<'a>, GitSummary>>
3644 for PathTarget<'_>
3645{
3646 fn cmp(
3647 &self,
3648 cursor_location: &Dimensions<TraversalProgress<'a>, GitSummary>,
3649 _: (),
3650 ) -> Ordering {
3651 self.cmp_path(cursor_location.0.max_path)
3652 }
3653}
3654
3655impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for PathKey {
3656 fn zero(_: S::Context<'_>) -> Self {
3657 Default::default()
3658 }
3659
3660 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3661 self.0 = summary.max_path.clone();
3662 }
3663}
3664
3665impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary<S>> for TraversalProgress<'a> {
3666 fn zero(_cx: S::Context<'_>) -> Self {
3667 Default::default()
3668 }
3669
3670 fn add_summary(&mut self, summary: &'a PathSummary<S>, _: S::Context<'_>) {
3671 self.max_path = summary.max_path.as_ref();
3672 }
3673}
3674
3675impl Entry {
3676 fn new(
3677 path: Arc<RelPath>,
3678 metadata: &fs::Metadata,
3679 id: ProjectEntryId,
3680 root_char_bag: CharBag,
3681 canonical_path: Option<Arc<Path>>,
3682 ) -> Self {
3683 let char_bag = char_bag_for_path(root_char_bag, &path);
3684 Self {
3685 id,
3686 kind: if metadata.is_dir {
3687 EntryKind::PendingDir
3688 } else {
3689 EntryKind::File
3690 },
3691 path,
3692 inode: metadata.inode,
3693 mtime: Some(metadata.mtime),
3694 size: metadata.len,
3695 canonical_path,
3696 is_ignored: false,
3697 is_hidden: false,
3698 is_always_included: false,
3699 is_external: false,
3700 is_private: false,
3701 char_bag,
3702 is_fifo: metadata.is_fifo,
3703 }
3704 }
3705
3706 pub fn is_created(&self) -> bool {
3707 self.mtime.is_some()
3708 }
3709
3710 pub fn is_dir(&self) -> bool {
3711 self.kind.is_dir()
3712 }
3713
3714 pub fn is_file(&self) -> bool {
3715 self.kind.is_file()
3716 }
3717}
3718
3719impl EntryKind {
3720 pub fn is_dir(&self) -> bool {
3721 matches!(
3722 self,
3723 EntryKind::Dir | EntryKind::PendingDir | EntryKind::UnloadedDir
3724 )
3725 }
3726
3727 pub fn is_unloaded(&self) -> bool {
3728 matches!(self, EntryKind::UnloadedDir)
3729 }
3730
3731 pub fn is_file(&self) -> bool {
3732 matches!(self, EntryKind::File)
3733 }
3734}
3735
3736impl sum_tree::Item for Entry {
3737 type Summary = EntrySummary;
3738
3739 fn summary(&self, _cx: ()) -> Self::Summary {
3740 let non_ignored_count = if self.is_ignored && !self.is_always_included {
3741 0
3742 } else {
3743 1
3744 };
3745 let file_count;
3746 let non_ignored_file_count;
3747 if self.is_file() {
3748 file_count = 1;
3749 non_ignored_file_count = non_ignored_count;
3750 } else {
3751 file_count = 0;
3752 non_ignored_file_count = 0;
3753 }
3754
3755 EntrySummary {
3756 max_path: self.path.clone(),
3757 count: 1,
3758 non_ignored_count,
3759 file_count,
3760 non_ignored_file_count,
3761 }
3762 }
3763}
3764
3765impl sum_tree::KeyedItem for Entry {
3766 type Key = PathKey;
3767
3768 fn key(&self) -> Self::Key {
3769 PathKey(self.path.clone())
3770 }
3771}
3772
3773#[derive(Clone, Debug)]
3774pub struct EntrySummary {
3775 max_path: Arc<RelPath>,
3776 count: usize,
3777 non_ignored_count: usize,
3778 file_count: usize,
3779 non_ignored_file_count: usize,
3780}
3781
3782impl Default for EntrySummary {
3783 fn default() -> Self {
3784 Self {
3785 max_path: Arc::from(RelPath::empty()),
3786 count: 0,
3787 non_ignored_count: 0,
3788 file_count: 0,
3789 non_ignored_file_count: 0,
3790 }
3791 }
3792}
3793
3794impl sum_tree::ContextLessSummary for EntrySummary {
3795 fn zero() -> Self {
3796 Default::default()
3797 }
3798
3799 fn add_summary(&mut self, rhs: &Self) {
3800 self.max_path = rhs.max_path.clone();
3801 self.count += rhs.count;
3802 self.non_ignored_count += rhs.non_ignored_count;
3803 self.file_count += rhs.file_count;
3804 self.non_ignored_file_count += rhs.non_ignored_file_count;
3805 }
3806}
3807
3808#[derive(Clone, Debug)]
3809struct PathEntry {
3810 id: ProjectEntryId,
3811 path: Arc<RelPath>,
3812 is_ignored: bool,
3813 scan_id: usize,
3814}
3815
3816impl sum_tree::Item for PathEntry {
3817 type Summary = PathEntrySummary;
3818
3819 fn summary(&self, _cx: ()) -> Self::Summary {
3820 PathEntrySummary { max_id: self.id }
3821 }
3822}
3823
3824impl sum_tree::KeyedItem for PathEntry {
3825 type Key = ProjectEntryId;
3826
3827 fn key(&self) -> Self::Key {
3828 self.id
3829 }
3830}
3831
3832#[derive(Clone, Debug, Default)]
3833struct PathEntrySummary {
3834 max_id: ProjectEntryId,
3835}
3836
3837impl sum_tree::ContextLessSummary for PathEntrySummary {
3838 fn zero() -> Self {
3839 Default::default()
3840 }
3841
3842 fn add_summary(&mut self, summary: &Self) {
3843 self.max_id = summary.max_id;
3844 }
3845}
3846
3847impl<'a> sum_tree::Dimension<'a, PathEntrySummary> for ProjectEntryId {
3848 fn zero(_cx: ()) -> Self {
3849 Default::default()
3850 }
3851
3852 fn add_summary(&mut self, summary: &'a PathEntrySummary, _: ()) {
3853 *self = summary.max_id;
3854 }
3855}
3856
3857#[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
3858pub struct PathKey(pub Arc<RelPath>);
3859
3860impl Default for PathKey {
3861 fn default() -> Self {
3862 Self(RelPath::empty().into())
3863 }
3864}
3865
3866impl<'a> sum_tree::Dimension<'a, EntrySummary> for PathKey {
3867 fn zero(_cx: ()) -> Self {
3868 Default::default()
3869 }
3870
3871 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
3872 self.0 = summary.max_path.clone();
3873 }
3874}
3875
3876struct BackgroundScanner {
3877 state: async_lock::Mutex<BackgroundScannerState>,
3878 fs: Arc<dyn Fs>,
3879 fs_case_sensitive: bool,
3880 status_updates_tx: UnboundedSender<ScanState>,
3881 executor: BackgroundExecutor,
3882 scan_requests_rx: channel::Receiver<ScanRequest>,
3883 path_prefixes_to_scan_rx: channel::Receiver<PathPrefixScanRequest>,
3884 next_entry_id: Arc<AtomicUsize>,
3885 phase: BackgroundScannerPhase,
3886 watcher: Arc<dyn Watcher>,
3887 settings: WorktreeSettings,
3888 share_private_files: bool,
3889 /// Whether this is a single-file worktree (root is a file, not a directory).
3890 /// Used to determine if we should give up after repeated canonicalization failures.
3891 is_single_file: bool,
3892}
3893
3894#[derive(Copy, Clone, PartialEq)]
3895enum BackgroundScannerPhase {
3896 InitialScan,
3897 EventsReceivedDuringInitialScan,
3898 Events,
3899}
3900
3901impl BackgroundScanner {
3902 async fn run(&mut self, mut fs_events_rx: Pin<Box<dyn Send + Stream<Item = Vec<PathEvent>>>>) {
3903 let root_abs_path;
3904 let scanning_enabled;
3905 {
3906 let state = self.state.lock().await;
3907 root_abs_path = state.snapshot.abs_path.clone();
3908 scanning_enabled = state.scanning_enabled;
3909 }
3910
3911 // If the worktree root does not contain a git repository, then find
3912 // the git repository in an ancestor directory. Find any gitignore files
3913 // in ancestor directories.
3914 let repo = if scanning_enabled {
3915 let (ignores, exclude, repo) =
3916 discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await;
3917 self.state
3918 .lock()
3919 .await
3920 .snapshot
3921 .ignores_by_parent_abs_path
3922 .extend(ignores);
3923 if let Some(exclude) = exclude {
3924 self.state
3925 .lock()
3926 .await
3927 .snapshot
3928 .repo_exclude_by_work_dir_abs_path
3929 .insert(root_abs_path.as_path().into(), (exclude, false));
3930 }
3931
3932 repo
3933 } else {
3934 None
3935 };
3936
3937 let containing_git_repository = if let Some((ancestor_dot_git, work_directory)) = repo
3938 && scanning_enabled
3939 {
3940 maybe!(async {
3941 self.state
3942 .lock()
3943 .await
3944 .insert_git_repository_for_path(
3945 work_directory,
3946 ancestor_dot_git.clone().into(),
3947 self.fs.as_ref(),
3948 self.watcher.as_ref(),
3949 )
3950 .await
3951 .log_err()?;
3952 Some(ancestor_dot_git)
3953 })
3954 .await
3955 } else {
3956 None
3957 };
3958
3959 log::trace!("containing git repository: {containing_git_repository:?}");
3960
3961 let global_gitignore_file = paths::global_gitignore_path();
3962 let mut global_gitignore_events = if let Some(global_gitignore_path) =
3963 &global_gitignore_file
3964 && scanning_enabled
3965 {
3966 let is_file = self.fs.is_file(&global_gitignore_path).await;
3967 self.state.lock().await.snapshot.global_gitignore = if is_file {
3968 build_gitignore(global_gitignore_path, self.fs.as_ref())
3969 .await
3970 .ok()
3971 .map(Arc::new)
3972 } else {
3973 None
3974 };
3975 if is_file {
3976 self.fs
3977 .watch(global_gitignore_path, FS_WATCH_LATENCY)
3978 .await
3979 .0
3980 } else {
3981 Box::pin(futures::stream::pending())
3982 }
3983 } else {
3984 self.state.lock().await.snapshot.global_gitignore = None;
3985 Box::pin(futures::stream::pending())
3986 };
3987
3988 let (scan_job_tx, scan_job_rx) = channel::unbounded();
3989 {
3990 let mut state = self.state.lock().await;
3991 state.snapshot.scan_id += 1;
3992 if let Some(mut root_entry) = state.snapshot.root_entry().cloned() {
3993 let ignore_stack = state
3994 .snapshot
3995 .ignore_stack_for_abs_path(root_abs_path.as_path(), true, self.fs.as_ref())
3996 .await;
3997 if ignore_stack.is_abs_path_ignored(root_abs_path.as_path(), true) {
3998 root_entry.is_ignored = true;
3999 let mut root_entry = root_entry.clone();
4000 state.reuse_entry_id(&mut root_entry);
4001 state
4002 .insert_entry(root_entry, self.fs.as_ref(), self.watcher.as_ref())
4003 .await;
4004 }
4005 if root_entry.is_dir() && state.scanning_enabled {
4006 state
4007 .enqueue_scan_dir(
4008 root_abs_path.as_path().into(),
4009 &root_entry,
4010 &scan_job_tx,
4011 self.fs.as_ref(),
4012 )
4013 .await;
4014 }
4015 }
4016 };
4017
4018 // Perform an initial scan of the directory.
4019 drop(scan_job_tx);
4020 self.scan_dirs(true, scan_job_rx).await;
4021 {
4022 let mut state = self.state.lock().await;
4023 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4024 }
4025
4026 self.send_status_update(false, SmallVec::new(), &[]).await;
4027
4028 // Process any any FS events that occurred while performing the initial scan.
4029 // For these events, update events cannot be as precise, because we didn't
4030 // have the previous state loaded yet.
4031 self.phase = BackgroundScannerPhase::EventsReceivedDuringInitialScan;
4032 if let Poll::Ready(Some(mut paths)) = futures::poll!(fs_events_rx.next()) {
4033 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
4034 paths.extend(more_paths);
4035 }
4036 self.process_events(
4037 paths
4038 .into_iter()
4039 .filter(|event| event.kind.is_some())
4040 .collect(),
4041 )
4042 .await;
4043 }
4044 if let Some(abs_path) = containing_git_repository {
4045 self.process_events(vec![PathEvent {
4046 path: abs_path,
4047 kind: Some(fs::PathEventKind::Changed),
4048 }])
4049 .await;
4050 }
4051
4052 // Continue processing events until the worktree is dropped.
4053 self.phase = BackgroundScannerPhase::Events;
4054
4055 loop {
4056 select_biased! {
4057 // Process any path refresh requests from the worktree. Prioritize
4058 // these before handling changes reported by the filesystem.
4059 request = self.next_scan_request().fuse() => {
4060 let Ok(request) = request else { break };
4061 if !self.process_scan_request(request, false).await {
4062 return;
4063 }
4064 }
4065
4066 path_prefix_request = self.path_prefixes_to_scan_rx.recv().fuse() => {
4067 let Ok(request) = path_prefix_request else { break };
4068 log::trace!("adding path prefix {:?}", request.path);
4069
4070 let did_scan = self.forcibly_load_paths(std::slice::from_ref(&request.path)).await;
4071 if did_scan {
4072 let abs_path =
4073 {
4074 let mut state = self.state.lock().await;
4075 state.path_prefixes_to_scan.insert(request.path.clone());
4076 state.snapshot.absolutize(&request.path)
4077 };
4078
4079 if let Some(abs_path) = self.fs.canonicalize(&abs_path).await.log_err() {
4080 self.process_events(vec![PathEvent {
4081 path: abs_path,
4082 kind: Some(fs::PathEventKind::Changed),
4083 }])
4084 .await;
4085 }
4086 }
4087 self.send_status_update(false, request.done, &[]).await;
4088 }
4089
4090 paths = fs_events_rx.next().fuse() => {
4091 let Some(mut paths) = paths else { break };
4092 while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) {
4093 paths.extend(more_paths);
4094 }
4095 self.process_events(paths.into_iter().filter(|event| event.kind.is_some()).collect()).await;
4096 }
4097
4098 _ = global_gitignore_events.next().fuse() => {
4099 if let Some(path) = &global_gitignore_file {
4100 self.update_global_gitignore(&path).await;
4101 }
4102 }
4103 }
4104 }
4105 }
4106
4107 async fn process_scan_request(&self, mut request: ScanRequest, scanning: bool) -> bool {
4108 log::debug!("rescanning paths {:?}", request.relative_paths);
4109
4110 request.relative_paths.sort_unstable();
4111 self.forcibly_load_paths(&request.relative_paths).await;
4112
4113 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4114 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4115 let root_canonical_path = match &root_canonical_path {
4116 Ok(path) => SanitizedPath::new(path),
4117 Err(err) => {
4118 log::error!("failed to canonicalize root path {root_path:?}: {err:#}");
4119 return true;
4120 }
4121 };
4122 let abs_paths = request
4123 .relative_paths
4124 .iter()
4125 .map(|path| {
4126 if path.file_name().is_some() {
4127 root_canonical_path.as_path().join(path.as_std_path())
4128 } else {
4129 root_canonical_path.as_path().to_path_buf()
4130 }
4131 })
4132 .collect::<Vec<_>>();
4133
4134 {
4135 let mut state = self.state.lock().await;
4136 let is_idle = state.snapshot.completed_scan_id == state.snapshot.scan_id;
4137 state.snapshot.scan_id += 1;
4138 if is_idle {
4139 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4140 }
4141 }
4142
4143 self.reload_entries_for_paths(
4144 &root_path,
4145 &root_canonical_path,
4146 &request.relative_paths,
4147 abs_paths,
4148 None,
4149 )
4150 .await;
4151
4152 self.send_status_update(scanning, request.done, &[]).await
4153 }
4154
4155 async fn process_events(&self, mut events: Vec<PathEvent>) {
4156 let root_path = self.state.lock().await.snapshot.abs_path.clone();
4157 let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await;
4158 let root_canonical_path = match &root_canonical_path {
4159 Ok(path) => SanitizedPath::new(path),
4160 Err(err) => {
4161 let new_path = self
4162 .state
4163 .lock()
4164 .await
4165 .snapshot
4166 .root_file_handle
4167 .clone()
4168 .and_then(|handle| match handle.current_path(&self.fs) {
4169 Ok(new_path) => Some(new_path),
4170 Err(e) => {
4171 log::error!("Failed to refresh worktree root path: {e:#}");
4172 None
4173 }
4174 })
4175 .map(|path| SanitizedPath::new_arc(&path))
4176 .filter(|new_path| *new_path != root_path);
4177
4178 if let Some(new_path) = new_path {
4179 log::info!(
4180 "root renamed from {:?} to {:?}",
4181 root_path.as_path(),
4182 new_path.as_path(),
4183 );
4184 self.status_updates_tx
4185 .unbounded_send(ScanState::RootUpdated { new_path })
4186 .ok();
4187 } else {
4188 log::error!("root path could not be canonicalized: {err:#}");
4189
4190 // For single-file worktrees, if we can't canonicalize and the file handle
4191 // fallback also failed, the file is gone - close the worktree
4192 if self.is_single_file {
4193 log::info!(
4194 "single-file worktree root {:?} no longer exists, marking as deleted",
4195 root_path.as_path()
4196 );
4197 self.status_updates_tx
4198 .unbounded_send(ScanState::RootDeleted)
4199 .ok();
4200 }
4201 }
4202 return;
4203 }
4204 };
4205
4206 // Certain directories may have FS changes, but do not lead to git data changes that Zed cares about.
4207 // Ignore these, to avoid Zed unnecessarily rescanning git metadata.
4208 let skipped_files_in_dot_git = [COMMIT_MESSAGE, INDEX_LOCK];
4209 let skipped_dirs_in_dot_git = [FSMONITOR_DAEMON, LFS_DIR];
4210
4211 let mut relative_paths = Vec::with_capacity(events.len());
4212 let mut dot_git_abs_paths = Vec::new();
4213 let mut work_dirs_needing_exclude_update = Vec::new();
4214 events.sort_unstable_by(|left, right| left.path.cmp(&right.path));
4215 events.dedup_by(|left, right| {
4216 if left.path == right.path {
4217 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4218 right.kind = left.kind;
4219 }
4220 true
4221 } else if left.path.starts_with(&right.path) {
4222 if matches!(left.kind, Some(fs::PathEventKind::Rescan)) {
4223 right.kind = left.kind;
4224 }
4225 true
4226 } else {
4227 false
4228 }
4229 });
4230 {
4231 let snapshot = &self.state.lock().await.snapshot;
4232
4233 let mut ranges_to_drop = SmallVec::<[Range<usize>; 4]>::new();
4234
4235 fn skip_ix(ranges: &mut SmallVec<[Range<usize>; 4]>, ix: usize) {
4236 if let Some(last_range) = ranges.last_mut()
4237 && last_range.end == ix
4238 {
4239 last_range.end += 1;
4240 } else {
4241 ranges.push(ix..ix + 1);
4242 }
4243 }
4244
4245 for (ix, event) in events.iter().enumerate() {
4246 let abs_path = SanitizedPath::new(&event.path);
4247
4248 let mut is_git_related = false;
4249 let mut dot_git_paths = None;
4250
4251 for ancestor in abs_path.as_path().ancestors() {
4252 if is_git_dir(ancestor, self.fs.as_ref()).await {
4253 let path_in_git_dir = abs_path
4254 .as_path()
4255 .strip_prefix(ancestor)
4256 .expect("stripping off the ancestor");
4257 dot_git_paths = Some((ancestor.to_owned(), path_in_git_dir.to_owned()));
4258 break;
4259 }
4260 }
4261
4262 if let Some((dot_git_abs_path, path_in_git_dir)) = dot_git_paths {
4263 // We ignore `""` as well, as that is going to be the
4264 // `.git` folder itself. WE do not care about it, if
4265 // there are changes within we will see them, we need
4266 // this ignore to prevent us from accidentally observing
4267 // the ignored created file due to the events not being
4268 // empty after filtering.
4269
4270 let is_dot_git_changed = {
4271 path_in_git_dir == Path::new("")
4272 && event.kind == Some(PathEventKind::Changed)
4273 && abs_path
4274 .strip_prefix(root_canonical_path)
4275 .ok()
4276 .and_then(|it| RelPath::new(it, PathStyle::local()).ok())
4277 .is_some_and(|it| {
4278 snapshot
4279 .entry_for_path(&it)
4280 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4281 })
4282 };
4283 let condition = skipped_files_in_dot_git.iter().any(|skipped| {
4284 OsStr::new(skipped) == path_in_git_dir.as_path().as_os_str()
4285 }) || skipped_dirs_in_dot_git
4286 .iter()
4287 .any(|skipped_git_subdir| path_in_git_dir.starts_with(skipped_git_subdir))
4288 || is_dot_git_changed;
4289 if condition {
4290 log::debug!(
4291 "ignoring event {abs_path:?} as it's in the .git directory among skipped files or directories"
4292 );
4293 skip_ix(&mut ranges_to_drop, ix);
4294 continue;
4295 }
4296
4297 is_git_related = true;
4298 if !dot_git_abs_paths.contains(&dot_git_abs_path) {
4299 dot_git_abs_paths.push(dot_git_abs_path);
4300 }
4301 }
4302
4303 let relative_path = if let Ok(path) = abs_path.strip_prefix(&root_canonical_path)
4304 && let Ok(path) = RelPath::new(path, PathStyle::local())
4305 {
4306 path
4307 } else {
4308 if is_git_related {
4309 log::debug!(
4310 "ignoring event {abs_path:?}, since it's in git dir outside of root path {root_canonical_path:?}",
4311 );
4312 } else {
4313 log::error!(
4314 "ignoring event {abs_path:?} outside of root path {root_canonical_path:?}",
4315 );
4316 }
4317 skip_ix(&mut ranges_to_drop, ix);
4318 continue;
4319 };
4320
4321 let absolute_path = abs_path.to_path_buf();
4322 if absolute_path.ends_with(Path::new(DOT_GIT).join(REPO_EXCLUDE)) {
4323 if let Some(repository) = snapshot
4324 .git_repositories
4325 .values()
4326 .find(|repo| repo.common_dir_abs_path.join(REPO_EXCLUDE) == absolute_path)
4327 {
4328 work_dirs_needing_exclude_update
4329 .push(repository.work_directory_abs_path.clone());
4330 }
4331 }
4332
4333 if abs_path.file_name() == Some(OsStr::new(GITIGNORE)) {
4334 for (_, repo) in snapshot
4335 .git_repositories
4336 .iter()
4337 .filter(|(_, repo)| repo.directory_contains(&relative_path))
4338 {
4339 if !dot_git_abs_paths.iter().any(|dot_git_abs_path| {
4340 dot_git_abs_path == repo.common_dir_abs_path.as_ref()
4341 }) {
4342 dot_git_abs_paths.push(repo.common_dir_abs_path.to_path_buf());
4343 }
4344 }
4345 }
4346
4347 let parent_dir_is_loaded = relative_path.parent().is_none_or(|parent| {
4348 snapshot
4349 .entry_for_path(parent)
4350 .is_some_and(|entry| entry.kind == EntryKind::Dir)
4351 });
4352 if !parent_dir_is_loaded {
4353 log::debug!("ignoring event {relative_path:?} within unloaded directory");
4354 skip_ix(&mut ranges_to_drop, ix);
4355 continue;
4356 }
4357
4358 if self.settings.is_path_excluded(&relative_path) {
4359 if !is_git_related {
4360 log::debug!("ignoring FS event for excluded path {relative_path:?}");
4361 }
4362 skip_ix(&mut ranges_to_drop, ix);
4363 continue;
4364 }
4365
4366 relative_paths.push(EventRoot {
4367 path: relative_path.into_arc(),
4368 was_rescanned: matches!(event.kind, Some(fs::PathEventKind::Rescan)),
4369 });
4370 }
4371
4372 for range_to_drop in ranges_to_drop.into_iter().rev() {
4373 events.drain(range_to_drop);
4374 }
4375 }
4376
4377 if relative_paths.is_empty() && dot_git_abs_paths.is_empty() {
4378 return;
4379 }
4380
4381 if !work_dirs_needing_exclude_update.is_empty() {
4382 let mut state = self.state.lock().await;
4383 for work_dir_abs_path in work_dirs_needing_exclude_update {
4384 if let Some((_, needs_update)) = state
4385 .snapshot
4386 .repo_exclude_by_work_dir_abs_path
4387 .get_mut(&work_dir_abs_path)
4388 {
4389 *needs_update = true;
4390 }
4391 }
4392 }
4393
4394 self.state.lock().await.snapshot.scan_id += 1;
4395
4396 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4397 log::debug!(
4398 "received fs events {:?}",
4399 relative_paths
4400 .iter()
4401 .map(|event_root| &event_root.path)
4402 .collect::<Vec<_>>()
4403 );
4404 self.reload_entries_for_paths(
4405 &root_path,
4406 &root_canonical_path,
4407 &relative_paths
4408 .iter()
4409 .map(|event_root| event_root.path.clone())
4410 .collect::<Vec<_>>(),
4411 events
4412 .into_iter()
4413 .map(|event| event.path)
4414 .collect::<Vec<_>>(),
4415 Some(scan_job_tx.clone()),
4416 )
4417 .await;
4418
4419 let affected_repo_roots = if !dot_git_abs_paths.is_empty() {
4420 self.update_git_repositories(dot_git_abs_paths).await
4421 } else {
4422 Vec::new()
4423 };
4424
4425 {
4426 let mut ignores_to_update = self.ignores_needing_update().await;
4427 ignores_to_update.extend(affected_repo_roots);
4428 let ignores_to_update = self.order_ignores(ignores_to_update).await;
4429 let snapshot = self.state.lock().await.snapshot.clone();
4430 self.update_ignore_statuses_for_paths(scan_job_tx, snapshot, ignores_to_update)
4431 .await;
4432 self.scan_dirs(false, scan_job_rx).await;
4433 }
4434
4435 {
4436 let mut state = self.state.lock().await;
4437 state.snapshot.completed_scan_id = state.snapshot.scan_id;
4438 for (_, entry) in mem::take(&mut state.removed_entries) {
4439 state.scanned_dirs.remove(&entry.id);
4440 }
4441 }
4442 self.send_status_update(false, SmallVec::new(), &relative_paths)
4443 .await;
4444 }
4445
4446 async fn update_global_gitignore(&self, abs_path: &Path) {
4447 let ignore = build_gitignore(abs_path, self.fs.as_ref())
4448 .await
4449 .log_err()
4450 .map(Arc::new);
4451 let (prev_snapshot, ignore_stack, abs_path) = {
4452 let mut state = self.state.lock().await;
4453 state.snapshot.global_gitignore = ignore;
4454 let abs_path = state.snapshot.abs_path().clone();
4455 let ignore_stack = state
4456 .snapshot
4457 .ignore_stack_for_abs_path(&abs_path, true, self.fs.as_ref())
4458 .await;
4459 (state.snapshot.clone(), ignore_stack, abs_path)
4460 };
4461 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4462 self.update_ignore_statuses_for_paths(
4463 scan_job_tx,
4464 prev_snapshot,
4465 vec![(abs_path, ignore_stack)],
4466 )
4467 .await;
4468 self.scan_dirs(false, scan_job_rx).await;
4469 self.send_status_update(false, SmallVec::new(), &[]).await;
4470 }
4471
4472 async fn forcibly_load_paths(&self, paths: &[Arc<RelPath>]) -> bool {
4473 let (scan_job_tx, scan_job_rx) = channel::unbounded();
4474 {
4475 let mut state = self.state.lock().await;
4476 let root_path = state.snapshot.abs_path.clone();
4477 for path in paths {
4478 for ancestor in path.ancestors() {
4479 if let Some(entry) = state.snapshot.entry_for_path(ancestor)
4480 && entry.kind == EntryKind::UnloadedDir
4481 {
4482 let abs_path = root_path.join(ancestor.as_std_path());
4483 state
4484 .enqueue_scan_dir(
4485 abs_path.into(),
4486 entry,
4487 &scan_job_tx,
4488 self.fs.as_ref(),
4489 )
4490 .await;
4491 state.paths_to_scan.insert(path.clone());
4492 break;
4493 }
4494 }
4495 }
4496 drop(scan_job_tx);
4497 }
4498 while let Ok(job) = scan_job_rx.recv().await {
4499 self.scan_dir(&job).await.log_err();
4500 }
4501
4502 !mem::take(&mut self.state.lock().await.paths_to_scan).is_empty()
4503 }
4504
4505 async fn scan_dirs(
4506 &self,
4507 enable_progress_updates: bool,
4508 scan_jobs_rx: channel::Receiver<ScanJob>,
4509 ) {
4510 if self
4511 .status_updates_tx
4512 .unbounded_send(ScanState::Started)
4513 .is_err()
4514 {
4515 return;
4516 }
4517
4518 let progress_update_count = AtomicUsize::new(0);
4519 self.executor
4520 .scoped_priority(Priority::Low, |scope| {
4521 for _ in 0..self.executor.num_cpus() {
4522 scope.spawn(async {
4523 let mut last_progress_update_count = 0;
4524 let progress_update_timer = self.progress_timer(enable_progress_updates).fuse();
4525 futures::pin_mut!(progress_update_timer);
4526
4527 loop {
4528 select_biased! {
4529 // Process any path refresh requests before moving on to process
4530 // the scan queue, so that user operations are prioritized.
4531 request = self.next_scan_request().fuse() => {
4532 let Ok(request) = request else { break };
4533 if !self.process_scan_request(request, true).await {
4534 return;
4535 }
4536 }
4537
4538 // Send periodic progress updates to the worktree. Use an atomic counter
4539 // to ensure that only one of the workers sends a progress update after
4540 // the update interval elapses.
4541 _ = progress_update_timer => {
4542 match progress_update_count.compare_exchange(
4543 last_progress_update_count,
4544 last_progress_update_count + 1,
4545 SeqCst,
4546 SeqCst
4547 ) {
4548 Ok(_) => {
4549 last_progress_update_count += 1;
4550 self.send_status_update(true, SmallVec::new(), &[])
4551 .await;
4552 }
4553 Err(count) => {
4554 last_progress_update_count = count;
4555 }
4556 }
4557 progress_update_timer.set(self.progress_timer(enable_progress_updates).fuse());
4558 }
4559
4560 // Recursively load directories from the file system.
4561 job = scan_jobs_rx.recv().fuse() => {
4562 let Ok(job) = job else { break };
4563 if let Err(err) = self.scan_dir(&job).await
4564 && job.path.is_empty() {
4565 log::error!("error scanning directory {:?}: {}", job.abs_path, err);
4566 }
4567 }
4568 }
4569 }
4570 });
4571 }
4572 })
4573 .await;
4574 }
4575
4576 async fn send_status_update(
4577 &self,
4578 scanning: bool,
4579 barrier: SmallVec<[barrier::Sender; 1]>,
4580 event_roots: &[EventRoot],
4581 ) -> bool {
4582 let mut state = self.state.lock().await;
4583 if state.changed_paths.is_empty() && event_roots.is_empty() && scanning {
4584 return true;
4585 }
4586
4587 let merged_event_roots = merge_event_roots(&state.changed_paths, event_roots);
4588
4589 let new_snapshot = state.snapshot.clone();
4590 let old_snapshot = mem::replace(&mut state.prev_snapshot, new_snapshot.snapshot.clone());
4591 let changes = build_diff(
4592 self.phase,
4593 &old_snapshot,
4594 &new_snapshot,
4595 &merged_event_roots,
4596 );
4597 state.changed_paths.clear();
4598
4599 self.status_updates_tx
4600 .unbounded_send(ScanState::Updated {
4601 snapshot: new_snapshot,
4602 changes,
4603 scanning,
4604 barrier,
4605 })
4606 .is_ok()
4607 }
4608
4609 async fn scan_dir(&self, job: &ScanJob) -> Result<()> {
4610 let root_abs_path;
4611 let root_char_bag;
4612 {
4613 let snapshot = &self.state.lock().await.snapshot;
4614 if self.settings.is_path_excluded(&job.path) {
4615 log::error!("skipping excluded directory {:?}", job.path);
4616 return Ok(());
4617 }
4618 log::trace!("scanning directory {:?}", job.path);
4619 root_abs_path = snapshot.abs_path().clone();
4620 root_char_bag = snapshot.root_char_bag;
4621 }
4622
4623 let next_entry_id = self.next_entry_id.clone();
4624 let mut ignore_stack = job.ignore_stack.clone();
4625 let mut new_ignore = None;
4626 let mut root_canonical_path = None;
4627 let mut new_entries: Vec<Entry> = Vec::new();
4628 let mut new_jobs: Vec<Option<ScanJob>> = Vec::new();
4629 let mut child_paths = self
4630 .fs
4631 .read_dir(&job.abs_path)
4632 .await?
4633 .filter_map(|entry| async {
4634 match entry {
4635 Ok(entry) => Some(entry),
4636 Err(error) => {
4637 log::error!("error processing entry {:?}", error);
4638 None
4639 }
4640 }
4641 })
4642 .collect::<Vec<_>>()
4643 .await;
4644
4645 // Ensure that .git and .gitignore are processed first.
4646 swap_to_front(&mut child_paths, GITIGNORE);
4647 swap_to_front(&mut child_paths, DOT_GIT);
4648
4649 if let Some(path) = child_paths.first()
4650 && path.ends_with(DOT_GIT)
4651 {
4652 ignore_stack.repo_root = Some(job.abs_path.clone());
4653 }
4654
4655 for child_abs_path in child_paths {
4656 let child_abs_path: Arc<Path> = child_abs_path.into();
4657 let child_name = child_abs_path.file_name().unwrap();
4658 let Some(child_path) = child_name
4659 .to_str()
4660 .and_then(|name| Some(job.path.join(RelPath::unix(name).ok()?)))
4661 else {
4662 continue;
4663 };
4664
4665 if child_name == DOT_GIT {
4666 let mut state = self.state.lock().await;
4667 state
4668 .insert_git_repository(
4669 child_path.clone(),
4670 self.fs.as_ref(),
4671 self.watcher.as_ref(),
4672 )
4673 .await;
4674 } else if child_name == GITIGNORE {
4675 match build_gitignore(&child_abs_path, self.fs.as_ref()).await {
4676 Ok(ignore) => {
4677 let ignore = Arc::new(ignore);
4678 ignore_stack = ignore_stack
4679 .append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
4680 new_ignore = Some(ignore);
4681 }
4682 Err(error) => {
4683 log::error!(
4684 "error loading .gitignore file {:?} - {:?}",
4685 child_name,
4686 error
4687 );
4688 }
4689 }
4690 }
4691
4692 if self.settings.is_path_excluded(&child_path) {
4693 log::debug!("skipping excluded child entry {child_path:?}");
4694 self.state
4695 .lock()
4696 .await
4697 .remove_path(&child_path, self.watcher.as_ref());
4698 continue;
4699 }
4700
4701 let child_metadata = match self.fs.metadata(&child_abs_path).await {
4702 Ok(Some(metadata)) => metadata,
4703 Ok(None) => continue,
4704 Err(err) => {
4705 log::error!("error processing {:?}: {err:#}", child_abs_path.display());
4706 continue;
4707 }
4708 };
4709
4710 let mut child_entry = Entry::new(
4711 child_path.clone(),
4712 &child_metadata,
4713 ProjectEntryId::new(&next_entry_id),
4714 root_char_bag,
4715 None,
4716 );
4717
4718 if job.is_external {
4719 child_entry.is_external = true;
4720 } else if child_metadata.is_symlink {
4721 let canonical_path = match self.fs.canonicalize(&child_abs_path).await {
4722 Ok(path) => path,
4723 Err(err) => {
4724 log::error!("error reading target of symlink {child_abs_path:?}: {err:#}",);
4725 continue;
4726 }
4727 };
4728
4729 // lazily canonicalize the root path in order to determine if
4730 // symlinks point outside of the worktree.
4731 let root_canonical_path = match &root_canonical_path {
4732 Some(path) => path,
4733 None => match self.fs.canonicalize(&root_abs_path).await {
4734 Ok(path) => root_canonical_path.insert(path),
4735 Err(err) => {
4736 log::error!("error canonicalizing root {:?}: {:?}", root_abs_path, err);
4737 continue;
4738 }
4739 },
4740 };
4741
4742 if !canonical_path.starts_with(root_canonical_path) {
4743 child_entry.is_external = true;
4744 }
4745
4746 child_entry.canonical_path = Some(canonical_path.into());
4747 }
4748
4749 if child_entry.is_dir() {
4750 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, true);
4751 child_entry.is_always_included =
4752 self.settings.is_path_always_included(&child_path, true);
4753
4754 // Avoid recursing until crash in the case of a recursive symlink
4755 if job.ancestor_inodes.contains(&child_entry.inode) {
4756 new_jobs.push(None);
4757 } else {
4758 let mut ancestor_inodes = job.ancestor_inodes.clone();
4759 ancestor_inodes.insert(child_entry.inode);
4760
4761 new_jobs.push(Some(ScanJob {
4762 abs_path: child_abs_path.clone(),
4763 path: child_path,
4764 is_external: child_entry.is_external,
4765 ignore_stack: if child_entry.is_ignored {
4766 IgnoreStack::all()
4767 } else {
4768 ignore_stack.clone()
4769 },
4770 ancestor_inodes,
4771 scan_queue: job.scan_queue.clone(),
4772 }));
4773 }
4774 } else {
4775 child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, false);
4776 child_entry.is_always_included =
4777 self.settings.is_path_always_included(&child_path, false);
4778 }
4779
4780 {
4781 let relative_path = job
4782 .path
4783 .join(RelPath::unix(child_name.to_str().unwrap()).unwrap());
4784 if self.is_path_private(&relative_path) {
4785 log::debug!("detected private file: {relative_path:?}");
4786 child_entry.is_private = true;
4787 }
4788 if self.settings.is_path_hidden(&relative_path) {
4789 log::debug!("detected hidden file: {relative_path:?}");
4790 child_entry.is_hidden = true;
4791 }
4792 }
4793
4794 new_entries.push(child_entry);
4795 }
4796
4797 let mut state = self.state.lock().await;
4798
4799 // Identify any subdirectories that should not be scanned.
4800 let mut job_ix = 0;
4801 for entry in &mut new_entries {
4802 state.reuse_entry_id(entry);
4803 if entry.is_dir() {
4804 if state.should_scan_directory(entry) {
4805 job_ix += 1;
4806 } else {
4807 log::debug!("defer scanning directory {:?}", entry.path);
4808 entry.kind = EntryKind::UnloadedDir;
4809 new_jobs.remove(job_ix);
4810 }
4811 }
4812 if entry.is_always_included {
4813 state
4814 .snapshot
4815 .always_included_entries
4816 .push(entry.path.clone());
4817 }
4818 }
4819
4820 state.populate_dir(job.path.clone(), new_entries, new_ignore);
4821 self.watcher.add(job.abs_path.as_ref()).log_err();
4822
4823 for new_job in new_jobs.into_iter().flatten() {
4824 job.scan_queue
4825 .try_send(new_job)
4826 .expect("channel is unbounded");
4827 }
4828
4829 Ok(())
4830 }
4831
4832 /// All list arguments should be sorted before calling this function
4833 async fn reload_entries_for_paths(
4834 &self,
4835 root_abs_path: &SanitizedPath,
4836 root_canonical_path: &SanitizedPath,
4837 relative_paths: &[Arc<RelPath>],
4838 abs_paths: Vec<PathBuf>,
4839 scan_queue_tx: Option<Sender<ScanJob>>,
4840 ) {
4841 // grab metadata for all requested paths
4842 let metadata = futures::future::join_all(
4843 abs_paths
4844 .iter()
4845 .map(|abs_path| async move {
4846 let metadata = self.fs.metadata(abs_path).await?;
4847 if let Some(metadata) = metadata {
4848 let canonical_path = self.fs.canonicalize(abs_path).await?;
4849
4850 // If we're on a case-insensitive filesystem (default on macOS), we want
4851 // to only ignore metadata for non-symlink files if their absolute-path matches
4852 // the canonical-path.
4853 // Because if not, this might be a case-only-renaming (`mv test.txt TEST.TXT`)
4854 // and we want to ignore the metadata for the old path (`test.txt`) so it's
4855 // treated as removed.
4856 if !self.fs_case_sensitive && !metadata.is_symlink {
4857 let canonical_file_name = canonical_path.file_name();
4858 let file_name = abs_path.file_name();
4859 if canonical_file_name != file_name {
4860 return Ok(None);
4861 }
4862 }
4863
4864 anyhow::Ok(Some((metadata, SanitizedPath::new_arc(&canonical_path))))
4865 } else {
4866 Ok(None)
4867 }
4868 })
4869 .collect::<Vec<_>>(),
4870 )
4871 .await;
4872
4873 let mut new_ancestor_repo = if relative_paths.iter().any(|path| path.is_empty()) {
4874 Some(discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await)
4875 } else {
4876 None
4877 };
4878
4879 let mut state = self.state.lock().await;
4880 let doing_recursive_update = scan_queue_tx.is_some();
4881
4882 // Remove any entries for paths that no longer exist or are being recursively
4883 // refreshed. Do this before adding any new entries, so that renames can be
4884 // detected regardless of the order of the paths.
4885 for (path, metadata) in relative_paths.iter().zip(metadata.iter()) {
4886 if matches!(metadata, Ok(None)) || doing_recursive_update {
4887 state.remove_path(path, self.watcher.as_ref());
4888 }
4889 }
4890
4891 for (path, metadata) in relative_paths.iter().zip(metadata.into_iter()) {
4892 let abs_path: Arc<Path> = root_abs_path.join(path.as_std_path()).into();
4893 match metadata {
4894 Ok(Some((metadata, canonical_path))) => {
4895 let ignore_stack = state
4896 .snapshot
4897 .ignore_stack_for_abs_path(&abs_path, metadata.is_dir, self.fs.as_ref())
4898 .await;
4899 let is_external = !canonical_path.starts_with(&root_canonical_path);
4900 let entry_id = state.entry_id_for(self.next_entry_id.as_ref(), path, &metadata);
4901 let mut fs_entry = Entry::new(
4902 path.clone(),
4903 &metadata,
4904 entry_id,
4905 state.snapshot.root_char_bag,
4906 if metadata.is_symlink {
4907 Some(canonical_path.as_path().to_path_buf().into())
4908 } else {
4909 None
4910 },
4911 );
4912
4913 let is_dir = fs_entry.is_dir();
4914 fs_entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, is_dir);
4915 fs_entry.is_external = is_external;
4916 fs_entry.is_private = self.is_path_private(path);
4917 fs_entry.is_always_included =
4918 self.settings.is_path_always_included(path, is_dir);
4919 fs_entry.is_hidden = self.settings.is_path_hidden(path);
4920
4921 if let (Some(scan_queue_tx), true) = (&scan_queue_tx, is_dir) {
4922 if state.should_scan_directory(&fs_entry)
4923 || (fs_entry.path.is_empty()
4924 && abs_path.file_name() == Some(OsStr::new(DOT_GIT)))
4925 {
4926 state
4927 .enqueue_scan_dir(
4928 abs_path,
4929 &fs_entry,
4930 scan_queue_tx,
4931 self.fs.as_ref(),
4932 )
4933 .await;
4934 } else {
4935 fs_entry.kind = EntryKind::UnloadedDir;
4936 }
4937 }
4938
4939 state
4940 .insert_entry(fs_entry.clone(), self.fs.as_ref(), self.watcher.as_ref())
4941 .await;
4942
4943 if path.is_empty()
4944 && let Some((ignores, exclude, repo)) = new_ancestor_repo.take()
4945 {
4946 log::trace!("updating ancestor git repository");
4947 state.snapshot.ignores_by_parent_abs_path.extend(ignores);
4948 if let Some((ancestor_dot_git, work_directory)) = repo {
4949 if let Some(exclude) = exclude {
4950 let work_directory_abs_path = self
4951 .state
4952 .lock()
4953 .await
4954 .snapshot
4955 .work_directory_abs_path(&work_directory);
4956
4957 state
4958 .snapshot
4959 .repo_exclude_by_work_dir_abs_path
4960 .insert(work_directory_abs_path.into(), (exclude, false));
4961 }
4962 state
4963 .insert_git_repository_for_path(
4964 work_directory,
4965 ancestor_dot_git.into(),
4966 self.fs.as_ref(),
4967 self.watcher.as_ref(),
4968 )
4969 .await
4970 .log_err();
4971 }
4972 }
4973 }
4974 Ok(None) => {
4975 self.remove_repo_path(path.clone(), &mut state.snapshot);
4976 }
4977 Err(err) => {
4978 log::error!("error reading file {abs_path:?} on event: {err:#}");
4979 }
4980 }
4981 }
4982
4983 util::extend_sorted(
4984 &mut state.changed_paths,
4985 relative_paths.iter().cloned(),
4986 usize::MAX,
4987 Ord::cmp,
4988 );
4989 }
4990
4991 fn remove_repo_path(&self, path: Arc<RelPath>, snapshot: &mut LocalSnapshot) -> Option<()> {
4992 if !path.components().any(|component| component == DOT_GIT)
4993 && let Some(local_repo) = snapshot.local_repo_for_work_directory_path(&path)
4994 {
4995 let id = local_repo.work_directory_id;
4996 log::debug!("remove repo path: {:?}", path);
4997 snapshot.git_repositories.remove(&id);
4998 return Some(());
4999 }
5000
5001 Some(())
5002 }
5003
5004 async fn update_ignore_statuses_for_paths(
5005 &self,
5006 scan_job_tx: Sender<ScanJob>,
5007 prev_snapshot: LocalSnapshot,
5008 ignores_to_update: Vec<(Arc<Path>, IgnoreStack)>,
5009 ) {
5010 let (ignore_queue_tx, ignore_queue_rx) = channel::unbounded();
5011 {
5012 for (parent_abs_path, ignore_stack) in ignores_to_update {
5013 ignore_queue_tx
5014 .send_blocking(UpdateIgnoreStatusJob {
5015 abs_path: parent_abs_path,
5016 ignore_stack,
5017 ignore_queue: ignore_queue_tx.clone(),
5018 scan_queue: scan_job_tx.clone(),
5019 })
5020 .unwrap();
5021 }
5022 }
5023 drop(ignore_queue_tx);
5024
5025 self.executor
5026 .scoped(|scope| {
5027 for _ in 0..self.executor.num_cpus() {
5028 scope.spawn(async {
5029 loop {
5030 select_biased! {
5031 // Process any path refresh requests before moving on to process
5032 // the queue of ignore statuses.
5033 request = self.next_scan_request().fuse() => {
5034 let Ok(request) = request else { break };
5035 if !self.process_scan_request(request, true).await {
5036 return;
5037 }
5038 }
5039
5040 // Recursively process directories whose ignores have changed.
5041 job = ignore_queue_rx.recv().fuse() => {
5042 let Ok(job) = job else { break };
5043 self.update_ignore_status(job, &prev_snapshot).await;
5044 }
5045 }
5046 }
5047 });
5048 }
5049 })
5050 .await;
5051 }
5052
5053 async fn ignores_needing_update(&self) -> Vec<Arc<Path>> {
5054 let mut ignores_to_update = Vec::new();
5055 let mut excludes_to_load: Vec<(Arc<Path>, PathBuf)> = Vec::new();
5056
5057 // First pass: collect updates and drop stale entries without awaiting.
5058 {
5059 let snapshot = &mut self.state.lock().await.snapshot;
5060 let abs_path = snapshot.abs_path.clone();
5061 let mut repo_exclude_keys_to_remove: Vec<Arc<Path>> = Vec::new();
5062
5063 for (work_dir_abs_path, (_, needs_update)) in
5064 snapshot.repo_exclude_by_work_dir_abs_path.iter_mut()
5065 {
5066 let repository = snapshot
5067 .git_repositories
5068 .iter()
5069 .find(|(_, repo)| &repo.work_directory_abs_path == work_dir_abs_path);
5070
5071 if *needs_update {
5072 *needs_update = false;
5073 ignores_to_update.push(work_dir_abs_path.clone());
5074
5075 if let Some((_, repository)) = repository {
5076 let exclude_abs_path = repository.common_dir_abs_path.join(REPO_EXCLUDE);
5077 excludes_to_load.push((work_dir_abs_path.clone(), exclude_abs_path));
5078 }
5079 }
5080
5081 if repository.is_none() {
5082 repo_exclude_keys_to_remove.push(work_dir_abs_path.clone());
5083 }
5084 }
5085
5086 for key in repo_exclude_keys_to_remove {
5087 snapshot.repo_exclude_by_work_dir_abs_path.remove(&key);
5088 }
5089
5090 snapshot
5091 .ignores_by_parent_abs_path
5092 .retain(|parent_abs_path, (_, needs_update)| {
5093 if let Ok(parent_path) = parent_abs_path.strip_prefix(abs_path.as_path())
5094 && let Some(parent_path) =
5095 RelPath::new(&parent_path, PathStyle::local()).log_err()
5096 {
5097 if *needs_update {
5098 *needs_update = false;
5099 if snapshot.snapshot.entry_for_path(&parent_path).is_some() {
5100 ignores_to_update.push(parent_abs_path.clone());
5101 }
5102 }
5103
5104 let ignore_path = parent_path.join(RelPath::unix(GITIGNORE).unwrap());
5105 if snapshot.snapshot.entry_for_path(&ignore_path).is_none() {
5106 return false;
5107 }
5108 }
5109 true
5110 });
5111 }
5112
5113 // Load gitignores asynchronously (outside the lock)
5114 let mut loaded_excludes: Vec<(Arc<Path>, Arc<Gitignore>)> = Vec::new();
5115 for (work_dir_abs_path, exclude_abs_path) in excludes_to_load {
5116 if let Ok(current_exclude) = build_gitignore(&exclude_abs_path, self.fs.as_ref()).await
5117 {
5118 loaded_excludes.push((work_dir_abs_path, Arc::new(current_exclude)));
5119 }
5120 }
5121
5122 // Second pass: apply updates.
5123 if !loaded_excludes.is_empty() {
5124 let snapshot = &mut self.state.lock().await.snapshot;
5125
5126 for (work_dir_abs_path, exclude) in loaded_excludes {
5127 if let Some((existing_exclude, _)) = snapshot
5128 .repo_exclude_by_work_dir_abs_path
5129 .get_mut(&work_dir_abs_path)
5130 {
5131 *existing_exclude = exclude;
5132 }
5133 }
5134 }
5135
5136 ignores_to_update
5137 }
5138
5139 async fn order_ignores(&self, mut ignores: Vec<Arc<Path>>) -> Vec<(Arc<Path>, IgnoreStack)> {
5140 let fs = self.fs.clone();
5141 let snapshot = self.state.lock().await.snapshot.clone();
5142 ignores.sort_unstable();
5143 let mut ignores_to_update = ignores.into_iter().peekable();
5144
5145 let mut result = vec![];
5146 while let Some(parent_abs_path) = ignores_to_update.next() {
5147 while ignores_to_update
5148 .peek()
5149 .map_or(false, |p| p.starts_with(&parent_abs_path))
5150 {
5151 ignores_to_update.next().unwrap();
5152 }
5153 let ignore_stack = snapshot
5154 .ignore_stack_for_abs_path(&parent_abs_path, true, fs.as_ref())
5155 .await;
5156 result.push((parent_abs_path, ignore_stack));
5157 }
5158
5159 result
5160 }
5161
5162 async fn update_ignore_status(&self, job: UpdateIgnoreStatusJob, snapshot: &LocalSnapshot) {
5163 log::trace!("update ignore status {:?}", job.abs_path);
5164
5165 let mut ignore_stack = job.ignore_stack;
5166 if let Some((ignore, _)) = snapshot.ignores_by_parent_abs_path.get(&job.abs_path) {
5167 ignore_stack =
5168 ignore_stack.append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone());
5169 }
5170
5171 let mut entries_by_id_edits = Vec::new();
5172 let mut entries_by_path_edits = Vec::new();
5173 let Some(path) = job
5174 .abs_path
5175 .strip_prefix(snapshot.abs_path.as_path())
5176 .map_err(|_| {
5177 anyhow::anyhow!(
5178 "Failed to strip prefix '{}' from path '{}'",
5179 snapshot.abs_path.as_path().display(),
5180 job.abs_path.display()
5181 )
5182 })
5183 .log_err()
5184 else {
5185 return;
5186 };
5187
5188 let Some(path) = RelPath::new(&path, PathStyle::local()).log_err() else {
5189 return;
5190 };
5191
5192 if let Ok(Some(metadata)) = self.fs.metadata(&job.abs_path.join(DOT_GIT)).await
5193 && metadata.is_dir
5194 {
5195 ignore_stack.repo_root = Some(job.abs_path.clone());
5196 }
5197
5198 for mut entry in snapshot.child_entries(&path).cloned() {
5199 let was_ignored = entry.is_ignored;
5200 let abs_path: Arc<Path> = snapshot.absolutize(&entry.path).into();
5201 entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, entry.is_dir());
5202
5203 if entry.is_dir() {
5204 let child_ignore_stack = if entry.is_ignored {
5205 IgnoreStack::all()
5206 } else {
5207 ignore_stack.clone()
5208 };
5209
5210 // Scan any directories that were previously ignored and weren't previously scanned.
5211 if was_ignored && !entry.is_ignored && entry.kind.is_unloaded() {
5212 let state = self.state.lock().await;
5213 if state.should_scan_directory(&entry) {
5214 state
5215 .enqueue_scan_dir(
5216 abs_path.clone(),
5217 &entry,
5218 &job.scan_queue,
5219 self.fs.as_ref(),
5220 )
5221 .await;
5222 }
5223 }
5224
5225 job.ignore_queue
5226 .send(UpdateIgnoreStatusJob {
5227 abs_path: abs_path.clone(),
5228 ignore_stack: child_ignore_stack,
5229 ignore_queue: job.ignore_queue.clone(),
5230 scan_queue: job.scan_queue.clone(),
5231 })
5232 .await
5233 .unwrap();
5234 }
5235
5236 if entry.is_ignored != was_ignored {
5237 let mut path_entry = snapshot.entries_by_id.get(&entry.id, ()).unwrap().clone();
5238 path_entry.scan_id = snapshot.scan_id;
5239 path_entry.is_ignored = entry.is_ignored;
5240 entries_by_id_edits.push(Edit::Insert(path_entry));
5241 entries_by_path_edits.push(Edit::Insert(entry));
5242 }
5243 }
5244
5245 let state = &mut self.state.lock().await;
5246 for edit in &entries_by_path_edits {
5247 if let Edit::Insert(entry) = edit
5248 && let Err(ix) = state.changed_paths.binary_search(&entry.path)
5249 {
5250 state.changed_paths.insert(ix, entry.path.clone());
5251 }
5252 }
5253
5254 state
5255 .snapshot
5256 .entries_by_path
5257 .edit(entries_by_path_edits, ());
5258 state.snapshot.entries_by_id.edit(entries_by_id_edits, ());
5259 }
5260
5261 async fn update_git_repositories(&self, dot_git_paths: Vec<PathBuf>) -> Vec<Arc<Path>> {
5262 log::trace!("reloading repositories: {dot_git_paths:?}");
5263 let mut state = self.state.lock().await;
5264 let scan_id = state.snapshot.scan_id;
5265 let mut affected_repo_roots = Vec::new();
5266 for dot_git_dir in dot_git_paths {
5267 let existing_repository_entry =
5268 state
5269 .snapshot
5270 .git_repositories
5271 .iter()
5272 .find_map(|(_, repo)| {
5273 let dot_git_dir = SanitizedPath::new(&dot_git_dir);
5274 if SanitizedPath::new(repo.common_dir_abs_path.as_ref()) == dot_git_dir
5275 || SanitizedPath::new(repo.repository_dir_abs_path.as_ref())
5276 == dot_git_dir
5277 {
5278 Some(repo.clone())
5279 } else {
5280 None
5281 }
5282 });
5283
5284 match existing_repository_entry {
5285 None => {
5286 let Ok(relative) = dot_git_dir.strip_prefix(state.snapshot.abs_path()) else {
5287 // This can happen legitimately when `.git` is a
5288 // gitfile (e.g. in a linked worktree or submodule)
5289 // pointing to a directory outside the worktree root.
5290 // Skip it — the repository was already registered
5291 // during the initial scan via `discover_git_paths`.
5292 debug_assert!(
5293 self.fs.is_file(&dot_git_dir).await,
5294 "update_git_repositories: .git path outside worktree root \
5295 is not a gitfile: {dot_git_dir:?}",
5296 );
5297 continue;
5298 };
5299 affected_repo_roots.push(dot_git_dir.parent().unwrap().into());
5300 state
5301 .insert_git_repository(
5302 RelPath::new(relative, PathStyle::local())
5303 .unwrap()
5304 .into_arc(),
5305 self.fs.as_ref(),
5306 self.watcher.as_ref(),
5307 )
5308 .await;
5309 }
5310 Some(local_repository) => {
5311 state.snapshot.git_repositories.update(
5312 &local_repository.work_directory_id,
5313 |entry| {
5314 entry.git_dir_scan_id = scan_id;
5315 },
5316 );
5317 }
5318 };
5319 }
5320
5321 // Remove any git repositories whose .git entry no longer exists.
5322 let snapshot = &mut state.snapshot;
5323 let mut ids_to_preserve = HashSet::default();
5324 for (&work_directory_id, entry) in snapshot.git_repositories.iter() {
5325 let exists_in_snapshot =
5326 snapshot
5327 .entry_for_id(work_directory_id)
5328 .is_some_and(|entry| {
5329 snapshot
5330 .entry_for_path(&entry.path.join(RelPath::unix(DOT_GIT).unwrap()))
5331 .is_some()
5332 });
5333
5334 if exists_in_snapshot
5335 || matches!(
5336 self.fs.metadata(&entry.common_dir_abs_path).await,
5337 Ok(Some(_))
5338 )
5339 {
5340 ids_to_preserve.insert(work_directory_id);
5341 }
5342 }
5343
5344 snapshot
5345 .git_repositories
5346 .retain(|work_directory_id, entry| {
5347 let preserve = ids_to_preserve.contains(work_directory_id);
5348 if !preserve {
5349 affected_repo_roots.push(entry.dot_git_abs_path.parent().unwrap().into());
5350 snapshot
5351 .repo_exclude_by_work_dir_abs_path
5352 .remove(&entry.work_directory_abs_path);
5353 }
5354 preserve
5355 });
5356
5357 affected_repo_roots
5358 }
5359
5360 async fn progress_timer(&self, running: bool) {
5361 if !running {
5362 return futures::future::pending().await;
5363 }
5364
5365 #[cfg(feature = "test-support")]
5366 if self.fs.is_fake() {
5367 return self.executor.simulate_random_delay().await;
5368 }
5369
5370 self.executor.timer(FS_WATCH_LATENCY).await
5371 }
5372
5373 fn is_path_private(&self, path: &RelPath) -> bool {
5374 !self.share_private_files && self.settings.is_path_private(path)
5375 }
5376
5377 async fn next_scan_request(&self) -> Result<ScanRequest> {
5378 let mut request = self.scan_requests_rx.recv().await?;
5379 while let Ok(next_request) = self.scan_requests_rx.try_recv() {
5380 request.relative_paths.extend(next_request.relative_paths);
5381 request.done.extend(next_request.done);
5382 }
5383 Ok(request)
5384 }
5385}
5386
5387async fn discover_ancestor_git_repo(
5388 fs: Arc<dyn Fs>,
5389 root_abs_path: &SanitizedPath,
5390) -> (
5391 HashMap<Arc<Path>, (Arc<Gitignore>, bool)>,
5392 Option<Arc<Gitignore>>,
5393 Option<(PathBuf, WorkDirectory)>,
5394) {
5395 let mut exclude = None;
5396 let mut ignores = HashMap::default();
5397 for (index, ancestor) in root_abs_path.as_path().ancestors().enumerate() {
5398 if index != 0 {
5399 if ancestor == paths::home_dir() {
5400 // Unless $HOME is itself the worktree root, don't consider it as a
5401 // containing git repository---expensive and likely unwanted.
5402 break;
5403 } else if let Ok(ignore) = build_gitignore(&ancestor.join(GITIGNORE), fs.as_ref()).await
5404 {
5405 ignores.insert(ancestor.into(), (ignore.into(), false));
5406 }
5407 }
5408
5409 let ancestor_dot_git = ancestor.join(DOT_GIT);
5410 log::trace!("considering ancestor: {ancestor_dot_git:?}");
5411 // Check whether the directory or file called `.git` exists (in the
5412 // case of worktrees it's a file.)
5413 if fs
5414 .metadata(&ancestor_dot_git)
5415 .await
5416 .is_ok_and(|metadata| metadata.is_some())
5417 {
5418 if index != 0 {
5419 // We canonicalize, since the FS events use the canonicalized path.
5420 if let Some(ancestor_dot_git) = fs.canonicalize(&ancestor_dot_git).await.log_err() {
5421 let location_in_repo = root_abs_path
5422 .as_path()
5423 .strip_prefix(ancestor)
5424 .unwrap()
5425 .into();
5426 log::info!("inserting parent git repo for this worktree: {location_in_repo:?}");
5427 // We associate the external git repo with our root folder and
5428 // also mark where in the git repo the root folder is located.
5429 return (
5430 ignores,
5431 exclude,
5432 Some((
5433 ancestor_dot_git,
5434 WorkDirectory::AboveProject {
5435 absolute_path: ancestor.into(),
5436 location_in_repo,
5437 },
5438 )),
5439 );
5440 };
5441 }
5442
5443 let repo_exclude_abs_path = ancestor_dot_git.join(REPO_EXCLUDE);
5444 if let Ok(repo_exclude) = build_gitignore(&repo_exclude_abs_path, fs.as_ref()).await {
5445 exclude = Some(Arc::new(repo_exclude));
5446 }
5447
5448 // Reached root of git repository.
5449 break;
5450 }
5451 }
5452
5453 (ignores, exclude, None)
5454}
5455
5456fn merge_event_roots(changed_paths: &[Arc<RelPath>], event_roots: &[EventRoot]) -> Vec<EventRoot> {
5457 let mut merged_event_roots = Vec::with_capacity(changed_paths.len() + event_roots.len());
5458 let mut changed_paths = changed_paths.iter().peekable();
5459 let mut event_roots = event_roots.iter().peekable();
5460 while let (Some(path), Some(event_root)) = (changed_paths.peek(), event_roots.peek()) {
5461 match path.cmp(&&event_root.path) {
5462 Ordering::Less => {
5463 merged_event_roots.push(EventRoot {
5464 path: (*changed_paths.next().expect("peeked changed path")).clone(),
5465 was_rescanned: false,
5466 });
5467 }
5468 Ordering::Equal => {
5469 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5470 changed_paths.next();
5471 }
5472 Ordering::Greater => {
5473 merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone());
5474 }
5475 }
5476 }
5477 merged_event_roots.extend(changed_paths.map(|path| EventRoot {
5478 path: path.clone(),
5479 was_rescanned: false,
5480 }));
5481 merged_event_roots.extend(event_roots.cloned());
5482 merged_event_roots
5483}
5484
5485fn build_diff(
5486 phase: BackgroundScannerPhase,
5487 old_snapshot: &Snapshot,
5488 new_snapshot: &Snapshot,
5489 event_roots: &[EventRoot],
5490) -> UpdatedEntriesSet {
5491 use BackgroundScannerPhase::*;
5492 use PathChange::{Added, AddedOrUpdated, Loaded, Removed, Updated};
5493
5494 // Identify which paths have changed. Use the known set of changed
5495 // parent paths to optimize the search.
5496 let mut changes = Vec::new();
5497
5498 let mut old_paths = old_snapshot.entries_by_path.cursor::<PathKey>(());
5499 let mut new_paths = new_snapshot.entries_by_path.cursor::<PathKey>(());
5500 let mut last_newly_loaded_dir_path = None;
5501 old_paths.next();
5502 new_paths.next();
5503 for event_root in event_roots {
5504 let path = PathKey(event_root.path.clone());
5505 if old_paths.item().is_some_and(|e| e.path < path.0) {
5506 old_paths.seek_forward(&path, Bias::Left);
5507 }
5508 if new_paths.item().is_some_and(|e| e.path < path.0) {
5509 new_paths.seek_forward(&path, Bias::Left);
5510 }
5511 loop {
5512 match (old_paths.item(), new_paths.item()) {
5513 (Some(old_entry), Some(new_entry)) => {
5514 if old_entry.path > path.0
5515 && new_entry.path > path.0
5516 && !old_entry.path.starts_with(&path.0)
5517 && !new_entry.path.starts_with(&path.0)
5518 {
5519 break;
5520 }
5521
5522 match Ord::cmp(&old_entry.path, &new_entry.path) {
5523 Ordering::Less => {
5524 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5525 old_paths.next();
5526 }
5527 Ordering::Equal => {
5528 if phase == EventsReceivedDuringInitialScan {
5529 if old_entry.id != new_entry.id {
5530 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5531 }
5532 // If the worktree was not fully initialized when this event was generated,
5533 // we can't know whether this entry was added during the scan or whether
5534 // it was merely updated.
5535 changes.push((
5536 new_entry.path.clone(),
5537 new_entry.id,
5538 AddedOrUpdated,
5539 ));
5540 } else if old_entry.id != new_entry.id {
5541 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5542 changes.push((new_entry.path.clone(), new_entry.id, Added));
5543 } else if old_entry != new_entry {
5544 if old_entry.kind.is_unloaded() {
5545 last_newly_loaded_dir_path = Some(&new_entry.path);
5546 changes.push((new_entry.path.clone(), new_entry.id, Loaded));
5547 } else {
5548 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5549 }
5550 } else if event_root.was_rescanned {
5551 changes.push((new_entry.path.clone(), new_entry.id, Updated));
5552 }
5553 old_paths.next();
5554 new_paths.next();
5555 }
5556 Ordering::Greater => {
5557 let is_newly_loaded = phase == InitialScan
5558 || last_newly_loaded_dir_path
5559 .as_ref()
5560 .is_some_and(|dir| new_entry.path.starts_with(dir));
5561 changes.push((
5562 new_entry.path.clone(),
5563 new_entry.id,
5564 if is_newly_loaded { Loaded } else { Added },
5565 ));
5566 new_paths.next();
5567 }
5568 }
5569 }
5570 (Some(old_entry), None) => {
5571 changes.push((old_entry.path.clone(), old_entry.id, Removed));
5572 old_paths.next();
5573 }
5574 (None, Some(new_entry)) => {
5575 let is_newly_loaded = phase == InitialScan
5576 || last_newly_loaded_dir_path
5577 .as_ref()
5578 .is_some_and(|dir| new_entry.path.starts_with(dir));
5579 changes.push((
5580 new_entry.path.clone(),
5581 new_entry.id,
5582 if is_newly_loaded { Loaded } else { Added },
5583 ));
5584 new_paths.next();
5585 }
5586 (None, None) => break,
5587 }
5588 }
5589 }
5590
5591 changes.into()
5592}
5593
5594fn swap_to_front(child_paths: &mut Vec<PathBuf>, file: &str) {
5595 let position = child_paths
5596 .iter()
5597 .position(|path| path.file_name().unwrap() == file);
5598 if let Some(position) = position {
5599 let temp = child_paths.remove(position);
5600 child_paths.insert(0, temp);
5601 }
5602}
5603
5604fn char_bag_for_path(root_char_bag: CharBag, path: &RelPath) -> CharBag {
5605 let mut result = root_char_bag;
5606 result.extend(path.as_unix_str().chars().map(|c| c.to_ascii_lowercase()));
5607 result
5608}
5609
5610#[derive(Debug)]
5611struct ScanJob {
5612 abs_path: Arc<Path>,
5613 path: Arc<RelPath>,
5614 ignore_stack: IgnoreStack,
5615 scan_queue: Sender<ScanJob>,
5616 ancestor_inodes: TreeSet<u64>,
5617 is_external: bool,
5618}
5619
5620struct UpdateIgnoreStatusJob {
5621 abs_path: Arc<Path>,
5622 ignore_stack: IgnoreStack,
5623 ignore_queue: Sender<UpdateIgnoreStatusJob>,
5624 scan_queue: Sender<ScanJob>,
5625}
5626
5627pub trait WorktreeModelHandle {
5628 #[cfg(feature = "test-support")]
5629 fn flush_fs_events<'a>(
5630 &self,
5631 cx: &'a mut gpui::TestAppContext,
5632 ) -> futures::future::LocalBoxFuture<'a, ()>;
5633
5634 #[cfg(feature = "test-support")]
5635 fn flush_fs_events_in_root_git_repository<'a>(
5636 &self,
5637 cx: &'a mut gpui::TestAppContext,
5638 ) -> futures::future::LocalBoxFuture<'a, ()>;
5639}
5640
5641impl WorktreeModelHandle for Entity<Worktree> {
5642 // When the worktree's FS event stream sometimes delivers "redundant" events for FS changes that
5643 // occurred before the worktree was constructed. These events can cause the worktree to perform
5644 // extra directory scans, and emit extra scan-state notifications.
5645 //
5646 // This function mutates the worktree's directory and waits for those mutations to be picked up,
5647 // to ensure that all redundant FS events have already been processed.
5648 #[cfg(feature = "test-support")]
5649 fn flush_fs_events<'a>(
5650 &self,
5651 cx: &'a mut gpui::TestAppContext,
5652 ) -> futures::future::LocalBoxFuture<'a, ()> {
5653 let file_name = "fs-event-sentinel";
5654
5655 let tree = self.clone();
5656 let (fs, root_path) = self.read_with(cx, |tree, _| {
5657 let tree = tree.as_local().unwrap();
5658 (tree.fs.clone(), tree.abs_path.clone())
5659 });
5660
5661 async move {
5662 // Subscribe to events BEFORE creating the file to avoid race condition
5663 // where events fire before subscription is set up
5664 let mut events = cx.events(&tree);
5665
5666 fs.create_file(&root_path.join(file_name), Default::default())
5667 .await
5668 .unwrap();
5669
5670 // Check if condition is already met before waiting for events
5671 let file_exists = || {
5672 tree.read_with(cx, |tree, _| {
5673 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5674 .is_some()
5675 })
5676 };
5677
5678 // Use select to avoid blocking indefinitely if events are delayed
5679 while !file_exists() {
5680 futures::select_biased! {
5681 _ = events.next() => {}
5682 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5683 }
5684 }
5685
5686 fs.remove_file(&root_path.join(file_name), Default::default())
5687 .await
5688 .unwrap();
5689
5690 // Check if condition is already met before waiting for events
5691 let file_gone = || {
5692 tree.read_with(cx, |tree, _| {
5693 tree.entry_for_path(RelPath::unix(file_name).unwrap())
5694 .is_none()
5695 })
5696 };
5697
5698 // Use select to avoid blocking indefinitely if events are delayed
5699 while !file_gone() {
5700 futures::select_biased! {
5701 _ = events.next() => {}
5702 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5703 }
5704 }
5705
5706 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5707 .await;
5708 }
5709 .boxed_local()
5710 }
5711
5712 // This function is similar to flush_fs_events, except that it waits for events to be flushed in
5713 // the .git folder of the root repository.
5714 // The reason for its existence is that a repository's .git folder might live *outside* of the
5715 // worktree and thus its FS events might go through a different path.
5716 // In order to flush those, we need to create artificial events in the .git folder and wait
5717 // for the repository to be reloaded.
5718 #[cfg(feature = "test-support")]
5719 fn flush_fs_events_in_root_git_repository<'a>(
5720 &self,
5721 cx: &'a mut gpui::TestAppContext,
5722 ) -> futures::future::LocalBoxFuture<'a, ()> {
5723 let file_name = "fs-event-sentinel";
5724
5725 let tree = self.clone();
5726 let (fs, root_path, mut git_dir_scan_id) = self.read_with(cx, |tree, _| {
5727 let tree = tree.as_local().unwrap();
5728 let local_repo_entry = tree
5729 .git_repositories
5730 .values()
5731 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5732 .unwrap();
5733 (
5734 tree.fs.clone(),
5735 local_repo_entry.common_dir_abs_path.clone(),
5736 local_repo_entry.git_dir_scan_id,
5737 )
5738 });
5739
5740 let scan_id_increased = |tree: &mut Worktree, git_dir_scan_id: &mut usize| {
5741 let tree = tree.as_local().unwrap();
5742 // let repository = tree.repositories.first().unwrap();
5743 let local_repo_entry = tree
5744 .git_repositories
5745 .values()
5746 .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone())
5747 .unwrap();
5748
5749 if local_repo_entry.git_dir_scan_id > *git_dir_scan_id {
5750 *git_dir_scan_id = local_repo_entry.git_dir_scan_id;
5751 true
5752 } else {
5753 false
5754 }
5755 };
5756
5757 async move {
5758 // Subscribe to events BEFORE creating the file to avoid race condition
5759 // where events fire before subscription is set up
5760 let mut events = cx.events(&tree);
5761
5762 fs.create_file(&root_path.join(file_name), Default::default())
5763 .await
5764 .unwrap();
5765
5766 // Use select to avoid blocking indefinitely if events are delayed
5767 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5768 futures::select_biased! {
5769 _ = events.next() => {}
5770 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5771 }
5772 }
5773
5774 fs.remove_file(&root_path.join(file_name), Default::default())
5775 .await
5776 .unwrap();
5777
5778 // Use select to avoid blocking indefinitely if events are delayed
5779 while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) {
5780 futures::select_biased! {
5781 _ = events.next() => {}
5782 _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => {}
5783 }
5784 }
5785
5786 cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete())
5787 .await;
5788 }
5789 .boxed_local()
5790 }
5791}
5792
5793#[derive(Clone, Debug)]
5794struct TraversalProgress<'a> {
5795 max_path: &'a RelPath,
5796 count: usize,
5797 non_ignored_count: usize,
5798 file_count: usize,
5799 non_ignored_file_count: usize,
5800}
5801
5802impl TraversalProgress<'_> {
5803 fn count(&self, include_files: bool, include_dirs: bool, include_ignored: bool) -> usize {
5804 match (include_files, include_dirs, include_ignored) {
5805 (true, true, true) => self.count,
5806 (true, true, false) => self.non_ignored_count,
5807 (true, false, true) => self.file_count,
5808 (true, false, false) => self.non_ignored_file_count,
5809 (false, true, true) => self.count - self.file_count,
5810 (false, true, false) => self.non_ignored_count - self.non_ignored_file_count,
5811 (false, false, _) => 0,
5812 }
5813 }
5814}
5815
5816impl<'a> sum_tree::Dimension<'a, EntrySummary> for TraversalProgress<'a> {
5817 fn zero(_cx: ()) -> Self {
5818 Default::default()
5819 }
5820
5821 fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) {
5822 self.max_path = summary.max_path.as_ref();
5823 self.count += summary.count;
5824 self.non_ignored_count += summary.non_ignored_count;
5825 self.file_count += summary.file_count;
5826 self.non_ignored_file_count += summary.non_ignored_file_count;
5827 }
5828}
5829
5830impl Default for TraversalProgress<'_> {
5831 fn default() -> Self {
5832 Self {
5833 max_path: RelPath::empty(),
5834 count: 0,
5835 non_ignored_count: 0,
5836 file_count: 0,
5837 non_ignored_file_count: 0,
5838 }
5839 }
5840}
5841
5842#[derive(Debug)]
5843pub struct Traversal<'a> {
5844 snapshot: &'a Snapshot,
5845 cursor: sum_tree::Cursor<'a, 'static, Entry, TraversalProgress<'a>>,
5846 include_ignored: bool,
5847 include_files: bool,
5848 include_dirs: bool,
5849}
5850
5851impl<'a> Traversal<'a> {
5852 fn new(
5853 snapshot: &'a Snapshot,
5854 include_files: bool,
5855 include_dirs: bool,
5856 include_ignored: bool,
5857 start_path: &RelPath,
5858 ) -> Self {
5859 let mut cursor = snapshot.entries_by_path.cursor(());
5860 cursor.seek(&TraversalTarget::path(start_path), Bias::Left);
5861 let mut traversal = Self {
5862 snapshot,
5863 cursor,
5864 include_files,
5865 include_dirs,
5866 include_ignored,
5867 };
5868 if traversal.end_offset() == traversal.start_offset() {
5869 traversal.next();
5870 }
5871 traversal
5872 }
5873
5874 pub fn advance(&mut self) -> bool {
5875 self.advance_by(1)
5876 }
5877
5878 pub fn advance_by(&mut self, count: usize) -> bool {
5879 self.cursor.seek_forward(
5880 &TraversalTarget::Count {
5881 count: self.end_offset() + count,
5882 include_dirs: self.include_dirs,
5883 include_files: self.include_files,
5884 include_ignored: self.include_ignored,
5885 },
5886 Bias::Left,
5887 )
5888 }
5889
5890 pub fn advance_to_sibling(&mut self) -> bool {
5891 while let Some(entry) = self.cursor.item() {
5892 self.cursor
5893 .seek_forward(&TraversalTarget::successor(&entry.path), Bias::Left);
5894 if let Some(entry) = self.cursor.item()
5895 && (self.include_files || !entry.is_file())
5896 && (self.include_dirs || !entry.is_dir())
5897 && (self.include_ignored || !entry.is_ignored || entry.is_always_included)
5898 {
5899 return true;
5900 }
5901 }
5902 false
5903 }
5904
5905 pub fn back_to_parent(&mut self) -> bool {
5906 let Some(parent_path) = self.cursor.item().and_then(|entry| entry.path.parent()) else {
5907 return false;
5908 };
5909 self.cursor
5910 .seek(&TraversalTarget::path(parent_path), Bias::Left)
5911 }
5912
5913 pub fn entry(&self) -> Option<&'a Entry> {
5914 self.cursor.item()
5915 }
5916
5917 pub fn snapshot(&self) -> &'a Snapshot {
5918 self.snapshot
5919 }
5920
5921 pub fn start_offset(&self) -> usize {
5922 self.cursor
5923 .start()
5924 .count(self.include_files, self.include_dirs, self.include_ignored)
5925 }
5926
5927 pub fn end_offset(&self) -> usize {
5928 self.cursor
5929 .end()
5930 .count(self.include_files, self.include_dirs, self.include_ignored)
5931 }
5932}
5933
5934impl<'a> Iterator for Traversal<'a> {
5935 type Item = &'a Entry;
5936
5937 fn next(&mut self) -> Option<Self::Item> {
5938 if let Some(item) = self.entry() {
5939 self.advance();
5940 Some(item)
5941 } else {
5942 None
5943 }
5944 }
5945}
5946
5947#[derive(Debug, Clone, Copy)]
5948pub enum PathTarget<'a> {
5949 Path(&'a RelPath),
5950 Successor(&'a RelPath),
5951}
5952
5953impl PathTarget<'_> {
5954 fn cmp_path(&self, other: &RelPath) -> Ordering {
5955 match self {
5956 PathTarget::Path(path) => path.cmp(&other),
5957 PathTarget::Successor(path) => {
5958 if other.starts_with(path) {
5959 Ordering::Greater
5960 } else {
5961 Ordering::Equal
5962 }
5963 }
5964 }
5965 }
5966}
5967
5968impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, PathProgress<'a>> for PathTarget<'_> {
5969 fn cmp(&self, cursor_location: &PathProgress<'a>, _: S::Context<'_>) -> Ordering {
5970 self.cmp_path(cursor_location.max_path)
5971 }
5972}
5973
5974impl<'a, S: Summary> SeekTarget<'a, PathSummary<S>, TraversalProgress<'a>> for PathTarget<'_> {
5975 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: S::Context<'_>) -> Ordering {
5976 self.cmp_path(cursor_location.max_path)
5977 }
5978}
5979
5980#[derive(Debug)]
5981enum TraversalTarget<'a> {
5982 Path(PathTarget<'a>),
5983 Count {
5984 count: usize,
5985 include_files: bool,
5986 include_ignored: bool,
5987 include_dirs: bool,
5988 },
5989}
5990
5991impl<'a> TraversalTarget<'a> {
5992 fn path(path: &'a RelPath) -> Self {
5993 Self::Path(PathTarget::Path(path))
5994 }
5995
5996 fn successor(path: &'a RelPath) -> Self {
5997 Self::Path(PathTarget::Successor(path))
5998 }
5999
6000 fn cmp_progress(&self, progress: &TraversalProgress) -> Ordering {
6001 match self {
6002 TraversalTarget::Path(path) => path.cmp_path(progress.max_path),
6003 TraversalTarget::Count {
6004 count,
6005 include_files,
6006 include_dirs,
6007 include_ignored,
6008 } => Ord::cmp(
6009 count,
6010 &progress.count(*include_files, *include_dirs, *include_ignored),
6011 ),
6012 }
6013 }
6014}
6015
6016impl<'a> SeekTarget<'a, EntrySummary, TraversalProgress<'a>> for TraversalTarget<'_> {
6017 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
6018 self.cmp_progress(cursor_location)
6019 }
6020}
6021
6022impl<'a> SeekTarget<'a, PathSummary<sum_tree::NoSummary>, TraversalProgress<'a>>
6023 for TraversalTarget<'_>
6024{
6025 fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering {
6026 self.cmp_progress(cursor_location)
6027 }
6028}
6029
6030pub struct ChildEntriesOptions {
6031 pub include_files: bool,
6032 pub include_dirs: bool,
6033 pub include_ignored: bool,
6034}
6035
6036pub struct ChildEntriesIter<'a> {
6037 parent_path: &'a RelPath,
6038 traversal: Traversal<'a>,
6039}
6040
6041impl<'a> Iterator for ChildEntriesIter<'a> {
6042 type Item = &'a Entry;
6043
6044 fn next(&mut self) -> Option<Self::Item> {
6045 if let Some(item) = self.traversal.entry()
6046 && item.path.starts_with(self.parent_path)
6047 {
6048 self.traversal.advance_to_sibling();
6049 return Some(item);
6050 }
6051 None
6052 }
6053}
6054
6055impl<'a> From<&'a Entry> for proto::Entry {
6056 fn from(entry: &'a Entry) -> Self {
6057 Self {
6058 id: entry.id.to_proto(),
6059 is_dir: entry.is_dir(),
6060 path: entry.path.as_ref().to_proto(),
6061 inode: entry.inode,
6062 mtime: entry.mtime.map(|time| time.into()),
6063 is_ignored: entry.is_ignored,
6064 is_hidden: entry.is_hidden,
6065 is_external: entry.is_external,
6066 is_fifo: entry.is_fifo,
6067 size: Some(entry.size),
6068 canonical_path: entry
6069 .canonical_path
6070 .as_ref()
6071 .map(|path| path.to_string_lossy().into_owned()),
6072 }
6073 }
6074}
6075
6076impl TryFrom<(&CharBag, &PathMatcher, proto::Entry)> for Entry {
6077 type Error = anyhow::Error;
6078
6079 fn try_from(
6080 (root_char_bag, always_included, entry): (&CharBag, &PathMatcher, proto::Entry),
6081 ) -> Result<Self> {
6082 let kind = if entry.is_dir {
6083 EntryKind::Dir
6084 } else {
6085 EntryKind::File
6086 };
6087
6088 let path =
6089 RelPath::from_proto(&entry.path).context("invalid relative path in proto message")?;
6090 let char_bag = char_bag_for_path(*root_char_bag, &path);
6091 let is_always_included = always_included.is_match(&path);
6092 Ok(Entry {
6093 id: ProjectEntryId::from_proto(entry.id),
6094 kind,
6095 path,
6096 inode: entry.inode,
6097 mtime: entry.mtime.map(|time| time.into()),
6098 size: entry.size.unwrap_or(0),
6099 canonical_path: entry
6100 .canonical_path
6101 .map(|path_string| Arc::from(PathBuf::from(path_string))),
6102 is_ignored: entry.is_ignored,
6103 is_hidden: entry.is_hidden,
6104 is_always_included,
6105 is_external: entry.is_external,
6106 is_private: false,
6107 char_bag,
6108 is_fifo: entry.is_fifo,
6109 })
6110 }
6111}
6112
6113#[derive(Clone, Copy, Debug, Default, Hash, PartialEq, Eq, PartialOrd, Ord)]
6114pub struct ProjectEntryId(usize);
6115
6116impl ProjectEntryId {
6117 pub const MAX: Self = Self(usize::MAX);
6118 pub const MIN: Self = Self(usize::MIN);
6119
6120 pub fn new(counter: &AtomicUsize) -> Self {
6121 Self(counter.fetch_add(1, SeqCst))
6122 }
6123
6124 pub fn from_proto(id: u64) -> Self {
6125 Self(id as usize)
6126 }
6127
6128 pub fn to_proto(self) -> u64 {
6129 self.0 as u64
6130 }
6131
6132 pub fn from_usize(id: usize) -> Self {
6133 ProjectEntryId(id)
6134 }
6135
6136 pub fn to_usize(self) -> usize {
6137 self.0
6138 }
6139}
6140
6141#[cfg(feature = "test-support")]
6142impl CreatedEntry {
6143 pub fn into_included(self) -> Option<Entry> {
6144 match self {
6145 CreatedEntry::Included(entry) => Some(entry),
6146 CreatedEntry::Excluded { .. } => None,
6147 }
6148 }
6149}
6150
6151fn parse_gitfile(content: &str) -> anyhow::Result<&Path> {
6152 let path = content
6153 .strip_prefix("gitdir:")
6154 .with_context(|| format!("parsing gitfile content {content:?}"))?;
6155 Ok(Path::new(path.trim()))
6156}
6157
6158pub async fn discover_root_repo_common_dir(root_abs_path: &Path, fs: &dyn Fs) -> Option<Arc<Path>> {
6159 let root_dot_git = root_abs_path.join(DOT_GIT);
6160 if !fs.metadata(&root_dot_git).await.is_ok_and(|m| m.is_some()) {
6161 return None;
6162 }
6163 let dot_git_path: Arc<Path> = root_dot_git.into();
6164 let (_, common_dir) = discover_git_paths(&dot_git_path, fs).await;
6165 Some(common_dir)
6166}
6167
6168async fn discover_git_paths(dot_git_abs_path: &Arc<Path>, fs: &dyn Fs) -> (Arc<Path>, Arc<Path>) {
6169 let mut repository_dir_abs_path = dot_git_abs_path.clone();
6170 let mut common_dir_abs_path = dot_git_abs_path.clone();
6171
6172 if let Some(path) = fs
6173 .load(dot_git_abs_path)
6174 .await
6175 .ok()
6176 .as_ref()
6177 .and_then(|contents| parse_gitfile(contents).log_err())
6178 {
6179 let path = dot_git_abs_path
6180 .parent()
6181 .unwrap_or(Path::new(""))
6182 .join(path);
6183 if let Some(path) = fs.canonicalize(&path).await.log_err() {
6184 repository_dir_abs_path = Path::new(&path).into();
6185 common_dir_abs_path = repository_dir_abs_path.clone();
6186
6187 if let Some(commondir_contents) = fs.load(&path.join("commondir")).await.ok()
6188 && let Some(commondir_path) = fs
6189 .canonicalize(&path.join(commondir_contents.trim()))
6190 .await
6191 .log_err()
6192 {
6193 common_dir_abs_path = commondir_path.as_path().into();
6194 }
6195 }
6196 };
6197 (repository_dir_abs_path, common_dir_abs_path)
6198}
6199
6200struct NullWatcher;
6201
6202impl fs::Watcher for NullWatcher {
6203 fn add(&self, _path: &Path) -> Result<()> {
6204 Ok(())
6205 }
6206
6207 fn remove(&self, _path: &Path) -> Result<()> {
6208 Ok(())
6209 }
6210}
6211
6212const FILE_ANALYSIS_BYTES: usize = 1024;
6213
6214async fn decode_file_text(
6215 fs: &dyn Fs,
6216 abs_path: &Path,
6217) -> Result<(String, &'static Encoding, bool)> {
6218 let mut file = fs
6219 .open_sync(&abs_path)
6220 .await
6221 .with_context(|| format!("opening file {abs_path:?}"))?;
6222
6223 // First, read the beginning of the file to determine its kind and encoding.
6224 // We do not want to load an entire large blob into memory only to discard it.
6225 let mut file_first_bytes = Vec::with_capacity(FILE_ANALYSIS_BYTES);
6226 let mut buf = [0u8; FILE_ANALYSIS_BYTES];
6227 let mut reached_eof = false;
6228 loop {
6229 if file_first_bytes.len() >= FILE_ANALYSIS_BYTES {
6230 break;
6231 }
6232 let n = file
6233 .read(&mut buf)
6234 .with_context(|| format!("reading bytes of the file {abs_path:?}"))?;
6235 if n == 0 {
6236 reached_eof = true;
6237 break;
6238 }
6239 file_first_bytes.extend_from_slice(&buf[..n]);
6240 }
6241 let (bom_encoding, byte_content) = decode_byte_header(&file_first_bytes);
6242 anyhow::ensure!(
6243 byte_content != ByteContent::Binary,
6244 "Binary files are not supported"
6245 );
6246
6247 // If the file is eligible for opening, read the rest of the file.
6248 let mut content = file_first_bytes;
6249 if !reached_eof {
6250 let mut buf = [0u8; 8 * 1024];
6251 loop {
6252 let n = file
6253 .read(&mut buf)
6254 .with_context(|| format!("reading remaining bytes of the file {abs_path:?}"))?;
6255 if n == 0 {
6256 break;
6257 }
6258 content.extend_from_slice(&buf[..n]);
6259 }
6260 }
6261 decode_byte_full(content, bom_encoding, byte_content)
6262}
6263
6264fn decode_byte_header(prefix: &[u8]) -> (Option<&'static Encoding>, ByteContent) {
6265 if let Some((encoding, _bom_len)) = Encoding::for_bom(prefix) {
6266 return (Some(encoding), ByteContent::Unknown);
6267 }
6268 (None, analyze_byte_content(prefix))
6269}
6270
6271fn decode_byte_full(
6272 bytes: Vec<u8>,
6273 bom_encoding: Option<&'static Encoding>,
6274 byte_content: ByteContent,
6275) -> Result<(String, &'static Encoding, bool)> {
6276 if let Some(encoding) = bom_encoding {
6277 let (cow, _) = encoding.decode_with_bom_removal(&bytes);
6278 return Ok((cow.into_owned(), encoding, true));
6279 }
6280
6281 match byte_content {
6282 ByteContent::Utf16Le => {
6283 let encoding = encoding_rs::UTF_16LE;
6284 let (cow, _, _) = encoding.decode(&bytes);
6285 return Ok((cow.into_owned(), encoding, false));
6286 }
6287 ByteContent::Utf16Be => {
6288 let encoding = encoding_rs::UTF_16BE;
6289 let (cow, _, _) = encoding.decode(&bytes);
6290 return Ok((cow.into_owned(), encoding, false));
6291 }
6292 ByteContent::Binary => {
6293 anyhow::bail!("Binary files are not supported");
6294 }
6295 ByteContent::Unknown => {}
6296 }
6297
6298 fn detect_encoding(bytes: Vec<u8>) -> (String, &'static Encoding) {
6299 let mut detector = EncodingDetector::new();
6300 detector.feed(&bytes, true);
6301
6302 let encoding = detector.guess(None, true); // Use None for TLD hint to ensure neutral detection logic.
6303
6304 let (cow, _, _) = encoding.decode(&bytes);
6305 (cow.into_owned(), encoding)
6306 }
6307
6308 match String::from_utf8(bytes) {
6309 Ok(text) => {
6310 // ISO-2022-JP (and other ISO-2022 variants) consists entirely of 7-bit ASCII bytes,
6311 // so it is valid UTF-8. However, it contains escape sequences starting with '\x1b'.
6312 // If we find an escape character, we double-check the encoding to prevent
6313 // displaying raw escape sequences instead of the correct characters.
6314 if text.contains('\x1b') {
6315 let (s, enc) = detect_encoding(text.into_bytes());
6316 Ok((s, enc, false))
6317 } else {
6318 Ok((text, encoding_rs::UTF_8, false))
6319 }
6320 }
6321 Err(e) => {
6322 let (s, enc) = detect_encoding(e.into_bytes());
6323 Ok((s, enc, false))
6324 }
6325 }
6326}
6327
6328#[derive(Debug, PartialEq)]
6329enum ByteContent {
6330 Utf16Le,
6331 Utf16Be,
6332 Binary,
6333 Unknown,
6334}
6335
6336// Heuristic check using null byte distribution plus a generic text-likeness
6337// heuristic. This prefers UTF-16 when many bytes are NUL and otherwise
6338// distinguishes between text-like and binary-like content.
6339fn analyze_byte_content(bytes: &[u8]) -> ByteContent {
6340 if bytes.len() < 2 {
6341 return ByteContent::Unknown;
6342 }
6343
6344 if is_known_binary_header(bytes) {
6345 return ByteContent::Binary;
6346 }
6347
6348 let limit = bytes.len().min(FILE_ANALYSIS_BYTES);
6349 let mut even_null_count = 0usize;
6350 let mut odd_null_count = 0usize;
6351 let mut non_text_like_count = 0usize;
6352
6353 for (i, &byte) in bytes[..limit].iter().enumerate() {
6354 if byte == 0 {
6355 if i % 2 == 0 {
6356 even_null_count += 1;
6357 } else {
6358 odd_null_count += 1;
6359 }
6360 non_text_like_count += 1;
6361 continue;
6362 }
6363
6364 let is_text_like = match byte {
6365 b'\t' | b'\n' | b'\r' | 0x0C => true,
6366 0x20..=0x7E => true,
6367 // Treat bytes that are likely part of UTF-8 or single-byte encodings as text-like.
6368 0x80..=0xBF | 0xC2..=0xF4 => true,
6369 _ => false,
6370 };
6371
6372 if !is_text_like {
6373 non_text_like_count += 1;
6374 }
6375 }
6376
6377 let total_null_count = even_null_count + odd_null_count;
6378
6379 // If there are no NUL bytes at all, this is overwhelmingly likely to be text.
6380 if total_null_count == 0 {
6381 return ByteContent::Unknown;
6382 }
6383
6384 let has_significant_nulls = total_null_count >= limit / 16;
6385 let nulls_skew_to_even = even_null_count > odd_null_count * 4;
6386 let nulls_skew_to_odd = odd_null_count > even_null_count * 4;
6387
6388 if has_significant_nulls {
6389 let sample = &bytes[..limit];
6390
6391 // UTF-16BE ASCII: [0x00, char] — nulls at even positions (high byte first)
6392 // UTF-16LE ASCII: [char, 0x00] — nulls at odd positions (low byte first)
6393
6394 if nulls_skew_to_even && is_plausible_utf16_text(sample, false) {
6395 return ByteContent::Utf16Be;
6396 }
6397
6398 if nulls_skew_to_odd && is_plausible_utf16_text(sample, true) {
6399 return ByteContent::Utf16Le;
6400 }
6401
6402 return ByteContent::Binary;
6403 }
6404
6405 if non_text_like_count * 100 < limit * 8 {
6406 ByteContent::Unknown
6407 } else {
6408 ByteContent::Binary
6409 }
6410}
6411
6412fn is_known_binary_header(bytes: &[u8]) -> bool {
6413 bytes.starts_with(b"%PDF-") // PDF
6414 || bytes.starts_with(b"PK\x03\x04") // ZIP local header
6415 || bytes.starts_with(b"PK\x05\x06") // ZIP end of central directory
6416 || bytes.starts_with(b"PK\x07\x08") // ZIP spanning/splitting
6417 || bytes.starts_with(b"\x89PNG\r\n\x1a\n") // PNG
6418 || bytes.starts_with(b"\xFF\xD8\xFF") // JPEG
6419 || bytes.starts_with(b"GIF87a") // GIF87a
6420 || bytes.starts_with(b"GIF89a") // GIF89a
6421 || bytes.starts_with(b"IWAD") // Doom IWAD archive
6422 || bytes.starts_with(b"PWAD") // Doom PWAD archive
6423 || bytes.starts_with(b"RIFF") // WAV, AVI, WebP
6424 || bytes.starts_with(b"OggS") // OGG (Vorbis, Opus, FLAC)
6425 || bytes.starts_with(b"fLaC") // FLAC
6426 || bytes.starts_with(b"ID3") // MP3 with ID3v2 tag
6427 || bytes.starts_with(b"\xFF\xFB") // MP3 frame sync (MPEG1 Layer3)
6428 || bytes.starts_with(b"\xFF\xFA") // MP3 frame sync (MPEG1 Layer3)
6429 || bytes.starts_with(b"\xFF\xF3") // MP3 frame sync (MPEG2 Layer3)
6430 || bytes.starts_with(b"\xFF\xF2") // MP3 frame sync (MPEG2 Layer3)
6431}
6432
6433// Null byte skew alone is not enough to identify UTF-16 -- binary formats with
6434// small 16-bit values (like PCM audio) produce the same pattern. Decode the
6435// bytes as UTF-16 and reject if too many code units land in control character
6436// ranges or form unpaired surrogates, which real text almost never contains.
6437fn is_plausible_utf16_text(bytes: &[u8], little_endian: bool) -> bool {
6438 let mut suspicious_count = 0usize;
6439 let mut total = 0usize;
6440
6441 let mut i = 0;
6442 while let Some(code_unit) = read_u16(bytes, i, little_endian) {
6443 total += 1;
6444
6445 match code_unit {
6446 0x0009 | 0x000A | 0x000C | 0x000D => {}
6447 // C0/C1 control characters and non-characters
6448 0x0000..=0x001F | 0x007F..=0x009F | 0xFFFE | 0xFFFF => suspicious_count += 1,
6449 0xD800..=0xDBFF => {
6450 let next_offset = i + 2;
6451 let has_low_surrogate = read_u16(bytes, next_offset, little_endian)
6452 .is_some_and(|next| (0xDC00..=0xDFFF).contains(&next));
6453 if has_low_surrogate {
6454 total += 1;
6455 i += 2;
6456 } else {
6457 suspicious_count += 1;
6458 }
6459 }
6460 // Lone low surrogate without a preceding high surrogate
6461 0xDC00..=0xDFFF => suspicious_count += 1,
6462 _ => {}
6463 }
6464
6465 i += 2;
6466 }
6467
6468 if total == 0 {
6469 return false;
6470 }
6471
6472 // Real UTF-16 text has near-zero control characters; binary data with
6473 // small 16-bit values typically exceeds 5%. 2% provides a safe margin.
6474 suspicious_count * 100 < total * 2
6475}
6476
6477fn read_u16(bytes: &[u8], offset: usize, little_endian: bool) -> Option<u16> {
6478 let pair = [*bytes.get(offset)?, *bytes.get(offset + 1)?];
6479 if little_endian {
6480 return Some(u16::from_le_bytes(pair));
6481 }
6482 Some(u16::from_be_bytes(pair))
6483}
6484
6485#[cfg(test)]
6486mod tests {
6487 use super::*;
6488
6489 /// reproduction of issue #50785
6490 fn build_pcm16_wav_bytes() -> Vec<u8> {
6491 let header: Vec<u8> = vec![
6492 /* RIFF header */
6493 0x52, 0x49, 0x46, 0x46, // "RIFF"
6494 0xc6, 0xcf, 0x00, 0x00, // file size: 8
6495 0x57, 0x41, 0x56, 0x45, // "WAVE"
6496 /* fmt chunk */
6497 0x66, 0x6d, 0x74, 0x20, // "fmt "
6498 0x10, 0x00, 0x00, 0x00, // chunk size: 16
6499 0x01, 0x00, // format: PCM (1)
6500 0x01, 0x00, // channels: 1 (mono)
6501 0x80, 0x3e, 0x00, 0x00, // sample rate: 16000
6502 0x00, 0x7d, 0x00, 0x00, // byte rate: 32000
6503 0x02, 0x00, // block align: 2
6504 0x10, 0x00, // bits per sample: 16
6505 /* LIST chunk */
6506 0x4c, 0x49, 0x53, 0x54, // "LIST"
6507 0x1a, 0x00, 0x00, 0x00, // chunk size: 26
6508 0x49, 0x4e, 0x46, 0x4f, // "INFO"
6509 0x49, 0x53, 0x46, 0x54, // "ISFT"
6510 0x0d, 0x00, 0x00, 0x00, // sub-chunk size: 13
6511 0x4c, 0x61, 0x76, 0x66, 0x36, 0x32, 0x2e, 0x33, // "Lavf62.3"
6512 0x2e, 0x31, 0x30, 0x30, 0x00, // ".100\0"
6513 /* padding byte for word alignment */
6514 0x00, // data chunk header
6515 0x64, 0x61, 0x74, 0x61, // "data"
6516 0x80, 0xcf, 0x00, 0x00, // chunk size
6517 ];
6518
6519 let mut bytes = header;
6520
6521 // fill remaining space up to `FILE_ANALYSIS_BYTES` with synthetic PCM
6522 let audio_bytes_needed = FILE_ANALYSIS_BYTES - bytes.len();
6523 for i in 0..(audio_bytes_needed / 2) {
6524 let sample = (i & 0xFF) as u8;
6525 bytes.push(sample); // low byte: varies
6526 bytes.push(0x00); // high byte: zero for small values
6527 }
6528
6529 bytes
6530 }
6531
6532 #[test]
6533 fn test_pcm16_wav_detected_as_binary() {
6534 let wav_bytes = build_pcm16_wav_bytes();
6535 assert_eq!(wav_bytes.len(), FILE_ANALYSIS_BYTES);
6536
6537 let result = analyze_byte_content(&wav_bytes);
6538 assert_eq!(
6539 result,
6540 ByteContent::Binary,
6541 "PCM 16-bit WAV should be detected as Binary via RIFF header"
6542 );
6543 }
6544
6545 #[test]
6546 fn test_le16_binary_not_misdetected_as_utf16le() {
6547 let mut bytes = b"FAKE".to_vec();
6548 while bytes.len() < FILE_ANALYSIS_BYTES {
6549 let sample = (bytes.len() & 0xFF) as u8;
6550 bytes.push(sample);
6551 bytes.push(0x00);
6552 }
6553 bytes.truncate(FILE_ANALYSIS_BYTES);
6554
6555 let result = analyze_byte_content(&bytes);
6556 assert_eq!(
6557 result,
6558 ByteContent::Binary,
6559 "LE 16-bit binary with control characters should be detected as Binary"
6560 );
6561 }
6562
6563 #[test]
6564 fn test_be16_binary_not_misdetected_as_utf16be() {
6565 let mut bytes = b"FAKE".to_vec();
6566 while bytes.len() < FILE_ANALYSIS_BYTES {
6567 bytes.push(0x00);
6568 let sample = (bytes.len() & 0xFF) as u8;
6569 bytes.push(sample);
6570 }
6571 bytes.truncate(FILE_ANALYSIS_BYTES);
6572
6573 let result = analyze_byte_content(&bytes);
6574 assert_eq!(
6575 result,
6576 ByteContent::Binary,
6577 "BE 16-bit binary with control characters should be detected as Binary"
6578 );
6579 }
6580
6581 #[test]
6582 fn test_utf16le_text_detected_as_utf16le() {
6583 let text = "Hello, world! This is a UTF-16 test string. ";
6584 let mut bytes = Vec::new();
6585 while bytes.len() < FILE_ANALYSIS_BYTES {
6586 bytes.extend(text.encode_utf16().flat_map(|u| u.to_le_bytes()));
6587 }
6588 bytes.truncate(FILE_ANALYSIS_BYTES);
6589
6590 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Le);
6591 }
6592
6593 #[test]
6594 fn test_utf16be_text_detected_as_utf16be() {
6595 let text = "Hello, world! This is a UTF-16 test string. ";
6596 let mut bytes = Vec::new();
6597 while bytes.len() < FILE_ANALYSIS_BYTES {
6598 bytes.extend(text.encode_utf16().flat_map(|u| u.to_be_bytes()));
6599 }
6600 bytes.truncate(FILE_ANALYSIS_BYTES);
6601
6602 assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Be);
6603 }
6604
6605 #[test]
6606 fn test_known_binary_headers() {
6607 let cases: &[(&[u8], &str)] = &[
6608 (b"RIFF\x00\x00\x00\x00WAVE", "WAV"),
6609 (b"RIFF\x00\x00\x00\x00AVI ", "AVI"),
6610 (b"OggS\x00\x02", "OGG"),
6611 (b"fLaC\x00\x00", "FLAC"),
6612 (b"ID3\x03\x00", "MP3 ID3v2"),
6613 (b"\xFF\xFB\x90\x00", "MP3 MPEG1 Layer3"),
6614 (b"\xFF\xF3\x90\x00", "MP3 MPEG2 Layer3"),
6615 ];
6616
6617 for (header, label) in cases {
6618 let mut bytes = header.to_vec();
6619 bytes.resize(FILE_ANALYSIS_BYTES, 0x41); // pad with 'A'
6620 assert_eq!(
6621 analyze_byte_content(&bytes),
6622 ByteContent::Binary,
6623 "{label} should be detected as Binary"
6624 );
6625 }
6626 }
6627}