mod ignore; mod worktree_settings; use ::ignore::gitignore::{Gitignore, GitignoreBuilder}; use anyhow::{Context as _, Result, anyhow}; use clock::ReplicaId; use collections::{BTreeMap, HashMap, HashSet, VecDeque}; use encoding_rs::Encoding; use fs::{ Fs, MTime, PathEvent, PathEventKind, RemoveOptions, TrashId, Watcher, copy_recursive, read_dir_items, }; use futures::{ FutureExt as _, Stream, StreamExt, channel::{ mpsc::{self, UnboundedSender}, oneshot, }, select_biased, stream, task::Poll, }; use futures_lite::future::yield_now; use fuzzy::CharBag; use git::{ BISECT_LOG, COMMIT_MESSAGE, DOT_GIT, FETCH_HEAD, FSMONITOR_DAEMON, GC_PID, GITIGNORE, HOOKS_DIR, INFO_DIR, LFS_DIR, LOGS_DIR, LOGS_REF_STASH, OBJECTS_DIR, ORIG_HEAD, REBASE_APPLY_DIR, REBASE_MERGE_DIR, REFS_DIR, REFTABLE_DIR, REPO_EXCLUDE, SEQUENCER_DIR, status::GitSummary, }; use gpui::{ App, AppContext as _, AsyncApp, BackgroundExecutor, Context, Entity, EventEmitter, Priority, Task, }; pub use ignore::{IgnoreKind, IgnoreStack}; use language::{ ByteContent, DiskState, FILE_ANALYSIS_BYTES, analyze_byte_content, decode_text, encode_text, }; use async_channel::{self, Sender}; use parking_lot::Mutex; use paths::{local_settings_folder_name, local_vscode_folder_name}; use postage::{ barrier, prelude::{Sink as _, Stream as _}, watch, }; use rpc::{ AnyProtoClient, proto::{self, split_worktree_update}, }; pub use settings::WorktreeId; use settings::{Settings, SettingsLocation, SettingsStore}; use smallvec::{SmallVec, smallvec}; use std::{ any::Any, borrow::Borrow as _, cmp::Ordering, collections::hash_map, convert::TryFrom, ffi::OsStr, fmt, future::Future, io::Read, mem::{self}, ops::{Deref, DerefMut, Range}, path::{Path, PathBuf}, pin::Pin, sync::{ Arc, atomic::{AtomicUsize, Ordering::SeqCst}, }, time::{Duration, Instant}, }; use sum_tree::{Bias, Dimensions, Edit, KeyedItem, SeekTarget, SumTree, Summary, TreeMap, TreeSet}; use text::{LineEnding, Rope}; use util::{ ResultExt, maybe, paths::{PathMatcher, PathStyle, SanitizedPath, home_dir}, rel_path::RelPath, }; pub use worktree_settings::WorktreeSettings; pub const FS_WATCH_LATENCY: Duration = Duration::from_millis(100); /// How often the background scanner verifies that the worktree root still /// exists at its recorded path. Native watchers report the root itself being /// renamed or deleted, but not a rename of one of its ancestors, which leaves /// the watcher silently attached to a path that no longer exists. pub const ROOT_PATH_CHECK_INTERVAL: Duration = Duration::from_secs(5); /// A set of local or remote files that are being opened as part of a project. /// Responsible for tracking related FS (for local)/collab (for remote) events and corresponding updates. /// Stores git repositories data and the diagnostics for the file(s). /// /// Has an absolute path, and may be set to be visible in Zed UI or not. /// May correspond to a directory or a single file. /// Possible examples: /// * a drag and dropped file — may be added as an invisible, "ephemeral" entry to the current worktree /// * a directory opened in Zed — may be added as a visible entry to the current worktree /// /// Uses [`Entry`] to track the state of each file/directory, can look up absolute paths for entries. pub enum Worktree { Local(LocalWorktree), Remote(RemoteWorktree), } /// An entry, created in the worktree. #[derive(Debug)] pub enum CreatedEntry { /// Got created and indexed by the worktree, receiving a corresponding entry. Included(Entry), /// Got created, but not indexed due to falling under exclusion filters. Excluded { abs_path: PathBuf }, } #[derive(Debug)] pub struct LoadedFile { pub file: Arc, pub text: Rope, pub line_ending: LineEnding, pub encoding: &'static Encoding, pub has_bom: bool, pub is_writable: bool, } pub struct LoadedBinaryFile { pub file: Arc, pub content: Vec, } impl fmt::Debug for LoadedBinaryFile { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { f.debug_struct("LoadedBinaryFile") .field("file", &self.file) .field("content_bytes", &self.content.len()) .finish() } } pub struct LocalWorktree { snapshot: LocalSnapshot, scan_requests_tx: async_channel::Sender, path_prefixes_to_scan_tx: async_channel::Sender, is_scanning: (watch::Sender, watch::Receiver), snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>, _background_scanner_tasks: Vec>, update_observer: Option, fs: Arc, fs_case_sensitive: bool, visible: bool, next_entry_id: Arc, settings: WorktreeSettings, share_private_files: bool, scanning_enabled: bool, force_defer_watch: bool, } pub struct PathPrefixScanRequest { path: Arc, done: SmallVec<[barrier::Sender; 1]>, } struct ScanRequest { relative_paths: Vec>, done: SmallVec<[barrier::Sender; 1]>, } pub struct RemoteWorktree { snapshot: Snapshot, background_snapshot: Arc)>>, project_id: u64, client: AnyProtoClient, file_scan_inclusions: PathMatcher, updates_tx: Option>, update_observer: Option>, snapshot_subscriptions: VecDeque<(usize, oneshot::Sender<()>)>, replica_id: ReplicaId, visible: bool, disconnected: bool, received_initial_update: bool, } #[derive(Clone)] pub struct Snapshot { id: WorktreeId, /// The absolute path of the worktree root. abs_path: Arc, path_style: PathStyle, root_name: Arc, root_char_bag: CharBag, entries_by_path: SumTree, entries_by_id: SumTree, root_repo_common_dir: Option>, root_repo_is_linked_worktree: bool, /// A number that increases every time the worktree begins scanning /// a set of paths from the filesystem. This scanning could be caused /// by some operation performed on the worktree, such as reading or /// writing a file, or by an event reported by the filesystem. scan_id: usize, /// The latest scan id that has completed, and whose preceding scans /// have all completed. The current `scan_id` could be more than one /// greater than the `completed_scan_id` if operations are performed /// on the worktree while it is processing a file-system event. completed_scan_id: usize, } /// This path corresponds to the 'content path' of a repository in relation /// to Zed's project root. /// In the majority of the cases, this is the folder that contains the .git folder. /// But if a sub-folder of a git repository is opened, this corresponds to the /// project root and the .git folder is located in a parent directory. #[derive(Clone, Debug, Ord, PartialOrd, Eq, PartialEq, Hash)] pub enum WorkDirectory { InProject { relative_path: Arc, }, AboveProject { absolute_path: Arc, location_in_repo: Arc, }, } impl WorkDirectory { fn path_key(&self) -> PathKey { match self { WorkDirectory::InProject { relative_path } => PathKey(relative_path.clone()), WorkDirectory::AboveProject { .. } => PathKey(RelPath::empty_arc()), } } /// Returns true if the given path is a child of the work directory. /// /// Note that the path may not be a member of this repository, if there /// is a repository in a directory between these two paths /// external .git folder in a parent folder of the project root. #[track_caller] pub fn directory_contains(&self, path: &RelPath) -> bool { match self { WorkDirectory::InProject { relative_path } => path.starts_with(relative_path), WorkDirectory::AboveProject { .. } => true, } } } impl Default for WorkDirectory { fn default() -> Self { Self::InProject { relative_path: Arc::from(RelPath::empty()), } } } #[derive(Clone)] pub struct LocalSnapshot { snapshot: Snapshot, global_gitignore: Option>, /// Exclude files for all git repositories in the worktree, indexed by their absolute path. /// The boolean indicates whether the gitignore needs to be updated. repo_exclude_by_work_dir_abs_path: HashMap, (Arc, bool)>, /// All of the gitignore files in the worktree, indexed by their absolute path. /// The boolean indicates whether the gitignore needs to be updated. ignores_by_parent_abs_path: HashMap, (Arc, bool)>, /// All of the git repositories in the worktree, indexed by the project entry /// id of their parent directory. git_repositories: TreeMap, /// The file handle of the worktree root /// (so we can find it after it's been moved) root_file_handle: Option>, /// Maps canonical absolute paths of externally watched symlinked directories /// to their relative paths within the worktree, used to translate FSEvents /// canonical-path events back to worktree-relative paths. external_canonical_to_relative: BTreeMap, Arc>, } struct BackgroundScannerState { snapshot: LocalSnapshot, symlink_paths_by_target: HashMap, SmallVec<[Arc; 1]>>, scanned_dirs: HashSet, watched_dir_abs_paths_by_entry_id: HashMap>, path_prefixes_to_scan: HashSet>, paths_to_scan: HashSet>, removed_entries: RemovedEntries, changed_paths: Vec>, prev_snapshot: Snapshot, scanning_enabled: bool, } /// The entries that were removed from the snapshot as part of the current /// update. Their entry ids may be re-used if the same inode is discovered /// at a new path, or if the given path is re-created after being deleted. /// /// Symlink aliases inside the worktree share their inode (and usually mtime) /// with the symlink target, so an inode may correspond to several entries. /// The path index allows an exact match to take precedence over the /// inode-based rename heuristics in that case. #[derive(Default)] struct RemovedEntries { current: RemovedEntriesGeneration, previous: RemovedEntriesGeneration, } #[derive(Default)] struct RemovedEntriesGeneration { by_inode: HashMap, by_path: HashMap, Entry>, } impl RemovedEntries { fn insert(&mut self, entry: &Entry) { self.current.insert(entry); } fn take_by_path(&mut self, path: &RelPath, inode: u64) -> Option { self.current .take_by_path(path, inode) .or_else(|| self.previous.take_by_path(path, inode)) } fn take_by_inode(&mut self, inode: u64) -> Option { self.current .take_by_inode(inode) .or_else(|| self.previous.take_by_inode(inode)) } fn rotate(&mut self) -> impl Iterator { let dropped = mem::replace(&mut self.previous, mem::take(&mut self.current)); dropped .by_inode .into_values() .chain(dropped.by_path.into_values()) } } impl RemovedEntriesGeneration { fn insert(&mut self, entry: &Entry) { self.by_path.insert(entry.path.clone(), entry.clone()); match self.by_inode.entry(entry.inode) { hash_map::Entry::Occupied(mut o) => { if entry.id > o.get().id { o.insert(entry.clone()); } } hash_map::Entry::Vacant(v) => { v.insert(entry.clone()); } } } fn take_by_path(&mut self, path: &RelPath, inode: u64) -> Option { if self.by_path.get(path)?.inode != inode { return None; } let removed = self.by_path.remove(path)?; if let hash_map::Entry::Occupied(o) = self.by_inode.entry(removed.inode) && o.get().id == removed.id { o.remove(); } Some(removed) } fn take_by_inode(&mut self, inode: u64) -> Option { let removed = self.by_inode.remove(&inode)?; if let hash_map::Entry::Occupied(o) = self.by_path.entry(removed.path.clone()) && o.get().id == removed.id { o.remove(); } Some(removed) } } #[derive(Clone, Debug, Eq, PartialEq)] struct EventRoot { path: Arc, was_rescanned: bool, } #[derive(Debug, Clone)] struct LocalRepositoryEntry { work_directory_id: ProjectEntryId, work_directory: WorkDirectory, work_directory_abs_path: Arc, git_dir_scan_id: usize, /// Absolute path to the original .git entry that caused us to create this repository. /// /// This is normally a directory, but may be a "gitfile" that points to a directory elsewhere /// (whose path we then store in `repository_dir_abs_path`). dot_git_abs_path: Arc, /// Absolute path to the "commondir" for this repository. /// /// This is always a directory. For a normal repository, this is the same as /// `dot_git_abs_path`. For a linked worktree, this is the main repo's `.git` /// directory (resolved from the worktree's `commondir` file). For a submodule, /// this equals `repository_dir_abs_path` (submodules don't have a `commondir` /// file). common_dir_abs_path: Arc, /// Absolute path to the directory holding the repository's state. /// /// For a normal repository, this is a directory and coincides with `dot_git_abs_path` and /// `common_dir_abs_path`. For a submodule or worktree, this is some subdirectory of the /// commondir like `/project/.git/modules/foo`. repository_dir_abs_path: Arc, } impl sum_tree::Item for LocalRepositoryEntry { type Summary = PathSummary; fn summary(&self, _: ::Context<'_>) -> Self::Summary { PathSummary { max_path: self.work_directory.path_key().0, item_summary: sum_tree::NoSummary, } } } impl KeyedItem for LocalRepositoryEntry { type Key = PathKey; fn key(&self) -> Self::Key { self.work_directory.path_key() } } impl Deref for LocalRepositoryEntry { type Target = WorkDirectory; fn deref(&self) -> &Self::Target { &self.work_directory } } impl Deref for LocalSnapshot { type Target = Snapshot; fn deref(&self) -> &Self::Target { &self.snapshot } } impl DerefMut for LocalSnapshot { fn deref_mut(&mut self) -> &mut Self::Target { &mut self.snapshot } } enum ScanState { Started, Updated { snapshot: LocalSnapshot, changes: UpdatedEntriesSet, barrier: SmallVec<[barrier::Sender; 1]>, scanning: bool, }, RootUpdated { new_path: Arc, }, RootDeleted, } struct UpdateObservationState { snapshots_tx: mpsc::UnboundedSender<(LocalSnapshot, UpdatedEntriesSet)>, resume_updates: watch::Sender<()>, _maintain_remote_snapshot: Task>, } #[derive(Debug, Clone)] pub enum Event { UpdatedEntries(UpdatedEntriesSet), UpdatedGitRepositories(UpdatedGitRepositoriesSet), UpdatedRootRepoCommonDir { old: Option>, }, DeletedEntry(ProjectEntryId), /// The worktree root itself has been deleted (for single-file worktrees) Deleted, } impl EventEmitter for Worktree {} impl Worktree { pub async fn local( path: impl Into>, visible: bool, fs: Arc, next_entry_id: Arc, scanning_enabled: bool, worktree_id: WorktreeId, cx: &mut AsyncApp, ) -> Result> { let abs_path = path.into(); let metadata = fs .metadata(&abs_path) .await .context("failed to stat worktree path")?; let fs_case_sensitive = fs.is_case_sensitive().await; let root_file_handle = if metadata.as_ref().is_some() { fs.open_handle(&abs_path) .await .with_context(|| { format!( "failed to open local worktree root at {}", abs_path.display() ) }) .log_err() } else { None }; let (root_repo_common_dir, root_repo_is_linked_worktree) = if visible { discover_root_repo_metadata(&abs_path, fs.as_ref()) .await .map(|(common_dir, is_linked_worktree)| { ( Some(SanitizedPath::from_arc(common_dir)), is_linked_worktree, ) }) .unwrap_or((None, false)) } else { (None, false) }; Ok(cx.new(move |cx: &mut Context| { let mut snapshot = LocalSnapshot { ignores_by_parent_abs_path: Default::default(), global_gitignore: Default::default(), repo_exclude_by_work_dir_abs_path: Default::default(), git_repositories: Default::default(), external_canonical_to_relative: Default::default(), snapshot: Snapshot::new( worktree_id, abs_path .file_name() .and_then(|f| f.to_str()) .map_or(RelPath::empty_arc(), |f| { RelPath::from_unix_str(f).unwrap().into() }), abs_path.clone(), PathStyle::local(), ), root_file_handle, }; snapshot.root_repo_common_dir = root_repo_common_dir; snapshot.root_repo_is_linked_worktree = root_repo_is_linked_worktree; let worktree_id = snapshot.id(); let settings_location = Some(SettingsLocation { worktree_id, path: RelPath::empty(), }); let settings = WorktreeSettings::get(settings_location, cx).clone(); cx.observe_global::(move |this, cx| { if let Self::Local(this) = this { let settings = WorktreeSettings::get(settings_location, cx).clone(); if this.settings != settings { this.settings = settings; this.restart_background_scanners(cx); } } }) .detach(); let share_private_files = false; if let Some(metadata) = metadata { let mut entry = Entry::new( RelPath::empty_arc(), &metadata, ProjectEntryId::new(&next_entry_id), snapshot.root_char_bag, None, ); if metadata.is_dir { if !scanning_enabled { entry.kind = EntryKind::UnloadedDir; } } else { if let Some(file_name) = abs_path.file_name() && let Some(file_name) = file_name.to_str() && let Ok(path) = RelPath::from_unix_str(file_name) { entry.is_private = !share_private_files && settings.is_path_private(path); entry.is_hidden = settings.is_path_hidden(path); } } cx.foreground_executor() .block_on(snapshot.insert_entry(entry, fs.as_ref())); } let (scan_requests_tx, scan_requests_rx) = async_channel::unbounded(); let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = async_channel::unbounded(); let mut worktree = LocalWorktree { share_private_files, next_entry_id, snapshot, is_scanning: watch::channel_with(true), snapshot_subscriptions: Default::default(), update_observer: None, scan_requests_tx, path_prefixes_to_scan_tx, _background_scanner_tasks: Vec::new(), fs, fs_case_sensitive, visible, settings, scanning_enabled, force_defer_watch: false, }; worktree.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx); Worktree::Local(worktree) })) } pub fn remote( project_id: u64, replica_id: ReplicaId, worktree: proto::WorktreeMetadata, client: AnyProtoClient, path_style: PathStyle, cx: &mut App, ) -> Entity { cx.new(|cx: &mut Context| { let mut snapshot = Snapshot::new( WorktreeId::from_proto(worktree.id), RelPath::from_unix_str(&worktree.root_name) .map_or_else(|_| RelPath::empty_arc(), Into::into), Path::new(&worktree.abs_path).into(), path_style, ); snapshot.root_repo_common_dir = worktree .root_repo_common_dir .map(|p| SanitizedPath::new_arc(Path::new(&p))); snapshot.root_repo_is_linked_worktree = worktree.root_repo_is_linked_worktree; let background_snapshot = Arc::new(Mutex::new(( snapshot.clone(), Vec::::new(), ))); let (background_updates_tx, mut background_updates_rx) = mpsc::unbounded::(); let (mut snapshot_updated_tx, mut snapshot_updated_rx) = watch::channel(); let worktree_id = snapshot.id(); let settings_location = Some(SettingsLocation { worktree_id, path: RelPath::empty(), }); let settings = WorktreeSettings::get(settings_location, cx).clone(); let worktree = RemoteWorktree { client, project_id, replica_id, snapshot, file_scan_inclusions: settings.parent_dir_scan_inclusions.clone(), background_snapshot: background_snapshot.clone(), updates_tx: Some(background_updates_tx), update_observer: None, snapshot_subscriptions: Default::default(), visible: worktree.visible, disconnected: false, received_initial_update: false, }; // Apply updates to a separate snapshot in a background task, then // send them to a foreground task which updates the model. cx.background_spawn(async move { while let Some(update) = background_updates_rx.next().await { { let mut lock = background_snapshot.lock(); lock.0.apply_remote_update( update.clone(), &settings.parent_dir_scan_inclusions, ); lock.1.push(update); } snapshot_updated_tx.send(()).await.ok(); } }) .detach(); // On the foreground task, update to the latest snapshot and notify // any update observer of all updates that led to that snapshot. cx.spawn(async move |this, cx| { while (snapshot_updated_rx.recv().await).is_some() { this.update(cx, |this, cx| { let this = this.as_remote_mut().unwrap(); // The watch channel delivers an initial signal before // any real updates arrive. Skip these spurious wakeups. if this.background_snapshot.lock().1.is_empty() { return; } let old_root_repo_common_dir = this.snapshot.root_repo_common_dir.clone(); let old_root_repo_is_linked_worktree = this.snapshot.root_repo_is_linked_worktree; let mut changed_entries: Vec<(Arc, ProjectEntryId, PathChange)> = Vec::new(); { let mut lock = this.background_snapshot.lock(); // Replace the snapshot, keeping the previous one around so we can // resolve the paths of removed entries (the new snapshot no longer // contains them, and the wire format only carries their ids). let old_snapshot = mem::replace(&mut this.snapshot, lock.0.clone()); for update in lock.1.drain(..) { for entry_id in &update.removed_entries { let entry_id = ProjectEntryId::from_proto(*entry_id); if let Some(entry) = old_snapshot.entry_for_id(entry_id) { changed_entries.push(( entry.path.clone(), entry_id, PathChange::Removed, )); } } for entry in &update.updated_entries { // Remote updates don't distinguish creation from // modification, so report `AddedOrUpdated`. if let Some(path) = RelPath::from_unix_str(&entry.path).log_err() { changed_entries.push(( path.into(), ProjectEntryId::from_proto(entry.id), PathChange::AddedOrUpdated, )); } } if let Some(tx) = &this.update_observer { tx.unbounded_send(update).ok(); } } }; if !changed_entries.is_empty() { cx.emit(Event::UpdatedEntries(changed_entries.into())); } let is_first_update = !this.received_initial_update; this.received_initial_update = true; if this.snapshot.root_repo_common_dir != old_root_repo_common_dir || this.snapshot.root_repo_is_linked_worktree != old_root_repo_is_linked_worktree || (is_first_update && this.snapshot.root_repo_common_dir.is_none()) { cx.emit(Event::UpdatedRootRepoCommonDir { old: old_root_repo_common_dir, }); } cx.notify(); while let Some((scan_id, _)) = this.snapshot_subscriptions.front() { if this.observed_snapshot(*scan_id) { let (_, tx) = this.snapshot_subscriptions.pop_front().unwrap(); let _ = tx.send(()); } else { break; } } })?; } anyhow::Ok(()) }) .detach(); Worktree::Remote(worktree) }) } pub fn as_local(&self) -> Option<&LocalWorktree> { if let Worktree::Local(worktree) = self { Some(worktree) } else { None } } pub fn as_remote(&self) -> Option<&RemoteWorktree> { if let Worktree::Remote(worktree) = self { Some(worktree) } else { None } } pub fn as_local_mut(&mut self) -> Option<&mut LocalWorktree> { if let Worktree::Local(worktree) = self { Some(worktree) } else { None } } pub fn as_remote_mut(&mut self) -> Option<&mut RemoteWorktree> { if let Worktree::Remote(worktree) = self { Some(worktree) } else { None } } pub fn is_local(&self) -> bool { matches!(self, Worktree::Local(_)) } pub fn is_remote(&self) -> bool { !self.is_local() } pub fn settings_location(&self, _: &Context) -> SettingsLocation<'static> { SettingsLocation { worktree_id: self.id(), path: RelPath::empty(), } } pub fn snapshot(&self) -> Snapshot { match self { Worktree::Local(worktree) => worktree.snapshot.snapshot.clone(), Worktree::Remote(worktree) => worktree.snapshot.clone(), } } pub fn scan_id(&self) -> usize { match self { Worktree::Local(worktree) => worktree.snapshot.scan_id, Worktree::Remote(worktree) => worktree.snapshot.scan_id, } } pub fn metadata_proto(&self) -> proto::WorktreeMetadata { proto::WorktreeMetadata { id: self.id().to_proto(), root_name: self.root_name().as_unix_str().to_owned(), visible: self.is_visible(), abs_path: self.abs_path().to_string_lossy().into_owned(), root_repo_common_dir: self .root_repo_common_dir() .map(|p| p.to_string_lossy().into_owned()), root_repo_is_linked_worktree: self.root_repo_is_linked_worktree(), } } pub fn completed_scan_id(&self) -> usize { match self { Worktree::Local(worktree) => worktree.snapshot.completed_scan_id, Worktree::Remote(worktree) => worktree.snapshot.completed_scan_id, } } pub fn is_visible(&self) -> bool { match self { Worktree::Local(worktree) => worktree.visible, Worktree::Remote(worktree) => worktree.visible, } } pub fn replica_id(&self) -> ReplicaId { match self { Worktree::Local(_) => ReplicaId::LOCAL, Worktree::Remote(worktree) => worktree.replica_id, } } pub fn abs_path(&self) -> Arc { match self { Worktree::Local(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()), Worktree::Remote(worktree) => SanitizedPath::cast_arc(worktree.abs_path.clone()), } } pub fn root_file(&self, cx: &Context) -> Option> { let entry = self.root_entry()?; Some(File::for_entry(entry.clone(), cx.entity())) } pub fn observe_updates(&mut self, project_id: u64, cx: &Context, callback: F) where F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut, Fut: 'static + Send + Future, { match self { Worktree::Local(this) => this.observe_updates(project_id, cx, callback), Worktree::Remote(this) => this.observe_updates(project_id, cx, callback), } } pub fn stop_observing_updates(&mut self) { match self { Worktree::Local(this) => { this.update_observer.take(); } Worktree::Remote(this) => { this.update_observer.take(); } } } pub fn wait_for_snapshot( &mut self, scan_id: usize, ) -> impl Future> + use<> { match self { Worktree::Local(this) => this.wait_for_snapshot(scan_id).boxed(), Worktree::Remote(this) => this.wait_for_snapshot(scan_id).boxed(), } } #[cfg(feature = "test-support")] pub fn has_update_observer(&self) -> bool { match self { Worktree::Local(this) => this.update_observer.is_some(), Worktree::Remote(this) => this.update_observer.is_some(), } } pub fn load_file(&self, path: &RelPath, cx: &Context) -> Task> { match self { Worktree::Local(this) => this.load_file(path, cx), Worktree::Remote(_) => { Task::ready(Err(anyhow!("remote worktrees can't yet load files"))) } } } pub fn load_binary_file( &self, path: &RelPath, cx: &Context, ) -> Task> { match self { Worktree::Local(this) => this.load_binary_file(path, cx), Worktree::Remote(_) => { Task::ready(Err(anyhow!("remote worktrees can't yet load binary files"))) } } } pub fn write_file( &self, path: Arc, text: Rope, line_ending: LineEnding, encoding: &'static Encoding, has_bom: bool, cx: &Context, ) -> Task>> { match self { Worktree::Local(this) => { this.write_file(path, text, line_ending, encoding, has_bom, cx) } Worktree::Remote(_) => { Task::ready(Err(anyhow!("remote worktree can't yet write files"))) } } } pub fn create_entry( &mut self, path: Arc, is_directory: bool, content: Option>, cx: &Context, ) -> Task> { let worktree_id = self.id(); match self { Worktree::Local(this) => this.create_entry(path, is_directory, content, cx), Worktree::Remote(this) => { let project_id = this.project_id; let request = this.client.request(proto::CreateProjectEntry { worktree_id: worktree_id.to_proto(), project_id, path: path.as_ref().as_unix_str().to_owned(), content, is_directory, }); cx.spawn(async move |this, cx| { let response = request.await?; match response.entry { Some(entry) => this .update(cx, |worktree, cx| { worktree.as_remote_mut().unwrap().insert_entry( entry, response.worktree_scan_id as usize, cx, ) })? .await .map(CreatedEntry::Included), None => { let abs_path = this.read_with(cx, |worktree, _| worktree.absolutize(&path))?; Ok(CreatedEntry::Excluded { abs_path }) } } }) } } } pub fn trash_entry( &mut self, entry_id: ProjectEntryId, cx: &mut Context, ) -> Option>> { let entry = match self { Worktree::Local(this) => this.entry_for_id(entry_id), Worktree::Remote(this) => this.entry_for_id(entry_id), }? .clone(); let task = match self { Worktree::Local(this) => this.trash_entry(entry.clone(), cx), Worktree::Remote(this) => this.trash_entry(entry_id, cx), }; let mut ids = vec![entry_id]; self.get_children_ids_recursive(&entry.path, &mut ids); for id in ids { cx.emit(Event::DeletedEntry(id)); } Some(task) } pub fn delete_entry( &mut self, entry_id: ProjectEntryId, cx: &mut Context, ) -> Option>> { let entry = match self { Worktree::Local(this) => this.entry_for_id(entry_id), Worktree::Remote(this) => this.entry_for_id(entry_id), }? .clone(); let task = match self { Worktree::Local(this) => this.delete_entry(entry.clone(), cx), Worktree::Remote(this) => this.delete_entry(entry_id, cx), }; let mut ids = vec![entry_id]; let path = entry.path; self.get_children_ids_recursive(&path, &mut ids); for id in ids { cx.emit(Event::DeletedEntry(id)); } Some(task) } pub fn restore_entry( &mut self, trash_id: TrashId, cx: &mut Context<'_, Worktree>, ) -> Task> { match self { Worktree::Local(this) => this.restore_entry(trash_id, cx), Worktree::Remote(this) => this.restore_entry(trash_id, cx), } } fn get_children_ids_recursive(&self, path: &RelPath, ids: &mut Vec) { let children_iter = self.child_entries(path); for child in children_iter { ids.push(child.id); self.get_children_ids_recursive(&child.path, ids); } } pub fn copy_external_entries( &mut self, target_directory: Arc, paths: Vec>, fs: Arc, cx: &Context, ) -> Task>> { match self { Worktree::Local(this) => this.copy_external_entries(target_directory, paths, cx), Worktree::Remote(this) => this.copy_external_entries(target_directory, paths, fs, cx), } } pub fn expand_entry( &mut self, entry_id: ProjectEntryId, cx: &Context, ) -> Option>> { match self { Worktree::Local(this) => this.expand_entry(entry_id, cx), Worktree::Remote(this) => { let response = this.client.request(proto::ExpandProjectEntry { project_id: this.project_id, entry_id: entry_id.to_proto(), }); Some(cx.spawn(async move |this, cx| { let response = response.await?; this.update(cx, |this, _| { this.as_remote_mut() .unwrap() .wait_for_snapshot(response.worktree_scan_id as usize) })? .await?; Ok(()) })) } } } pub fn expand_all_for_entry( &mut self, entry_id: ProjectEntryId, cx: &Context, ) -> Option>> { match self { Worktree::Local(this) => this.expand_all_for_entry(entry_id, cx), Worktree::Remote(this) => { let response = this.client.request(proto::ExpandAllForProjectEntry { project_id: this.project_id, entry_id: entry_id.to_proto(), }); Some(cx.spawn(async move |this, cx| { let response = response.await?; this.update(cx, |this, _| { this.as_remote_mut() .unwrap() .wait_for_snapshot(response.worktree_scan_id as usize) })? .await?; Ok(()) })) } } } pub async fn handle_create_entry( this: Entity, request: proto::CreateProjectEntry, mut cx: AsyncApp, ) -> Result { let (scan_id, entry) = this.update(&mut cx, |this, cx| { anyhow::Ok(( this.scan_id(), this.create_entry( RelPath::from_unix_str(&request.path) .with_context(|| { format!("received invalid relative path {:?}", request.path) })? .into(), request.is_directory, request.content, cx, ), )) })?; Ok(proto::ProjectEntryResponse { entry: match &entry.await? { CreatedEntry::Included(entry) => Some(entry.into()), CreatedEntry::Excluded { .. } => None, }, worktree_scan_id: scan_id as u64, }) } pub async fn handle_trash_entry( this: Entity, request: proto::TrashProjectEntry, mut cx: AsyncApp, ) -> Result { let (scan_id, task) = this.update(&mut cx, |this, cx| { ( this.scan_id(), this.trash_entry(ProjectEntryId::from_proto(request.entry_id), cx), ) }); let trash_id = task .ok_or_else(|| anyhow::anyhow!("invalid entry"))? .await?; Ok(proto::TrashProjectEntryResponse { trash_id: trash_id.to_proto(), worktree_scan_id: scan_id as u64, }) } pub async fn handle_delete_entry( this: Entity, request: proto::DeleteProjectEntry, mut cx: AsyncApp, ) -> Result { let (scan_id, task) = this.update(&mut cx, |this, cx| { // While the `use_trash` field is deprecated but not removed, we // still need to support either trashing or deleting the file. // Otherwise, if an older client sends the `DeleteProjectEntry { // use_trash: true }` rather than the newer `TrashProjectEntry`, and // the flag was ignored, we'd permanently delete a file that was // actually meant to be trashed. #[allow(deprecated)] let task = if request.use_trash { this.trash_entry(ProjectEntryId::from_proto(request.entry_id), cx) .map(|task| cx.background_spawn(async move { task.await.map(|_| ()) })) } else { this.delete_entry(ProjectEntryId::from_proto(request.entry_id), cx) }; (this.scan_id(), task) }); task.ok_or_else(|| anyhow::anyhow!("invalid entry"))? .await?; Ok(proto::ProjectEntryResponse { entry: None, worktree_scan_id: scan_id as u64, }) } pub async fn handle_restore_entry( this: Entity, request: proto::RestoreProjectEntry, mut cx: AsyncApp, ) -> Result { let (scan_id, task) = this.update(&mut cx, |this, cx| { ( this.scan_id(), this.restore_entry(TrashId::from_proto(request.trash_id), cx), ) }); let entry = task.await?; Ok(proto::RestoreProjectEntryResponse { entry: Some(proto::Entry::from(&entry)), worktree_scan_id: scan_id as u64, }) } pub async fn handle_expand_entry( this: Entity, request: proto::ExpandProjectEntry, mut cx: AsyncApp, ) -> Result { let task = this.update(&mut cx, |this, cx| { this.expand_entry(ProjectEntryId::from_proto(request.entry_id), cx) }); task.ok_or_else(|| anyhow::anyhow!("no such entry"))? .await?; let scan_id = this.read_with(&cx, |this, _| this.scan_id()); Ok(proto::ExpandProjectEntryResponse { worktree_scan_id: scan_id as u64, }) } pub async fn handle_expand_all_for_entry( this: Entity, request: proto::ExpandAllForProjectEntry, mut cx: AsyncApp, ) -> Result { let task = this.update(&mut cx, |this, cx| { this.expand_all_for_entry(ProjectEntryId::from_proto(request.entry_id), cx) }); task.ok_or_else(|| anyhow::anyhow!("no such entry"))? .await?; let scan_id = this.read_with(&cx, |this, _| this.scan_id()); Ok(proto::ExpandAllForProjectEntryResponse { worktree_scan_id: scan_id as u64, }) } pub fn is_single_file(&self) -> bool { self.root_dir().is_none() } /// For visible worktrees, returns the path with the worktree name as the first component. /// Otherwise, returns an absolute path. pub fn full_path(&self, worktree_relative_path: &RelPath) -> PathBuf { if self.is_visible() { self.root_name() .join(worktree_relative_path) .display(self.path_style) .to_string() .into() } else { let full_path = self.abs_path(); let mut full_path_string = if self.is_local() && let Ok(stripped) = full_path.strip_prefix(home_dir()) { self.path_style .join("~", &*stripped.to_string_lossy()) .unwrap() } else { full_path.to_string_lossy().into_owned() }; if worktree_relative_path.components().next().is_some() { full_path_string.push_str(self.path_style.primary_separator()); full_path_string.push_str(&worktree_relative_path.display(self.path_style)); } full_path_string.into() } } } impl LocalWorktree { pub fn fs(&self) -> &Arc { &self.fs } pub fn is_path_private(&self, path: &RelPath) -> bool { !self.share_private_files && self.settings.is_path_private(path) } pub fn fs_is_case_sensitive(&self) -> bool { self.fs_case_sensitive } fn restart_background_scanners(&mut self, cx: &Context) { let (scan_requests_tx, scan_requests_rx) = async_channel::unbounded(); let (path_prefixes_to_scan_tx, path_prefixes_to_scan_rx) = async_channel::unbounded(); self.scan_requests_tx = scan_requests_tx; self.path_prefixes_to_scan_tx = path_prefixes_to_scan_tx; self.start_background_scanner(scan_requests_rx, path_prefixes_to_scan_rx, cx); let always_included_entries = self .snapshot .entries(true, 0) .filter(|entry| entry.is_always_included) .map(|entry| entry.path.clone()) .collect::>(); log::debug!( "refreshing entries for the following always included paths: {:?}", always_included_entries ); // Cleans up old always included entries to ensure they get updated properly. Otherwise, // nested always included entries may not get updated and will result in out-of-date info. self.refresh_entries_for_paths(always_included_entries); } fn start_background_scanner( &mut self, scan_requests_rx: async_channel::Receiver, path_prefixes_to_scan_rx: async_channel::Receiver, cx: &Context, ) { let snapshot = self.snapshot(); let share_private_files = self.share_private_files; let next_entry_id = self.next_entry_id.clone(); let fs = self.fs.clone(); let scanning_enabled = self.scanning_enabled; let force_defer_watch = self.force_defer_watch; let track_git_repositories = self.visible; let settings = self.settings.clone(); let (scan_states_tx, mut scan_states_rx) = mpsc::unbounded(); let background_scanner = cx.background_spawn({ let abs_path = snapshot.abs_path.as_path().to_path_buf(); let background = cx.background_executor().clone(); async move { let defer_watch = force_defer_watch || (scanning_enabled && fs.requires_poll_watcher(&abs_path)); let (events, watcher) = if scanning_enabled && !defer_watch { fs.watch(&abs_path, FS_WATCH_LATENCY).await } else { (Box::pin(stream::pending()) as _, Arc::new(NullWatcher) as _) }; let fs_case_sensitive = fs.is_case_sensitive().await; let is_single_file = snapshot.snapshot.root_dir().is_none(); let mut scanner = BackgroundScanner { fs, fs_case_sensitive, status_updates_tx: scan_states_tx, executor: background, scan_requests_rx, path_prefixes_to_scan_rx, next_entry_id, state: async_lock::Mutex::new(BackgroundScannerState { prev_snapshot: snapshot.snapshot.clone(), snapshot, symlink_paths_by_target: Default::default(), scanned_dirs: Default::default(), watched_dir_abs_paths_by_entry_id: Default::default(), scanning_enabled, path_prefixes_to_scan: Default::default(), paths_to_scan: Default::default(), removed_entries: RemovedEntries::default(), changed_paths: Default::default(), }), phase: BackgroundScannerPhase::InitialScan, share_private_files, settings, watcher, track_git_repositories, is_single_file, defer_watch, }; scanner.run(events).await; } }); let scan_state_updater = cx.spawn(async move |this, cx| { while let Some((state, this)) = scan_states_rx.next().await.zip(this.upgrade()) { this.update(cx, |this, cx| { let this = this.as_local_mut().unwrap(); match state { ScanState::Started => { *this.is_scanning.0.borrow_mut() = true; } ScanState::Updated { snapshot, changes, barrier, scanning, } => { *this.is_scanning.0.borrow_mut() = scanning; this.set_snapshot(snapshot, changes, cx); drop(barrier); } ScanState::RootUpdated { new_path } => { this.update_abs_path_and_refresh(new_path, cx); } ScanState::RootDeleted => { log::info!( "worktree root {} no longer exists, closing worktree", this.abs_path().display() ); cx.emit(Event::Deleted); } } }); } }); self._background_scanner_tasks = vec![background_scanner, scan_state_updater]; *self.is_scanning.0.borrow_mut() = true; } fn set_snapshot( &mut self, mut new_snapshot: LocalSnapshot, entry_changes: UpdatedEntriesSet, cx: &mut Context, ) { let repo_changes = self.changed_repos(&self.snapshot, &mut new_snapshot); if let Some((common_dir, is_linked_worktree)) = new_snapshot .local_repo_for_work_directory_path(RelPath::empty()) .map(|repo| { ( SanitizedPath::from_arc(repo.common_dir_abs_path.clone()), repo.repository_dir_abs_path != repo.common_dir_abs_path, ) }) { new_snapshot.root_repo_common_dir = Some(common_dir); new_snapshot.root_repo_is_linked_worktree = is_linked_worktree; } else { new_snapshot.root_repo_common_dir = None; new_snapshot.root_repo_is_linked_worktree = false; } let root_repo_metadata_changed = self.snapshot.root_repo_common_dir != new_snapshot.root_repo_common_dir || self.snapshot.root_repo_is_linked_worktree != new_snapshot.root_repo_is_linked_worktree; let old_root_repo_common_dir = root_repo_metadata_changed.then(|| self.snapshot.root_repo_common_dir.clone()); self.snapshot = new_snapshot; if let Some(share) = self.update_observer.as_mut() { share .snapshots_tx .unbounded_send((self.snapshot.clone(), entry_changes.clone())) .ok(); } if !entry_changes.is_empty() { cx.emit(Event::UpdatedEntries(entry_changes)); } if !repo_changes.is_empty() { cx.emit(Event::UpdatedGitRepositories(repo_changes)); } if let Some(old) = old_root_repo_common_dir { cx.emit(Event::UpdatedRootRepoCommonDir { old }); } while let Some((scan_id, _)) = self.snapshot_subscriptions.front() { if self.snapshot.completed_scan_id >= *scan_id { let (_, tx) = self.snapshot_subscriptions.pop_front().unwrap(); tx.send(()).ok(); } else { break; } } } fn changed_repos( &self, old_snapshot: &LocalSnapshot, new_snapshot: &mut LocalSnapshot, ) -> UpdatedGitRepositoriesSet { let mut changes = Vec::new(); let mut old_repos = old_snapshot.git_repositories.iter().peekable(); let new_repos = new_snapshot.git_repositories.clone(); let mut new_repos = new_repos.iter().peekable(); loop { match (new_repos.peek().map(clone), old_repos.peek().map(clone)) { (Some((new_entry_id, new_repo)), Some((old_entry_id, old_repo))) => { match Ord::cmp(&new_entry_id, &old_entry_id) { Ordering::Less => { changes.push(UpdatedGitRepository { work_directory_id: new_entry_id, old_work_directory_abs_path: None, new_work_directory_abs_path: Some( new_repo.work_directory_abs_path.clone(), ), dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()), repository_dir_abs_path: Some( new_repo.repository_dir_abs_path.clone(), ), common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()), }); new_repos.next(); } Ordering::Equal => { // A change to a repository's git state is signaled by bumping // `git_dir_scan_id`, and the diff below detects it via `!=`. If the // value ever regresses (e.g. a rescan re-inserting the repository // with a fresh scan id of 0), a bump from the same cycle is wiped // out and the corresponding git update is silently lost. debug_assert!( new_repo.git_dir_scan_id >= old_repo.git_dir_scan_id, "git_dir_scan_id for repository at {:?} regressed from {} to {}", new_repo.work_directory_abs_path, old_repo.git_dir_scan_id, new_repo.git_dir_scan_id, ); if new_repo.git_dir_scan_id != old_repo.git_dir_scan_id || new_repo.work_directory_abs_path != old_repo.work_directory_abs_path { changes.push(UpdatedGitRepository { work_directory_id: new_entry_id, old_work_directory_abs_path: Some( old_repo.work_directory_abs_path.clone(), ), new_work_directory_abs_path: Some( new_repo.work_directory_abs_path.clone(), ), dot_git_abs_path: Some(new_repo.dot_git_abs_path.clone()), repository_dir_abs_path: Some( new_repo.repository_dir_abs_path.clone(), ), common_dir_abs_path: Some(new_repo.common_dir_abs_path.clone()), }); } new_repos.next(); old_repos.next(); } Ordering::Greater => { changes.push(UpdatedGitRepository { work_directory_id: old_entry_id, old_work_directory_abs_path: Some( old_repo.work_directory_abs_path.clone(), ), new_work_directory_abs_path: None, dot_git_abs_path: None, repository_dir_abs_path: None, common_dir_abs_path: None, }); old_repos.next(); } } } (Some((entry_id, repo)), None) => { changes.push(UpdatedGitRepository { work_directory_id: entry_id, old_work_directory_abs_path: None, new_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()), dot_git_abs_path: Some(repo.dot_git_abs_path.clone()), repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()), common_dir_abs_path: Some(repo.common_dir_abs_path.clone()), }); new_repos.next(); } (None, Some((entry_id, repo))) => { changes.push(UpdatedGitRepository { work_directory_id: entry_id, old_work_directory_abs_path: Some(repo.work_directory_abs_path.clone()), new_work_directory_abs_path: None, dot_git_abs_path: Some(repo.dot_git_abs_path.clone()), repository_dir_abs_path: Some(repo.repository_dir_abs_path.clone()), common_dir_abs_path: Some(repo.common_dir_abs_path.clone()), }); old_repos.next(); } (None, None) => break, } } fn clone(value: &(&T, &U)) -> (T, U) { (value.0.clone(), value.1.clone()) } changes.into() } pub fn scan_complete(&self) -> impl Future + use<> { let mut is_scanning_rx = self.is_scanning.1.clone(); async move { let mut is_scanning = *is_scanning_rx.borrow(); while is_scanning { if let Some(value) = is_scanning_rx.recv().await { is_scanning = value; } else { break; } } } } pub fn wait_for_snapshot( &mut self, scan_id: usize, ) -> impl Future> + use<> { let (tx, rx) = oneshot::channel(); if self.snapshot.completed_scan_id >= scan_id { tx.send(()).ok(); } else { match self .snapshot_subscriptions .binary_search_by_key(&scan_id, |probe| probe.0) { Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)), } } async move { rx.await?; Ok(()) } } pub fn snapshot(&self) -> LocalSnapshot { self.snapshot.clone() } pub fn settings(&self) -> WorktreeSettings { self.settings.clone() } fn load_binary_file( &self, path: &RelPath, cx: &Context, ) -> Task> { let path = Arc::from(path); let abs_path = self.absolutize(&path); let fs = self.fs.clone(); let entry = self.refresh_entry(path.clone(), None, cx); let is_private = self.is_path_private(&path); let worktree = cx.weak_entity(); cx.background_spawn(async move { let content = fs.load_bytes(&abs_path).await?; let worktree = worktree.upgrade().context("worktree was dropped")?; let file = match entry.await? { Some(entry) => File::for_entry(entry, worktree), None => { let metadata = fs .metadata(&abs_path) .await .with_context(|| { format!("Loading metadata for excluded file {abs_path:?}") })? .with_context(|| { format!("Excluded file {abs_path:?} got removed during loading") })?; Arc::new(File { entry_id: None, worktree, path, disk_state: DiskState::Present { mtime: metadata.mtime, size: metadata.len, }, is_local: true, is_private, }) } }; Ok(LoadedBinaryFile { file, content }) }) } #[ztracing::instrument(skip_all)] fn load_file(&self, path: &RelPath, cx: &Context) -> Task> { let path = Arc::from(path); let abs_path = self.absolutize(&path); let fs = self.fs.clone(); let entry = self.refresh_entry(path.clone(), None, cx); let is_private = self.is_path_private(path.as_ref()); let this = cx.weak_entity(); cx.background_spawn(async move { // WARN: Temporary workaround for #27283. // We are not efficient with our memory usage per file, and use in excess of 64GB for a 10GB file // Therefore, as a temporary workaround to prevent system freezes, we just bail before opening a file // if it is too large // 5GB seems to be more reasonable, peaking at ~16GB, while 6GB jumps up to >24GB which seems like a // reasonable limit const FILE_SIZE_MAX: u64 = 6 * 1024 * 1024 * 1024; // 6GB let metadata = fs.metadata(&abs_path).await?; if let Some(metadata) = metadata.as_ref() && metadata.len >= FILE_SIZE_MAX { anyhow::bail!("File is too large to load"); } let (text, line_ending, encoding, has_bom) = decode_file_text_to_rope(fs.as_ref(), &abs_path).await?; let is_writable = metadata.is_some_and(|metadata| metadata.is_writable); let worktree = this.upgrade().context("worktree was dropped")?; let file = match entry.await? { Some(entry) => File::for_entry(entry, worktree), None => { let metadata = fs .metadata(&abs_path) .await .with_context(|| { format!("Loading metadata for excluded file {abs_path:?}") })? .with_context(|| { format!("Excluded file {abs_path:?} got removed during loading") })?; Arc::new(File { entry_id: None, worktree, path, disk_state: DiskState::Present { mtime: metadata.mtime, size: metadata.len, }, is_local: true, is_private, }) } }; Ok(LoadedFile { file, text, line_ending, encoding, has_bom, is_writable, }) }) } /// Find the lowest path in the worktree's datastructures that is an ancestor fn lowest_ancestor(&self, path: &RelPath) -> Arc { let mut lowest_ancestor = None; for path in path.ancestors() { if self.entry_for_path(path).is_some() { lowest_ancestor = Some(path.into()); break; } } lowest_ancestor.unwrap_or_else(|| RelPath::empty_arc()) } pub fn create_entry( &self, path: Arc, is_dir: bool, content: Option>, cx: &Context, ) -> Task> { let abs_path = self.absolutize(&path); let path_excluded = self.settings.is_path_excluded(&path); let fs = self.fs.clone(); let task_abs_path = abs_path.clone(); let write = cx.background_spawn(async move { if is_dir { fs.create_dir(&task_abs_path) .await .with_context(|| format!("creating directory {task_abs_path:?}")) } else { fs.write(&task_abs_path, content.as_deref().unwrap_or(&[])) .await .with_context(|| format!("creating file {task_abs_path:?}")) } }); let lowest_ancestor = self.lowest_ancestor(&path); cx.spawn(async move |this, cx| { write.await?; if path_excluded { return Ok(CreatedEntry::Excluded { abs_path }); } let (result, refreshes) = this.update(cx, |this, cx| { let mut refreshes = Vec::new(); let refresh_paths = path.strip_prefix(&lowest_ancestor).unwrap(); for refresh_path in refresh_paths.ancestors() { if refresh_path == RelPath::empty() { continue; } let refresh_full_path = lowest_ancestor.join(refresh_path); refreshes.push(this.as_local_mut().unwrap().refresh_entry( refresh_full_path.into(), None, cx, )); } ( this.as_local_mut().unwrap().refresh_entry(path, None, cx), refreshes, ) })?; for refresh in refreshes { refresh.await.log_err(); } Ok(result .await? .map(CreatedEntry::Included) .unwrap_or_else(|| CreatedEntry::Excluded { abs_path })) }) } pub fn write_file( &self, path: Arc, text: Rope, line_ending: LineEnding, encoding: &'static Encoding, has_bom: bool, cx: &Context, ) -> Task>> { let fs = self.fs.clone(); let is_private = self.is_path_private(&path); let abs_path = self.absolutize(&path); let write = cx.background_spawn({ let fs = fs.clone(); let abs_path = abs_path.clone(); async move { // For UTF-8, use the optimized `fs.save` which writes Rope chunks directly to disk // without allocating a contiguous string. if encoding == encoding_rs::UTF_8 && !has_bom { return fs.save(&abs_path, &text, line_ending).await; } // For legacy encodings (e.g. Shift-JIS), we fall back to converting the entire Rope // to a String/Bytes in memory before writing. // // Note: This is inefficient for very large files compared to the streaming approach above, // but supporting streaming writes for arbitrary encodings would require a significant // refactor of the `fs` crate to expose a Writer interface. let text_string = text.to_string(); let normalized_text = match line_ending { LineEnding::Unix => text_string, LineEnding::Windows => text_string.replace('\n', "\r\n"), }; let bytes = encode_text(normalized_text, encoding, has_bom); fs.write(&abs_path, &bytes).await } }); cx.spawn(async move |this, cx| { write.await?; let entry = this .update(cx, |this, cx| { this.as_local_mut() .unwrap() .refresh_entry(path.clone(), None, cx) })? .await?; let worktree = this.upgrade().context("worktree dropped")?; if let Some(entry) = entry { Ok(File::for_entry(entry, worktree)) } else { let metadata = fs .metadata(&abs_path) .await .with_context(|| { format!("Fetching metadata after saving the excluded buffer {abs_path:?}") })? .with_context(|| { format!("Excluded buffer {path:?} got removed during saving") })?; Ok(Arc::new(File { worktree, path, disk_state: DiskState::Present { mtime: metadata.mtime, size: metadata.len, }, entry_id: None, is_local: true, is_private, })) } }) } pub fn trash_entry(&self, entry: Entry, cx: &Context) -> Task> { let abs_path = self.absolutize(&entry.path); let fs = self.fs.clone(); cx.spawn(async move |this, cx| { let trash_id = if entry.is_file() { fs.trash(&abs_path, Default::default()).await? } else { fs.trash( &abs_path, RemoveOptions { recursive: true, ignore_if_not_exists: false, }, ) .await? }; this.update(cx, |this, _| { this.as_local_mut() .unwrap() .refresh_entries_for_paths(vec![entry.path]) })? .recv() .await; Ok(trash_id) }) } pub fn delete_entry(&self, entry: Entry, cx: &Context) -> Task> { let abs_path = self.absolutize(&entry.path); let fs = self.fs.clone(); cx.spawn(async move |this, cx| { if entry.is_file() { fs.remove_file(&abs_path, Default::default()).await? } else { fs.remove_dir( &abs_path, RemoveOptions { recursive: true, ignore_if_not_exists: false, }, ) .await? }; this.update(cx, |this, _| { this.as_local_mut() .unwrap() .refresh_entries_for_paths(vec![entry.path]) })? .recv() .await; Ok(()) }) } pub fn restore_entry( &mut self, trash_id: TrashId, cx: &mut Context<'_, Worktree>, ) -> Task> { let fs = self.fs.clone(); let worktree_abs_path = self.abs_path().clone(); let path_style = self.path_style(); cx.spawn(async move |this, cx| { let path_buf = fs.restore(trash_id).await?; let path = path_buf .strip_prefix(worktree_abs_path) .context("Could not strip prefix")?; let path = Arc::from(RelPath::new(&path, path_style)?.as_ref()); let entry = this .update(cx, |this, cx| { this.as_local_mut().unwrap().refresh_entry(path, None, cx) })? .await? .context("Entry not found after restore")?; Ok(entry) }) } pub fn copy_external_entries( &self, target_directory: Arc, paths: Vec>, cx: &Context, ) -> Task>> { let target_directory = self.absolutize(&target_directory); let worktree_path = self.abs_path().clone(); let fs = self.fs.clone(); let paths = paths .into_iter() .filter_map(|source| { let file_name = source.file_name()?; let mut target = target_directory.clone(); target.push(file_name); // Do not allow copying the same file to itself. if source.as_ref() != target.as_path() { Some((source, target)) } else { None } }) .collect::>(); let paths_to_refresh = paths .iter() .filter_map(|(_, target)| { RelPath::new( target.strip_prefix(&worktree_path).ok()?, PathStyle::local(), ) .ok() .map(|path| path.into_arc()) }) .collect::>(); cx.spawn(async move |this, cx| { cx.background_spawn(async move { for (source, target) in paths { copy_recursive( fs.as_ref(), &source, &target, fs::CopyOptions { overwrite: true, ..Default::default() }, ) .await .with_context(|| { format!("Failed to copy file from {source:?} to {target:?}") })?; } anyhow::Ok(()) }) .await .log_err(); let mut refresh = cx.read_entity( &this.upgrade().with_context(|| "Dropped worktree")?, |this, _| { anyhow::Ok::( this.as_local() .with_context(|| "Worktree is not local")? .refresh_entries_for_paths(paths_to_refresh.clone()), ) }, )?; cx.background_spawn(async move { refresh.next().await; anyhow::Ok(()) }) .await .log_err(); let this = this.upgrade().with_context(|| "Dropped worktree")?; Ok(cx.read_entity(&this, |this, _| { paths_to_refresh .iter() .filter_map(|path| Some(this.entry_for_path(path)?.id)) .collect() })) }) } fn expand_entry( &self, entry_id: ProjectEntryId, cx: &Context, ) -> Option>> { let path = self.entry_for_id(entry_id)?.path.clone(); let mut refresh = self.refresh_entries_for_paths(vec![path]); Some(cx.background_spawn(async move { refresh.next().await; Ok(()) })) } fn expand_all_for_entry( &self, entry_id: ProjectEntryId, cx: &Context, ) -> Option>> { let path = self.entry_for_id(entry_id).unwrap().path.clone(); let mut rx = self.add_path_prefix_to_scan(path); Some(cx.background_spawn(async move { rx.next().await; Ok(()) })) } pub fn refresh_entries_for_paths(&self, paths: Vec>) -> barrier::Receiver { let (tx, rx) = barrier::channel(); self.scan_requests_tx .try_send(ScanRequest { relative_paths: paths, done: smallvec![tx], }) .ok(); rx } #[cfg(feature = "test-support")] pub fn manually_refresh_entries_for_paths( &self, paths: Vec>, ) -> barrier::Receiver { self.refresh_entries_for_paths(paths) } pub fn add_path_prefix_to_scan(&self, path_prefix: Arc) -> barrier::Receiver { let (tx, rx) = barrier::channel(); self.path_prefixes_to_scan_tx .try_send(PathPrefixScanRequest { path: path_prefix, done: smallvec![tx], }) .ok(); rx } pub fn refresh_entry( &self, path: Arc, old_path: Option>, cx: &Context, ) -> Task>> { if self.settings.is_path_excluded(&path) { return Task::ready(Ok(None)); } let paths = if let Some(old_path) = old_path.as_ref() { vec![old_path.clone(), path.clone()] } else { vec![path.clone()] }; let t0 = Instant::now(); let mut refresh = self.refresh_entries_for_paths(paths); // todo(lw): Hot foreground spawn cx.spawn(async move |this, cx| { refresh.recv().await; log::trace!("refreshed entry {path:?} in {:?}", t0.elapsed()); let new_entry = this.read_with(cx, |this, _| { this.entry_for_path(&path).cloned().with_context(|| { format!("Could not find entry in worktree for {path:?} after refresh") }) })??; Ok(Some(new_entry)) }) } pub fn observe_updates(&mut self, project_id: u64, cx: &Context, callback: F) where F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut, Fut: 'static + Send + Future, { if let Some(observer) = self.update_observer.as_mut() { *observer.resume_updates.borrow_mut() = (); return; } let (resume_updates_tx, mut resume_updates_rx) = watch::channel::<()>(); let (snapshots_tx, mut snapshots_rx) = mpsc::unbounded::<(LocalSnapshot, UpdatedEntriesSet)>(); snapshots_tx .unbounded_send((self.snapshot(), Arc::default())) .ok(); let worktree_id = self.id.to_proto(); let _maintain_remote_snapshot = cx.background_spawn(async move { let mut is_first = true; while let Some((snapshot, entry_changes)) = snapshots_rx.next().await { let update = if is_first { is_first = false; snapshot.build_initial_update(project_id, worktree_id) } else { snapshot.build_update(project_id, worktree_id, entry_changes) }; for update in proto::split_worktree_update(update) { let _ = resume_updates_rx.try_recv(); loop { let result = callback(update.clone()); if result.await { break; } else { log::info!("waiting to resume updates"); if resume_updates_rx.next().await.is_none() { return Some(()); } } } } } Some(()) }); self.update_observer = Some(UpdateObservationState { snapshots_tx, resume_updates: resume_updates_tx, _maintain_remote_snapshot, }); } pub fn share_private_files(&mut self, cx: &Context) { self.share_private_files = true; self.restart_background_scanners(cx); } pub fn update_abs_path_and_refresh( &mut self, new_path: Arc, cx: &Context, ) { self.snapshot.git_repositories = Default::default(); self.snapshot.ignores_by_parent_abs_path = Default::default(); let root_name = new_path .as_path() .file_name() .and_then(|f| f.to_str()) .map_or(RelPath::empty_arc(), |f| { RelPath::from_unix_str(f).unwrap().into() }); self.snapshot.update_abs_path(new_path, root_name); self.restart_background_scanners(cx); } #[cfg(feature = "test-support")] pub fn set_defer_watch(&mut self, defer: bool, cx: &mut Context) { self.force_defer_watch = defer; self.restart_background_scanners(cx); } #[cfg(feature = "test-support")] pub fn repositories(&self) -> Vec> { self.git_repositories .values() .map(|entry| entry.work_directory_abs_path.clone()) .collect::>() } } impl RemoteWorktree { pub fn project_id(&self) -> u64 { self.project_id } pub fn client(&self) -> AnyProtoClient { self.client.clone() } pub fn disconnected_from_host(&mut self) { self.updates_tx.take(); self.snapshot_subscriptions.clear(); self.disconnected = true; } pub fn update_from_remote(&self, update: proto::UpdateWorktree) { if let Some(updates_tx) = &self.updates_tx { updates_tx .unbounded_send(update) .expect("consumer runs to completion"); } } fn observe_updates(&mut self, project_id: u64, cx: &Context, callback: F) where F: 'static + Send + Fn(proto::UpdateWorktree) -> Fut, Fut: 'static + Send + Future, { let (tx, mut rx) = mpsc::unbounded(); let initial_update = self .snapshot .build_initial_update(project_id, self.id().to_proto()); self.update_observer = Some(tx); cx.spawn(async move |this, cx| { let mut update = initial_update; 'outer: loop { // SSH projects use a special project ID of 0, and we need to // remap it to the correct one here. update.project_id = project_id; for chunk in split_worktree_update(update) { if !callback(chunk).await { break 'outer; } } if let Some(next_update) = rx.next().await { update = next_update; } else { break; } } this.update(cx, |this, _| { let this = this.as_remote_mut().unwrap(); this.update_observer.take(); }) }) .detach(); } fn observed_snapshot(&self, scan_id: usize) -> bool { self.completed_scan_id >= scan_id } pub fn wait_for_snapshot( &mut self, scan_id: usize, ) -> impl Future> + use<> { let (tx, rx) = oneshot::channel(); if self.observed_snapshot(scan_id) { let _ = tx.send(()); } else if self.disconnected { drop(tx); } else { match self .snapshot_subscriptions .binary_search_by_key(&scan_id, |probe| probe.0) { Ok(ix) | Err(ix) => self.snapshot_subscriptions.insert(ix, (scan_id, tx)), } } async move { rx.await?; Ok(()) } } pub fn insert_entry( &mut self, entry: proto::Entry, scan_id: usize, cx: &Context, ) -> Task> { let wait_for_snapshot = self.wait_for_snapshot(scan_id); cx.spawn(async move |this, cx| { wait_for_snapshot.await?; this.update(cx, |worktree, _| { let worktree = worktree.as_remote_mut().unwrap(); let snapshot = &mut worktree.background_snapshot.lock().0; let entry = snapshot.insert_entry(entry, &worktree.file_scan_inclusions); worktree.snapshot = snapshot.clone(); entry })? }) } fn trash_entry( &self, entry_id: ProjectEntryId, cx: &Context, ) -> Task> { let response = self.client.request(proto::TrashProjectEntry { project_id: self.project_id, entry_id: entry_id.to_proto(), }); cx.spawn(async move |this, cx| { let response = response.await?; let scan_id = response.worktree_scan_id as usize; let trash_id = response.trash_id; this.update(cx, move |this, _| { this.as_remote_mut().unwrap().wait_for_snapshot(scan_id) })? .await?; this.update(cx, |this, _| { let this = this.as_remote_mut().unwrap(); let snapshot = &mut this.background_snapshot.lock().0; snapshot.delete_entry(entry_id); this.snapshot = snapshot.clone(); })?; Ok(TrashId::from_proto(trash_id)) }) } fn delete_entry(&self, entry_id: ProjectEntryId, cx: &Context) -> Task> { let response = self.client.request(proto::DeleteProjectEntry { project_id: self.project_id, entry_id: entry_id.to_proto(), // The `use_trash` field is being deprecated but it's still required // in the message, hence the `#[allow(deprecated)]` attribute. #[allow(deprecated)] use_trash: false, }); cx.spawn(async move |this, cx| { let response = response.await?; let scan_id = response.worktree_scan_id as usize; this.update(cx, move |this, _| { this.as_remote_mut().unwrap().wait_for_snapshot(scan_id) })? .await?; this.update(cx, |this, _| { let this = this.as_remote_mut().unwrap(); let snapshot = &mut this.background_snapshot.lock().0; snapshot.delete_entry(entry_id); this.snapshot = snapshot.clone(); }) }) } fn restore_entry(&mut self, trash_id: TrashId, cx: &Context) -> Task> { let project_id = self.project_id(); let worktree_id = self.id().to_proto(); let trash_id = trash_id.to_proto(); let request = self.client.request(proto::RestoreProjectEntry { project_id, worktree_id, trash_id, }); cx.spawn(async move |this, cx| { let response = request.await?; let scan_id = response.worktree_scan_id as usize; let proto_entry = response.entry.context("Missing entry in in response")?; this.update(cx, move |worktree, cx| { worktree .as_remote_mut() .unwrap() .insert_entry(proto_entry, scan_id, cx) })? .await }) } fn copy_external_entries( &self, target_directory: Arc, paths_to_copy: Vec>, local_fs: Arc, cx: &Context, ) -> Task>> { let client = self.client.clone(); let worktree_id = self.id().to_proto(); let project_id = self.project_id; cx.background_spawn(async move { let mut requests = Vec::new(); for root_path_to_copy in paths_to_copy { let Some(filename) = root_path_to_copy .file_name() .and_then(|name| name.to_str()) .and_then(|filename| RelPath::from_unix_str(filename).ok()) else { continue; }; for (abs_path, is_directory) in read_dir_items(local_fs.as_ref(), &root_path_to_copy).await? { let Some(relative_path) = abs_path .strip_prefix(&root_path_to_copy) .map_err(|e| anyhow::Error::from(e)) .and_then(|relative_path| RelPath::new(relative_path, PathStyle::local())) .log_err() else { continue; }; let content = if is_directory { None } else { Some(local_fs.load_bytes(&abs_path).await?) }; let mut target_path = target_directory.join(filename); if relative_path.file_name().is_some() { target_path = target_path.join(&relative_path); } requests.push(proto::CreateProjectEntry { project_id, worktree_id, path: target_path.as_unix_str().to_owned(), is_directory, content, }); } } requests.sort_unstable_by(|a, b| a.path.cmp(&b.path)); requests.dedup(); let mut copied_entry_ids = Vec::new(); for request in requests { let response = client.request(request).await?; copied_entry_ids.extend(response.entry.map(|e| ProjectEntryId::from_proto(e.id))); } Ok(copied_entry_ids) }) } } impl Snapshot { pub fn new( id: WorktreeId, root_name: Arc, abs_path: Arc, path_style: PathStyle, ) -> Self { Snapshot { id, abs_path: SanitizedPath::from_arc(abs_path), path_style, root_char_bag: root_name .as_unix_str() .chars() .map(|c| c.to_ascii_lowercase()) .collect(), root_name, entries_by_path: Default::default(), entries_by_id: Default::default(), root_repo_common_dir: None, root_repo_is_linked_worktree: false, scan_id: 1, completed_scan_id: 0, } } pub fn id(&self) -> WorktreeId { self.id } // TODO: // Consider the following: // // ```rust // let abs_path: Arc = snapshot.abs_path(); // e.g. "C:\Users\user\Desktop\project" // let some_non_trimmed_path = Path::new("\\\\?\\C:\\Users\\user\\Desktop\\project\\main.rs"); // // The caller perform some actions here: // some_non_trimmed_path.strip_prefix(abs_path); // This fails // some_non_trimmed_path.starts_with(abs_path); // This fails too // ``` // // This is definitely a bug, but it's not clear if we should handle it here or not. pub fn abs_path(&self) -> &Arc { SanitizedPath::cast_arc_ref(&self.abs_path) } pub fn root_repo_common_dir(&self) -> Option<&Arc> { self.root_repo_common_dir .as_ref() .map(SanitizedPath::cast_arc_ref) } pub fn root_repo_is_linked_worktree(&self) -> bool { self.root_repo_is_linked_worktree } fn build_initial_update(&self, project_id: u64, worktree_id: u64) -> proto::UpdateWorktree { let mut updated_entries = self .entries_by_path .iter() .map(proto::Entry::from) .collect::>(); updated_entries.sort_unstable_by_key(|e| e.id); proto::UpdateWorktree { project_id, worktree_id, abs_path: self.abs_path().to_string_lossy().into_owned(), root_name: self.root_name().as_unix_str().to_owned(), root_repo_common_dir: self .root_repo_common_dir() .map(|p| p.to_string_lossy().into_owned()), root_repo_is_linked_worktree: self.root_repo_is_linked_worktree, updated_entries, removed_entries: Vec::new(), scan_id: self.scan_id as u64, is_last_update: self.completed_scan_id == self.scan_id, // Sent in separate messages. updated_repositories: Vec::new(), removed_repositories: Vec::new(), } } pub fn work_directory_abs_path(&self, work_directory: &WorkDirectory) -> PathBuf { match work_directory { WorkDirectory::InProject { relative_path } => self.absolutize(relative_path), WorkDirectory::AboveProject { absolute_path, .. } => absolute_path.as_ref().to_owned(), } } pub fn absolutize(&self, path: &RelPath) -> PathBuf { if path.file_name().is_some() { let mut abs_path = self.abs_path.to_string(); for component in path.components() { if !abs_path.ends_with(self.path_style.primary_separator()) { abs_path.push_str(self.path_style.primary_separator()); } abs_path.push_str(component); } PathBuf::from(abs_path) } else { self.abs_path.as_path().to_path_buf() } } pub fn contains_entry(&self, entry_id: ProjectEntryId) -> bool { self.entries_by_id.get(&entry_id, ()).is_some() } fn insert_entry( &mut self, entry: proto::Entry, always_included_paths: &PathMatcher, ) -> Result { let entry = Entry::try_from((&self.root_char_bag, always_included_paths, entry))?; let old_entry = self.entries_by_id.insert_or_replace( PathEntry { id: entry.id, path: entry.path.clone(), is_ignored: entry.is_ignored, scan_id: 0, }, (), ); if let Some(old_entry) = old_entry { self.entries_by_path.remove(&PathKey(old_entry.path), ()); } self.entries_by_path.insert_or_replace(entry.clone(), ()); Ok(entry) } fn delete_entry(&mut self, entry_id: ProjectEntryId) -> Option> { let removed_entry = self.entries_by_id.remove(&entry_id, ())?; self.entries_by_path = { let mut cursor = self.entries_by_path.cursor::(()); let mut new_entries_by_path = cursor.slice(&TraversalTarget::path(&removed_entry.path), Bias::Left); while let Some(entry) = cursor.item() { if entry.path.starts_with(&removed_entry.path) { self.entries_by_id.remove(&entry.id, ()); cursor.next(); } else { break; } } new_entries_by_path.append(cursor.suffix(), ()); new_entries_by_path }; Some(removed_entry.path) } fn update_abs_path(&mut self, abs_path: Arc, root_name: Arc) { self.abs_path = abs_path; if root_name != self.root_name { self.root_char_bag = root_name .as_unix_str() .chars() .map(|c| c.to_ascii_lowercase()) .collect(); self.root_name = root_name; } } pub fn apply_remote_update( &mut self, update: proto::UpdateWorktree, always_included_paths: &PathMatcher, ) { log::debug!( "applying remote worktree update. {} entries updated, {} removed", update.updated_entries.len(), update.removed_entries.len() ); if let Some(root_name) = RelPath::from_unix_str(&update.root_name).log_err() { self.update_abs_path( SanitizedPath::new_arc(&Path::new(&update.abs_path)), root_name.into(), ); } let mut entries_by_path_edits = Vec::new(); let mut entries_by_id_edits = Vec::new(); for entry_id in update.removed_entries { let entry_id = ProjectEntryId::from_proto(entry_id); entries_by_id_edits.push(Edit::Remove(entry_id)); if let Some(entry) = self.entry_for_id(entry_id) { entries_by_path_edits.push(Edit::Remove(PathKey(entry.path.clone()))); } } for entry in update.updated_entries { let Some(entry) = Entry::try_from((&self.root_char_bag, always_included_paths, entry)).log_err() else { continue; }; if let Some(PathEntry { path, .. }) = self.entries_by_id.get(&entry.id, ()) { entries_by_path_edits.push(Edit::Remove(PathKey(path.clone()))); } if let Some(old_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ()) && old_entry.id != entry.id { entries_by_id_edits.push(Edit::Remove(old_entry.id)); } entries_by_id_edits.push(Edit::Insert(PathEntry { id: entry.id, path: entry.path.clone(), is_ignored: entry.is_ignored, scan_id: 0, })); entries_by_path_edits.push(Edit::Insert(entry)); } self.entries_by_path.edit(entries_by_path_edits, ()); self.entries_by_id.edit(entries_by_id_edits, ()); // A `None` from a completed scan is a real repo removal, whereas a `None` // mid-scan may just mean the sender hasn't registered the root repo yet. match update .root_repo_common_dir .map(|p| SanitizedPath::new_arc(Path::new(&p))) { Some(dir) => { self.root_repo_common_dir = Some(dir); self.root_repo_is_linked_worktree = update.root_repo_is_linked_worktree; } None if update.is_last_update => { self.root_repo_common_dir = None; self.root_repo_is_linked_worktree = false; } None => {} } self.scan_id = update.scan_id as usize; if update.is_last_update { self.completed_scan_id = update.scan_id as usize; } } pub fn entry_count(&self) -> usize { self.entries_by_path.summary().count } pub fn visible_entry_count(&self) -> usize { self.entries_by_path.summary().non_ignored_count } pub fn dir_count(&self) -> usize { let summary = self.entries_by_path.summary(); summary.count - summary.file_count } pub fn visible_dir_count(&self) -> usize { let summary = self.entries_by_path.summary(); summary.non_ignored_count - summary.non_ignored_file_count } pub fn file_count(&self) -> usize { self.entries_by_path.summary().file_count } pub fn visible_file_count(&self) -> usize { self.entries_by_path.summary().non_ignored_file_count } pub fn deferred_scan_dir_count(&self) -> usize { self.entries_by_path.summary().deferred_scan_dir_count } fn traverse_from_offset( &self, include_files: bool, include_dirs: bool, include_ignored: bool, start_offset: usize, ) -> Traversal<'_> { let mut cursor = self.entries_by_path.cursor(()); cursor.seek( &TraversalTarget::Count { count: start_offset, include_files, include_dirs, include_ignored, }, Bias::Right, ); Traversal { snapshot: self, cursor, include_files, include_dirs, include_ignored, } } pub fn traverse_from_path( &self, include_files: bool, include_dirs: bool, include_ignored: bool, path: &RelPath, ) -> Traversal<'_> { Traversal::new(self, include_files, include_dirs, include_ignored, path) } pub fn files(&self, include_ignored: bool, start: usize) -> Traversal<'_> { self.traverse_from_offset(true, false, include_ignored, start) } pub fn directories(&self, include_ignored: bool, start: usize) -> Traversal<'_> { self.traverse_from_offset(false, true, include_ignored, start) } pub fn entries(&self, include_ignored: bool, start: usize) -> Traversal<'_> { self.traverse_from_offset(true, true, include_ignored, start) } pub fn paths(&self) -> impl Iterator { self.entries_by_path .cursor::<()>(()) .filter(move |entry| !entry.path.is_empty()) .map(|entry| entry.path.as_ref()) } pub fn child_entries<'a>(&'a self, parent_path: &'a RelPath) -> ChildEntriesIter<'a> { let options = ChildEntriesOptions { include_files: true, include_dirs: true, include_ignored: true, }; self.child_entries_with_options(parent_path, options) } pub fn child_entries_with_options<'a>( &'a self, parent_path: &'a RelPath, options: ChildEntriesOptions, ) -> ChildEntriesIter<'a> { let mut cursor = self.entries_by_path.cursor(()); cursor.seek(&TraversalTarget::path(parent_path), Bias::Right); let traversal = Traversal { snapshot: self, cursor, include_files: options.include_files, include_dirs: options.include_dirs, include_ignored: options.include_ignored, }; ChildEntriesIter { traversal, parent_path, } } pub fn root_entry(&self) -> Option<&Entry> { self.entries_by_path.first() } /// Returns `None` for a single file worktree, or `Some(self.abs_path())` if /// it is a directory. pub fn root_dir(&self) -> Option> { self.root_entry() .filter(|entry| entry.is_dir()) .map(|_| self.abs_path().clone()) } pub fn root_name(&self) -> &RelPath { &self.root_name } pub fn root_name_str(&self) -> &str { self.root_name.as_unix_str() } pub fn scan_id(&self) -> usize { self.scan_id } pub fn entry_for_path(&self, path: &RelPath) -> Option<&Entry> { let entry = self.traverse_from_path(true, true, true, path).entry(); entry.filter(|&entry| entry.path.as_ref() == path) } /// Whether `path` is gitignored, or lies inside a gitignored directory. /// /// The contents of ignored directories aren't scanned until explicitly /// expanded, so when `path` has no entry this falls back to the ignore /// status of its nearest scanned ancestor. pub fn is_path_ignored(&self, path: &RelPath) -> bool { path.ancestors() .find_map(|ancestor| self.entry_for_path(ancestor)) .is_some_and(|entry| entry.is_ignored) } /// Resolves a path to an executable using the following heuristics: /// /// 1. If the path starts with `~`, it is expanded to the user's home directory. /// 2. If the path is relative and contains more than one component, /// it is joined to the worktree root path. /// 3. If the path is relative and exists in the worktree /// (even if falls under an exclusion filter), /// it is joined to the worktree root path. /// 4. Otherwise the path is returned unmodified. /// /// Relative paths that do not exist in the worktree may /// still be found using the `PATH` environment variable. pub fn resolve_relative_path(&self, path: PathBuf) -> PathBuf { if let Some(path_str) = path.to_str() { if let Some(remaining_path) = path_str.strip_prefix("~/") { return home_dir().join(remaining_path); } else if path_str == "~" { return home_dir().to_path_buf(); } } if let Ok(rel_path) = RelPath::new(&path, self.path_style) && (path.components().count() > 1 || self.entry_for_path(&rel_path).is_some()) { self.abs_path().join(path) } else { path } } pub fn entry_for_id(&self, id: ProjectEntryId) -> Option<&Entry> { let entry = self.entries_by_id.get(&id, ())?; self.entry_for_path(&entry.path) } pub fn path_style(&self) -> PathStyle { self.path_style } } impl LocalSnapshot { fn local_repo_for_work_directory_path(&self, path: &RelPath) -> Option<&LocalRepositoryEntry> { self.git_repositories .iter() .map(|(_, entry)| entry) .find(|entry| entry.work_directory.path_key() == PathKey(path.into())) } fn build_update( &self, project_id: u64, worktree_id: u64, entry_changes: UpdatedEntriesSet, ) -> proto::UpdateWorktree { let mut updated_entries = Vec::new(); let mut removed_entries = Vec::new(); for (_, entry_id, path_change) in entry_changes.iter() { if let PathChange::Removed = path_change { removed_entries.push(entry_id.0 as u64); } else if let Some(entry) = self.entry_for_id(*entry_id) { updated_entries.push(proto::Entry::from(entry)); } } removed_entries.sort_unstable(); updated_entries.sort_unstable_by_key(|e| e.id); // TODO - optimize, knowing that removed_entries are sorted. removed_entries.retain(|id| updated_entries.binary_search_by_key(id, |e| e.id).is_err()); proto::UpdateWorktree { project_id, worktree_id, abs_path: self.abs_path().to_string_lossy().into_owned(), root_name: self.root_name().as_unix_str().to_owned(), root_repo_common_dir: self .root_repo_common_dir() .map(|p| p.to_string_lossy().into_owned()), root_repo_is_linked_worktree: self.root_repo_is_linked_worktree, updated_entries, removed_entries, scan_id: self.scan_id as u64, is_last_update: self.completed_scan_id == self.scan_id, // Sent in separate messages. updated_repositories: Vec::new(), removed_repositories: Vec::new(), } } async fn insert_entry(&mut self, mut entry: Entry, fs: &dyn Fs) -> Entry { log::trace!("insert entry {:?}", entry.path); if entry.is_file() && entry.path.file_name() == Some(&GITIGNORE) { let abs_path = self.absolutize(&entry.path); match build_gitignore(&abs_path, fs).await { Ok(ignore) => { self.ignores_by_parent_abs_path .insert(abs_path.parent().unwrap().into(), (Arc::new(ignore), true)); } Err(error) => { log::error!( "error loading .gitignore file {:?} - {:?}", entry.path, error ); } } } if entry.kind == EntryKind::PendingDir && let Some(existing_entry) = self.entries_by_path.get(&PathKey(entry.path.clone()), ()) { entry.kind = existing_entry.kind; } let scan_id = self.scan_id; let removed = self.entries_by_path.insert_or_replace(entry.clone(), ()); if let Some(removed) = removed && removed.id != entry.id { self.entries_by_id.remove(&removed.id, ()); } self.entries_by_id.insert_or_replace( PathEntry { id: entry.id, path: entry.path.clone(), is_ignored: entry.is_ignored, scan_id, }, (), ); entry } fn ancestor_inodes_for_path(&self, path: &RelPath) -> TreeSet { let mut inodes = TreeSet::default(); for ancestor in path.ancestors().skip(1) { if let Some(entry) = self.entry_for_path(ancestor) { inodes.insert(entry.inode); } } inodes } async fn ignore_stack_for_abs_path( &self, abs_path: &Path, is_dir: bool, fs: &dyn Fs, ) -> IgnoreStack { let mut new_ignores = Vec::new(); let mut repo_excludes = Vec::new(); let mut repo_root = None; for (index, ancestor) in abs_path.ancestors().enumerate() { if index > 0 { if let Some((ignore, _)) = self.ignores_by_parent_abs_path.get(ancestor) { new_ignores.push((ancestor, Some(ignore.clone()))); } else { new_ignores.push((ancestor, None)); } } // Collect the `info/exclude` rules of every containing repository, not just // the innermost one: a nested repository's files are still governed by the // exclude rules of the outer repository that contains it. if let Some((repo_exclude, _)) = self.repo_exclude_by_work_dir_abs_path.get(ancestor) { repo_excludes.push(repo_exclude.clone()); } let is_repo_root = fs .metadata(&ancestor.join(DOT_GIT)) .await .is_ok_and(|metadata| metadata.is_some()); if is_repo_root { if repo_root.is_none() { repo_root = Some(Arc::from(ancestor)); } // Stop at the repository containing the worktree root, but not at // ones nested below it, where its rules still apply. if self.abs_path.as_path().starts_with(ancestor) { break; } } } let mut ignore_stack = if let Some(global_gitignore) = self.global_gitignore.clone() { IgnoreStack::global(global_gitignore) } else { IgnoreStack::none() }; for repo_exclude in repo_excludes.into_iter().rev() { ignore_stack = ignore_stack.append(IgnoreKind::RepoExclude, repo_exclude); } ignore_stack.global_ignore_root = Some(repo_root.clone().unwrap_or_else(|| self.abs_path().clone())); ignore_stack.repo_root = repo_root; let mut ancestor_ignore_stack = ignore_stack.clone(); for (parent_abs_path, ignore) in new_ignores.into_iter().rev() { if ancestor_ignore_stack.is_abs_path_ignored(parent_abs_path, true) { ignore_stack = IgnoreStack::all(); break; } else if let Some(ignore) = ignore { let kind = IgnoreKind::Gitignore(parent_abs_path.into()); ancestor_ignore_stack = ancestor_ignore_stack.append(kind, ignore.clone()); ignore_stack = ignore_stack.append(IgnoreKind::Gitignore(parent_abs_path.into()), ignore); } } if ignore_stack.is_abs_path_ignored(abs_path, is_dir) { ignore_stack = IgnoreStack::all(); } ignore_stack } #[cfg(feature = "test-support")] pub fn expanded_entries(&self, file_scan_depth: Option) -> impl Iterator { let file_scan_depth = if self.root_repo_common_dir.is_some() { None } else { file_scan_depth }; self.entries_by_path.cursor::<()>(()).filter(move |entry| { entry.kind == EntryKind::Dir && (entry.is_external || entry.is_ignored || (!entry.is_always_included && is_beyond_scan_depth(file_scan_depth, &entry.path) && !self .git_repositories .values() .any(|repo| repo.work_directory.directory_contains(&entry.path)))) }) } #[cfg(feature = "test-support")] pub fn check_invariants(&self, git_state: bool) { use pretty_assertions::assert_eq; assert_eq!( self.entries_by_path .cursor::<()>(()) .map(|e| (&e.path, e.id)) .collect::>(), self.entries_by_id .cursor::<()>(()) .map(|e| (&e.path, e.id)) .collect::>() .into_iter() .collect::>(), "entries_by_path and entries_by_id are inconsistent" ); let mut files = self.files(true, 0); let mut visible_files = self.files(false, 0); for entry in self.entries_by_path.cursor::<()>(()) { if entry.is_file() { assert_eq!(files.next().unwrap().inode, entry.inode); if !entry.is_ignored || entry.is_always_included { assert_eq!(visible_files.next().unwrap().inode, entry.inode); } } } assert!(files.next().is_none()); assert!(visible_files.next().is_none()); let mut bfs_paths = Vec::new(); let mut stack = self .root_entry() .map(|e| e.path.as_ref()) .into_iter() .collect::>(); while let Some(path) = stack.pop() { bfs_paths.push(path); let ix = stack.len(); for child_entry in self.child_entries(path) { stack.insert(ix, &child_entry.path); } } let dfs_paths_via_iter = self .entries_by_path .cursor::<()>(()) .map(|e| e.path.as_ref()) .collect::>(); assert_eq!(bfs_paths, dfs_paths_via_iter); let dfs_paths_via_traversal = self .entries(true, 0) .map(|e| e.path.as_ref()) .collect::>(); assert_eq!(dfs_paths_via_traversal, dfs_paths_via_iter); if git_state { for ignore_parent_abs_path in self.ignores_by_parent_abs_path.keys() { let ignore_parent_path = &RelPath::new( ignore_parent_abs_path .strip_prefix(self.abs_path.as_path()) .unwrap(), PathStyle::local(), ) .unwrap(); assert!(self.entry_for_path(ignore_parent_path).is_some()); assert!( self.entry_for_path( &ignore_parent_path.join(RelPath::from_unix_str(GITIGNORE).unwrap()) ) .is_some() ); } } } #[cfg(feature = "test-support")] pub fn entries_without_ids(&self, include_ignored: bool) -> Vec<(&RelPath, u64, bool)> { let mut paths = Vec::new(); for entry in self.entries_by_path.cursor::<()>(()) { if include_ignored || !entry.is_ignored { paths.push((entry.path.as_ref(), entry.inode, entry.is_ignored)); } } paths.sort_by(|a, b| a.0.cmp(b.0)); paths } } impl BackgroundScannerState { async fn enqueue_scan_dir( &self, abs_path: Arc, entry: &Entry, scan_job_tx: &Sender, fs: &dyn Fs, ) { let path = entry.path.clone(); let ignore_stack = self .snapshot .ignore_stack_for_abs_path(&abs_path, true, fs) .await; let mut ancestor_inodes = self.snapshot.ancestor_inodes_for_path(&path); if !ancestor_inodes.contains(&entry.inode) { ancestor_inodes.insert(entry.inode); scan_job_tx .try_send(ScanJob { abs_path, path, ignore_stack, scan_queue: scan_job_tx.clone(), ancestor_inodes, is_external: entry.is_external, }) .unwrap(); } } fn reuse_entry_id(&mut self, entry: &mut Entry) { let Some(mtime) = entry.mtime else { return; }; if let Some(entry_id) = self.reused_entry_id(&entry.path, entry.inode, mtime) { entry.id = entry_id; } } fn entry_id_for( &mut self, next_entry_id: &AtomicUsize, path: &RelPath, metadata: &fs::Metadata, ) -> ProjectEntryId { self.reused_entry_id(path, metadata.inode, metadata.mtime) .unwrap_or_else(|| ProjectEntryId::new(next_entry_id)) } fn reused_entry_id( &mut self, path: &RelPath, inode: u64, mtime: MTime, ) -> Option { if let Some(removed_entry) = self.removed_entries.take_by_path(path, inode) { return Some(removed_entry.id); } // If an entry with the same inode was removed from the worktree during this scan, // then it *might* represent the same file or directory. But the OS might also have // re-used the inode for a completely different file or directory. // // Conditionally reuse the old entry's id: // * if the mtime is the same, the file was probably been renamed. // * if the path is the same, the file may just have been updated if let Some(removed_entry) = self.removed_entries.take_by_inode(inode) { (removed_entry.mtime == Some(mtime) || *removed_entry.path == *path) .then_some(removed_entry.id) } else { Some(self.snapshot.entry_for_path(path)?.id) } } async fn insert_entry(&mut self, entry: Entry, fs: &dyn Fs, watcher: &dyn Watcher) -> Entry { let entry = self.snapshot.insert_entry(entry, fs).await; if entry.path.file_name() == Some(&DOT_GIT) { self.insert_git_repository(entry.path.clone(), fs, watcher) .await; } #[cfg(feature = "test-support")] self.snapshot.check_invariants(false); entry } fn populate_dir( &mut self, parent_path: Arc, entries: impl IntoIterator, ignore: Option>, ) { let mut parent_entry = if let Some(parent_entry) = self .snapshot .entries_by_path .get(&PathKey(parent_path.clone()), ()) { parent_entry.clone() } else { log::warn!( "populating a directory {:?} that has been removed", parent_path ); return; }; match parent_entry.kind { EntryKind::PendingDir | EntryKind::UnloadedDir => parent_entry.kind = EntryKind::Dir, EntryKind::Dir => {} _ => return, } if let Some(ignore) = ignore { let abs_parent_path = self .snapshot .abs_path .as_path() .join(parent_path.as_std_path()) .into(); self.snapshot .ignores_by_parent_abs_path .insert(abs_parent_path, (ignore, false)); } let parent_entry_id = parent_entry.id; self.scanned_dirs.insert(parent_entry_id); let mut entries_by_path_edits = vec![Edit::Insert(parent_entry)]; let mut entries_by_id_edits = Vec::new(); for entry in entries { entries_by_id_edits.push(Edit::Insert(PathEntry { id: entry.id, path: entry.path.clone(), is_ignored: entry.is_ignored, scan_id: self.snapshot.scan_id, })); entries_by_path_edits.push(Edit::Insert(entry)); } self.snapshot .entries_by_path .edit(entries_by_path_edits, ()); self.snapshot.entries_by_id.edit(entries_by_id_edits, ()); if let Err(ix) = self.changed_paths.binary_search(&parent_path) { self.changed_paths.insert(ix, parent_path.clone()); } #[cfg(feature = "test-support")] self.snapshot.check_invariants(false); } fn remove_path_from_snapshot_and_unwatch( &mut self, path: &RelPath, watcher: &dyn Watcher, preserve_repository_watches: bool, ) { // When the caller preserves repository watches, the removal must not // prune git repositories: either the subtree is about to be re-scanned, // or its entries are being unloaded by depth deferral while the // repositories stay active. Pruning here would transiently drop and // then re-create them with fresh `RepositoryId`s. let prune_repositories = !preserve_repository_watches; let removed_descendant_abs_paths = self.remove_path_from_snapshot(path, prune_repositories); self.unwatch_path( watcher, path, removed_descendant_abs_paths, preserve_repository_watches, ); } fn unwatch_path( &mut self, watcher: &dyn Watcher, path: &RelPath, removed_descendant_abs_paths: Vec, preserve_repository_watches: bool, ) { let mut repository_watches_to_preserve = HashSet::>::default(); if preserve_repository_watches { for repository in self.snapshot.git_repositories.values() { repository_watches_to_preserve.insert(repository.common_dir_abs_path.clone()); repository_watches_to_preserve.insert(repository.repository_dir_abs_path.clone()); } } for removed_dir_abs_path in removed_descendant_abs_paths { if repository_watches_to_preserve.contains(removed_dir_abs_path.as_path()) { continue; } watcher.remove(&removed_dir_abs_path).log_err(); } self.snapshot .external_canonical_to_relative .retain(|canonical, relative| { if relative.starts_with(path) { if !repository_watches_to_preserve.contains(canonical.as_ref()) { watcher.remove(canonical.as_ref()).log_err(); } false } else { true } }); } fn remove_path_from_snapshot( &mut self, path: &RelPath, prune_repositories: bool, ) -> Vec { log::trace!("background scanner removing path {path:?}"); let mut new_entries; let removed_entries; { let mut cursor = self .snapshot .entries_by_path .cursor::(()); new_entries = cursor.slice(&TraversalTarget::path(path), Bias::Left); removed_entries = cursor.slice(&TraversalTarget::successor(path), Bias::Left); new_entries.append(cursor.suffix(), ()); } self.snapshot.entries_by_path = new_entries; let mut removed_ids = Vec::with_capacity(removed_entries.summary().count); let mut removed_dir_abs_paths = Vec::new(); for entry in removed_entries.cursor::<()>(()) { if entry.is_dir() { let watch_path = self .watched_dir_abs_paths_by_entry_id .remove(&entry.id) .map(|path| path.as_ref().to_path_buf()) .unwrap_or_else(|| self.snapshot.absolutize(&entry.path)); removed_dir_abs_paths.push(watch_path); } self.removed_entries.insert(entry); if entry.path.file_name() == Some(GITIGNORE) { let abs_parent_path = self.snapshot.absolutize(&entry.path.parent().unwrap()); if let Some((_, needs_update)) = self .snapshot .ignores_by_parent_abs_path .get_mut(abs_parent_path.as_path()) { *needs_update = true; } } if let Err(ix) = removed_ids.binary_search(&entry.id) { removed_ids.insert(ix, entry.id); } } self.snapshot .entries_by_id .edit(removed_ids.iter().map(|&id| Edit::Remove(id)).collect(), ()); // Only prune git repositories when the entries are being genuinely // removed. During a recursive refresh (e.g. a watcher-forced rescan), // the subtree is removed and immediately re-scanned; dropping the // repositories here would make them flap, causing the GitStore to // tear them down and re-create them with fresh `RepositoryId`s. Stale // repositories are instead reaped authoritatively (against the actual // filesystem) in `update_git_repositories`. if prune_repositories { self.snapshot .git_repositories .retain(|id, _| removed_ids.binary_search(id).is_err()); } #[cfg(feature = "test-support")] self.snapshot.check_invariants(false); removed_dir_abs_paths } async fn insert_git_repository( &mut self, dot_git_path: Arc, fs: &dyn Fs, watcher: &dyn Watcher, ) { let work_dir_path: Arc = match dot_git_path.parent() { Some(parent_dir) => { // Guard against repositories inside the repository metadata if parent_dir .components() .any(|component| component == DOT_GIT) { log::debug!( "not building git repository for nested `.git` directory, `.git` path in the worktree: {dot_git_path:?}" ); return; }; parent_dir.into() } None => { // `dot_git_path.parent().is_none()` means `.git` directory is the opened worktree itself, // no files inside that directory are tracked by git, so no need to build the repo around it log::debug!( "not building git repository for the worktree itself, `.git` path in the worktree: {dot_git_path:?}" ); return; } }; let dot_git_abs_path = Arc::from(self.snapshot.absolutize(&dot_git_path).as_ref()); self.insert_git_repository_for_path( WorkDirectory::InProject { relative_path: work_dir_path, }, dot_git_abs_path, fs, watcher, ) .await .log_err(); } async fn insert_git_repository_for_path( &mut self, work_directory: WorkDirectory, dot_git_abs_path: Arc, fs: &dyn Fs, watcher: &dyn Watcher, ) -> Result { let work_dir_entry = self .snapshot .entry_for_path(&work_directory.path_key().0) .with_context(|| { format!( "working directory `{}` not indexed", work_directory .path_key() .0 .display(self.snapshot.path_style) ) })?; let work_directory_abs_path = self.snapshot.work_directory_abs_path(&work_directory); let (repository_dir_abs_path, common_dir_abs_path) = discover_git_paths(&dot_git_abs_path, fs).await; watcher .add(&common_dir_abs_path) .context("failed to add common directory to watcher") .log_err(); watcher .add(&repository_dir_abs_path) .context("failed to add repository directory to watcher") .log_err(); watch_git_dir_subdirectories(&common_dir_abs_path, fs, watcher).await; if repository_dir_abs_path != common_dir_abs_path { watch_git_dir_subdirectories(&repository_dir_abs_path, fs, watcher).await; } let work_directory_id = work_dir_entry.id; // A repository can be re-inserted when its `.git` entry is re-discovered, e.g. // during a watcher-forced rescan. Carry the existing `git_dir_scan_id` forward // in that case: the snapshot diff detects git changes by comparing scan ids, so // resetting the id would wipe out a bump made earlier in the same scan cycle, // silently dropping the corresponding git update. Deliberately don't bump the // id here either: re-insertion is snapshot bookkeeping, not evidence that git // state changed. It also happens on non-lossy paths (explicit refreshes, // path-prefix scans) where we know nothing in `.git` changed, and a bump would // trigger a spurious full reload of the repository's git state. Only // `update_git_repositories` claims that git state changed, by stamping the // current scan id. let git_dir_scan_id = self .snapshot .git_repositories .get(&work_directory_id) .map_or(0, |existing_repository| existing_repository.git_dir_scan_id); let local_repository = LocalRepositoryEntry { work_directory_id, work_directory, work_directory_abs_path: work_directory_abs_path.as_path().into(), git_dir_scan_id, dot_git_abs_path, common_dir_abs_path, repository_dir_abs_path, }; self.snapshot .git_repositories .insert(work_directory_id, local_repository.clone()); log::trace!("inserting new local git repository"); Ok(local_repository) } } /// Watches the directories inside a git directory that git writes ref updates to. /// /// On Linux and FreeBSD the native file watcher is non-recursive, so a watch on the git /// directory itself does not report changes to files nested below it, such as the loose /// refs that git updates on commit, fetch, and branch operations. Watch the `refs` tree /// (its directories are watched individually because branch names may contain slashes) /// and, for repositories using the reftable backend, the `reftable` directory. On /// platforms with recursive watchers these calls are deduplicated against the existing /// recursive registration, making them effectively free. async fn watch_git_dir_subdirectories(git_dir_abs_path: &Path, fs: &dyn Fs, watcher: &dyn Watcher) { let reftable_dir_abs_path = git_dir_abs_path.join(REFTABLE_DIR); if fs.is_dir(&reftable_dir_abs_path).await { watcher .add(&reftable_dir_abs_path) .context("failed to add reftable directory to watcher") .log_err(); } watch_dir_tree(git_dir_abs_path.join(REFS_DIR), fs, watcher).await; } /// Watches a directory and all of its descendant directories. /// /// Each directory is watched before its children are enumerated, so that a child /// created concurrently is either seen by the enumeration or reported by the watch. async fn watch_dir_tree(root_abs_path: PathBuf, fs: &dyn Fs, watcher: &dyn Watcher) { let mut dirs_to_watch = vec![root_abs_path]; while let Some(dir_abs_path) = dirs_to_watch.pop() { if !fs.is_dir(&dir_abs_path).await { continue; } watcher .add(&dir_abs_path) .with_context(|| format!("failed to watch directory {dir_abs_path:?}")) .log_err(); let Some(mut children) = fs.read_dir(&dir_abs_path).await.log_err() else { continue; }; while let Some(child_abs_path) = children.next().await { let Some(child_abs_path) = child_abs_path.log_err() else { continue; }; if fs.is_dir(&child_abs_path).await { dirs_to_watch.push(child_abs_path); } } } } async fn is_dot_git(path: &Path, fs: &dyn Fs) -> bool { if let Some(file_name) = path.file_name() && file_name == DOT_GIT { return true; } // If we're in a bare repository, we are not inside a `.git` folder. In a // bare repository, the root folder contains what would normally be in the // `.git` folder. let head_metadata = fs.metadata(&path.join("HEAD")).await; if !matches!(head_metadata, Ok(Some(_))) { return false; } let config_metadata = fs.metadata(&path.join("config")).await; matches!(config_metadata, Ok(Some(_))) } async fn build_gitignore(abs_path: &Path, fs: &dyn Fs) -> Result { let parent = abs_path.parent().unwrap_or_else(|| Path::new("/")); build_gitignore_with_root(abs_path, parent, fs).await } async fn build_gitignore_with_root(abs_path: &Path, root: &Path, fs: &dyn Fs) -> Result { let contents = fs .load(abs_path) .await .with_context(|| format!("failed to load gitignore file at {}", abs_path.display()))?; let mut builder = GitignoreBuilder::new(root); for line in contents.lines() { builder.add_line(Some(abs_path.into()), line)?; } Ok(builder.build()?) } impl Deref for Worktree { type Target = Snapshot; fn deref(&self) -> &Self::Target { match self { Worktree::Local(worktree) => &worktree.snapshot, Worktree::Remote(worktree) => &worktree.snapshot, } } } impl Deref for LocalWorktree { type Target = LocalSnapshot; fn deref(&self) -> &Self::Target { &self.snapshot } } impl Deref for RemoteWorktree { type Target = Snapshot; fn deref(&self) -> &Self::Target { &self.snapshot } } impl fmt::Debug for LocalWorktree { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { self.snapshot.fmt(f) } } impl fmt::Debug for Snapshot { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { struct EntriesById<'a>(&'a SumTree); struct EntriesByPath<'a>(&'a SumTree); impl fmt::Debug for EntriesByPath<'_> { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { f.debug_map() .entries(self.0.iter().map(|entry| (&entry.path, entry.id))) .finish() } } impl fmt::Debug for EntriesById<'_> { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { f.debug_list().entries(self.0.iter()).finish() } } f.debug_struct("Snapshot") .field("id", &self.id) .field("root_name", &self.root_name) .field("entries_by_path", &EntriesByPath(&self.entries_by_path)) .field("entries_by_id", &EntriesById(&self.entries_by_id)) .finish() } } #[derive(Debug, Clone, PartialEq)] pub struct File { pub worktree: Entity, pub path: Arc, pub disk_state: DiskState, pub entry_id: Option, pub is_local: bool, pub is_private: bool, } impl language::File for File { fn as_local(&self) -> Option<&dyn language::LocalFile> { if self.is_local { Some(self) } else { None } } fn disk_state(&self) -> DiskState { self.disk_state } fn path(&self) -> &Arc { &self.path } fn full_path(&self, cx: &App) -> PathBuf { self.worktree.read(cx).full_path(&self.path) } /// Returns the last component of this handle's absolute path. If this handle refers to the root /// of its worktree, then this method will return the name of the worktree itself. fn file_name<'a>(&'a self, cx: &'a App) -> &'a str { self.path .file_name() .unwrap_or_else(|| self.worktree.read(cx).root_name_str()) } fn worktree_id(&self, cx: &App) -> WorktreeId { self.worktree.read(cx).id() } fn to_proto(&self, cx: &App) -> rpc::proto::File { rpc::proto::File { worktree_id: self.worktree.read(cx).id().to_proto(), entry_id: self.entry_id.map(|id| id.to_proto()), path: self.path.as_ref().as_unix_str().to_owned(), mtime: self.disk_state.mtime().map(|time| time.into()), is_deleted: self.disk_state.is_deleted(), is_historic: matches!(self.disk_state, DiskState::Historic { .. }), } } fn is_private(&self) -> bool { self.is_private } fn path_style(&self, cx: &App) -> PathStyle { self.worktree.read(cx).path_style() } fn can_open(&self) -> bool { true } } impl language::LocalFile for File { fn abs_path(&self, cx: &App) -> PathBuf { self.worktree.read(cx).absolutize(&self.path) } fn load(&self, cx: &App) -> Task> { let worktree = self.worktree.read(cx).as_local().unwrap(); let abs_path = worktree.absolutize(&self.path); let fs = worktree.fs.clone(); cx.background_spawn(async move { fs.load(&abs_path).await }) } fn load_bytes(&self, cx: &App) -> Task>> { let worktree = self.worktree.read(cx).as_local().unwrap(); let abs_path = worktree.absolutize(&self.path); let fs = worktree.fs.clone(); cx.background_spawn(async move { fs.load_bytes(&abs_path).await }) } } impl File { pub fn for_entry(entry: Entry, worktree: Entity) -> Arc { Arc::new(Self { worktree, path: entry.path.clone(), disk_state: if let Some(mtime) = entry.mtime { DiskState::Present { mtime, size: entry.size, } } else { DiskState::New }, entry_id: Some(entry.id), is_local: true, is_private: entry.is_private, }) } pub fn from_proto( proto: rpc::proto::File, worktree: Entity, cx: &App, ) -> Result { let worktree_id = worktree.read(cx).as_remote().context("not remote")?.id(); anyhow::ensure!( worktree_id.to_proto() == proto.worktree_id, "worktree id does not match file" ); let disk_state = if proto.is_historic { DiskState::Historic { was_deleted: proto.is_deleted, } } else if proto.is_deleted { DiskState::Deleted } else if let Some(mtime) = proto.mtime.map(&Into::into) { DiskState::Present { mtime, size: 0 } } else { DiskState::New }; Ok(Self { worktree, path: RelPath::from_unix_str(&proto.path) .context("invalid path in file protobuf")? .into(), disk_state, entry_id: proto.entry_id.map(ProjectEntryId::from_proto), is_local: false, is_private: false, }) } pub fn from_dyn(file: Option<&Arc>) -> Option<&Self> { file.and_then(|f| { let f: &dyn language::File = f.borrow(); let f: &dyn Any = f; f.downcast_ref() }) } pub fn worktree_id(&self, cx: &App) -> WorktreeId { self.worktree.read(cx).id() } pub fn project_entry_id(&self) -> Option { match self.disk_state { DiskState::Deleted => None, _ => self.entry_id, } } } #[derive(Clone, Debug, PartialEq, Eq)] pub struct Entry { pub id: ProjectEntryId, pub kind: EntryKind, pub path: Arc, pub inode: u64, pub mtime: Option, pub canonical_path: Option>, /// Whether this entry is ignored by Git. /// /// We only scan ignored entries once the directory is expanded and /// exclude them from searches. pub is_ignored: bool, /// Whether this entry is hidden or inside hidden directory. /// /// We only scan hidden entries once the directory is expanded. pub is_hidden: bool, /// Whether this entry is always included in searches. /// /// This is used for entries that are always included in searches, even /// if they are ignored by git. Overridden by file_scan_exclusions. pub is_always_included: bool, /// Whether this entry's canonical path is outside of the worktree. /// This means the entry is only accessible from the worktree root via a /// symlink. /// /// We only scan entries outside of the worktree once the symlinked /// directory is expanded. pub is_external: bool, /// Whether this entry is considered to be a `.env` file. pub is_private: bool, /// The entry's size on disk, in bytes. pub size: u64, pub char_bag: CharBag, pub is_fifo: bool, } #[derive(Clone, Copy, Debug, PartialEq, Eq)] pub enum EntryKind { UnloadedDir, PendingDir, Dir, File, } #[derive(Clone, Copy, Debug, PartialEq)] pub enum PathChange { /// A filesystem entry was was created. Added, /// A filesystem entry was removed. Removed, /// A filesystem entry was updated. Updated, /// A filesystem entry was either updated or added. We don't know /// whether or not it already existed, because the path had not /// been loaded before the event. AddedOrUpdated, /// A filesystem entry was found during the initial scan of the worktree. Loaded, } #[derive(Clone, Debug, PartialEq, Eq)] pub struct UpdatedGitRepository { /// ID of the repository's working directory. /// /// For a repo that's above the worktree root, this is the ID of the worktree root, and hence not unique. /// It's included here to aid the GitStore in detecting when a repository's working directory is renamed. pub work_directory_id: ProjectEntryId, pub old_work_directory_abs_path: Option>, pub new_work_directory_abs_path: Option>, /// For a normal git repository checkout, the absolute path to the .git directory. /// For a worktree, the absolute path to the worktree's subdirectory inside the .git directory. pub dot_git_abs_path: Option>, pub repository_dir_abs_path: Option>, pub common_dir_abs_path: Option>, } pub type UpdatedEntriesSet = Arc<[(Arc, ProjectEntryId, PathChange)]>; pub type UpdatedGitRepositoriesSet = Arc<[UpdatedGitRepository]>; #[derive(Clone, Debug)] pub struct PathProgress<'a> { pub max_path: &'a RelPath, } #[derive(Clone, Debug)] pub struct PathSummary { pub max_path: Arc, pub item_summary: S, } impl Summary for PathSummary { type Context<'a> = S::Context<'a>; fn zero(cx: Self::Context<'_>) -> Self { Self { max_path: RelPath::empty_arc(), item_summary: S::zero(cx), } } fn add_summary(&mut self, rhs: &Self, cx: Self::Context<'_>) { self.max_path = rhs.max_path.clone(); self.item_summary.add_summary(&rhs.item_summary, cx); } } impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary> for PathProgress<'a> { fn zero(_: as Summary>::Context<'_>) -> Self { Self { max_path: RelPath::empty(), } } fn add_summary( &mut self, summary: &'a PathSummary, _: as Summary>::Context<'_>, ) { self.max_path = summary.max_path.as_ref() } } impl<'a> sum_tree::Dimension<'a, PathSummary> for GitSummary { fn zero(_cx: ()) -> Self { Default::default() } fn add_summary(&mut self, summary: &'a PathSummary, _: ()) { *self += summary.item_summary } } impl<'a> sum_tree::SeekTarget<'a, PathSummary, Dimensions, GitSummary>> for PathTarget<'_> { fn cmp( &self, cursor_location: &Dimensions, GitSummary>, _: (), ) -> Ordering { self.cmp_path(cursor_location.0.max_path) } } impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary> for PathKey { fn zero(_: S::Context<'_>) -> Self { Default::default() } fn add_summary(&mut self, summary: &'a PathSummary, _: S::Context<'_>) { self.0 = summary.max_path.clone(); } } impl<'a, S: Summary> sum_tree::Dimension<'a, PathSummary> for TraversalProgress<'a> { fn zero(_cx: S::Context<'_>) -> Self { Default::default() } fn add_summary(&mut self, summary: &'a PathSummary, _: S::Context<'_>) { self.max_path = summary.max_path.as_ref(); } } impl Entry { fn new( path: Arc, metadata: &fs::Metadata, id: ProjectEntryId, root_char_bag: CharBag, canonical_path: Option>, ) -> Self { let char_bag = char_bag_for_path(root_char_bag, &path); Self { id, kind: if metadata.is_dir { EntryKind::PendingDir } else { EntryKind::File }, path, inode: metadata.inode, mtime: Some(metadata.mtime), size: metadata.len, canonical_path, is_ignored: false, is_hidden: false, is_always_included: false, is_external: false, is_private: false, char_bag, is_fifo: metadata.is_fifo, } } pub fn is_created(&self) -> bool { self.mtime.is_some() } pub fn is_dir(&self) -> bool { self.kind.is_dir() } pub fn is_file(&self) -> bool { self.kind.is_file() } } impl EntryKind { pub fn is_dir(&self) -> bool { matches!( self, EntryKind::Dir | EntryKind::PendingDir | EntryKind::UnloadedDir ) } pub fn is_unloaded(&self) -> bool { matches!(self, EntryKind::UnloadedDir) } pub fn is_file(&self) -> bool { matches!(self, EntryKind::File) } } impl sum_tree::Item for Entry { type Summary = EntrySummary; fn summary(&self, _cx: ()) -> Self::Summary { let non_ignored_count = if self.is_ignored && !self.is_always_included { 0 } else { 1 }; let file_count; let non_ignored_file_count; if self.is_file() { file_count = 1; non_ignored_file_count = non_ignored_count; } else { file_count = 0; non_ignored_file_count = 0; } EntrySummary { max_path: self.path.clone(), count: 1, non_ignored_count, file_count, non_ignored_file_count, deferred_scan_dir_count: usize::from( self.kind == EntryKind::UnloadedDir && !self.is_ignored && !self.is_external && !self.path.is_empty(), ), } } } impl sum_tree::KeyedItem for Entry { type Key = PathKey; fn key(&self) -> Self::Key { PathKey(self.path.clone()) } } #[derive(Clone, Debug)] pub struct EntrySummary { max_path: Arc, count: usize, non_ignored_count: usize, file_count: usize, non_ignored_file_count: usize, deferred_scan_dir_count: usize, } impl Default for EntrySummary { fn default() -> Self { Self { max_path: Arc::from(RelPath::empty()), count: 0, non_ignored_count: 0, file_count: 0, non_ignored_file_count: 0, deferred_scan_dir_count: 0, } } } impl sum_tree::ContextLessSummary for EntrySummary { fn zero() -> Self { Default::default() } fn add_summary(&mut self, rhs: &Self) { self.max_path = rhs.max_path.clone(); self.count += rhs.count; self.non_ignored_count += rhs.non_ignored_count; self.file_count += rhs.file_count; self.non_ignored_file_count += rhs.non_ignored_file_count; self.deferred_scan_dir_count += rhs.deferred_scan_dir_count; } } #[derive(Clone, Debug)] struct PathEntry { id: ProjectEntryId, path: Arc, is_ignored: bool, scan_id: usize, } impl sum_tree::Item for PathEntry { type Summary = PathEntrySummary; fn summary(&self, _cx: ()) -> Self::Summary { PathEntrySummary { max_id: self.id } } } impl sum_tree::KeyedItem for PathEntry { type Key = ProjectEntryId; fn key(&self) -> Self::Key { self.id } } #[derive(Clone, Debug, Default)] struct PathEntrySummary { max_id: ProjectEntryId, } impl sum_tree::ContextLessSummary for PathEntrySummary { fn zero() -> Self { Default::default() } fn add_summary(&mut self, summary: &Self) { self.max_id = summary.max_id; } } impl<'a> sum_tree::Dimension<'a, PathEntrySummary> for ProjectEntryId { fn zero(_cx: ()) -> Self { Default::default() } fn add_summary(&mut self, summary: &'a PathEntrySummary, _: ()) { *self = summary.max_id; } } #[derive(Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)] pub struct PathKey(pub Arc); impl Default for PathKey { fn default() -> Self { Self(RelPath::empty_arc()) } } impl<'a> sum_tree::Dimension<'a, EntrySummary> for PathKey { fn zero(_cx: ()) -> Self { Default::default() } fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) { self.0 = summary.max_path.clone(); } } struct BackgroundScanner { state: async_lock::Mutex, fs: Arc, fs_case_sensitive: bool, status_updates_tx: UnboundedSender, executor: BackgroundExecutor, scan_requests_rx: async_channel::Receiver, path_prefixes_to_scan_rx: async_channel::Receiver, next_entry_id: Arc, phase: BackgroundScannerPhase, watcher: Arc, settings: WorktreeSettings, share_private_files: bool, track_git_repositories: bool, /// Whether this is a single-file worktree (root is a file, not a directory). /// Used to determine if we should give up after repeated canonicalization failures. is_single_file: bool, defer_watch: bool, } #[derive(Copy, Clone, PartialEq)] enum BackgroundScannerPhase { InitialScan, EventsReceivedDuringInitialScan, Events, } impl BackgroundScanner { async fn run(&mut self, mut fs_events_rx: Pin>>>) { let root_abs_path; let scanning_enabled; { let state = self.state.lock().await; root_abs_path = state.snapshot.abs_path.clone(); scanning_enabled = state.scanning_enabled; } // If the worktree root does not contain a git repository, then find // the git repository in an ancestor directory. Find any gitignore files // in ancestor directories. let repo = if scanning_enabled && self.track_git_repositories { let (ignores, exclude, repo) = discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await; let mut state = self.state.lock().await; state.snapshot.ignores_by_parent_abs_path.extend(ignores); if let Some(exclude) = exclude { let work_directory_abs_path: Arc = repo .as_ref() .map(|(_, work_directory)| { state .snapshot .work_directory_abs_path(work_directory) .into() }) .unwrap_or_else(|| root_abs_path.as_path().into()); state .snapshot .repo_exclude_by_work_dir_abs_path .insert(work_directory_abs_path, (exclude, false)); } repo } else { None }; let containing_git_repository = if let Some((ancestor_dot_git, work_directory)) = repo && scanning_enabled && self.track_git_repositories { maybe!(async { self.state .lock() .await .insert_git_repository_for_path( work_directory, ancestor_dot_git.clone().into(), self.fs.as_ref(), self.watcher.as_ref(), ) .await .log_err()?; Some(ancestor_dot_git) }) .await } else { None }; log::trace!("containing git repository: {containing_git_repository:?}"); let global_gitignore_file = paths::global_gitignore_path(); let mut global_gitignore_events = if let Some(global_gitignore_path) = &global_gitignore_file && scanning_enabled && self.track_git_repositories { let is_file = self.fs.is_file(&global_gitignore_path).await; self.state.lock().await.snapshot.global_gitignore = if is_file { build_gitignore(global_gitignore_path, self.fs.as_ref()) .await .ok() .map(Arc::new) } else { None }; if is_file { self.fs .watch(global_gitignore_path, FS_WATCH_LATENCY) .await .0 } else { Box::pin(futures::stream::pending()) } } else { self.state.lock().await.snapshot.global_gitignore = None; Box::pin(futures::stream::pending()) }; let (scan_job_tx, scan_job_rx) = async_channel::unbounded(); { let mut state = self.state.lock().await; state.snapshot.scan_id += 1; if let Some(mut root_entry) = state.snapshot.root_entry().cloned() { let ignore_stack = state .snapshot .ignore_stack_for_abs_path(root_abs_path.as_path(), true, self.fs.as_ref()) .await; if ignore_stack.is_abs_path_ignored(root_abs_path.as_path(), true) { root_entry.is_ignored = true; let mut root_entry = root_entry.clone(); state.reuse_entry_id(&mut root_entry); state .insert_entry(root_entry, self.fs.as_ref(), self.watcher.as_ref()) .await; } if root_entry.is_dir() && state.scanning_enabled { state .enqueue_scan_dir( root_abs_path.as_path().into(), &root_entry, &scan_job_tx, self.fs.as_ref(), ) .await; } } }; // Perform an initial scan of the directory. drop(scan_job_tx); self.scan_dirs(true, scan_job_rx).await; { let mut state = self.state.lock().await; state.snapshot.completed_scan_id = state.snapshot.scan_id; if state.scanning_enabled { let deferred_scan_dir_count = state.snapshot.deferred_scan_dir_count(); if deferred_scan_dir_count > 0 && let Some(file_scan_depth) = self.settings.file_scan_depth { log::info!( "deferred indexing of {deferred_scan_dir_count} directories in {:?} that are outside of git repositories and deeper than file_scan_depth={file_scan_depth}", root_abs_path.as_path(), ); } } } self.send_status_update(false, SmallVec::new(), &[]).await; if self.defer_watch { let (events, watcher) = self .fs .watch(root_abs_path.as_path(), FS_WATCH_LATENCY) .await; self.watcher = watcher; fs_events_rx = Box::pin(events.map(|events| events.into_iter().collect())); let state = self.state.lock().await; for target in state.symlink_paths_by_target.keys() { if !target.starts_with(root_abs_path.as_path()) { self.watcher.add(target).log_err(); } } for repo in state.snapshot.git_repositories.values() { if !repo .common_dir_abs_path .starts_with(root_abs_path.as_path()) { self.watcher.add(&repo.common_dir_abs_path).log_err(); } if !repo .repository_dir_abs_path .starts_with(root_abs_path.as_path()) { self.watcher.add(&repo.repository_dir_abs_path).log_err(); } } drop(state); } // Process any any FS events that occurred while performing the initial scan. // For these events, update events cannot be as precise, because we didn't // have the previous state loaded yet. self.phase = BackgroundScannerPhase::EventsReceivedDuringInitialScan; if let Poll::Ready(Some(mut paths)) = futures::poll!(fs_events_rx.next()) { while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) { paths.extend(more_paths); } self.process_events( paths .into_iter() .filter(|event| event.kind.is_some()) .collect(), ) .await; } if let Some(abs_path) = containing_git_repository { self.process_events(vec![PathEvent { path: abs_path, kind: Some(fs::PathEventKind::Changed), }]) .await; } // Continue processing events until the worktree is dropped. self.phase = BackgroundScannerPhase::Events; let root_path_check_timer = self.executor.timer(ROOT_PATH_CHECK_INTERVAL).fuse(); futures::pin_mut!(root_path_check_timer); let mut root_path_missing = false; loop { select_biased! { // Process any path refresh requests from the worktree. Prioritize // these before handling changes reported by the filesystem. request = self.next_scan_request().fuse() => { let Ok(request) = request else { break }; if !self.process_scan_request(request, false).await { return; } } path_prefix_request = self.path_prefixes_to_scan_rx.recv().fuse() => { let Ok(request) = path_prefix_request else { break }; if self.state.lock().await.path_prefixes_to_scan.contains(&request.path) { self.send_status_update(false, request.done, &[]).await; continue; } log::trace!("adding path prefix {:?}", request.path); let did_scan = self.forcibly_load_paths(std::slice::from_ref(&request.path)).await; if did_scan { let abs_path = { let mut state = self.state.lock().await; state.path_prefixes_to_scan.insert(request.path.clone()); state.snapshot.absolutize(&request.path) }; if let Some(abs_path) = self.fs.canonicalize(&abs_path).await.log_err() { self.process_events(vec![PathEvent { path: abs_path, kind: Some(fs::PathEventKind::Changed), }]) .await; } } self.send_status_update(false, request.done, &[]).await; } paths = fs_events_rx.next().fuse() => { let Some(mut paths) = paths else { break }; while let Poll::Ready(Some(more_paths)) = futures::poll!(fs_events_rx.next()) { paths.extend(more_paths); } self.process_events(paths.into_iter().filter(|event| event.kind.is_some()).collect()).await; } _ = global_gitignore_events.next().fuse() => { if let Some(path) = &global_gitignore_file { self.update_global_gitignore(&path).await; } } _ = root_path_check_timer => { let root_path = self.state.lock().await.snapshot.abs_path.clone(); match self.fs.canonicalize(root_path.as_path()).await { Ok(_) => root_path_missing = false, Err(error) => { if !root_path_missing { root_path_missing = true; self.report_root_moved_or_deleted(&root_path, &error).await; } } } root_path_check_timer.set(self.executor.timer(ROOT_PATH_CHECK_INTERVAL).fuse()); } } } } async fn process_scan_request(&self, mut request: ScanRequest, scanning: bool) -> bool { log::debug!("rescanning paths {:?}", request.relative_paths); { let state = self.state.lock().await; let mut missing_ancestors = Vec::new(); for path in &request.relative_paths { let mut candidates = Vec::new(); let mut covered_by_existing_ancestor = false; for ancestor in path.ancestors().skip(1) { if let Some(entry) = state.snapshot.entry_for_path(ancestor) { covered_by_existing_ancestor = entry.kind == EntryKind::UnloadedDir || entry.kind == EntryKind::File; break; } candidates.push(ancestor.into_arc()); } if !covered_by_existing_ancestor { missing_ancestors.extend(candidates); } } request.relative_paths.extend(missing_ancestors); } request.relative_paths.sort_unstable(); request.relative_paths.dedup(); self.forcibly_load_paths(&request.relative_paths).await; let root_path = self.state.lock().await.snapshot.abs_path.clone(); let root_canonical_path = self.fs.canonicalize(root_path.as_path()).await; let root_canonical_path = match &root_canonical_path { Ok(path) => SanitizedPath::new(path), Err(err) => { log::error!("failed to canonicalize root path {root_path:?}: {err:#}"); return true; } }; let abs_paths = request .relative_paths .iter() .map(|path| { if path.file_name().is_some() { root_canonical_path.as_path().join(path.as_std_path()) } else { root_canonical_path.as_path().to_path_buf() } }) .collect::>(); { let mut state = self.state.lock().await; let is_idle = state.snapshot.completed_scan_id == state.snapshot.scan_id; state.snapshot.scan_id += 1; if is_idle { state.snapshot.completed_scan_id = state.snapshot.scan_id; } } self.reload_entries_for_paths( &root_path, &root_canonical_path, &request.relative_paths, abs_paths, None, ) .await; self.send_status_update(scanning, request.done, &[]).await } fn normalized_events_for_worktree( state: &BackgroundScannerState, root_canonical_path: &SanitizedPath, mut events: Vec, ) -> Vec { if state.symlink_paths_by_target.is_empty() { return events; } let mut mapped_events = Vec::new(); events.retain(|event| { let abs_path = SanitizedPath::new(&event.path); let mut best_match: Option<(&Arc, &SmallVec<[Arc; 1]>)> = None; let mut best_depth = 0; for (target_root, symlink_paths) in &state.symlink_paths_by_target { if abs_path.as_path().starts_with(target_root.as_ref()) { let depth = target_root.as_ref().components().count(); if depth > best_depth { best_depth = depth; best_match = Some((target_root, symlink_paths)); } } } let Some((target_root, symlink_paths)) = best_match else { return true; }; let Ok(suffix) = abs_path.as_path().strip_prefix(target_root.as_ref()) else { return true; }; // If the symlink's real target is outside this worktree, the original path // isn't visible to the worktree. Keep only the remapped symlink events. let keep_original = target_root.starts_with(root_canonical_path.as_path()); for symlink_path in symlink_paths { let mapped_path = if suffix.as_os_str().is_empty() { root_canonical_path .as_path() .join(symlink_path.as_std_path()) } else { root_canonical_path .as_path() .join(symlink_path.as_std_path()) .join(suffix) }; if mapped_path != event.path { mapped_events.push(PathEvent { path: mapped_path, kind: event.kind, }); } } keep_original }); events.extend(mapped_events); events } /// Called when the worktree root can no longer be canonicalized. Uses the /// open handle on the root to find where it moved to, and reports the new /// location (or, for single-file worktrees, the deletion) to the worktree. async fn report_root_moved_or_deleted( &self, root_path: &Arc, canonicalize_error: &anyhow::Error, ) { let new_path = self .state .lock() .await .snapshot .root_file_handle .clone() .and_then(|handle| match handle.current_path(&self.fs) { Ok(new_path) => Some(new_path), Err(e) => { log::error!("Failed to refresh worktree root path: {e:#}"); None } }) .map(|path| SanitizedPath::new_arc(&path)) .filter(|new_path| new_path != root_path); if let Some(new_path) = new_path { log::info!( "root renamed from {:?} to {:?}", root_path.as_path(), new_path.as_path(), ); self.status_updates_tx .unbounded_send(ScanState::RootUpdated { new_path }) .ok(); } else { log::error!("root path could not be canonicalized: {canonicalize_error:#}"); // For single-file worktrees, if we can't canonicalize and the file handle // fallback also failed, the file is gone - close the worktree if self.is_single_file { log::info!( "single-file worktree root {:?} no longer exists, marking as deleted", root_path.as_path() ); self.status_updates_tx .unbounded_send(ScanState::RootDeleted) .ok(); } } } async fn process_events(&self, mut events: Vec) { let root_path = self.state.lock().await.snapshot.abs_path.clone(); let root_canonical_path = match self.fs.canonicalize(root_path.as_path()).await { Ok(path) => path, Err(error) => { self.report_root_moved_or_deleted(&root_path, &error).await; return; } }; let root_canonical_path = SanitizedPath::new(&root_canonical_path); { let state = self.state.lock().await; events = Self::normalized_events_for_worktree(&state, &root_canonical_path, events); } log::debug!("raw events for process_events: {events:?}"); fn skip_ix(ranges: &mut SmallVec<[Range; 4]>, ix: usize) { if let Some(last_range) = ranges.last_mut() && last_range.end == ix { last_range.end += 1; } else { ranges.push(ix..ix + 1); } } // Check for events inside .git directories, so that we know which repositories need their git state reloaded. // // Certain directories may have FS changes, but do not lead to git data changes that Zed cares about. // Ignore these, to avoid Zed unnecessarily rescanning git metadata. let skipped_file_names_in_dot_git = [COMMIT_MESSAGE, FETCH_HEAD, ORIG_HEAD, BISECT_LOG, GC_PID]; let skipped_dirs_in_dot_git = [ FSMONITOR_DAEMON, LFS_DIR, OBJECTS_DIR, HOOKS_DIR, REBASE_MERGE_DIR, REBASE_APPLY_DIR, SEQUENCER_DIR, ]; let mut dot_git_abs_paths = Vec::new(); let mut work_dirs_needing_exclude_update = Vec::new(); { let snapshot = &self.state.lock().await.snapshot; let mut ranges_to_drop = SmallVec::<[Range; 4]>::new(); // On Windows, creating or deleting a file directly inside `.git` // updates the directory's last-write time, so the watcher reports a // bare `.git` Changed event alongside the file's own event. When the // file event is filtered out (e.g. git's transient `index.lock`), the // bare event carries no information either, so acting on it would // turn every ignored lock file into a git rescan. Since a rescan's // own `git diff` can take `index.lock`, that feeds back into an // infinite loop of rescans. So bare `.git` events are deferred here // and only rescanned when nothing in the batch explains them; a // standalone bare event (macOS coalescing) still triggers a rescan. let mut bare_dot_git_abs_paths = Vec::new(); let mut dot_git_dirs_with_ignored_events = Vec::new(); for (ix, event) in events.iter().enumerate() { let abs_path = SanitizedPath::new(&event.path); let mut dot_git_paths = None; if self.track_git_repositories { for ancestor in abs_path.as_path().ancestors() { if is_dot_git(ancestor, self.fs.as_ref()).await { let path_in_git_dir = abs_path .as_path() .strip_prefix(ancestor) .expect("stripping off the ancestor"); dot_git_paths = Some((ancestor.to_owned(), path_in_git_dir.to_owned())); break; } } } if let Some((dot_git_abs_path, path_in_git_dir)) = dot_git_paths { let is_ignored = skipped_file_names_in_dot_git.iter().any(|skipped| { path_in_git_dir .file_name() .is_some_and(|file_name| file_name == OsStr::new(skipped)) }) || (path_in_git_dir.starts_with(LOGS_DIR) && path_in_git_dir != Path::new(LOGS_REF_STASH)) || (path_in_git_dir.starts_with(INFO_DIR) && path_in_git_dir != Path::new(REPO_EXCLUDE)) || skipped_dirs_in_dot_git.iter().any(|skipped_git_subdir| { path_in_git_dir.starts_with(skipped_git_subdir) }) || path_in_git_dir.extension().is_some_and(|ext| ext == "lock") || (path_in_git_dir.components().count() == 1 && path_in_git_dir .extension() .is_some_and(|ext| ext == "new" || ext == "tmp")); let is_dot_git = path_in_git_dir == Path::new("") && matches!(event.kind, Some(PathEventKind::Changed)) && self.fs.is_dir(&dot_git_abs_path).await; if is_ignored { log::debug!( "ignoring event {abs_path:?} as it's in the .git directory among skipped files or directories" ); if !dot_git_dirs_with_ignored_events.contains(&dot_git_abs_path) { dot_git_dirs_with_ignored_events.push(dot_git_abs_path); } skip_ix(&mut ranges_to_drop, ix); continue; } if is_dot_git { log::debug!( "ignoring event {abs_path:?} for .git directory itself (kind: {:?})", event.kind ); if !bare_dot_git_abs_paths.contains(&dot_git_abs_path) { bare_dot_git_abs_paths.push(dot_git_abs_path); } skip_ix(&mut ranges_to_drop, ix); continue; } if !dot_git_abs_paths.contains(&dot_git_abs_path) { log::debug!( "detected update within git repo at {dot_git_abs_path:?}: {abs_path:?}" ); dot_git_abs_paths.push(dot_git_abs_path); } // New directories can appear under the `refs` tree at any time, e.g. when a // remote is added or a branch name contains slashes. On platforms where the // native watcher is non-recursive they need their own watches, or subsequent // ref updates inside them would go unnoticed. The subtree is walked because // nested directories may have been created before this watch took effect. if matches!(event.kind, Some(PathEventKind::Created)) && path_in_git_dir .components() .any(|component| component.as_os_str() == OsStr::new(REFS_DIR)) { watch_dir_tree( abs_path.as_path().to_path_buf(), self.fs.as_ref(), self.watcher.as_ref(), ) .await; } } // A rescan event means the watcher lost sync and events under the // rescanned path were dropped, possibly including events inside `.git` // directories. Reload the git state of every repository with a git // directory under the rescanned path, since changes there may have // gone unseen. if self.track_git_repositories && matches!(event.kind, Some(PathEventKind::Rescan)) { for repository in snapshot.git_repositories.values() { let affected_by_rescan = [ &repository.dot_git_abs_path, &repository.common_dir_abs_path, &repository.repository_dir_abs_path, ] .iter() .any(|git_dir_abs_path| git_dir_abs_path.starts_with(abs_path.as_path())); let dot_git_abs_path = repository.dot_git_abs_path.to_path_buf(); if affected_by_rescan && !dot_git_abs_paths.contains(&dot_git_abs_path) { log::debug!( "reloading git repo at {dot_git_abs_path:?} due to rescan of {abs_path:?}" ); dot_git_abs_paths.push(dot_git_abs_path); } } } if self.track_git_repositories && abs_path .as_path() .ends_with(Path::new(DOT_GIT).join(REPO_EXCLUDE)) { if let Some(repository) = snapshot.git_repositories.values().find(|repo| { repo.common_dir_abs_path.join(REPO_EXCLUDE) == abs_path.as_path() }) { work_dirs_needing_exclude_update .push(repository.work_directory_abs_path.clone()); } } } for dot_git_abs_path in bare_dot_git_abs_paths { if dot_git_dirs_with_ignored_events.contains(&dot_git_abs_path) { log::debug!( "not reloading git repo at {dot_git_abs_path:?}: the bare .git event is explained by filtered events in the same batch" ); } else if !dot_git_abs_paths.contains(&dot_git_abs_path) { log::debug!( "detected update within git repo at {dot_git_abs_path:?}: bare .git directory event" ); dot_git_abs_paths.push(dot_git_abs_path); } } for range_to_drop in ranges_to_drop.into_iter().rev() { events.drain(range_to_drop); } } events.sort_unstable_by(|left, right| left.path.cmp(&right.path)); events.dedup_by(|left, right| { if left.path == right.path { if matches!(left.kind, Some(fs::PathEventKind::Rescan)) { right.kind = left.kind; } true } else if left.path.starts_with(&right.path) { if matches!(left.kind, Some(fs::PathEventKind::Rescan)) { right.kind = left.kind; } true } else { false } }); let mut relative_paths = Vec::with_capacity(events.len()); { let snapshot = &self.state.lock().await.snapshot; let mut ranges_to_drop = SmallVec::<[Range; 4]>::new(); for (ix, event) in events.iter().enumerate() { let abs_path = SanitizedPath::new(&event.path); // TODO: this strips the root case-sensitively, so on a case-insensitive // volume an event whose casing differs from the canonical root is // dropped. Once `fs` exposes per-volume case-sensitivity (e.g. on the // `Fs` trait, with a per-volume cache + `FakeFs` support), fold this // comparison on case-insensitive volumes. let relative_path = if let Ok(path) = abs_path.strip_prefix(&root_canonical_path) && let Ok(path) = RelPath::new(path, PathStyle::local()) { path } else if let Ok(path) = abs_path.strip_prefix(&root_path) && let Ok(path) = RelPath::new(path, PathStyle::local()) { path } else if let Some(path) = snapshot.external_canonical_to_relative.iter().find_map( |(canonical, relative)| { abs_path .as_path() .strip_prefix(canonical.as_ref()) .ok() .and_then(|suffix| { RelPath::new(suffix, PathStyle::local()) .ok() .map(|suffix_rel| { std::borrow::Cow::Owned( relative.join(&suffix_rel).to_rel_path_buf(), ) }) }) }, ) { path } else { skip_ix(&mut ranges_to_drop, ix); continue; }; if self.track_git_repositories && abs_path.file_name() == Some(OsStr::new(GITIGNORE)) { for (_, repo) in snapshot .git_repositories .iter() .filter(|(_, repo)| repo.directory_contains(&relative_path)) { if !dot_git_abs_paths.iter().any(|dot_git_abs_path| { dot_git_abs_path == repo.common_dir_abs_path.as_ref() }) { dot_git_abs_paths.push(repo.common_dir_abs_path.to_path_buf()); } } } let parent_dir_is_loaded = relative_path.parent().is_none_or(|parent| { snapshot .entry_for_path(parent) .is_some_and(|entry| entry.kind == EntryKind::Dir) }); if !parent_dir_is_loaded { log::debug!("filtering event {relative_path:?} within unloaded directory"); skip_ix(&mut ranges_to_drop, ix); continue; } if self.settings.is_path_excluded(&relative_path) { skip_ix(&mut ranges_to_drop, ix); continue; } relative_paths.push(EventRoot { path: relative_path.into_arc(), was_rescanned: matches!(event.kind, Some(fs::PathEventKind::Rescan)), }); } for range_to_drop in ranges_to_drop.into_iter().rev() { events.drain(range_to_drop); } } if relative_paths.is_empty() && dot_git_abs_paths.is_empty() { return; } if !work_dirs_needing_exclude_update.is_empty() { let mut state = self.state.lock().await; for work_dir_abs_path in work_dirs_needing_exclude_update { if let Some((_, needs_update)) = state .snapshot .repo_exclude_by_work_dir_abs_path .get_mut(&work_dir_abs_path) { *needs_update = true; } } } self.state.lock().await.snapshot.scan_id += 1; let (scan_job_tx, scan_job_rx) = async_channel::unbounded(); if !relative_paths.is_empty() { log::debug!( "will update project paths {:?}", relative_paths .iter() .map(|event_root| &event_root.path) .collect::>() ); } self.reload_entries_for_paths( &root_path, &root_canonical_path, &relative_paths .iter() .map(|event_root| event_root.path.clone()) .collect::>(), events .into_iter() .map(|event| event.path) .collect::>(), Some(scan_job_tx.clone()), ) .await; let affected_repo_roots = if !dot_git_abs_paths.is_empty() { self.update_git_repositories(dot_git_abs_paths).await } else { Vec::new() }; { let mut ignores_to_update = self.ignores_needing_update().await; ignores_to_update.extend(affected_repo_roots); let ignores_to_update = self.order_ignores(ignores_to_update).await; let snapshot = self.state.lock().await.snapshot.clone(); self.update_ignore_statuses_for_paths(scan_job_tx, snapshot, ignores_to_update) .await; self.scan_dirs(false, scan_job_rx).await; } { let mut state = self.state.lock().await; state.snapshot.completed_scan_id = state.snapshot.scan_id; let dropped_entries = state.removed_entries.rotate().collect::>(); for entry in dropped_entries { state.scanned_dirs.remove(&entry.id); } } self.send_status_update(false, SmallVec::new(), &relative_paths) .await; } async fn update_global_gitignore(&self, abs_path: &Path) { let ignore = build_gitignore(abs_path, self.fs.as_ref()) .await .log_err() .map(Arc::new); let (prev_snapshot, ignore_stack, abs_path) = { let mut state = self.state.lock().await; state.snapshot.global_gitignore = ignore; let abs_path = state.snapshot.abs_path().clone(); let ignore_stack = state .snapshot .ignore_stack_for_abs_path(&abs_path, true, self.fs.as_ref()) .await; (state.snapshot.clone(), ignore_stack, abs_path) }; let (scan_job_tx, scan_job_rx) = async_channel::unbounded(); self.update_ignore_statuses_for_paths( scan_job_tx, prev_snapshot, vec![(abs_path, ignore_stack)], ) .await; self.scan_dirs(false, scan_job_rx).await; self.send_status_update(false, SmallVec::new(), &[]).await; } async fn forcibly_load_paths(&self, paths: &[Arc]) -> bool { let (scan_job_tx, scan_job_rx) = async_channel::unbounded(); { let mut state = self.state.lock().await; let root_path = state.snapshot.abs_path.clone(); for path in paths { for ancestor in path.ancestors() { if let Some(entry) = state.snapshot.entry_for_path(ancestor) && entry.kind == EntryKind::UnloadedDir { let abs_path = if entry.is_external { entry .canonical_path .as_ref() .map(|path| path.as_ref().to_path_buf()) .unwrap_or_else(|| root_path.join(ancestor.as_std_path())) } else { root_path.join(ancestor.as_std_path()) }; state .enqueue_scan_dir( abs_path.into(), entry, &scan_job_tx, self.fs.as_ref(), ) .await; state.paths_to_scan.insert(path.clone()); break; } } } drop(scan_job_tx); } while let Ok(job) = scan_job_rx.recv().await { self.scan_dir(&job).await.log_err(); } !mem::take(&mut self.state.lock().await.paths_to_scan).is_empty() } async fn scan_dirs( &self, enable_progress_updates: bool, scan_jobs_rx: async_channel::Receiver, ) { if self .status_updates_tx .unbounded_send(ScanState::Started) .is_err() { return; } let progress_update_count = AtomicUsize::new(0); self.executor .scoped_priority(Priority::Low, |scope| { for _ in 0..self.executor.num_cpus() { scope.spawn(async { let mut last_progress_update_count = 0; let progress_update_timer = self.progress_timer(enable_progress_updates).fuse(); futures::pin_mut!(progress_update_timer); loop { select_biased! { // Process any path refresh requests before moving on to process // the scan queue, so that user operations are prioritized. request = self.next_scan_request().fuse() => { let Ok(request) = request else { break }; if !self.process_scan_request(request, true).await { return; } } // Send periodic progress updates to the worktree. Use an atomic counter // to ensure that only one of the workers sends a progress update after // the update interval elapses. _ = progress_update_timer => { match progress_update_count.compare_exchange( last_progress_update_count, last_progress_update_count + 1, SeqCst, SeqCst ) { Ok(_) => { last_progress_update_count += 1; self.send_status_update(true, SmallVec::new(), &[]) .await; } Err(count) => { last_progress_update_count = count; } } progress_update_timer.set(self.progress_timer(enable_progress_updates).fuse()); } // Recursively load directories from the file system. job = scan_jobs_rx.recv().fuse() => { let Ok(job) = job else { break }; if let Err(err) = self.scan_dir(&job).await && job.path.is_empty() { log::error!("error scanning directory {:?}: {}", job.abs_path, err); } } } } }); } }) .await; } async fn send_status_update( &self, scanning: bool, barrier: SmallVec<[barrier::Sender; 1]>, event_roots: &[EventRoot], ) -> bool { let mut state = self.state.lock().await; if state.changed_paths.is_empty() && event_roots.is_empty() && scanning { return true; } let merged_event_roots = merge_event_roots(&state.changed_paths, event_roots); let new_snapshot = state.snapshot.clone(); let old_snapshot = mem::replace(&mut state.prev_snapshot, new_snapshot.snapshot.clone()); let changes = build_diff( self.phase, &old_snapshot, &new_snapshot, &merged_event_roots, ); state.changed_paths.clear(); self.status_updates_tx .unbounded_send(ScanState::Updated { snapshot: new_snapshot, changes, scanning, barrier, }) .is_ok() } async fn scan_dir(&self, job: &ScanJob) -> Result<()> { let root_abs_path; let root_char_bag; { let snapshot = &self.state.lock().await.snapshot; if self.settings.is_path_excluded(&job.path) { log::error!("skipping excluded directory {:?}", job.path); return Ok(()); } log::trace!("scanning directory {:?}", job.path); root_abs_path = snapshot.abs_path().clone(); root_char_bag = snapshot.root_char_bag; } let next_entry_id = self.next_entry_id.clone(); let mut ignore_stack = job.ignore_stack.clone(); let mut new_ignore = None; let mut root_canonical_path = None; let mut new_entries: Vec = Vec::new(); let mut new_jobs: Vec> = Vec::new(); let mut child_paths = self .fs .read_dir(&job.abs_path) .await? .filter_map(|entry| async { match entry { Ok(entry) => Some(entry), Err(error) => { log::error!("error processing entry {:?}", error); None } } }) .collect::>() .await; // Ensure that .git and .gitignore are processed first. swap_to_front(&mut child_paths, GITIGNORE); swap_to_front(&mut child_paths, DOT_GIT); if let Some(path) = child_paths.first() && path.ends_with(DOT_GIT) { ignore_stack.repo_root = Some(job.abs_path.clone()); ignore_stack.global_ignore_root = Some(job.abs_path.clone()); } for child_abs_path in child_paths { let child_abs_path: Arc = child_abs_path.into(); let child_name = child_abs_path.file_name().unwrap(); let Some(child_path) = child_name .to_str() .and_then(|name| Some(job.path.join(RelPath::from_unix_str(name).ok()?))) else { continue; }; let child_path: Arc = child_path.into(); if self.track_git_repositories { if child_name == DOT_GIT { let mut state = self.state.lock().await; state .insert_git_repository( child_path.clone(), self.fs.as_ref(), self.watcher.as_ref(), ) .await; } else if child_name == GITIGNORE { match build_gitignore(&child_abs_path, self.fs.as_ref()).await { Ok(ignore) => { let ignore = Arc::new(ignore); ignore_stack = ignore_stack.append( IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone(), ); new_ignore = Some(ignore); } Err(error) => { log::error!( "error loading .gitignore file {:?} - {:?}", child_name, error ); } } } } if self.settings.is_path_excluded(&child_path) { log::debug!("skipping excluded child entry {child_path:?}"); self.state .lock() .await .remove_path_from_snapshot_and_unwatch( &child_path, self.watcher.as_ref(), true, ); continue; } let child_metadata = match self.fs.metadata(&child_abs_path).await { Ok(Some(metadata)) => metadata, Ok(None) => continue, Err(err) => { log::error!("error processing {:?}: {err:#}", child_abs_path.display()); continue; } }; let mut child_entry = Entry::new( child_path.clone(), &child_metadata, ProjectEntryId::new(&next_entry_id), root_char_bag, None, ); if job.is_external { child_entry.is_external = true; } else if child_metadata.is_symlink { let canonical_path = match self.fs.canonicalize(&child_abs_path).await { Ok(path) => path, Err(err) => { log::error!("error reading target of symlink {child_abs_path:?}: {err:#}",); continue; } }; // lazily canonicalize the root path in order to determine if // symlinks point outside of the worktree. let root_canonical_path = match &root_canonical_path { Some(path) => path, None => match self.fs.canonicalize(&root_abs_path).await { Ok(path) => root_canonical_path.insert(path), Err(err) => { log::error!("error canonicalizing root {:?}: {:?}", root_abs_path, err); continue; } }, }; if !canonical_path.starts_with(root_canonical_path) { child_entry.is_external = true; } if child_metadata.is_dir { let mut state = self.state.lock().await; let paths = state .symlink_paths_by_target .entry(Arc::from(canonical_path.clone())) .or_default(); if !paths.iter().any(|path| path == &child_path) { paths.push(child_path.clone()); } } child_entry.canonical_path = Some(canonical_path.into()); } if child_entry.is_dir() { child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, true); child_entry.is_always_included = self.settings.is_path_always_included(&child_path, true); // Avoid recursing until crash in the case of a recursive symlink if job.ancestor_inodes.contains(&child_entry.inode) { new_jobs.push(None); } else { let mut ancestor_inodes = job.ancestor_inodes.clone(); ancestor_inodes.insert(child_entry.inode); new_jobs.push(Some(ScanJob { abs_path: child_abs_path.clone(), path: child_path, is_external: child_entry.is_external, ignore_stack: if child_entry.is_ignored { IgnoreStack::all() } else { ignore_stack.clone() }, ancestor_inodes, scan_queue: job.scan_queue.clone(), })); } } else { child_entry.is_ignored = ignore_stack.is_abs_path_ignored(&child_abs_path, false); child_entry.is_always_included = self.settings.is_path_always_included(&child_path, false); } { let relative_path = job .path .join(RelPath::from_unix_str(child_name.to_str().unwrap()).unwrap()); if self.is_path_private(&relative_path) { log::debug!("detected private file: {relative_path:?}"); child_entry.is_private = true; } if self.settings.is_path_hidden(&relative_path) { log::debug!("detected hidden file: {relative_path:?}"); child_entry.is_hidden = true; } } new_entries.push(child_entry); } let mut state = self.state.lock().await; // Identify any subdirectories that should not be scanned. let mut job_ix = 0; for entry in &mut new_entries { state.reuse_entry_id(entry); if entry.is_dir() { if !self.should_scan_directory(&state, entry, ignore_stack.repo_root.is_some()) { log::debug!("defer scanning directory {:?}", entry.path); entry.kind = EntryKind::UnloadedDir; new_jobs[job_ix] = None; if !entry.is_ignored && !entry.is_external && state.snapshot.child_entries(&entry.path).next().is_some() { state.remove_path_from_snapshot_and_unwatch( &entry.path, self.watcher.as_ref(), true, ); } } job_ix += 1; } } state.populate_dir(job.path.clone(), new_entries, new_ignore); // For external entries, watch the canonical (resolved) path so OS-level // FS events on the real filesystem location are observed. The same // canonical path is stored in both `external_canonical_to_relative` // (for translating canonical-path FS events back to worktree-relative // paths) and `watched_dir_abs_paths_by_entry_id` (used by `remove_path` // to know which abs path to unwatch), so both cleanup paths agree on // the path the watcher was actually registered on. // // `canonicalize` is an async filesystem operation that may suspend, so // the lock must not be held across the await point below. drop(state); let watched_abs_path: Option> = if job.is_external { self.fs .canonicalize(job.abs_path.as_ref()) .await .ok() .map(|canonical| { let canonical: Arc = canonical.into(); self.watcher.add(&canonical).log_err(); canonical }) } else { self.watcher.add(job.abs_path.as_ref()).log_err(); Some(job.abs_path.clone()) }; let mut state = self.state.lock().await; if let Some(watched_abs_path) = &watched_abs_path { if job.is_external { state .snapshot .external_canonical_to_relative .insert(watched_abs_path.clone(), job.path.clone()); } if let Some(entry_id) = state .snapshot .entry_for_path(&job.path) .map(|entry| entry.id) { state .watched_dir_abs_paths_by_entry_id .insert(entry_id, watched_abs_path.clone()); } } for new_job in new_jobs.into_iter().flatten() { job.scan_queue .try_send(new_job) .expect("channel is unbounded"); } Ok(()) } /// All list arguments should be sorted before calling this function async fn reload_entries_for_paths( &self, root_abs_path: &SanitizedPath, root_canonical_path: &SanitizedPath, relative_paths: &[Arc], abs_paths: Vec, scan_queue_tx: Option>, ) { // grab metadata for all requested paths let metadata = futures::future::join_all( abs_paths .iter() .map(|abs_path| async move { let metadata = self.fs.metadata(abs_path).await?; if let Some(metadata) = metadata { let canonical_path = self.fs.canonicalize(abs_path).await?; // If we're on a case-insensitive filesystem (default on macOS), we want // to only ignore metadata for non-symlink files if their absolute-path matches // the canonical-path. // Because if not, this might be a case-only-renaming (`mv test.txt TEST.TXT`) // and we want to ignore the metadata for the old path (`test.txt`) so it's // treated as removed. if !self.fs_case_sensitive && !metadata.is_symlink { let canonical_file_name = canonical_path.file_name(); let file_name = abs_path.file_name(); if canonical_file_name != file_name { return Ok(None); } } anyhow::Ok(Some((metadata, SanitizedPath::new_arc(&canonical_path)))) } else { Ok(None) } }) .collect::>(), ) .await; let mut new_ancestor_repo = if self.track_git_repositories && relative_paths.iter().any(|path| path.is_empty()) { Some(discover_ancestor_git_repo(self.fs.clone(), &root_abs_path).await) } else { None }; let mut state = self.state.lock().await; let doing_recursive_update = scan_queue_tx.is_some(); // Remove any entries for paths that no longer exist or are being recursively // refreshed. Do this before adding any new entries, so that renames can be // detected regardless of the order of the paths. let mut paths_to_process = Vec::with_capacity(relative_paths.len()); for (path, metadata) in relative_paths.iter().zip(metadata.iter()) { let path_was_removed = matches!(metadata, Ok(None)); let removed_descendant_paths = if path_was_removed || doing_recursive_update { state.remove_path_from_snapshot(path, path_was_removed) } else { Vec::new() }; paths_to_process.push((path, metadata, removed_descendant_paths)); } for (path, metadata, removed_descendant_abs_paths) in paths_to_process { let abs_path: Arc = root_abs_path.join(path.as_std_path()).into(); match metadata { Ok(Some((metadata, canonical_path))) => { let ignore_stack = state .snapshot .ignore_stack_for_abs_path(&abs_path, metadata.is_dir, self.fs.as_ref()) .await; let is_external = !canonical_path.starts_with(&root_canonical_path); let entry_id = state.entry_id_for(self.next_entry_id.as_ref(), path, &metadata); let mut fs_entry = Entry::new( path.clone(), &metadata, entry_id, state.snapshot.root_char_bag, if metadata.is_symlink { Some(canonical_path.as_path().to_path_buf().into()) } else { None }, ); let is_dir = fs_entry.is_dir(); fs_entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, is_dir); fs_entry.is_external = is_external; fs_entry.is_private = self.is_path_private(path); fs_entry.is_always_included = self.settings.is_path_always_included(path, is_dir); fs_entry.is_hidden = self.settings.is_path_hidden(path); if let (Some(scan_queue_tx), true) = (&scan_queue_tx, is_dir) { if self.should_scan_directory( &state, &fs_entry, ignore_stack.repo_root.is_some(), ) || (self.track_git_repositories && fs_entry.path.is_empty() && abs_path.file_name() == Some(OsStr::new(DOT_GIT))) { state .enqueue_scan_dir( abs_path, &fs_entry, scan_queue_tx, self.fs.as_ref(), ) .await; } else { fs_entry.kind = EntryKind::UnloadedDir; } } state .insert_entry(fs_entry.clone(), self.fs.as_ref(), self.watcher.as_ref()) .await; if path.is_empty() && let Some((ignores, exclude, repo)) = new_ancestor_repo.take() { log::trace!("updating ancestor git repository"); state.snapshot.ignores_by_parent_abs_path.extend(ignores); if let Some((ancestor_dot_git, work_directory)) = repo { if let Some(exclude) = exclude { let work_directory_abs_path = state.snapshot.work_directory_abs_path(&work_directory); state .snapshot .repo_exclude_by_work_dir_abs_path .insert(work_directory_abs_path.into(), (exclude, false)); } state .insert_git_repository_for_path( work_directory, ancestor_dot_git.into(), self.fs.as_ref(), self.watcher.as_ref(), ) .await .log_err(); } } } Ok(None) => { self.remove_repo_path(path.clone(), &mut state.snapshot); state.unwatch_path( self.watcher.as_ref(), path, removed_descendant_abs_paths, false, ); } Err(err) => { log::error!("error reading file {abs_path:?} on event: {err:#}"); state.unwatch_path( self.watcher.as_ref(), path, removed_descendant_abs_paths, false, ); } } } util::extend_sorted( &mut state.changed_paths, relative_paths.iter().cloned(), usize::MAX, Ord::cmp, ); } fn remove_repo_path(&self, path: Arc, snapshot: &mut LocalSnapshot) -> Option<()> { if !path.components().any(|component| component == DOT_GIT) && let Some(local_repo) = snapshot.local_repo_for_work_directory_path(&path) { let id = local_repo.work_directory_id; log::debug!("remove repo path: {:?}", path); snapshot.git_repositories.remove(&id); return Some(()); } Some(()) } async fn update_ignore_statuses_for_paths( &self, scan_job_tx: Sender, prev_snapshot: LocalSnapshot, ignores_to_update: Vec<(Arc, IgnoreStack)>, ) { let (ignore_queue_tx, ignore_queue_rx) = async_channel::unbounded(); { for (parent_abs_path, ignore_stack) in ignores_to_update { ignore_queue_tx .send_blocking(UpdateIgnoreStatusJob { abs_path: parent_abs_path, ignore_stack, ignore_queue: ignore_queue_tx.clone(), scan_queue: scan_job_tx.clone(), }) .unwrap(); } } drop(ignore_queue_tx); self.executor .scoped(|scope| { for _ in 0..self.executor.num_cpus() { scope.spawn(async { loop { select_biased! { // Process any path refresh requests before moving on to process // the queue of ignore statuses. request = self.next_scan_request().fuse() => { let Ok(request) = request else { break }; if !self.process_scan_request(request, true).await { return; } } // Recursively process directories whose ignores have changed. job = ignore_queue_rx.recv().fuse() => { let Ok(job) = job else { break }; self.update_ignore_status(job, &prev_snapshot).await; } } } }); } }) .await; } async fn ignores_needing_update(&self) -> Vec> { let mut ignores_to_update = Vec::new(); let mut excludes_to_load: Vec<(Arc, PathBuf)> = Vec::new(); // First pass: collect updates and drop stale entries without awaiting. { let snapshot = &mut self.state.lock().await.snapshot; let abs_path = snapshot.abs_path.clone(); let mut repo_exclude_keys_to_remove: Vec> = Vec::new(); for (work_dir_abs_path, (_, needs_update)) in snapshot.repo_exclude_by_work_dir_abs_path.iter_mut() { let repository = snapshot .git_repositories .iter() .find(|(_, repo)| &repo.work_directory_abs_path == work_dir_abs_path); if *needs_update { *needs_update = false; if work_dir_abs_path.starts_with(abs_path.as_path()) { ignores_to_update.push(work_dir_abs_path.clone()); } else { ignores_to_update.push(abs_path.as_path().into()); } if let Some((_, repository)) = repository { let exclude_abs_path = repository.common_dir_abs_path.join(REPO_EXCLUDE); excludes_to_load.push((work_dir_abs_path.clone(), exclude_abs_path)); } } if repository.is_none() { repo_exclude_keys_to_remove.push(work_dir_abs_path.clone()); } } for key in repo_exclude_keys_to_remove { snapshot.repo_exclude_by_work_dir_abs_path.remove(&key); } snapshot .ignores_by_parent_abs_path .retain(|parent_abs_path, (_, needs_update)| { if let Ok(parent_path) = parent_abs_path.strip_prefix(abs_path.as_path()) && let Some(parent_path) = RelPath::new(&parent_path, PathStyle::local()).log_err() { if *needs_update { *needs_update = false; if snapshot.snapshot.entry_for_path(&parent_path).is_some() { ignores_to_update.push(parent_abs_path.clone()); } } let ignore_path = parent_path.join(RelPath::from_unix_str(GITIGNORE).unwrap()); if snapshot.snapshot.entry_for_path(&ignore_path).is_none() { return false; } } true }); } // Load gitignores asynchronously (outside the lock) let mut loaded_excludes: Vec<(Arc, Arc)> = Vec::new(); for (work_dir_abs_path, exclude_abs_path) in excludes_to_load { if let Ok(current_exclude) = build_gitignore_with_root(&exclude_abs_path, &work_dir_abs_path, self.fs.as_ref()) .await { loaded_excludes.push((work_dir_abs_path, Arc::new(current_exclude))); } } // Second pass: apply updates. if !loaded_excludes.is_empty() { let snapshot = &mut self.state.lock().await.snapshot; for (work_dir_abs_path, exclude) in loaded_excludes { if let Some((existing_exclude, _)) = snapshot .repo_exclude_by_work_dir_abs_path .get_mut(&work_dir_abs_path) { *existing_exclude = exclude; } } } ignores_to_update } async fn order_ignores(&self, mut ignores: Vec>) -> Vec<(Arc, IgnoreStack)> { let fs = self.fs.clone(); let snapshot = self.state.lock().await.snapshot.clone(); ignores.sort_unstable(); let mut ignores_to_update = ignores.into_iter().peekable(); let mut result = vec![]; while let Some(parent_abs_path) = ignores_to_update.next() { while ignores_to_update .peek() .map_or(false, |p| p.starts_with(&parent_abs_path)) { ignores_to_update.next().unwrap(); } let ignore_stack = snapshot .ignore_stack_for_abs_path(&parent_abs_path, true, fs.as_ref()) .await; result.push((parent_abs_path, ignore_stack)); } result } async fn update_ignore_status(&self, job: UpdateIgnoreStatusJob, snapshot: &LocalSnapshot) { log::trace!("update ignore status {:?}", job.abs_path); let mut ignore_stack = job.ignore_stack; if let Some((ignore, _)) = snapshot.ignores_by_parent_abs_path.get(&job.abs_path) { ignore_stack = ignore_stack.append(IgnoreKind::Gitignore(job.abs_path.clone()), ignore.clone()); } let mut entries_by_id_edits = Vec::new(); let mut entries_by_path_edits = Vec::new(); let Some(path) = job .abs_path .strip_prefix(snapshot.abs_path.as_path()) .map_err(|_| { anyhow::anyhow!( "Failed to strip prefix '{}' from path '{}'", snapshot.abs_path.as_path().display(), job.abs_path.display() ) }) .log_err() else { return; }; let Some(path) = RelPath::new(&path, PathStyle::local()).log_err() else { return; }; if let Ok(Some(_)) = self.fs.metadata(&job.abs_path.join(DOT_GIT)).await { ignore_stack.repo_root = Some(job.abs_path.clone()); ignore_stack.global_ignore_root = Some(job.abs_path.clone()); } for mut entry in snapshot.child_entries(&path).cloned() { let was_ignored = entry.is_ignored; let abs_path: Arc = snapshot.absolutize(&entry.path).into(); entry.is_ignored = ignore_stack.is_abs_path_ignored(&abs_path, entry.is_dir()); if entry.is_dir() { let child_ignore_stack = if entry.is_ignored { IgnoreStack::all() } else { ignore_stack.clone() }; // Scan any unloaded directories that became scannable: no longer // ignored, or newly inside a repository that exempts them from // the scan depth limit. if !entry.is_ignored && entry.kind.is_unloaded() && (was_ignored || ignore_stack.repo_root.is_some()) { let state = self.state.lock().await; if self.should_scan_directory(&state, &entry, ignore_stack.repo_root.is_some()) { state .enqueue_scan_dir( abs_path.clone(), &entry, &job.scan_queue, self.fs.as_ref(), ) .await; } } job.ignore_queue .send(UpdateIgnoreStatusJob { abs_path: abs_path.clone(), ignore_stack: child_ignore_stack, ignore_queue: job.ignore_queue.clone(), scan_queue: job.scan_queue.clone(), }) .await .unwrap(); } if entry.is_ignored != was_ignored { let mut path_entry = snapshot.entries_by_id.get(&entry.id, ()).unwrap().clone(); path_entry.scan_id = snapshot.scan_id; path_entry.is_ignored = entry.is_ignored; entries_by_id_edits.push(Edit::Insert(path_entry)); entries_by_path_edits.push(Edit::Insert(entry)); } } let state = &mut self.state.lock().await; for edit in &entries_by_path_edits { if let Edit::Insert(entry) = edit && let Err(ix) = state.changed_paths.binary_search(&entry.path) { state.changed_paths.insert(ix, entry.path.clone()); } } state .snapshot .entries_by_path .edit(entries_by_path_edits, ()); state.snapshot.entries_by_id.edit(entries_by_id_edits, ()); } async fn update_git_repositories(&self, dot_git_paths: Vec) -> Vec> { log::trace!("reloading repositories: {dot_git_paths:?}"); let mut state = self.state.lock().await; let scan_id = state.snapshot.scan_id; let mut affected_repo_roots = Vec::new(); for dot_git_dir in dot_git_paths { // Several repositories can share a git directory: a linked worktree's // commondir is the main checkout's `.git`, so a ref update there must // refresh every repository that reads from it. let existing_work_directory_ids = state .snapshot .git_repositories .iter() .filter_map(|(&work_directory_id, repo)| { let dot_git_dir = SanitizedPath::new(&dot_git_dir); if SanitizedPath::new(repo.common_dir_abs_path.as_ref()) == dot_git_dir || SanitizedPath::new(repo.repository_dir_abs_path.as_ref()) == dot_git_dir || SanitizedPath::new(repo.dot_git_abs_path.as_ref()) == dot_git_dir { Some(work_directory_id) } else { None } }) .collect::>(); if existing_work_directory_ids.is_empty() { let Ok(relative) = dot_git_dir.strip_prefix(state.snapshot.abs_path()) else { // A `.git` path outside the worktree root is not // ours to register. This happens legitimately when // `.git` is a gitfile pointing outside the worktree // (linked worktrees and submodules), and also when // a rescan of a linked worktree's commondir arrives // after the worktree's repository has already been // unregistered. continue; }; affected_repo_roots.push(dot_git_dir.parent().unwrap().into()); state .insert_git_repository( RelPath::new(relative, PathStyle::local()) .unwrap() .into_arc(), self.fs.as_ref(), self.watcher.as_ref(), ) .await; } else { for work_directory_id in existing_work_directory_ids { state .snapshot .git_repositories .update(&work_directory_id, |entry| { entry.git_dir_scan_id = scan_id; }); } } } // Remove any git repositories whose .git entry no longer exists. let snapshot = &mut state.snapshot; let mut ids_to_preserve = HashSet::default(); for (&work_directory_id, entry) in snapshot.git_repositories.iter() { let exists_in_snapshot = snapshot .entry_for_id(work_directory_id) .is_some_and(|entry| { snapshot .entry_for_path( &entry.path.join(RelPath::from_unix_str(DOT_GIT).unwrap()), ) .is_some() }); // Only drop a repository when we can positively confirm that its git // directory is gone. `metadata` returns `Ok(None)` for a confirmed // absence, but `Err(_)` for a transient failure (which can happen // under heavy filesystem churn). Treating an error as a deletion // makes the repository flap out of and back into the snapshot, // causing the GitStore to repeatedly tear it down and re-create it // with a fresh `RepositoryId`. So preserve the repository unless the // `.git` entry is confirmed absent. let dot_git_present = !matches!(self.fs.metadata(&entry.dot_git_abs_path).await, Ok(None)); if exists_in_snapshot || dot_git_present { ids_to_preserve.insert(work_directory_id); } } snapshot .git_repositories .retain(|work_directory_id, entry| { let preserve = ids_to_preserve.contains(work_directory_id); if !preserve { affected_repo_roots.push(entry.dot_git_abs_path.parent().unwrap().into()); snapshot .repo_exclude_by_work_dir_abs_path .remove(&entry.work_directory_abs_path); } preserve }); affected_repo_roots } async fn progress_timer(&self, running: bool) { if !running { return futures::future::pending().await; } #[cfg(feature = "test-support")] if self.fs.is_fake() { return self.executor.simulate_random_delay().await; } self.executor.timer(FS_WATCH_LATENCY).await } fn is_path_private(&self, path: &RelPath) -> bool { !self.share_private_files && self.settings.is_path_private(path) } fn should_scan_directory( &self, state: &BackgroundScannerState, entry: &Entry, in_repo: bool, ) -> bool { let beyond_scan_depth = !in_repo && is_beyond_scan_depth(self.settings.file_scan_depth, &entry.path); let scannable = state.scanning_enabled && (!entry.is_external || self.settings.scan_symlinks == settings::ScanSymlinksSetting::Always) && (!(entry.is_ignored || beyond_scan_depth) || entry.is_always_included); scannable || entry.path.file_name() == Some(DOT_GIT) || entry.path.file_name() == Some(local_settings_folder_name()) || entry.path.file_name() == Some(local_vscode_folder_name()) || state.scanned_dirs.contains(&entry.id) // If we've ever scanned it, keep scanning || state .paths_to_scan .iter() .any(|p| p.starts_with(&entry.path)) || state .path_prefixes_to_scan .iter() .any(|p| entry.path.starts_with(p)) } async fn next_scan_request(&self) -> Result { let mut request = self.scan_requests_rx.recv().await?; while let Ok(next_request) = self.scan_requests_rx.try_recv() { request.relative_paths.extend(next_request.relative_paths); request.done.extend(next_request.done); } Ok(request) } } async fn discover_ancestor_git_repo( fs: Arc, root_abs_path: &SanitizedPath, ) -> ( HashMap, (Arc, bool)>, Option>, Option<(PathBuf, WorkDirectory)>, ) { let mut exclude = None; let mut ignores = HashMap::default(); for (index, ancestor) in root_abs_path.as_path().ancestors().enumerate() { if index != 0 { if ancestor == paths::home_dir() { // Unless $HOME is itself the worktree root, don't consider it as a // containing git repository---expensive and likely unwanted. break; } else if let Ok(ignore) = build_gitignore(&ancestor.join(GITIGNORE), fs.as_ref()).await { ignores.insert(ancestor.into(), (ignore.into(), false)); } } let ancestor_dot_git = ancestor.join(DOT_GIT); log::trace!("considering ancestor: {ancestor_dot_git:?}"); // Check whether the directory or file called `.git` exists (in the // case of worktrees it's a file.) if fs .metadata(&ancestor_dot_git) .await .is_ok_and(|metadata| metadata.is_some()) { let dot_git_abs_path = if index != 0 { // We canonicalize, since the FS events use the canonicalized path. match fs.canonicalize(&ancestor_dot_git).await.log_err() { Some(path) => path, None => continue, } } else { ancestor_dot_git.clone() }; let dot_git_abs_path: Arc = dot_git_abs_path.as_path().into(); let (_, common_dir_abs_path) = discover_git_paths(&dot_git_abs_path, fs.as_ref()).await; let repo_exclude_abs_path = common_dir_abs_path.join(REPO_EXCLUDE); if let Ok(repo_exclude) = build_gitignore_with_root(&repo_exclude_abs_path, ancestor, fs.as_ref()).await { exclude = Some(Arc::new(repo_exclude)); } if index != 0 { let location_in_repo = root_abs_path .as_path() .strip_prefix(ancestor) .unwrap() .into(); log::info!("inserting parent git repo for this worktree: {location_in_repo:?}"); // We associate the external git repo with our root folder and // also mark where in the git repo the root folder is located. return ( ignores, exclude, Some(( dot_git_abs_path.as_ref().into(), WorkDirectory::AboveProject { absolute_path: ancestor.into(), location_in_repo, }, )), ); } break; } } (ignores, exclude, None) } fn merge_event_roots(changed_paths: &[Arc], event_roots: &[EventRoot]) -> Vec { let mut merged_event_roots = Vec::with_capacity(changed_paths.len() + event_roots.len()); let mut changed_paths = changed_paths.iter().peekable(); let mut event_roots = event_roots.iter().peekable(); while let (Some(path), Some(event_root)) = (changed_paths.peek(), event_roots.peek()) { match path.cmp(&&event_root.path) { Ordering::Less => { merged_event_roots.push(EventRoot { path: (*changed_paths.next().expect("peeked changed path")).clone(), was_rescanned: false, }); } Ordering::Equal => { merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone()); changed_paths.next(); } Ordering::Greater => { merged_event_roots.push((*event_roots.next().expect("peeked event root")).clone()); } } } merged_event_roots.extend(changed_paths.map(|path| EventRoot { path: path.clone(), was_rescanned: false, })); merged_event_roots.extend(event_roots.cloned()); merged_event_roots } fn build_diff( phase: BackgroundScannerPhase, old_snapshot: &Snapshot, new_snapshot: &Snapshot, event_roots: &[EventRoot], ) -> UpdatedEntriesSet { use BackgroundScannerPhase::*; use PathChange::{Added, AddedOrUpdated, Loaded, Removed, Updated}; // Identify which paths have changed. Use the known set of changed // parent paths to optimize the search. let mut changes = Vec::new(); let mut old_paths = old_snapshot.entries_by_path.cursor::(()); let mut new_paths = new_snapshot.entries_by_path.cursor::(()); let mut last_newly_loaded_dir_path = None; old_paths.next(); new_paths.next(); for event_root in event_roots { let path = PathKey(event_root.path.clone()); if old_paths.item().is_some_and(|e| e.path < path.0) { old_paths.seek_forward(&path, Bias::Left); } if new_paths.item().is_some_and(|e| e.path < path.0) { new_paths.seek_forward(&path, Bias::Left); } loop { match (old_paths.item(), new_paths.item()) { (Some(old_entry), Some(new_entry)) => { if old_entry.path > path.0 && new_entry.path > path.0 && !old_entry.path.starts_with(&path.0) && !new_entry.path.starts_with(&path.0) { break; } match Ord::cmp(&old_entry.path, &new_entry.path) { Ordering::Less => { changes.push((old_entry.path.clone(), old_entry.id, Removed)); old_paths.next(); } Ordering::Equal => { if phase == EventsReceivedDuringInitialScan { if old_entry.id != new_entry.id { changes.push((old_entry.path.clone(), old_entry.id, Removed)); } // If the worktree was not fully initialized when this event was generated, // we can't know whether this entry was added during the scan or whether // it was merely updated. changes.push(( new_entry.path.clone(), new_entry.id, AddedOrUpdated, )); } else if old_entry.id != new_entry.id { changes.push((old_entry.path.clone(), old_entry.id, Removed)); changes.push((new_entry.path.clone(), new_entry.id, Added)); } else if old_entry != new_entry { if old_entry.kind.is_unloaded() { last_newly_loaded_dir_path = Some(&new_entry.path); changes.push((new_entry.path.clone(), new_entry.id, Loaded)); } else { changes.push((new_entry.path.clone(), new_entry.id, Updated)); } } else if event_root.was_rescanned { changes.push((new_entry.path.clone(), new_entry.id, Updated)); } old_paths.next(); new_paths.next(); } Ordering::Greater => { let is_newly_loaded = phase == InitialScan || last_newly_loaded_dir_path .as_ref() .is_some_and(|dir| new_entry.path.starts_with(dir)); changes.push(( new_entry.path.clone(), new_entry.id, if is_newly_loaded { Loaded } else { Added }, )); new_paths.next(); } } } (Some(old_entry), None) => { changes.push((old_entry.path.clone(), old_entry.id, Removed)); old_paths.next(); } (None, Some(new_entry)) => { let is_newly_loaded = phase == InitialScan || last_newly_loaded_dir_path .as_ref() .is_some_and(|dir| new_entry.path.starts_with(dir)); changes.push(( new_entry.path.clone(), new_entry.id, if is_newly_loaded { Loaded } else { Added }, )); new_paths.next(); } (None, None) => break, } } } changes.into() } fn is_beyond_scan_depth(file_scan_depth: Option, path: &RelPath) -> bool { file_scan_depth.is_some_and(|depth| path.components().count() >= depth as usize) } fn swap_to_front(child_paths: &mut Vec, file: &str) { let position = child_paths .iter() .position(|path| path.file_name().unwrap() == file); if let Some(position) = position { let temp = child_paths.remove(position); child_paths.insert(0, temp); } } fn char_bag_for_path(root_char_bag: CharBag, path: &RelPath) -> CharBag { let mut result = root_char_bag; result.extend(path.as_unix_str().chars().map(|c| c.to_ascii_lowercase())); result } #[derive(Debug)] struct ScanJob { abs_path: Arc, path: Arc, ignore_stack: IgnoreStack, scan_queue: Sender, ancestor_inodes: TreeSet, is_external: bool, } struct UpdateIgnoreStatusJob { abs_path: Arc, ignore_stack: IgnoreStack, ignore_queue: Sender, scan_queue: Sender, } pub trait WorktreeModelHandle { #[cfg(feature = "test-support")] fn flush_fs_events<'a>( &self, cx: &'a mut gpui::TestAppContext, ) -> futures::future::LocalBoxFuture<'a, ()>; #[cfg(feature = "test-support")] fn flush_fs_events_in_root_git_repository<'a>( &self, cx: &'a mut gpui::TestAppContext, ) -> futures::future::LocalBoxFuture<'a, ()>; } impl WorktreeModelHandle for Entity { // When the worktree's FS event stream sometimes delivers "redundant" events for FS changes that // occurred before the worktree was constructed. These events can cause the worktree to perform // extra directory scans, and emit extra scan-state notifications. // // This function mutates the worktree's directory and waits for those mutations to be picked up, // to ensure that all redundant FS events have already been processed. #[cfg(feature = "test-support")] fn flush_fs_events<'a>( &self, cx: &'a mut gpui::TestAppContext, ) -> futures::future::LocalBoxFuture<'a, ()> { let file_name = "fs-event-sentinel"; let tree = self.clone(); let (fs, root_path) = self.read_with(cx, |tree, _| { let tree = tree.as_local().unwrap(); (tree.fs.clone(), tree.abs_path.clone()) }); async move { // Subscribe to events BEFORE creating the file to avoid race condition // where events fire before subscription is set up let mut events = cx.events(&tree); fs.create_file(&root_path.join(file_name), Default::default()) .await .unwrap(); // Check if condition is already met before waiting for events let file_exists = || { tree.read_with(cx, |tree, _| { tree.entry_for_path(RelPath::from_unix_str(file_name).unwrap()) .is_some() }) }; // Use select to avoid blocking indefinitely if events are delayed let mut ticks = 0; while !file_exists() { futures::select_biased! { _ = events.next() => {} _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => { ticks += 1; if ticks % SENTINEL_RETRY_TICKS == 0 { retouch_sentinel(fs.as_ref(), &root_path.join(file_name)).await; } } } } fs.remove_file(&root_path.join(file_name), Default::default()) .await .unwrap(); // Check if condition is already met before waiting for events let file_gone = || { tree.read_with(cx, |tree, _| { tree.entry_for_path(RelPath::from_unix_str(file_name).unwrap()) .is_none() }) }; // Use select to avoid blocking indefinitely if events are delayed let mut ticks = 0; while !file_gone() { futures::select_biased! { _ = events.next() => {} _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => { ticks += 1; if ticks % SENTINEL_RETRY_TICKS == 0 { retouch_and_remove_sentinel(fs.as_ref(), &root_path.join(file_name)).await; } } } } cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete()) .await; } .boxed_local() } // This function is similar to flush_fs_events, except that it waits for events to be flushed in // the .git folder of the root repository. // The reason for its existence is that a repository's .git folder might live *outside* of the // worktree and thus its FS events might go through a different path. // In order to flush those, we need to create artificial events in the .git folder and wait // for the repository to be reloaded. #[cfg(feature = "test-support")] fn flush_fs_events_in_root_git_repository<'a>( &self, cx: &'a mut gpui::TestAppContext, ) -> futures::future::LocalBoxFuture<'a, ()> { let file_name = "fs-event-sentinel"; let tree = self.clone(); let (fs, root_path, mut git_dir_scan_id) = self.read_with(cx, |tree, _| { let tree = tree.as_local().unwrap(); let local_repo_entry = tree .git_repositories .values() .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone()) .unwrap(); ( tree.fs.clone(), local_repo_entry.common_dir_abs_path.clone(), local_repo_entry.git_dir_scan_id, ) }); let scan_id_increased = |tree: &mut Worktree, git_dir_scan_id: &mut usize| { let tree = tree.as_local().unwrap(); // let repository = tree.repositories.first().unwrap(); let local_repo_entry = tree .git_repositories .values() .min_by_key(|local_repo_entry| local_repo_entry.work_directory.clone()) .unwrap(); if local_repo_entry.git_dir_scan_id > *git_dir_scan_id { *git_dir_scan_id = local_repo_entry.git_dir_scan_id; true } else { false } }; async move { // Subscribe to events BEFORE creating the file to avoid race condition // where events fire before subscription is set up let mut events = cx.events(&tree); fs.create_file(&root_path.join(file_name), Default::default()) .await .unwrap(); // Use select to avoid blocking indefinitely if events are delayed let mut ticks = 0; while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) { futures::select_biased! { _ = events.next() => {} _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => { ticks += 1; if ticks % SENTINEL_RETRY_TICKS == 0 { retouch_sentinel(fs.as_ref(), &root_path.join(file_name)).await; } } } } fs.remove_file(&root_path.join(file_name), Default::default()) .await .unwrap(); // Use select to avoid blocking indefinitely if events are delayed let mut ticks = 0; while !tree.update(cx, |tree, _| scan_id_increased(tree, &mut git_dir_scan_id)) { futures::select_biased! { _ = events.next() => {} _ = futures::FutureExt::fuse(cx.background_executor.timer(std::time::Duration::from_millis(10))) => { ticks += 1; if ticks % SENTINEL_RETRY_TICKS == 0 { retouch_and_remove_sentinel(fs.as_ref(), &root_path.join(file_name)).await; } } } } cx.update(|cx| tree.read(cx).as_local().unwrap().scan_complete()) .await; } .boxed_local() } } #[cfg(feature = "test-support")] const SENTINEL_RETRY_TICKS: usize = 10; // On macOS an FS event can rarely be dropped while notify's shared FSEventStream // is being swapped out by a concurrent watch/unwatch: events already queued for // the old stream's callback are discarded when it is stopped. A dropped sentinel // event would park the test forever, so touch the sentinel again to emit a fresh // event on the current stream. #[cfg(feature = "test-support")] async fn retouch_sentinel(fs: &dyn Fs, abs_path: &std::path::Path) { fs.create_file( abs_path, fs::CreateOptions { overwrite: true, ignore_if_exists: false, }, ) .await .unwrap(); } #[cfg(feature = "test-support")] async fn retouch_and_remove_sentinel(fs: &dyn Fs, abs_path: &std::path::Path) { retouch_sentinel(fs, abs_path).await; fs.remove_file( abs_path, RemoveOptions { recursive: false, ignore_if_not_exists: true, }, ) .await .unwrap(); } #[derive(Clone, Debug)] struct TraversalProgress<'a> { max_path: &'a RelPath, count: usize, non_ignored_count: usize, file_count: usize, non_ignored_file_count: usize, } impl TraversalProgress<'_> { fn count(&self, include_files: bool, include_dirs: bool, include_ignored: bool) -> usize { match (include_files, include_dirs, include_ignored) { (true, true, true) => self.count, (true, true, false) => self.non_ignored_count, (true, false, true) => self.file_count, (true, false, false) => self.non_ignored_file_count, (false, true, true) => self.count - self.file_count, (false, true, false) => self.non_ignored_count - self.non_ignored_file_count, (false, false, _) => 0, } } } impl<'a> sum_tree::Dimension<'a, EntrySummary> for TraversalProgress<'a> { fn zero(_cx: ()) -> Self { Default::default() } fn add_summary(&mut self, summary: &'a EntrySummary, _: ()) { self.max_path = summary.max_path.as_ref(); self.count += summary.count; self.non_ignored_count += summary.non_ignored_count; self.file_count += summary.file_count; self.non_ignored_file_count += summary.non_ignored_file_count; } } impl Default for TraversalProgress<'_> { fn default() -> Self { Self { max_path: RelPath::empty(), count: 0, non_ignored_count: 0, file_count: 0, non_ignored_file_count: 0, } } } #[derive(Debug)] pub struct Traversal<'a> { snapshot: &'a Snapshot, cursor: sum_tree::Cursor<'a, 'static, Entry, TraversalProgress<'a>>, include_ignored: bool, include_files: bool, include_dirs: bool, } impl<'a> Traversal<'a> { fn new( snapshot: &'a Snapshot, include_files: bool, include_dirs: bool, include_ignored: bool, start_path: &RelPath, ) -> Self { let mut cursor = snapshot.entries_by_path.cursor(()); cursor.seek(&TraversalTarget::path(start_path), Bias::Left); let mut traversal = Self { snapshot, cursor, include_files, include_dirs, include_ignored, }; if traversal.end_offset() == traversal.start_offset() { traversal.next(); } traversal } pub fn advance(&mut self) -> bool { self.advance_by(1) } pub fn advance_by(&mut self, count: usize) -> bool { self.cursor.seek_forward( &TraversalTarget::Count { count: self.end_offset() + count, include_dirs: self.include_dirs, include_files: self.include_files, include_ignored: self.include_ignored, }, Bias::Left, ) } pub fn advance_to_sibling(&mut self) -> bool { while let Some(entry) = self.cursor.item() { self.cursor .seek_forward(&TraversalTarget::successor(&entry.path), Bias::Left); if let Some(entry) = self.cursor.item() && (self.include_files || !entry.is_file()) && (self.include_dirs || !entry.is_dir()) && (self.include_ignored || !entry.is_ignored || entry.is_always_included) { return true; } } false } pub fn back_to_parent(&mut self) -> bool { let Some(parent_path) = self.cursor.item().and_then(|entry| entry.path.parent()) else { return false; }; self.cursor .seek(&TraversalTarget::path(parent_path), Bias::Left) } pub fn entry(&self) -> Option<&'a Entry> { self.cursor.item() } pub fn snapshot(&self) -> &'a Snapshot { self.snapshot } pub fn start_offset(&self) -> usize { self.cursor .start() .count(self.include_files, self.include_dirs, self.include_ignored) } pub fn end_offset(&self) -> usize { self.cursor .end() .count(self.include_files, self.include_dirs, self.include_ignored) } } impl<'a> Iterator for Traversal<'a> { type Item = &'a Entry; fn next(&mut self) -> Option { if let Some(item) = self.entry() { self.advance(); Some(item) } else { None } } } #[derive(Debug, Clone, Copy)] pub enum PathTarget<'a> { Path(&'a RelPath), Successor(&'a RelPath), } impl PathTarget<'_> { fn cmp_path(&self, other: &RelPath) -> Ordering { match self { PathTarget::Path(path) => path.cmp(&other), PathTarget::Successor(path) => { if other.starts_with(path) { Ordering::Greater } else { Ordering::Equal } } } } } impl<'a, S: Summary> SeekTarget<'a, PathSummary, PathProgress<'a>> for PathTarget<'_> { fn cmp(&self, cursor_location: &PathProgress<'a>, _: S::Context<'_>) -> Ordering { self.cmp_path(cursor_location.max_path) } } impl<'a, S: Summary> SeekTarget<'a, PathSummary, TraversalProgress<'a>> for PathTarget<'_> { fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: S::Context<'_>) -> Ordering { self.cmp_path(cursor_location.max_path) } } #[derive(Debug)] enum TraversalTarget<'a> { Path(PathTarget<'a>), Count { count: usize, include_files: bool, include_ignored: bool, include_dirs: bool, }, } impl<'a> TraversalTarget<'a> { fn path(path: &'a RelPath) -> Self { Self::Path(PathTarget::Path(path)) } fn successor(path: &'a RelPath) -> Self { Self::Path(PathTarget::Successor(path)) } fn cmp_progress(&self, progress: &TraversalProgress) -> Ordering { match self { TraversalTarget::Path(path) => path.cmp_path(progress.max_path), TraversalTarget::Count { count, include_files, include_dirs, include_ignored, } => Ord::cmp( count, &progress.count(*include_files, *include_dirs, *include_ignored), ), } } } impl<'a> SeekTarget<'a, EntrySummary, TraversalProgress<'a>> for TraversalTarget<'_> { fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering { self.cmp_progress(cursor_location) } } impl<'a> SeekTarget<'a, PathSummary, TraversalProgress<'a>> for TraversalTarget<'_> { fn cmp(&self, cursor_location: &TraversalProgress<'a>, _: ()) -> Ordering { self.cmp_progress(cursor_location) } } pub struct ChildEntriesOptions { pub include_files: bool, pub include_dirs: bool, pub include_ignored: bool, } pub struct ChildEntriesIter<'a> { parent_path: &'a RelPath, traversal: Traversal<'a>, } impl<'a> Iterator for ChildEntriesIter<'a> { type Item = &'a Entry; fn next(&mut self) -> Option { if let Some(item) = self.traversal.entry() && item.path.starts_with(self.parent_path) { self.traversal.advance_to_sibling(); return Some(item); } None } } impl<'a> From<&'a Entry> for proto::Entry { fn from(entry: &'a Entry) -> Self { Self { id: entry.id.to_proto(), is_dir: entry.is_dir(), path: entry.path.as_ref().as_unix_str().to_owned(), inode: entry.inode, mtime: entry.mtime.map(|time| time.into()), is_ignored: entry.is_ignored, is_hidden: entry.is_hidden, is_external: entry.is_external, is_fifo: entry.is_fifo, size: Some(entry.size), canonical_path: entry .canonical_path .as_ref() .map(|path| path.to_string_lossy().into_owned()), is_unloaded: entry.kind == EntryKind::UnloadedDir, } } } impl TryFrom<(&CharBag, &PathMatcher, proto::Entry)> for Entry { type Error = anyhow::Error; fn try_from( (root_char_bag, always_included, entry): (&CharBag, &PathMatcher, proto::Entry), ) -> Result { let kind = if entry.is_dir { if entry.is_unloaded { EntryKind::UnloadedDir } else { EntryKind::Dir } } else { EntryKind::File }; let path = RelPath::from_unix_str(&entry.path) .context("invalid relative path in proto message")?; let char_bag = char_bag_for_path(*root_char_bag, &path); let is_always_included = always_included.is_match(&path); Ok(Entry { id: ProjectEntryId::from_proto(entry.id), kind, path: path.into(), inode: entry.inode, mtime: entry.mtime.map(|time| time.into()), size: entry.size.unwrap_or(0), canonical_path: entry .canonical_path .map(|path_string| Arc::from(PathBuf::from(path_string))), is_ignored: entry.is_ignored, is_hidden: entry.is_hidden, is_always_included, is_external: entry.is_external, is_private: false, char_bag, is_fifo: entry.is_fifo, }) } } #[derive(Clone, Copy, Debug, Default, Hash, PartialEq, Eq, PartialOrd, Ord)] pub struct ProjectEntryId(usize); impl ProjectEntryId { pub const MAX: Self = Self(usize::MAX); pub const MIN: Self = Self(usize::MIN); pub fn new(counter: &AtomicUsize) -> Self { Self(counter.fetch_add(1, SeqCst)) } pub fn from_proto(id: u64) -> Self { Self(id as usize) } pub fn to_proto(self) -> u64 { self.0 as u64 } pub fn from_usize(id: usize) -> Self { ProjectEntryId(id) } pub fn to_usize(self) -> usize { self.0 } } #[cfg(feature = "test-support")] impl CreatedEntry { pub fn into_included(self) -> Option { match self { CreatedEntry::Included(entry) => Some(entry), CreatedEntry::Excluded { .. } => None, } } } fn parse_gitfile(content: &str) -> anyhow::Result<&Path> { let path = content .strip_prefix("gitdir:") .with_context(|| format!("parsing gitfile content {content:?}"))?; Ok(Path::new(path.trim())) } fn resolve_gitfile_path(dot_git_abs_path: &Path, gitfile_path: &Path) -> PathBuf { if gitfile_path.is_absolute() { gitfile_path.into() } else { dot_git_abs_path .parent() .unwrap_or_else(|| Path::new("")) .join(gitfile_path) } } fn resolve_commondir_path(repository_dir_abs_path: &Path, commondir_path: &str) -> PathBuf { let commondir_path = Path::new(commondir_path.trim()); if commondir_path.is_absolute() { commondir_path.into() } else { repository_dir_abs_path.join(commondir_path) } } pub async fn discover_root_repo_common_dir(root_abs_path: &Path, fs: &dyn Fs) -> Option> { discover_root_repo_metadata(root_abs_path, fs) .await .map(|(common_dir, _)| common_dir) } async fn discover_root_repo_metadata( root_abs_path: &Path, fs: &dyn Fs, ) -> Option<(Arc, bool)> { let root_dot_git = root_abs_path.join(DOT_GIT); if !fs.metadata(&root_dot_git).await.is_ok_and(|m| m.is_some()) { return None; } let dot_git_path: Arc = root_dot_git.into(); let (repository_dir, common_dir) = discover_git_paths(&dot_git_path, fs).await; let is_linked_worktree = repository_dir != common_dir; Some((common_dir, is_linked_worktree)) } async fn discover_git_paths(dot_git_abs_path: &Arc, fs: &dyn Fs) -> (Arc, Arc) { let mut repository_dir_abs_path = dot_git_abs_path.clone(); let mut common_dir_abs_path = dot_git_abs_path.clone(); if let Some(path) = fs .load(dot_git_abs_path) .await .ok() .as_ref() .and_then(|contents| parse_gitfile(contents).log_err()) { let path = resolve_gitfile_path(dot_git_abs_path, path); if let Some(path) = fs.canonicalize(&path).await.log_err() { repository_dir_abs_path = Path::new(&path).into(); common_dir_abs_path = repository_dir_abs_path.clone(); if let Some(commondir_contents) = fs.load(&path.join("commondir")).await.ok() && let Some(commondir_path) = fs .canonicalize(&resolve_commondir_path(&path, &commondir_contents)) .await .log_err() { common_dir_abs_path = commondir_path.as_path().into(); } } }; (repository_dir_abs_path, common_dir_abs_path) } struct NullWatcher; impl fs::Watcher for NullWatcher { fn add(&self, _path: &Path) -> Result<()> { Ok(()) } fn remove(&self, _path: &Path) -> Result<()> { Ok(()) } } /// Reads the beginning of `file` to determine its kind and encoding, returning /// the bytes consumed and whether the file ended within them. fn read_file_header(file: &mut dyn Read, abs_path: &Path) -> Result<(Vec, bool)> { let mut header = Vec::with_capacity(FILE_ANALYSIS_BYTES); let mut buf = [0u8; FILE_ANALYSIS_BYTES]; let mut reached_eof = false; while header.len() < FILE_ANALYSIS_BYTES { let n = file .read(&mut buf) .with_context(|| format!("reading bytes of the file {abs_path:?}"))?; if n == 0 { reached_eof = true; break; } header.extend_from_slice(&buf[..n]); } Ok((header, reached_eof)) } const STREAM_BLOCK_BYTES: usize = 1024 * 1024; async fn read_file_to_end( file: &mut (dyn Read + Send), content: &mut Vec, abs_path: &Path, ) -> Result<()> { let mut buf = vec![0u8; STREAM_BLOCK_BYTES]; loop { let mut block_len = 0; while block_len < buf.len() { let n = file .read(&mut buf[block_len..]) .with_context(|| format!("reading remaining bytes of the file {abs_path:?}"))?; if n == 0 { break; } block_len += n; } if block_len == 0 { break; } content.extend_from_slice(&buf[..block_len]); if block_len < buf.len() { break; } yield_now().await; } Ok(()) } pub async fn decode_file_text( fs: &dyn Fs, abs_path: &Path, ) -> Result<(String, &'static Encoding, bool)> { let mut file = fs .open_sync(&abs_path) .await .with_context(|| format!("opening file {abs_path:?}"))?; let (file_first_bytes, reached_eof) = read_file_header(&mut *file, abs_path)?; let (_, byte_content) = decode_byte_header(&file_first_bytes); anyhow::ensure!( byte_content != ByteContent::Binary, "Binary files are not supported" ); // If the file is eligible for opening, read the rest of the file. let mut content = file_first_bytes; if !reached_eof { read_file_to_end(&mut *file, &mut content, abs_path).await?; } let decoded = decode_text(content)?; Ok((decoded.text, decoded.encoding, decoded.has_bom)) } /// Reads and decodes a file straight into a [`Rope`]. /// The returned rope has already had its line endings normalized, the /// [`LineEnding`] detected before normalizing is returned alongside it. pub async fn decode_file_text_to_rope( fs: &dyn Fs, abs_path: &Path, ) -> Result<(Rope, LineEnding, &'static Encoding, bool)> { let mut file = fs .open_sync(abs_path) .await .with_context(|| format!("opening file {abs_path:?}"))?; let (prefix, reached_eof) = read_file_header(&mut *file, abs_path)?; let (bom_encoding, byte_content) = decode_byte_header(&prefix); anyhow::ensure!( byte_content != ByteContent::Binary, "Binary files are not supported" ); // Only BOM-less, non-UTF-16 files are candidates for streaming: everything // else needs the whole byte buffer in hand to decode or to detect encoding. if bom_encoding.is_none() && byte_content == ByteContent::Unknown && let Some((rope, line_ending)) = stream_utf8_into_rope(&mut *file, prefix, reached_eof, abs_path).await? { return Ok((rope, line_ending, encoding_rs::UTF_8, false)); } // Not plain UTF-8 after all. Re-read the file and decode it all at once. let (mut text, encoding, has_bom) = decode_file_text(fs, abs_path).await?; let line_ending = LineEnding::detect(&text); LineEnding::normalize(&mut text); Ok((Rope::from(text), line_ending, encoding, has_bom)) } /// Streams a presumed-UTF-8 file into a [`Rope`], normalizing line endings as it /// goes. /// /// Returns `None` if the file turns out not to be plain UTF-8, in which case the /// caller re-reads it and decodes it the slow way. `prefix` is the portion of /// the file already consumed from `file` for encoding detection. async fn stream_utf8_into_rope( file: &mut (dyn Read + Send), prefix: Vec, reached_eof: bool, abs_path: &Path, ) -> Result> { let mut rope = Rope::new(); let mut line_ending = None; let mut scratch = String::new(); let mut pending = prefix; let mut buf = vec![0u8; STREAM_BLOCK_BYTES]; let mut eof = reached_eof; loop { // Fill a whole block before decoding, so that each `Rope::push` gets a // slice large enough to build its chunks in parallel. while !eof && pending.len() < STREAM_BLOCK_BYTES { let n = file .read(&mut buf) .with_context(|| format!("reading bytes of the file {abs_path:?}"))?; if n == 0 { eof = true; } else { pending.extend_from_slice(&buf[..n]); } } // Decode as much of `pending` as forms complete UTF-8. let valid_len = match std::str::from_utf8(&pending) { Ok(_) => pending.len(), // A multi-byte character straddling the block boundary; the rest of // it arrives with the next block. Err(e) if e.error_len().is_none() && !eof => e.valid_up_to(), // Genuinely not UTF-8, or truncated at EOF: fall back. Err(_) => return Ok(None), }; // Hold back a trailing carriage return until we know what follows it. let emit_len = if !eof && valid_len > 0 && pending[valid_len - 1] == b'\r' { valid_len - 1 } else { valid_len }; let text = std::str::from_utf8(&pending[..emit_len]) .expect("a prefix of validated UTF-8 is itself valid UTF-8"); // ISO-2022-JP and friends are valid UTF-8 but carry escape sequences, so // they need the full-file encoding detector rather than this fast path. if text.contains('\x1b') { return Ok(None); } // `LineEnding::detect` only inspects the first 1000 bytes, so the first // block gives the same answer the whole file would. if line_ending.is_none() && !text.is_empty() { line_ending = Some(LineEnding::detect(text)); } push_normalized(&mut rope, text, &mut scratch); pending.drain(..emit_len); if eof { break; } yield_now().await; } // At EOF everything should have been consumed. Anything left over is a // truncated multi-byte sequence, which means this is not valid UTF-8. if !pending.is_empty() { return Ok(None); } Ok(Some((rope, line_ending.unwrap_or_default()))) } /// Appends `text` to `rope`, rewriting CRLF and lone CR as LF, matching /// [`LineEnding::normalize`]. `scratch` is reused across calls. fn push_normalized(rope: &mut Rope, text: &str, scratch: &mut String) { if !text.contains('\r') { rope.push(text); return; } scratch.clear(); scratch.reserve(text.len()); let bytes = text.as_bytes(); let mut start = 0; let mut ix = 0; while ix < bytes.len() { if bytes[ix] == b'\r' { scratch.push_str(&text[start..ix]); scratch.push('\n'); ix += if bytes.get(ix + 1) == Some(&b'\n') { 2 } else { 1 }; start = ix; } else { ix += 1; } } scratch.push_str(&text[start..]); rope.push(scratch); } pub fn decode_byte_header(prefix: &[u8]) -> (Option<&'static Encoding>, ByteContent) { if let Some((encoding, _bom_len)) = Encoding::for_bom(prefix) { return (Some(encoding), ByteContent::Unknown); } (None, analyze_byte_content(prefix)) } #[cfg(test)] mod tests { use super::*; /// Streams `bytes` the way `decode_file_text_to_rope` would, returning the /// decoded text and detected line ending, or `None` if the fast path bailed. async fn stream(bytes: &[u8]) -> Option<(String, LineEnding)> { let mut reader = std::io::Cursor::new(bytes.to_vec()); stream_utf8_into_rope(&mut reader, Vec::new(), false, Path::new("test")) .await .unwrap() .map(|(rope, line_ending)| (rope.to_string(), line_ending)) } #[test] fn test_stream_utf8_yields_between_blocks() { let mut reader = std::io::Cursor::new(vec![b'a'; STREAM_BLOCK_BYTES * 2]); assert!( stream_utf8_into_rope(&mut reader, Vec::new(), false, Path::new("test")) .now_or_never() .is_none() ); assert_eq!(reader.position(), STREAM_BLOCK_BYTES as u64); } #[test] fn test_file_reading_yields_between_blocks() { let mut reader = std::io::Cursor::new(vec![b'a'; STREAM_BLOCK_BYTES * 2]); let mut content = Vec::new(); assert!( read_file_to_end(&mut reader, &mut content, Path::new("test")) .now_or_never() .is_none() ); assert_eq!(reader.position(), STREAM_BLOCK_BYTES as u64); assert_eq!(content.len(), STREAM_BLOCK_BYTES); } #[gpui::test] async fn test_stream_utf8_normalizes_line_endings() { let crlf = "one\r\ntwo\r\nthree\r\n".repeat(40); let (text, line_ending) = stream(crlf.as_bytes()).await.unwrap(); assert_eq!(text, crlf.replace("\r\n", "\n")); assert_eq!(line_ending, LineEnding::Windows); let cr = "one\rtwo\rthree\r".repeat(40); assert_eq!( stream(cr.as_bytes()).await.unwrap().0, cr.replace('\r', "\n") ); } #[gpui::test] async fn test_stream_utf8_block_boundaries() { // A carriage return landing on the last byte of a block, with and // without its newline arriving in the next one. for (suffix, expected) in [("\r\ntail\n", "\ntail\n"), ("\rtail", "\ntail")] { let filler = "a".repeat(STREAM_BLOCK_BYTES - 1); let source = format!("{filler}{suffix}"); assert_eq!( stream(source.as_bytes()).await.unwrap().0, format!("{filler}{expected}"), "suffix = {suffix:?}" ); } // Multi-byte characters straddling the boundary at every split point. for ch in ['\u{20ac}', '\u{1f600}'] { for split in 1..=ch.len_utf8() { let filler = "a".repeat(STREAM_BLOCK_BYTES - split); let source = format!("{filler}{ch}tail"); assert_eq!( stream(source.as_bytes()).await.unwrap().0, source, "ch = {ch:?}, split = {split}" ); } } } #[gpui::test] async fn test_stream_utf8_falls_back_on_non_utf8() { // Each of these must bail so the caller re-reads and decodes the slow // way, rather than silently mangling the file. assert_eq!(stream(b"hello \xff\xfeA").await, None, "invalid utf-8"); assert_eq!(stream(b"hello \xe2\x82").await, None, "truncated at eof"); assert_eq!(stream(b"plain \x1b$B text").await, None, "iso-2022 escape"); } /// reproduction of issue #50785 fn build_pcm16_wav_bytes() -> Vec { let header: Vec = vec![ /* RIFF header */ 0x52, 0x49, 0x46, 0x46, // "RIFF" 0xc6, 0xcf, 0x00, 0x00, // file size: 8 0x57, 0x41, 0x56, 0x45, // "WAVE" /* fmt chunk */ 0x66, 0x6d, 0x74, 0x20, // "fmt " 0x10, 0x00, 0x00, 0x00, // chunk size: 16 0x01, 0x00, // format: PCM (1) 0x01, 0x00, // channels: 1 (mono) 0x80, 0x3e, 0x00, 0x00, // sample rate: 16000 0x00, 0x7d, 0x00, 0x00, // byte rate: 32000 0x02, 0x00, // block align: 2 0x10, 0x00, // bits per sample: 16 /* LIST chunk */ 0x4c, 0x49, 0x53, 0x54, // "LIST" 0x1a, 0x00, 0x00, 0x00, // chunk size: 26 0x49, 0x4e, 0x46, 0x4f, // "INFO" 0x49, 0x53, 0x46, 0x54, // "ISFT" 0x0d, 0x00, 0x00, 0x00, // sub-chunk size: 13 0x4c, 0x61, 0x76, 0x66, 0x36, 0x32, 0x2e, 0x33, // "Lavf62.3" 0x2e, 0x31, 0x30, 0x30, 0x00, // ".100\0" /* padding byte for word alignment */ 0x00, // data chunk header 0x64, 0x61, 0x74, 0x61, // "data" 0x80, 0xcf, 0x00, 0x00, // chunk size ]; let mut bytes = header; // fill remaining space up to `FILE_ANALYSIS_BYTES` with synthetic PCM let audio_bytes_needed = FILE_ANALYSIS_BYTES - bytes.len(); for i in 0..(audio_bytes_needed / 2) { let sample = (i & 0xFF) as u8; bytes.push(sample); // low byte: varies bytes.push(0x00); // high byte: zero for small values } bytes } #[test] fn test_pcm16_wav_detected_as_binary() { let wav_bytes = build_pcm16_wav_bytes(); assert_eq!(wav_bytes.len(), FILE_ANALYSIS_BYTES); let result = analyze_byte_content(&wav_bytes); assert_eq!( result, ByteContent::Binary, "PCM 16-bit WAV should be detected as Binary via RIFF header" ); } #[test] fn test_le16_binary_not_misdetected_as_utf16le() { let mut bytes = b"FAKE".to_vec(); while bytes.len() < FILE_ANALYSIS_BYTES { let sample = (bytes.len() & 0xFF) as u8; bytes.push(sample); bytes.push(0x00); } bytes.truncate(FILE_ANALYSIS_BYTES); let result = analyze_byte_content(&bytes); assert_eq!( result, ByteContent::Binary, "LE 16-bit binary with control characters should be detected as Binary" ); } #[test] fn test_be16_binary_not_misdetected_as_utf16be() { let mut bytes = b"FAKE".to_vec(); while bytes.len() < FILE_ANALYSIS_BYTES { bytes.push(0x00); let sample = (bytes.len() & 0xFF) as u8; bytes.push(sample); } bytes.truncate(FILE_ANALYSIS_BYTES); let result = analyze_byte_content(&bytes); assert_eq!( result, ByteContent::Binary, "BE 16-bit binary with control characters should be detected as Binary" ); } // Mimics binary formats that interleave short ASCII fragments with small // length/type fields (as seen in some game/asset binary formats, e.g. // Tibia-style OTBM maps): most high bytes are zero, matching UTF-16LE's // null-byte pattern for ASCII, but the low bytes are mostly non-word // "tag" values rather than real letters/digits/spaces. fn build_tag_interleaved_binary_bytes() -> Vec { let mut bytes = Vec::new(); let tags: [u8; 6] = [0xFE, 0xFF, 0x25, 0x2B, 0xA3, 0xC5]; let mut i = 0; while bytes.len() < FILE_ANALYSIS_BYTES { bytes.push(tags[i % tags.len()]); bytes.push(0x00); i += 1; } bytes.truncate(FILE_ANALYSIS_BYTES); bytes } #[test] fn test_tag_interleaved_binary_not_misdetected_as_utf16le() { let bytes = build_tag_interleaved_binary_bytes(); assert_eq!(bytes.len(), FILE_ANALYSIS_BYTES); let result = analyze_byte_content(&bytes); assert_eq!( result, ByteContent::Binary, "binary data with sparse non-word low bytes and null high bytes \ should not be misdetected as UTF-16LE text" ); } #[test] fn test_utf16le_text_detected_as_utf16le() { let text = "Hello, world! This is a UTF-16 test string. "; let mut bytes = Vec::new(); while bytes.len() < FILE_ANALYSIS_BYTES { bytes.extend(text.encode_utf16().flat_map(|u| u.to_le_bytes())); } bytes.truncate(FILE_ANALYSIS_BYTES); assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Le); } #[test] fn test_utf16be_text_detected_as_utf16be() { let text = "Hello, world! This is a UTF-16 test string. "; let mut bytes = Vec::new(); while bytes.len() < FILE_ANALYSIS_BYTES { bytes.extend(text.encode_utf16().flat_map(|u| u.to_be_bytes())); } bytes.truncate(FILE_ANALYSIS_BYTES); assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Be); } #[test] fn test_utf16le_cyrillic_text_detected_as_utf16le() { let text = "Привет, мир! Это тестовая строка в UTF-16. "; let mut bytes = Vec::new(); while bytes.len() < FILE_ANALYSIS_BYTES { bytes.extend(text.encode_utf16().flat_map(|u| u.to_le_bytes())); } bytes.truncate(FILE_ANALYSIS_BYTES); assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Le); } #[test] fn test_utf16be_greek_text_detected_as_utf16be() { let text = "Γεια σου κόσμε! Αυτή είναι μια δοκιμαστική συμβολοσειρά. "; let mut bytes = Vec::new(); while bytes.len() < FILE_ANALYSIS_BYTES { bytes.extend(text.encode_utf16().flat_map(|u| u.to_be_bytes())); } bytes.truncate(FILE_ANALYSIS_BYTES); assert_eq!(analyze_byte_content(&bytes), ByteContent::Utf16Be); } #[test] fn test_known_binary_headers() { let cases: &[(&[u8], &str)] = &[ (b"RIFF\x00\x00\x00\x00WAVE", "WAV"), (b"RIFF\x00\x00\x00\x00AVI ", "AVI"), (b"OggS\x00\x02", "OGG"), (b"fLaC\x00\x00", "FLAC"), (b"ID3\x03\x00", "MP3 ID3v2"), (b"\xFF\xFB\x90\x00", "MP3 MPEG1 Layer3"), (b"\xFF\xF3\x90\x00", "MP3 MPEG2 Layer3"), ]; for (header, label) in cases { let mut bytes = header.to_vec(); bytes.resize(FILE_ANALYSIS_BYTES, 0x41); // pad with 'A' assert_eq!( analyze_byte_content(&bytes), ByteContent::Binary, "{label} should be detected as Binary" ); } } }