1pub mod row_chunk;
2
3use crate::{
4 DebuggerTextObject, LanguageScope, ModelineSettings, Outline, OutlineConfig, PLAIN_TEXT,
5 RunnableCapture, RunnableTag, TextObject, TreeSitterOptions,
6 diagnostic_set::{DiagnosticEntry, DiagnosticEntryRef, DiagnosticGroup},
7 language_settings::{AutoIndentMode, LanguageSettings},
8 outline::OutlineItem,
9 row_chunk::RowChunks,
10 syntax_map::{
11 MAX_BYTES_TO_QUERY, SyntaxLayer, SyntaxMap, SyntaxMapCapture, SyntaxMapCaptures,
12 SyntaxMapMatch, SyntaxMapMatches, SyntaxSnapshot, ToTreeSitterPoint,
13 },
14 task_context::RunnableRange,
15 text_diff::text_diff,
16 unified_diff_with_offsets,
17};
18pub use crate::{
19 Grammar, Language, LanguageRegistry,
20 diagnostic_set::DiagnosticSet,
21 highlight_map::{HighlightId, HighlightMap},
22 proto,
23};
24use anyhow::{Context as _, Result};
25use clock::Lamport;
26pub use clock::ReplicaId;
27use collections::{HashMap, HashSet};
28use encoding_rs::Encoding;
29use fs::MTime;
30use futures::channel::oneshot;
31use gpui::{
32 App, AppContext as _, Context, Entity, EventEmitter, HighlightStyle, SharedString, StyledText,
33 Task, TextStyle,
34};
35
36use lsp::{LanguageServerId, NumberOrString};
37use parking_lot::Mutex;
38use serde::{Deserialize, Serialize};
39use serde_json::Value;
40use settings::WorktreeId;
41use smallvec::SmallVec;
42use smol::future::yield_now;
43use std::{
44 any::Any,
45 borrow::Cow,
46 cell::Cell,
47 cmp::{self, Ordering, Reverse},
48 collections::{BTreeMap, BTreeSet},
49 future::Future,
50 iter::{self, Iterator, Peekable},
51 mem,
52 num::NonZeroU32,
53 ops::{Deref, Range},
54 path::PathBuf,
55 rc,
56 sync::Arc,
57 time::{Duration, Instant},
58 vec,
59};
60use sum_tree::TreeMap;
61use text::operation_queue::OperationQueue;
62use text::*;
63pub use text::{
64 Anchor, Bias, Buffer as TextBuffer, BufferId, BufferSnapshot as TextBufferSnapshot, Edit,
65 LineIndent, OffsetRangeExt, OffsetUtf16, Patch, Point, PointUtf16, Rope, Selection,
66 SelectionGoal, Subscription, TextDimension, TextSummary, ToOffset, ToOffsetUtf16, ToPoint,
67 ToPointUtf16, Transaction, TransactionId, Unclipped,
68};
69use theme::{ActiveTheme as _, SyntaxTheme};
70#[cfg(any(test, feature = "test-support"))]
71use util::RandomCharIter;
72use util::{RangeExt, debug_panic, maybe, paths::PathStyle, rel_path::RelPath};
73
74#[cfg(any(test, feature = "test-support"))]
75pub use {tree_sitter_python, tree_sitter_rust, tree_sitter_typescript};
76
77pub use lsp::DiagnosticSeverity;
78
79/// Indicate whether a [`Buffer`] has permissions to edit.
80#[derive(PartialEq, Clone, Copy, Debug)]
81pub enum Capability {
82 /// The buffer is a mutable replica.
83 ReadWrite,
84 /// The buffer is a mutable replica, but toggled to be only readable.
85 Read,
86 /// The buffer is a read-only replica.
87 ReadOnly,
88}
89
90impl Capability {
91 /// Returns `true` if the capability is `ReadWrite`.
92 pub fn editable(self) -> bool {
93 matches!(self, Capability::ReadWrite)
94 }
95}
96
97pub type BufferRow = u32;
98
99/// An in-memory representation of a source code file, including its text,
100/// syntax trees, git status, and diagnostics.
101pub struct Buffer {
102 text: TextBuffer,
103 branch_state: Option<BufferBranchState>,
104 /// Filesystem state, `None` when there is no path.
105 file: Option<Arc<dyn File>>,
106 /// The mtime of the file when this buffer was last loaded from
107 /// or saved to disk.
108 saved_mtime: Option<MTime>,
109 /// The version vector when this buffer was last loaded from
110 /// or saved to disk.
111 saved_version: clock::Global,
112 preview_version: clock::Global,
113 transaction_depth: usize,
114 was_dirty_before_starting_transaction: Option<bool>,
115 reload_task: Option<Task<Result<()>>>,
116 language: Option<Arc<Language>>,
117 autoindent_requests: Vec<Arc<AutoindentRequest>>,
118 wait_for_autoindent_txs: Vec<oneshot::Sender<()>>,
119 pending_autoindent: Option<Task<()>>,
120 sync_parse_timeout: Option<Duration>,
121 syntax_map: Mutex<SyntaxMap>,
122 reparse: Option<Task<()>>,
123 parse_status: (watch::Sender<ParseStatus>, watch::Receiver<ParseStatus>),
124 non_text_state_update_count: usize,
125 diagnostics: TreeMap<LanguageServerId, DiagnosticSet>,
126 remote_selections: TreeMap<ReplicaId, SelectionSet>,
127 diagnostics_timestamp: clock::Lamport,
128 completion_triggers: BTreeSet<String>,
129 completion_triggers_per_language_server: HashMap<LanguageServerId, BTreeSet<String>>,
130 completion_triggers_timestamp: clock::Lamport,
131 deferred_ops: OperationQueue<Operation>,
132 capability: Capability,
133 has_conflict: bool,
134 /// Memoize calls to has_changes_since(saved_version).
135 /// The contents of a cell are (self.version, has_changes) at the time of a last call.
136 has_unsaved_edits: Cell<(clock::Global, bool)>,
137 change_bits: Vec<rc::Weak<Cell<bool>>>,
138 modeline: Option<Arc<ModelineSettings>>,
139 _subscriptions: Vec<gpui::Subscription>,
140 tree_sitter_data: Arc<TreeSitterData>,
141 encoding: &'static Encoding,
142 has_bom: bool,
143 reload_with_encoding_txns: HashMap<TransactionId, (&'static Encoding, bool)>,
144}
145
146#[derive(Debug)]
147pub struct TreeSitterData {
148 chunks: RowChunks,
149 brackets_by_chunks: Mutex<Vec<Option<Vec<BracketMatch<usize>>>>>,
150}
151
152const MAX_ROWS_IN_A_CHUNK: u32 = 50;
153
154impl TreeSitterData {
155 fn clear(&mut self, snapshot: &text::BufferSnapshot) {
156 self.chunks = RowChunks::new(&snapshot, MAX_ROWS_IN_A_CHUNK);
157 self.brackets_by_chunks.get_mut().clear();
158 self.brackets_by_chunks
159 .get_mut()
160 .resize(self.chunks.len(), None);
161 }
162
163 fn new(snapshot: &text::BufferSnapshot) -> Self {
164 let chunks = RowChunks::new(&snapshot, MAX_ROWS_IN_A_CHUNK);
165 Self {
166 brackets_by_chunks: Mutex::new(vec![None; chunks.len()]),
167 chunks,
168 }
169 }
170
171 fn version(&self) -> &clock::Global {
172 self.chunks.version()
173 }
174}
175
176#[derive(Copy, Clone, Debug, PartialEq, Eq)]
177pub enum ParseStatus {
178 Idle,
179 Parsing,
180}
181
182struct BufferBranchState {
183 base_buffer: Entity<Buffer>,
184 merged_operations: Vec<Lamport>,
185}
186
187/// An immutable, cheaply cloneable representation of a fixed
188/// state of a buffer.
189pub struct BufferSnapshot {
190 pub text: text::BufferSnapshot,
191 pub(crate) syntax: SyntaxSnapshot,
192 tree_sitter_data: Arc<TreeSitterData>,
193 diagnostics: TreeMap<LanguageServerId, DiagnosticSet>,
194 remote_selections: TreeMap<ReplicaId, SelectionSet>,
195 language: Option<Arc<Language>>,
196 file: Option<Arc<dyn File>>,
197 non_text_state_update_count: usize,
198 pub capability: Capability,
199 modeline: Option<Arc<ModelineSettings>>,
200}
201
202/// The kind and amount of indentation in a particular line. For now,
203/// assumes that indentation is all the same character.
204#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, Default)]
205pub struct IndentSize {
206 /// The number of bytes that comprise the indentation.
207 pub len: u32,
208 /// The kind of whitespace used for indentation.
209 pub kind: IndentKind,
210}
211
212/// A whitespace character that's used for indentation.
213#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, Default)]
214pub enum IndentKind {
215 /// An ASCII space character.
216 #[default]
217 Space,
218 /// An ASCII tab character.
219 Tab,
220}
221
222/// The shape of a selection cursor.
223#[derive(Copy, Clone, Debug, Default, PartialEq, Eq)]
224pub enum CursorShape {
225 /// A vertical bar
226 #[default]
227 Bar,
228 /// A block that surrounds the following character
229 Block,
230 /// An underline that runs along the following character
231 Underline,
232 /// A box drawn around the following character
233 Hollow,
234}
235
236impl From<settings::CursorShape> for CursorShape {
237 fn from(shape: settings::CursorShape) -> Self {
238 match shape {
239 settings::CursorShape::Bar => CursorShape::Bar,
240 settings::CursorShape::Block => CursorShape::Block,
241 settings::CursorShape::Underline => CursorShape::Underline,
242 settings::CursorShape::Hollow => CursorShape::Hollow,
243 }
244 }
245}
246
247#[derive(Clone, Debug)]
248struct SelectionSet {
249 line_mode: bool,
250 cursor_shape: CursorShape,
251 selections: Arc<[Selection<Anchor>]>,
252 lamport_timestamp: clock::Lamport,
253}
254
255/// A diagnostic associated with a certain range of a buffer.
256#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize)]
257pub struct Diagnostic {
258 /// The name of the service that produced this diagnostic.
259 pub source: Option<String>,
260 /// The ID provided by the dynamic registration that produced this diagnostic.
261 pub registration_id: Option<SharedString>,
262 /// A machine-readable code that identifies this diagnostic.
263 pub code: Option<NumberOrString>,
264 pub code_description: Option<lsp::Uri>,
265 /// Whether this diagnostic is a hint, warning, or error.
266 pub severity: DiagnosticSeverity,
267 /// The human-readable message associated with this diagnostic.
268 pub message: String,
269 /// The human-readable message (in markdown format)
270 pub markdown: Option<String>,
271 /// An id that identifies the group to which this diagnostic belongs.
272 ///
273 /// When a language server produces a diagnostic with
274 /// one or more associated diagnostics, those diagnostics are all
275 /// assigned a single group ID.
276 pub group_id: usize,
277 /// Whether this diagnostic is the primary diagnostic for its group.
278 ///
279 /// In a given group, the primary diagnostic is the top-level diagnostic
280 /// returned by the language server. The non-primary diagnostics are the
281 /// associated diagnostics.
282 pub is_primary: bool,
283 /// Whether this diagnostic is considered to originate from an analysis of
284 /// files on disk, as opposed to any unsaved buffer contents. This is a
285 /// property of a given diagnostic source, and is configured for a given
286 /// language server via the [`LspAdapter::disk_based_diagnostic_sources`](crate::LspAdapter::disk_based_diagnostic_sources) method
287 /// for the language server.
288 pub is_disk_based: bool,
289 /// Whether this diagnostic marks unnecessary code.
290 pub is_unnecessary: bool,
291 /// Quick separation of diagnostics groups based by their source.
292 pub source_kind: DiagnosticSourceKind,
293 /// Data from language server that produced this diagnostic. Passed back to the LS when we request code actions for this diagnostic.
294 pub data: Option<Value>,
295 /// Whether to underline the corresponding text range in the editor.
296 pub underline: bool,
297}
298
299#[derive(Clone, Copy, Debug, PartialEq, Eq, Serialize, Deserialize)]
300pub enum DiagnosticSourceKind {
301 Pulled,
302 Pushed,
303 Other,
304}
305
306/// An operation used to synchronize this buffer with its other replicas.
307#[derive(Clone, Debug, PartialEq)]
308pub enum Operation {
309 /// A text operation.
310 Buffer(text::Operation),
311
312 /// An update to the buffer's diagnostics.
313 UpdateDiagnostics {
314 /// The id of the language server that produced the new diagnostics.
315 server_id: LanguageServerId,
316 /// The diagnostics.
317 diagnostics: Arc<[DiagnosticEntry<Anchor>]>,
318 /// The buffer's lamport timestamp.
319 lamport_timestamp: clock::Lamport,
320 },
321
322 /// An update to the most recent selections in this buffer.
323 UpdateSelections {
324 /// The selections.
325 selections: Arc<[Selection<Anchor>]>,
326 /// The buffer's lamport timestamp.
327 lamport_timestamp: clock::Lamport,
328 /// Whether the selections are in 'line mode'.
329 line_mode: bool,
330 /// The [`CursorShape`] associated with these selections.
331 cursor_shape: CursorShape,
332 },
333
334 /// An update to the characters that should trigger autocompletion
335 /// for this buffer.
336 UpdateCompletionTriggers {
337 /// The characters that trigger autocompletion.
338 triggers: Vec<String>,
339 /// The buffer's lamport timestamp.
340 lamport_timestamp: clock::Lamport,
341 /// The language server ID.
342 server_id: LanguageServerId,
343 },
344
345 /// An update to the line ending type of this buffer.
346 UpdateLineEnding {
347 /// The line ending type.
348 line_ending: LineEnding,
349 /// The buffer's lamport timestamp.
350 lamport_timestamp: clock::Lamport,
351 },
352}
353
354/// An event that occurs in a buffer.
355#[derive(Clone, Debug, PartialEq)]
356pub enum BufferEvent {
357 /// The buffer was changed in a way that must be
358 /// propagated to its other replicas.
359 Operation {
360 operation: Operation,
361 is_local: bool,
362 },
363 /// The buffer was edited.
364 Edited { is_local: bool },
365 /// The buffer's `dirty` bit changed.
366 DirtyChanged,
367 /// The buffer was saved.
368 Saved,
369 /// The buffer's file was changed on disk.
370 FileHandleChanged,
371 /// The buffer was reloaded.
372 Reloaded,
373 /// The buffer is in need of a reload
374 ReloadNeeded,
375 /// The buffer's language was changed.
376 /// The boolean indicates whether this buffer did not have a language before, but does now.
377 LanguageChanged(bool),
378 /// The buffer's syntax trees were updated.
379 Reparsed,
380 /// The buffer's diagnostics were updated.
381 DiagnosticsUpdated,
382 /// The buffer gained or lost editing capabilities.
383 CapabilityChanged,
384}
385
386/// The file associated with a buffer.
387pub trait File: Send + Sync + Any {
388 /// Returns the [`LocalFile`] associated with this file, if the
389 /// file is local.
390 fn as_local(&self) -> Option<&dyn LocalFile>;
391
392 /// Returns whether this file is local.
393 fn is_local(&self) -> bool {
394 self.as_local().is_some()
395 }
396
397 /// Returns whether the file is new, exists in storage, or has been deleted. Includes metadata
398 /// only available in some states, such as modification time.
399 fn disk_state(&self) -> DiskState;
400
401 /// Returns the path of this file relative to the worktree's root directory.
402 fn path(&self) -> &Arc<RelPath>;
403
404 /// Returns the path of this file relative to the worktree's parent directory (this means it
405 /// includes the name of the worktree's root folder).
406 fn full_path(&self, cx: &App) -> PathBuf;
407
408 /// Returns the path style of this file.
409 fn path_style(&self, cx: &App) -> PathStyle;
410
411 /// Returns the last component of this handle's absolute path. If this handle refers to the root
412 /// of its worktree, then this method will return the name of the worktree itself.
413 fn file_name<'a>(&'a self, cx: &'a App) -> &'a str;
414
415 /// Returns the id of the worktree to which this file belongs.
416 ///
417 /// This is needed for looking up project-specific settings.
418 fn worktree_id(&self, cx: &App) -> WorktreeId;
419
420 /// Converts this file into a protobuf message.
421 fn to_proto(&self, cx: &App) -> rpc::proto::File;
422
423 /// Return whether Zed considers this to be a private file.
424 fn is_private(&self) -> bool;
425
426 fn can_open(&self) -> bool {
427 !self.is_local()
428 }
429}
430
431/// The file's storage status - whether it's stored (`Present`), and if so when it was last
432/// modified. In the case where the file is not stored, it can be either `New` or `Deleted`. In the
433/// UI these two states are distinguished. For example, the buffer tab does not display a deletion
434/// indicator for new files.
435#[derive(Copy, Clone, Debug, PartialEq)]
436pub enum DiskState {
437 /// File created in Zed that has not been saved.
438 New,
439 /// File present on the filesystem.
440 Present { mtime: MTime, size: u64 },
441 /// Deleted file that was previously present.
442 Deleted,
443 /// An old version of a file that was previously present
444 /// usually from a version control system. e.g. A git blob
445 Historic { was_deleted: bool },
446}
447
448impl DiskState {
449 /// Returns the file's last known modification time on disk.
450 pub fn mtime(self) -> Option<MTime> {
451 match self {
452 DiskState::New => None,
453 DiskState::Present { mtime, .. } => Some(mtime),
454 DiskState::Deleted => None,
455 DiskState::Historic { .. } => None,
456 }
457 }
458
459 /// Returns the file's size on disk in bytes.
460 pub fn size(self) -> Option<u64> {
461 match self {
462 DiskState::New => None,
463 DiskState::Present { size, .. } => Some(size),
464 DiskState::Deleted => None,
465 DiskState::Historic { .. } => None,
466 }
467 }
468
469 pub fn exists(&self) -> bool {
470 match self {
471 DiskState::New => false,
472 DiskState::Present { .. } => true,
473 DiskState::Deleted => false,
474 DiskState::Historic { .. } => false,
475 }
476 }
477
478 /// Returns true if this state represents a deleted file.
479 pub fn is_deleted(&self) -> bool {
480 match self {
481 DiskState::Deleted => true,
482 DiskState::Historic { was_deleted } => *was_deleted,
483 _ => false,
484 }
485 }
486}
487
488/// The file associated with a buffer, in the case where the file is on the local disk.
489pub trait LocalFile: File {
490 /// Returns the absolute path of this file
491 fn abs_path(&self, cx: &App) -> PathBuf;
492
493 /// Loads the file contents from disk and returns them as a UTF-8 encoded string.
494 fn load(&self, cx: &App) -> Task<Result<String>>;
495
496 /// Loads the file's contents from disk.
497 fn load_bytes(&self, cx: &App) -> Task<Result<Vec<u8>>>;
498}
499
500/// The auto-indent behavior associated with an editing operation.
501/// For some editing operations, each affected line of text has its
502/// indentation recomputed. For other operations, the entire block
503/// of edited text is adjusted uniformly.
504#[derive(Clone, Debug)]
505pub enum AutoindentMode {
506 /// Indent each line of inserted text.
507 EachLine,
508 /// Apply the same indentation adjustment to all of the lines
509 /// in a given insertion.
510 Block {
511 /// The original indentation column of the first line of each
512 /// insertion, if it has been copied.
513 ///
514 /// Knowing this makes it possible to preserve the relative indentation
515 /// of every line in the insertion from when it was copied.
516 ///
517 /// If the original indent column is `a`, and the first line of insertion
518 /// is then auto-indented to column `b`, then every other line of
519 /// the insertion will be auto-indented to column `b - a`
520 original_indent_columns: Vec<Option<u32>>,
521 },
522}
523
524#[derive(Clone)]
525struct AutoindentRequest {
526 before_edit: BufferSnapshot,
527 entries: Vec<AutoindentRequestEntry>,
528 is_block_mode: bool,
529 ignore_empty_lines: bool,
530}
531
532#[derive(Debug, Clone)]
533struct AutoindentRequestEntry {
534 /// A range of the buffer whose indentation should be adjusted.
535 range: Range<Anchor>,
536 /// The row of the edit start in the buffer before the edit was applied.
537 /// This is stored here because the anchor in range is created after
538 /// the edit, so it cannot be used with the before_edit snapshot.
539 old_row: Option<u32>,
540 indent_size: IndentSize,
541 original_indent_column: Option<u32>,
542}
543
544#[derive(Debug)]
545struct IndentSuggestion {
546 basis_row: u32,
547 delta: Ordering,
548 within_error: bool,
549}
550
551struct BufferChunkHighlights<'a> {
552 captures: SyntaxMapCaptures<'a>,
553 next_capture: Option<SyntaxMapCapture<'a>>,
554 stack: Vec<(usize, HighlightId)>,
555 highlight_maps: Vec<HighlightMap>,
556}
557
558/// An iterator that yields chunks of a buffer's text, along with their
559/// syntax highlights and diagnostic status.
560pub struct BufferChunks<'a> {
561 buffer_snapshot: Option<&'a BufferSnapshot>,
562 range: Range<usize>,
563 chunks: text::Chunks<'a>,
564 diagnostic_endpoints: Option<Peekable<vec::IntoIter<DiagnosticEndpoint>>>,
565 error_depth: usize,
566 warning_depth: usize,
567 information_depth: usize,
568 hint_depth: usize,
569 unnecessary_depth: usize,
570 underline: bool,
571 highlights: Option<BufferChunkHighlights<'a>>,
572}
573
574/// A chunk of a buffer's text, along with its syntax highlight and
575/// diagnostic status.
576#[derive(Clone, Debug, Default)]
577pub struct Chunk<'a> {
578 /// The text of the chunk.
579 pub text: &'a str,
580 /// The syntax highlighting style of the chunk.
581 pub syntax_highlight_id: Option<HighlightId>,
582 /// The highlight style that has been applied to this chunk in
583 /// the editor.
584 pub highlight_style: Option<HighlightStyle>,
585 /// The severity of diagnostic associated with this chunk, if any.
586 pub diagnostic_severity: Option<DiagnosticSeverity>,
587 /// A bitset of which characters are tabs in this string.
588 pub tabs: u128,
589 /// Bitmap of character indices in this chunk
590 pub chars: u128,
591 /// Bitmap of newline indices in this chunk
592 pub newlines: u128,
593 /// Whether this chunk of text is marked as unnecessary.
594 pub is_unnecessary: bool,
595 /// Whether this chunk of text was originally a tab character.
596 pub is_tab: bool,
597 /// Whether this chunk of text was originally an inlay.
598 pub is_inlay: bool,
599 /// Whether to underline the corresponding text range in the editor.
600 pub underline: bool,
601}
602
603/// A set of edits to a given version of a buffer, computed asynchronously.
604#[derive(Debug, Clone)]
605pub struct Diff {
606 pub base_version: clock::Global,
607 pub line_ending: LineEnding,
608 pub edits: Vec<(Range<usize>, Arc<str>)>,
609}
610
611#[derive(Debug, Clone, Copy)]
612pub(crate) struct DiagnosticEndpoint {
613 offset: usize,
614 is_start: bool,
615 underline: bool,
616 severity: DiagnosticSeverity,
617 is_unnecessary: bool,
618}
619
620/// A class of characters, used for characterizing a run of text.
621#[derive(Copy, Clone, Eq, PartialEq, PartialOrd, Ord, Debug)]
622pub enum CharKind {
623 /// Whitespace.
624 Whitespace,
625 /// Punctuation.
626 Punctuation,
627 /// Word.
628 Word,
629}
630
631/// Context for character classification within a specific scope.
632#[derive(Copy, Clone, Eq, PartialEq, Debug)]
633pub enum CharScopeContext {
634 /// Character classification for completion queries.
635 ///
636 /// This context treats certain characters as word constituents that would
637 /// normally be considered punctuation, such as '-' in Tailwind classes
638 /// ("bg-yellow-100") or '.' in import paths ("foo.ts").
639 Completion,
640 /// Character classification for linked edits.
641 ///
642 /// This context handles characters that should be treated as part of
643 /// identifiers during linked editing operations, such as '.' in JSX
644 /// component names like `<Animated.View>`.
645 LinkedEdit,
646}
647
648/// A runnable is a set of data about a region that could be resolved into a task
649pub struct Runnable {
650 pub tags: SmallVec<[RunnableTag; 1]>,
651 pub language: Arc<Language>,
652 pub buffer: BufferId,
653}
654
655#[derive(Default, Clone, Debug)]
656pub struct HighlightedText {
657 pub text: SharedString,
658 pub highlights: Vec<(Range<usize>, HighlightStyle)>,
659}
660
661#[derive(Default, Debug)]
662struct HighlightedTextBuilder {
663 pub text: String,
664 highlights: Vec<(Range<usize>, HighlightStyle)>,
665}
666
667impl HighlightedText {
668 pub fn from_buffer_range<T: ToOffset>(
669 range: Range<T>,
670 snapshot: &text::BufferSnapshot,
671 syntax_snapshot: &SyntaxSnapshot,
672 override_style: Option<HighlightStyle>,
673 syntax_theme: &SyntaxTheme,
674 ) -> Self {
675 let mut highlighted_text = HighlightedTextBuilder::default();
676 highlighted_text.add_text_from_buffer_range(
677 range,
678 snapshot,
679 syntax_snapshot,
680 override_style,
681 syntax_theme,
682 );
683 highlighted_text.build()
684 }
685
686 pub fn to_styled_text(&self, default_style: &TextStyle) -> StyledText {
687 gpui::StyledText::new(self.text.clone())
688 .with_default_highlights(default_style, self.highlights.iter().cloned())
689 }
690
691 /// Returns the first line without leading whitespace unless highlighted
692 /// and a boolean indicating if there are more lines after
693 pub fn first_line_preview(self) -> (Self, bool) {
694 let newline_ix = self.text.find('\n').unwrap_or(self.text.len());
695 let first_line = &self.text[..newline_ix];
696
697 // Trim leading whitespace, unless an edit starts prior to it.
698 let mut preview_start_ix = first_line.len() - first_line.trim_start().len();
699 if let Some((first_highlight_range, _)) = self.highlights.first() {
700 preview_start_ix = preview_start_ix.min(first_highlight_range.start);
701 }
702
703 let preview_text = &first_line[preview_start_ix..];
704 let preview_highlights = self
705 .highlights
706 .into_iter()
707 .skip_while(|(range, _)| range.end <= preview_start_ix)
708 .take_while(|(range, _)| range.start < newline_ix)
709 .filter_map(|(mut range, highlight)| {
710 range.start = range.start.saturating_sub(preview_start_ix);
711 range.end = range.end.min(newline_ix).saturating_sub(preview_start_ix);
712 if range.is_empty() {
713 None
714 } else {
715 Some((range, highlight))
716 }
717 });
718
719 let preview = Self {
720 text: SharedString::new(preview_text),
721 highlights: preview_highlights.collect(),
722 };
723
724 (preview, self.text.len() > newline_ix)
725 }
726}
727
728impl HighlightedTextBuilder {
729 pub fn build(self) -> HighlightedText {
730 HighlightedText {
731 text: self.text.into(),
732 highlights: self.highlights,
733 }
734 }
735
736 pub fn add_text_from_buffer_range<T: ToOffset>(
737 &mut self,
738 range: Range<T>,
739 snapshot: &text::BufferSnapshot,
740 syntax_snapshot: &SyntaxSnapshot,
741 override_style: Option<HighlightStyle>,
742 syntax_theme: &SyntaxTheme,
743 ) {
744 let range = range.to_offset(snapshot);
745 for chunk in Self::highlighted_chunks(range, snapshot, syntax_snapshot) {
746 let start = self.text.len();
747 self.text.push_str(chunk.text);
748 let end = self.text.len();
749
750 if let Some(highlight_style) = chunk
751 .syntax_highlight_id
752 .and_then(|id| id.style(syntax_theme))
753 {
754 let highlight_style = override_style.map_or(highlight_style, |override_style| {
755 highlight_style.highlight(override_style)
756 });
757 self.highlights.push((start..end, highlight_style));
758 } else if let Some(override_style) = override_style {
759 self.highlights.push((start..end, override_style));
760 }
761 }
762 }
763
764 fn highlighted_chunks<'a>(
765 range: Range<usize>,
766 snapshot: &'a text::BufferSnapshot,
767 syntax_snapshot: &'a SyntaxSnapshot,
768 ) -> BufferChunks<'a> {
769 let captures = syntax_snapshot.captures(range.clone(), snapshot, |grammar| {
770 grammar
771 .highlights_config
772 .as_ref()
773 .map(|config| &config.query)
774 });
775
776 let highlight_maps = captures
777 .grammars()
778 .iter()
779 .map(|grammar| grammar.highlight_map())
780 .collect();
781
782 BufferChunks::new(
783 snapshot.as_rope(),
784 range,
785 Some((captures, highlight_maps)),
786 false,
787 None,
788 )
789 }
790}
791
792#[derive(Clone)]
793pub struct EditPreview {
794 old_snapshot: text::BufferSnapshot,
795 applied_edits_snapshot: text::BufferSnapshot,
796 syntax_snapshot: SyntaxSnapshot,
797}
798
799impl EditPreview {
800 pub fn as_unified_diff(
801 &self,
802 file: Option<&Arc<dyn File>>,
803 edits: &[(Range<Anchor>, impl AsRef<str>)],
804 ) -> Option<String> {
805 let (first, _) = edits.first()?;
806 let (last, _) = edits.last()?;
807
808 let start = first.start.to_point(&self.old_snapshot);
809 let old_end = last.end.to_point(&self.old_snapshot);
810 let new_end = last
811 .end
812 .bias_right(&self.old_snapshot)
813 .to_point(&self.applied_edits_snapshot);
814
815 let start = Point::new(start.row.saturating_sub(3), 0);
816 let old_end = Point::new(old_end.row + 4, 0).min(self.old_snapshot.max_point());
817 let new_end = Point::new(new_end.row + 4, 0).min(self.applied_edits_snapshot.max_point());
818
819 let diff_body = unified_diff_with_offsets(
820 &self
821 .old_snapshot
822 .text_for_range(start..old_end)
823 .collect::<String>(),
824 &self
825 .applied_edits_snapshot
826 .text_for_range(start..new_end)
827 .collect::<String>(),
828 start.row,
829 start.row,
830 );
831
832 let path = file.map(|f| f.path().as_unix_str());
833 let header = match path {
834 Some(p) => format!("--- a/{}\n+++ b/{}\n", p, p),
835 None => String::new(),
836 };
837
838 Some(format!("{}{}", header, diff_body))
839 }
840
841 pub fn highlight_edits(
842 &self,
843 current_snapshot: &BufferSnapshot,
844 edits: &[(Range<Anchor>, impl AsRef<str>)],
845 include_deletions: bool,
846 cx: &App,
847 ) -> HighlightedText {
848 let Some(visible_range_in_preview_snapshot) = self.compute_visible_range(edits) else {
849 return HighlightedText::default();
850 };
851
852 let mut highlighted_text = HighlightedTextBuilder::default();
853
854 let visible_range_in_preview_snapshot =
855 visible_range_in_preview_snapshot.to_offset(&self.applied_edits_snapshot);
856 let mut offset_in_preview_snapshot = visible_range_in_preview_snapshot.start;
857
858 let insertion_highlight_style = HighlightStyle {
859 background_color: Some(cx.theme().status().created_background),
860 ..Default::default()
861 };
862 let deletion_highlight_style = HighlightStyle {
863 background_color: Some(cx.theme().status().deleted_background),
864 ..Default::default()
865 };
866 let syntax_theme = cx.theme().syntax();
867
868 for (range, edit_text) in edits {
869 let edit_new_end_in_preview_snapshot = range
870 .end
871 .bias_right(&self.old_snapshot)
872 .to_offset(&self.applied_edits_snapshot);
873 let edit_start_in_preview_snapshot =
874 edit_new_end_in_preview_snapshot - edit_text.as_ref().len();
875
876 let unchanged_range_in_preview_snapshot =
877 offset_in_preview_snapshot..edit_start_in_preview_snapshot;
878 if !unchanged_range_in_preview_snapshot.is_empty() {
879 highlighted_text.add_text_from_buffer_range(
880 unchanged_range_in_preview_snapshot,
881 &self.applied_edits_snapshot,
882 &self.syntax_snapshot,
883 None,
884 syntax_theme,
885 );
886 }
887
888 let range_in_current_snapshot = range.to_offset(current_snapshot);
889 if include_deletions && !range_in_current_snapshot.is_empty() {
890 highlighted_text.add_text_from_buffer_range(
891 range_in_current_snapshot,
892 ¤t_snapshot.text,
893 ¤t_snapshot.syntax,
894 Some(deletion_highlight_style),
895 syntax_theme,
896 );
897 }
898
899 if !edit_text.as_ref().is_empty() {
900 highlighted_text.add_text_from_buffer_range(
901 edit_start_in_preview_snapshot..edit_new_end_in_preview_snapshot,
902 &self.applied_edits_snapshot,
903 &self.syntax_snapshot,
904 Some(insertion_highlight_style),
905 syntax_theme,
906 );
907 }
908
909 offset_in_preview_snapshot = edit_new_end_in_preview_snapshot;
910 }
911
912 highlighted_text.add_text_from_buffer_range(
913 offset_in_preview_snapshot..visible_range_in_preview_snapshot.end,
914 &self.applied_edits_snapshot,
915 &self.syntax_snapshot,
916 None,
917 syntax_theme,
918 );
919
920 highlighted_text.build()
921 }
922
923 pub fn build_result_buffer(&self, cx: &mut App) -> Entity<Buffer> {
924 cx.new(|cx| {
925 let mut buffer = Buffer::local_normalized(
926 self.applied_edits_snapshot.as_rope().clone(),
927 self.applied_edits_snapshot.line_ending(),
928 cx,
929 );
930 buffer.set_language_async(self.syntax_snapshot.root_language(), cx);
931 buffer
932 })
933 }
934
935 pub fn anchor_to_offset_in_result(&self, anchor: Anchor) -> usize {
936 anchor
937 .bias_right(&self.old_snapshot)
938 .to_offset(&self.applied_edits_snapshot)
939 }
940
941 pub fn compute_visible_range<T>(&self, edits: &[(Range<Anchor>, T)]) -> Option<Range<Point>> {
942 let (first, _) = edits.first()?;
943 let (last, _) = edits.last()?;
944
945 let start = first
946 .start
947 .bias_left(&self.old_snapshot)
948 .to_point(&self.applied_edits_snapshot);
949 let end = last
950 .end
951 .bias_right(&self.old_snapshot)
952 .to_point(&self.applied_edits_snapshot);
953
954 // Ensure that the first line of the first edit and the last line of the last edit are always fully visible
955 let range = Point::new(start.row, 0)
956 ..Point::new(end.row, self.applied_edits_snapshot.line_len(end.row));
957
958 Some(range)
959 }
960}
961
962#[derive(Clone, Debug, PartialEq, Eq)]
963pub struct BracketMatch<T> {
964 pub open_range: Range<T>,
965 pub close_range: Range<T>,
966 pub newline_only: bool,
967 pub syntax_layer_depth: usize,
968 pub color_index: Option<usize>,
969}
970
971impl<T> BracketMatch<T> {
972 pub fn bracket_ranges(self) -> (Range<T>, Range<T>) {
973 (self.open_range, self.close_range)
974 }
975}
976
977impl Buffer {
978 /// Create a new buffer with the given base text.
979 pub fn local<T: Into<String>>(base_text: T, cx: &Context<Self>) -> Self {
980 Self::build(
981 TextBuffer::new(
982 ReplicaId::LOCAL,
983 cx.entity_id().as_non_zero_u64().into(),
984 base_text.into(),
985 ),
986 None,
987 Capability::ReadWrite,
988 )
989 }
990
991 /// Create a new buffer with the given base text that has proper line endings and other normalization applied.
992 pub fn local_normalized(
993 base_text_normalized: Rope,
994 line_ending: LineEnding,
995 cx: &Context<Self>,
996 ) -> Self {
997 Self::build(
998 TextBuffer::new_normalized(
999 ReplicaId::LOCAL,
1000 cx.entity_id().as_non_zero_u64().into(),
1001 line_ending,
1002 base_text_normalized,
1003 ),
1004 None,
1005 Capability::ReadWrite,
1006 )
1007 }
1008
1009 /// Create a new buffer that is a replica of a remote buffer.
1010 pub fn remote(
1011 remote_id: BufferId,
1012 replica_id: ReplicaId,
1013 capability: Capability,
1014 base_text: impl Into<String>,
1015 ) -> Self {
1016 Self::build(
1017 TextBuffer::new(replica_id, remote_id, base_text.into()),
1018 None,
1019 capability,
1020 )
1021 }
1022
1023 /// Create a new buffer that is a replica of a remote buffer, populating its
1024 /// state from the given protobuf message.
1025 pub fn from_proto(
1026 replica_id: ReplicaId,
1027 capability: Capability,
1028 message: proto::BufferState,
1029 file: Option<Arc<dyn File>>,
1030 ) -> Result<Self> {
1031 let buffer_id = BufferId::new(message.id).context("Could not deserialize buffer_id")?;
1032 let buffer = TextBuffer::new(replica_id, buffer_id, message.base_text);
1033 let mut this = Self::build(buffer, file, capability);
1034 this.text.set_line_ending(proto::deserialize_line_ending(
1035 rpc::proto::LineEnding::from_i32(message.line_ending).context("missing line_ending")?,
1036 ));
1037 this.saved_version = proto::deserialize_version(&message.saved_version);
1038 this.saved_mtime = message.saved_mtime.map(|time| time.into());
1039 Ok(this)
1040 }
1041
1042 /// Serialize the buffer's state to a protobuf message.
1043 pub fn to_proto(&self, cx: &App) -> proto::BufferState {
1044 proto::BufferState {
1045 id: self.remote_id().into(),
1046 file: self.file.as_ref().map(|f| f.to_proto(cx)),
1047 base_text: self.base_text().to_string(),
1048 line_ending: proto::serialize_line_ending(self.line_ending()) as i32,
1049 saved_version: proto::serialize_version(&self.saved_version),
1050 saved_mtime: self.saved_mtime.map(|time| time.into()),
1051 }
1052 }
1053
1054 /// Serialize as protobufs all of the changes to the buffer since the given version.
1055 pub fn serialize_ops(
1056 &self,
1057 since: Option<clock::Global>,
1058 cx: &App,
1059 ) -> Task<Vec<proto::Operation>> {
1060 let mut operations = Vec::new();
1061 operations.extend(self.deferred_ops.iter().map(proto::serialize_operation));
1062
1063 operations.extend(self.remote_selections.iter().map(|(_, set)| {
1064 proto::serialize_operation(&Operation::UpdateSelections {
1065 selections: set.selections.clone(),
1066 lamport_timestamp: set.lamport_timestamp,
1067 line_mode: set.line_mode,
1068 cursor_shape: set.cursor_shape,
1069 })
1070 }));
1071
1072 for (server_id, diagnostics) in self.diagnostics.iter() {
1073 operations.push(proto::serialize_operation(&Operation::UpdateDiagnostics {
1074 lamport_timestamp: self.diagnostics_timestamp,
1075 server_id: *server_id,
1076 diagnostics: diagnostics.iter().cloned().collect(),
1077 }));
1078 }
1079
1080 for (server_id, completions) in &self.completion_triggers_per_language_server {
1081 operations.push(proto::serialize_operation(
1082 &Operation::UpdateCompletionTriggers {
1083 triggers: completions.iter().cloned().collect(),
1084 lamport_timestamp: self.completion_triggers_timestamp,
1085 server_id: *server_id,
1086 },
1087 ));
1088 }
1089
1090 let text_operations = self.text.operations().clone();
1091 cx.background_spawn(async move {
1092 let since = since.unwrap_or_default();
1093 operations.extend(
1094 text_operations
1095 .iter()
1096 .filter(|(_, op)| !since.observed(op.timestamp()))
1097 .map(|(_, op)| proto::serialize_operation(&Operation::Buffer(op.clone()))),
1098 );
1099 operations.sort_unstable_by_key(proto::lamport_timestamp_for_operation);
1100 operations
1101 })
1102 }
1103
1104 /// Assign a language to the buffer, returning the buffer.
1105 pub fn with_language_async(mut self, language: Arc<Language>, cx: &mut Context<Self>) -> Self {
1106 self.set_language_async(Some(language), cx);
1107 self
1108 }
1109
1110 /// Assign a language to the buffer, blocking for up to 1ms to reparse the buffer, returning the buffer.
1111 #[ztracing::instrument(skip_all, fields(lang = language.config.name.0.as_str()))]
1112 pub fn with_language(mut self, language: Arc<Language>, cx: &mut Context<Self>) -> Self {
1113 self.set_language(Some(language), cx);
1114 self
1115 }
1116
1117 /// Returns the [`Capability`] of this buffer.
1118 pub fn capability(&self) -> Capability {
1119 self.capability
1120 }
1121
1122 /// Whether this buffer can only be read.
1123 pub fn read_only(&self) -> bool {
1124 !self.capability.editable()
1125 }
1126
1127 /// Builds a [`Buffer`] with the given underlying [`TextBuffer`], diff base, [`File`] and [`Capability`].
1128 pub fn build(buffer: TextBuffer, file: Option<Arc<dyn File>>, capability: Capability) -> Self {
1129 let saved_mtime = file.as_ref().and_then(|file| file.disk_state().mtime());
1130 let snapshot = buffer.snapshot();
1131 let syntax_map = Mutex::new(SyntaxMap::new(&snapshot));
1132 let tree_sitter_data = TreeSitterData::new(snapshot);
1133 Self {
1134 saved_mtime,
1135 tree_sitter_data: Arc::new(tree_sitter_data),
1136 saved_version: buffer.version(),
1137 preview_version: buffer.version(),
1138 reload_task: None,
1139 transaction_depth: 0,
1140 was_dirty_before_starting_transaction: None,
1141 has_unsaved_edits: Cell::new((buffer.version(), false)),
1142 text: buffer,
1143 branch_state: None,
1144 file,
1145 capability,
1146 syntax_map,
1147 reparse: None,
1148 non_text_state_update_count: 0,
1149 sync_parse_timeout: if cfg!(any(test, feature = "test-support")) {
1150 Some(Duration::from_millis(10))
1151 } else {
1152 Some(Duration::from_millis(1))
1153 },
1154 parse_status: watch::channel(ParseStatus::Idle),
1155 autoindent_requests: Default::default(),
1156 wait_for_autoindent_txs: Default::default(),
1157 pending_autoindent: Default::default(),
1158 language: None,
1159 remote_selections: Default::default(),
1160 diagnostics: Default::default(),
1161 diagnostics_timestamp: Lamport::MIN,
1162 completion_triggers: Default::default(),
1163 completion_triggers_per_language_server: Default::default(),
1164 completion_triggers_timestamp: Lamport::MIN,
1165 deferred_ops: OperationQueue::new(),
1166 has_conflict: false,
1167 change_bits: Default::default(),
1168 modeline: None,
1169 _subscriptions: Vec::new(),
1170 encoding: encoding_rs::UTF_8,
1171 has_bom: false,
1172 reload_with_encoding_txns: HashMap::default(),
1173 }
1174 }
1175
1176 #[ztracing::instrument(skip_all)]
1177 pub fn build_snapshot(
1178 text: Rope,
1179 language: Option<Arc<Language>>,
1180 language_registry: Option<Arc<LanguageRegistry>>,
1181 modeline: Option<Arc<ModelineSettings>>,
1182 cx: &mut App,
1183 ) -> impl Future<Output = BufferSnapshot> + use<> {
1184 let entity_id = cx.reserve_entity::<Self>().entity_id();
1185 let buffer_id = entity_id.as_non_zero_u64().into();
1186 async move {
1187 let text =
1188 TextBuffer::new_normalized(ReplicaId::LOCAL, buffer_id, Default::default(), text);
1189 let text = text.into_snapshot();
1190 let mut syntax = SyntaxMap::new(&text).snapshot();
1191 if let Some(language) = language.clone() {
1192 let language_registry = language_registry.clone();
1193 syntax.reparse(&text, language_registry, language);
1194 }
1195 let tree_sitter_data = TreeSitterData::new(&text);
1196 BufferSnapshot {
1197 text,
1198 syntax,
1199 file: None,
1200 diagnostics: Default::default(),
1201 remote_selections: Default::default(),
1202 tree_sitter_data: Arc::new(tree_sitter_data),
1203 language,
1204 non_text_state_update_count: 0,
1205 capability: Capability::ReadOnly,
1206 modeline,
1207 }
1208 }
1209 }
1210
1211 pub fn build_empty_snapshot(cx: &mut App) -> BufferSnapshot {
1212 let entity_id = cx.reserve_entity::<Self>().entity_id();
1213 let buffer_id = entity_id.as_non_zero_u64().into();
1214 let text = TextBuffer::new_normalized(
1215 ReplicaId::LOCAL,
1216 buffer_id,
1217 Default::default(),
1218 Rope::new(),
1219 );
1220 let text = text.into_snapshot();
1221 let syntax = SyntaxMap::new(&text).snapshot();
1222 let tree_sitter_data = TreeSitterData::new(&text);
1223 BufferSnapshot {
1224 text,
1225 syntax,
1226 tree_sitter_data: Arc::new(tree_sitter_data),
1227 file: None,
1228 diagnostics: Default::default(),
1229 remote_selections: Default::default(),
1230 language: None,
1231 non_text_state_update_count: 0,
1232 capability: Capability::ReadOnly,
1233 modeline: None,
1234 }
1235 }
1236
1237 #[cfg(any(test, feature = "test-support"))]
1238 pub fn build_snapshot_sync(
1239 text: Rope,
1240 language: Option<Arc<Language>>,
1241 language_registry: Option<Arc<LanguageRegistry>>,
1242 cx: &mut App,
1243 ) -> BufferSnapshot {
1244 let entity_id = cx.reserve_entity::<Self>().entity_id();
1245 let buffer_id = entity_id.as_non_zero_u64().into();
1246 let text =
1247 TextBuffer::new_normalized(ReplicaId::LOCAL, buffer_id, Default::default(), text)
1248 .into_snapshot();
1249 let mut syntax = SyntaxMap::new(&text).snapshot();
1250 if let Some(language) = language.clone() {
1251 syntax.reparse(&text, language_registry, language);
1252 }
1253 let tree_sitter_data = TreeSitterData::new(&text);
1254 BufferSnapshot {
1255 text,
1256 syntax,
1257 tree_sitter_data: Arc::new(tree_sitter_data),
1258 file: None,
1259 diagnostics: Default::default(),
1260 remote_selections: Default::default(),
1261 language,
1262 non_text_state_update_count: 0,
1263 capability: Capability::ReadOnly,
1264 modeline: None,
1265 }
1266 }
1267
1268 /// Retrieve a snapshot of the buffer's current state. This is computationally
1269 /// cheap, and allows reading from the buffer on a background thread.
1270 pub fn snapshot(&self) -> BufferSnapshot {
1271 let text = self.text.snapshot();
1272
1273 let syntax = {
1274 let mut syntax_map = self.syntax_map.lock();
1275 syntax_map.interpolate(text);
1276 syntax_map.snapshot()
1277 };
1278
1279 let tree_sitter_data = if self.text.version() != *self.tree_sitter_data.version() {
1280 Arc::new(TreeSitterData::new(text))
1281 } else {
1282 self.tree_sitter_data.clone()
1283 };
1284
1285 BufferSnapshot {
1286 text: text.clone(),
1287 syntax,
1288 tree_sitter_data,
1289 file: self.file.clone(),
1290 remote_selections: self.remote_selections.clone(),
1291 diagnostics: self.diagnostics.clone(),
1292 language: self.language.clone(),
1293 non_text_state_update_count: self.non_text_state_update_count,
1294 capability: self.capability,
1295 modeline: self.modeline.clone(),
1296 }
1297 }
1298
1299 pub fn branch(&mut self, cx: &mut Context<Self>) -> Entity<Self> {
1300 let this = cx.entity();
1301 cx.new(|cx| {
1302 let mut branch = Self {
1303 branch_state: Some(BufferBranchState {
1304 base_buffer: this.clone(),
1305 merged_operations: Default::default(),
1306 }),
1307 language: self.language.clone(),
1308 has_conflict: self.has_conflict,
1309 has_unsaved_edits: Cell::new(self.has_unsaved_edits.get_mut().clone()),
1310 _subscriptions: vec![cx.subscribe(&this, Self::on_base_buffer_event)],
1311 ..Self::build(self.text.branch(), self.file.clone(), self.capability())
1312 };
1313 if let Some(language_registry) = self.language_registry() {
1314 branch.set_language_registry(language_registry);
1315 }
1316
1317 // Reparse the branch buffer so that we get syntax highlighting immediately.
1318 branch.reparse(cx, true);
1319
1320 branch
1321 })
1322 }
1323
1324 #[ztracing::instrument(skip_all)]
1325 pub fn preview_edits(
1326 &self,
1327 edits: Arc<[(Range<Anchor>, Arc<str>)]>,
1328 cx: &App,
1329 ) -> Task<EditPreview> {
1330 let registry = self.language_registry();
1331 let language = self.language().cloned();
1332 let old_snapshot = self.text.snapshot().clone();
1333 let mut branch_buffer = self.text.branch();
1334 let mut syntax_snapshot = self.syntax_map.lock().snapshot();
1335 cx.background_spawn(async move {
1336 if !edits.is_empty() {
1337 if let Some(language) = language.clone() {
1338 syntax_snapshot.reparse(&old_snapshot, registry.clone(), language);
1339 }
1340
1341 branch_buffer.edit(edits.iter().cloned());
1342 let snapshot = branch_buffer.snapshot();
1343 syntax_snapshot.interpolate(&snapshot);
1344
1345 if let Some(language) = language {
1346 syntax_snapshot.reparse(&snapshot, registry, language);
1347 }
1348 }
1349 EditPreview {
1350 old_snapshot,
1351 applied_edits_snapshot: branch_buffer.into_snapshot(),
1352 syntax_snapshot,
1353 }
1354 })
1355 }
1356
1357 /// Applies all of the changes in this buffer that intersect any of the
1358 /// given `ranges` to its base buffer.
1359 ///
1360 /// If `ranges` is empty, then all changes will be applied. This buffer must
1361 /// be a branch buffer to call this method.
1362 pub fn merge_into_base(&mut self, ranges: Vec<Range<usize>>, cx: &mut Context<Self>) {
1363 let Some(base_buffer) = self.base_buffer() else {
1364 debug_panic!("not a branch buffer");
1365 return;
1366 };
1367
1368 let mut ranges = if ranges.is_empty() {
1369 &[0..usize::MAX]
1370 } else {
1371 ranges.as_slice()
1372 }
1373 .iter()
1374 .peekable();
1375
1376 let mut edits = Vec::new();
1377 for edit in self.edits_since::<usize>(&base_buffer.read(cx).version()) {
1378 let mut is_included = false;
1379 while let Some(range) = ranges.peek() {
1380 if range.end < edit.new.start {
1381 ranges.next().unwrap();
1382 } else {
1383 if range.start <= edit.new.end {
1384 is_included = true;
1385 }
1386 break;
1387 }
1388 }
1389
1390 if is_included {
1391 edits.push((
1392 edit.old.clone(),
1393 self.text_for_range(edit.new.clone()).collect::<String>(),
1394 ));
1395 }
1396 }
1397
1398 let operation = base_buffer.update(cx, |base_buffer, cx| {
1399 // cx.emit(BufferEvent::DiffBaseChanged);
1400 base_buffer.edit(edits, None, cx)
1401 });
1402
1403 if let Some(operation) = operation
1404 && let Some(BufferBranchState {
1405 merged_operations, ..
1406 }) = &mut self.branch_state
1407 {
1408 merged_operations.push(operation);
1409 }
1410 }
1411
1412 fn on_base_buffer_event(
1413 &mut self,
1414 _: Entity<Buffer>,
1415 event: &BufferEvent,
1416 cx: &mut Context<Self>,
1417 ) {
1418 let BufferEvent::Operation { operation, .. } = event else {
1419 return;
1420 };
1421 let Some(BufferBranchState {
1422 merged_operations, ..
1423 }) = &mut self.branch_state
1424 else {
1425 return;
1426 };
1427
1428 let mut operation_to_undo = None;
1429 if let Operation::Buffer(text::Operation::Edit(operation)) = &operation
1430 && let Ok(ix) = merged_operations.binary_search(&operation.timestamp)
1431 {
1432 merged_operations.remove(ix);
1433 operation_to_undo = Some(operation.timestamp);
1434 }
1435
1436 self.apply_ops([operation.clone()], cx);
1437
1438 if let Some(timestamp) = operation_to_undo {
1439 let counts = [(timestamp, u32::MAX)].into_iter().collect();
1440 self.undo_operations(counts, cx);
1441 }
1442 }
1443
1444 pub fn as_text_snapshot(&self) -> &text::BufferSnapshot {
1445 &self.text
1446 }
1447
1448 /// Retrieve a snapshot of the buffer's raw text, without any
1449 /// language-related state like the syntax tree or diagnostics.
1450 #[ztracing::instrument(skip_all)]
1451 pub fn text_snapshot(&self) -> text::BufferSnapshot {
1452 // todo lw
1453 self.text.snapshot().clone()
1454 }
1455
1456 /// The file associated with the buffer, if any.
1457 pub fn file(&self) -> Option<&Arc<dyn File>> {
1458 self.file.as_ref()
1459 }
1460
1461 /// The version of the buffer that was last saved or reloaded from disk.
1462 pub fn saved_version(&self) -> &clock::Global {
1463 &self.saved_version
1464 }
1465
1466 /// The mtime of the buffer's file when the buffer was last saved or reloaded from disk.
1467 pub fn saved_mtime(&self) -> Option<MTime> {
1468 self.saved_mtime
1469 }
1470
1471 /// Returns the character encoding of the buffer's file.
1472 pub fn encoding(&self) -> &'static Encoding {
1473 self.encoding
1474 }
1475
1476 /// Sets the character encoding of the buffer.
1477 pub fn set_encoding(&mut self, encoding: &'static Encoding) {
1478 self.encoding = encoding;
1479 }
1480
1481 /// Returns whether the buffer has a Byte Order Mark.
1482 pub fn has_bom(&self) -> bool {
1483 self.has_bom
1484 }
1485
1486 /// Sets whether the buffer has a Byte Order Mark.
1487 pub fn set_has_bom(&mut self, has_bom: bool) {
1488 self.has_bom = has_bom;
1489 }
1490
1491 /// Assign a language to the buffer.
1492 pub fn set_language_async(&mut self, language: Option<Arc<Language>>, cx: &mut Context<Self>) {
1493 self.set_language_(language, cfg!(any(test, feature = "test-support")), cx);
1494 }
1495
1496 /// Assign a language to the buffer, blocking for up to 1ms to reparse the buffer.
1497 pub fn set_language(&mut self, language: Option<Arc<Language>>, cx: &mut Context<Self>) {
1498 self.set_language_(language, true, cx);
1499 }
1500
1501 #[ztracing::instrument(skip_all)]
1502 fn set_language_(
1503 &mut self,
1504 language: Option<Arc<Language>>,
1505 may_block: bool,
1506 cx: &mut Context<Self>,
1507 ) {
1508 if language == self.language {
1509 return;
1510 }
1511 self.non_text_state_update_count += 1;
1512 self.syntax_map.lock().clear(&self.text);
1513 let old_language = std::mem::replace(&mut self.language, language);
1514 self.was_changed();
1515 self.reparse(cx, may_block);
1516 let has_fresh_language =
1517 self.language.is_some() && old_language.is_none_or(|old| old == *PLAIN_TEXT);
1518 cx.emit(BufferEvent::LanguageChanged(has_fresh_language));
1519 }
1520
1521 /// Assign a language registry to the buffer. This allows the buffer to retrieve
1522 /// other languages if parts of the buffer are written in different languages.
1523 pub fn set_language_registry(&self, language_registry: Arc<LanguageRegistry>) {
1524 self.syntax_map
1525 .lock()
1526 .set_language_registry(language_registry);
1527 }
1528
1529 pub fn language_registry(&self) -> Option<Arc<LanguageRegistry>> {
1530 self.syntax_map.lock().language_registry()
1531 }
1532
1533 /// Assign the line ending type to the buffer.
1534 pub fn set_line_ending(&mut self, line_ending: LineEnding, cx: &mut Context<Self>) {
1535 self.text.set_line_ending(line_ending);
1536
1537 let lamport_timestamp = self.text.lamport_clock.tick();
1538 self.send_operation(
1539 Operation::UpdateLineEnding {
1540 line_ending,
1541 lamport_timestamp,
1542 },
1543 true,
1544 cx,
1545 );
1546 }
1547
1548 /// Assign the buffer [`ModelineSettings`].
1549 pub fn set_modeline(&mut self, modeline: Option<ModelineSettings>) -> bool {
1550 if modeline.as_ref() != self.modeline.as_deref() {
1551 self.modeline = modeline.map(Arc::new);
1552 true
1553 } else {
1554 false
1555 }
1556 }
1557
1558 /// Returns the [`ModelineSettings`].
1559 pub fn modeline(&self) -> Option<&Arc<ModelineSettings>> {
1560 self.modeline.as_ref()
1561 }
1562
1563 /// Assign the buffer a new [`Capability`].
1564 pub fn set_capability(&mut self, capability: Capability, cx: &mut Context<Self>) {
1565 if self.capability != capability {
1566 self.capability = capability;
1567 cx.emit(BufferEvent::CapabilityChanged)
1568 }
1569 }
1570
1571 /// This method is called to signal that the buffer has been saved.
1572 pub fn did_save(
1573 &mut self,
1574 version: clock::Global,
1575 mtime: Option<MTime>,
1576 cx: &mut Context<Self>,
1577 ) {
1578 self.saved_version = version.clone();
1579 self.has_unsaved_edits.set((version, false));
1580 self.has_conflict = false;
1581 self.saved_mtime = mtime;
1582 self.was_changed();
1583 cx.emit(BufferEvent::Saved);
1584 cx.notify();
1585 }
1586
1587 /// Reloads the contents of the buffer from disk.
1588 pub fn reload(&mut self, cx: &Context<Self>) -> oneshot::Receiver<Option<Transaction>> {
1589 self.reload_impl(None, cx)
1590 }
1591
1592 /// Reloads the contents of the buffer from disk using the specified encoding.
1593 ///
1594 /// This bypasses automatic encoding detection heuristics (like BOM checks) for non-Unicode encodings,
1595 /// allowing users to force a specific interpretation of the bytes.
1596 pub fn reload_with_encoding(
1597 &mut self,
1598 encoding: &'static Encoding,
1599 cx: &Context<Self>,
1600 ) -> oneshot::Receiver<Option<Transaction>> {
1601 self.reload_impl(Some(encoding), cx)
1602 }
1603
1604 fn reload_impl(
1605 &mut self,
1606 force_encoding: Option<&'static Encoding>,
1607 cx: &Context<Self>,
1608 ) -> oneshot::Receiver<Option<Transaction>> {
1609 let (tx, rx) = futures::channel::oneshot::channel();
1610 let prev_version = self.text.version();
1611
1612 self.reload_task = Some(cx.spawn(async move |this, cx| {
1613 let Some((new_mtime, load_bytes_task, current_encoding)) =
1614 this.update(cx, |this, cx| {
1615 let file = this.file.as_ref()?.as_local()?;
1616 Some((
1617 file.disk_state().mtime(),
1618 file.load_bytes(cx),
1619 this.encoding,
1620 ))
1621 })?
1622 else {
1623 return Ok(());
1624 };
1625
1626 let target_encoding = force_encoding.unwrap_or(current_encoding);
1627
1628 let is_unicode = target_encoding == encoding_rs::UTF_8
1629 || target_encoding == encoding_rs::UTF_16LE
1630 || target_encoding == encoding_rs::UTF_16BE;
1631
1632 let (new_text, has_bom, encoding_used) = if force_encoding.is_some() && !is_unicode {
1633 let bytes = load_bytes_task.await?;
1634 let (cow, _had_errors) = target_encoding.decode_without_bom_handling(&bytes);
1635 (cow.into_owned(), false, target_encoding)
1636 } else {
1637 let bytes = load_bytes_task.await?;
1638 let (cow, used_enc, _had_errors) = target_encoding.decode(&bytes);
1639
1640 let actual_has_bom = if used_enc == encoding_rs::UTF_8 {
1641 bytes.starts_with(&[0xEF, 0xBB, 0xBF])
1642 } else if used_enc == encoding_rs::UTF_16LE {
1643 bytes.starts_with(&[0xFF, 0xFE])
1644 } else if used_enc == encoding_rs::UTF_16BE {
1645 bytes.starts_with(&[0xFE, 0xFF])
1646 } else {
1647 false
1648 };
1649 (cow.into_owned(), actual_has_bom, used_enc)
1650 };
1651
1652 let diff = this.update(cx, |this, cx| this.diff(new_text, cx))?.await;
1653 this.update(cx, |this, cx| {
1654 if this.version() == diff.base_version {
1655 this.finalize_last_transaction();
1656 let old_encoding = this.encoding;
1657 let old_has_bom = this.has_bom;
1658 this.apply_diff(diff, cx);
1659 this.encoding = encoding_used;
1660 this.has_bom = has_bom;
1661 let transaction = this.finalize_last_transaction().cloned();
1662 if let Some(ref txn) = transaction {
1663 if old_encoding != encoding_used || old_has_bom != has_bom {
1664 this.reload_with_encoding_txns
1665 .insert(txn.id, (old_encoding, old_has_bom));
1666 }
1667 }
1668 tx.send(transaction).ok();
1669 this.has_conflict = false;
1670 this.did_reload(this.version(), this.line_ending(), new_mtime, cx);
1671 } else {
1672 if !diff.edits.is_empty()
1673 || this
1674 .edits_since::<usize>(&diff.base_version)
1675 .next()
1676 .is_some()
1677 {
1678 this.has_conflict = true;
1679 }
1680
1681 this.did_reload(prev_version, this.line_ending(), this.saved_mtime, cx);
1682 }
1683
1684 this.reload_task.take();
1685 })
1686 }));
1687 rx
1688 }
1689
1690 /// This method is called to signal that the buffer has been reloaded.
1691 pub fn did_reload(
1692 &mut self,
1693 version: clock::Global,
1694 line_ending: LineEnding,
1695 mtime: Option<MTime>,
1696 cx: &mut Context<Self>,
1697 ) {
1698 self.saved_version = version;
1699 self.has_unsaved_edits
1700 .set((self.saved_version.clone(), false));
1701 self.text.set_line_ending(line_ending);
1702 self.saved_mtime = mtime;
1703 cx.emit(BufferEvent::Reloaded);
1704 cx.notify();
1705 }
1706
1707 /// Updates the [`File`] backing this buffer. This should be called when
1708 /// the file has changed or has been deleted.
1709 pub fn file_updated(&mut self, new_file: Arc<dyn File>, cx: &mut Context<Self>) {
1710 let was_dirty = self.is_dirty();
1711 let mut file_changed = false;
1712
1713 if let Some(old_file) = self.file.as_ref() {
1714 if new_file.path() != old_file.path() {
1715 file_changed = true;
1716 }
1717
1718 let old_state = old_file.disk_state();
1719 let new_state = new_file.disk_state();
1720 if old_state != new_state {
1721 file_changed = true;
1722 if !was_dirty && matches!(new_state, DiskState::Present { .. }) {
1723 cx.emit(BufferEvent::ReloadNeeded)
1724 }
1725 }
1726 } else {
1727 file_changed = true;
1728 };
1729
1730 self.file = Some(new_file);
1731 if file_changed {
1732 self.was_changed();
1733 self.non_text_state_update_count += 1;
1734 if was_dirty != self.is_dirty() {
1735 cx.emit(BufferEvent::DirtyChanged);
1736 }
1737 cx.emit(BufferEvent::FileHandleChanged);
1738 cx.notify();
1739 }
1740 }
1741
1742 pub fn base_buffer(&self) -> Option<Entity<Self>> {
1743 Some(self.branch_state.as_ref()?.base_buffer.clone())
1744 }
1745
1746 /// Returns the primary [`Language`] assigned to this [`Buffer`].
1747 pub fn language(&self) -> Option<&Arc<Language>> {
1748 self.language.as_ref()
1749 }
1750
1751 /// Returns the [`Language`] at the given location.
1752 pub fn language_at<D: ToOffset>(&self, position: D) -> Option<Arc<Language>> {
1753 let offset = position.to_offset(self);
1754 let text: &TextBufferSnapshot = &self.text;
1755 self.syntax_map
1756 .lock()
1757 .layers_for_range(offset..offset, text, false)
1758 .filter(|layer| {
1759 layer
1760 .included_sub_ranges
1761 .is_none_or(|ranges| offset_in_sub_ranges(ranges, offset, text))
1762 })
1763 .last()
1764 .map(|info| info.language.clone())
1765 .or_else(|| self.language.clone())
1766 }
1767
1768 /// Returns each [`Language`] for the active syntax layers at the given location.
1769 pub fn languages_at<D: ToOffset>(&self, position: D) -> Vec<Arc<Language>> {
1770 let offset = position.to_offset(self);
1771 let text: &TextBufferSnapshot = &self.text;
1772 let mut languages: Vec<Arc<Language>> = self
1773 .syntax_map
1774 .lock()
1775 .layers_for_range(offset..offset, text, false)
1776 .filter(|layer| {
1777 // For combined injections, check if offset is within the actual sub-ranges.
1778 layer
1779 .included_sub_ranges
1780 .is_none_or(|ranges| offset_in_sub_ranges(ranges, offset, text))
1781 })
1782 .map(|info| info.language.clone())
1783 .collect();
1784
1785 if languages.is_empty()
1786 && let Some(buffer_language) = self.language()
1787 {
1788 languages.push(buffer_language.clone());
1789 }
1790
1791 languages
1792 }
1793
1794 /// An integer version number that accounts for all updates besides
1795 /// the buffer's text itself (which is versioned via a version vector).
1796 pub fn non_text_state_update_count(&self) -> usize {
1797 self.non_text_state_update_count
1798 }
1799
1800 /// Whether the buffer is being parsed in the background.
1801 #[cfg(any(test, feature = "test-support"))]
1802 pub fn is_parsing(&self) -> bool {
1803 self.reparse.is_some()
1804 }
1805
1806 /// Indicates whether the buffer contains any regions that may be
1807 /// written in a language that hasn't been loaded yet.
1808 pub fn contains_unknown_injections(&self) -> bool {
1809 self.syntax_map.lock().contains_unknown_injections()
1810 }
1811
1812 /// Sets the sync parse timeout for this buffer.
1813 ///
1814 /// Setting this to `None` disables sync parsing entirely.
1815 pub fn set_sync_parse_timeout(&mut self, timeout: Option<Duration>) {
1816 self.sync_parse_timeout = timeout;
1817 }
1818
1819 fn invalidate_tree_sitter_data(
1820 tree_sitter_data: &mut Arc<TreeSitterData>,
1821 snapshot: &text::BufferSnapshot,
1822 ) {
1823 match Arc::get_mut(tree_sitter_data) {
1824 Some(tree_sitter_data) => tree_sitter_data.clear(snapshot),
1825 None => {
1826 let new_tree_sitter_data = TreeSitterData::new(snapshot);
1827 *tree_sitter_data = Arc::new(new_tree_sitter_data)
1828 }
1829 }
1830 }
1831
1832 /// Called after an edit to synchronize the buffer's main parse tree with
1833 /// the buffer's new underlying state.
1834 ///
1835 /// Locks the syntax map and interpolates the edits since the last reparse
1836 /// into the foreground syntax tree.
1837 ///
1838 /// Then takes a stable snapshot of the syntax map before unlocking it.
1839 /// The snapshot with the interpolated edits is sent to a background thread,
1840 /// where we ask Tree-sitter to perform an incremental parse.
1841 ///
1842 /// Meanwhile, in the foreground if `may_block` is true, we block the main
1843 /// thread for up to 1ms waiting on the parse to complete. As soon as it
1844 /// completes, we proceed synchronously, unless a 1ms timeout elapses.
1845 ///
1846 /// If we time out waiting on the parse, we spawn a second task waiting
1847 /// until the parse does complete and return with the interpolated tree still
1848 /// in the foreground. When the background parse completes, call back into
1849 /// the main thread and assign the foreground parse state.
1850 ///
1851 /// If the buffer or grammar changed since the start of the background parse,
1852 /// initiate an additional reparse recursively. To avoid concurrent parses
1853 /// for the same buffer, we only initiate a new parse if we are not already
1854 /// parsing in the background.
1855 #[ztracing::instrument(skip_all)]
1856 pub fn reparse(&mut self, cx: &mut Context<Self>, may_block: bool) {
1857 if self.text.version() != *self.tree_sitter_data.version() {
1858 Self::invalidate_tree_sitter_data(&mut self.tree_sitter_data, self.text.snapshot());
1859 }
1860 if self.reparse.is_some() {
1861 return;
1862 }
1863 let language = if let Some(language) = self.language.clone() {
1864 language
1865 } else {
1866 return;
1867 };
1868
1869 let text = self.text_snapshot();
1870 let parsed_version = self.version();
1871
1872 let mut syntax_map = self.syntax_map.lock();
1873 syntax_map.interpolate(&text);
1874 let language_registry = syntax_map.language_registry();
1875 let mut syntax_snapshot = syntax_map.snapshot();
1876 drop(syntax_map);
1877
1878 self.parse_status.0.send(ParseStatus::Parsing).unwrap();
1879 if may_block && let Some(sync_parse_timeout) = self.sync_parse_timeout {
1880 if let Ok(()) = syntax_snapshot.reparse_with_timeout(
1881 &text,
1882 language_registry.clone(),
1883 language.clone(),
1884 sync_parse_timeout,
1885 ) {
1886 self.did_finish_parsing(syntax_snapshot, Some(Duration::from_millis(300)), cx);
1887 self.reparse = None;
1888 return;
1889 }
1890 }
1891
1892 let parse_task = cx.background_spawn({
1893 let language = language.clone();
1894 let language_registry = language_registry.clone();
1895 async move {
1896 syntax_snapshot.reparse(&text, language_registry, language);
1897 syntax_snapshot
1898 }
1899 });
1900
1901 self.reparse = Some(cx.spawn(async move |this, cx| {
1902 let new_syntax_map = parse_task.await;
1903 this.update(cx, move |this, cx| {
1904 let grammar_changed = || {
1905 this.language
1906 .as_ref()
1907 .is_none_or(|current_language| !Arc::ptr_eq(&language, current_language))
1908 };
1909 let language_registry_changed = || {
1910 new_syntax_map.contains_unknown_injections()
1911 && language_registry.is_some_and(|registry| {
1912 registry.version() != new_syntax_map.language_registry_version()
1913 })
1914 };
1915 let parse_again = this.version.changed_since(&parsed_version)
1916 || language_registry_changed()
1917 || grammar_changed();
1918 this.did_finish_parsing(new_syntax_map, None, cx);
1919 this.reparse = None;
1920 if parse_again {
1921 this.reparse(cx, false);
1922 }
1923 })
1924 .ok();
1925 }));
1926 }
1927
1928 fn did_finish_parsing(
1929 &mut self,
1930 syntax_snapshot: SyntaxSnapshot,
1931 block_budget: Option<Duration>,
1932 cx: &mut Context<Self>,
1933 ) {
1934 self.non_text_state_update_count += 1;
1935 self.syntax_map.lock().did_parse(syntax_snapshot);
1936 self.was_changed();
1937 self.request_autoindent(cx, block_budget);
1938 self.parse_status.0.send(ParseStatus::Idle).unwrap();
1939 Self::invalidate_tree_sitter_data(&mut self.tree_sitter_data, &self.text.snapshot());
1940 cx.emit(BufferEvent::Reparsed);
1941 cx.notify();
1942 }
1943
1944 pub fn parse_status(&self) -> watch::Receiver<ParseStatus> {
1945 self.parse_status.1.clone()
1946 }
1947
1948 /// Wait until the buffer is no longer parsing
1949 pub fn parsing_idle(&self) -> impl Future<Output = ()> + use<> {
1950 let mut parse_status = self.parse_status();
1951 async move {
1952 while *parse_status.borrow() != ParseStatus::Idle {
1953 if parse_status.changed().await.is_err() {
1954 break;
1955 }
1956 }
1957 }
1958 }
1959
1960 /// Assign to the buffer a set of diagnostics created by a given language server.
1961 pub fn update_diagnostics(
1962 &mut self,
1963 server_id: LanguageServerId,
1964 diagnostics: DiagnosticSet,
1965 cx: &mut Context<Self>,
1966 ) {
1967 let lamport_timestamp = self.text.lamport_clock.tick();
1968 let op = Operation::UpdateDiagnostics {
1969 server_id,
1970 diagnostics: diagnostics.iter().cloned().collect(),
1971 lamport_timestamp,
1972 };
1973
1974 self.apply_diagnostic_update(server_id, diagnostics, lamport_timestamp, cx);
1975 self.send_operation(op, true, cx);
1976 }
1977
1978 pub fn buffer_diagnostics(
1979 &self,
1980 for_server: Option<LanguageServerId>,
1981 ) -> Vec<&DiagnosticEntry<Anchor>> {
1982 match for_server {
1983 Some(server_id) => self
1984 .diagnostics
1985 .get(&server_id)
1986 .map_or_else(Vec::new, |diagnostics| diagnostics.iter().collect()),
1987 None => self
1988 .diagnostics
1989 .iter()
1990 .flat_map(|(_, diagnostic_set)| diagnostic_set.iter())
1991 .collect(),
1992 }
1993 }
1994
1995 fn request_autoindent(&mut self, cx: &mut Context<Self>, block_budget: Option<Duration>) {
1996 if let Some(indent_sizes) = self.compute_autoindents() {
1997 let indent_sizes = cx.background_spawn(indent_sizes);
1998 let Some(block_budget) = block_budget else {
1999 self.pending_autoindent = Some(cx.spawn(async move |this, cx| {
2000 let indent_sizes = indent_sizes.await;
2001 this.update(cx, |this, cx| {
2002 this.apply_autoindents(indent_sizes, cx);
2003 })
2004 .ok();
2005 }));
2006 return;
2007 };
2008 match cx
2009 .foreground_executor()
2010 .block_with_timeout(block_budget, indent_sizes)
2011 {
2012 Ok(indent_sizes) => self.apply_autoindents(indent_sizes, cx),
2013 Err(indent_sizes) => {
2014 self.pending_autoindent = Some(cx.spawn(async move |this, cx| {
2015 let indent_sizes = indent_sizes.await;
2016 this.update(cx, |this, cx| {
2017 this.apply_autoindents(indent_sizes, cx);
2018 })
2019 .ok();
2020 }));
2021 }
2022 }
2023 } else {
2024 self.autoindent_requests.clear();
2025 for tx in self.wait_for_autoindent_txs.drain(..) {
2026 tx.send(()).ok();
2027 }
2028 }
2029 }
2030
2031 fn compute_autoindents(
2032 &self,
2033 ) -> Option<impl Future<Output = BTreeMap<u32, IndentSize>> + use<>> {
2034 let max_rows_between_yields = 100;
2035 let snapshot = self.snapshot();
2036 if snapshot.syntax.is_empty() || self.autoindent_requests.is_empty() {
2037 return None;
2038 }
2039
2040 let autoindent_requests = self.autoindent_requests.clone();
2041 Some(async move {
2042 let mut indent_sizes = BTreeMap::<u32, (IndentSize, bool)>::new();
2043 for request in autoindent_requests {
2044 // Resolve each edited range to its row in the current buffer and in the
2045 // buffer before this batch of edits.
2046 let mut row_ranges = Vec::new();
2047 let mut old_to_new_rows = BTreeMap::new();
2048 let mut language_indent_sizes_by_new_row = Vec::new();
2049 for entry in &request.entries {
2050 let position = entry.range.start;
2051 let new_row = position.to_point(&snapshot).row;
2052 let new_end_row = entry.range.end.to_point(&snapshot).row + 1;
2053 language_indent_sizes_by_new_row.push((new_row, entry.indent_size));
2054
2055 if let Some(old_row) = entry.old_row {
2056 old_to_new_rows.insert(old_row, new_row);
2057 }
2058 row_ranges.push((new_row..new_end_row, entry.original_indent_column));
2059 }
2060
2061 // Build a map containing the suggested indentation for each of the edited lines
2062 // with respect to the state of the buffer before these edits. This map is keyed
2063 // by the rows for these lines in the current state of the buffer.
2064 let mut old_suggestions = BTreeMap::<u32, (IndentSize, bool)>::default();
2065 let old_edited_ranges =
2066 contiguous_ranges(old_to_new_rows.keys().copied(), max_rows_between_yields);
2067 let mut language_indent_sizes = language_indent_sizes_by_new_row.iter().peekable();
2068 let mut language_indent_size = IndentSize::default();
2069 for old_edited_range in old_edited_ranges {
2070 let suggestions = request
2071 .before_edit
2072 .suggest_autoindents(old_edited_range.clone())
2073 .into_iter()
2074 .flatten();
2075 for (old_row, suggestion) in old_edited_range.zip(suggestions) {
2076 if let Some(suggestion) = suggestion {
2077 let new_row = *old_to_new_rows.get(&old_row).unwrap();
2078
2079 // Find the indent size based on the language for this row.
2080 while let Some((row, size)) = language_indent_sizes.peek() {
2081 if *row > new_row {
2082 break;
2083 }
2084 language_indent_size = *size;
2085 language_indent_sizes.next();
2086 }
2087
2088 let suggested_indent = old_to_new_rows
2089 .get(&suggestion.basis_row)
2090 .and_then(|from_row| {
2091 Some(old_suggestions.get(from_row).copied()?.0)
2092 })
2093 .unwrap_or_else(|| {
2094 request
2095 .before_edit
2096 .indent_size_for_line(suggestion.basis_row)
2097 })
2098 .with_delta(suggestion.delta, language_indent_size);
2099 old_suggestions
2100 .insert(new_row, (suggested_indent, suggestion.within_error));
2101 }
2102 }
2103 yield_now().await;
2104 }
2105
2106 // Compute new suggestions for each line, but only include them in the result
2107 // if they differ from the old suggestion for that line.
2108 let mut language_indent_sizes = language_indent_sizes_by_new_row.iter().peekable();
2109 let mut language_indent_size = IndentSize::default();
2110 for (row_range, original_indent_column) in row_ranges {
2111 let new_edited_row_range = if request.is_block_mode {
2112 row_range.start..row_range.start + 1
2113 } else {
2114 row_range.clone()
2115 };
2116
2117 let suggestions = snapshot
2118 .suggest_autoindents(new_edited_row_range.clone())
2119 .into_iter()
2120 .flatten();
2121 for (new_row, suggestion) in new_edited_row_range.zip(suggestions) {
2122 if let Some(suggestion) = suggestion {
2123 // Find the indent size based on the language for this row.
2124 while let Some((row, size)) = language_indent_sizes.peek() {
2125 if *row > new_row {
2126 break;
2127 }
2128 language_indent_size = *size;
2129 language_indent_sizes.next();
2130 }
2131
2132 let suggested_indent = indent_sizes
2133 .get(&suggestion.basis_row)
2134 .copied()
2135 .map(|e| e.0)
2136 .unwrap_or_else(|| {
2137 snapshot.indent_size_for_line(suggestion.basis_row)
2138 })
2139 .with_delta(suggestion.delta, language_indent_size);
2140
2141 if old_suggestions.get(&new_row).is_none_or(
2142 |(old_indentation, was_within_error)| {
2143 suggested_indent != *old_indentation
2144 && (!suggestion.within_error || *was_within_error)
2145 },
2146 ) {
2147 indent_sizes.insert(
2148 new_row,
2149 (suggested_indent, request.ignore_empty_lines),
2150 );
2151 }
2152 }
2153 }
2154
2155 if let (true, Some(original_indent_column)) =
2156 (request.is_block_mode, original_indent_column)
2157 {
2158 let new_indent =
2159 if let Some((indent, _)) = indent_sizes.get(&row_range.start) {
2160 *indent
2161 } else {
2162 snapshot.indent_size_for_line(row_range.start)
2163 };
2164 let delta = new_indent.len as i64 - original_indent_column as i64;
2165 if delta != 0 {
2166 for row in row_range.skip(1) {
2167 indent_sizes.entry(row).or_insert_with(|| {
2168 let mut size = snapshot.indent_size_for_line(row);
2169 if size.kind == new_indent.kind {
2170 match delta.cmp(&0) {
2171 Ordering::Greater => size.len += delta as u32,
2172 Ordering::Less => {
2173 size.len = size.len.saturating_sub(-delta as u32)
2174 }
2175 Ordering::Equal => {}
2176 }
2177 }
2178 (size, request.ignore_empty_lines)
2179 });
2180 }
2181 }
2182 }
2183
2184 yield_now().await;
2185 }
2186 }
2187
2188 indent_sizes
2189 .into_iter()
2190 .filter_map(|(row, (indent, ignore_empty_lines))| {
2191 if ignore_empty_lines && snapshot.line_len(row) == 0 {
2192 None
2193 } else {
2194 Some((row, indent))
2195 }
2196 })
2197 .collect()
2198 })
2199 }
2200
2201 fn apply_autoindents(
2202 &mut self,
2203 indent_sizes: BTreeMap<u32, IndentSize>,
2204 cx: &mut Context<Self>,
2205 ) {
2206 self.autoindent_requests.clear();
2207 for tx in self.wait_for_autoindent_txs.drain(..) {
2208 tx.send(()).ok();
2209 }
2210
2211 let edits: Vec<_> = indent_sizes
2212 .into_iter()
2213 .filter_map(|(row, indent_size)| {
2214 let current_size = indent_size_for_line(self, row);
2215 Self::edit_for_indent_size_adjustment(row, current_size, indent_size)
2216 })
2217 .collect();
2218
2219 let preserve_preview = self.preserve_preview();
2220 self.edit(edits, None, cx);
2221 if preserve_preview {
2222 self.refresh_preview();
2223 }
2224 }
2225
2226 /// Create a minimal edit that will cause the given row to be indented
2227 /// with the given size. After applying this edit, the length of the line
2228 /// will always be at least `new_size.len`.
2229 pub fn edit_for_indent_size_adjustment(
2230 row: u32,
2231 current_size: IndentSize,
2232 new_size: IndentSize,
2233 ) -> Option<(Range<Point>, String)> {
2234 if new_size.kind == current_size.kind {
2235 match new_size.len.cmp(¤t_size.len) {
2236 Ordering::Greater => {
2237 let point = Point::new(row, 0);
2238 Some((
2239 point..point,
2240 iter::repeat(new_size.char())
2241 .take((new_size.len - current_size.len) as usize)
2242 .collect::<String>(),
2243 ))
2244 }
2245
2246 Ordering::Less => Some((
2247 Point::new(row, 0)..Point::new(row, current_size.len - new_size.len),
2248 String::new(),
2249 )),
2250
2251 Ordering::Equal => None,
2252 }
2253 } else {
2254 Some((
2255 Point::new(row, 0)..Point::new(row, current_size.len),
2256 iter::repeat(new_size.char())
2257 .take(new_size.len as usize)
2258 .collect::<String>(),
2259 ))
2260 }
2261 }
2262
2263 /// Spawns a background task that asynchronously computes a `Diff` between the buffer's text
2264 /// and the given new text.
2265 pub fn diff<T>(&self, new_text: T, cx: &App) -> Task<Diff>
2266 where
2267 T: AsRef<str> + Send + 'static,
2268 {
2269 let old_text = self.as_rope().clone();
2270 let base_version = self.version();
2271 cx.background_spawn(async move {
2272 let old_text = old_text.to_string();
2273 let mut new_text = new_text.as_ref().to_owned();
2274 let line_ending = LineEnding::detect(&new_text);
2275 LineEnding::normalize(&mut new_text);
2276 let edits = text_diff(&old_text, &new_text);
2277 Diff {
2278 base_version,
2279 line_ending,
2280 edits,
2281 }
2282 })
2283 }
2284
2285 /// Spawns a background task that searches the buffer for any whitespace
2286 /// at the ends of a lines, and returns a `Diff` that removes that whitespace.
2287 pub fn remove_trailing_whitespace(&self, cx: &App) -> Task<Diff> {
2288 let old_text = self.as_rope().clone();
2289 let line_ending = self.line_ending();
2290 let base_version = self.version();
2291 cx.background_spawn(async move {
2292 let ranges = trailing_whitespace_ranges(&old_text);
2293 let empty = Arc::<str>::from("");
2294 Diff {
2295 base_version,
2296 line_ending,
2297 edits: ranges
2298 .into_iter()
2299 .map(|range| (range, empty.clone()))
2300 .collect(),
2301 }
2302 })
2303 }
2304
2305 /// Ensures that the buffer ends with a single newline character, and
2306 /// no other whitespace. Skips if the buffer is empty.
2307 pub fn ensure_final_newline(&mut self, cx: &mut Context<Self>) {
2308 let len = self.len();
2309 if len == 0 {
2310 return;
2311 }
2312 let mut offset = len;
2313 for chunk in self.as_rope().reversed_chunks_in_range(0..len) {
2314 let non_whitespace_len = chunk
2315 .trim_end_matches(|c: char| c.is_ascii_whitespace())
2316 .len();
2317 offset -= chunk.len();
2318 offset += non_whitespace_len;
2319 if non_whitespace_len != 0 {
2320 if offset == len - 1 && chunk.get(non_whitespace_len..) == Some("\n") {
2321 return;
2322 }
2323 break;
2324 }
2325 }
2326 self.edit([(offset..len, "\n")], None, cx);
2327 }
2328
2329 /// Applies a diff to the buffer. If the buffer has changed since the given diff was
2330 /// calculated, then adjust the diff to account for those changes, and discard any
2331 /// parts of the diff that conflict with those changes.
2332 pub fn apply_diff(&mut self, diff: Diff, cx: &mut Context<Self>) -> Option<TransactionId> {
2333 let snapshot = self.snapshot();
2334 let mut edits_since = snapshot.edits_since::<usize>(&diff.base_version).peekable();
2335 let mut delta = 0;
2336 let adjusted_edits = diff.edits.into_iter().filter_map(|(range, new_text)| {
2337 while let Some(edit_since) = edits_since.peek() {
2338 // If the edit occurs after a diff hunk, then it does not
2339 // affect that hunk.
2340 if edit_since.old.start > range.end {
2341 break;
2342 }
2343 // If the edit precedes the diff hunk, then adjust the hunk
2344 // to reflect the edit.
2345 else if edit_since.old.end < range.start {
2346 delta += edit_since.new_len() as i64 - edit_since.old_len() as i64;
2347 edits_since.next();
2348 }
2349 // If the edit intersects a diff hunk, then discard that hunk.
2350 else {
2351 return None;
2352 }
2353 }
2354
2355 let start = (range.start as i64 + delta) as usize;
2356 let end = (range.end as i64 + delta) as usize;
2357 Some((start..end, new_text))
2358 });
2359
2360 self.start_transaction();
2361 self.text.set_line_ending(diff.line_ending);
2362 self.edit(adjusted_edits, None, cx);
2363 self.end_transaction(cx)
2364 }
2365
2366 pub fn has_unsaved_edits(&self) -> bool {
2367 let (last_version, has_unsaved_edits) = self.has_unsaved_edits.take();
2368
2369 if last_version == self.version {
2370 self.has_unsaved_edits
2371 .set((last_version, has_unsaved_edits));
2372 return has_unsaved_edits;
2373 }
2374
2375 let has_edits = self.has_edits_since(&self.saved_version);
2376 self.has_unsaved_edits
2377 .set((self.version.clone(), has_edits));
2378 has_edits
2379 }
2380
2381 /// Checks if the buffer has unsaved changes.
2382 pub fn is_dirty(&self) -> bool {
2383 if self.capability == Capability::ReadOnly {
2384 return false;
2385 }
2386 if self.has_conflict {
2387 return true;
2388 }
2389 match self.file.as_ref().map(|f| f.disk_state()) {
2390 Some(DiskState::New) | Some(DiskState::Deleted) => {
2391 !self.is_empty() && self.has_unsaved_edits()
2392 }
2393 _ => self.has_unsaved_edits(),
2394 }
2395 }
2396
2397 /// Marks the buffer as having a conflict regardless of current buffer state.
2398 pub fn set_conflict(&mut self) {
2399 self.has_conflict = true;
2400 }
2401
2402 /// Checks if the buffer and its file have both changed since the buffer
2403 /// was last saved or reloaded.
2404 pub fn has_conflict(&self) -> bool {
2405 if self.has_conflict {
2406 return true;
2407 }
2408 let Some(file) = self.file.as_ref() else {
2409 return false;
2410 };
2411 match file.disk_state() {
2412 DiskState::New => false,
2413 DiskState::Present { mtime, .. } => match self.saved_mtime {
2414 Some(saved_mtime) => {
2415 mtime.bad_is_greater_than(saved_mtime) && self.has_unsaved_edits()
2416 }
2417 None => true,
2418 },
2419 DiskState::Deleted => false,
2420 DiskState::Historic { .. } => false,
2421 }
2422 }
2423
2424 /// Gets a [`Subscription`] that tracks all of the changes to the buffer's text.
2425 pub fn subscribe(&mut self) -> Subscription<usize> {
2426 self.text.subscribe()
2427 }
2428
2429 /// Adds a bit to the list of bits that are set when the buffer's text changes.
2430 ///
2431 /// This allows downstream code to check if the buffer's text has changed without
2432 /// waiting for an effect cycle, which would be required if using eents.
2433 pub fn record_changes(&mut self, bit: rc::Weak<Cell<bool>>) {
2434 if let Err(ix) = self
2435 .change_bits
2436 .binary_search_by_key(&rc::Weak::as_ptr(&bit), rc::Weak::as_ptr)
2437 {
2438 self.change_bits.insert(ix, bit);
2439 }
2440 }
2441
2442 /// Set the change bit for all "listeners".
2443 fn was_changed(&mut self) {
2444 self.change_bits.retain(|change_bit| {
2445 change_bit
2446 .upgrade()
2447 .inspect(|bit| {
2448 _ = bit.replace(true);
2449 })
2450 .is_some()
2451 });
2452 }
2453
2454 /// Starts a transaction, if one is not already in-progress. When undoing or
2455 /// redoing edits, all of the edits performed within a transaction are undone
2456 /// or redone together.
2457 pub fn start_transaction(&mut self) -> Option<TransactionId> {
2458 self.start_transaction_at(Instant::now())
2459 }
2460
2461 /// Starts a transaction, providing the current time. Subsequent transactions
2462 /// that occur within a short period of time will be grouped together. This
2463 /// is controlled by the buffer's undo grouping duration.
2464 pub fn start_transaction_at(&mut self, now: Instant) -> Option<TransactionId> {
2465 self.transaction_depth += 1;
2466 if self.was_dirty_before_starting_transaction.is_none() {
2467 self.was_dirty_before_starting_transaction = Some(self.is_dirty());
2468 }
2469 self.text.start_transaction_at(now)
2470 }
2471
2472 /// Terminates the current transaction, if this is the outermost transaction.
2473 pub fn end_transaction(&mut self, cx: &mut Context<Self>) -> Option<TransactionId> {
2474 self.end_transaction_at(Instant::now(), cx)
2475 }
2476
2477 /// Terminates the current transaction, providing the current time. Subsequent transactions
2478 /// that occur within a short period of time will be grouped together. This
2479 /// is controlled by the buffer's undo grouping duration.
2480 pub fn end_transaction_at(
2481 &mut self,
2482 now: Instant,
2483 cx: &mut Context<Self>,
2484 ) -> Option<TransactionId> {
2485 assert!(self.transaction_depth > 0);
2486 self.transaction_depth -= 1;
2487 let was_dirty = if self.transaction_depth == 0 {
2488 self.was_dirty_before_starting_transaction.take().unwrap()
2489 } else {
2490 false
2491 };
2492 if let Some((transaction_id, start_version)) = self.text.end_transaction_at(now) {
2493 self.did_edit(&start_version, was_dirty, true, cx);
2494 Some(transaction_id)
2495 } else {
2496 None
2497 }
2498 }
2499
2500 /// Manually add a transaction to the buffer's undo history.
2501 pub fn push_transaction(&mut self, transaction: Transaction, now: Instant) {
2502 self.text.push_transaction(transaction, now);
2503 }
2504
2505 /// Differs from `push_transaction` in that it does not clear the redo
2506 /// stack. Intended to be used to create a parent transaction to merge
2507 /// potential child transactions into.
2508 ///
2509 /// The caller is responsible for removing it from the undo history using
2510 /// `forget_transaction` if no edits are merged into it. Otherwise, if edits
2511 /// are merged into this transaction, the caller is responsible for ensuring
2512 /// the redo stack is cleared. The easiest way to ensure the redo stack is
2513 /// cleared is to create transactions with the usual `start_transaction` and
2514 /// `end_transaction` methods and merging the resulting transactions into
2515 /// the transaction created by this method
2516 pub fn push_empty_transaction(&mut self, now: Instant) -> TransactionId {
2517 self.text.push_empty_transaction(now)
2518 }
2519
2520 /// Prevent the last transaction from being grouped with any subsequent transactions,
2521 /// even if they occur with the buffer's undo grouping duration.
2522 pub fn finalize_last_transaction(&mut self) -> Option<&Transaction> {
2523 self.text.finalize_last_transaction()
2524 }
2525
2526 /// Manually group all changes since a given transaction.
2527 pub fn group_until_transaction(&mut self, transaction_id: TransactionId) {
2528 self.text.group_until_transaction(transaction_id);
2529 }
2530
2531 /// Manually remove a transaction from the buffer's undo history
2532 pub fn forget_transaction(&mut self, transaction_id: TransactionId) -> Option<Transaction> {
2533 self.text.forget_transaction(transaction_id)
2534 }
2535
2536 /// Retrieve a transaction from the buffer's undo history
2537 pub fn get_transaction(&self, transaction_id: TransactionId) -> Option<&Transaction> {
2538 self.text.get_transaction(transaction_id)
2539 }
2540
2541 /// Manually merge two transactions in the buffer's undo history.
2542 pub fn merge_transactions(&mut self, transaction: TransactionId, destination: TransactionId) {
2543 self.text.merge_transactions(transaction, destination);
2544 }
2545
2546 /// Waits for the buffer to receive operations with the given timestamps.
2547 pub fn wait_for_edits<It: IntoIterator<Item = clock::Lamport>>(
2548 &mut self,
2549 edit_ids: It,
2550 ) -> impl Future<Output = Result<()>> + use<It> {
2551 self.text.wait_for_edits(edit_ids)
2552 }
2553
2554 /// Waits for the buffer to receive the operations necessary for resolving the given anchors.
2555 pub fn wait_for_anchors<It: IntoIterator<Item = Anchor>>(
2556 &mut self,
2557 anchors: It,
2558 ) -> impl 'static + Future<Output = Result<()>> + use<It> {
2559 self.text.wait_for_anchors(anchors)
2560 }
2561
2562 /// Waits for the buffer to receive operations up to the given version.
2563 pub fn wait_for_version(
2564 &mut self,
2565 version: clock::Global,
2566 ) -> impl Future<Output = Result<()>> + use<> {
2567 self.text.wait_for_version(version)
2568 }
2569
2570 /// Forces all futures returned by [`Buffer::wait_for_version`], [`Buffer::wait_for_edits`], or
2571 /// [`Buffer::wait_for_version`] to resolve with an error.
2572 pub fn give_up_waiting(&mut self) {
2573 self.text.give_up_waiting();
2574 }
2575
2576 pub fn wait_for_autoindent_applied(&mut self) -> Option<oneshot::Receiver<()>> {
2577 let mut rx = None;
2578 if !self.autoindent_requests.is_empty() {
2579 let channel = oneshot::channel();
2580 self.wait_for_autoindent_txs.push(channel.0);
2581 rx = Some(channel.1);
2582 }
2583 rx
2584 }
2585
2586 /// Stores a set of selections that should be broadcasted to all of the buffer's replicas.
2587 pub fn set_active_selections(
2588 &mut self,
2589 selections: Arc<[Selection<Anchor>]>,
2590 line_mode: bool,
2591 cursor_shape: CursorShape,
2592 cx: &mut Context<Self>,
2593 ) {
2594 let lamport_timestamp = self.text.lamport_clock.tick();
2595 self.remote_selections.insert(
2596 self.text.replica_id(),
2597 SelectionSet {
2598 selections: selections.clone(),
2599 lamport_timestamp,
2600 line_mode,
2601 cursor_shape,
2602 },
2603 );
2604 self.send_operation(
2605 Operation::UpdateSelections {
2606 selections,
2607 line_mode,
2608 lamport_timestamp,
2609 cursor_shape,
2610 },
2611 true,
2612 cx,
2613 );
2614 self.non_text_state_update_count += 1;
2615 cx.notify();
2616 }
2617
2618 /// Clears the selections, so that other replicas of the buffer do not see any selections for
2619 /// this replica.
2620 pub fn remove_active_selections(&mut self, cx: &mut Context<Self>) {
2621 if self
2622 .remote_selections
2623 .get(&self.text.replica_id())
2624 .is_none_or(|set| !set.selections.is_empty())
2625 {
2626 self.set_active_selections(Arc::default(), false, Default::default(), cx);
2627 }
2628 }
2629
2630 pub fn set_agent_selections(
2631 &mut self,
2632 selections: Arc<[Selection<Anchor>]>,
2633 line_mode: bool,
2634 cursor_shape: CursorShape,
2635 cx: &mut Context<Self>,
2636 ) {
2637 let lamport_timestamp = self.text.lamport_clock.tick();
2638 self.remote_selections.insert(
2639 ReplicaId::AGENT,
2640 SelectionSet {
2641 selections,
2642 lamport_timestamp,
2643 line_mode,
2644 cursor_shape,
2645 },
2646 );
2647 self.non_text_state_update_count += 1;
2648 cx.notify();
2649 }
2650
2651 pub fn remove_agent_selections(&mut self, cx: &mut Context<Self>) {
2652 self.set_agent_selections(Arc::default(), false, Default::default(), cx);
2653 }
2654
2655 /// Replaces the buffer's entire text.
2656 pub fn set_text<T>(&mut self, text: T, cx: &mut Context<Self>) -> Option<clock::Lamport>
2657 where
2658 T: Into<Arc<str>>,
2659 {
2660 self.autoindent_requests.clear();
2661 self.edit([(0..self.len(), text)], None, cx)
2662 }
2663
2664 /// Appends the given text to the end of the buffer.
2665 pub fn append<T>(&mut self, text: T, cx: &mut Context<Self>) -> Option<clock::Lamport>
2666 where
2667 T: Into<Arc<str>>,
2668 {
2669 self.edit([(self.len()..self.len(), text)], None, cx)
2670 }
2671
2672 /// Applies the given edits to the buffer. Each edit is specified as a range of text to
2673 /// delete, and a string of text to insert at that location. Adjacent edits are coalesced.
2674 ///
2675 /// If an [`AutoindentMode`] is provided, then the buffer will enqueue an auto-indent
2676 /// request for the edited ranges, which will be processed when the buffer finishes
2677 /// parsing.
2678 ///
2679 /// Parsing takes place at the end of a transaction, and may compute synchronously
2680 /// or asynchronously, depending on the changes.
2681 pub fn edit<I, S, T>(
2682 &mut self,
2683 edits_iter: I,
2684 autoindent_mode: Option<AutoindentMode>,
2685 cx: &mut Context<Self>,
2686 ) -> Option<clock::Lamport>
2687 where
2688 I: IntoIterator<Item = (Range<S>, T)>,
2689 S: ToOffset,
2690 T: Into<Arc<str>>,
2691 {
2692 self.edit_internal(edits_iter, autoindent_mode, true, cx)
2693 }
2694
2695 /// Like [`edit`](Self::edit), but does not coalesce adjacent edits.
2696 pub fn edit_non_coalesce<I, S, T>(
2697 &mut self,
2698 edits_iter: I,
2699 autoindent_mode: Option<AutoindentMode>,
2700 cx: &mut Context<Self>,
2701 ) -> Option<clock::Lamport>
2702 where
2703 I: IntoIterator<Item = (Range<S>, T)>,
2704 S: ToOffset,
2705 T: Into<Arc<str>>,
2706 {
2707 self.edit_internal(edits_iter, autoindent_mode, false, cx)
2708 }
2709
2710 fn edit_internal<I, S, T>(
2711 &mut self,
2712 edits_iter: I,
2713 autoindent_mode: Option<AutoindentMode>,
2714 coalesce_adjacent: bool,
2715 cx: &mut Context<Self>,
2716 ) -> Option<clock::Lamport>
2717 where
2718 I: IntoIterator<Item = (Range<S>, T)>,
2719 S: ToOffset,
2720 T: Into<Arc<str>>,
2721 {
2722 // Skip invalid edits and coalesce contiguous ones.
2723 let mut edits: Vec<(Range<usize>, Arc<str>)> = Vec::new();
2724
2725 for (range, new_text) in edits_iter {
2726 let mut range = range.start.to_offset(self)..range.end.to_offset(self);
2727
2728 if range.start > range.end {
2729 mem::swap(&mut range.start, &mut range.end);
2730 }
2731 let new_text = new_text.into();
2732 if !new_text.is_empty() || !range.is_empty() {
2733 let prev_edit = edits.last_mut();
2734 let should_coalesce = prev_edit.as_ref().is_some_and(|(prev_range, _)| {
2735 if coalesce_adjacent {
2736 prev_range.end >= range.start
2737 } else {
2738 prev_range.end > range.start
2739 }
2740 });
2741
2742 if let Some((prev_range, prev_text)) = prev_edit
2743 && should_coalesce
2744 {
2745 prev_range.end = cmp::max(prev_range.end, range.end);
2746 *prev_text = format!("{prev_text}{new_text}").into();
2747 } else {
2748 edits.push((range, new_text));
2749 }
2750 }
2751 }
2752 if edits.is_empty() {
2753 return None;
2754 }
2755
2756 self.start_transaction();
2757 self.pending_autoindent.take();
2758 let autoindent_request = autoindent_mode
2759 .and_then(|mode| self.language.as_ref().map(|_| (self.snapshot(), mode)));
2760
2761 let edit_operation = self.text.edit(edits.iter().cloned());
2762 let edit_id = edit_operation.timestamp();
2763
2764 if let Some((before_edit, mode)) = autoindent_request {
2765 let mut delta = 0isize;
2766 let mut previous_setting = None;
2767 let entries: Vec<_> = edits
2768 .into_iter()
2769 .enumerate()
2770 .zip(&edit_operation.as_edit().unwrap().new_text)
2771 .filter(|((_, (range, _)), _)| {
2772 let language = before_edit.language_at(range.start);
2773 let language_id = language.map(|l| l.id());
2774 if let Some((cached_language_id, apply_syntax_indent)) = previous_setting
2775 && cached_language_id == language_id
2776 {
2777 apply_syntax_indent
2778 } else {
2779 // The auto-indent setting is not present in editorconfigs, hence
2780 // we can avoid passing the file here.
2781 let auto_indent_mode = LanguageSettings::resolve(
2782 None,
2783 language.map(|l| l.name()).as_ref(),
2784 cx,
2785 )
2786 .auto_indent;
2787 let apply_syntax_indent = auto_indent_mode == AutoIndentMode::SyntaxAware;
2788 previous_setting = Some((language_id, apply_syntax_indent));
2789 apply_syntax_indent
2790 }
2791 })
2792 .map(|((ix, (range, _)), new_text)| {
2793 let new_text_length = new_text.len();
2794 let old_start = range.start.to_point(&before_edit);
2795 let new_start = (delta + range.start as isize) as usize;
2796 let range_len = range.end - range.start;
2797 delta += new_text_length as isize - range_len as isize;
2798
2799 // Decide what range of the insertion to auto-indent, and whether
2800 // the first line of the insertion should be considered a newly-inserted line
2801 // or an edit to an existing line.
2802 let mut range_of_insertion_to_indent = 0..new_text_length;
2803 let mut first_line_is_new = true;
2804
2805 let old_line_start = before_edit.indent_size_for_line(old_start.row).len;
2806 let old_line_end = before_edit.line_len(old_start.row);
2807
2808 if old_start.column > old_line_start {
2809 first_line_is_new = false;
2810 }
2811
2812 if !new_text.contains('\n')
2813 && (old_start.column + (range_len as u32) < old_line_end
2814 || old_line_end == old_line_start)
2815 {
2816 first_line_is_new = false;
2817 }
2818
2819 // When inserting text starting with a newline, avoid auto-indenting the
2820 // previous line.
2821 if new_text.starts_with('\n') {
2822 range_of_insertion_to_indent.start += 1;
2823 first_line_is_new = true;
2824 }
2825
2826 let mut original_indent_column = None;
2827 if let AutoindentMode::Block {
2828 original_indent_columns,
2829 } = &mode
2830 {
2831 original_indent_column = Some(if new_text.starts_with('\n') {
2832 indent_size_for_text(
2833 new_text[range_of_insertion_to_indent.clone()].chars(),
2834 )
2835 .len
2836 } else {
2837 original_indent_columns
2838 .get(ix)
2839 .copied()
2840 .flatten()
2841 .unwrap_or_else(|| {
2842 indent_size_for_text(
2843 new_text[range_of_insertion_to_indent.clone()].chars(),
2844 )
2845 .len
2846 })
2847 });
2848
2849 // Avoid auto-indenting the line after the edit.
2850 if new_text[range_of_insertion_to_indent.clone()].ends_with('\n') {
2851 range_of_insertion_to_indent.end -= 1;
2852 }
2853 }
2854
2855 AutoindentRequestEntry {
2856 original_indent_column,
2857 old_row: if first_line_is_new {
2858 None
2859 } else {
2860 Some(old_start.row)
2861 },
2862 indent_size: before_edit.language_indent_size_at(range.start, cx),
2863 range: self.anchor_before(new_start + range_of_insertion_to_indent.start)
2864 ..self.anchor_after(new_start + range_of_insertion_to_indent.end),
2865 }
2866 })
2867 .collect();
2868
2869 if !entries.is_empty() {
2870 self.autoindent_requests.push(Arc::new(AutoindentRequest {
2871 before_edit,
2872 entries,
2873 is_block_mode: matches!(mode, AutoindentMode::Block { .. }),
2874 ignore_empty_lines: false,
2875 }));
2876 }
2877 }
2878
2879 self.end_transaction(cx);
2880 self.send_operation(Operation::Buffer(edit_operation), true, cx);
2881 Some(edit_id)
2882 }
2883
2884 fn did_edit(
2885 &mut self,
2886 old_version: &clock::Global,
2887 was_dirty: bool,
2888 is_local: bool,
2889 cx: &mut Context<Self>,
2890 ) {
2891 self.was_changed();
2892
2893 if self.edits_since::<usize>(old_version).next().is_none() {
2894 return;
2895 }
2896
2897 self.reparse(cx, true);
2898 cx.emit(BufferEvent::Edited { is_local });
2899 let is_dirty = self.is_dirty();
2900 if was_dirty != is_dirty {
2901 cx.emit(BufferEvent::DirtyChanged);
2902 }
2903 if was_dirty && !is_dirty {
2904 if let Some(file) = self.file.as_ref() {
2905 if matches!(file.disk_state(), DiskState::Present { .. })
2906 && file.disk_state().mtime() != self.saved_mtime
2907 {
2908 cx.emit(BufferEvent::ReloadNeeded);
2909 }
2910 }
2911 }
2912 cx.notify();
2913 }
2914
2915 pub fn autoindent_ranges<I, T>(&mut self, ranges: I, cx: &mut Context<Self>)
2916 where
2917 I: IntoIterator<Item = Range<T>>,
2918 T: ToOffset + Copy,
2919 {
2920 let before_edit = self.snapshot();
2921 let entries = ranges
2922 .into_iter()
2923 .map(|range| AutoindentRequestEntry {
2924 range: before_edit.anchor_before(range.start)..before_edit.anchor_after(range.end),
2925 old_row: None,
2926 indent_size: before_edit.language_indent_size_at(range.start, cx),
2927 original_indent_column: None,
2928 })
2929 .collect();
2930 self.autoindent_requests.push(Arc::new(AutoindentRequest {
2931 before_edit,
2932 entries,
2933 is_block_mode: false,
2934 ignore_empty_lines: true,
2935 }));
2936 self.request_autoindent(cx, Some(Duration::from_micros(300)));
2937 }
2938
2939 // Inserts newlines at the given position to create an empty line, returning the start of the new line.
2940 // You can also request the insertion of empty lines above and below the line starting at the returned point.
2941 pub fn insert_empty_line(
2942 &mut self,
2943 position: impl ToPoint,
2944 space_above: bool,
2945 space_below: bool,
2946 cx: &mut Context<Self>,
2947 ) -> Point {
2948 let mut position = position.to_point(self);
2949
2950 self.start_transaction();
2951
2952 self.edit(
2953 [(position..position, "\n")],
2954 Some(AutoindentMode::EachLine),
2955 cx,
2956 );
2957
2958 if position.column > 0 {
2959 position += Point::new(1, 0);
2960 }
2961
2962 if !self.is_line_blank(position.row) {
2963 self.edit(
2964 [(position..position, "\n")],
2965 Some(AutoindentMode::EachLine),
2966 cx,
2967 );
2968 }
2969
2970 if space_above && position.row > 0 && !self.is_line_blank(position.row - 1) {
2971 self.edit(
2972 [(position..position, "\n")],
2973 Some(AutoindentMode::EachLine),
2974 cx,
2975 );
2976 position.row += 1;
2977 }
2978
2979 if space_below
2980 && (position.row == self.max_point().row || !self.is_line_blank(position.row + 1))
2981 {
2982 self.edit(
2983 [(position..position, "\n")],
2984 Some(AutoindentMode::EachLine),
2985 cx,
2986 );
2987 }
2988
2989 self.end_transaction(cx);
2990
2991 position
2992 }
2993
2994 /// Applies the given remote operations to the buffer.
2995 pub fn apply_ops<I: IntoIterator<Item = Operation>>(&mut self, ops: I, cx: &mut Context<Self>) {
2996 self.pending_autoindent.take();
2997 let was_dirty = self.is_dirty();
2998 let old_version = self.version.clone();
2999 let mut deferred_ops = Vec::new();
3000 let buffer_ops = ops
3001 .into_iter()
3002 .filter_map(|op| match op {
3003 Operation::Buffer(op) => Some(op),
3004 _ => {
3005 if self.can_apply_op(&op) {
3006 self.apply_op(op, cx);
3007 } else {
3008 deferred_ops.push(op);
3009 }
3010 None
3011 }
3012 })
3013 .collect::<Vec<_>>();
3014 for operation in buffer_ops.iter() {
3015 self.send_operation(Operation::Buffer(operation.clone()), false, cx);
3016 }
3017 self.text.apply_ops(buffer_ops);
3018 self.deferred_ops.insert(deferred_ops);
3019 self.flush_deferred_ops(cx);
3020 self.did_edit(&old_version, was_dirty, false, cx);
3021 // Notify independently of whether the buffer was edited as the operations could include a
3022 // selection update.
3023 cx.notify();
3024 }
3025
3026 fn flush_deferred_ops(&mut self, cx: &mut Context<Self>) {
3027 let mut deferred_ops = Vec::new();
3028 for op in self.deferred_ops.drain().iter().cloned() {
3029 if self.can_apply_op(&op) {
3030 self.apply_op(op, cx);
3031 } else {
3032 deferred_ops.push(op);
3033 }
3034 }
3035 self.deferred_ops.insert(deferred_ops);
3036 }
3037
3038 pub fn has_deferred_ops(&self) -> bool {
3039 !self.deferred_ops.is_empty() || self.text.has_deferred_ops()
3040 }
3041
3042 fn can_apply_op(&self, operation: &Operation) -> bool {
3043 match operation {
3044 Operation::Buffer(_) => {
3045 unreachable!("buffer operations should never be applied at this layer")
3046 }
3047 Operation::UpdateDiagnostics {
3048 diagnostics: diagnostic_set,
3049 ..
3050 } => diagnostic_set.iter().all(|diagnostic| {
3051 self.text.can_resolve(&diagnostic.range.start)
3052 && self.text.can_resolve(&diagnostic.range.end)
3053 }),
3054 Operation::UpdateSelections { selections, .. } => selections
3055 .iter()
3056 .all(|s| self.can_resolve(&s.start) && self.can_resolve(&s.end)),
3057 Operation::UpdateCompletionTriggers { .. } | Operation::UpdateLineEnding { .. } => true,
3058 }
3059 }
3060
3061 fn apply_op(&mut self, operation: Operation, cx: &mut Context<Self>) {
3062 match operation {
3063 Operation::Buffer(_) => {
3064 unreachable!("buffer operations should never be applied at this layer")
3065 }
3066 Operation::UpdateDiagnostics {
3067 server_id,
3068 diagnostics: diagnostic_set,
3069 lamport_timestamp,
3070 } => {
3071 let snapshot = self.snapshot();
3072 self.apply_diagnostic_update(
3073 server_id,
3074 DiagnosticSet::from_sorted_entries(diagnostic_set.iter().cloned(), &snapshot),
3075 lamport_timestamp,
3076 cx,
3077 );
3078 }
3079 Operation::UpdateSelections {
3080 selections,
3081 lamport_timestamp,
3082 line_mode,
3083 cursor_shape,
3084 } => {
3085 if let Some(set) = self.remote_selections.get(&lamport_timestamp.replica_id)
3086 && set.lamport_timestamp > lamport_timestamp
3087 {
3088 return;
3089 }
3090
3091 self.remote_selections.insert(
3092 lamport_timestamp.replica_id,
3093 SelectionSet {
3094 selections,
3095 lamport_timestamp,
3096 line_mode,
3097 cursor_shape,
3098 },
3099 );
3100 self.text.lamport_clock.observe(lamport_timestamp);
3101 self.non_text_state_update_count += 1;
3102 }
3103 Operation::UpdateCompletionTriggers {
3104 triggers,
3105 lamport_timestamp,
3106 server_id,
3107 } => {
3108 if triggers.is_empty() {
3109 self.completion_triggers_per_language_server
3110 .remove(&server_id);
3111 self.completion_triggers = self
3112 .completion_triggers_per_language_server
3113 .values()
3114 .flat_map(|triggers| triggers.iter().cloned())
3115 .collect();
3116 } else {
3117 self.completion_triggers_per_language_server
3118 .insert(server_id, triggers.iter().cloned().collect());
3119 self.completion_triggers.extend(triggers);
3120 }
3121 self.text.lamport_clock.observe(lamport_timestamp);
3122 }
3123 Operation::UpdateLineEnding {
3124 line_ending,
3125 lamport_timestamp,
3126 } => {
3127 self.text.set_line_ending(line_ending);
3128 self.text.lamport_clock.observe(lamport_timestamp);
3129 }
3130 }
3131 }
3132
3133 fn apply_diagnostic_update(
3134 &mut self,
3135 server_id: LanguageServerId,
3136 diagnostics: DiagnosticSet,
3137 lamport_timestamp: clock::Lamport,
3138 cx: &mut Context<Self>,
3139 ) {
3140 if lamport_timestamp > self.diagnostics_timestamp {
3141 if diagnostics.is_empty() {
3142 self.diagnostics.remove(&server_id);
3143 } else {
3144 self.diagnostics.insert(server_id, diagnostics);
3145 }
3146 self.diagnostics_timestamp = lamport_timestamp;
3147 self.non_text_state_update_count += 1;
3148 self.text.lamport_clock.observe(lamport_timestamp);
3149 cx.notify();
3150 cx.emit(BufferEvent::DiagnosticsUpdated);
3151 }
3152 }
3153
3154 fn send_operation(&mut self, operation: Operation, is_local: bool, cx: &mut Context<Self>) {
3155 self.was_changed();
3156 cx.emit(BufferEvent::Operation {
3157 operation,
3158 is_local,
3159 });
3160 }
3161
3162 /// Removes the selections for a given peer.
3163 pub fn remove_peer(&mut self, replica_id: ReplicaId, cx: &mut Context<Self>) {
3164 self.remote_selections.remove(&replica_id);
3165 cx.notify();
3166 }
3167
3168 /// Undoes the most recent transaction.
3169 pub fn undo(&mut self, cx: &mut Context<Self>) -> Option<TransactionId> {
3170 let was_dirty = self.is_dirty();
3171 let old_version = self.version.clone();
3172
3173 if let Some((transaction_id, operation)) = self.text.undo() {
3174 self.send_operation(Operation::Buffer(operation), true, cx);
3175 self.did_edit(&old_version, was_dirty, true, cx);
3176 self.restore_encoding_for_transaction(transaction_id, was_dirty);
3177 Some(transaction_id)
3178 } else {
3179 None
3180 }
3181 }
3182
3183 /// Manually undoes a specific transaction in the buffer's undo history.
3184 pub fn undo_transaction(
3185 &mut self,
3186 transaction_id: TransactionId,
3187 cx: &mut Context<Self>,
3188 ) -> bool {
3189 let was_dirty = self.is_dirty();
3190 let old_version = self.version.clone();
3191 if let Some(operation) = self.text.undo_transaction(transaction_id) {
3192 self.send_operation(Operation::Buffer(operation), true, cx);
3193 self.did_edit(&old_version, was_dirty, true, cx);
3194 true
3195 } else {
3196 false
3197 }
3198 }
3199
3200 /// Manually undoes all changes after a given transaction in the buffer's undo history.
3201 pub fn undo_to_transaction(
3202 &mut self,
3203 transaction_id: TransactionId,
3204 cx: &mut Context<Self>,
3205 ) -> bool {
3206 let was_dirty = self.is_dirty();
3207 let old_version = self.version.clone();
3208
3209 let operations = self.text.undo_to_transaction(transaction_id);
3210 let undone = !operations.is_empty();
3211 for operation in operations {
3212 self.send_operation(Operation::Buffer(operation), true, cx);
3213 }
3214 if undone {
3215 self.did_edit(&old_version, was_dirty, true, cx)
3216 }
3217 undone
3218 }
3219
3220 pub fn undo_operations(&mut self, counts: HashMap<Lamport, u32>, cx: &mut Context<Buffer>) {
3221 let was_dirty = self.is_dirty();
3222 let operation = self.text.undo_operations(counts);
3223 let old_version = self.version.clone();
3224 self.send_operation(Operation::Buffer(operation), true, cx);
3225 self.did_edit(&old_version, was_dirty, true, cx);
3226 }
3227
3228 /// Manually redoes a specific transaction in the buffer's redo history.
3229 pub fn redo(&mut self, cx: &mut Context<Self>) -> Option<TransactionId> {
3230 let was_dirty = self.is_dirty();
3231 let old_version = self.version.clone();
3232
3233 if let Some((transaction_id, operation)) = self.text.redo() {
3234 self.send_operation(Operation::Buffer(operation), true, cx);
3235 self.did_edit(&old_version, was_dirty, true, cx);
3236 self.restore_encoding_for_transaction(transaction_id, was_dirty);
3237 Some(transaction_id)
3238 } else {
3239 None
3240 }
3241 }
3242
3243 fn restore_encoding_for_transaction(&mut self, transaction_id: TransactionId, was_dirty: bool) {
3244 if let Some((old_encoding, old_has_bom)) =
3245 self.reload_with_encoding_txns.get(&transaction_id)
3246 {
3247 let current_encoding = self.encoding;
3248 let current_has_bom = self.has_bom;
3249 self.encoding = *old_encoding;
3250 self.has_bom = *old_has_bom;
3251 if !was_dirty {
3252 self.saved_version = self.version.clone();
3253 self.has_unsaved_edits
3254 .set((self.saved_version.clone(), false));
3255 }
3256 self.reload_with_encoding_txns
3257 .insert(transaction_id, (current_encoding, current_has_bom));
3258 }
3259 }
3260
3261 /// Manually undoes all changes until a given transaction in the buffer's redo history.
3262 pub fn redo_to_transaction(
3263 &mut self,
3264 transaction_id: TransactionId,
3265 cx: &mut Context<Self>,
3266 ) -> bool {
3267 let was_dirty = self.is_dirty();
3268 let old_version = self.version.clone();
3269
3270 let operations = self.text.redo_to_transaction(transaction_id);
3271 let redone = !operations.is_empty();
3272 for operation in operations {
3273 self.send_operation(Operation::Buffer(operation), true, cx);
3274 }
3275 if redone {
3276 self.did_edit(&old_version, was_dirty, true, cx)
3277 }
3278 redone
3279 }
3280
3281 /// Override current completion triggers with the user-provided completion triggers.
3282 pub fn set_completion_triggers(
3283 &mut self,
3284 server_id: LanguageServerId,
3285 triggers: BTreeSet<String>,
3286 cx: &mut Context<Self>,
3287 ) {
3288 self.completion_triggers_timestamp = self.text.lamport_clock.tick();
3289 if triggers.is_empty() {
3290 self.completion_triggers_per_language_server
3291 .remove(&server_id);
3292 self.completion_triggers = self
3293 .completion_triggers_per_language_server
3294 .values()
3295 .flat_map(|triggers| triggers.iter().cloned())
3296 .collect();
3297 } else {
3298 self.completion_triggers_per_language_server
3299 .insert(server_id, triggers.clone());
3300 self.completion_triggers.extend(triggers.iter().cloned());
3301 }
3302 self.send_operation(
3303 Operation::UpdateCompletionTriggers {
3304 triggers: triggers.into_iter().collect(),
3305 lamport_timestamp: self.completion_triggers_timestamp,
3306 server_id,
3307 },
3308 true,
3309 cx,
3310 );
3311 cx.notify();
3312 }
3313
3314 /// Returns a list of strings which trigger a completion menu for this language.
3315 /// Usually this is driven by LSP server which returns a list of trigger characters for completions.
3316 pub fn completion_triggers(&self) -> &BTreeSet<String> {
3317 &self.completion_triggers
3318 }
3319
3320 /// Call this directly after performing edits to prevent the preview tab
3321 /// from being dismissed by those edits. It causes `should_dismiss_preview`
3322 /// to return false until there are additional edits.
3323 pub fn refresh_preview(&mut self) {
3324 self.preview_version = self.version.clone();
3325 }
3326
3327 /// Whether we should preserve the preview status of a tab containing this buffer.
3328 pub fn preserve_preview(&self) -> bool {
3329 !self.has_edits_since(&self.preview_version)
3330 }
3331}
3332
3333#[doc(hidden)]
3334#[cfg(any(test, feature = "test-support"))]
3335impl Buffer {
3336 pub fn edit_via_marked_text(
3337 &mut self,
3338 marked_string: &str,
3339 autoindent_mode: Option<AutoindentMode>,
3340 cx: &mut Context<Self>,
3341 ) {
3342 let edits = self.edits_for_marked_text(marked_string);
3343 self.edit(edits, autoindent_mode, cx);
3344 }
3345
3346 pub fn set_group_interval(&mut self, group_interval: Duration) {
3347 self.text.set_group_interval(group_interval);
3348 }
3349
3350 pub fn randomly_edit<T>(&mut self, rng: &mut T, old_range_count: usize, cx: &mut Context<Self>)
3351 where
3352 T: rand::Rng,
3353 {
3354 let mut edits: Vec<(Range<usize>, String)> = Vec::new();
3355 let mut last_end = None;
3356 for _ in 0..old_range_count {
3357 if last_end.is_some_and(|last_end| last_end >= self.len()) {
3358 break;
3359 }
3360
3361 let new_start = last_end.map_or(0, |last_end| last_end + 1);
3362 let mut range = self.random_byte_range(new_start, rng);
3363 if rng.random_bool(0.2) {
3364 mem::swap(&mut range.start, &mut range.end);
3365 }
3366 last_end = Some(range.end);
3367
3368 let new_text_len = rng.random_range(0..10);
3369 let mut new_text: String = RandomCharIter::new(&mut *rng).take(new_text_len).collect();
3370 new_text = new_text.to_uppercase();
3371
3372 edits.push((range, new_text));
3373 }
3374 log::info!("mutating buffer {:?} with {:?}", self.replica_id(), edits);
3375 self.edit(edits, None, cx);
3376 }
3377
3378 pub fn randomly_undo_redo(&mut self, rng: &mut impl rand::Rng, cx: &mut Context<Self>) {
3379 let was_dirty = self.is_dirty();
3380 let old_version = self.version.clone();
3381
3382 let ops = self.text.randomly_undo_redo(rng);
3383 if !ops.is_empty() {
3384 for op in ops {
3385 self.send_operation(Operation::Buffer(op), true, cx);
3386 self.did_edit(&old_version, was_dirty, true, cx);
3387 }
3388 }
3389 }
3390}
3391
3392impl EventEmitter<BufferEvent> for Buffer {}
3393
3394fn offset_in_sub_ranges(
3395 sub_ranges: &[Range<Anchor>],
3396 offset: usize,
3397 snapshot: &TextBufferSnapshot,
3398) -> bool {
3399 let start_anchor = snapshot.anchor_before(offset);
3400 let end_anchor = snapshot.anchor_after(offset);
3401
3402 sub_ranges.iter().any(|sub_range| {
3403 let is_before_start = sub_range.end.cmp(&start_anchor, snapshot).is_lt();
3404 let is_after_end = sub_range.start.cmp(&end_anchor, snapshot).is_gt();
3405 !is_before_start && !is_after_end
3406 })
3407}
3408
3409impl Deref for Buffer {
3410 type Target = TextBuffer;
3411
3412 fn deref(&self) -> &Self::Target {
3413 &self.text
3414 }
3415}
3416
3417impl BufferSnapshot {
3418 /// Returns [`IndentSize`] for a given line that respects user settings and
3419 /// language preferences.
3420 pub fn indent_size_for_line(&self, row: u32) -> IndentSize {
3421 indent_size_for_line(self, row)
3422 }
3423
3424 /// Returns [`IndentSize`] for a given position that respects user settings
3425 /// and language preferences.
3426 pub fn language_indent_size_at<T: ToOffset>(&self, position: T, cx: &App) -> IndentSize {
3427 let settings = self.settings_at(position, cx);
3428 if settings.hard_tabs {
3429 IndentSize::tab()
3430 } else {
3431 IndentSize::spaces(settings.tab_size.get())
3432 }
3433 }
3434
3435 /// Retrieve the suggested indent size for all of the given rows. The unit of indentation
3436 /// is passed in as `single_indent_size`.
3437 pub fn suggested_indents(
3438 &self,
3439 rows: impl Iterator<Item = u32>,
3440 single_indent_size: IndentSize,
3441 ) -> BTreeMap<u32, IndentSize> {
3442 let mut result = BTreeMap::new();
3443
3444 for row_range in contiguous_ranges(rows, 10) {
3445 let suggestions = match self.suggest_autoindents(row_range.clone()) {
3446 Some(suggestions) => suggestions,
3447 _ => break,
3448 };
3449
3450 for (row, suggestion) in row_range.zip(suggestions) {
3451 let indent_size = if let Some(suggestion) = suggestion {
3452 result
3453 .get(&suggestion.basis_row)
3454 .copied()
3455 .unwrap_or_else(|| self.indent_size_for_line(suggestion.basis_row))
3456 .with_delta(suggestion.delta, single_indent_size)
3457 } else {
3458 self.indent_size_for_line(row)
3459 };
3460
3461 result.insert(row, indent_size);
3462 }
3463 }
3464
3465 result
3466 }
3467
3468 fn suggest_autoindents(
3469 &self,
3470 row_range: Range<u32>,
3471 ) -> Option<impl Iterator<Item = Option<IndentSuggestion>> + '_> {
3472 let config = &self.language.as_ref()?.config;
3473 let prev_non_blank_row = self.prev_non_blank_row(row_range.start);
3474
3475 #[derive(Debug, Clone)]
3476 struct StartPosition {
3477 start: Point,
3478 suffix: SharedString,
3479 language: Arc<Language>,
3480 }
3481
3482 // Find the suggested indentation ranges based on the syntax tree.
3483 let start = Point::new(prev_non_blank_row.unwrap_or(row_range.start), 0);
3484 let end = Point::new(row_range.end, 0);
3485 let range = (start..end).to_offset(&self.text);
3486 let mut matches = self.syntax.matches_with_options(
3487 range.clone(),
3488 &self.text,
3489 TreeSitterOptions {
3490 max_bytes_to_query: Some(MAX_BYTES_TO_QUERY),
3491 max_start_depth: None,
3492 },
3493 |grammar| Some(&grammar.indents_config.as_ref()?.query),
3494 );
3495 let indent_configs = matches
3496 .grammars()
3497 .iter()
3498 .map(|grammar| grammar.indents_config.as_ref().unwrap())
3499 .collect::<Vec<_>>();
3500
3501 let mut indent_ranges = Vec::<Range<Point>>::new();
3502 let mut start_positions = Vec::<StartPosition>::new();
3503 let mut outdent_positions = Vec::<Point>::new();
3504 while let Some(mat) = matches.peek() {
3505 let mut start: Option<Point> = None;
3506 let mut end: Option<Point> = None;
3507
3508 let config = indent_configs[mat.grammar_index];
3509 for capture in mat.captures {
3510 if capture.index == config.indent_capture_ix {
3511 start.get_or_insert(Point::from_ts_point(capture.node.start_position()));
3512 end.get_or_insert(Point::from_ts_point(capture.node.end_position()));
3513 } else if Some(capture.index) == config.start_capture_ix {
3514 start = Some(Point::from_ts_point(capture.node.end_position()));
3515 } else if Some(capture.index) == config.end_capture_ix {
3516 end = Some(Point::from_ts_point(capture.node.start_position()));
3517 } else if Some(capture.index) == config.outdent_capture_ix {
3518 outdent_positions.push(Point::from_ts_point(capture.node.start_position()));
3519 } else if let Some(suffix) = config.suffixed_start_captures.get(&capture.index) {
3520 start_positions.push(StartPosition {
3521 start: Point::from_ts_point(capture.node.start_position()),
3522 suffix: suffix.clone(),
3523 language: mat.language.clone(),
3524 });
3525 }
3526 }
3527
3528 matches.advance();
3529 if let Some((start, end)) = start.zip(end) {
3530 if start.row == end.row {
3531 continue;
3532 }
3533 let range = start..end;
3534 match indent_ranges.binary_search_by_key(&range.start, |r| r.start) {
3535 Err(ix) => indent_ranges.insert(ix, range),
3536 Ok(ix) => {
3537 let prev_range = &mut indent_ranges[ix];
3538 prev_range.end = prev_range.end.max(range.end);
3539 }
3540 }
3541 }
3542 }
3543
3544 let mut error_ranges = Vec::<Range<Point>>::new();
3545 let mut matches = self
3546 .syntax
3547 .matches(range, &self.text, |grammar| grammar.error_query.as_ref());
3548 while let Some(mat) = matches.peek() {
3549 let node = mat.captures[0].node;
3550 let start = Point::from_ts_point(node.start_position());
3551 let end = Point::from_ts_point(node.end_position());
3552 let range = start..end;
3553 let ix = match error_ranges.binary_search_by_key(&range.start, |r| r.start) {
3554 Ok(ix) | Err(ix) => ix,
3555 };
3556 let mut end_ix = ix;
3557 while let Some(existing_range) = error_ranges.get(end_ix) {
3558 if existing_range.end < end {
3559 end_ix += 1;
3560 } else {
3561 break;
3562 }
3563 }
3564 error_ranges.splice(ix..end_ix, [range]);
3565 matches.advance();
3566 }
3567
3568 outdent_positions.sort();
3569 for outdent_position in outdent_positions {
3570 // find the innermost indent range containing this outdent_position
3571 // set its end to the outdent position
3572 if let Some(range_to_truncate) = indent_ranges
3573 .iter_mut()
3574 .rfind(|indent_range| indent_range.contains(&outdent_position))
3575 {
3576 range_to_truncate.end = outdent_position;
3577 }
3578 }
3579
3580 start_positions.sort_by_key(|b| b.start);
3581
3582 // Find the suggested indentation increases and decreased based on regexes.
3583 let mut regex_outdent_map = HashMap::default();
3584 let mut last_seen_suffix: HashMap<String, Vec<StartPosition>> = HashMap::default();
3585 let mut start_positions_iter = start_positions.iter().peekable();
3586
3587 let mut indent_change_rows = Vec::<(u32, Ordering)>::new();
3588 self.for_each_line(
3589 Point::new(prev_non_blank_row.unwrap_or(row_range.start), 0)
3590 ..Point::new(row_range.end, 0),
3591 |row, line| {
3592 let indent_len = self.indent_size_for_line(row).len;
3593 let row_language = self.language_at(Point::new(row, indent_len)).cloned();
3594 let row_language_config = row_language
3595 .as_ref()
3596 .map(|lang| lang.config())
3597 .unwrap_or(config);
3598
3599 if row_language_config
3600 .decrease_indent_pattern
3601 .as_ref()
3602 .is_some_and(|regex| regex.is_match(line))
3603 {
3604 indent_change_rows.push((row, Ordering::Less));
3605 }
3606 if row_language_config
3607 .increase_indent_pattern
3608 .as_ref()
3609 .is_some_and(|regex| regex.is_match(line))
3610 {
3611 indent_change_rows.push((row + 1, Ordering::Greater));
3612 }
3613 while let Some(pos) = start_positions_iter.peek() {
3614 if pos.start.row < row {
3615 let pos = start_positions_iter.next().unwrap().clone();
3616 last_seen_suffix
3617 .entry(pos.suffix.to_string())
3618 .or_default()
3619 .push(pos);
3620 } else {
3621 break;
3622 }
3623 }
3624 for rule in &row_language_config.decrease_indent_patterns {
3625 if rule.pattern.as_ref().is_some_and(|r| r.is_match(line)) {
3626 let row_start_column = self.indent_size_for_line(row).len;
3627 let basis_row = rule
3628 .valid_after
3629 .iter()
3630 .filter_map(|valid_suffix| last_seen_suffix.get(valid_suffix))
3631 .flatten()
3632 .filter(|pos| {
3633 row_language
3634 .as_ref()
3635 .or(self.language.as_ref())
3636 .is_some_and(|lang| Arc::ptr_eq(lang, &pos.language))
3637 })
3638 .filter(|pos| pos.start.column <= row_start_column)
3639 .max_by_key(|pos| pos.start.row);
3640 if let Some(outdent_to) = basis_row {
3641 regex_outdent_map.insert(row, outdent_to.start.row);
3642 }
3643 break;
3644 }
3645 }
3646 },
3647 );
3648
3649 let mut indent_changes = indent_change_rows.into_iter().peekable();
3650 let mut prev_row = if config.auto_indent_using_last_non_empty_line {
3651 prev_non_blank_row.unwrap_or(0)
3652 } else {
3653 row_range.start.saturating_sub(1)
3654 };
3655
3656 let mut prev_row_start = Point::new(prev_row, self.indent_size_for_line(prev_row).len);
3657 Some(row_range.map(move |row| {
3658 let row_start = Point::new(row, self.indent_size_for_line(row).len);
3659
3660 let mut indent_from_prev_row = false;
3661 let mut outdent_from_prev_row = false;
3662 let mut outdent_to_row = u32::MAX;
3663 let mut from_regex = false;
3664
3665 while let Some((indent_row, delta)) = indent_changes.peek() {
3666 match indent_row.cmp(&row) {
3667 Ordering::Equal => match delta {
3668 Ordering::Less => {
3669 from_regex = true;
3670 outdent_from_prev_row = true
3671 }
3672 Ordering::Greater => {
3673 indent_from_prev_row = true;
3674 from_regex = true
3675 }
3676 _ => {}
3677 },
3678
3679 Ordering::Greater => break,
3680 Ordering::Less => {}
3681 }
3682
3683 indent_changes.next();
3684 }
3685
3686 for range in &indent_ranges {
3687 if range.start.row >= row {
3688 break;
3689 }
3690 if range.start.row == prev_row && range.end > row_start {
3691 indent_from_prev_row = true;
3692 }
3693 if range.end > prev_row_start && range.end <= row_start {
3694 outdent_to_row = outdent_to_row.min(range.start.row);
3695 }
3696 }
3697
3698 if let Some(basis_row) = regex_outdent_map.get(&row) {
3699 indent_from_prev_row = false;
3700 outdent_to_row = *basis_row;
3701 from_regex = true;
3702 }
3703
3704 let within_error = error_ranges
3705 .iter()
3706 .any(|e| e.start.row < row && e.end > row_start);
3707
3708 let suggestion = if outdent_to_row == prev_row
3709 || (outdent_from_prev_row && indent_from_prev_row)
3710 {
3711 Some(IndentSuggestion {
3712 basis_row: prev_row,
3713 delta: Ordering::Equal,
3714 within_error: within_error && !from_regex,
3715 })
3716 } else if indent_from_prev_row {
3717 Some(IndentSuggestion {
3718 basis_row: prev_row,
3719 delta: Ordering::Greater,
3720 within_error: within_error && !from_regex,
3721 })
3722 } else if outdent_to_row < prev_row {
3723 Some(IndentSuggestion {
3724 basis_row: outdent_to_row,
3725 delta: Ordering::Equal,
3726 within_error: within_error && !from_regex,
3727 })
3728 } else if outdent_from_prev_row {
3729 Some(IndentSuggestion {
3730 basis_row: prev_row,
3731 delta: Ordering::Less,
3732 within_error: within_error && !from_regex,
3733 })
3734 } else if config.auto_indent_using_last_non_empty_line || !self.is_line_blank(prev_row)
3735 {
3736 Some(IndentSuggestion {
3737 basis_row: prev_row,
3738 delta: Ordering::Equal,
3739 within_error: within_error && !from_regex,
3740 })
3741 } else {
3742 None
3743 };
3744
3745 prev_row = row;
3746 prev_row_start = row_start;
3747 suggestion
3748 }))
3749 }
3750
3751 fn prev_non_blank_row(&self, mut row: u32) -> Option<u32> {
3752 while row > 0 {
3753 row -= 1;
3754 if !self.is_line_blank(row) {
3755 return Some(row);
3756 }
3757 }
3758 None
3759 }
3760
3761 pub fn captures(
3762 &self,
3763 range: Range<usize>,
3764 query: fn(&Grammar) -> Option<&tree_sitter::Query>,
3765 ) -> SyntaxMapCaptures<'_> {
3766 self.syntax.captures(range, &self.text, query)
3767 }
3768
3769 #[ztracing::instrument(skip_all)]
3770 fn get_highlights(&self, range: Range<usize>) -> (SyntaxMapCaptures<'_>, Vec<HighlightMap>) {
3771 let captures = self.syntax.captures(range, &self.text, |grammar| {
3772 grammar
3773 .highlights_config
3774 .as_ref()
3775 .map(|config| &config.query)
3776 });
3777 let highlight_maps = captures
3778 .grammars()
3779 .iter()
3780 .map(|grammar| grammar.highlight_map())
3781 .collect();
3782 (captures, highlight_maps)
3783 }
3784
3785 /// Iterates over chunks of text in the given range of the buffer. Text is chunked
3786 /// in an arbitrary way due to being stored in a [`Rope`](text::Rope). The text is also
3787 /// returned in chunks where each chunk has a single syntax highlighting style and
3788 /// diagnostic status.
3789 #[ztracing::instrument(skip_all)]
3790 pub fn chunks<T: ToOffset>(&self, range: Range<T>, language_aware: bool) -> BufferChunks<'_> {
3791 let range = range.start.to_offset(self)..range.end.to_offset(self);
3792
3793 let mut syntax = None;
3794 if language_aware {
3795 syntax = Some(self.get_highlights(range.clone()));
3796 }
3797 // We want to look at diagnostic spans only when iterating over language-annotated chunks.
3798 let diagnostics = language_aware;
3799 BufferChunks::new(self.text.as_rope(), range, syntax, diagnostics, Some(self))
3800 }
3801
3802 pub fn highlighted_text_for_range<T: ToOffset>(
3803 &self,
3804 range: Range<T>,
3805 override_style: Option<HighlightStyle>,
3806 syntax_theme: &SyntaxTheme,
3807 ) -> HighlightedText {
3808 HighlightedText::from_buffer_range(
3809 range,
3810 &self.text,
3811 &self.syntax,
3812 override_style,
3813 syntax_theme,
3814 )
3815 }
3816
3817 /// Invokes the given callback for each line of text in the given range of the buffer.
3818 /// Uses callback to avoid allocating a string for each line.
3819 fn for_each_line(&self, range: Range<Point>, mut callback: impl FnMut(u32, &str)) {
3820 let mut line = String::new();
3821 let mut row = range.start.row;
3822 for chunk in self
3823 .as_rope()
3824 .chunks_in_range(range.to_offset(self))
3825 .chain(["\n"])
3826 {
3827 for (newline_ix, text) in chunk.split('\n').enumerate() {
3828 if newline_ix > 0 {
3829 callback(row, &line);
3830 row += 1;
3831 line.clear();
3832 }
3833 line.push_str(text);
3834 }
3835 }
3836 }
3837
3838 /// Iterates over every [`SyntaxLayer`] in the buffer.
3839 pub fn syntax_layers(&self) -> impl Iterator<Item = SyntaxLayer<'_>> + '_ {
3840 self.syntax_layers_for_range(0..self.len(), true)
3841 }
3842
3843 pub fn syntax_layer_at<D: ToOffset>(&self, position: D) -> Option<SyntaxLayer<'_>> {
3844 let offset = position.to_offset(self);
3845 self.syntax_layers_for_range(offset..offset, false)
3846 .filter(|l| {
3847 if let Some(ranges) = l.included_sub_ranges {
3848 ranges.iter().any(|range| {
3849 let start = range.start.to_offset(self);
3850 start <= offset && {
3851 let end = range.end.to_offset(self);
3852 offset < end
3853 }
3854 })
3855 } else {
3856 l.node().start_byte() <= offset && l.node().end_byte() > offset
3857 }
3858 })
3859 .last()
3860 }
3861
3862 pub fn syntax_layers_for_range<D: ToOffset>(
3863 &self,
3864 range: Range<D>,
3865 include_hidden: bool,
3866 ) -> impl Iterator<Item = SyntaxLayer<'_>> + '_ {
3867 self.syntax
3868 .layers_for_range(range, &self.text, include_hidden)
3869 }
3870
3871 pub fn syntax_layers_languages(&self) -> impl Iterator<Item = &Arc<Language>> {
3872 self.syntax.languages(&self, true)
3873 }
3874
3875 pub fn smallest_syntax_layer_containing<D: ToOffset>(
3876 &self,
3877 range: Range<D>,
3878 ) -> Option<SyntaxLayer<'_>> {
3879 let range = range.to_offset(self);
3880 self.syntax
3881 .layers_for_range(range, &self.text, false)
3882 .max_by(|a, b| {
3883 if a.depth != b.depth {
3884 a.depth.cmp(&b.depth)
3885 } else if a.offset.0 != b.offset.0 {
3886 a.offset.0.cmp(&b.offset.0)
3887 } else {
3888 a.node().end_byte().cmp(&b.node().end_byte()).reverse()
3889 }
3890 })
3891 }
3892
3893 /// Returns the [`ModelineSettings`].
3894 pub fn modeline(&self) -> Option<&Arc<ModelineSettings>> {
3895 self.modeline.as_ref()
3896 }
3897
3898 /// Returns the main [`Language`].
3899 pub fn language(&self) -> Option<&Arc<Language>> {
3900 self.language.as_ref()
3901 }
3902
3903 /// Returns the [`Language`] at the given location.
3904 pub fn language_at<D: ToOffset>(&self, position: D) -> Option<&Arc<Language>> {
3905 self.syntax_layer_at(position)
3906 .map(|info| info.language)
3907 .or(self.language.as_ref())
3908 }
3909
3910 /// Returns the settings for the language at the given location.
3911 pub fn settings_at<'a, D: ToOffset>(
3912 &'a self,
3913 position: D,
3914 cx: &'a App,
3915 ) -> Cow<'a, LanguageSettings> {
3916 LanguageSettings::for_buffer_snapshot(self, Some(position.to_offset(self)), cx)
3917 }
3918
3919 pub fn char_classifier_at<T: ToOffset>(&self, point: T) -> CharClassifier {
3920 CharClassifier::new(self.language_scope_at(point))
3921 }
3922
3923 /// Returns the [`LanguageScope`] at the given location.
3924 pub fn language_scope_at<D: ToOffset>(&self, position: D) -> Option<LanguageScope> {
3925 let offset = position.to_offset(self);
3926 let mut scope = None;
3927 let mut smallest_range_and_depth: Option<(Range<usize>, usize)> = None;
3928 let text: &TextBufferSnapshot = self;
3929
3930 // Use the layer that has the smallest node intersecting the given point.
3931 for layer in self
3932 .syntax
3933 .layers_for_range(offset..offset, &self.text, false)
3934 {
3935 if let Some(ranges) = layer.included_sub_ranges
3936 && !offset_in_sub_ranges(ranges, offset, text)
3937 {
3938 continue;
3939 }
3940
3941 let mut cursor = layer.node().walk();
3942
3943 let mut range = None;
3944 loop {
3945 let child_range = cursor.node().byte_range();
3946 if !child_range.contains(&offset) {
3947 break;
3948 }
3949
3950 range = Some(child_range);
3951 if cursor.goto_first_child_for_byte(offset).is_none() {
3952 break;
3953 }
3954 }
3955
3956 if let Some(range) = range
3957 && smallest_range_and_depth.as_ref().is_none_or(
3958 |(smallest_range, smallest_range_depth)| {
3959 if layer.depth > *smallest_range_depth {
3960 true
3961 } else if layer.depth == *smallest_range_depth {
3962 range.len() < smallest_range.len()
3963 } else {
3964 false
3965 }
3966 },
3967 )
3968 {
3969 smallest_range_and_depth = Some((range, layer.depth));
3970 scope = Some(LanguageScope {
3971 language: layer.language.clone(),
3972 override_id: layer.override_id(offset, &self.text),
3973 });
3974 }
3975 }
3976
3977 scope.or_else(|| {
3978 self.language.clone().map(|language| LanguageScope {
3979 language,
3980 override_id: None,
3981 })
3982 })
3983 }
3984
3985 /// Returns a tuple of the range and character kind of the word
3986 /// surrounding the given position.
3987 pub fn surrounding_word<T: ToOffset>(
3988 &self,
3989 start: T,
3990 scope_context: Option<CharScopeContext>,
3991 ) -> (Range<usize>, Option<CharKind>) {
3992 let mut start = start.to_offset(self);
3993 let mut end = start;
3994 let mut next_chars = self.chars_at(start).take(128).peekable();
3995 let mut prev_chars = self.reversed_chars_at(start).take(128).peekable();
3996
3997 let classifier = self.char_classifier_at(start).scope_context(scope_context);
3998 let word_kind = cmp::max(
3999 prev_chars.peek().copied().map(|c| classifier.kind(c)),
4000 next_chars.peek().copied().map(|c| classifier.kind(c)),
4001 );
4002
4003 for ch in prev_chars {
4004 if Some(classifier.kind(ch)) == word_kind && ch != '\n' {
4005 start -= ch.len_utf8();
4006 } else {
4007 break;
4008 }
4009 }
4010
4011 for ch in next_chars {
4012 if Some(classifier.kind(ch)) == word_kind && ch != '\n' {
4013 end += ch.len_utf8();
4014 } else {
4015 break;
4016 }
4017 }
4018
4019 (start..end, word_kind)
4020 }
4021
4022 /// Moves the TreeCursor to the smallest descendant or ancestor syntax node enclosing the given
4023 /// range. When `require_larger` is true, the node found must be larger than the query range.
4024 ///
4025 /// Returns true if a node was found, and false otherwise. In the `false` case the cursor will
4026 /// be moved to the root of the tree.
4027 fn goto_node_enclosing_range(
4028 cursor: &mut tree_sitter::TreeCursor,
4029 query_range: &Range<usize>,
4030 require_larger: bool,
4031 ) -> bool {
4032 let mut ascending = false;
4033 loop {
4034 let mut range = cursor.node().byte_range();
4035 if query_range.is_empty() {
4036 // When the query range is empty and the current node starts after it, move to the
4037 // previous sibling to find the node the containing node.
4038 if range.start > query_range.start {
4039 cursor.goto_previous_sibling();
4040 range = cursor.node().byte_range();
4041 }
4042 } else {
4043 // When the query range is non-empty and the current node ends exactly at the start,
4044 // move to the next sibling to find a node that extends beyond the start.
4045 if range.end == query_range.start {
4046 cursor.goto_next_sibling();
4047 range = cursor.node().byte_range();
4048 }
4049 }
4050
4051 let encloses = range.contains_inclusive(query_range)
4052 && (!require_larger || range.len() > query_range.len());
4053 if !encloses {
4054 ascending = true;
4055 if !cursor.goto_parent() {
4056 return false;
4057 }
4058 continue;
4059 } else if ascending {
4060 return true;
4061 }
4062
4063 // Descend into the current node.
4064 if cursor
4065 .goto_first_child_for_byte(query_range.start)
4066 .is_none()
4067 {
4068 return true;
4069 }
4070 }
4071 }
4072
4073 pub fn syntax_ancestor<'a, T: ToOffset>(
4074 &'a self,
4075 range: Range<T>,
4076 ) -> Option<tree_sitter::Node<'a>> {
4077 let range = range.start.to_offset(self)..range.end.to_offset(self);
4078 let mut result: Option<tree_sitter::Node<'a>> = None;
4079 for layer in self
4080 .syntax
4081 .layers_for_range(range.clone(), &self.text, true)
4082 {
4083 let mut cursor = layer.node().walk();
4084
4085 // Find the node that both contains the range and is larger than it.
4086 if !Self::goto_node_enclosing_range(&mut cursor, &range, true) {
4087 continue;
4088 }
4089
4090 let left_node = cursor.node();
4091 let mut layer_result = left_node;
4092
4093 // For an empty range, try to find another node immediately to the right of the range.
4094 if left_node.end_byte() == range.start {
4095 let mut right_node = None;
4096 while !cursor.goto_next_sibling() {
4097 if !cursor.goto_parent() {
4098 break;
4099 }
4100 }
4101
4102 while cursor.node().start_byte() == range.start {
4103 right_node = Some(cursor.node());
4104 if !cursor.goto_first_child() {
4105 break;
4106 }
4107 }
4108
4109 // If there is a candidate node on both sides of the (empty) range, then
4110 // decide between the two by favoring a named node over an anonymous token.
4111 // If both nodes are the same in that regard, favor the right one.
4112 if let Some(right_node) = right_node
4113 && (right_node.is_named() || !left_node.is_named())
4114 {
4115 layer_result = right_node;
4116 }
4117 }
4118
4119 if let Some(previous_result) = &result
4120 && previous_result.byte_range().len() < layer_result.byte_range().len()
4121 {
4122 continue;
4123 }
4124 result = Some(layer_result);
4125 }
4126
4127 result
4128 }
4129
4130 /// Find the previous sibling syntax node at the given range.
4131 ///
4132 /// This function locates the syntax node that precedes the node containing
4133 /// the given range. It searches hierarchically by:
4134 /// 1. Finding the node that contains the given range
4135 /// 2. Looking for the previous sibling at the same tree level
4136 /// 3. If no sibling is found, moving up to parent levels and searching for siblings
4137 ///
4138 /// Returns `None` if there is no previous sibling at any ancestor level.
4139 pub fn syntax_prev_sibling<'a, T: ToOffset>(
4140 &'a self,
4141 range: Range<T>,
4142 ) -> Option<tree_sitter::Node<'a>> {
4143 let range = range.start.to_offset(self)..range.end.to_offset(self);
4144 let mut result: Option<tree_sitter::Node<'a>> = None;
4145
4146 for layer in self
4147 .syntax
4148 .layers_for_range(range.clone(), &self.text, true)
4149 {
4150 let mut cursor = layer.node().walk();
4151
4152 // Find the node that contains the range
4153 if !Self::goto_node_enclosing_range(&mut cursor, &range, false) {
4154 continue;
4155 }
4156
4157 // Look for the previous sibling, moving up ancestor levels if needed
4158 loop {
4159 if cursor.goto_previous_sibling() {
4160 let layer_result = cursor.node();
4161
4162 if let Some(previous_result) = &result {
4163 if previous_result.byte_range().end < layer_result.byte_range().end {
4164 continue;
4165 }
4166 }
4167 result = Some(layer_result);
4168 break;
4169 }
4170
4171 // No sibling found at this level, try moving up to parent
4172 if !cursor.goto_parent() {
4173 break;
4174 }
4175 }
4176 }
4177
4178 result
4179 }
4180
4181 /// Find the next sibling syntax node at the given range.
4182 ///
4183 /// This function locates the syntax node that follows the node containing
4184 /// the given range. It searches hierarchically by:
4185 /// 1. Finding the node that contains the given range
4186 /// 2. Looking for the next sibling at the same tree level
4187 /// 3. If no sibling is found, moving up to parent levels and searching for siblings
4188 ///
4189 /// Returns `None` if there is no next sibling at any ancestor level.
4190 pub fn syntax_next_sibling<'a, T: ToOffset>(
4191 &'a self,
4192 range: Range<T>,
4193 ) -> Option<tree_sitter::Node<'a>> {
4194 let range = range.start.to_offset(self)..range.end.to_offset(self);
4195 let mut result: Option<tree_sitter::Node<'a>> = None;
4196
4197 for layer in self
4198 .syntax
4199 .layers_for_range(range.clone(), &self.text, true)
4200 {
4201 let mut cursor = layer.node().walk();
4202
4203 // Find the node that contains the range
4204 if !Self::goto_node_enclosing_range(&mut cursor, &range, false) {
4205 continue;
4206 }
4207
4208 // Look for the next sibling, moving up ancestor levels if needed
4209 loop {
4210 if cursor.goto_next_sibling() {
4211 let layer_result = cursor.node();
4212
4213 if let Some(previous_result) = &result {
4214 if previous_result.byte_range().start > layer_result.byte_range().start {
4215 continue;
4216 }
4217 }
4218 result = Some(layer_result);
4219 break;
4220 }
4221
4222 // No sibling found at this level, try moving up to parent
4223 if !cursor.goto_parent() {
4224 break;
4225 }
4226 }
4227 }
4228
4229 result
4230 }
4231
4232 /// Returns the root syntax node within the given row
4233 pub fn syntax_root_ancestor(&self, position: Anchor) -> Option<tree_sitter::Node<'_>> {
4234 let start_offset = position.to_offset(self);
4235
4236 let row = self.summary_for_anchor::<text::PointUtf16>(&position).row as usize;
4237
4238 let layer = self
4239 .syntax
4240 .layers_for_range(start_offset..start_offset, &self.text, true)
4241 .next()?;
4242
4243 let mut cursor = layer.node().walk();
4244
4245 // Descend to the first leaf that touches the start of the range.
4246 while cursor.goto_first_child_for_byte(start_offset).is_some() {
4247 if cursor.node().end_byte() == start_offset {
4248 cursor.goto_next_sibling();
4249 }
4250 }
4251
4252 // Ascend to the root node within the same row.
4253 while cursor.goto_parent() {
4254 if cursor.node().start_position().row != row {
4255 break;
4256 }
4257 }
4258
4259 Some(cursor.node())
4260 }
4261
4262 /// Returns the outline for the buffer.
4263 ///
4264 /// This method allows passing an optional [`SyntaxTheme`] to
4265 /// syntax-highlight the returned symbols.
4266 pub fn outline(&self, theme: Option<&SyntaxTheme>) -> Outline<Anchor> {
4267 Outline::new(self.outline_items_containing(0..self.len(), true, theme))
4268 }
4269
4270 /// Returns all the symbols that contain the given position.
4271 ///
4272 /// This method allows passing an optional [`SyntaxTheme`] to
4273 /// syntax-highlight the returned symbols.
4274 pub fn symbols_containing<T: ToOffset>(
4275 &self,
4276 position: T,
4277 theme: Option<&SyntaxTheme>,
4278 ) -> Vec<OutlineItem<Anchor>> {
4279 let position = position.to_offset(self);
4280 let start = self.clip_offset(position.saturating_sub(1), Bias::Left);
4281 let end = self.clip_offset(position + 1, Bias::Right);
4282 let mut items = self.outline_items_containing(start..end, false, theme);
4283 let mut prev_depth = None;
4284 items.retain(|item| {
4285 let result = prev_depth.is_none_or(|prev_depth| item.depth > prev_depth);
4286 prev_depth = Some(item.depth);
4287 result
4288 });
4289 items
4290 }
4291
4292 pub fn outline_range_containing<T: ToOffset>(&self, range: Range<T>) -> Option<Range<Point>> {
4293 let range = range.to_offset(self);
4294 let mut matches = self.syntax.matches(range.clone(), &self.text, |grammar| {
4295 grammar.outline_config.as_ref().map(|c| &c.query)
4296 });
4297 let configs = matches
4298 .grammars()
4299 .iter()
4300 .map(|g| g.outline_config.as_ref().unwrap())
4301 .collect::<Vec<_>>();
4302
4303 while let Some(mat) = matches.peek() {
4304 let config = &configs[mat.grammar_index];
4305 let containing_item_node = maybe!({
4306 let item_node = mat.captures.iter().find_map(|cap| {
4307 if cap.index == config.item_capture_ix {
4308 Some(cap.node)
4309 } else {
4310 None
4311 }
4312 })?;
4313
4314 let item_byte_range = item_node.byte_range();
4315 if item_byte_range.end < range.start || item_byte_range.start > range.end {
4316 None
4317 } else {
4318 Some(item_node)
4319 }
4320 });
4321
4322 if let Some(item_node) = containing_item_node {
4323 return Some(
4324 Point::from_ts_point(item_node.start_position())
4325 ..Point::from_ts_point(item_node.end_position()),
4326 );
4327 }
4328
4329 matches.advance();
4330 }
4331 None
4332 }
4333
4334 pub fn outline_items_containing<T: ToOffset>(
4335 &self,
4336 range: Range<T>,
4337 include_extra_context: bool,
4338 theme: Option<&SyntaxTheme>,
4339 ) -> Vec<OutlineItem<Anchor>> {
4340 self.outline_items_containing_internal(
4341 range,
4342 include_extra_context,
4343 theme,
4344 |this, range| this.anchor_after(range.start)..this.anchor_before(range.end),
4345 )
4346 }
4347
4348 pub fn outline_items_as_points_containing<T: ToOffset>(
4349 &self,
4350 range: Range<T>,
4351 include_extra_context: bool,
4352 theme: Option<&SyntaxTheme>,
4353 ) -> Vec<OutlineItem<Point>> {
4354 self.outline_items_containing_internal(range, include_extra_context, theme, |_, range| {
4355 range
4356 })
4357 }
4358
4359 pub fn outline_items_as_offsets_containing<T: ToOffset>(
4360 &self,
4361 range: Range<T>,
4362 include_extra_context: bool,
4363 theme: Option<&SyntaxTheme>,
4364 ) -> Vec<OutlineItem<usize>> {
4365 self.outline_items_containing_internal(
4366 range,
4367 include_extra_context,
4368 theme,
4369 |buffer, range| range.to_offset(buffer),
4370 )
4371 }
4372
4373 fn outline_items_containing_internal<T: ToOffset, U>(
4374 &self,
4375 range: Range<T>,
4376 include_extra_context: bool,
4377 theme: Option<&SyntaxTheme>,
4378 range_callback: fn(&Self, Range<Point>) -> Range<U>,
4379 ) -> Vec<OutlineItem<U>> {
4380 let range = range.to_offset(self);
4381 let mut matches = self.syntax.matches(range.clone(), &self.text, |grammar| {
4382 grammar.outline_config.as_ref().map(|c| &c.query)
4383 });
4384
4385 let mut items = Vec::new();
4386 let mut annotation_row_ranges: Vec<Range<u32>> = Vec::new();
4387 while let Some(mat) = matches.peek() {
4388 let config = matches.grammars()[mat.grammar_index]
4389 .outline_config
4390 .as_ref()
4391 .unwrap();
4392 if let Some(item) =
4393 self.next_outline_item(config, &mat, &range, include_extra_context, theme)
4394 {
4395 items.push(item);
4396 } else if let Some(capture) = mat
4397 .captures
4398 .iter()
4399 .find(|capture| Some(capture.index) == config.annotation_capture_ix)
4400 {
4401 let capture_range = capture.node.start_position()..capture.node.end_position();
4402 let mut capture_row_range =
4403 capture_range.start.row as u32..capture_range.end.row as u32;
4404 if capture_range.end.row > capture_range.start.row && capture_range.end.column == 0
4405 {
4406 capture_row_range.end -= 1;
4407 }
4408 if let Some(last_row_range) = annotation_row_ranges.last_mut() {
4409 if last_row_range.end >= capture_row_range.start.saturating_sub(1) {
4410 last_row_range.end = capture_row_range.end;
4411 } else {
4412 annotation_row_ranges.push(capture_row_range);
4413 }
4414 } else {
4415 annotation_row_ranges.push(capture_row_range);
4416 }
4417 }
4418 matches.advance();
4419 }
4420
4421 items.sort_by_key(|item| (item.range.start, Reverse(item.range.end)));
4422
4423 // Assign depths based on containment relationships and convert to anchors.
4424 let mut item_ends_stack = Vec::<Point>::new();
4425 let mut anchor_items = Vec::new();
4426 let mut annotation_row_ranges = annotation_row_ranges.into_iter().peekable();
4427 for item in items {
4428 while let Some(last_end) = item_ends_stack.last().copied() {
4429 if last_end < item.range.end {
4430 item_ends_stack.pop();
4431 } else {
4432 break;
4433 }
4434 }
4435
4436 let mut annotation_row_range = None;
4437 while let Some(next_annotation_row_range) = annotation_row_ranges.peek() {
4438 let row_preceding_item = item.range.start.row.saturating_sub(1);
4439 if next_annotation_row_range.end < row_preceding_item {
4440 annotation_row_ranges.next();
4441 } else {
4442 if next_annotation_row_range.end == row_preceding_item {
4443 annotation_row_range = Some(next_annotation_row_range.clone());
4444 annotation_row_ranges.next();
4445 }
4446 break;
4447 }
4448 }
4449
4450 anchor_items.push(OutlineItem {
4451 depth: item_ends_stack.len(),
4452 range: range_callback(self, item.range.clone()),
4453 source_range_for_text: range_callback(self, item.source_range_for_text.clone()),
4454 text: item.text,
4455 highlight_ranges: item.highlight_ranges,
4456 name_ranges: item.name_ranges,
4457 body_range: item.body_range.map(|r| range_callback(self, r)),
4458 annotation_range: annotation_row_range.map(|annotation_range| {
4459 let point_range = Point::new(annotation_range.start, 0)
4460 ..Point::new(annotation_range.end, self.line_len(annotation_range.end));
4461 range_callback(self, point_range)
4462 }),
4463 });
4464 item_ends_stack.push(item.range.end);
4465 }
4466
4467 anchor_items
4468 }
4469
4470 fn next_outline_item(
4471 &self,
4472 config: &OutlineConfig,
4473 mat: &SyntaxMapMatch,
4474 range: &Range<usize>,
4475 include_extra_context: bool,
4476 theme: Option<&SyntaxTheme>,
4477 ) -> Option<OutlineItem<Point>> {
4478 let item_node = mat.captures.iter().find_map(|cap| {
4479 if cap.index == config.item_capture_ix {
4480 Some(cap.node)
4481 } else {
4482 None
4483 }
4484 })?;
4485
4486 let item_byte_range = item_node.byte_range();
4487 if item_byte_range.end < range.start || item_byte_range.start > range.end {
4488 return None;
4489 }
4490 let item_point_range = Point::from_ts_point(item_node.start_position())
4491 ..Point::from_ts_point(item_node.end_position());
4492
4493 let mut open_point = None;
4494 let mut close_point = None;
4495
4496 let mut buffer_ranges = Vec::new();
4497 let mut add_to_buffer_ranges = |node: tree_sitter::Node, node_is_name| {
4498 let mut range = node.start_byte()..node.end_byte();
4499 let start = node.start_position();
4500 if node.end_position().row > start.row {
4501 range.end = range.start + self.line_len(start.row as u32) as usize - start.column;
4502 }
4503
4504 if !range.is_empty() {
4505 buffer_ranges.push((range, node_is_name));
4506 }
4507 };
4508
4509 for capture in mat.captures {
4510 if capture.index == config.name_capture_ix {
4511 add_to_buffer_ranges(capture.node, true);
4512 } else if Some(capture.index) == config.context_capture_ix
4513 || (Some(capture.index) == config.extra_context_capture_ix && include_extra_context)
4514 {
4515 add_to_buffer_ranges(capture.node, false);
4516 } else {
4517 if Some(capture.index) == config.open_capture_ix {
4518 open_point = Some(Point::from_ts_point(capture.node.end_position()));
4519 } else if Some(capture.index) == config.close_capture_ix {
4520 close_point = Some(Point::from_ts_point(capture.node.start_position()));
4521 }
4522 }
4523 }
4524
4525 if buffer_ranges.is_empty() {
4526 return None;
4527 }
4528 let source_range_for_text =
4529 buffer_ranges.first().unwrap().0.start..buffer_ranges.last().unwrap().0.end;
4530
4531 let mut text = String::new();
4532 let mut highlight_ranges = Vec::new();
4533 let mut name_ranges = Vec::new();
4534 let mut chunks = self.chunks(source_range_for_text.clone(), true);
4535 let mut last_buffer_range_end = 0;
4536 for (buffer_range, is_name) in buffer_ranges {
4537 let space_added = !text.is_empty() && buffer_range.start > last_buffer_range_end;
4538 if space_added {
4539 text.push(' ');
4540 }
4541 let before_append_len = text.len();
4542 let mut offset = buffer_range.start;
4543 chunks.seek(buffer_range.clone());
4544 for mut chunk in chunks.by_ref() {
4545 if chunk.text.len() > buffer_range.end - offset {
4546 chunk.text = &chunk.text[0..(buffer_range.end - offset)];
4547 offset = buffer_range.end;
4548 } else {
4549 offset += chunk.text.len();
4550 }
4551 let style = chunk
4552 .syntax_highlight_id
4553 .zip(theme)
4554 .and_then(|(highlight, theme)| highlight.style(theme));
4555 if let Some(style) = style {
4556 let start = text.len();
4557 let end = start + chunk.text.len();
4558 highlight_ranges.push((start..end, style));
4559 }
4560 text.push_str(chunk.text);
4561 if offset >= buffer_range.end {
4562 break;
4563 }
4564 }
4565 if is_name {
4566 let after_append_len = text.len();
4567 let start = if space_added && !name_ranges.is_empty() {
4568 before_append_len - 1
4569 } else {
4570 before_append_len
4571 };
4572 name_ranges.push(start..after_append_len);
4573 }
4574 last_buffer_range_end = buffer_range.end;
4575 }
4576
4577 Some(OutlineItem {
4578 depth: 0, // We'll calculate the depth later
4579 range: item_point_range,
4580 source_range_for_text: source_range_for_text.to_point(self),
4581 text,
4582 highlight_ranges,
4583 name_ranges,
4584 body_range: open_point.zip(close_point).map(|(start, end)| start..end),
4585 annotation_range: None,
4586 })
4587 }
4588
4589 pub fn function_body_fold_ranges<T: ToOffset>(
4590 &self,
4591 within: Range<T>,
4592 ) -> impl Iterator<Item = Range<usize>> + '_ {
4593 self.text_object_ranges(within, TreeSitterOptions::default())
4594 .filter_map(|(range, obj)| (obj == TextObject::InsideFunction).then_some(range))
4595 }
4596
4597 /// For each grammar in the language, runs the provided
4598 /// [`tree_sitter::Query`] against the given range.
4599 pub fn matches(
4600 &self,
4601 range: Range<usize>,
4602 query: fn(&Grammar) -> Option<&tree_sitter::Query>,
4603 ) -> SyntaxMapMatches<'_> {
4604 self.syntax.matches(range, self, query)
4605 }
4606
4607 /// Finds all [`RowChunks`] applicable to the given range, then returns all bracket pairs that intersect with those chunks.
4608 /// Hence, may return more bracket pairs than the range contains.
4609 ///
4610 /// Will omit known chunks.
4611 /// The resulting bracket match collections are not ordered.
4612 pub fn fetch_bracket_ranges(
4613 &self,
4614 range: Range<usize>,
4615 known_chunks: Option<&HashSet<Range<BufferRow>>>,
4616 ) -> HashMap<Range<BufferRow>, Vec<BracketMatch<usize>>> {
4617 let mut all_bracket_matches = HashMap::default();
4618
4619 for chunk in self
4620 .tree_sitter_data
4621 .chunks
4622 .applicable_chunks(&[range.to_point(self)])
4623 {
4624 if known_chunks.is_some_and(|chunks| chunks.contains(&chunk.row_range())) {
4625 continue;
4626 }
4627 let chunk_range = chunk.anchor_range();
4628 let chunk_range = chunk_range.to_offset(&self);
4629
4630 if let Some(cached_brackets) =
4631 &self.tree_sitter_data.brackets_by_chunks.lock()[chunk.id]
4632 {
4633 all_bracket_matches.insert(chunk.row_range(), cached_brackets.clone());
4634 continue;
4635 }
4636
4637 let mut all_brackets: Vec<(BracketMatch<usize>, usize, bool)> = Vec::new();
4638 let mut opens = Vec::new();
4639 let mut color_pairs = Vec::new();
4640
4641 let mut matches = self.syntax.matches_with_options(
4642 chunk_range.clone(),
4643 &self.text,
4644 TreeSitterOptions {
4645 max_bytes_to_query: Some(MAX_BYTES_TO_QUERY),
4646 max_start_depth: None,
4647 },
4648 |grammar| grammar.brackets_config.as_ref().map(|c| &c.query),
4649 );
4650 let configs = matches
4651 .grammars()
4652 .iter()
4653 .map(|grammar| grammar.brackets_config.as_ref().unwrap())
4654 .collect::<Vec<_>>();
4655
4656 // Group matches by open range so we can either trust grammar output
4657 // or repair it by picking a single closest close per open.
4658 let mut open_to_close_ranges = BTreeMap::new();
4659 while let Some(mat) = matches.peek() {
4660 let mut open = None;
4661 let mut close = None;
4662 let syntax_layer_depth = mat.depth;
4663 let pattern_index = mat.pattern_index;
4664 let config = configs[mat.grammar_index];
4665 let pattern = &config.patterns[pattern_index];
4666 for capture in mat.captures {
4667 if capture.index == config.open_capture_ix {
4668 open = Some(capture.node.byte_range());
4669 } else if capture.index == config.close_capture_ix {
4670 close = Some(capture.node.byte_range());
4671 }
4672 }
4673
4674 matches.advance();
4675
4676 let Some((open_range, close_range)) = open.zip(close) else {
4677 continue;
4678 };
4679
4680 let bracket_range = open_range.start..=close_range.end;
4681 if !bracket_range.overlaps(&chunk_range) {
4682 continue;
4683 }
4684
4685 open_to_close_ranges
4686 .entry((open_range.start, open_range.end, pattern_index))
4687 .or_insert_with(BTreeMap::new)
4688 .insert(
4689 (close_range.start, close_range.end),
4690 BracketMatch {
4691 open_range: open_range.clone(),
4692 close_range: close_range.clone(),
4693 syntax_layer_depth,
4694 newline_only: pattern.newline_only,
4695 color_index: None,
4696 },
4697 );
4698
4699 all_brackets.push((
4700 BracketMatch {
4701 open_range,
4702 close_range,
4703 syntax_layer_depth,
4704 newline_only: pattern.newline_only,
4705 color_index: None,
4706 },
4707 pattern_index,
4708 pattern.rainbow_exclude,
4709 ));
4710 }
4711
4712 let has_bogus_matches = open_to_close_ranges
4713 .iter()
4714 .any(|(_, end_ranges)| end_ranges.len() > 1);
4715 if has_bogus_matches {
4716 // Grammar is producing bogus matches where one open is paired with multiple
4717 // closes. Build a valid stack by walking through positions in order.
4718 // For each close, we know the expected open_len from tree-sitter matches.
4719
4720 // Map each close to its expected open length (for inferring opens)
4721 let close_to_open_len: HashMap<(usize, usize, usize), usize> = all_brackets
4722 .iter()
4723 .map(|(bracket_match, pattern_index, _)| {
4724 (
4725 (
4726 bracket_match.close_range.start,
4727 bracket_match.close_range.end,
4728 *pattern_index,
4729 ),
4730 bracket_match.open_range.len(),
4731 )
4732 })
4733 .collect();
4734
4735 // Collect unique opens and closes within this chunk
4736 let mut unique_opens: HashSet<(usize, usize, usize)> = all_brackets
4737 .iter()
4738 .map(|(bracket_match, pattern_index, _)| {
4739 (
4740 bracket_match.open_range.start,
4741 bracket_match.open_range.end,
4742 *pattern_index,
4743 )
4744 })
4745 .filter(|(start, _, _)| chunk_range.contains(start))
4746 .collect();
4747
4748 let mut unique_closes: Vec<(usize, usize, usize)> = all_brackets
4749 .iter()
4750 .map(|(bracket_match, pattern_index, _)| {
4751 (
4752 bracket_match.close_range.start,
4753 bracket_match.close_range.end,
4754 *pattern_index,
4755 )
4756 })
4757 .filter(|(start, _, _)| chunk_range.contains(start))
4758 .collect();
4759 unique_closes.sort();
4760 unique_closes.dedup();
4761
4762 // Build valid pairs by walking through closes in order
4763 let mut unique_opens_vec: Vec<_> = unique_opens.iter().copied().collect();
4764 unique_opens_vec.sort();
4765
4766 let mut valid_pairs: HashSet<((usize, usize, usize), (usize, usize, usize))> =
4767 HashSet::default();
4768 let mut open_stacks: HashMap<usize, Vec<(usize, usize)>> = HashMap::default();
4769 let mut open_idx = 0;
4770
4771 for close in &unique_closes {
4772 // Push all opens before this close onto stack
4773 while open_idx < unique_opens_vec.len()
4774 && unique_opens_vec[open_idx].0 < close.0
4775 {
4776 let (start, end, pattern_index) = unique_opens_vec[open_idx];
4777 open_stacks
4778 .entry(pattern_index)
4779 .or_default()
4780 .push((start, end));
4781 open_idx += 1;
4782 }
4783
4784 // Try to match with most recent open
4785 let (close_start, close_end, pattern_index) = *close;
4786 if let Some(open) = open_stacks
4787 .get_mut(&pattern_index)
4788 .and_then(|open_stack| open_stack.pop())
4789 {
4790 valid_pairs.insert(((open.0, open.1, pattern_index), *close));
4791 } else if let Some(&open_len) = close_to_open_len.get(close) {
4792 // No open on stack - infer one based on expected open_len
4793 if close_start >= open_len {
4794 let inferred = (close_start - open_len, close_start, pattern_index);
4795 unique_opens.insert(inferred);
4796 valid_pairs.insert((inferred, *close));
4797 all_brackets.push((
4798 BracketMatch {
4799 open_range: inferred.0..inferred.1,
4800 close_range: close_start..close_end,
4801 newline_only: false,
4802 syntax_layer_depth: 0,
4803 color_index: None,
4804 },
4805 pattern_index,
4806 false,
4807 ));
4808 }
4809 }
4810 }
4811
4812 all_brackets.retain(|(bracket_match, pattern_index, _)| {
4813 let open = (
4814 bracket_match.open_range.start,
4815 bracket_match.open_range.end,
4816 *pattern_index,
4817 );
4818 let close = (
4819 bracket_match.close_range.start,
4820 bracket_match.close_range.end,
4821 *pattern_index,
4822 );
4823 valid_pairs.contains(&(open, close))
4824 });
4825 }
4826
4827 let mut all_brackets = all_brackets
4828 .into_iter()
4829 .enumerate()
4830 .map(|(index, (bracket_match, _, rainbow_exclude))| {
4831 // Certain languages have "brackets" that are not brackets, e.g. tags. and such
4832 // bracket will match the entire tag with all text inside.
4833 // For now, avoid highlighting any pair that has more than single char in each bracket.
4834 // We need to colorize `<Element/>` bracket pairs, so cannot make this check stricter.
4835 let should_color = !rainbow_exclude
4836 && (bracket_match.open_range.len() == 1
4837 || bracket_match.close_range.len() == 1);
4838 if should_color {
4839 opens.push(bracket_match.open_range.clone());
4840 color_pairs.push((
4841 bracket_match.open_range.clone(),
4842 bracket_match.close_range.clone(),
4843 index,
4844 ));
4845 }
4846 bracket_match
4847 })
4848 .collect::<Vec<_>>();
4849
4850 opens.sort_by_key(|r| (r.start, r.end));
4851 opens.dedup_by(|a, b| a.start == b.start && a.end == b.end);
4852 color_pairs.sort_by_key(|(_, close, _)| close.end);
4853
4854 let mut open_stack = Vec::new();
4855 let mut open_index = 0;
4856 for (open, close, index) in color_pairs {
4857 while open_index < opens.len() && opens[open_index].start < close.start {
4858 open_stack.push(opens[open_index].clone());
4859 open_index += 1;
4860 }
4861
4862 if open_stack.last() == Some(&open) {
4863 let depth_index = open_stack.len() - 1;
4864 all_brackets[index].color_index = Some(depth_index);
4865 open_stack.pop();
4866 }
4867 }
4868
4869 all_brackets.sort_by_key(|bracket_match| {
4870 (bracket_match.open_range.start, bracket_match.open_range.end)
4871 });
4872
4873 if let empty_slot @ None =
4874 &mut self.tree_sitter_data.brackets_by_chunks.lock()[chunk.id]
4875 {
4876 *empty_slot = Some(all_brackets.clone());
4877 }
4878 all_bracket_matches.insert(chunk.row_range(), all_brackets);
4879 }
4880
4881 all_bracket_matches
4882 }
4883
4884 pub fn all_bracket_ranges(
4885 &self,
4886 range: Range<usize>,
4887 ) -> impl Iterator<Item = BracketMatch<usize>> {
4888 self.fetch_bracket_ranges(range.clone(), None)
4889 .into_values()
4890 .flatten()
4891 .filter(move |bracket_match| {
4892 let bracket_range = bracket_match.open_range.start..bracket_match.close_range.end;
4893 bracket_range.overlaps(&range)
4894 })
4895 }
4896
4897 /// Returns bracket range pairs overlapping or adjacent to `range`
4898 pub fn bracket_ranges<T: ToOffset>(
4899 &self,
4900 range: Range<T>,
4901 ) -> impl Iterator<Item = BracketMatch<usize>> + '_ {
4902 // Find bracket pairs that *inclusively* contain the given range.
4903 let range = range.start.to_previous_offset(self)..range.end.to_next_offset(self);
4904 self.all_bracket_ranges(range)
4905 .filter(|pair| !pair.newline_only)
4906 }
4907
4908 pub fn debug_variables_query<T: ToOffset>(
4909 &self,
4910 range: Range<T>,
4911 ) -> impl Iterator<Item = (Range<usize>, DebuggerTextObject)> + '_ {
4912 let range = range.start.to_previous_offset(self)..range.end.to_next_offset(self);
4913
4914 let mut matches = self.syntax.matches_with_options(
4915 range.clone(),
4916 &self.text,
4917 TreeSitterOptions::default(),
4918 |grammar| grammar.debug_variables_config.as_ref().map(|c| &c.query),
4919 );
4920
4921 let configs = matches
4922 .grammars()
4923 .iter()
4924 .map(|grammar| grammar.debug_variables_config.as_ref())
4925 .collect::<Vec<_>>();
4926
4927 let mut captures = Vec::<(Range<usize>, DebuggerTextObject)>::new();
4928
4929 iter::from_fn(move || {
4930 loop {
4931 while let Some(capture) = captures.pop() {
4932 if capture.0.overlaps(&range) {
4933 return Some(capture);
4934 }
4935 }
4936
4937 let mat = matches.peek()?;
4938
4939 let Some(config) = configs[mat.grammar_index].as_ref() else {
4940 matches.advance();
4941 continue;
4942 };
4943
4944 for capture in mat.captures {
4945 let Some(ix) = config
4946 .objects_by_capture_ix
4947 .binary_search_by_key(&capture.index, |e| e.0)
4948 .ok()
4949 else {
4950 continue;
4951 };
4952 let text_object = config.objects_by_capture_ix[ix].1;
4953 let byte_range = capture.node.byte_range();
4954
4955 let mut found = false;
4956 for (range, existing) in captures.iter_mut() {
4957 if existing == &text_object {
4958 range.start = range.start.min(byte_range.start);
4959 range.end = range.end.max(byte_range.end);
4960 found = true;
4961 break;
4962 }
4963 }
4964
4965 if !found {
4966 captures.push((byte_range, text_object));
4967 }
4968 }
4969
4970 matches.advance();
4971 }
4972 })
4973 }
4974
4975 pub fn text_object_ranges<T: ToOffset>(
4976 &self,
4977 range: Range<T>,
4978 options: TreeSitterOptions,
4979 ) -> impl Iterator<Item = (Range<usize>, TextObject)> + '_ {
4980 let range =
4981 range.start.to_previous_offset(self)..self.len().min(range.end.to_next_offset(self));
4982
4983 let mut matches =
4984 self.syntax
4985 .matches_with_options(range.clone(), &self.text, options, |grammar| {
4986 grammar.text_object_config.as_ref().map(|c| &c.query)
4987 });
4988
4989 let configs = matches
4990 .grammars()
4991 .iter()
4992 .map(|grammar| grammar.text_object_config.as_ref())
4993 .collect::<Vec<_>>();
4994
4995 let mut captures = Vec::<(Range<usize>, TextObject)>::new();
4996
4997 iter::from_fn(move || {
4998 loop {
4999 while let Some(capture) = captures.pop() {
5000 if capture.0.overlaps(&range) {
5001 return Some(capture);
5002 }
5003 }
5004
5005 let mat = matches.peek()?;
5006
5007 let Some(config) = configs[mat.grammar_index].as_ref() else {
5008 matches.advance();
5009 continue;
5010 };
5011
5012 for capture in mat.captures {
5013 let Some(ix) = config
5014 .text_objects_by_capture_ix
5015 .binary_search_by_key(&capture.index, |e| e.0)
5016 .ok()
5017 else {
5018 continue;
5019 };
5020 let text_object = config.text_objects_by_capture_ix[ix].1;
5021 let byte_range = capture.node.byte_range();
5022
5023 let mut found = false;
5024 for (range, existing) in captures.iter_mut() {
5025 if existing == &text_object {
5026 range.start = range.start.min(byte_range.start);
5027 range.end = range.end.max(byte_range.end);
5028 found = true;
5029 break;
5030 }
5031 }
5032
5033 if !found {
5034 captures.push((byte_range, text_object));
5035 }
5036 }
5037
5038 matches.advance();
5039 }
5040 })
5041 }
5042
5043 /// Returns enclosing bracket ranges containing the given range
5044 pub fn enclosing_bracket_ranges<T: ToOffset>(
5045 &self,
5046 range: Range<T>,
5047 ) -> impl Iterator<Item = BracketMatch<usize>> + '_ {
5048 let range = range.start.to_offset(self)..range.end.to_offset(self);
5049
5050 let result: Vec<_> = self.bracket_ranges(range.clone()).collect();
5051 let max_depth = result
5052 .iter()
5053 .map(|mat| mat.syntax_layer_depth)
5054 .max()
5055 .unwrap_or(0);
5056 result.into_iter().filter(move |pair| {
5057 pair.open_range.start <= range.start
5058 && pair.close_range.end >= range.end
5059 && pair.syntax_layer_depth == max_depth
5060 })
5061 }
5062
5063 /// Returns the smallest enclosing bracket ranges containing the given range or None if no brackets contain range
5064 ///
5065 /// Can optionally pass a range_filter to filter the ranges of brackets to consider
5066 pub fn innermost_enclosing_bracket_ranges<T: ToOffset>(
5067 &self,
5068 range: Range<T>,
5069 range_filter: Option<&dyn Fn(Range<usize>, Range<usize>) -> bool>,
5070 ) -> Option<(Range<usize>, Range<usize>)> {
5071 let range = range.start.to_offset(self)..range.end.to_offset(self);
5072
5073 // Get the ranges of the innermost pair of brackets.
5074 let mut result: Option<(Range<usize>, Range<usize>)> = None;
5075
5076 for pair in self.enclosing_bracket_ranges(range) {
5077 if let Some(range_filter) = range_filter
5078 && !range_filter(pair.open_range.clone(), pair.close_range.clone())
5079 {
5080 continue;
5081 }
5082
5083 let len = pair.close_range.end - pair.open_range.start;
5084
5085 if let Some((existing_open, existing_close)) = &result {
5086 let existing_len = existing_close.end - existing_open.start;
5087 if len > existing_len {
5088 continue;
5089 }
5090 }
5091
5092 result = Some((pair.open_range, pair.close_range));
5093 }
5094
5095 result
5096 }
5097
5098 /// Returns anchor ranges for any matches of the redaction query.
5099 /// The buffer can be associated with multiple languages, and the redaction query associated with each
5100 /// will be run on the relevant section of the buffer.
5101 pub fn redacted_ranges<T: ToOffset>(
5102 &self,
5103 range: Range<T>,
5104 ) -> impl Iterator<Item = Range<usize>> + '_ {
5105 let offset_range = range.start.to_offset(self)..range.end.to_offset(self);
5106 let mut syntax_matches = self.syntax.matches(offset_range, self, |grammar| {
5107 grammar
5108 .redactions_config
5109 .as_ref()
5110 .map(|config| &config.query)
5111 });
5112
5113 let configs = syntax_matches
5114 .grammars()
5115 .iter()
5116 .map(|grammar| grammar.redactions_config.as_ref())
5117 .collect::<Vec<_>>();
5118
5119 iter::from_fn(move || {
5120 let redacted_range = syntax_matches
5121 .peek()
5122 .and_then(|mat| {
5123 configs[mat.grammar_index].and_then(|config| {
5124 mat.captures
5125 .iter()
5126 .find(|capture| capture.index == config.redaction_capture_ix)
5127 })
5128 })
5129 .map(|mat| mat.node.byte_range());
5130 syntax_matches.advance();
5131 redacted_range
5132 })
5133 }
5134
5135 pub fn injections_intersecting_range<T: ToOffset>(
5136 &self,
5137 range: Range<T>,
5138 ) -> impl Iterator<Item = (Range<usize>, &Arc<Language>)> + '_ {
5139 let offset_range = range.start.to_offset(self)..range.end.to_offset(self);
5140
5141 let mut syntax_matches = self.syntax.matches(offset_range, self, |grammar| {
5142 grammar
5143 .injection_config
5144 .as_ref()
5145 .map(|config| &config.query)
5146 });
5147
5148 let configs = syntax_matches
5149 .grammars()
5150 .iter()
5151 .map(|grammar| grammar.injection_config.as_ref())
5152 .collect::<Vec<_>>();
5153
5154 iter::from_fn(move || {
5155 let ranges = syntax_matches.peek().and_then(|mat| {
5156 let config = &configs[mat.grammar_index]?;
5157 let content_capture_range = mat.captures.iter().find_map(|capture| {
5158 if capture.index == config.content_capture_ix {
5159 Some(capture.node.byte_range())
5160 } else {
5161 None
5162 }
5163 })?;
5164 let language = self.language_at(content_capture_range.start)?;
5165 Some((content_capture_range, language))
5166 });
5167 syntax_matches.advance();
5168 ranges
5169 })
5170 }
5171
5172 pub fn runnable_ranges(
5173 &self,
5174 offset_range: Range<usize>,
5175 ) -> impl Iterator<Item = RunnableRange> + '_ {
5176 let mut syntax_matches = self.syntax.matches(offset_range, self, |grammar| {
5177 grammar.runnable_config.as_ref().map(|config| &config.query)
5178 });
5179
5180 let test_configs = syntax_matches
5181 .grammars()
5182 .iter()
5183 .map(|grammar| grammar.runnable_config.as_ref())
5184 .collect::<Vec<_>>();
5185
5186 iter::from_fn(move || {
5187 loop {
5188 let mat = syntax_matches.peek()?;
5189
5190 let test_range = test_configs[mat.grammar_index].and_then(|test_configs| {
5191 let mut run_range = None;
5192 let full_range = mat.captures.iter().fold(
5193 Range {
5194 start: usize::MAX,
5195 end: 0,
5196 },
5197 |mut acc, next| {
5198 let byte_range = next.node.byte_range();
5199 if acc.start > byte_range.start {
5200 acc.start = byte_range.start;
5201 }
5202 if acc.end < byte_range.end {
5203 acc.end = byte_range.end;
5204 }
5205 acc
5206 },
5207 );
5208 if full_range.start > full_range.end {
5209 // We did not find a full spanning range of this match.
5210 return None;
5211 }
5212 let extra_captures: SmallVec<[_; 1]> =
5213 SmallVec::from_iter(mat.captures.iter().filter_map(|capture| {
5214 test_configs
5215 .extra_captures
5216 .get(capture.index as usize)
5217 .cloned()
5218 .and_then(|tag_name| match tag_name {
5219 RunnableCapture::Named(name) => {
5220 Some((capture.node.byte_range(), name))
5221 }
5222 RunnableCapture::Run => {
5223 let _ = run_range.insert(capture.node.byte_range());
5224 None
5225 }
5226 })
5227 }));
5228 let run_range = run_range?;
5229 let tags = test_configs
5230 .query
5231 .property_settings(mat.pattern_index)
5232 .iter()
5233 .filter_map(|property| {
5234 if *property.key == *"tag" {
5235 property
5236 .value
5237 .as_ref()
5238 .map(|value| RunnableTag(value.to_string().into()))
5239 } else {
5240 None
5241 }
5242 })
5243 .collect();
5244 let extra_captures = extra_captures
5245 .into_iter()
5246 .map(|(range, name)| {
5247 (
5248 name.to_string(),
5249 self.text_for_range(range).collect::<String>(),
5250 )
5251 })
5252 .collect();
5253 // All tags should have the same range.
5254 Some(RunnableRange {
5255 run_range,
5256 full_range,
5257 runnable: Runnable {
5258 tags,
5259 language: mat.language,
5260 buffer: self.remote_id(),
5261 },
5262 extra_captures,
5263 buffer_id: self.remote_id(),
5264 })
5265 });
5266
5267 syntax_matches.advance();
5268 if test_range.is_some() {
5269 // It's fine for us to short-circuit on .peek()? returning None. We don't want to return None from this iter if we
5270 // had a capture that did not contain a run marker, hence we'll just loop around for the next capture.
5271 return test_range;
5272 }
5273 }
5274 })
5275 }
5276
5277 /// Returns selections for remote peers intersecting the given range.
5278 #[allow(clippy::type_complexity)]
5279 pub fn selections_in_range(
5280 &self,
5281 range: Range<Anchor>,
5282 include_local: bool,
5283 ) -> impl Iterator<
5284 Item = (
5285 ReplicaId,
5286 bool,
5287 CursorShape,
5288 impl Iterator<Item = &Selection<Anchor>> + '_,
5289 ),
5290 > + '_ {
5291 self.remote_selections
5292 .iter()
5293 .filter(move |(replica_id, set)| {
5294 (include_local || **replica_id != self.text.replica_id())
5295 && !set.selections.is_empty()
5296 })
5297 .map(move |(replica_id, set)| {
5298 let start_ix = match set.selections.binary_search_by(|probe| {
5299 probe.end.cmp(&range.start, self).then(Ordering::Greater)
5300 }) {
5301 Ok(ix) | Err(ix) => ix,
5302 };
5303 let end_ix = match set.selections.binary_search_by(|probe| {
5304 probe.start.cmp(&range.end, self).then(Ordering::Less)
5305 }) {
5306 Ok(ix) | Err(ix) => ix,
5307 };
5308
5309 (
5310 *replica_id,
5311 set.line_mode,
5312 set.cursor_shape,
5313 set.selections[start_ix..end_ix].iter(),
5314 )
5315 })
5316 }
5317
5318 /// Returns if the buffer contains any diagnostics.
5319 pub fn has_diagnostics(&self) -> bool {
5320 !self.diagnostics.is_empty()
5321 }
5322
5323 /// Returns all the diagnostics intersecting the given range.
5324 pub fn diagnostics_in_range<'a, T, O>(
5325 &'a self,
5326 search_range: Range<T>,
5327 reversed: bool,
5328 ) -> impl 'a + Iterator<Item = DiagnosticEntryRef<'a, O>>
5329 where
5330 T: 'a + Clone + ToOffset,
5331 O: 'a + FromAnchor,
5332 {
5333 let mut iterators: Vec<_> = self
5334 .diagnostics
5335 .iter()
5336 .map(|(_, collection)| {
5337 collection
5338 .range::<T, text::Anchor>(search_range.clone(), self, true, reversed)
5339 .peekable()
5340 })
5341 .collect();
5342
5343 std::iter::from_fn(move || {
5344 let (next_ix, _) = iterators
5345 .iter_mut()
5346 .enumerate()
5347 .flat_map(|(ix, iter)| Some((ix, iter.peek()?)))
5348 .min_by(|(_, a), (_, b)| {
5349 let cmp = a
5350 .range
5351 .start
5352 .cmp(&b.range.start, self)
5353 // when range is equal, sort by diagnostic severity
5354 .then(a.diagnostic.severity.cmp(&b.diagnostic.severity))
5355 // and stabilize order with group_id
5356 .then(a.diagnostic.group_id.cmp(&b.diagnostic.group_id));
5357 if reversed { cmp.reverse() } else { cmp }
5358 })?;
5359 iterators[next_ix]
5360 .next()
5361 .map(
5362 |DiagnosticEntryRef { range, diagnostic }| DiagnosticEntryRef {
5363 diagnostic,
5364 range: FromAnchor::from_anchor(&range.start, self)
5365 ..FromAnchor::from_anchor(&range.end, self),
5366 },
5367 )
5368 })
5369 }
5370
5371 /// Returns all the diagnostic groups associated with the given
5372 /// language server ID. If no language server ID is provided,
5373 /// all diagnostics groups are returned.
5374 pub fn diagnostic_groups(
5375 &self,
5376 language_server_id: Option<LanguageServerId>,
5377 ) -> Vec<(LanguageServerId, DiagnosticGroup<'_, Anchor>)> {
5378 let mut groups = Vec::new();
5379
5380 if let Some(language_server_id) = language_server_id {
5381 if let Some(set) = self.diagnostics.get(&language_server_id) {
5382 set.groups(language_server_id, &mut groups, self);
5383 }
5384 } else {
5385 for (language_server_id, diagnostics) in self.diagnostics.iter() {
5386 diagnostics.groups(*language_server_id, &mut groups, self);
5387 }
5388 }
5389
5390 groups.sort_by(|(id_a, group_a), (id_b, group_b)| {
5391 let a_start = &group_a.entries[group_a.primary_ix].range.start;
5392 let b_start = &group_b.entries[group_b.primary_ix].range.start;
5393 a_start.cmp(b_start, self).then_with(|| id_a.cmp(id_b))
5394 });
5395
5396 groups
5397 }
5398
5399 /// Returns an iterator over the diagnostics for the given group.
5400 pub fn diagnostic_group<O>(
5401 &self,
5402 group_id: usize,
5403 ) -> impl Iterator<Item = DiagnosticEntryRef<'_, O>> + use<'_, O>
5404 where
5405 O: FromAnchor + 'static,
5406 {
5407 self.diagnostics
5408 .iter()
5409 .flat_map(move |(_, set)| set.group(group_id, self))
5410 }
5411
5412 /// An integer version number that accounts for all updates besides
5413 /// the buffer's text itself (which is versioned via a version vector).
5414 pub fn non_text_state_update_count(&self) -> usize {
5415 self.non_text_state_update_count
5416 }
5417
5418 /// An integer version that changes when the buffer's syntax changes.
5419 pub fn syntax_update_count(&self) -> usize {
5420 self.syntax.update_count()
5421 }
5422
5423 /// Returns a snapshot of underlying file.
5424 pub fn file(&self) -> Option<&Arc<dyn File>> {
5425 self.file.as_ref()
5426 }
5427
5428 pub fn resolve_file_path(&self, include_root: bool, cx: &App) -> Option<String> {
5429 if let Some(file) = self.file() {
5430 if file.path().file_name().is_none() || include_root {
5431 Some(file.full_path(cx).to_string_lossy().into_owned())
5432 } else {
5433 Some(file.path().display(file.path_style(cx)).to_string())
5434 }
5435 } else {
5436 None
5437 }
5438 }
5439
5440 pub fn words_in_range(&self, query: WordsQuery) -> BTreeMap<String, Range<Anchor>> {
5441 let query_str = query.fuzzy_contents;
5442 if query_str.is_some_and(|query| query.is_empty()) {
5443 return BTreeMap::default();
5444 }
5445
5446 let classifier = CharClassifier::new(self.language.clone().map(|language| LanguageScope {
5447 language,
5448 override_id: None,
5449 }));
5450
5451 let mut query_ix = 0;
5452 let query_chars = query_str.map(|query| query.chars().collect::<Vec<_>>());
5453 let query_len = query_chars.as_ref().map_or(0, |query| query.len());
5454
5455 let mut words = BTreeMap::default();
5456 let mut current_word_start_ix = None;
5457 let mut chunk_ix = query.range.start;
5458 for chunk in self.chunks(query.range, false) {
5459 for (i, c) in chunk.text.char_indices() {
5460 let ix = chunk_ix + i;
5461 if classifier.is_word(c) {
5462 if current_word_start_ix.is_none() {
5463 current_word_start_ix = Some(ix);
5464 }
5465
5466 if let Some(query_chars) = &query_chars
5467 && query_ix < query_len
5468 && c.to_lowercase().eq(query_chars[query_ix].to_lowercase())
5469 {
5470 query_ix += 1;
5471 }
5472 continue;
5473 } else if let Some(word_start) = current_word_start_ix.take()
5474 && query_ix == query_len
5475 {
5476 let word_range = self.anchor_before(word_start)..self.anchor_after(ix);
5477 let mut word_text = self.text_for_range(word_start..ix).peekable();
5478 let first_char = word_text
5479 .peek()
5480 .and_then(|first_chunk| first_chunk.chars().next());
5481 // Skip empty and "words" starting with digits as a heuristic to reduce useless completions
5482 if !query.skip_digits
5483 || first_char.is_none_or(|first_char| !first_char.is_digit(10))
5484 {
5485 words.insert(word_text.collect(), word_range);
5486 }
5487 }
5488 query_ix = 0;
5489 }
5490 chunk_ix += chunk.text.len();
5491 }
5492
5493 words
5494 }
5495}
5496
5497pub struct WordsQuery<'a> {
5498 /// Only returns words with all chars from the fuzzy string in them.
5499 pub fuzzy_contents: Option<&'a str>,
5500 /// Skips words that start with a digit.
5501 pub skip_digits: bool,
5502 /// Buffer offset range, to look for words.
5503 pub range: Range<usize>,
5504}
5505
5506fn indent_size_for_line(text: &text::BufferSnapshot, row: u32) -> IndentSize {
5507 indent_size_for_text(text.chars_at(Point::new(row, 0)))
5508}
5509
5510fn indent_size_for_text(text: impl Iterator<Item = char>) -> IndentSize {
5511 let mut result = IndentSize::spaces(0);
5512 for c in text {
5513 let kind = match c {
5514 ' ' => IndentKind::Space,
5515 '\t' => IndentKind::Tab,
5516 _ => break,
5517 };
5518 if result.len == 0 {
5519 result.kind = kind;
5520 }
5521 result.len += 1;
5522 }
5523 result
5524}
5525
5526impl Clone for BufferSnapshot {
5527 fn clone(&self) -> Self {
5528 Self {
5529 text: self.text.clone(),
5530 syntax: self.syntax.clone(),
5531 file: self.file.clone(),
5532 remote_selections: self.remote_selections.clone(),
5533 diagnostics: self.diagnostics.clone(),
5534 language: self.language.clone(),
5535 tree_sitter_data: self.tree_sitter_data.clone(),
5536 non_text_state_update_count: self.non_text_state_update_count,
5537 capability: self.capability,
5538 modeline: self.modeline.clone(),
5539 }
5540 }
5541}
5542
5543impl Deref for BufferSnapshot {
5544 type Target = text::BufferSnapshot;
5545
5546 fn deref(&self) -> &Self::Target {
5547 &self.text
5548 }
5549}
5550
5551unsafe impl Send for BufferChunks<'_> {}
5552
5553impl<'a> BufferChunks<'a> {
5554 pub(crate) fn new(
5555 text: &'a Rope,
5556 range: Range<usize>,
5557 syntax: Option<(SyntaxMapCaptures<'a>, Vec<HighlightMap>)>,
5558 diagnostics: bool,
5559 buffer_snapshot: Option<&'a BufferSnapshot>,
5560 ) -> Self {
5561 let mut highlights = None;
5562 if let Some((captures, highlight_maps)) = syntax {
5563 highlights = Some(BufferChunkHighlights {
5564 captures,
5565 next_capture: None,
5566 stack: Default::default(),
5567 highlight_maps,
5568 })
5569 }
5570
5571 let diagnostic_endpoints = diagnostics.then(|| Vec::new().into_iter().peekable());
5572 let chunks = text.chunks_in_range(range.clone());
5573
5574 let mut this = BufferChunks {
5575 range,
5576 buffer_snapshot,
5577 chunks,
5578 diagnostic_endpoints,
5579 error_depth: 0,
5580 warning_depth: 0,
5581 information_depth: 0,
5582 hint_depth: 0,
5583 unnecessary_depth: 0,
5584 underline: true,
5585 highlights,
5586 };
5587 this.initialize_diagnostic_endpoints();
5588 this
5589 }
5590
5591 /// Seeks to the given byte offset in the buffer.
5592 pub fn seek(&mut self, range: Range<usize>) {
5593 let old_range = std::mem::replace(&mut self.range, range.clone());
5594 self.chunks.set_range(self.range.clone());
5595 if let Some(highlights) = self.highlights.as_mut() {
5596 if old_range.start <= self.range.start && old_range.end >= self.range.end {
5597 // Reuse existing highlights stack, as the new range is a subrange of the old one.
5598 highlights
5599 .stack
5600 .retain(|(end_offset, _)| *end_offset > range.start);
5601 if let Some(capture) = &highlights.next_capture
5602 && range.start >= capture.node.start_byte()
5603 {
5604 let next_capture_end = capture.node.end_byte();
5605 if range.start < next_capture_end {
5606 highlights.stack.push((
5607 next_capture_end,
5608 highlights.highlight_maps[capture.grammar_index].get(capture.index),
5609 ));
5610 }
5611 highlights.next_capture.take();
5612 }
5613 } else if let Some(snapshot) = self.buffer_snapshot {
5614 let (captures, highlight_maps) = snapshot.get_highlights(self.range.clone());
5615 *highlights = BufferChunkHighlights {
5616 captures,
5617 next_capture: None,
5618 stack: Default::default(),
5619 highlight_maps,
5620 };
5621 } else {
5622 // We cannot obtain new highlights for a language-aware buffer iterator, as we don't have a buffer snapshot.
5623 // Seeking such BufferChunks is not supported.
5624 debug_assert!(
5625 false,
5626 "Attempted to seek on a language-aware buffer iterator without associated buffer snapshot"
5627 );
5628 }
5629
5630 highlights.captures.set_byte_range(self.range.clone());
5631 self.initialize_diagnostic_endpoints();
5632 }
5633 }
5634
5635 fn initialize_diagnostic_endpoints(&mut self) {
5636 if let Some(diagnostics) = self.diagnostic_endpoints.as_mut()
5637 && let Some(buffer) = self.buffer_snapshot
5638 {
5639 let mut diagnostic_endpoints = Vec::new();
5640 for entry in buffer.diagnostics_in_range::<_, usize>(self.range.clone(), false) {
5641 diagnostic_endpoints.push(DiagnosticEndpoint {
5642 offset: entry.range.start,
5643 is_start: true,
5644 severity: entry.diagnostic.severity,
5645 is_unnecessary: entry.diagnostic.is_unnecessary,
5646 underline: entry.diagnostic.underline,
5647 });
5648 diagnostic_endpoints.push(DiagnosticEndpoint {
5649 offset: entry.range.end,
5650 is_start: false,
5651 severity: entry.diagnostic.severity,
5652 is_unnecessary: entry.diagnostic.is_unnecessary,
5653 underline: entry.diagnostic.underline,
5654 });
5655 }
5656 diagnostic_endpoints
5657 .sort_unstable_by_key(|endpoint| (endpoint.offset, !endpoint.is_start));
5658 *diagnostics = diagnostic_endpoints.into_iter().peekable();
5659 self.hint_depth = 0;
5660 self.error_depth = 0;
5661 self.warning_depth = 0;
5662 self.information_depth = 0;
5663 }
5664 }
5665
5666 /// The current byte offset in the buffer.
5667 pub fn offset(&self) -> usize {
5668 self.range.start
5669 }
5670
5671 pub fn range(&self) -> Range<usize> {
5672 self.range.clone()
5673 }
5674
5675 fn update_diagnostic_depths(&mut self, endpoint: DiagnosticEndpoint) {
5676 let depth = match endpoint.severity {
5677 DiagnosticSeverity::ERROR => &mut self.error_depth,
5678 DiagnosticSeverity::WARNING => &mut self.warning_depth,
5679 DiagnosticSeverity::INFORMATION => &mut self.information_depth,
5680 DiagnosticSeverity::HINT => &mut self.hint_depth,
5681 _ => return,
5682 };
5683 if endpoint.is_start {
5684 *depth += 1;
5685 } else {
5686 *depth -= 1;
5687 }
5688
5689 if endpoint.is_unnecessary {
5690 if endpoint.is_start {
5691 self.unnecessary_depth += 1;
5692 } else {
5693 self.unnecessary_depth -= 1;
5694 }
5695 }
5696 }
5697
5698 fn current_diagnostic_severity(&self) -> Option<DiagnosticSeverity> {
5699 if self.error_depth > 0 {
5700 Some(DiagnosticSeverity::ERROR)
5701 } else if self.warning_depth > 0 {
5702 Some(DiagnosticSeverity::WARNING)
5703 } else if self.information_depth > 0 {
5704 Some(DiagnosticSeverity::INFORMATION)
5705 } else if self.hint_depth > 0 {
5706 Some(DiagnosticSeverity::HINT)
5707 } else {
5708 None
5709 }
5710 }
5711
5712 fn current_code_is_unnecessary(&self) -> bool {
5713 self.unnecessary_depth > 0
5714 }
5715}
5716
5717impl<'a> Iterator for BufferChunks<'a> {
5718 type Item = Chunk<'a>;
5719
5720 fn next(&mut self) -> Option<Self::Item> {
5721 let mut next_capture_start = usize::MAX;
5722 let mut next_diagnostic_endpoint = usize::MAX;
5723
5724 if let Some(highlights) = self.highlights.as_mut() {
5725 while let Some((parent_capture_end, _)) = highlights.stack.last() {
5726 if *parent_capture_end <= self.range.start {
5727 highlights.stack.pop();
5728 } else {
5729 break;
5730 }
5731 }
5732
5733 if highlights.next_capture.is_none() {
5734 highlights.next_capture = highlights.captures.next();
5735 }
5736
5737 while let Some(capture) = highlights.next_capture.as_ref() {
5738 if self.range.start < capture.node.start_byte() {
5739 next_capture_start = capture.node.start_byte();
5740 break;
5741 } else {
5742 let highlight_id =
5743 highlights.highlight_maps[capture.grammar_index].get(capture.index);
5744 highlights
5745 .stack
5746 .push((capture.node.end_byte(), highlight_id));
5747 highlights.next_capture = highlights.captures.next();
5748 }
5749 }
5750 }
5751
5752 let mut diagnostic_endpoints = std::mem::take(&mut self.diagnostic_endpoints);
5753 if let Some(diagnostic_endpoints) = diagnostic_endpoints.as_mut() {
5754 while let Some(endpoint) = diagnostic_endpoints.peek().copied() {
5755 if endpoint.offset <= self.range.start {
5756 self.update_diagnostic_depths(endpoint);
5757 diagnostic_endpoints.next();
5758 self.underline = endpoint.underline;
5759 } else {
5760 next_diagnostic_endpoint = endpoint.offset;
5761 break;
5762 }
5763 }
5764 }
5765 self.diagnostic_endpoints = diagnostic_endpoints;
5766
5767 if let Some(ChunkBitmaps {
5768 text: chunk,
5769 chars: chars_map,
5770 tabs,
5771 newlines,
5772 }) = self.chunks.peek_with_bitmaps()
5773 {
5774 let chunk_start = self.range.start;
5775 let mut chunk_end = (self.chunks.offset() + chunk.len())
5776 .min(next_capture_start)
5777 .min(next_diagnostic_endpoint);
5778 let mut highlight_id = None;
5779 if let Some(highlights) = self.highlights.as_ref()
5780 && let Some((parent_capture_end, parent_highlight_id)) = highlights.stack.last()
5781 {
5782 chunk_end = chunk_end.min(*parent_capture_end);
5783 highlight_id = Some(*parent_highlight_id);
5784 }
5785 let bit_start = chunk_start - self.chunks.offset();
5786 let bit_end = chunk_end - self.chunks.offset();
5787
5788 let slice = &chunk[bit_start..bit_end];
5789
5790 let mask = 1u128.unbounded_shl(bit_end as u32).wrapping_sub(1);
5791 let tabs = (tabs >> bit_start) & mask;
5792 let chars = (chars_map >> bit_start) & mask;
5793 let newlines = (newlines >> bit_start) & mask;
5794
5795 self.range.start = chunk_end;
5796 if self.range.start == self.chunks.offset() + chunk.len() {
5797 self.chunks.next().unwrap();
5798 }
5799
5800 Some(Chunk {
5801 text: slice,
5802 syntax_highlight_id: highlight_id,
5803 underline: self.underline,
5804 diagnostic_severity: self.current_diagnostic_severity(),
5805 is_unnecessary: self.current_code_is_unnecessary(),
5806 tabs,
5807 chars,
5808 newlines,
5809 ..Chunk::default()
5810 })
5811 } else {
5812 None
5813 }
5814 }
5815}
5816
5817impl operation_queue::Operation for Operation {
5818 fn lamport_timestamp(&self) -> clock::Lamport {
5819 match self {
5820 Operation::Buffer(_) => {
5821 unreachable!("buffer operations should never be deferred at this layer")
5822 }
5823 Operation::UpdateDiagnostics {
5824 lamport_timestamp, ..
5825 }
5826 | Operation::UpdateSelections {
5827 lamport_timestamp, ..
5828 }
5829 | Operation::UpdateCompletionTriggers {
5830 lamport_timestamp, ..
5831 }
5832 | Operation::UpdateLineEnding {
5833 lamport_timestamp, ..
5834 } => *lamport_timestamp,
5835 }
5836 }
5837}
5838
5839impl Default for Diagnostic {
5840 fn default() -> Self {
5841 Self {
5842 source: Default::default(),
5843 source_kind: DiagnosticSourceKind::Other,
5844 code: None,
5845 code_description: None,
5846 severity: DiagnosticSeverity::ERROR,
5847 message: Default::default(),
5848 markdown: None,
5849 group_id: 0,
5850 is_primary: false,
5851 is_disk_based: false,
5852 is_unnecessary: false,
5853 underline: true,
5854 data: None,
5855 registration_id: None,
5856 }
5857 }
5858}
5859
5860impl IndentSize {
5861 /// Returns an [`IndentSize`] representing the given spaces.
5862 pub fn spaces(len: u32) -> Self {
5863 Self {
5864 len,
5865 kind: IndentKind::Space,
5866 }
5867 }
5868
5869 /// Returns an [`IndentSize`] representing a tab.
5870 pub fn tab() -> Self {
5871 Self {
5872 len: 1,
5873 kind: IndentKind::Tab,
5874 }
5875 }
5876
5877 /// An iterator over the characters represented by this [`IndentSize`].
5878 pub fn chars(&self) -> impl Iterator<Item = char> {
5879 iter::repeat(self.char()).take(self.len as usize)
5880 }
5881
5882 /// The character representation of this [`IndentSize`].
5883 pub fn char(&self) -> char {
5884 match self.kind {
5885 IndentKind::Space => ' ',
5886 IndentKind::Tab => '\t',
5887 }
5888 }
5889
5890 /// Consumes the current [`IndentSize`] and returns a new one that has
5891 /// been shrunk or enlarged by the given size along the given direction.
5892 pub fn with_delta(mut self, direction: Ordering, size: IndentSize) -> Self {
5893 match direction {
5894 Ordering::Less => {
5895 if self.kind == size.kind && self.len >= size.len {
5896 self.len -= size.len;
5897 }
5898 }
5899 Ordering::Equal => {}
5900 Ordering::Greater => {
5901 if self.len == 0 {
5902 self = size;
5903 } else if self.kind == size.kind {
5904 self.len += size.len;
5905 }
5906 }
5907 }
5908 self
5909 }
5910
5911 pub fn len_with_expanded_tabs(&self, tab_size: NonZeroU32) -> usize {
5912 match self.kind {
5913 IndentKind::Space => self.len as usize,
5914 IndentKind::Tab => self.len as usize * tab_size.get() as usize,
5915 }
5916 }
5917}
5918
5919#[cfg(any(test, feature = "test-support"))]
5920pub struct TestFile {
5921 pub path: Arc<RelPath>,
5922 pub root_name: String,
5923 pub local_root: Option<PathBuf>,
5924}
5925
5926#[cfg(any(test, feature = "test-support"))]
5927impl File for TestFile {
5928 fn path(&self) -> &Arc<RelPath> {
5929 &self.path
5930 }
5931
5932 fn full_path(&self, _: &gpui::App) -> PathBuf {
5933 PathBuf::from(self.root_name.clone()).join(self.path.as_std_path())
5934 }
5935
5936 fn as_local(&self) -> Option<&dyn LocalFile> {
5937 if self.local_root.is_some() {
5938 Some(self)
5939 } else {
5940 None
5941 }
5942 }
5943
5944 fn disk_state(&self) -> DiskState {
5945 unimplemented!()
5946 }
5947
5948 fn file_name<'a>(&'a self, _: &'a gpui::App) -> &'a str {
5949 self.path().file_name().unwrap_or(self.root_name.as_ref())
5950 }
5951
5952 fn worktree_id(&self, _: &App) -> WorktreeId {
5953 WorktreeId::from_usize(0)
5954 }
5955
5956 fn to_proto(&self, _: &App) -> rpc::proto::File {
5957 unimplemented!()
5958 }
5959
5960 fn is_private(&self) -> bool {
5961 false
5962 }
5963
5964 fn path_style(&self, _cx: &App) -> PathStyle {
5965 PathStyle::local()
5966 }
5967}
5968
5969#[cfg(any(test, feature = "test-support"))]
5970impl LocalFile for TestFile {
5971 fn abs_path(&self, _cx: &App) -> PathBuf {
5972 PathBuf::from(self.local_root.as_ref().unwrap())
5973 .join(&self.root_name)
5974 .join(self.path.as_std_path())
5975 }
5976
5977 fn load(&self, _cx: &App) -> Task<Result<String>> {
5978 unimplemented!()
5979 }
5980
5981 fn load_bytes(&self, _cx: &App) -> Task<Result<Vec<u8>>> {
5982 unimplemented!()
5983 }
5984}
5985
5986pub(crate) fn contiguous_ranges(
5987 values: impl Iterator<Item = u32>,
5988 max_len: usize,
5989) -> impl Iterator<Item = Range<u32>> {
5990 let mut values = values;
5991 let mut current_range: Option<Range<u32>> = None;
5992 std::iter::from_fn(move || {
5993 loop {
5994 if let Some(value) = values.next() {
5995 if let Some(range) = &mut current_range
5996 && value == range.end
5997 && range.len() < max_len
5998 {
5999 range.end += 1;
6000 continue;
6001 }
6002
6003 let prev_range = current_range.clone();
6004 current_range = Some(value..(value + 1));
6005 if prev_range.is_some() {
6006 return prev_range;
6007 }
6008 } else {
6009 return current_range.take();
6010 }
6011 }
6012 })
6013}
6014
6015#[derive(Default, Debug)]
6016pub struct CharClassifier {
6017 scope: Option<LanguageScope>,
6018 scope_context: Option<CharScopeContext>,
6019 ignore_punctuation: bool,
6020}
6021
6022impl CharClassifier {
6023 pub fn new(scope: Option<LanguageScope>) -> Self {
6024 Self {
6025 scope,
6026 scope_context: None,
6027 ignore_punctuation: false,
6028 }
6029 }
6030
6031 pub fn scope_context(self, scope_context: Option<CharScopeContext>) -> Self {
6032 Self {
6033 scope_context,
6034 ..self
6035 }
6036 }
6037
6038 pub fn ignore_punctuation(self, ignore_punctuation: bool) -> Self {
6039 Self {
6040 ignore_punctuation,
6041 ..self
6042 }
6043 }
6044
6045 pub fn is_whitespace(&self, c: char) -> bool {
6046 self.kind(c) == CharKind::Whitespace
6047 }
6048
6049 pub fn is_word(&self, c: char) -> bool {
6050 self.kind(c) == CharKind::Word
6051 }
6052
6053 pub fn is_punctuation(&self, c: char) -> bool {
6054 self.kind(c) == CharKind::Punctuation
6055 }
6056
6057 pub fn kind_with(&self, c: char, ignore_punctuation: bool) -> CharKind {
6058 if c.is_alphanumeric() || c == '_' {
6059 return CharKind::Word;
6060 }
6061
6062 if let Some(scope) = &self.scope {
6063 let characters = match self.scope_context {
6064 Some(CharScopeContext::Completion) => scope.completion_query_characters(),
6065 Some(CharScopeContext::LinkedEdit) => scope.linked_edit_characters(),
6066 None => scope.word_characters(),
6067 };
6068 if let Some(characters) = characters
6069 && characters.contains(&c)
6070 {
6071 return CharKind::Word;
6072 }
6073 }
6074
6075 if c.is_whitespace() {
6076 return CharKind::Whitespace;
6077 }
6078
6079 if ignore_punctuation {
6080 CharKind::Word
6081 } else {
6082 CharKind::Punctuation
6083 }
6084 }
6085
6086 pub fn kind(&self, c: char) -> CharKind {
6087 self.kind_with(c, self.ignore_punctuation)
6088 }
6089}
6090
6091/// Find all of the ranges of whitespace that occur at the ends of lines
6092/// in the given rope.
6093///
6094/// This could also be done with a regex search, but this implementation
6095/// avoids copying text.
6096pub fn trailing_whitespace_ranges(rope: &Rope) -> Vec<Range<usize>> {
6097 let mut ranges = Vec::new();
6098
6099 let mut offset = 0;
6100 let mut prev_chunk_trailing_whitespace_range = 0..0;
6101 for chunk in rope.chunks() {
6102 let mut prev_line_trailing_whitespace_range = 0..0;
6103 for (i, line) in chunk.split('\n').enumerate() {
6104 let line_end_offset = offset + line.len();
6105 let trimmed_line_len = line.trim_end_matches([' ', '\t']).len();
6106 let mut trailing_whitespace_range = (offset + trimmed_line_len)..line_end_offset;
6107
6108 if i == 0 && trimmed_line_len == 0 {
6109 trailing_whitespace_range.start = prev_chunk_trailing_whitespace_range.start;
6110 }
6111 if !prev_line_trailing_whitespace_range.is_empty() {
6112 ranges.push(prev_line_trailing_whitespace_range);
6113 }
6114
6115 offset = line_end_offset + 1;
6116 prev_line_trailing_whitespace_range = trailing_whitespace_range;
6117 }
6118
6119 offset -= 1;
6120 prev_chunk_trailing_whitespace_range = prev_line_trailing_whitespace_range;
6121 }
6122
6123 if !prev_chunk_trailing_whitespace_range.is_empty() {
6124 ranges.push(prev_chunk_trailing_whitespace_range);
6125 }
6126
6127 ranges
6128}