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