1//! The `language` crate provides a large chunk of Zed's language-related
2//! features (the other big contributors being project and lsp crates that revolve around LSP features).
3//! Namely, this crate:
4//! - Provides [`Language`], [`Grammar`] and [`LanguageRegistry`] types that
5//! use Tree-sitter to provide syntax highlighting to the editor; note though that `language` doesn't perform the highlighting by itself. It only maps ranges in a buffer to colors. Treesitter is also used for buffer outlines (lists of symbols in a buffer)
6//! - Exposes [`LanguageConfig`] that describes how constructs (like brackets or line comments) should be handled by the editor for a source file of a particular language.
7//!
8//! Notably we do *not* assign a single language to a single file; in real world a single file can consist of multiple programming languages - HTML is a good example of that - and `language` crate tends to reflect that status quo in it's API.
9mod buffer;
10mod diagnostic_set;
11mod highlight_map;
12pub mod language_settings;
13mod outline;
14pub mod proto;
15mod syntax_map;
16
17#[cfg(test)]
18mod buffer_tests;
19pub mod markdown;
20
21use anyhow::{anyhow, Context, Result};
22use async_trait::async_trait;
23use collections::{HashMap, HashSet};
24use futures::{
25 channel::{mpsc, oneshot},
26 future::Shared,
27 FutureExt, TryFutureExt as _,
28};
29use gpui::{AppContext, AsyncAppContext, BackgroundExecutor, Task};
30pub use highlight_map::HighlightMap;
31use lazy_static::lazy_static;
32use lsp::{CodeActionKind, LanguageServerBinary};
33use parking_lot::{Mutex, RwLock};
34use postage::watch;
35use regex::Regex;
36use serde::{de, Deserialize, Deserializer};
37use serde_json::Value;
38use std::{
39 any::Any,
40 borrow::Cow,
41 cell::RefCell,
42 fmt::Debug,
43 hash::Hash,
44 mem,
45 ops::{Not, Range},
46 path::{Path, PathBuf},
47 str,
48 sync::{
49 atomic::{AtomicU64, AtomicUsize, Ordering::SeqCst},
50 Arc,
51 },
52};
53use syntax_map::SyntaxSnapshot;
54use theme::{SyntaxTheme, Theme};
55use tree_sitter::{self, wasmtime, Query, WasmStore};
56use unicase::UniCase;
57use util::{http::HttpClient, paths::PathExt};
58use util::{post_inc, ResultExt, TryFutureExt as _, UnwrapFuture};
59
60pub use buffer::Operation;
61pub use buffer::*;
62pub use diagnostic_set::DiagnosticEntry;
63pub use lsp::LanguageServerId;
64pub use outline::{Outline, OutlineItem};
65pub use syntax_map::{OwnedSyntaxLayer, SyntaxLayer};
66pub use text::LineEnding;
67pub use tree_sitter::{Parser, Tree};
68
69/// Initializes the `language` crate.
70///
71/// This should be called before making use of items from the create.
72pub fn init(cx: &mut AppContext) {
73 language_settings::init(cx);
74}
75
76#[derive(Clone, Default)]
77struct LspBinaryStatusSender {
78 txs: Arc<Mutex<Vec<mpsc::UnboundedSender<(Arc<Language>, LanguageServerBinaryStatus)>>>>,
79}
80
81impl LspBinaryStatusSender {
82 fn subscribe(&self) -> mpsc::UnboundedReceiver<(Arc<Language>, LanguageServerBinaryStatus)> {
83 let (tx, rx) = mpsc::unbounded();
84 self.txs.lock().push(tx);
85 rx
86 }
87
88 fn send(&self, language: Arc<Language>, status: LanguageServerBinaryStatus) {
89 let mut txs = self.txs.lock();
90 txs.retain(|tx| {
91 tx.unbounded_send((language.clone(), status.clone()))
92 .is_ok()
93 });
94 }
95}
96
97thread_local! {
98 static PARSER: RefCell<Parser> = {
99 let mut parser = Parser::new();
100 parser.set_wasm_store(WasmStore::new(WASM_ENGINE.clone()).unwrap()).unwrap();
101 RefCell::new(parser)
102 };
103}
104
105lazy_static! {
106 pub(crate) static ref NEXT_GRAMMAR_ID: AtomicUsize = Default::default();
107 /// A shared grammar for plain text, exposed for reuse by downstream crates.
108 #[doc(hidden)]
109 pub static ref WASM_ENGINE: wasmtime::Engine = wasmtime::Engine::default();
110 pub static ref PLAIN_TEXT: Arc<Language> = Arc::new(Language::new(
111 LanguageConfig {
112 name: "Plain Text".into(),
113 ..Default::default()
114 },
115 None,
116 ));
117}
118
119/// Types that represent a position in a buffer, and can be converted into
120/// an LSP position, to send to a language server.
121pub trait ToLspPosition {
122 /// Converts the value into an LSP position.
123 fn to_lsp_position(self) -> lsp::Position;
124}
125
126/// A name of a language server.
127#[derive(Clone, Debug, PartialEq, Eq, Hash)]
128pub struct LanguageServerName(pub Arc<str>);
129
130/// Represents a Language Server, with certain cached sync properties.
131/// Uses [`LspAdapter`] under the hood, but calls all 'static' methods
132/// once at startup, and caches the results.
133pub struct CachedLspAdapter {
134 pub name: LanguageServerName,
135 pub short_name: &'static str,
136 pub disk_based_diagnostic_sources: Vec<String>,
137 pub disk_based_diagnostics_progress_token: Option<String>,
138 pub language_ids: HashMap<String, String>,
139 pub adapter: Arc<dyn LspAdapter>,
140 pub reinstall_attempt_count: AtomicU64,
141}
142
143impl CachedLspAdapter {
144 pub async fn new(adapter: Arc<dyn LspAdapter>) -> Arc<Self> {
145 let name = adapter.name();
146 let short_name = adapter.short_name();
147 let disk_based_diagnostic_sources = adapter.disk_based_diagnostic_sources();
148 let disk_based_diagnostics_progress_token = adapter.disk_based_diagnostics_progress_token();
149 let language_ids = adapter.language_ids();
150
151 Arc::new(CachedLspAdapter {
152 name,
153 short_name,
154 disk_based_diagnostic_sources,
155 disk_based_diagnostics_progress_token,
156 language_ids,
157 adapter,
158 reinstall_attempt_count: AtomicU64::new(0),
159 })
160 }
161
162 pub async fn fetch_latest_server_version(
163 &self,
164 delegate: &dyn LspAdapterDelegate,
165 ) -> Result<Box<dyn 'static + Send + Any>> {
166 self.adapter.fetch_latest_server_version(delegate).await
167 }
168
169 pub fn will_fetch_server(
170 &self,
171 delegate: &Arc<dyn LspAdapterDelegate>,
172 cx: &mut AsyncAppContext,
173 ) -> Option<Task<Result<()>>> {
174 self.adapter.will_fetch_server(delegate, cx)
175 }
176
177 pub fn will_start_server(
178 &self,
179 delegate: &Arc<dyn LspAdapterDelegate>,
180 cx: &mut AsyncAppContext,
181 ) -> Option<Task<Result<()>>> {
182 self.adapter.will_start_server(delegate, cx)
183 }
184
185 pub async fn fetch_server_binary(
186 &self,
187 version: Box<dyn 'static + Send + Any>,
188 container_dir: PathBuf,
189 delegate: &dyn LspAdapterDelegate,
190 ) -> Result<LanguageServerBinary> {
191 self.adapter
192 .fetch_server_binary(version, container_dir, delegate)
193 .await
194 }
195
196 pub async fn cached_server_binary(
197 &self,
198 container_dir: PathBuf,
199 delegate: &dyn LspAdapterDelegate,
200 ) -> Option<LanguageServerBinary> {
201 self.adapter
202 .cached_server_binary(container_dir, delegate)
203 .await
204 }
205
206 pub fn can_be_reinstalled(&self) -> bool {
207 self.adapter.can_be_reinstalled()
208 }
209
210 pub async fn installation_test_binary(
211 &self,
212 container_dir: PathBuf,
213 ) -> Option<LanguageServerBinary> {
214 self.adapter.installation_test_binary(container_dir).await
215 }
216
217 pub fn code_action_kinds(&self) -> Option<Vec<CodeActionKind>> {
218 self.adapter.code_action_kinds()
219 }
220
221 pub fn workspace_configuration(&self, workspace_root: &Path, cx: &mut AppContext) -> Value {
222 self.adapter.workspace_configuration(workspace_root, cx)
223 }
224
225 pub fn process_diagnostics(&self, params: &mut lsp::PublishDiagnosticsParams) {
226 self.adapter.process_diagnostics(params)
227 }
228
229 pub async fn process_completion(&self, completion_item: &mut lsp::CompletionItem) {
230 self.adapter.process_completion(completion_item).await
231 }
232
233 pub async fn label_for_completion(
234 &self,
235 completion_item: &lsp::CompletionItem,
236 language: &Arc<Language>,
237 ) -> Option<CodeLabel> {
238 self.adapter
239 .label_for_completion(completion_item, language)
240 .await
241 }
242
243 pub async fn label_for_symbol(
244 &self,
245 name: &str,
246 kind: lsp::SymbolKind,
247 language: &Arc<Language>,
248 ) -> Option<CodeLabel> {
249 self.adapter.label_for_symbol(name, kind, language).await
250 }
251
252 pub fn prettier_plugins(&self) -> &[&'static str] {
253 self.adapter.prettier_plugins()
254 }
255}
256
257/// [`LspAdapterDelegate`] allows [`LspAdapter]` implementations to interface with the application
258// e.g. to display a notification or fetch data from the web.
259pub trait LspAdapterDelegate: Send + Sync {
260 fn show_notification(&self, message: &str, cx: &mut AppContext);
261 fn http_client(&self) -> Arc<dyn HttpClient>;
262}
263
264#[async_trait]
265pub trait LspAdapter: 'static + Send + Sync {
266 fn name(&self) -> LanguageServerName;
267
268 fn short_name(&self) -> &'static str;
269
270 async fn fetch_latest_server_version(
271 &self,
272 delegate: &dyn LspAdapterDelegate,
273 ) -> Result<Box<dyn 'static + Send + Any>>;
274
275 fn will_fetch_server(
276 &self,
277 _: &Arc<dyn LspAdapterDelegate>,
278 _: &mut AsyncAppContext,
279 ) -> Option<Task<Result<()>>> {
280 None
281 }
282
283 fn will_start_server(
284 &self,
285 _: &Arc<dyn LspAdapterDelegate>,
286 _: &mut AsyncAppContext,
287 ) -> Option<Task<Result<()>>> {
288 None
289 }
290
291 async fn fetch_server_binary(
292 &self,
293 version: Box<dyn 'static + Send + Any>,
294 container_dir: PathBuf,
295 delegate: &dyn LspAdapterDelegate,
296 ) -> Result<LanguageServerBinary>;
297
298 async fn cached_server_binary(
299 &self,
300 container_dir: PathBuf,
301 delegate: &dyn LspAdapterDelegate,
302 ) -> Option<LanguageServerBinary>;
303
304 /// Returns `true` if a language server can be reinstalled.
305 ///
306 /// If language server initialization fails, a reinstallation will be attempted unless the value returned from this method is `false`.
307 ///
308 /// Implementations that rely on software already installed on user's system
309 /// should have [`can_be_reinstalled`](Self::can_be_reinstalled) return `false`.
310 fn can_be_reinstalled(&self) -> bool {
311 true
312 }
313
314 async fn installation_test_binary(
315 &self,
316 container_dir: PathBuf,
317 ) -> Option<LanguageServerBinary>;
318
319 fn process_diagnostics(&self, _: &mut lsp::PublishDiagnosticsParams) {}
320
321 /// A callback called for each [`lsp::CompletionItem`] obtained from LSP server.
322 /// Some LspAdapter implementations might want to modify the obtained item to
323 /// change how it's displayed.
324 async fn process_completion(&self, _: &mut lsp::CompletionItem) {}
325
326 async fn label_for_completion(
327 &self,
328 _: &lsp::CompletionItem,
329 _: &Arc<Language>,
330 ) -> Option<CodeLabel> {
331 None
332 }
333
334 async fn label_for_symbol(
335 &self,
336 _: &str,
337 _: lsp::SymbolKind,
338 _: &Arc<Language>,
339 ) -> Option<CodeLabel> {
340 None
341 }
342
343 /// Returns initialization options that are going to be sent to a LSP server as a part of [`lsp::InitializeParams`]
344 fn initialization_options(&self) -> Option<Value> {
345 None
346 }
347
348 fn workspace_configuration(&self, _workspace_root: &Path, _cx: &mut AppContext) -> Value {
349 serde_json::json!({})
350 }
351
352 /// Returns a list of code actions supported by a given LspAdapter
353 fn code_action_kinds(&self) -> Option<Vec<CodeActionKind>> {
354 Some(vec![
355 CodeActionKind::EMPTY,
356 CodeActionKind::QUICKFIX,
357 CodeActionKind::REFACTOR,
358 CodeActionKind::REFACTOR_EXTRACT,
359 CodeActionKind::SOURCE,
360 ])
361 }
362
363 fn disk_based_diagnostic_sources(&self) -> Vec<String> {
364 Default::default()
365 }
366
367 fn disk_based_diagnostics_progress_token(&self) -> Option<String> {
368 None
369 }
370
371 fn language_ids(&self) -> HashMap<String, String> {
372 Default::default()
373 }
374
375 fn prettier_plugins(&self) -> &[&'static str] {
376 &[]
377 }
378}
379
380#[derive(Clone, Debug, PartialEq, Eq)]
381pub struct CodeLabel {
382 /// The text to display.
383 pub text: String,
384 /// Syntax highlighting runs.
385 pub runs: Vec<(Range<usize>, HighlightId)>,
386 /// The portion of the text that should be used in fuzzy filtering.
387 pub filter_range: Range<usize>,
388}
389
390#[derive(Clone, Deserialize)]
391pub struct LanguageConfig {
392 /// Human-readable name of the language.
393 pub name: Arc<str>,
394 // The name of the grammar in a WASM bundle (experimental).
395 pub grammar_name: Option<Arc<str>>,
396 /// Given a list of `LanguageConfig`'s, the language of a file can be determined based on the path extension matching any of the `path_suffixes`.
397 pub path_suffixes: Vec<String>,
398 /// List of bracket types in a language.
399 pub brackets: BracketPairConfig,
400 /// A regex pattern that determines whether the language should be assigned to a file or not.
401 #[serde(default, deserialize_with = "deserialize_regex")]
402 pub first_line_pattern: Option<Regex>,
403 /// If set to true, auto indentation uses last non empty line to determine
404 /// the indentation level for a new line.
405 #[serde(default = "auto_indent_using_last_non_empty_line_default")]
406 pub auto_indent_using_last_non_empty_line: bool,
407 /// A regex that is used to determine whether the indentation level should be
408 /// increased in the following line.
409 #[serde(default, deserialize_with = "deserialize_regex")]
410 pub increase_indent_pattern: Option<Regex>,
411 /// A regex that is used to determine whether the indentation level should be
412 /// decreased in the following line.
413 #[serde(default, deserialize_with = "deserialize_regex")]
414 pub decrease_indent_pattern: Option<Regex>,
415 /// A list of characters that trigger the automatic insertion of a closing
416 /// bracket when they immediately precede the point where an opening
417 /// bracket is inserted.
418 #[serde(default)]
419 pub autoclose_before: String,
420 /// A placeholder used internally by Semantic Index.
421 #[serde(default)]
422 pub collapsed_placeholder: String,
423 /// A line comment string that is inserted in e.g. `toggle comments` action.
424 /// A language can have multiple flavours of line comments. All of the provided line comments are
425 /// used for comment continuations on the next line, but only the first one is used for Editor::ToggleComments.
426 #[serde(default)]
427 pub line_comments: Vec<Arc<str>>,
428 /// Starting and closing characters of a block comment.
429 #[serde(default)]
430 pub block_comment: Option<(Arc<str>, Arc<str>)>,
431 /// A list of language servers that are allowed to run on subranges of a given language.
432 #[serde(default)]
433 pub scope_opt_in_language_servers: Vec<String>,
434 #[serde(default)]
435 pub overrides: HashMap<String, LanguageConfigOverride>,
436 /// A list of characters that Zed should treat as word characters for the
437 /// purpose of features that operate on word boundaries, like 'move to next word end'
438 /// or a whole-word search in buffer search.
439 #[serde(default)]
440 pub word_characters: HashSet<char>,
441 /// The name of a Prettier parser that should be used for this language.
442 #[serde(default)]
443 pub prettier_parser_name: Option<String>,
444}
445
446/// Tree-sitter language queries for a given language.
447#[derive(Debug, Default)]
448pub struct LanguageQueries {
449 pub highlights: Option<Cow<'static, str>>,
450 pub brackets: Option<Cow<'static, str>>,
451 pub indents: Option<Cow<'static, str>>,
452 pub outline: Option<Cow<'static, str>>,
453 pub embedding: Option<Cow<'static, str>>,
454 pub injections: Option<Cow<'static, str>>,
455 pub overrides: Option<Cow<'static, str>>,
456}
457
458/// Represents a language for the given range. Some languages (e.g. HTML)
459/// interleave several languages together, thus a single buffer might actually contain
460/// several nested scopes.
461#[derive(Clone, Debug)]
462pub struct LanguageScope {
463 language: Arc<Language>,
464 override_id: Option<u32>,
465}
466
467#[derive(Clone, Deserialize, Default, Debug)]
468pub struct LanguageConfigOverride {
469 #[serde(default)]
470 pub line_comments: Override<Vec<Arc<str>>>,
471 #[serde(default)]
472 pub block_comment: Override<(Arc<str>, Arc<str>)>,
473 #[serde(skip_deserializing)]
474 pub disabled_bracket_ixs: Vec<u16>,
475 #[serde(default)]
476 pub word_characters: Override<HashSet<char>>,
477 #[serde(default)]
478 pub opt_into_language_servers: Vec<String>,
479}
480
481#[derive(Clone, Deserialize, Debug)]
482#[serde(untagged)]
483pub enum Override<T> {
484 Remove { remove: bool },
485 Set(T),
486}
487
488impl<T> Default for Override<T> {
489 fn default() -> Self {
490 Override::Remove { remove: false }
491 }
492}
493
494impl<T> Override<T> {
495 fn as_option<'a>(this: Option<&'a Self>, original: Option<&'a T>) -> Option<&'a T> {
496 match this {
497 Some(Self::Set(value)) => Some(value),
498 Some(Self::Remove { remove: true }) => None,
499 Some(Self::Remove { remove: false }) | None => original,
500 }
501 }
502}
503
504impl Default for LanguageConfig {
505 fn default() -> Self {
506 Self {
507 name: "".into(),
508 grammar_name: None,
509 path_suffixes: Default::default(),
510 brackets: Default::default(),
511 auto_indent_using_last_non_empty_line: auto_indent_using_last_non_empty_line_default(),
512 first_line_pattern: Default::default(),
513 increase_indent_pattern: Default::default(),
514 decrease_indent_pattern: Default::default(),
515 autoclose_before: Default::default(),
516 line_comments: Default::default(),
517 block_comment: Default::default(),
518 scope_opt_in_language_servers: Default::default(),
519 overrides: Default::default(),
520 word_characters: Default::default(),
521 prettier_parser_name: None,
522 collapsed_placeholder: Default::default(),
523 }
524 }
525}
526
527fn auto_indent_using_last_non_empty_line_default() -> bool {
528 true
529}
530
531fn deserialize_regex<'de, D: Deserializer<'de>>(d: D) -> Result<Option<Regex>, D::Error> {
532 let source = Option::<String>::deserialize(d)?;
533 if let Some(source) = source {
534 Ok(Some(regex::Regex::new(&source).map_err(de::Error::custom)?))
535 } else {
536 Ok(None)
537 }
538}
539
540#[doc(hidden)]
541#[cfg(any(test, feature = "test-support"))]
542pub struct FakeLspAdapter {
543 pub name: &'static str,
544 pub initialization_options: Option<Value>,
545 pub capabilities: lsp::ServerCapabilities,
546 pub initializer: Option<Box<dyn 'static + Send + Sync + Fn(&mut lsp::FakeLanguageServer)>>,
547 pub disk_based_diagnostics_progress_token: Option<String>,
548 pub disk_based_diagnostics_sources: Vec<String>,
549 pub prettier_plugins: Vec<&'static str>,
550}
551
552/// Configuration of handling bracket pairs for a given language.
553///
554/// This struct includes settings for defining which pairs of characters are considered brackets and
555/// also specifies any language-specific scopes where these pairs should be ignored for bracket matching purposes.
556#[derive(Clone, Debug, Default)]
557pub struct BracketPairConfig {
558 /// A list of character pairs that should be treated as brackets in the context of a given language.
559 pub pairs: Vec<BracketPair>,
560 /// A list of tree-sitter scopes for which a given bracket should not be active.
561 /// N-th entry in `[Self::disabled_scopes_by_bracket_ix]` contains a list of disabled scopes for an n-th entry in `[Self::pairs]`
562 pub disabled_scopes_by_bracket_ix: Vec<Vec<String>>,
563}
564
565impl<'de> Deserialize<'de> for BracketPairConfig {
566 fn deserialize<D>(deserializer: D) -> std::result::Result<Self, D::Error>
567 where
568 D: Deserializer<'de>,
569 {
570 #[derive(Deserialize)]
571 pub struct Entry {
572 #[serde(flatten)]
573 pub bracket_pair: BracketPair,
574 #[serde(default)]
575 pub not_in: Vec<String>,
576 }
577
578 let result = Vec::<Entry>::deserialize(deserializer)?;
579 let mut brackets = Vec::with_capacity(result.len());
580 let mut disabled_scopes_by_bracket_ix = Vec::with_capacity(result.len());
581 for entry in result {
582 brackets.push(entry.bracket_pair);
583 disabled_scopes_by_bracket_ix.push(entry.not_in);
584 }
585
586 Ok(BracketPairConfig {
587 pairs: brackets,
588 disabled_scopes_by_bracket_ix,
589 })
590 }
591}
592
593/// Describes a single bracket pair and how an editor should react to e.g. inserting
594/// an opening bracket or to a newline character insertion in between `start` and `end` characters.
595#[derive(Clone, Debug, Default, Deserialize, PartialEq)]
596pub struct BracketPair {
597 /// Starting substring for a bracket.
598 pub start: String,
599 /// Ending substring for a bracket.
600 pub end: String,
601 /// True if `end` should be automatically inserted right after `start` characters.
602 pub close: bool,
603 /// True if an extra newline should be inserted while the cursor is in the middle
604 /// of that bracket pair.
605 pub newline: bool,
606}
607
608pub struct Language {
609 pub(crate) config: LanguageConfig,
610 pub(crate) grammar: Option<Arc<Grammar>>,
611 pub(crate) adapters: Vec<Arc<CachedLspAdapter>>,
612
613 #[cfg(any(test, feature = "test-support"))]
614 fake_adapter: Option<(
615 mpsc::UnboundedSender<lsp::FakeLanguageServer>,
616 Arc<FakeLspAdapter>,
617 )>,
618}
619
620pub struct Grammar {
621 id: usize,
622 pub ts_language: tree_sitter::Language,
623 pub(crate) error_query: Query,
624 pub(crate) highlights_query: Option<Query>,
625 pub(crate) brackets_config: Option<BracketConfig>,
626 pub(crate) indents_config: Option<IndentConfig>,
627 pub outline_config: Option<OutlineConfig>,
628 pub embedding_config: Option<EmbeddingConfig>,
629 pub(crate) injection_config: Option<InjectionConfig>,
630 pub(crate) override_config: Option<OverrideConfig>,
631 pub(crate) highlight_map: Mutex<HighlightMap>,
632}
633
634struct IndentConfig {
635 query: Query,
636 indent_capture_ix: u32,
637 start_capture_ix: Option<u32>,
638 end_capture_ix: Option<u32>,
639 outdent_capture_ix: Option<u32>,
640}
641
642pub struct OutlineConfig {
643 pub query: Query,
644 pub item_capture_ix: u32,
645 pub name_capture_ix: u32,
646 pub context_capture_ix: Option<u32>,
647 pub extra_context_capture_ix: Option<u32>,
648}
649
650#[derive(Debug)]
651pub struct EmbeddingConfig {
652 pub query: Query,
653 pub item_capture_ix: u32,
654 pub name_capture_ix: Option<u32>,
655 pub context_capture_ix: Option<u32>,
656 pub collapse_capture_ix: Option<u32>,
657 pub keep_capture_ix: Option<u32>,
658}
659
660struct InjectionConfig {
661 query: Query,
662 content_capture_ix: u32,
663 language_capture_ix: Option<u32>,
664 patterns: Vec<InjectionPatternConfig>,
665}
666
667struct OverrideConfig {
668 query: Query,
669 values: HashMap<u32, (String, LanguageConfigOverride)>,
670}
671
672#[derive(Default, Clone)]
673struct InjectionPatternConfig {
674 language: Option<Box<str>>,
675 combined: bool,
676}
677
678struct BracketConfig {
679 query: Query,
680 open_capture_ix: u32,
681 close_capture_ix: u32,
682}
683
684#[derive(Clone)]
685pub enum LanguageServerBinaryStatus {
686 CheckingForUpdate,
687 Downloading,
688 Downloaded,
689 Cached,
690 Failed { error: String },
691}
692
693type AvailableLanguageId = usize;
694
695#[derive(Clone)]
696struct AvailableLanguage {
697 id: AvailableLanguageId,
698 config: LanguageConfig,
699 grammar: AvailableGrammar,
700 lsp_adapters: Vec<Arc<dyn LspAdapter>>,
701 loaded: bool,
702}
703
704#[derive(Clone)]
705enum AvailableGrammar {
706 Native {
707 grammar: tree_sitter::Language,
708 asset_dir: &'static str,
709 get_queries: fn(&str) -> LanguageQueries,
710 },
711 Wasm {
712 path: Arc<Path>,
713 get_queries: fn(&Path) -> LanguageQueries,
714 },
715}
716
717pub struct LanguageRegistry {
718 state: RwLock<LanguageRegistryState>,
719 language_server_download_dir: Option<Arc<Path>>,
720 login_shell_env_loaded: Shared<Task<()>>,
721 #[allow(clippy::type_complexity)]
722 lsp_binary_paths: Mutex<
723 HashMap<LanguageServerName, Shared<Task<Result<LanguageServerBinary, Arc<anyhow::Error>>>>>,
724 >,
725 executor: Option<BackgroundExecutor>,
726 lsp_binary_status_tx: LspBinaryStatusSender,
727}
728
729struct LanguageRegistryState {
730 next_language_server_id: usize,
731 languages: Vec<Arc<Language>>,
732 available_languages: Vec<AvailableLanguage>,
733 next_available_language_id: AvailableLanguageId,
734 loading_languages: HashMap<AvailableLanguageId, Vec<oneshot::Sender<Result<Arc<Language>>>>>,
735 subscription: (watch::Sender<()>, watch::Receiver<()>),
736 theme: Option<Arc<Theme>>,
737 version: usize,
738 reload_count: usize,
739}
740
741pub struct PendingLanguageServer {
742 pub server_id: LanguageServerId,
743 pub task: Task<Result<lsp::LanguageServer>>,
744 pub container_dir: Option<Arc<Path>>,
745}
746
747impl LanguageRegistry {
748 pub fn new(login_shell_env_loaded: Task<()>) -> Self {
749 Self {
750 state: RwLock::new(LanguageRegistryState {
751 next_language_server_id: 0,
752 languages: vec![PLAIN_TEXT.clone()],
753 available_languages: Default::default(),
754 next_available_language_id: 0,
755 loading_languages: Default::default(),
756 subscription: watch::channel(),
757 theme: Default::default(),
758 version: 0,
759 reload_count: 0,
760 }),
761 language_server_download_dir: None,
762 login_shell_env_loaded: login_shell_env_loaded.shared(),
763 lsp_binary_paths: Default::default(),
764 executor: None,
765 lsp_binary_status_tx: Default::default(),
766 }
767 }
768
769 #[cfg(any(test, feature = "test-support"))]
770 pub fn test() -> Self {
771 Self::new(Task::ready(()))
772 }
773
774 pub fn set_executor(&mut self, executor: BackgroundExecutor) {
775 self.executor = Some(executor);
776 }
777
778 /// Clear out all of the loaded languages and reload them from scratch.
779 pub fn reload(&self) {
780 self.state.write().reload();
781 }
782
783 pub fn register(
784 &self,
785 asset_dir: &'static str,
786 config: LanguageConfig,
787 grammar: tree_sitter::Language,
788 lsp_adapters: Vec<Arc<dyn LspAdapter>>,
789 get_queries: fn(&str) -> LanguageQueries,
790 ) {
791 let state = &mut *self.state.write();
792 state.available_languages.push(AvailableLanguage {
793 id: post_inc(&mut state.next_available_language_id),
794 config,
795 grammar: AvailableGrammar::Native {
796 grammar,
797 get_queries,
798 asset_dir,
799 },
800 lsp_adapters,
801 loaded: false,
802 });
803 }
804
805 pub fn register_wasm(
806 &self,
807 path: Arc<Path>,
808 config: LanguageConfig,
809 get_queries: fn(&Path) -> LanguageQueries,
810 ) {
811 let state = &mut *self.state.write();
812 state.available_languages.push(AvailableLanguage {
813 id: post_inc(&mut state.next_available_language_id),
814 config,
815 grammar: AvailableGrammar::Wasm { path, get_queries },
816 lsp_adapters: Vec::new(),
817 loaded: false,
818 });
819 }
820
821 pub fn language_names(&self) -> Vec<String> {
822 let state = self.state.read();
823 let mut result = state
824 .available_languages
825 .iter()
826 .filter_map(|l| l.loaded.not().then_some(l.config.name.to_string()))
827 .chain(state.languages.iter().map(|l| l.config.name.to_string()))
828 .collect::<Vec<_>>();
829 result.sort_unstable_by_key(|language_name| language_name.to_lowercase());
830 result
831 }
832
833 pub fn add(&self, language: Arc<Language>) {
834 self.state.write().add(language);
835 }
836
837 pub fn subscribe(&self) -> watch::Receiver<()> {
838 self.state.read().subscription.1.clone()
839 }
840
841 /// The number of times that the registry has been changed,
842 /// by adding languages or reloading.
843 pub fn version(&self) -> usize {
844 self.state.read().version
845 }
846
847 /// The number of times that the registry has been reloaded.
848 pub fn reload_count(&self) -> usize {
849 self.state.read().reload_count
850 }
851
852 pub fn set_theme(&self, theme: Arc<Theme>) {
853 let mut state = self.state.write();
854 state.theme = Some(theme.clone());
855 for language in &state.languages {
856 language.set_theme(theme.syntax());
857 }
858 }
859
860 pub fn set_language_server_download_dir(&mut self, path: impl Into<Arc<Path>>) {
861 self.language_server_download_dir = Some(path.into());
862 }
863
864 pub fn language_for_name(
865 self: &Arc<Self>,
866 name: &str,
867 ) -> UnwrapFuture<oneshot::Receiver<Result<Arc<Language>>>> {
868 let name = UniCase::new(name);
869 self.get_or_load_language(|config| UniCase::new(config.name.as_ref()) == name)
870 }
871
872 pub fn language_for_name_or_extension(
873 self: &Arc<Self>,
874 string: &str,
875 ) -> UnwrapFuture<oneshot::Receiver<Result<Arc<Language>>>> {
876 let string = UniCase::new(string);
877 self.get_or_load_language(|config| {
878 UniCase::new(config.name.as_ref()) == string
879 || config
880 .path_suffixes
881 .iter()
882 .any(|suffix| UniCase::new(suffix) == string)
883 })
884 }
885
886 pub fn language_for_file(
887 self: &Arc<Self>,
888 path: impl AsRef<Path>,
889 content: Option<&Rope>,
890 ) -> UnwrapFuture<oneshot::Receiver<Result<Arc<Language>>>> {
891 let path = path.as_ref();
892 let filename = path.file_name().and_then(|name| name.to_str());
893 let extension = path.extension_or_hidden_file_name();
894 let path_suffixes = [extension, filename];
895 self.get_or_load_language(|config| {
896 let path_matches = config
897 .path_suffixes
898 .iter()
899 .any(|suffix| path_suffixes.contains(&Some(suffix.as_str())));
900 let content_matches = content.zip(config.first_line_pattern.as_ref()).map_or(
901 false,
902 |(content, pattern)| {
903 let end = content.clip_point(Point::new(0, 256), Bias::Left);
904 let end = content.point_to_offset(end);
905 let text = content.chunks_in_range(0..end).collect::<String>();
906 pattern.is_match(&text)
907 },
908 );
909 path_matches || content_matches
910 })
911 }
912
913 fn get_or_load_language(
914 self: &Arc<Self>,
915 callback: impl Fn(&LanguageConfig) -> bool,
916 ) -> UnwrapFuture<oneshot::Receiver<Result<Arc<Language>>>> {
917 let (tx, rx) = oneshot::channel();
918
919 let mut state = self.state.write();
920 if let Some(language) = state
921 .languages
922 .iter()
923 .find(|language| callback(&language.config))
924 {
925 let _ = tx.send(Ok(language.clone()));
926 } else if let Some(executor) = self.executor.clone() {
927 if let Some(language) = state
928 .available_languages
929 .iter()
930 .find(|l| !l.loaded && callback(&l.config))
931 .cloned()
932 {
933 let txs = state
934 .loading_languages
935 .entry(language.id)
936 .or_insert_with(|| {
937 let this = self.clone();
938 executor
939 .spawn(async move {
940 let id = language.id;
941 let name = language.config.name.clone();
942 let language = async {
943 let (grammar, queries) = match language.grammar {
944 AvailableGrammar::Native {
945 grammar,
946 asset_dir,
947 get_queries,
948 } => (grammar, (get_queries)(asset_dir)),
949 AvailableGrammar::Wasm { path, get_queries } => {
950 let grammar_name =
951 &language.config.grammar_name.as_ref().ok_or_else(
952 || anyhow!("missing grammar name"),
953 )?;
954 let mut wasm_path = path.join(grammar_name.as_ref());
955 wasm_path.set_extension("wasm");
956 let wasm_bytes = std::fs::read(&wasm_path)?;
957 let grammar = PARSER.with(|parser| {
958 let mut parser = parser.borrow_mut();
959 let mut store = parser.take_wasm_store().unwrap();
960 let grammar =
961 store.load_language(&grammar_name, &wasm_bytes);
962 parser.set_wasm_store(store).unwrap();
963 grammar
964 })?;
965 (grammar, get_queries(path.as_ref()))
966 }
967 };
968 Language::new(language.config, Some(grammar))
969 .with_lsp_adapters(language.lsp_adapters)
970 .await
971 .with_queries(queries)
972 }
973 .await;
974
975 match language {
976 Ok(language) => {
977 let language = Arc::new(language);
978 let mut state = this.state.write();
979
980 state.add(language.clone());
981 state.mark_language_loaded(id);
982 if let Some(mut txs) = state.loading_languages.remove(&id) {
983 for tx in txs.drain(..) {
984 let _ = tx.send(Ok(language.clone()));
985 }
986 }
987 }
988 Err(e) => {
989 log::error!("failed to load language {name}:\n{:?}", e);
990 let mut state = this.state.write();
991 state.mark_language_loaded(id);
992 if let Some(mut txs) = state.loading_languages.remove(&id) {
993 for tx in txs.drain(..) {
994 let _ = tx.send(Err(anyhow!(
995 "failed to load language {}: {}",
996 name,
997 e
998 )));
999 }
1000 }
1001 }
1002 };
1003 })
1004 .detach();
1005
1006 Vec::new()
1007 });
1008 txs.push(tx);
1009 } else {
1010 let _ = tx.send(Err(anyhow!("language not found")));
1011 }
1012 } else {
1013 let _ = tx.send(Err(anyhow!("executor does not exist")));
1014 }
1015
1016 rx.unwrap()
1017 }
1018
1019 pub fn to_vec(&self) -> Vec<Arc<Language>> {
1020 self.state.read().languages.iter().cloned().collect()
1021 }
1022
1023 pub fn create_pending_language_server(
1024 self: &Arc<Self>,
1025 stderr_capture: Arc<Mutex<Option<String>>>,
1026 language: Arc<Language>,
1027 adapter: Arc<CachedLspAdapter>,
1028 root_path: Arc<Path>,
1029 delegate: Arc<dyn LspAdapterDelegate>,
1030 cx: &mut AppContext,
1031 ) -> Option<PendingLanguageServer> {
1032 let server_id = self.state.write().next_language_server_id();
1033 log::info!(
1034 "starting language server {:?}, path: {root_path:?}, id: {server_id}",
1035 adapter.name.0
1036 );
1037
1038 #[cfg(any(test, feature = "test-support"))]
1039 if language.fake_adapter.is_some() {
1040 let task = cx.spawn(|cx| async move {
1041 let (servers_tx, fake_adapter) = language.fake_adapter.as_ref().unwrap();
1042 let (server, mut fake_server) = lsp::FakeLanguageServer::new(
1043 fake_adapter.name.to_string(),
1044 fake_adapter.capabilities.clone(),
1045 cx.clone(),
1046 );
1047
1048 if let Some(initializer) = &fake_adapter.initializer {
1049 initializer(&mut fake_server);
1050 }
1051
1052 let servers_tx = servers_tx.clone();
1053 cx.background_executor()
1054 .spawn(async move {
1055 if fake_server
1056 .try_receive_notification::<lsp::notification::Initialized>()
1057 .await
1058 .is_some()
1059 {
1060 servers_tx.unbounded_send(fake_server).ok();
1061 }
1062 })
1063 .detach();
1064
1065 Ok(server)
1066 });
1067
1068 return Some(PendingLanguageServer {
1069 server_id,
1070 task,
1071 container_dir: None,
1072 });
1073 }
1074
1075 let download_dir = self
1076 .language_server_download_dir
1077 .clone()
1078 .ok_or_else(|| anyhow!("language server download directory has not been assigned before starting server"))
1079 .log_err()?;
1080 let this = self.clone();
1081 let language = language.clone();
1082 let container_dir: Arc<Path> = Arc::from(download_dir.join(adapter.name.0.as_ref()));
1083 let root_path = root_path.clone();
1084 let adapter = adapter.clone();
1085 let login_shell_env_loaded = self.login_shell_env_loaded.clone();
1086 let lsp_binary_statuses = self.lsp_binary_status_tx.clone();
1087
1088 let task = {
1089 let container_dir = container_dir.clone();
1090 cx.spawn(move |mut cx| async move {
1091 login_shell_env_loaded.await;
1092
1093 let entry = this
1094 .lsp_binary_paths
1095 .lock()
1096 .entry(adapter.name.clone())
1097 .or_insert_with(|| {
1098 let adapter = adapter.clone();
1099 let language = language.clone();
1100 let delegate = delegate.clone();
1101 cx.spawn(|cx| {
1102 get_binary(
1103 adapter,
1104 language,
1105 delegate,
1106 container_dir,
1107 lsp_binary_statuses,
1108 cx,
1109 )
1110 .map_err(Arc::new)
1111 })
1112 .shared()
1113 })
1114 .clone();
1115
1116 let binary = match entry.await {
1117 Ok(binary) => binary,
1118 Err(err) => anyhow::bail!("{err}"),
1119 };
1120
1121 if let Some(task) = adapter.will_start_server(&delegate, &mut cx) {
1122 task.await?;
1123 }
1124
1125 lsp::LanguageServer::new(
1126 stderr_capture,
1127 server_id,
1128 binary,
1129 &root_path,
1130 adapter.code_action_kinds(),
1131 cx,
1132 )
1133 })
1134 };
1135
1136 Some(PendingLanguageServer {
1137 server_id,
1138 task,
1139 container_dir: Some(container_dir),
1140 })
1141 }
1142
1143 pub fn language_server_binary_statuses(
1144 &self,
1145 ) -> mpsc::UnboundedReceiver<(Arc<Language>, LanguageServerBinaryStatus)> {
1146 self.lsp_binary_status_tx.subscribe()
1147 }
1148
1149 pub fn delete_server_container(
1150 &self,
1151 adapter: Arc<CachedLspAdapter>,
1152 cx: &mut AppContext,
1153 ) -> Task<()> {
1154 log::info!("deleting server container");
1155
1156 let mut lock = self.lsp_binary_paths.lock();
1157 lock.remove(&adapter.name);
1158
1159 let download_dir = self
1160 .language_server_download_dir
1161 .clone()
1162 .expect("language server download directory has not been assigned before deleting server container");
1163
1164 cx.spawn(|_| async move {
1165 let container_dir = download_dir.join(adapter.name.0.as_ref());
1166 smol::fs::remove_dir_all(container_dir)
1167 .await
1168 .context("server container removal")
1169 .log_err();
1170 })
1171 }
1172
1173 pub fn next_language_server_id(&self) -> LanguageServerId {
1174 self.state.write().next_language_server_id()
1175 }
1176}
1177
1178impl LanguageRegistryState {
1179 fn next_language_server_id(&mut self) -> LanguageServerId {
1180 LanguageServerId(post_inc(&mut self.next_language_server_id))
1181 }
1182
1183 fn add(&mut self, language: Arc<Language>) {
1184 if let Some(theme) = self.theme.as_ref() {
1185 language.set_theme(theme.syntax());
1186 }
1187 self.languages.push(language);
1188 self.version += 1;
1189 *self.subscription.0.borrow_mut() = ();
1190 }
1191
1192 fn reload(&mut self) {
1193 self.languages.clear();
1194 self.version += 1;
1195 self.reload_count += 1;
1196 for language in &mut self.available_languages {
1197 language.loaded = false;
1198 }
1199 *self.subscription.0.borrow_mut() = ();
1200 }
1201
1202 /// Mark the given language a having been loaded, so that the
1203 /// language registry won't try to load it again.
1204 fn mark_language_loaded(&mut self, id: AvailableLanguageId) {
1205 for language in &mut self.available_languages {
1206 if language.id == id {
1207 language.loaded = true;
1208 break;
1209 }
1210 }
1211 }
1212}
1213
1214#[cfg(any(test, feature = "test-support"))]
1215impl Default for LanguageRegistry {
1216 fn default() -> Self {
1217 Self::test()
1218 }
1219}
1220
1221async fn get_binary(
1222 adapter: Arc<CachedLspAdapter>,
1223 language: Arc<Language>,
1224 delegate: Arc<dyn LspAdapterDelegate>,
1225 container_dir: Arc<Path>,
1226 statuses: LspBinaryStatusSender,
1227 mut cx: AsyncAppContext,
1228) -> Result<LanguageServerBinary> {
1229 if !container_dir.exists() {
1230 smol::fs::create_dir_all(&container_dir)
1231 .await
1232 .context("failed to create container directory")?;
1233 }
1234
1235 if let Some(task) = adapter.will_fetch_server(&delegate, &mut cx) {
1236 task.await?;
1237 }
1238
1239 let binary = fetch_latest_binary(
1240 adapter.clone(),
1241 language.clone(),
1242 delegate.as_ref(),
1243 &container_dir,
1244 statuses.clone(),
1245 )
1246 .await;
1247
1248 if let Err(error) = binary.as_ref() {
1249 if let Some(binary) = adapter
1250 .cached_server_binary(container_dir.to_path_buf(), delegate.as_ref())
1251 .await
1252 {
1253 statuses.send(language.clone(), LanguageServerBinaryStatus::Cached);
1254 return Ok(binary);
1255 } else {
1256 statuses.send(
1257 language.clone(),
1258 LanguageServerBinaryStatus::Failed {
1259 error: format!("{:?}", error),
1260 },
1261 );
1262 }
1263 }
1264
1265 binary
1266}
1267
1268async fn fetch_latest_binary(
1269 adapter: Arc<CachedLspAdapter>,
1270 language: Arc<Language>,
1271 delegate: &dyn LspAdapterDelegate,
1272 container_dir: &Path,
1273 lsp_binary_statuses_tx: LspBinaryStatusSender,
1274) -> Result<LanguageServerBinary> {
1275 let container_dir: Arc<Path> = container_dir.into();
1276 lsp_binary_statuses_tx.send(
1277 language.clone(),
1278 LanguageServerBinaryStatus::CheckingForUpdate,
1279 );
1280
1281 let version_info = adapter.fetch_latest_server_version(delegate).await?;
1282 lsp_binary_statuses_tx.send(language.clone(), LanguageServerBinaryStatus::Downloading);
1283
1284 let binary = adapter
1285 .fetch_server_binary(version_info, container_dir.to_path_buf(), delegate)
1286 .await?;
1287 lsp_binary_statuses_tx.send(language.clone(), LanguageServerBinaryStatus::Downloaded);
1288
1289 Ok(binary)
1290}
1291
1292impl Language {
1293 pub fn new(config: LanguageConfig, ts_language: Option<tree_sitter::Language>) -> Self {
1294 Self {
1295 config,
1296 grammar: ts_language.map(|ts_language| {
1297 Arc::new(Grammar {
1298 id: NEXT_GRAMMAR_ID.fetch_add(1, SeqCst),
1299 highlights_query: None,
1300 brackets_config: None,
1301 outline_config: None,
1302 embedding_config: None,
1303 indents_config: None,
1304 injection_config: None,
1305 override_config: None,
1306 error_query: Query::new(&ts_language, "(ERROR) @error").unwrap(),
1307 ts_language,
1308 highlight_map: Default::default(),
1309 })
1310 }),
1311 adapters: Vec::new(),
1312
1313 #[cfg(any(test, feature = "test-support"))]
1314 fake_adapter: None,
1315 }
1316 }
1317
1318 pub fn lsp_adapters(&self) -> &[Arc<CachedLspAdapter>] {
1319 &self.adapters
1320 }
1321
1322 pub fn id(&self) -> Option<usize> {
1323 self.grammar.as_ref().map(|g| g.id)
1324 }
1325
1326 pub fn with_queries(mut self, queries: LanguageQueries) -> Result<Self> {
1327 if let Some(query) = queries.highlights {
1328 self = self
1329 .with_highlights_query(query.as_ref())
1330 .context("Error loading highlights query")?;
1331 }
1332 if let Some(query) = queries.brackets {
1333 self = self
1334 .with_brackets_query(query.as_ref())
1335 .context("Error loading brackets query")?;
1336 }
1337 if let Some(query) = queries.indents {
1338 self = self
1339 .with_indents_query(query.as_ref())
1340 .context("Error loading indents query")?;
1341 }
1342 if let Some(query) = queries.outline {
1343 self = self
1344 .with_outline_query(query.as_ref())
1345 .context("Error loading outline query")?;
1346 }
1347 if let Some(query) = queries.embedding {
1348 self = self
1349 .with_embedding_query(query.as_ref())
1350 .context("Error loading embedding query")?;
1351 }
1352 if let Some(query) = queries.injections {
1353 self = self
1354 .with_injection_query(query.as_ref())
1355 .context("Error loading injection query")?;
1356 }
1357 if let Some(query) = queries.overrides {
1358 self = self
1359 .with_override_query(query.as_ref())
1360 .context("Error loading override query")?;
1361 }
1362 Ok(self)
1363 }
1364
1365 pub fn with_highlights_query(mut self, source: &str) -> Result<Self> {
1366 let grammar = self.grammar_mut();
1367 grammar.highlights_query = Some(Query::new(&grammar.ts_language, source)?);
1368 Ok(self)
1369 }
1370
1371 pub fn with_outline_query(mut self, source: &str) -> Result<Self> {
1372 let grammar = self.grammar_mut();
1373 let query = Query::new(&grammar.ts_language, source)?;
1374 let mut item_capture_ix = None;
1375 let mut name_capture_ix = None;
1376 let mut context_capture_ix = None;
1377 let mut extra_context_capture_ix = None;
1378 get_capture_indices(
1379 &query,
1380 &mut [
1381 ("item", &mut item_capture_ix),
1382 ("name", &mut name_capture_ix),
1383 ("context", &mut context_capture_ix),
1384 ("context.extra", &mut extra_context_capture_ix),
1385 ],
1386 );
1387 if let Some((item_capture_ix, name_capture_ix)) = item_capture_ix.zip(name_capture_ix) {
1388 grammar.outline_config = Some(OutlineConfig {
1389 query,
1390 item_capture_ix,
1391 name_capture_ix,
1392 context_capture_ix,
1393 extra_context_capture_ix,
1394 });
1395 }
1396 Ok(self)
1397 }
1398
1399 pub fn with_embedding_query(mut self, source: &str) -> Result<Self> {
1400 let grammar = self.grammar_mut();
1401 let query = Query::new(&grammar.ts_language, source)?;
1402 let mut item_capture_ix = None;
1403 let mut name_capture_ix = None;
1404 let mut context_capture_ix = None;
1405 let mut collapse_capture_ix = None;
1406 let mut keep_capture_ix = None;
1407 get_capture_indices(
1408 &query,
1409 &mut [
1410 ("item", &mut item_capture_ix),
1411 ("name", &mut name_capture_ix),
1412 ("context", &mut context_capture_ix),
1413 ("keep", &mut keep_capture_ix),
1414 ("collapse", &mut collapse_capture_ix),
1415 ],
1416 );
1417 if let Some(item_capture_ix) = item_capture_ix {
1418 grammar.embedding_config = Some(EmbeddingConfig {
1419 query,
1420 item_capture_ix,
1421 name_capture_ix,
1422 context_capture_ix,
1423 collapse_capture_ix,
1424 keep_capture_ix,
1425 });
1426 }
1427 Ok(self)
1428 }
1429
1430 pub fn with_brackets_query(mut self, source: &str) -> Result<Self> {
1431 let grammar = self.grammar_mut();
1432 let query = Query::new(&grammar.ts_language, source)?;
1433 let mut open_capture_ix = None;
1434 let mut close_capture_ix = None;
1435 get_capture_indices(
1436 &query,
1437 &mut [
1438 ("open", &mut open_capture_ix),
1439 ("close", &mut close_capture_ix),
1440 ],
1441 );
1442 if let Some((open_capture_ix, close_capture_ix)) = open_capture_ix.zip(close_capture_ix) {
1443 grammar.brackets_config = Some(BracketConfig {
1444 query,
1445 open_capture_ix,
1446 close_capture_ix,
1447 });
1448 }
1449 Ok(self)
1450 }
1451
1452 pub fn with_indents_query(mut self, source: &str) -> Result<Self> {
1453 let grammar = self.grammar_mut();
1454 let query = Query::new(&grammar.ts_language, source)?;
1455 let mut indent_capture_ix = None;
1456 let mut start_capture_ix = None;
1457 let mut end_capture_ix = None;
1458 let mut outdent_capture_ix = None;
1459 get_capture_indices(
1460 &query,
1461 &mut [
1462 ("indent", &mut indent_capture_ix),
1463 ("start", &mut start_capture_ix),
1464 ("end", &mut end_capture_ix),
1465 ("outdent", &mut outdent_capture_ix),
1466 ],
1467 );
1468 if let Some(indent_capture_ix) = indent_capture_ix {
1469 grammar.indents_config = Some(IndentConfig {
1470 query,
1471 indent_capture_ix,
1472 start_capture_ix,
1473 end_capture_ix,
1474 outdent_capture_ix,
1475 });
1476 }
1477 Ok(self)
1478 }
1479
1480 pub fn with_injection_query(mut self, source: &str) -> Result<Self> {
1481 let grammar = self.grammar_mut();
1482 let query = Query::new(&grammar.ts_language, source)?;
1483 let mut language_capture_ix = None;
1484 let mut content_capture_ix = None;
1485 get_capture_indices(
1486 &query,
1487 &mut [
1488 ("language", &mut language_capture_ix),
1489 ("content", &mut content_capture_ix),
1490 ],
1491 );
1492 let patterns = (0..query.pattern_count())
1493 .map(|ix| {
1494 let mut config = InjectionPatternConfig::default();
1495 for setting in query.property_settings(ix) {
1496 match setting.key.as_ref() {
1497 "language" => {
1498 config.language = setting.value.clone();
1499 }
1500 "combined" => {
1501 config.combined = true;
1502 }
1503 _ => {}
1504 }
1505 }
1506 config
1507 })
1508 .collect();
1509 if let Some(content_capture_ix) = content_capture_ix {
1510 grammar.injection_config = Some(InjectionConfig {
1511 query,
1512 language_capture_ix,
1513 content_capture_ix,
1514 patterns,
1515 });
1516 }
1517 Ok(self)
1518 }
1519
1520 pub fn with_override_query(mut self, source: &str) -> anyhow::Result<Self> {
1521 let query = Query::new(&self.grammar_mut().ts_language, source)?;
1522
1523 let mut override_configs_by_id = HashMap::default();
1524 for (ix, name) in query.capture_names().iter().enumerate() {
1525 if !name.starts_with('_') {
1526 let value = self.config.overrides.remove(*name).unwrap_or_default();
1527 for server_name in &value.opt_into_language_servers {
1528 if !self
1529 .config
1530 .scope_opt_in_language_servers
1531 .contains(server_name)
1532 {
1533 util::debug_panic!("Server {server_name:?} has been opted-in by scope {name:?} but has not been marked as an opt-in server");
1534 }
1535 }
1536
1537 override_configs_by_id.insert(ix as u32, (name.to_string(), value));
1538 }
1539 }
1540
1541 if !self.config.overrides.is_empty() {
1542 let keys = self.config.overrides.keys().collect::<Vec<_>>();
1543 Err(anyhow!(
1544 "language {:?} has overrides in config not in query: {keys:?}",
1545 self.config.name
1546 ))?;
1547 }
1548
1549 for disabled_scope_name in self
1550 .config
1551 .brackets
1552 .disabled_scopes_by_bracket_ix
1553 .iter()
1554 .flatten()
1555 {
1556 if !override_configs_by_id
1557 .values()
1558 .any(|(scope_name, _)| scope_name == disabled_scope_name)
1559 {
1560 Err(anyhow!(
1561 "language {:?} has overrides in config not in query: {disabled_scope_name:?}",
1562 self.config.name
1563 ))?;
1564 }
1565 }
1566
1567 for (name, override_config) in override_configs_by_id.values_mut() {
1568 override_config.disabled_bracket_ixs = self
1569 .config
1570 .brackets
1571 .disabled_scopes_by_bracket_ix
1572 .iter()
1573 .enumerate()
1574 .filter_map(|(ix, disabled_scope_names)| {
1575 if disabled_scope_names.contains(name) {
1576 Some(ix as u16)
1577 } else {
1578 None
1579 }
1580 })
1581 .collect();
1582 }
1583
1584 self.config.brackets.disabled_scopes_by_bracket_ix.clear();
1585 self.grammar_mut().override_config = Some(OverrideConfig {
1586 query,
1587 values: override_configs_by_id,
1588 });
1589 Ok(self)
1590 }
1591
1592 fn grammar_mut(&mut self) -> &mut Grammar {
1593 Arc::get_mut(self.grammar.as_mut().unwrap()).unwrap()
1594 }
1595
1596 pub async fn with_lsp_adapters(mut self, lsp_adapters: Vec<Arc<dyn LspAdapter>>) -> Self {
1597 for adapter in lsp_adapters {
1598 self.adapters.push(CachedLspAdapter::new(adapter).await);
1599 }
1600 self
1601 }
1602
1603 #[cfg(any(test, feature = "test-support"))]
1604 pub async fn set_fake_lsp_adapter(
1605 &mut self,
1606 fake_lsp_adapter: Arc<FakeLspAdapter>,
1607 ) -> mpsc::UnboundedReceiver<lsp::FakeLanguageServer> {
1608 let (servers_tx, servers_rx) = mpsc::unbounded();
1609 self.fake_adapter = Some((servers_tx, fake_lsp_adapter.clone()));
1610 let adapter = CachedLspAdapter::new(Arc::new(fake_lsp_adapter)).await;
1611 self.adapters = vec![adapter];
1612 servers_rx
1613 }
1614
1615 pub fn name(&self) -> Arc<str> {
1616 self.config.name.clone()
1617 }
1618
1619 pub async fn disk_based_diagnostic_sources(&self) -> &[String] {
1620 match self.adapters.first().as_ref() {
1621 Some(adapter) => &adapter.disk_based_diagnostic_sources,
1622 None => &[],
1623 }
1624 }
1625
1626 pub async fn disk_based_diagnostics_progress_token(&self) -> Option<&str> {
1627 for adapter in &self.adapters {
1628 let token = adapter.disk_based_diagnostics_progress_token.as_deref();
1629 if token.is_some() {
1630 return token;
1631 }
1632 }
1633
1634 None
1635 }
1636
1637 pub async fn process_completion(self: &Arc<Self>, completion: &mut lsp::CompletionItem) {
1638 for adapter in &self.adapters {
1639 adapter.process_completion(completion).await;
1640 }
1641 }
1642
1643 pub async fn label_for_completion(
1644 self: &Arc<Self>,
1645 completion: &lsp::CompletionItem,
1646 ) -> Option<CodeLabel> {
1647 self.adapters
1648 .first()
1649 .as_ref()?
1650 .label_for_completion(completion, self)
1651 .await
1652 }
1653
1654 pub async fn label_for_symbol(
1655 self: &Arc<Self>,
1656 name: &str,
1657 kind: lsp::SymbolKind,
1658 ) -> Option<CodeLabel> {
1659 self.adapters
1660 .first()
1661 .as_ref()?
1662 .label_for_symbol(name, kind, self)
1663 .await
1664 }
1665
1666 pub fn highlight_text<'a>(
1667 self: &'a Arc<Self>,
1668 text: &'a Rope,
1669 range: Range<usize>,
1670 ) -> Vec<(Range<usize>, HighlightId)> {
1671 let mut result = Vec::new();
1672 if let Some(grammar) = &self.grammar {
1673 let tree = grammar.parse_text(text, None);
1674 let captures =
1675 SyntaxSnapshot::single_tree_captures(range.clone(), text, &tree, self, |grammar| {
1676 grammar.highlights_query.as_ref()
1677 });
1678 let highlight_maps = vec![grammar.highlight_map()];
1679 let mut offset = 0;
1680 for chunk in BufferChunks::new(text, range, Some((captures, highlight_maps)), vec![]) {
1681 let end_offset = offset + chunk.text.len();
1682 if let Some(highlight_id) = chunk.syntax_highlight_id {
1683 if !highlight_id.is_default() {
1684 result.push((offset..end_offset, highlight_id));
1685 }
1686 }
1687 offset = end_offset;
1688 }
1689 }
1690 result
1691 }
1692
1693 pub fn path_suffixes(&self) -> &[String] {
1694 &self.config.path_suffixes
1695 }
1696
1697 pub fn should_autoclose_before(&self, c: char) -> bool {
1698 c.is_whitespace() || self.config.autoclose_before.contains(c)
1699 }
1700
1701 pub fn set_theme(&self, theme: &SyntaxTheme) {
1702 if let Some(grammar) = self.grammar.as_ref() {
1703 if let Some(highlights_query) = &grammar.highlights_query {
1704 *grammar.highlight_map.lock() =
1705 HighlightMap::new(highlights_query.capture_names(), theme);
1706 }
1707 }
1708 }
1709
1710 pub fn grammar(&self) -> Option<&Arc<Grammar>> {
1711 self.grammar.as_ref()
1712 }
1713
1714 pub fn default_scope(self: &Arc<Self>) -> LanguageScope {
1715 LanguageScope {
1716 language: self.clone(),
1717 override_id: None,
1718 }
1719 }
1720
1721 pub fn prettier_parser_name(&self) -> Option<&str> {
1722 self.config.prettier_parser_name.as_deref()
1723 }
1724}
1725
1726impl LanguageScope {
1727 pub fn collapsed_placeholder(&self) -> &str {
1728 self.language.config.collapsed_placeholder.as_ref()
1729 }
1730
1731 /// Returns line prefix that is inserted in e.g. line continuations or
1732 /// in `toggle comments` action.
1733 pub fn line_comment_prefixes(&self) -> Option<&Vec<Arc<str>>> {
1734 Override::as_option(
1735 self.config_override().map(|o| &o.line_comments),
1736 Some(&self.language.config.line_comments),
1737 )
1738 }
1739
1740 pub fn block_comment_delimiters(&self) -> Option<(&Arc<str>, &Arc<str>)> {
1741 Override::as_option(
1742 self.config_override().map(|o| &o.block_comment),
1743 self.language.config.block_comment.as_ref(),
1744 )
1745 .map(|e| (&e.0, &e.1))
1746 }
1747
1748 /// Returns a list of language-specific word characters.
1749 ///
1750 /// By default, Zed treats alphanumeric characters (and '_') as word characters for
1751 /// the purpose of actions like 'move to next word end` or whole-word search.
1752 /// It additionally accounts for language's additional word characters.
1753 pub fn word_characters(&self) -> Option<&HashSet<char>> {
1754 Override::as_option(
1755 self.config_override().map(|o| &o.word_characters),
1756 Some(&self.language.config.word_characters),
1757 )
1758 }
1759
1760 /// Returns a list of bracket pairs for a given language with an additional
1761 /// piece of information about whether the particular bracket pair is currently active for a given language.
1762 pub fn brackets(&self) -> impl Iterator<Item = (&BracketPair, bool)> {
1763 let mut disabled_ids = self
1764 .config_override()
1765 .map_or(&[] as _, |o| o.disabled_bracket_ixs.as_slice());
1766 self.language
1767 .config
1768 .brackets
1769 .pairs
1770 .iter()
1771 .enumerate()
1772 .map(move |(ix, bracket)| {
1773 let mut is_enabled = true;
1774 if let Some(next_disabled_ix) = disabled_ids.first() {
1775 if ix == *next_disabled_ix as usize {
1776 disabled_ids = &disabled_ids[1..];
1777 is_enabled = false;
1778 }
1779 }
1780 (bracket, is_enabled)
1781 })
1782 }
1783
1784 pub fn should_autoclose_before(&self, c: char) -> bool {
1785 c.is_whitespace() || self.language.config.autoclose_before.contains(c)
1786 }
1787
1788 pub fn language_allowed(&self, name: &LanguageServerName) -> bool {
1789 let config = &self.language.config;
1790 let opt_in_servers = &config.scope_opt_in_language_servers;
1791 if opt_in_servers.iter().any(|o| *o == *name.0) {
1792 if let Some(over) = self.config_override() {
1793 over.opt_into_language_servers.iter().any(|o| *o == *name.0)
1794 } else {
1795 false
1796 }
1797 } else {
1798 true
1799 }
1800 }
1801
1802 fn config_override(&self) -> Option<&LanguageConfigOverride> {
1803 let id = self.override_id?;
1804 let grammar = self.language.grammar.as_ref()?;
1805 let override_config = grammar.override_config.as_ref()?;
1806 override_config.values.get(&id).map(|e| &e.1)
1807 }
1808}
1809
1810impl Hash for Language {
1811 fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
1812 self.id().hash(state)
1813 }
1814}
1815
1816impl PartialEq for Language {
1817 fn eq(&self, other: &Self) -> bool {
1818 self.id().eq(&other.id())
1819 }
1820}
1821
1822impl Eq for Language {}
1823
1824impl Debug for Language {
1825 fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
1826 f.debug_struct("Language")
1827 .field("name", &self.config.name)
1828 .finish()
1829 }
1830}
1831
1832impl Grammar {
1833 pub fn id(&self) -> usize {
1834 self.id
1835 }
1836
1837 fn parse_text(&self, text: &Rope, old_tree: Option<Tree>) -> Tree {
1838 PARSER.with(|parser| {
1839 let mut parser = parser.borrow_mut();
1840 parser
1841 .set_language(&self.ts_language)
1842 .expect("incompatible grammar");
1843 let mut chunks = text.chunks_in_range(0..text.len());
1844 parser
1845 .parse_with(
1846 &mut move |offset, _| {
1847 chunks.seek(offset);
1848 chunks.next().unwrap_or("").as_bytes()
1849 },
1850 old_tree.as_ref(),
1851 )
1852 .unwrap()
1853 })
1854 }
1855
1856 pub fn highlight_map(&self) -> HighlightMap {
1857 self.highlight_map.lock().clone()
1858 }
1859
1860 pub fn highlight_id_for_name(&self, name: &str) -> Option<HighlightId> {
1861 let capture_id = self
1862 .highlights_query
1863 .as_ref()?
1864 .capture_index_for_name(name)?;
1865 Some(self.highlight_map.lock().get(capture_id))
1866 }
1867}
1868
1869impl CodeLabel {
1870 pub fn plain(text: String, filter_text: Option<&str>) -> Self {
1871 let mut result = Self {
1872 runs: Vec::new(),
1873 filter_range: 0..text.len(),
1874 text,
1875 };
1876 if let Some(filter_text) = filter_text {
1877 if let Some(ix) = result.text.find(filter_text) {
1878 result.filter_range = ix..ix + filter_text.len();
1879 }
1880 }
1881 result
1882 }
1883}
1884
1885#[cfg(any(test, feature = "test-support"))]
1886impl Default for FakeLspAdapter {
1887 fn default() -> Self {
1888 Self {
1889 name: "the-fake-language-server",
1890 capabilities: lsp::LanguageServer::full_capabilities(),
1891 initializer: None,
1892 disk_based_diagnostics_progress_token: None,
1893 initialization_options: None,
1894 disk_based_diagnostics_sources: Vec::new(),
1895 prettier_plugins: Vec::new(),
1896 }
1897 }
1898}
1899
1900#[cfg(any(test, feature = "test-support"))]
1901#[async_trait]
1902impl LspAdapter for Arc<FakeLspAdapter> {
1903 fn name(&self) -> LanguageServerName {
1904 LanguageServerName(self.name.into())
1905 }
1906
1907 fn short_name(&self) -> &'static str {
1908 "FakeLspAdapter"
1909 }
1910
1911 async fn fetch_latest_server_version(
1912 &self,
1913 _: &dyn LspAdapterDelegate,
1914 ) -> Result<Box<dyn 'static + Send + Any>> {
1915 unreachable!();
1916 }
1917
1918 async fn fetch_server_binary(
1919 &self,
1920 _: Box<dyn 'static + Send + Any>,
1921 _: PathBuf,
1922 _: &dyn LspAdapterDelegate,
1923 ) -> Result<LanguageServerBinary> {
1924 unreachable!();
1925 }
1926
1927 async fn cached_server_binary(
1928 &self,
1929 _: PathBuf,
1930 _: &dyn LspAdapterDelegate,
1931 ) -> Option<LanguageServerBinary> {
1932 unreachable!();
1933 }
1934
1935 async fn installation_test_binary(&self, _: PathBuf) -> Option<LanguageServerBinary> {
1936 unreachable!();
1937 }
1938
1939 fn process_diagnostics(&self, _: &mut lsp::PublishDiagnosticsParams) {}
1940
1941 fn disk_based_diagnostic_sources(&self) -> Vec<String> {
1942 self.disk_based_diagnostics_sources.clone()
1943 }
1944
1945 fn disk_based_diagnostics_progress_token(&self) -> Option<String> {
1946 self.disk_based_diagnostics_progress_token.clone()
1947 }
1948
1949 fn initialization_options(&self) -> Option<Value> {
1950 self.initialization_options.clone()
1951 }
1952
1953 fn prettier_plugins(&self) -> &[&'static str] {
1954 &self.prettier_plugins
1955 }
1956}
1957
1958fn get_capture_indices(query: &Query, captures: &mut [(&str, &mut Option<u32>)]) {
1959 for (ix, name) in query.capture_names().iter().enumerate() {
1960 for (capture_name, index) in captures.iter_mut() {
1961 if capture_name == name {
1962 **index = Some(ix as u32);
1963 break;
1964 }
1965 }
1966 }
1967}
1968
1969pub fn point_to_lsp(point: PointUtf16) -> lsp::Position {
1970 lsp::Position::new(point.row, point.column)
1971}
1972
1973pub fn point_from_lsp(point: lsp::Position) -> Unclipped<PointUtf16> {
1974 Unclipped(PointUtf16::new(point.line, point.character))
1975}
1976
1977pub fn range_to_lsp(range: Range<PointUtf16>) -> lsp::Range {
1978 lsp::Range {
1979 start: point_to_lsp(range.start),
1980 end: point_to_lsp(range.end),
1981 }
1982}
1983
1984pub fn range_from_lsp(range: lsp::Range) -> Range<Unclipped<PointUtf16>> {
1985 let mut start = point_from_lsp(range.start);
1986 let mut end = point_from_lsp(range.end);
1987 if start > end {
1988 mem::swap(&mut start, &mut end);
1989 }
1990 start..end
1991}
1992
1993#[cfg(test)]
1994mod tests {
1995 use super::*;
1996 use gpui::TestAppContext;
1997
1998 #[gpui::test(iterations = 10)]
1999 async fn test_first_line_pattern(cx: &mut TestAppContext) {
2000 let mut languages = LanguageRegistry::test();
2001
2002 languages.set_executor(cx.executor());
2003 let languages = Arc::new(languages);
2004 languages.register(
2005 "/javascript",
2006 LanguageConfig {
2007 name: "JavaScript".into(),
2008 path_suffixes: vec!["js".into()],
2009 first_line_pattern: Some(Regex::new(r"\bnode\b").unwrap()),
2010 ..Default::default()
2011 },
2012 tree_sitter_typescript::language_tsx(),
2013 vec![],
2014 |_| Default::default(),
2015 );
2016
2017 languages
2018 .language_for_file("the/script", None)
2019 .await
2020 .unwrap_err();
2021 languages
2022 .language_for_file("the/script", Some(&"nothing".into()))
2023 .await
2024 .unwrap_err();
2025 assert_eq!(
2026 languages
2027 .language_for_file("the/script", Some(&"#!/bin/env node".into()))
2028 .await
2029 .unwrap()
2030 .name()
2031 .as_ref(),
2032 "JavaScript"
2033 );
2034 }
2035
2036 #[gpui::test(iterations = 10)]
2037 async fn test_language_loading(cx: &mut TestAppContext) {
2038 let mut languages = LanguageRegistry::test();
2039 languages.set_executor(cx.executor());
2040 let languages = Arc::new(languages);
2041 languages.register(
2042 "/JSON",
2043 LanguageConfig {
2044 name: "JSON".into(),
2045 path_suffixes: vec!["json".into()],
2046 ..Default::default()
2047 },
2048 tree_sitter_json::language(),
2049 vec![],
2050 |_| Default::default(),
2051 );
2052 languages.register(
2053 "/rust",
2054 LanguageConfig {
2055 name: "Rust".into(),
2056 path_suffixes: vec!["rs".into()],
2057 ..Default::default()
2058 },
2059 tree_sitter_rust::language(),
2060 vec![],
2061 |_| Default::default(),
2062 );
2063 assert_eq!(
2064 languages.language_names(),
2065 &[
2066 "JSON".to_string(),
2067 "Plain Text".to_string(),
2068 "Rust".to_string(),
2069 ]
2070 );
2071
2072 let rust1 = languages.language_for_name("Rust");
2073 let rust2 = languages.language_for_name("Rust");
2074
2075 // Ensure language is still listed even if it's being loaded.
2076 assert_eq!(
2077 languages.language_names(),
2078 &[
2079 "JSON".to_string(),
2080 "Plain Text".to_string(),
2081 "Rust".to_string(),
2082 ]
2083 );
2084
2085 let (rust1, rust2) = futures::join!(rust1, rust2);
2086 assert!(Arc::ptr_eq(&rust1.unwrap(), &rust2.unwrap()));
2087
2088 // Ensure language is still listed even after loading it.
2089 assert_eq!(
2090 languages.language_names(),
2091 &[
2092 "JSON".to_string(),
2093 "Plain Text".to_string(),
2094 "Rust".to_string(),
2095 ]
2096 );
2097
2098 // Loading an unknown language returns an error.
2099 assert!(languages.language_for_name("Unknown").await.is_err());
2100 }
2101}