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 pub redactions: Option<Cow<'static, str>>,
457}
458
459/// Represents a language for the given range. Some languages (e.g. HTML)
460/// interleave several languages together, thus a single buffer might actually contain
461/// several nested scopes.
462#[derive(Clone, Debug)]
463pub struct LanguageScope {
464 language: Arc<Language>,
465 override_id: Option<u32>,
466}
467
468#[derive(Clone, Deserialize, Default, Debug)]
469pub struct LanguageConfigOverride {
470 #[serde(default)]
471 pub line_comments: Override<Vec<Arc<str>>>,
472 #[serde(default)]
473 pub block_comment: Override<(Arc<str>, Arc<str>)>,
474 #[serde(skip_deserializing)]
475 pub disabled_bracket_ixs: Vec<u16>,
476 #[serde(default)]
477 pub word_characters: Override<HashSet<char>>,
478 #[serde(default)]
479 pub opt_into_language_servers: Vec<String>,
480}
481
482#[derive(Clone, Deserialize, Debug)]
483#[serde(untagged)]
484pub enum Override<T> {
485 Remove { remove: bool },
486 Set(T),
487}
488
489impl<T> Default for Override<T> {
490 fn default() -> Self {
491 Override::Remove { remove: false }
492 }
493}
494
495impl<T> Override<T> {
496 fn as_option<'a>(this: Option<&'a Self>, original: Option<&'a T>) -> Option<&'a T> {
497 match this {
498 Some(Self::Set(value)) => Some(value),
499 Some(Self::Remove { remove: true }) => None,
500 Some(Self::Remove { remove: false }) | None => original,
501 }
502 }
503}
504
505impl Default for LanguageConfig {
506 fn default() -> Self {
507 Self {
508 name: "".into(),
509 grammar_name: None,
510 path_suffixes: Default::default(),
511 brackets: Default::default(),
512 auto_indent_using_last_non_empty_line: auto_indent_using_last_non_empty_line_default(),
513 first_line_pattern: Default::default(),
514 increase_indent_pattern: Default::default(),
515 decrease_indent_pattern: Default::default(),
516 autoclose_before: Default::default(),
517 line_comments: Default::default(),
518 block_comment: Default::default(),
519 scope_opt_in_language_servers: Default::default(),
520 overrides: Default::default(),
521 word_characters: Default::default(),
522 prettier_parser_name: None,
523 collapsed_placeholder: Default::default(),
524 }
525 }
526}
527
528fn auto_indent_using_last_non_empty_line_default() -> bool {
529 true
530}
531
532fn deserialize_regex<'de, D: Deserializer<'de>>(d: D) -> Result<Option<Regex>, D::Error> {
533 let source = Option::<String>::deserialize(d)?;
534 if let Some(source) = source {
535 Ok(Some(regex::Regex::new(&source).map_err(de::Error::custom)?))
536 } else {
537 Ok(None)
538 }
539}
540
541#[doc(hidden)]
542#[cfg(any(test, feature = "test-support"))]
543pub struct FakeLspAdapter {
544 pub name: &'static str,
545 pub initialization_options: Option<Value>,
546 pub capabilities: lsp::ServerCapabilities,
547 pub initializer: Option<Box<dyn 'static + Send + Sync + Fn(&mut lsp::FakeLanguageServer)>>,
548 pub disk_based_diagnostics_progress_token: Option<String>,
549 pub disk_based_diagnostics_sources: Vec<String>,
550 pub prettier_plugins: Vec<&'static str>,
551}
552
553/// Configuration of handling bracket pairs for a given language.
554///
555/// This struct includes settings for defining which pairs of characters are considered brackets and
556/// also specifies any language-specific scopes where these pairs should be ignored for bracket matching purposes.
557#[derive(Clone, Debug, Default)]
558pub struct BracketPairConfig {
559 /// A list of character pairs that should be treated as brackets in the context of a given language.
560 pub pairs: Vec<BracketPair>,
561 /// A list of tree-sitter scopes for which a given bracket should not be active.
562 /// N-th entry in `[Self::disabled_scopes_by_bracket_ix]` contains a list of disabled scopes for an n-th entry in `[Self::pairs]`
563 pub disabled_scopes_by_bracket_ix: Vec<Vec<String>>,
564}
565
566impl<'de> Deserialize<'de> for BracketPairConfig {
567 fn deserialize<D>(deserializer: D) -> std::result::Result<Self, D::Error>
568 where
569 D: Deserializer<'de>,
570 {
571 #[derive(Deserialize)]
572 pub struct Entry {
573 #[serde(flatten)]
574 pub bracket_pair: BracketPair,
575 #[serde(default)]
576 pub not_in: Vec<String>,
577 }
578
579 let result = Vec::<Entry>::deserialize(deserializer)?;
580 let mut brackets = Vec::with_capacity(result.len());
581 let mut disabled_scopes_by_bracket_ix = Vec::with_capacity(result.len());
582 for entry in result {
583 brackets.push(entry.bracket_pair);
584 disabled_scopes_by_bracket_ix.push(entry.not_in);
585 }
586
587 Ok(BracketPairConfig {
588 pairs: brackets,
589 disabled_scopes_by_bracket_ix,
590 })
591 }
592}
593
594/// Describes a single bracket pair and how an editor should react to e.g. inserting
595/// an opening bracket or to a newline character insertion in between `start` and `end` characters.
596#[derive(Clone, Debug, Default, Deserialize, PartialEq)]
597pub struct BracketPair {
598 /// Starting substring for a bracket.
599 pub start: String,
600 /// Ending substring for a bracket.
601 pub end: String,
602 /// True if `end` should be automatically inserted right after `start` characters.
603 pub close: bool,
604 /// True if an extra newline should be inserted while the cursor is in the middle
605 /// of that bracket pair.
606 pub newline: bool,
607}
608
609pub struct Language {
610 pub(crate) config: LanguageConfig,
611 pub(crate) grammar: Option<Arc<Grammar>>,
612 pub(crate) adapters: Vec<Arc<CachedLspAdapter>>,
613
614 #[cfg(any(test, feature = "test-support"))]
615 fake_adapter: Option<(
616 mpsc::UnboundedSender<lsp::FakeLanguageServer>,
617 Arc<FakeLspAdapter>,
618 )>,
619}
620
621pub struct Grammar {
622 id: usize,
623 pub ts_language: tree_sitter::Language,
624 pub(crate) error_query: Query,
625 pub(crate) highlights_query: Option<Query>,
626 pub(crate) brackets_config: Option<BracketConfig>,
627 pub(crate) redactions_config: Option<RedactionConfig>,
628 pub(crate) indents_config: Option<IndentConfig>,
629 pub outline_config: Option<OutlineConfig>,
630 pub embedding_config: Option<EmbeddingConfig>,
631 pub(crate) injection_config: Option<InjectionConfig>,
632 pub(crate) override_config: Option<OverrideConfig>,
633 pub(crate) highlight_map: Mutex<HighlightMap>,
634}
635
636struct IndentConfig {
637 query: Query,
638 indent_capture_ix: u32,
639 start_capture_ix: Option<u32>,
640 end_capture_ix: Option<u32>,
641 outdent_capture_ix: Option<u32>,
642}
643
644pub struct OutlineConfig {
645 pub query: Query,
646 pub item_capture_ix: u32,
647 pub name_capture_ix: u32,
648 pub context_capture_ix: Option<u32>,
649 pub extra_context_capture_ix: Option<u32>,
650}
651
652#[derive(Debug)]
653pub struct EmbeddingConfig {
654 pub query: Query,
655 pub item_capture_ix: u32,
656 pub name_capture_ix: Option<u32>,
657 pub context_capture_ix: Option<u32>,
658 pub collapse_capture_ix: Option<u32>,
659 pub keep_capture_ix: Option<u32>,
660}
661
662struct InjectionConfig {
663 query: Query,
664 content_capture_ix: u32,
665 language_capture_ix: Option<u32>,
666 patterns: Vec<InjectionPatternConfig>,
667}
668
669struct RedactionConfig {
670 pub query: Query,
671 pub redaction_capture_ix: u32,
672}
673
674struct OverrideConfig {
675 query: Query,
676 values: HashMap<u32, (String, LanguageConfigOverride)>,
677}
678
679#[derive(Default, Clone)]
680struct InjectionPatternConfig {
681 language: Option<Box<str>>,
682 combined: bool,
683}
684
685struct BracketConfig {
686 query: Query,
687 open_capture_ix: u32,
688 close_capture_ix: u32,
689}
690
691#[derive(Clone)]
692pub enum LanguageServerBinaryStatus {
693 CheckingForUpdate,
694 Downloading,
695 Downloaded,
696 Cached,
697 Failed { error: String },
698}
699
700type AvailableLanguageId = usize;
701
702#[derive(Clone)]
703struct AvailableLanguage {
704 id: AvailableLanguageId,
705 config: LanguageConfig,
706 grammar: AvailableGrammar,
707 lsp_adapters: Vec<Arc<dyn LspAdapter>>,
708 loaded: bool,
709}
710
711#[derive(Clone)]
712enum AvailableGrammar {
713 Native {
714 grammar: tree_sitter::Language,
715 asset_dir: &'static str,
716 get_queries: fn(&str) -> LanguageQueries,
717 },
718 Wasm {
719 path: Arc<Path>,
720 get_queries: fn(&Path) -> LanguageQueries,
721 },
722}
723
724pub struct LanguageRegistry {
725 state: RwLock<LanguageRegistryState>,
726 language_server_download_dir: Option<Arc<Path>>,
727 login_shell_env_loaded: Shared<Task<()>>,
728 #[allow(clippy::type_complexity)]
729 lsp_binary_paths: Mutex<
730 HashMap<LanguageServerName, Shared<Task<Result<LanguageServerBinary, Arc<anyhow::Error>>>>>,
731 >,
732 executor: Option<BackgroundExecutor>,
733 lsp_binary_status_tx: LspBinaryStatusSender,
734}
735
736struct LanguageRegistryState {
737 next_language_server_id: usize,
738 languages: Vec<Arc<Language>>,
739 available_languages: Vec<AvailableLanguage>,
740 next_available_language_id: AvailableLanguageId,
741 loading_languages: HashMap<AvailableLanguageId, Vec<oneshot::Sender<Result<Arc<Language>>>>>,
742 subscription: (watch::Sender<()>, watch::Receiver<()>),
743 theme: Option<Arc<Theme>>,
744 version: usize,
745 reload_count: usize,
746}
747
748pub struct PendingLanguageServer {
749 pub server_id: LanguageServerId,
750 pub task: Task<Result<lsp::LanguageServer>>,
751 pub container_dir: Option<Arc<Path>>,
752}
753
754impl LanguageRegistry {
755 pub fn new(login_shell_env_loaded: Task<()>) -> Self {
756 Self {
757 state: RwLock::new(LanguageRegistryState {
758 next_language_server_id: 0,
759 languages: vec![PLAIN_TEXT.clone()],
760 available_languages: Default::default(),
761 next_available_language_id: 0,
762 loading_languages: Default::default(),
763 subscription: watch::channel(),
764 theme: Default::default(),
765 version: 0,
766 reload_count: 0,
767 }),
768 language_server_download_dir: None,
769 login_shell_env_loaded: login_shell_env_loaded.shared(),
770 lsp_binary_paths: Default::default(),
771 executor: None,
772 lsp_binary_status_tx: Default::default(),
773 }
774 }
775
776 #[cfg(any(test, feature = "test-support"))]
777 pub fn test() -> Self {
778 Self::new(Task::ready(()))
779 }
780
781 pub fn set_executor(&mut self, executor: BackgroundExecutor) {
782 self.executor = Some(executor);
783 }
784
785 /// Clear out all of the loaded languages and reload them from scratch.
786 pub fn reload(&self) {
787 self.state.write().reload();
788 }
789
790 pub fn register(
791 &self,
792 asset_dir: &'static str,
793 config: LanguageConfig,
794 grammar: tree_sitter::Language,
795 lsp_adapters: Vec<Arc<dyn LspAdapter>>,
796 get_queries: fn(&str) -> LanguageQueries,
797 ) {
798 let state = &mut *self.state.write();
799 state.available_languages.push(AvailableLanguage {
800 id: post_inc(&mut state.next_available_language_id),
801 config,
802 grammar: AvailableGrammar::Native {
803 grammar,
804 get_queries,
805 asset_dir,
806 },
807 lsp_adapters,
808 loaded: false,
809 });
810 }
811
812 pub fn register_wasm(
813 &self,
814 path: Arc<Path>,
815 config: LanguageConfig,
816 get_queries: fn(&Path) -> LanguageQueries,
817 ) {
818 let state = &mut *self.state.write();
819 state.available_languages.push(AvailableLanguage {
820 id: post_inc(&mut state.next_available_language_id),
821 config,
822 grammar: AvailableGrammar::Wasm { path, get_queries },
823 lsp_adapters: Vec::new(),
824 loaded: false,
825 });
826 }
827
828 pub fn language_names(&self) -> Vec<String> {
829 let state = self.state.read();
830 let mut result = state
831 .available_languages
832 .iter()
833 .filter_map(|l| l.loaded.not().then_some(l.config.name.to_string()))
834 .chain(state.languages.iter().map(|l| l.config.name.to_string()))
835 .collect::<Vec<_>>();
836 result.sort_unstable_by_key(|language_name| language_name.to_lowercase());
837 result
838 }
839
840 pub fn add(&self, language: Arc<Language>) {
841 self.state.write().add(language);
842 }
843
844 pub fn subscribe(&self) -> watch::Receiver<()> {
845 self.state.read().subscription.1.clone()
846 }
847
848 /// The number of times that the registry has been changed,
849 /// by adding languages or reloading.
850 pub fn version(&self) -> usize {
851 self.state.read().version
852 }
853
854 /// The number of times that the registry has been reloaded.
855 pub fn reload_count(&self) -> usize {
856 self.state.read().reload_count
857 }
858
859 pub fn set_theme(&self, theme: Arc<Theme>) {
860 let mut state = self.state.write();
861 state.theme = Some(theme.clone());
862 for language in &state.languages {
863 language.set_theme(theme.syntax());
864 }
865 }
866
867 pub fn set_language_server_download_dir(&mut self, path: impl Into<Arc<Path>>) {
868 self.language_server_download_dir = Some(path.into());
869 }
870
871 pub fn language_for_name(
872 self: &Arc<Self>,
873 name: &str,
874 ) -> UnwrapFuture<oneshot::Receiver<Result<Arc<Language>>>> {
875 let name = UniCase::new(name);
876 self.get_or_load_language(|config| UniCase::new(config.name.as_ref()) == name)
877 }
878
879 pub fn language_for_name_or_extension(
880 self: &Arc<Self>,
881 string: &str,
882 ) -> UnwrapFuture<oneshot::Receiver<Result<Arc<Language>>>> {
883 let string = UniCase::new(string);
884 self.get_or_load_language(|config| {
885 UniCase::new(config.name.as_ref()) == string
886 || config
887 .path_suffixes
888 .iter()
889 .any(|suffix| UniCase::new(suffix) == string)
890 })
891 }
892
893 pub fn language_for_file(
894 self: &Arc<Self>,
895 path: impl AsRef<Path>,
896 content: Option<&Rope>,
897 ) -> UnwrapFuture<oneshot::Receiver<Result<Arc<Language>>>> {
898 let path = path.as_ref();
899 let filename = path.file_name().and_then(|name| name.to_str());
900 let extension = path.extension_or_hidden_file_name();
901 let path_suffixes = [extension, filename];
902 self.get_or_load_language(|config| {
903 let path_matches = config
904 .path_suffixes
905 .iter()
906 .any(|suffix| path_suffixes.contains(&Some(suffix.as_str())));
907 let content_matches = content.zip(config.first_line_pattern.as_ref()).map_or(
908 false,
909 |(content, pattern)| {
910 let end = content.clip_point(Point::new(0, 256), Bias::Left);
911 let end = content.point_to_offset(end);
912 let text = content.chunks_in_range(0..end).collect::<String>();
913 pattern.is_match(&text)
914 },
915 );
916 path_matches || content_matches
917 })
918 }
919
920 fn get_or_load_language(
921 self: &Arc<Self>,
922 callback: impl Fn(&LanguageConfig) -> bool,
923 ) -> UnwrapFuture<oneshot::Receiver<Result<Arc<Language>>>> {
924 let (tx, rx) = oneshot::channel();
925
926 let mut state = self.state.write();
927 if let Some(language) = state
928 .languages
929 .iter()
930 .find(|language| callback(&language.config))
931 {
932 let _ = tx.send(Ok(language.clone()));
933 } else if let Some(executor) = self.executor.clone() {
934 if let Some(language) = state
935 .available_languages
936 .iter()
937 .find(|l| !l.loaded && callback(&l.config))
938 .cloned()
939 {
940 let txs = state
941 .loading_languages
942 .entry(language.id)
943 .or_insert_with(|| {
944 let this = self.clone();
945 executor
946 .spawn(async move {
947 let id = language.id;
948 let name = language.config.name.clone();
949 let language = async {
950 let (grammar, queries) = match language.grammar {
951 AvailableGrammar::Native {
952 grammar,
953 asset_dir,
954 get_queries,
955 } => (grammar, (get_queries)(asset_dir)),
956 AvailableGrammar::Wasm { path, get_queries } => {
957 let grammar_name =
958 &language.config.grammar_name.as_ref().ok_or_else(
959 || anyhow!("missing grammar name"),
960 )?;
961 let mut wasm_path = path.join(grammar_name.as_ref());
962 wasm_path.set_extension("wasm");
963 let wasm_bytes = std::fs::read(&wasm_path)?;
964 let grammar = PARSER.with(|parser| {
965 let mut parser = parser.borrow_mut();
966 let mut store = parser.take_wasm_store().unwrap();
967 let grammar =
968 store.load_language(&grammar_name, &wasm_bytes);
969 parser.set_wasm_store(store).unwrap();
970 grammar
971 })?;
972 (grammar, get_queries(path.as_ref()))
973 }
974 };
975 Language::new(language.config, Some(grammar))
976 .with_lsp_adapters(language.lsp_adapters)
977 .await
978 .with_queries(queries)
979 }
980 .await;
981
982 match language {
983 Ok(language) => {
984 let language = Arc::new(language);
985 let mut state = this.state.write();
986
987 state.add(language.clone());
988 state.mark_language_loaded(id);
989 if let Some(mut txs) = state.loading_languages.remove(&id) {
990 for tx in txs.drain(..) {
991 let _ = tx.send(Ok(language.clone()));
992 }
993 }
994 }
995 Err(e) => {
996 log::error!("failed to load language {name}:\n{:?}", e);
997 let mut state = this.state.write();
998 state.mark_language_loaded(id);
999 if let Some(mut txs) = state.loading_languages.remove(&id) {
1000 for tx in txs.drain(..) {
1001 let _ = tx.send(Err(anyhow!(
1002 "failed to load language {}: {}",
1003 name,
1004 e
1005 )));
1006 }
1007 }
1008 }
1009 };
1010 })
1011 .detach();
1012
1013 Vec::new()
1014 });
1015 txs.push(tx);
1016 } else {
1017 let _ = tx.send(Err(anyhow!("language not found")));
1018 }
1019 } else {
1020 let _ = tx.send(Err(anyhow!("executor does not exist")));
1021 }
1022
1023 rx.unwrap()
1024 }
1025
1026 pub fn to_vec(&self) -> Vec<Arc<Language>> {
1027 self.state.read().languages.iter().cloned().collect()
1028 }
1029
1030 pub fn create_pending_language_server(
1031 self: &Arc<Self>,
1032 stderr_capture: Arc<Mutex<Option<String>>>,
1033 language: Arc<Language>,
1034 adapter: Arc<CachedLspAdapter>,
1035 root_path: Arc<Path>,
1036 delegate: Arc<dyn LspAdapterDelegate>,
1037 cx: &mut AppContext,
1038 ) -> Option<PendingLanguageServer> {
1039 let server_id = self.state.write().next_language_server_id();
1040 log::info!(
1041 "starting language server {:?}, path: {root_path:?}, id: {server_id}",
1042 adapter.name.0
1043 );
1044
1045 #[cfg(any(test, feature = "test-support"))]
1046 if language.fake_adapter.is_some() {
1047 let task = cx.spawn(|cx| async move {
1048 let (servers_tx, fake_adapter) = language.fake_adapter.as_ref().unwrap();
1049 let (server, mut fake_server) = lsp::FakeLanguageServer::new(
1050 fake_adapter.name.to_string(),
1051 fake_adapter.capabilities.clone(),
1052 cx.clone(),
1053 );
1054
1055 if let Some(initializer) = &fake_adapter.initializer {
1056 initializer(&mut fake_server);
1057 }
1058
1059 let servers_tx = servers_tx.clone();
1060 cx.background_executor()
1061 .spawn(async move {
1062 if fake_server
1063 .try_receive_notification::<lsp::notification::Initialized>()
1064 .await
1065 .is_some()
1066 {
1067 servers_tx.unbounded_send(fake_server).ok();
1068 }
1069 })
1070 .detach();
1071
1072 Ok(server)
1073 });
1074
1075 return Some(PendingLanguageServer {
1076 server_id,
1077 task,
1078 container_dir: None,
1079 });
1080 }
1081
1082 let download_dir = self
1083 .language_server_download_dir
1084 .clone()
1085 .ok_or_else(|| anyhow!("language server download directory has not been assigned before starting server"))
1086 .log_err()?;
1087 let this = self.clone();
1088 let language = language.clone();
1089 let container_dir: Arc<Path> = Arc::from(download_dir.join(adapter.name.0.as_ref()));
1090 let root_path = root_path.clone();
1091 let adapter = adapter.clone();
1092 let login_shell_env_loaded = self.login_shell_env_loaded.clone();
1093 let lsp_binary_statuses = self.lsp_binary_status_tx.clone();
1094
1095 let task = {
1096 let container_dir = container_dir.clone();
1097 cx.spawn(move |mut cx| async move {
1098 login_shell_env_loaded.await;
1099
1100 let entry = this
1101 .lsp_binary_paths
1102 .lock()
1103 .entry(adapter.name.clone())
1104 .or_insert_with(|| {
1105 let adapter = adapter.clone();
1106 let language = language.clone();
1107 let delegate = delegate.clone();
1108 cx.spawn(|cx| {
1109 get_binary(
1110 adapter,
1111 language,
1112 delegate,
1113 container_dir,
1114 lsp_binary_statuses,
1115 cx,
1116 )
1117 .map_err(Arc::new)
1118 })
1119 .shared()
1120 })
1121 .clone();
1122
1123 let binary = match entry.await {
1124 Ok(binary) => binary,
1125 Err(err) => anyhow::bail!("{err}"),
1126 };
1127
1128 if let Some(task) = adapter.will_start_server(&delegate, &mut cx) {
1129 task.await?;
1130 }
1131
1132 lsp::LanguageServer::new(
1133 stderr_capture,
1134 server_id,
1135 binary,
1136 &root_path,
1137 adapter.code_action_kinds(),
1138 cx,
1139 )
1140 })
1141 };
1142
1143 Some(PendingLanguageServer {
1144 server_id,
1145 task,
1146 container_dir: Some(container_dir),
1147 })
1148 }
1149
1150 pub fn language_server_binary_statuses(
1151 &self,
1152 ) -> mpsc::UnboundedReceiver<(Arc<Language>, LanguageServerBinaryStatus)> {
1153 self.lsp_binary_status_tx.subscribe()
1154 }
1155
1156 pub fn delete_server_container(
1157 &self,
1158 adapter: Arc<CachedLspAdapter>,
1159 cx: &mut AppContext,
1160 ) -> Task<()> {
1161 log::info!("deleting server container");
1162
1163 let mut lock = self.lsp_binary_paths.lock();
1164 lock.remove(&adapter.name);
1165
1166 let download_dir = self
1167 .language_server_download_dir
1168 .clone()
1169 .expect("language server download directory has not been assigned before deleting server container");
1170
1171 cx.spawn(|_| async move {
1172 let container_dir = download_dir.join(adapter.name.0.as_ref());
1173 smol::fs::remove_dir_all(container_dir)
1174 .await
1175 .context("server container removal")
1176 .log_err();
1177 })
1178 }
1179
1180 pub fn next_language_server_id(&self) -> LanguageServerId {
1181 self.state.write().next_language_server_id()
1182 }
1183}
1184
1185impl LanguageRegistryState {
1186 fn next_language_server_id(&mut self) -> LanguageServerId {
1187 LanguageServerId(post_inc(&mut self.next_language_server_id))
1188 }
1189
1190 fn add(&mut self, language: Arc<Language>) {
1191 if let Some(theme) = self.theme.as_ref() {
1192 language.set_theme(theme.syntax());
1193 }
1194 self.languages.push(language);
1195 self.version += 1;
1196 *self.subscription.0.borrow_mut() = ();
1197 }
1198
1199 fn reload(&mut self) {
1200 self.languages.clear();
1201 self.version += 1;
1202 self.reload_count += 1;
1203 for language in &mut self.available_languages {
1204 language.loaded = false;
1205 }
1206 *self.subscription.0.borrow_mut() = ();
1207 }
1208
1209 /// Mark the given language a having been loaded, so that the
1210 /// language registry won't try to load it again.
1211 fn mark_language_loaded(&mut self, id: AvailableLanguageId) {
1212 for language in &mut self.available_languages {
1213 if language.id == id {
1214 language.loaded = true;
1215 break;
1216 }
1217 }
1218 }
1219}
1220
1221#[cfg(any(test, feature = "test-support"))]
1222impl Default for LanguageRegistry {
1223 fn default() -> Self {
1224 Self::test()
1225 }
1226}
1227
1228async fn get_binary(
1229 adapter: Arc<CachedLspAdapter>,
1230 language: Arc<Language>,
1231 delegate: Arc<dyn LspAdapterDelegate>,
1232 container_dir: Arc<Path>,
1233 statuses: LspBinaryStatusSender,
1234 mut cx: AsyncAppContext,
1235) -> Result<LanguageServerBinary> {
1236 if !container_dir.exists() {
1237 smol::fs::create_dir_all(&container_dir)
1238 .await
1239 .context("failed to create container directory")?;
1240 }
1241
1242 if let Some(task) = adapter.will_fetch_server(&delegate, &mut cx) {
1243 task.await?;
1244 }
1245
1246 let binary = fetch_latest_binary(
1247 adapter.clone(),
1248 language.clone(),
1249 delegate.as_ref(),
1250 &container_dir,
1251 statuses.clone(),
1252 )
1253 .await;
1254
1255 if let Err(error) = binary.as_ref() {
1256 if let Some(binary) = adapter
1257 .cached_server_binary(container_dir.to_path_buf(), delegate.as_ref())
1258 .await
1259 {
1260 statuses.send(language.clone(), LanguageServerBinaryStatus::Cached);
1261 return Ok(binary);
1262 } else {
1263 statuses.send(
1264 language.clone(),
1265 LanguageServerBinaryStatus::Failed {
1266 error: format!("{:?}", error),
1267 },
1268 );
1269 }
1270 }
1271
1272 binary
1273}
1274
1275async fn fetch_latest_binary(
1276 adapter: Arc<CachedLspAdapter>,
1277 language: Arc<Language>,
1278 delegate: &dyn LspAdapterDelegate,
1279 container_dir: &Path,
1280 lsp_binary_statuses_tx: LspBinaryStatusSender,
1281) -> Result<LanguageServerBinary> {
1282 let container_dir: Arc<Path> = container_dir.into();
1283 lsp_binary_statuses_tx.send(
1284 language.clone(),
1285 LanguageServerBinaryStatus::CheckingForUpdate,
1286 );
1287
1288 let version_info = adapter.fetch_latest_server_version(delegate).await?;
1289 lsp_binary_statuses_tx.send(language.clone(), LanguageServerBinaryStatus::Downloading);
1290
1291 let binary = adapter
1292 .fetch_server_binary(version_info, container_dir.to_path_buf(), delegate)
1293 .await?;
1294 lsp_binary_statuses_tx.send(language.clone(), LanguageServerBinaryStatus::Downloaded);
1295
1296 Ok(binary)
1297}
1298
1299impl Language {
1300 pub fn new(config: LanguageConfig, ts_language: Option<tree_sitter::Language>) -> Self {
1301 Self {
1302 config,
1303 grammar: ts_language.map(|ts_language| {
1304 Arc::new(Grammar {
1305 id: NEXT_GRAMMAR_ID.fetch_add(1, SeqCst),
1306 highlights_query: None,
1307 brackets_config: None,
1308 outline_config: None,
1309 embedding_config: None,
1310 indents_config: None,
1311 injection_config: None,
1312 override_config: None,
1313 redactions_config: None,
1314 error_query: Query::new(&ts_language, "(ERROR) @error").unwrap(),
1315 ts_language,
1316 highlight_map: Default::default(),
1317 })
1318 }),
1319 adapters: Vec::new(),
1320
1321 #[cfg(any(test, feature = "test-support"))]
1322 fake_adapter: None,
1323 }
1324 }
1325
1326 pub fn lsp_adapters(&self) -> &[Arc<CachedLspAdapter>] {
1327 &self.adapters
1328 }
1329
1330 pub fn id(&self) -> Option<usize> {
1331 self.grammar.as_ref().map(|g| g.id)
1332 }
1333
1334 pub fn with_queries(mut self, queries: LanguageQueries) -> Result<Self> {
1335 if let Some(query) = queries.highlights {
1336 self = self
1337 .with_highlights_query(query.as_ref())
1338 .context("Error loading highlights query")?;
1339 }
1340 if let Some(query) = queries.brackets {
1341 self = self
1342 .with_brackets_query(query.as_ref())
1343 .context("Error loading brackets query")?;
1344 }
1345 if let Some(query) = queries.indents {
1346 self = self
1347 .with_indents_query(query.as_ref())
1348 .context("Error loading indents query")?;
1349 }
1350 if let Some(query) = queries.outline {
1351 self = self
1352 .with_outline_query(query.as_ref())
1353 .context("Error loading outline query")?;
1354 }
1355 if let Some(query) = queries.embedding {
1356 self = self
1357 .with_embedding_query(query.as_ref())
1358 .context("Error loading embedding query")?;
1359 }
1360 if let Some(query) = queries.injections {
1361 self = self
1362 .with_injection_query(query.as_ref())
1363 .context("Error loading injection query")?;
1364 }
1365 if let Some(query) = queries.overrides {
1366 self = self
1367 .with_override_query(query.as_ref())
1368 .context("Error loading override query")?;
1369 }
1370 if let Some(query) = queries.redactions {
1371 self = self
1372 .with_redaction_query(query.as_ref())
1373 .context("Error loading redaction query")?;
1374 }
1375 Ok(self)
1376 }
1377
1378 pub fn with_highlights_query(mut self, source: &str) -> Result<Self> {
1379 let grammar = self.grammar_mut();
1380 grammar.highlights_query = Some(Query::new(&grammar.ts_language, source)?);
1381 Ok(self)
1382 }
1383
1384 pub fn with_outline_query(mut self, source: &str) -> Result<Self> {
1385 let grammar = self.grammar_mut();
1386 let query = Query::new(&grammar.ts_language, source)?;
1387 let mut item_capture_ix = None;
1388 let mut name_capture_ix = None;
1389 let mut context_capture_ix = None;
1390 let mut extra_context_capture_ix = None;
1391 get_capture_indices(
1392 &query,
1393 &mut [
1394 ("item", &mut item_capture_ix),
1395 ("name", &mut name_capture_ix),
1396 ("context", &mut context_capture_ix),
1397 ("context.extra", &mut extra_context_capture_ix),
1398 ],
1399 );
1400 if let Some((item_capture_ix, name_capture_ix)) = item_capture_ix.zip(name_capture_ix) {
1401 grammar.outline_config = Some(OutlineConfig {
1402 query,
1403 item_capture_ix,
1404 name_capture_ix,
1405 context_capture_ix,
1406 extra_context_capture_ix,
1407 });
1408 }
1409 Ok(self)
1410 }
1411
1412 pub fn with_embedding_query(mut self, source: &str) -> Result<Self> {
1413 let grammar = self.grammar_mut();
1414 let query = Query::new(&grammar.ts_language, source)?;
1415 let mut item_capture_ix = None;
1416 let mut name_capture_ix = None;
1417 let mut context_capture_ix = None;
1418 let mut collapse_capture_ix = None;
1419 let mut keep_capture_ix = None;
1420 get_capture_indices(
1421 &query,
1422 &mut [
1423 ("item", &mut item_capture_ix),
1424 ("name", &mut name_capture_ix),
1425 ("context", &mut context_capture_ix),
1426 ("keep", &mut keep_capture_ix),
1427 ("collapse", &mut collapse_capture_ix),
1428 ],
1429 );
1430 if let Some(item_capture_ix) = item_capture_ix {
1431 grammar.embedding_config = Some(EmbeddingConfig {
1432 query,
1433 item_capture_ix,
1434 name_capture_ix,
1435 context_capture_ix,
1436 collapse_capture_ix,
1437 keep_capture_ix,
1438 });
1439 }
1440 Ok(self)
1441 }
1442
1443 pub fn with_brackets_query(mut self, source: &str) -> Result<Self> {
1444 let grammar = self.grammar_mut();
1445 let query = Query::new(&grammar.ts_language, source)?;
1446 let mut open_capture_ix = None;
1447 let mut close_capture_ix = None;
1448 get_capture_indices(
1449 &query,
1450 &mut [
1451 ("open", &mut open_capture_ix),
1452 ("close", &mut close_capture_ix),
1453 ],
1454 );
1455 if let Some((open_capture_ix, close_capture_ix)) = open_capture_ix.zip(close_capture_ix) {
1456 grammar.brackets_config = Some(BracketConfig {
1457 query,
1458 open_capture_ix,
1459 close_capture_ix,
1460 });
1461 }
1462 Ok(self)
1463 }
1464
1465 pub fn with_indents_query(mut self, source: &str) -> Result<Self> {
1466 let grammar = self.grammar_mut();
1467 let query = Query::new(&grammar.ts_language, source)?;
1468 let mut indent_capture_ix = None;
1469 let mut start_capture_ix = None;
1470 let mut end_capture_ix = None;
1471 let mut outdent_capture_ix = None;
1472 get_capture_indices(
1473 &query,
1474 &mut [
1475 ("indent", &mut indent_capture_ix),
1476 ("start", &mut start_capture_ix),
1477 ("end", &mut end_capture_ix),
1478 ("outdent", &mut outdent_capture_ix),
1479 ],
1480 );
1481 if let Some(indent_capture_ix) = indent_capture_ix {
1482 grammar.indents_config = Some(IndentConfig {
1483 query,
1484 indent_capture_ix,
1485 start_capture_ix,
1486 end_capture_ix,
1487 outdent_capture_ix,
1488 });
1489 }
1490 Ok(self)
1491 }
1492
1493 pub fn with_injection_query(mut self, source: &str) -> Result<Self> {
1494 let grammar = self.grammar_mut();
1495 let query = Query::new(&grammar.ts_language, source)?;
1496 let mut language_capture_ix = None;
1497 let mut content_capture_ix = None;
1498 get_capture_indices(
1499 &query,
1500 &mut [
1501 ("language", &mut language_capture_ix),
1502 ("content", &mut content_capture_ix),
1503 ],
1504 );
1505 let patterns = (0..query.pattern_count())
1506 .map(|ix| {
1507 let mut config = InjectionPatternConfig::default();
1508 for setting in query.property_settings(ix) {
1509 match setting.key.as_ref() {
1510 "language" => {
1511 config.language = setting.value.clone();
1512 }
1513 "combined" => {
1514 config.combined = true;
1515 }
1516 _ => {}
1517 }
1518 }
1519 config
1520 })
1521 .collect();
1522 if let Some(content_capture_ix) = content_capture_ix {
1523 grammar.injection_config = Some(InjectionConfig {
1524 query,
1525 language_capture_ix,
1526 content_capture_ix,
1527 patterns,
1528 });
1529 }
1530 Ok(self)
1531 }
1532
1533 pub fn with_override_query(mut self, source: &str) -> anyhow::Result<Self> {
1534 let query = Query::new(&self.grammar_mut().ts_language, source)?;
1535
1536 let mut override_configs_by_id = HashMap::default();
1537 for (ix, name) in query.capture_names().iter().enumerate() {
1538 if !name.starts_with('_') {
1539 let value = self.config.overrides.remove(*name).unwrap_or_default();
1540 for server_name in &value.opt_into_language_servers {
1541 if !self
1542 .config
1543 .scope_opt_in_language_servers
1544 .contains(server_name)
1545 {
1546 util::debug_panic!("Server {server_name:?} has been opted-in by scope {name:?} but has not been marked as an opt-in server");
1547 }
1548 }
1549
1550 override_configs_by_id.insert(ix as u32, (name.to_string(), value));
1551 }
1552 }
1553
1554 if !self.config.overrides.is_empty() {
1555 let keys = self.config.overrides.keys().collect::<Vec<_>>();
1556 Err(anyhow!(
1557 "language {:?} has overrides in config not in query: {keys:?}",
1558 self.config.name
1559 ))?;
1560 }
1561
1562 for disabled_scope_name in self
1563 .config
1564 .brackets
1565 .disabled_scopes_by_bracket_ix
1566 .iter()
1567 .flatten()
1568 {
1569 if !override_configs_by_id
1570 .values()
1571 .any(|(scope_name, _)| scope_name == disabled_scope_name)
1572 {
1573 Err(anyhow!(
1574 "language {:?} has overrides in config not in query: {disabled_scope_name:?}",
1575 self.config.name
1576 ))?;
1577 }
1578 }
1579
1580 for (name, override_config) in override_configs_by_id.values_mut() {
1581 override_config.disabled_bracket_ixs = self
1582 .config
1583 .brackets
1584 .disabled_scopes_by_bracket_ix
1585 .iter()
1586 .enumerate()
1587 .filter_map(|(ix, disabled_scope_names)| {
1588 if disabled_scope_names.contains(name) {
1589 Some(ix as u16)
1590 } else {
1591 None
1592 }
1593 })
1594 .collect();
1595 }
1596
1597 self.config.brackets.disabled_scopes_by_bracket_ix.clear();
1598 self.grammar_mut().override_config = Some(OverrideConfig {
1599 query,
1600 values: override_configs_by_id,
1601 });
1602 Ok(self)
1603 }
1604
1605 pub fn with_redaction_query(mut self, source: &str) -> anyhow::Result<Self> {
1606 let grammar = self.grammar_mut();
1607 let query = Query::new(&grammar.ts_language, source)?;
1608 let mut redaction_capture_ix = None;
1609 get_capture_indices(&query, &mut [("redact", &mut redaction_capture_ix)]);
1610
1611 if let Some(redaction_capture_ix) = redaction_capture_ix {
1612 grammar.redactions_config = Some(RedactionConfig {
1613 query,
1614 redaction_capture_ix,
1615 });
1616 }
1617
1618 Ok(self)
1619 }
1620
1621 fn grammar_mut(&mut self) -> &mut Grammar {
1622 Arc::get_mut(self.grammar.as_mut().unwrap()).unwrap()
1623 }
1624
1625 pub async fn with_lsp_adapters(mut self, lsp_adapters: Vec<Arc<dyn LspAdapter>>) -> Self {
1626 for adapter in lsp_adapters {
1627 self.adapters.push(CachedLspAdapter::new(adapter).await);
1628 }
1629 self
1630 }
1631
1632 #[cfg(any(test, feature = "test-support"))]
1633 pub async fn set_fake_lsp_adapter(
1634 &mut self,
1635 fake_lsp_adapter: Arc<FakeLspAdapter>,
1636 ) -> mpsc::UnboundedReceiver<lsp::FakeLanguageServer> {
1637 let (servers_tx, servers_rx) = mpsc::unbounded();
1638 self.fake_adapter = Some((servers_tx, fake_lsp_adapter.clone()));
1639 let adapter = CachedLspAdapter::new(Arc::new(fake_lsp_adapter)).await;
1640 self.adapters = vec![adapter];
1641 servers_rx
1642 }
1643
1644 pub fn name(&self) -> Arc<str> {
1645 self.config.name.clone()
1646 }
1647
1648 pub async fn disk_based_diagnostic_sources(&self) -> &[String] {
1649 match self.adapters.first().as_ref() {
1650 Some(adapter) => &adapter.disk_based_diagnostic_sources,
1651 None => &[],
1652 }
1653 }
1654
1655 pub async fn disk_based_diagnostics_progress_token(&self) -> Option<&str> {
1656 for adapter in &self.adapters {
1657 let token = adapter.disk_based_diagnostics_progress_token.as_deref();
1658 if token.is_some() {
1659 return token;
1660 }
1661 }
1662
1663 None
1664 }
1665
1666 pub async fn process_completion(self: &Arc<Self>, completion: &mut lsp::CompletionItem) {
1667 for adapter in &self.adapters {
1668 adapter.process_completion(completion).await;
1669 }
1670 }
1671
1672 pub async fn label_for_completion(
1673 self: &Arc<Self>,
1674 completion: &lsp::CompletionItem,
1675 ) -> Option<CodeLabel> {
1676 self.adapters
1677 .first()
1678 .as_ref()?
1679 .label_for_completion(completion, self)
1680 .await
1681 }
1682
1683 pub async fn label_for_symbol(
1684 self: &Arc<Self>,
1685 name: &str,
1686 kind: lsp::SymbolKind,
1687 ) -> Option<CodeLabel> {
1688 self.adapters
1689 .first()
1690 .as_ref()?
1691 .label_for_symbol(name, kind, self)
1692 .await
1693 }
1694
1695 pub fn highlight_text<'a>(
1696 self: &'a Arc<Self>,
1697 text: &'a Rope,
1698 range: Range<usize>,
1699 ) -> Vec<(Range<usize>, HighlightId)> {
1700 let mut result = Vec::new();
1701 if let Some(grammar) = &self.grammar {
1702 let tree = grammar.parse_text(text, None);
1703 let captures =
1704 SyntaxSnapshot::single_tree_captures(range.clone(), text, &tree, self, |grammar| {
1705 grammar.highlights_query.as_ref()
1706 });
1707 let highlight_maps = vec![grammar.highlight_map()];
1708 let mut offset = 0;
1709 for chunk in BufferChunks::new(text, range, Some((captures, highlight_maps)), vec![]) {
1710 let end_offset = offset + chunk.text.len();
1711 if let Some(highlight_id) = chunk.syntax_highlight_id {
1712 if !highlight_id.is_default() {
1713 result.push((offset..end_offset, highlight_id));
1714 }
1715 }
1716 offset = end_offset;
1717 }
1718 }
1719 result
1720 }
1721
1722 pub fn path_suffixes(&self) -> &[String] {
1723 &self.config.path_suffixes
1724 }
1725
1726 pub fn should_autoclose_before(&self, c: char) -> bool {
1727 c.is_whitespace() || self.config.autoclose_before.contains(c)
1728 }
1729
1730 pub fn set_theme(&self, theme: &SyntaxTheme) {
1731 if let Some(grammar) = self.grammar.as_ref() {
1732 if let Some(highlights_query) = &grammar.highlights_query {
1733 *grammar.highlight_map.lock() =
1734 HighlightMap::new(highlights_query.capture_names(), theme);
1735 }
1736 }
1737 }
1738
1739 pub fn grammar(&self) -> Option<&Arc<Grammar>> {
1740 self.grammar.as_ref()
1741 }
1742
1743 pub fn default_scope(self: &Arc<Self>) -> LanguageScope {
1744 LanguageScope {
1745 language: self.clone(),
1746 override_id: None,
1747 }
1748 }
1749
1750 pub fn prettier_parser_name(&self) -> Option<&str> {
1751 self.config.prettier_parser_name.as_deref()
1752 }
1753}
1754
1755impl LanguageScope {
1756 pub fn collapsed_placeholder(&self) -> &str {
1757 self.language.config.collapsed_placeholder.as_ref()
1758 }
1759
1760 /// Returns line prefix that is inserted in e.g. line continuations or
1761 /// in `toggle comments` action.
1762 pub fn line_comment_prefixes(&self) -> Option<&Vec<Arc<str>>> {
1763 Override::as_option(
1764 self.config_override().map(|o| &o.line_comments),
1765 Some(&self.language.config.line_comments),
1766 )
1767 }
1768
1769 pub fn block_comment_delimiters(&self) -> Option<(&Arc<str>, &Arc<str>)> {
1770 Override::as_option(
1771 self.config_override().map(|o| &o.block_comment),
1772 self.language.config.block_comment.as_ref(),
1773 )
1774 .map(|e| (&e.0, &e.1))
1775 }
1776
1777 /// Returns a list of language-specific word characters.
1778 ///
1779 /// By default, Zed treats alphanumeric characters (and '_') as word characters for
1780 /// the purpose of actions like 'move to next word end` or whole-word search.
1781 /// It additionally accounts for language's additional word characters.
1782 pub fn word_characters(&self) -> Option<&HashSet<char>> {
1783 Override::as_option(
1784 self.config_override().map(|o| &o.word_characters),
1785 Some(&self.language.config.word_characters),
1786 )
1787 }
1788
1789 /// Returns a list of bracket pairs for a given language with an additional
1790 /// piece of information about whether the particular bracket pair is currently active for a given language.
1791 pub fn brackets(&self) -> impl Iterator<Item = (&BracketPair, bool)> {
1792 let mut disabled_ids = self
1793 .config_override()
1794 .map_or(&[] as _, |o| o.disabled_bracket_ixs.as_slice());
1795 self.language
1796 .config
1797 .brackets
1798 .pairs
1799 .iter()
1800 .enumerate()
1801 .map(move |(ix, bracket)| {
1802 let mut is_enabled = true;
1803 if let Some(next_disabled_ix) = disabled_ids.first() {
1804 if ix == *next_disabled_ix as usize {
1805 disabled_ids = &disabled_ids[1..];
1806 is_enabled = false;
1807 }
1808 }
1809 (bracket, is_enabled)
1810 })
1811 }
1812
1813 pub fn should_autoclose_before(&self, c: char) -> bool {
1814 c.is_whitespace() || self.language.config.autoclose_before.contains(c)
1815 }
1816
1817 pub fn language_allowed(&self, name: &LanguageServerName) -> bool {
1818 let config = &self.language.config;
1819 let opt_in_servers = &config.scope_opt_in_language_servers;
1820 if opt_in_servers.iter().any(|o| *o == *name.0) {
1821 if let Some(over) = self.config_override() {
1822 over.opt_into_language_servers.iter().any(|o| *o == *name.0)
1823 } else {
1824 false
1825 }
1826 } else {
1827 true
1828 }
1829 }
1830
1831 fn config_override(&self) -> Option<&LanguageConfigOverride> {
1832 let id = self.override_id?;
1833 let grammar = self.language.grammar.as_ref()?;
1834 let override_config = grammar.override_config.as_ref()?;
1835 override_config.values.get(&id).map(|e| &e.1)
1836 }
1837}
1838
1839impl Hash for Language {
1840 fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
1841 self.id().hash(state)
1842 }
1843}
1844
1845impl PartialEq for Language {
1846 fn eq(&self, other: &Self) -> bool {
1847 self.id().eq(&other.id())
1848 }
1849}
1850
1851impl Eq for Language {}
1852
1853impl Debug for Language {
1854 fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
1855 f.debug_struct("Language")
1856 .field("name", &self.config.name)
1857 .finish()
1858 }
1859}
1860
1861impl Grammar {
1862 pub fn id(&self) -> usize {
1863 self.id
1864 }
1865
1866 fn parse_text(&self, text: &Rope, old_tree: Option<Tree>) -> Tree {
1867 PARSER.with(|parser| {
1868 let mut parser = parser.borrow_mut();
1869 parser
1870 .set_language(&self.ts_language)
1871 .expect("incompatible grammar");
1872 let mut chunks = text.chunks_in_range(0..text.len());
1873 parser
1874 .parse_with(
1875 &mut move |offset, _| {
1876 chunks.seek(offset);
1877 chunks.next().unwrap_or("").as_bytes()
1878 },
1879 old_tree.as_ref(),
1880 )
1881 .unwrap()
1882 })
1883 }
1884
1885 pub fn highlight_map(&self) -> HighlightMap {
1886 self.highlight_map.lock().clone()
1887 }
1888
1889 pub fn highlight_id_for_name(&self, name: &str) -> Option<HighlightId> {
1890 let capture_id = self
1891 .highlights_query
1892 .as_ref()?
1893 .capture_index_for_name(name)?;
1894 Some(self.highlight_map.lock().get(capture_id))
1895 }
1896}
1897
1898impl CodeLabel {
1899 pub fn plain(text: String, filter_text: Option<&str>) -> Self {
1900 let mut result = Self {
1901 runs: Vec::new(),
1902 filter_range: 0..text.len(),
1903 text,
1904 };
1905 if let Some(filter_text) = filter_text {
1906 if let Some(ix) = result.text.find(filter_text) {
1907 result.filter_range = ix..ix + filter_text.len();
1908 }
1909 }
1910 result
1911 }
1912}
1913
1914#[cfg(any(test, feature = "test-support"))]
1915impl Default for FakeLspAdapter {
1916 fn default() -> Self {
1917 Self {
1918 name: "the-fake-language-server",
1919 capabilities: lsp::LanguageServer::full_capabilities(),
1920 initializer: None,
1921 disk_based_diagnostics_progress_token: None,
1922 initialization_options: None,
1923 disk_based_diagnostics_sources: Vec::new(),
1924 prettier_plugins: Vec::new(),
1925 }
1926 }
1927}
1928
1929#[cfg(any(test, feature = "test-support"))]
1930#[async_trait]
1931impl LspAdapter for Arc<FakeLspAdapter> {
1932 fn name(&self) -> LanguageServerName {
1933 LanguageServerName(self.name.into())
1934 }
1935
1936 fn short_name(&self) -> &'static str {
1937 "FakeLspAdapter"
1938 }
1939
1940 async fn fetch_latest_server_version(
1941 &self,
1942 _: &dyn LspAdapterDelegate,
1943 ) -> Result<Box<dyn 'static + Send + Any>> {
1944 unreachable!();
1945 }
1946
1947 async fn fetch_server_binary(
1948 &self,
1949 _: Box<dyn 'static + Send + Any>,
1950 _: PathBuf,
1951 _: &dyn LspAdapterDelegate,
1952 ) -> Result<LanguageServerBinary> {
1953 unreachable!();
1954 }
1955
1956 async fn cached_server_binary(
1957 &self,
1958 _: PathBuf,
1959 _: &dyn LspAdapterDelegate,
1960 ) -> Option<LanguageServerBinary> {
1961 unreachable!();
1962 }
1963
1964 async fn installation_test_binary(&self, _: PathBuf) -> Option<LanguageServerBinary> {
1965 unreachable!();
1966 }
1967
1968 fn process_diagnostics(&self, _: &mut lsp::PublishDiagnosticsParams) {}
1969
1970 fn disk_based_diagnostic_sources(&self) -> Vec<String> {
1971 self.disk_based_diagnostics_sources.clone()
1972 }
1973
1974 fn disk_based_diagnostics_progress_token(&self) -> Option<String> {
1975 self.disk_based_diagnostics_progress_token.clone()
1976 }
1977
1978 fn initialization_options(&self) -> Option<Value> {
1979 self.initialization_options.clone()
1980 }
1981
1982 fn prettier_plugins(&self) -> &[&'static str] {
1983 &self.prettier_plugins
1984 }
1985}
1986
1987fn get_capture_indices(query: &Query, captures: &mut [(&str, &mut Option<u32>)]) {
1988 for (ix, name) in query.capture_names().iter().enumerate() {
1989 for (capture_name, index) in captures.iter_mut() {
1990 if capture_name == name {
1991 **index = Some(ix as u32);
1992 break;
1993 }
1994 }
1995 }
1996}
1997
1998pub fn point_to_lsp(point: PointUtf16) -> lsp::Position {
1999 lsp::Position::new(point.row, point.column)
2000}
2001
2002pub fn point_from_lsp(point: lsp::Position) -> Unclipped<PointUtf16> {
2003 Unclipped(PointUtf16::new(point.line, point.character))
2004}
2005
2006pub fn range_to_lsp(range: Range<PointUtf16>) -> lsp::Range {
2007 lsp::Range {
2008 start: point_to_lsp(range.start),
2009 end: point_to_lsp(range.end),
2010 }
2011}
2012
2013pub fn range_from_lsp(range: lsp::Range) -> Range<Unclipped<PointUtf16>> {
2014 let mut start = point_from_lsp(range.start);
2015 let mut end = point_from_lsp(range.end);
2016 if start > end {
2017 mem::swap(&mut start, &mut end);
2018 }
2019 start..end
2020}
2021
2022#[cfg(test)]
2023mod tests {
2024 use super::*;
2025 use gpui::TestAppContext;
2026
2027 #[gpui::test(iterations = 10)]
2028 async fn test_first_line_pattern(cx: &mut TestAppContext) {
2029 let mut languages = LanguageRegistry::test();
2030
2031 languages.set_executor(cx.executor());
2032 let languages = Arc::new(languages);
2033 languages.register(
2034 "/javascript",
2035 LanguageConfig {
2036 name: "JavaScript".into(),
2037 path_suffixes: vec!["js".into()],
2038 first_line_pattern: Some(Regex::new(r"\bnode\b").unwrap()),
2039 ..Default::default()
2040 },
2041 tree_sitter_typescript::language_tsx(),
2042 vec![],
2043 |_| Default::default(),
2044 );
2045
2046 languages
2047 .language_for_file("the/script", None)
2048 .await
2049 .unwrap_err();
2050 languages
2051 .language_for_file("the/script", Some(&"nothing".into()))
2052 .await
2053 .unwrap_err();
2054 assert_eq!(
2055 languages
2056 .language_for_file("the/script", Some(&"#!/bin/env node".into()))
2057 .await
2058 .unwrap()
2059 .name()
2060 .as_ref(),
2061 "JavaScript"
2062 );
2063 }
2064
2065 #[gpui::test(iterations = 10)]
2066 async fn test_language_loading(cx: &mut TestAppContext) {
2067 let mut languages = LanguageRegistry::test();
2068 languages.set_executor(cx.executor());
2069 let languages = Arc::new(languages);
2070 languages.register(
2071 "/JSON",
2072 LanguageConfig {
2073 name: "JSON".into(),
2074 path_suffixes: vec!["json".into()],
2075 ..Default::default()
2076 },
2077 tree_sitter_json::language(),
2078 vec![],
2079 |_| Default::default(),
2080 );
2081 languages.register(
2082 "/rust",
2083 LanguageConfig {
2084 name: "Rust".into(),
2085 path_suffixes: vec!["rs".into()],
2086 ..Default::default()
2087 },
2088 tree_sitter_rust::language(),
2089 vec![],
2090 |_| Default::default(),
2091 );
2092 assert_eq!(
2093 languages.language_names(),
2094 &[
2095 "JSON".to_string(),
2096 "Plain Text".to_string(),
2097 "Rust".to_string(),
2098 ]
2099 );
2100
2101 let rust1 = languages.language_for_name("Rust");
2102 let rust2 = languages.language_for_name("Rust");
2103
2104 // Ensure language is still listed even if it's being loaded.
2105 assert_eq!(
2106 languages.language_names(),
2107 &[
2108 "JSON".to_string(),
2109 "Plain Text".to_string(),
2110 "Rust".to_string(),
2111 ]
2112 );
2113
2114 let (rust1, rust2) = futures::join!(rust1, rust2);
2115 assert!(Arc::ptr_eq(&rust1.unwrap(), &rust2.unwrap()));
2116
2117 // Ensure language is still listed even after loading it.
2118 assert_eq!(
2119 languages.language_names(),
2120 &[
2121 "JSON".to_string(),
2122 "Plain Text".to_string(),
2123 "Rust".to_string(),
2124 ]
2125 );
2126
2127 // Loading an unknown language returns an error.
2128 assert!(languages.language_for_name("Unknown").await.is_err());
2129 }
2130}