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