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