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 its API.
9mod buffer;
10mod diagnostic_set;
11mod highlight_map;
12mod language_registry;
13pub mod language_settings;
14mod outline;
15pub mod proto;
16mod syntax_map;
17mod task_context;
18
19#[cfg(test)]
20pub mod buffer_tests;
21pub mod markdown;
22
23use crate::language_settings::SoftWrap;
24use anyhow::{anyhow, Context, Result};
25use async_trait::async_trait;
26use collections::{HashMap, HashSet};
27use futures::Future;
28use gpui::{AppContext, AsyncAppContext, Model, SharedString, Task};
29pub use highlight_map::HighlightMap;
30use http_client::HttpClient;
31pub use language_registry::LanguageName;
32use lsp::{CodeActionKind, LanguageServerBinary};
33use parking_lot::Mutex;
34use regex::Regex;
35use schemars::{
36 gen::SchemaGenerator,
37 schema::{InstanceType, Schema, SchemaObject},
38 JsonSchema,
39};
40use serde::{de, Deserialize, Deserializer, Serialize, Serializer};
41use serde_json::Value;
42use settings::WorktreeId;
43use smol::future::FutureExt as _;
44use std::num::NonZeroU32;
45use std::{
46 any::Any,
47 ffi::OsStr,
48 fmt::Debug,
49 hash::Hash,
50 mem,
51 ops::{DerefMut, Range},
52 path::{Path, PathBuf},
53 pin::Pin,
54 str,
55 sync::{
56 atomic::{AtomicU64, AtomicUsize, Ordering::SeqCst},
57 Arc, LazyLock,
58 },
59};
60use syntax_map::{QueryCursorHandle, SyntaxSnapshot};
61use task::RunnableTag;
62pub use task_context::{ContextProvider, RunnableRange};
63use theme::SyntaxTheme;
64use tree_sitter::{self, wasmtime, Query, QueryCursor, WasmStore};
65use util::serde::default_true;
66
67pub use buffer::Operation;
68pub use buffer::*;
69pub use diagnostic_set::DiagnosticEntry;
70pub use language_registry::{
71 AvailableLanguage, LanguageNotFound, LanguageQueries, LanguageRegistry,
72 LanguageServerBinaryStatus, PendingLanguageServer, QUERY_FILENAME_PREFIXES,
73};
74pub use lsp::LanguageServerId;
75pub use outline::*;
76pub use syntax_map::{OwnedSyntaxLayer, SyntaxLayer};
77pub use text::{AnchorRangeExt, LineEnding};
78pub use tree_sitter::{Node, Parser, Tree, TreeCursor};
79
80/// Initializes the `language` crate.
81///
82/// This should be called before making use of items from the create.
83pub fn init(cx: &mut AppContext) {
84 language_settings::init(cx);
85}
86
87static QUERY_CURSORS: Mutex<Vec<QueryCursor>> = Mutex::new(vec![]);
88static PARSERS: Mutex<Vec<Parser>> = Mutex::new(vec![]);
89
90pub fn with_parser<F, R>(func: F) -> R
91where
92 F: FnOnce(&mut Parser) -> R,
93{
94 let mut parser = PARSERS.lock().pop().unwrap_or_else(|| {
95 let mut parser = Parser::new();
96 parser
97 .set_wasm_store(WasmStore::new(&WASM_ENGINE).unwrap())
98 .unwrap();
99 parser
100 });
101 parser.set_included_ranges(&[]).unwrap();
102 let result = func(&mut parser);
103 PARSERS.lock().push(parser);
104 result
105}
106
107pub fn with_query_cursor<F, R>(func: F) -> R
108where
109 F: FnOnce(&mut QueryCursor) -> R,
110{
111 let mut cursor = QueryCursorHandle::new();
112 func(cursor.deref_mut())
113}
114
115static NEXT_LANGUAGE_ID: LazyLock<AtomicUsize> = LazyLock::new(Default::default);
116static NEXT_GRAMMAR_ID: LazyLock<AtomicUsize> = LazyLock::new(Default::default);
117static WASM_ENGINE: LazyLock<wasmtime::Engine> = LazyLock::new(|| {
118 wasmtime::Engine::new(&wasmtime::Config::new()).expect("Failed to create Wasmtime engine")
119});
120
121/// A shared grammar for plain text, exposed for reuse by downstream crates.
122pub static PLAIN_TEXT: LazyLock<Arc<Language>> = LazyLock::new(|| {
123 Arc::new(Language::new(
124 LanguageConfig {
125 name: "Plain Text".into(),
126 soft_wrap: Some(SoftWrap::EditorWidth),
127 ..Default::default()
128 },
129 None,
130 ))
131});
132
133/// Types that represent a position in a buffer, and can be converted into
134/// an LSP position, to send to a language server.
135pub trait ToLspPosition {
136 /// Converts the value into an LSP position.
137 fn to_lsp_position(self) -> lsp::Position;
138}
139
140/// A name of a language server.
141#[derive(Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Deserialize, Serialize)]
142pub struct LanguageServerName(pub SharedString);
143
144impl std::fmt::Display for LanguageServerName {
145 fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
146 std::fmt::Display::fmt(&self.0, f)
147 }
148}
149
150impl AsRef<str> for LanguageServerName {
151 fn as_ref(&self) -> &str {
152 self.0.as_ref()
153 }
154}
155
156impl AsRef<OsStr> for LanguageServerName {
157 fn as_ref(&self) -> &OsStr {
158 self.0.as_ref().as_ref()
159 }
160}
161
162impl JsonSchema for LanguageServerName {
163 fn schema_name() -> String {
164 "LanguageServerName".into()
165 }
166
167 fn json_schema(_: &mut SchemaGenerator) -> Schema {
168 SchemaObject {
169 instance_type: Some(InstanceType::String.into()),
170 ..Default::default()
171 }
172 .into()
173 }
174}
175impl LanguageServerName {
176 pub const fn new_static(s: &'static str) -> Self {
177 Self(SharedString::new_static(s))
178 }
179
180 pub fn from_proto(s: String) -> Self {
181 Self(s.into())
182 }
183}
184
185impl<'a> From<&'a str> for LanguageServerName {
186 fn from(str: &'a str) -> LanguageServerName {
187 LanguageServerName(str.to_string().into())
188 }
189}
190
191#[derive(Debug, Clone, PartialEq, Eq, Hash)]
192pub struct Location {
193 pub buffer: Model<Buffer>,
194 pub range: Range<Anchor>,
195}
196
197/// Represents a Language Server, with certain cached sync properties.
198/// Uses [`LspAdapter`] under the hood, but calls all 'static' methods
199/// once at startup, and caches the results.
200pub struct CachedLspAdapter {
201 pub name: LanguageServerName,
202 pub disk_based_diagnostic_sources: Vec<String>,
203 pub disk_based_diagnostics_progress_token: Option<String>,
204 language_ids: HashMap<String, String>,
205 pub adapter: Arc<dyn LspAdapter>,
206 pub reinstall_attempt_count: AtomicU64,
207 cached_binary: futures::lock::Mutex<Option<LanguageServerBinary>>,
208}
209
210impl Debug for CachedLspAdapter {
211 fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
212 f.debug_struct("CachedLspAdapter")
213 .field("name", &self.name)
214 .field(
215 "disk_based_diagnostic_sources",
216 &self.disk_based_diagnostic_sources,
217 )
218 .field(
219 "disk_based_diagnostics_progress_token",
220 &self.disk_based_diagnostics_progress_token,
221 )
222 .field("language_ids", &self.language_ids)
223 .field("reinstall_attempt_count", &self.reinstall_attempt_count)
224 .finish_non_exhaustive()
225 }
226}
227
228impl CachedLspAdapter {
229 pub fn new(adapter: Arc<dyn LspAdapter>) -> Arc<Self> {
230 let name = adapter.name();
231 let disk_based_diagnostic_sources = adapter.disk_based_diagnostic_sources();
232 let disk_based_diagnostics_progress_token = adapter.disk_based_diagnostics_progress_token();
233 let language_ids = adapter.language_ids();
234
235 Arc::new(CachedLspAdapter {
236 name,
237 disk_based_diagnostic_sources,
238 disk_based_diagnostics_progress_token,
239 language_ids,
240 adapter,
241 cached_binary: Default::default(),
242 reinstall_attempt_count: AtomicU64::new(0),
243 })
244 }
245
246 pub fn name(&self) -> LanguageServerName {
247 self.adapter.name().clone()
248 }
249
250 pub async fn get_language_server_command(
251 self: Arc<Self>,
252 container_dir: Option<Arc<Path>>,
253 delegate: Arc<dyn LspAdapterDelegate>,
254 cx: &mut AsyncAppContext,
255 ) -> Result<LanguageServerBinary> {
256 let cached_binary = self.cached_binary.lock().await;
257 self.adapter
258 .clone()
259 .get_language_server_command(container_dir, delegate, cached_binary, cx)
260 .await
261 }
262
263 pub fn can_be_reinstalled(&self) -> bool {
264 self.adapter.can_be_reinstalled()
265 }
266
267 pub async fn installation_test_binary(
268 &self,
269 container_dir: PathBuf,
270 ) -> Option<LanguageServerBinary> {
271 self.adapter.installation_test_binary(container_dir).await
272 }
273
274 pub fn code_action_kinds(&self) -> Option<Vec<CodeActionKind>> {
275 self.adapter.code_action_kinds()
276 }
277
278 pub fn process_diagnostics(&self, params: &mut lsp::PublishDiagnosticsParams) {
279 self.adapter.process_diagnostics(params)
280 }
281
282 pub async fn process_completions(&self, completion_items: &mut [lsp::CompletionItem]) {
283 self.adapter.process_completions(completion_items).await
284 }
285
286 pub async fn labels_for_completions(
287 &self,
288 completion_items: &[lsp::CompletionItem],
289 language: &Arc<Language>,
290 ) -> Result<Vec<Option<CodeLabel>>> {
291 self.adapter
292 .clone()
293 .labels_for_completions(completion_items, language)
294 .await
295 }
296
297 pub async fn labels_for_symbols(
298 &self,
299 symbols: &[(String, lsp::SymbolKind)],
300 language: &Arc<Language>,
301 ) -> Result<Vec<Option<CodeLabel>>> {
302 self.adapter
303 .clone()
304 .labels_for_symbols(symbols, language)
305 .await
306 }
307
308 pub fn language_id(&self, language_name: &LanguageName) -> String {
309 self.language_ids
310 .get(language_name.0.as_ref())
311 .cloned()
312 .unwrap_or_else(|| language_name.lsp_id())
313 }
314}
315
316/// [`LspAdapterDelegate`] allows [`LspAdapter]` implementations to interface with the application
317// e.g. to display a notification or fetch data from the web.
318#[async_trait]
319pub trait LspAdapterDelegate: Send + Sync {
320 fn show_notification(&self, message: &str, cx: &mut AppContext);
321 fn http_client(&self) -> Arc<dyn HttpClient>;
322 fn worktree_id(&self) -> WorktreeId;
323 fn worktree_root_path(&self) -> &Path;
324 fn update_status(&self, language: LanguageServerName, status: LanguageServerBinaryStatus);
325
326 async fn which(&self, command: &OsStr) -> Option<PathBuf>;
327 async fn shell_env(&self) -> HashMap<String, String>;
328 async fn read_text_file(&self, path: PathBuf) -> Result<String>;
329 async fn try_exec(&self, binary: LanguageServerBinary) -> Result<()>;
330}
331
332#[async_trait(?Send)]
333pub trait LspAdapter: 'static + Send + Sync {
334 fn name(&self) -> LanguageServerName;
335
336 fn get_language_server_command<'a>(
337 self: Arc<Self>,
338 container_dir: Option<Arc<Path>>,
339 delegate: Arc<dyn LspAdapterDelegate>,
340 mut cached_binary: futures::lock::MutexGuard<'a, Option<LanguageServerBinary>>,
341 cx: &'a mut AsyncAppContext,
342 ) -> Pin<Box<dyn 'a + Future<Output = Result<LanguageServerBinary>>>> {
343 async move {
344 // First we check whether the adapter can give us a user-installed binary.
345 // If so, we do *not* want to cache that, because each worktree might give us a different
346 // binary:
347 //
348 // worktree 1: user-installed at `.bin/gopls`
349 // worktree 2: user-installed at `~/bin/gopls`
350 // worktree 3: no gopls found in PATH -> fallback to Zed installation
351 //
352 // We only want to cache when we fall back to the global one,
353 // because we don't want to download and overwrite our global one
354 // for each worktree we might have open.
355 if let Some(binary) = self.check_if_user_installed(delegate.as_ref(), cx).await {
356 log::info!(
357 "found user-installed language server for {}. path: {:?}, arguments: {:?}",
358 self.name().0,
359 binary.path,
360 binary.arguments
361 );
362 return Ok(binary);
363 }
364
365 if let Some(cached_binary) = cached_binary.as_ref() {
366 return Ok(cached_binary.clone());
367 }
368
369 let Some(container_dir) = container_dir else {
370 anyhow::bail!("cannot download language servers for remotes (yet)")
371 };
372
373 if !container_dir.exists() {
374 smol::fs::create_dir_all(&container_dir)
375 .await
376 .context("failed to create container directory")?;
377 }
378
379 let mut binary = try_fetch_server_binary(self.as_ref(), &delegate, container_dir.to_path_buf(), cx).await;
380
381 if let Err(error) = binary.as_ref() {
382 if let Some(prev_downloaded_binary) = self
383 .cached_server_binary(container_dir.to_path_buf(), delegate.as_ref())
384 .await
385 {
386 log::info!(
387 "failed to fetch newest version of language server {:?}. falling back to using {:?}",
388 self.name(),
389 prev_downloaded_binary.path
390 );
391 binary = Ok(prev_downloaded_binary);
392 } else {
393 delegate.update_status(
394 self.name(),
395 LanguageServerBinaryStatus::Failed {
396 error: format!("{error:?}"),
397 },
398 );
399 }
400 }
401
402 if let Ok(binary) = &binary {
403 *cached_binary = Some(binary.clone());
404 }
405
406 binary
407 }
408 .boxed_local()
409 }
410
411 async fn check_if_user_installed(
412 &self,
413 _: &dyn LspAdapterDelegate,
414 _: &AsyncAppContext,
415 ) -> Option<LanguageServerBinary> {
416 None
417 }
418
419 async fn fetch_latest_server_version(
420 &self,
421 delegate: &dyn LspAdapterDelegate,
422 ) -> Result<Box<dyn 'static + Send + Any>>;
423
424 fn will_fetch_server(
425 &self,
426 _: &Arc<dyn LspAdapterDelegate>,
427 _: &mut AsyncAppContext,
428 ) -> Option<Task<Result<()>>> {
429 None
430 }
431
432 async fn fetch_server_binary(
433 &self,
434 latest_version: Box<dyn 'static + Send + Any>,
435 container_dir: PathBuf,
436 delegate: &dyn LspAdapterDelegate,
437 ) -> Result<LanguageServerBinary>;
438
439 async fn cached_server_binary(
440 &self,
441 container_dir: PathBuf,
442 delegate: &dyn LspAdapterDelegate,
443 ) -> Option<LanguageServerBinary>;
444
445 /// Returns `true` if a language server can be reinstalled.
446 ///
447 /// If language server initialization fails, a reinstallation will be attempted unless the value returned from this method is `false`.
448 ///
449 /// Implementations that rely on software already installed on user's system
450 /// should have [`can_be_reinstalled`](Self::can_be_reinstalled) return `false`.
451 fn can_be_reinstalled(&self) -> bool {
452 true
453 }
454
455 async fn installation_test_binary(
456 &self,
457 container_dir: PathBuf,
458 ) -> Option<LanguageServerBinary>;
459
460 fn process_diagnostics(&self, _: &mut lsp::PublishDiagnosticsParams) {}
461
462 /// Post-processes completions provided by the language server.
463 async fn process_completions(&self, _: &mut [lsp::CompletionItem]) {}
464
465 async fn labels_for_completions(
466 self: Arc<Self>,
467 completions: &[lsp::CompletionItem],
468 language: &Arc<Language>,
469 ) -> Result<Vec<Option<CodeLabel>>> {
470 let mut labels = Vec::new();
471 for (ix, completion) in completions.iter().enumerate() {
472 let label = self.label_for_completion(completion, language).await;
473 if let Some(label) = label {
474 labels.resize(ix + 1, None);
475 *labels.last_mut().unwrap() = Some(label);
476 }
477 }
478 Ok(labels)
479 }
480
481 async fn label_for_completion(
482 &self,
483 _: &lsp::CompletionItem,
484 _: &Arc<Language>,
485 ) -> Option<CodeLabel> {
486 None
487 }
488
489 async fn labels_for_symbols(
490 self: Arc<Self>,
491 symbols: &[(String, lsp::SymbolKind)],
492 language: &Arc<Language>,
493 ) -> Result<Vec<Option<CodeLabel>>> {
494 let mut labels = Vec::new();
495 for (ix, (name, kind)) in symbols.iter().enumerate() {
496 let label = self.label_for_symbol(name, *kind, language).await;
497 if let Some(label) = label {
498 labels.resize(ix + 1, None);
499 *labels.last_mut().unwrap() = Some(label);
500 }
501 }
502 Ok(labels)
503 }
504
505 async fn label_for_symbol(
506 &self,
507 _: &str,
508 _: lsp::SymbolKind,
509 _: &Arc<Language>,
510 ) -> Option<CodeLabel> {
511 None
512 }
513
514 /// Returns initialization options that are going to be sent to a LSP server as a part of [`lsp::InitializeParams`]
515 async fn initialization_options(
516 self: Arc<Self>,
517 _: &Arc<dyn LspAdapterDelegate>,
518 ) -> Result<Option<Value>> {
519 Ok(None)
520 }
521
522 async fn workspace_configuration(
523 self: Arc<Self>,
524 _: &Arc<dyn LspAdapterDelegate>,
525 _cx: &mut AsyncAppContext,
526 ) -> Result<Value> {
527 Ok(serde_json::json!({}))
528 }
529
530 /// Returns a list of code actions supported by a given LspAdapter
531 fn code_action_kinds(&self) -> Option<Vec<CodeActionKind>> {
532 Some(vec![
533 CodeActionKind::EMPTY,
534 CodeActionKind::QUICKFIX,
535 CodeActionKind::REFACTOR,
536 CodeActionKind::REFACTOR_EXTRACT,
537 CodeActionKind::SOURCE,
538 ])
539 }
540
541 fn disk_based_diagnostic_sources(&self) -> Vec<String> {
542 Default::default()
543 }
544
545 fn disk_based_diagnostics_progress_token(&self) -> Option<String> {
546 None
547 }
548
549 fn language_ids(&self) -> HashMap<String, String> {
550 Default::default()
551 }
552}
553
554async fn try_fetch_server_binary<L: LspAdapter + 'static + Send + Sync + ?Sized>(
555 adapter: &L,
556 delegate: &Arc<dyn LspAdapterDelegate>,
557 container_dir: PathBuf,
558 cx: &mut AsyncAppContext,
559) -> Result<LanguageServerBinary> {
560 if let Some(task) = adapter.will_fetch_server(delegate, cx) {
561 task.await?;
562 }
563
564 let name = adapter.name();
565 log::info!("fetching latest version of language server {:?}", name.0);
566 delegate.update_status(name.clone(), LanguageServerBinaryStatus::CheckingForUpdate);
567 let latest_version = adapter
568 .fetch_latest_server_version(delegate.as_ref())
569 .await?;
570
571 log::info!("downloading language server {:?}", name.0);
572 delegate.update_status(adapter.name(), LanguageServerBinaryStatus::Downloading);
573 let binary = adapter
574 .fetch_server_binary(latest_version, container_dir, delegate.as_ref())
575 .await;
576
577 delegate.update_status(name.clone(), LanguageServerBinaryStatus::None);
578 binary
579}
580
581#[derive(Clone, Debug, Default, PartialEq, Eq)]
582pub struct CodeLabel {
583 /// The text to display.
584 pub text: String,
585 /// Syntax highlighting runs.
586 pub runs: Vec<(Range<usize>, HighlightId)>,
587 /// The portion of the text that should be used in fuzzy filtering.
588 pub filter_range: Range<usize>,
589}
590
591#[derive(Clone, Deserialize, JsonSchema)]
592pub struct LanguageConfig {
593 /// Human-readable name of the language.
594 pub name: LanguageName,
595 /// The name of this language for a Markdown code fence block
596 pub code_fence_block_name: Option<Arc<str>>,
597 // The name of the grammar in a WASM bundle (experimental).
598 pub grammar: Option<Arc<str>>,
599 /// The criteria for matching this language to a given file.
600 #[serde(flatten)]
601 pub matcher: LanguageMatcher,
602 /// List of bracket types in a language.
603 #[serde(default)]
604 #[schemars(schema_with = "bracket_pair_config_json_schema")]
605 pub brackets: BracketPairConfig,
606 /// If set to true, auto indentation uses last non empty line to determine
607 /// the indentation level for a new line.
608 #[serde(default = "auto_indent_using_last_non_empty_line_default")]
609 pub auto_indent_using_last_non_empty_line: bool,
610 /// A regex that is used to determine whether the indentation level should be
611 /// increased in the following line.
612 #[serde(default, deserialize_with = "deserialize_regex")]
613 #[schemars(schema_with = "regex_json_schema")]
614 pub increase_indent_pattern: Option<Regex>,
615 /// A regex that is used to determine whether the indentation level should be
616 /// decreased in the following line.
617 #[serde(default, deserialize_with = "deserialize_regex")]
618 #[schemars(schema_with = "regex_json_schema")]
619 pub decrease_indent_pattern: Option<Regex>,
620 /// A list of characters that trigger the automatic insertion of a closing
621 /// bracket when they immediately precede the point where an opening
622 /// bracket is inserted.
623 #[serde(default)]
624 pub autoclose_before: String,
625 /// A placeholder used internally by Semantic Index.
626 #[serde(default)]
627 pub collapsed_placeholder: String,
628 /// A line comment string that is inserted in e.g. `toggle comments` action.
629 /// A language can have multiple flavours of line comments. All of the provided line comments are
630 /// used for comment continuations on the next line, but only the first one is used for Editor::ToggleComments.
631 #[serde(default)]
632 pub line_comments: Vec<Arc<str>>,
633 /// Starting and closing characters of a block comment.
634 #[serde(default)]
635 pub block_comment: Option<(Arc<str>, Arc<str>)>,
636 /// A list of language servers that are allowed to run on subranges of a given language.
637 #[serde(default)]
638 pub scope_opt_in_language_servers: Vec<LanguageServerName>,
639 #[serde(default)]
640 pub overrides: HashMap<String, LanguageConfigOverride>,
641 /// A list of characters that Zed should treat as word characters for the
642 /// purpose of features that operate on word boundaries, like 'move to next word end'
643 /// or a whole-word search in buffer search.
644 #[serde(default)]
645 pub word_characters: HashSet<char>,
646 /// Whether to indent lines using tab characters, as opposed to multiple
647 /// spaces.
648 #[serde(default)]
649 pub hard_tabs: Option<bool>,
650 /// How many columns a tab should occupy.
651 #[serde(default)]
652 pub tab_size: Option<NonZeroU32>,
653 /// How to soft-wrap long lines of text.
654 #[serde(default)]
655 pub soft_wrap: Option<SoftWrap>,
656 /// The name of a Prettier parser that will be used for this language when no file path is available.
657 /// If there's a parser name in the language settings, that will be used instead.
658 #[serde(default)]
659 pub prettier_parser_name: Option<String>,
660 /// If true, this language is only for syntax highlighting via an injection into other
661 /// languages, but should not appear to the user as a distinct language.
662 #[serde(default)]
663 pub hidden: bool,
664}
665
666#[derive(Clone, Debug, Serialize, Deserialize, Default, JsonSchema)]
667pub struct LanguageMatcher {
668 /// 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`.
669 #[serde(default)]
670 pub path_suffixes: Vec<String>,
671 /// A regex pattern that determines whether the language should be assigned to a file or not.
672 #[serde(
673 default,
674 serialize_with = "serialize_regex",
675 deserialize_with = "deserialize_regex"
676 )]
677 #[schemars(schema_with = "regex_json_schema")]
678 pub first_line_pattern: Option<Regex>,
679}
680
681/// Represents a language for the given range. Some languages (e.g. HTML)
682/// interleave several languages together, thus a single buffer might actually contain
683/// several nested scopes.
684#[derive(Clone, Debug)]
685pub struct LanguageScope {
686 language: Arc<Language>,
687 override_id: Option<u32>,
688}
689
690#[derive(Clone, Deserialize, Default, Debug, JsonSchema)]
691pub struct LanguageConfigOverride {
692 #[serde(default)]
693 pub line_comments: Override<Vec<Arc<str>>>,
694 #[serde(default)]
695 pub block_comment: Override<(Arc<str>, Arc<str>)>,
696 #[serde(skip_deserializing)]
697 #[schemars(skip)]
698 pub disabled_bracket_ixs: Vec<u16>,
699 #[serde(default)]
700 pub word_characters: Override<HashSet<char>>,
701 #[serde(default)]
702 pub opt_into_language_servers: Vec<LanguageServerName>,
703}
704
705#[derive(Clone, Deserialize, Debug, Serialize, JsonSchema)]
706#[serde(untagged)]
707pub enum Override<T> {
708 Remove { remove: bool },
709 Set(T),
710}
711
712impl<T> Default for Override<T> {
713 fn default() -> Self {
714 Override::Remove { remove: false }
715 }
716}
717
718impl<T> Override<T> {
719 fn as_option<'a>(this: Option<&'a Self>, original: Option<&'a T>) -> Option<&'a T> {
720 match this {
721 Some(Self::Set(value)) => Some(value),
722 Some(Self::Remove { remove: true }) => None,
723 Some(Self::Remove { remove: false }) | None => original,
724 }
725 }
726}
727
728impl Default for LanguageConfig {
729 fn default() -> Self {
730 Self {
731 name: LanguageName::new(""),
732 code_fence_block_name: None,
733 grammar: None,
734 matcher: LanguageMatcher::default(),
735 brackets: Default::default(),
736 auto_indent_using_last_non_empty_line: auto_indent_using_last_non_empty_line_default(),
737 increase_indent_pattern: Default::default(),
738 decrease_indent_pattern: Default::default(),
739 autoclose_before: Default::default(),
740 line_comments: Default::default(),
741 block_comment: Default::default(),
742 scope_opt_in_language_servers: Default::default(),
743 overrides: Default::default(),
744 word_characters: Default::default(),
745 collapsed_placeholder: Default::default(),
746 hard_tabs: None,
747 tab_size: None,
748 soft_wrap: None,
749 prettier_parser_name: None,
750 hidden: false,
751 }
752 }
753}
754
755fn auto_indent_using_last_non_empty_line_default() -> bool {
756 true
757}
758
759fn deserialize_regex<'de, D: Deserializer<'de>>(d: D) -> Result<Option<Regex>, D::Error> {
760 let source = Option::<String>::deserialize(d)?;
761 if let Some(source) = source {
762 Ok(Some(regex::Regex::new(&source).map_err(de::Error::custom)?))
763 } else {
764 Ok(None)
765 }
766}
767
768fn regex_json_schema(_: &mut SchemaGenerator) -> Schema {
769 Schema::Object(SchemaObject {
770 instance_type: Some(InstanceType::String.into()),
771 ..Default::default()
772 })
773}
774
775fn serialize_regex<S>(regex: &Option<Regex>, serializer: S) -> Result<S::Ok, S::Error>
776where
777 S: Serializer,
778{
779 match regex {
780 Some(regex) => serializer.serialize_str(regex.as_str()),
781 None => serializer.serialize_none(),
782 }
783}
784
785#[doc(hidden)]
786#[cfg(any(test, feature = "test-support"))]
787pub struct FakeLspAdapter {
788 pub name: &'static str,
789 pub initialization_options: Option<Value>,
790 pub prettier_plugins: Vec<&'static str>,
791 pub disk_based_diagnostics_progress_token: Option<String>,
792 pub disk_based_diagnostics_sources: Vec<String>,
793 pub language_server_binary: LanguageServerBinary,
794
795 pub capabilities: lsp::ServerCapabilities,
796 pub initializer: Option<Box<dyn 'static + Send + Sync + Fn(&mut lsp::FakeLanguageServer)>>,
797}
798
799/// Configuration of handling bracket pairs for a given language.
800///
801/// This struct includes settings for defining which pairs of characters are considered brackets and
802/// also specifies any language-specific scopes where these pairs should be ignored for bracket matching purposes.
803#[derive(Clone, Debug, Default, JsonSchema)]
804pub struct BracketPairConfig {
805 /// A list of character pairs that should be treated as brackets in the context of a given language.
806 pub pairs: Vec<BracketPair>,
807 /// A list of tree-sitter scopes for which a given bracket should not be active.
808 /// N-th entry in `[Self::disabled_scopes_by_bracket_ix]` contains a list of disabled scopes for an n-th entry in `[Self::pairs]`
809 #[schemars(skip)]
810 pub disabled_scopes_by_bracket_ix: Vec<Vec<String>>,
811}
812
813fn bracket_pair_config_json_schema(gen: &mut SchemaGenerator) -> Schema {
814 Option::<Vec<BracketPairContent>>::json_schema(gen)
815}
816
817#[derive(Deserialize, JsonSchema)]
818pub struct BracketPairContent {
819 #[serde(flatten)]
820 pub bracket_pair: BracketPair,
821 #[serde(default)]
822 pub not_in: Vec<String>,
823}
824
825impl<'de> Deserialize<'de> for BracketPairConfig {
826 fn deserialize<D>(deserializer: D) -> std::result::Result<Self, D::Error>
827 where
828 D: Deserializer<'de>,
829 {
830 let result = Vec::<BracketPairContent>::deserialize(deserializer)?;
831 let mut brackets = Vec::with_capacity(result.len());
832 let mut disabled_scopes_by_bracket_ix = Vec::with_capacity(result.len());
833 for entry in result {
834 brackets.push(entry.bracket_pair);
835 disabled_scopes_by_bracket_ix.push(entry.not_in);
836 }
837
838 Ok(BracketPairConfig {
839 pairs: brackets,
840 disabled_scopes_by_bracket_ix,
841 })
842 }
843}
844
845/// Describes a single bracket pair and how an editor should react to e.g. inserting
846/// an opening bracket or to a newline character insertion in between `start` and `end` characters.
847#[derive(Clone, Debug, Default, Deserialize, PartialEq, JsonSchema)]
848pub struct BracketPair {
849 /// Starting substring for a bracket.
850 pub start: String,
851 /// Ending substring for a bracket.
852 pub end: String,
853 /// True if `end` should be automatically inserted right after `start` characters.
854 pub close: bool,
855 /// True if selected text should be surrounded by `start` and `end` characters.
856 #[serde(default = "default_true")]
857 pub surround: bool,
858 /// True if an extra newline should be inserted while the cursor is in the middle
859 /// of that bracket pair.
860 pub newline: bool,
861}
862
863#[derive(Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Clone, Copy)]
864pub(crate) struct LanguageId(usize);
865
866impl LanguageId {
867 pub(crate) fn new() -> Self {
868 Self(NEXT_LANGUAGE_ID.fetch_add(1, SeqCst))
869 }
870}
871
872pub struct Language {
873 pub(crate) id: LanguageId,
874 pub(crate) config: LanguageConfig,
875 pub(crate) grammar: Option<Arc<Grammar>>,
876 pub(crate) context_provider: Option<Arc<dyn ContextProvider>>,
877}
878
879#[derive(Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Clone, Copy)]
880pub struct GrammarId(pub usize);
881
882impl GrammarId {
883 pub(crate) fn new() -> Self {
884 Self(NEXT_GRAMMAR_ID.fetch_add(1, SeqCst))
885 }
886}
887
888pub struct Grammar {
889 id: GrammarId,
890 pub ts_language: tree_sitter::Language,
891 pub(crate) error_query: Query,
892 pub(crate) highlights_query: Option<Query>,
893 pub(crate) brackets_config: Option<BracketConfig>,
894 pub(crate) redactions_config: Option<RedactionConfig>,
895 pub(crate) runnable_config: Option<RunnableConfig>,
896 pub(crate) indents_config: Option<IndentConfig>,
897 pub outline_config: Option<OutlineConfig>,
898 pub embedding_config: Option<EmbeddingConfig>,
899 pub(crate) injection_config: Option<InjectionConfig>,
900 pub(crate) override_config: Option<OverrideConfig>,
901 pub(crate) highlight_map: Mutex<HighlightMap>,
902}
903
904struct IndentConfig {
905 query: Query,
906 indent_capture_ix: u32,
907 start_capture_ix: Option<u32>,
908 end_capture_ix: Option<u32>,
909 outdent_capture_ix: Option<u32>,
910}
911
912pub struct OutlineConfig {
913 pub query: Query,
914 pub item_capture_ix: u32,
915 pub name_capture_ix: u32,
916 pub context_capture_ix: Option<u32>,
917 pub extra_context_capture_ix: Option<u32>,
918 pub open_capture_ix: Option<u32>,
919 pub close_capture_ix: Option<u32>,
920 pub annotation_capture_ix: Option<u32>,
921}
922
923#[derive(Debug)]
924pub struct EmbeddingConfig {
925 pub query: Query,
926 pub item_capture_ix: u32,
927 pub name_capture_ix: Option<u32>,
928 pub context_capture_ix: Option<u32>,
929 pub collapse_capture_ix: Option<u32>,
930 pub keep_capture_ix: Option<u32>,
931}
932
933struct InjectionConfig {
934 query: Query,
935 content_capture_ix: u32,
936 language_capture_ix: Option<u32>,
937 patterns: Vec<InjectionPatternConfig>,
938}
939
940struct RedactionConfig {
941 pub query: Query,
942 pub redaction_capture_ix: u32,
943}
944
945#[derive(Clone, Debug, PartialEq)]
946enum RunnableCapture {
947 Named(SharedString),
948 Run,
949}
950
951struct RunnableConfig {
952 pub query: Query,
953 /// A mapping from capture indice to capture kind
954 pub extra_captures: Vec<RunnableCapture>,
955}
956
957struct OverrideConfig {
958 query: Query,
959 values: HashMap<u32, (String, LanguageConfigOverride)>,
960}
961
962#[derive(Default, Clone)]
963struct InjectionPatternConfig {
964 language: Option<Box<str>>,
965 combined: bool,
966}
967
968struct BracketConfig {
969 query: Query,
970 open_capture_ix: u32,
971 close_capture_ix: u32,
972}
973
974impl Language {
975 pub fn new(config: LanguageConfig, ts_language: Option<tree_sitter::Language>) -> Self {
976 Self::new_with_id(LanguageId::new(), config, ts_language)
977 }
978
979 fn new_with_id(
980 id: LanguageId,
981 config: LanguageConfig,
982 ts_language: Option<tree_sitter::Language>,
983 ) -> Self {
984 Self {
985 id,
986 config,
987 grammar: ts_language.map(|ts_language| {
988 Arc::new(Grammar {
989 id: GrammarId::new(),
990 highlights_query: None,
991 brackets_config: None,
992 outline_config: None,
993 embedding_config: None,
994 indents_config: None,
995 injection_config: None,
996 override_config: None,
997 redactions_config: None,
998 runnable_config: None,
999 error_query: Query::new(&ts_language, "(ERROR) @error").unwrap(),
1000 ts_language,
1001 highlight_map: Default::default(),
1002 })
1003 }),
1004 context_provider: None,
1005 }
1006 }
1007
1008 pub fn with_context_provider(mut self, provider: Option<Arc<dyn ContextProvider>>) -> Self {
1009 self.context_provider = provider;
1010 self
1011 }
1012
1013 pub fn with_queries(mut self, queries: LanguageQueries) -> Result<Self> {
1014 if let Some(query) = queries.highlights {
1015 self = self
1016 .with_highlights_query(query.as_ref())
1017 .context("Error loading highlights query")?;
1018 }
1019 if let Some(query) = queries.brackets {
1020 self = self
1021 .with_brackets_query(query.as_ref())
1022 .context("Error loading brackets query")?;
1023 }
1024 if let Some(query) = queries.indents {
1025 self = self
1026 .with_indents_query(query.as_ref())
1027 .context("Error loading indents query")?;
1028 }
1029 if let Some(query) = queries.outline {
1030 self = self
1031 .with_outline_query(query.as_ref())
1032 .context("Error loading outline query")?;
1033 }
1034 if let Some(query) = queries.embedding {
1035 self = self
1036 .with_embedding_query(query.as_ref())
1037 .context("Error loading embedding query")?;
1038 }
1039 if let Some(query) = queries.injections {
1040 self = self
1041 .with_injection_query(query.as_ref())
1042 .context("Error loading injection query")?;
1043 }
1044 if let Some(query) = queries.overrides {
1045 self = self
1046 .with_override_query(query.as_ref())
1047 .context("Error loading override query")?;
1048 }
1049 if let Some(query) = queries.redactions {
1050 self = self
1051 .with_redaction_query(query.as_ref())
1052 .context("Error loading redaction query")?;
1053 }
1054 if let Some(query) = queries.runnables {
1055 self = self
1056 .with_runnable_query(query.as_ref())
1057 .context("Error loading tests query")?;
1058 }
1059 Ok(self)
1060 }
1061
1062 pub fn with_highlights_query(mut self, source: &str) -> Result<Self> {
1063 let grammar = self
1064 .grammar_mut()
1065 .ok_or_else(|| anyhow!("cannot mutate grammar"))?;
1066 grammar.highlights_query = Some(Query::new(&grammar.ts_language, source)?);
1067 Ok(self)
1068 }
1069
1070 pub fn with_runnable_query(mut self, source: &str) -> Result<Self> {
1071 let grammar = self
1072 .grammar_mut()
1073 .ok_or_else(|| anyhow!("cannot mutate grammar"))?;
1074
1075 let query = Query::new(&grammar.ts_language, source)?;
1076 let mut extra_captures = Vec::with_capacity(query.capture_names().len());
1077
1078 for name in query.capture_names().iter() {
1079 let kind = if *name == "run" {
1080 RunnableCapture::Run
1081 } else {
1082 RunnableCapture::Named(name.to_string().into())
1083 };
1084 extra_captures.push(kind);
1085 }
1086
1087 grammar.runnable_config = Some(RunnableConfig {
1088 extra_captures,
1089 query,
1090 });
1091
1092 Ok(self)
1093 }
1094
1095 pub fn with_outline_query(mut self, source: &str) -> Result<Self> {
1096 let grammar = self
1097 .grammar_mut()
1098 .ok_or_else(|| anyhow!("cannot mutate grammar"))?;
1099 let query = Query::new(&grammar.ts_language, source)?;
1100 let mut item_capture_ix = None;
1101 let mut name_capture_ix = None;
1102 let mut context_capture_ix = None;
1103 let mut extra_context_capture_ix = None;
1104 let mut open_capture_ix = None;
1105 let mut close_capture_ix = None;
1106 let mut annotation_capture_ix = None;
1107 get_capture_indices(
1108 &query,
1109 &mut [
1110 ("item", &mut item_capture_ix),
1111 ("name", &mut name_capture_ix),
1112 ("context", &mut context_capture_ix),
1113 ("context.extra", &mut extra_context_capture_ix),
1114 ("open", &mut open_capture_ix),
1115 ("close", &mut close_capture_ix),
1116 ("annotation", &mut annotation_capture_ix),
1117 ],
1118 );
1119 if let Some((item_capture_ix, name_capture_ix)) = item_capture_ix.zip(name_capture_ix) {
1120 grammar.outline_config = Some(OutlineConfig {
1121 query,
1122 item_capture_ix,
1123 name_capture_ix,
1124 context_capture_ix,
1125 extra_context_capture_ix,
1126 open_capture_ix,
1127 close_capture_ix,
1128 annotation_capture_ix,
1129 });
1130 }
1131 Ok(self)
1132 }
1133
1134 pub fn with_embedding_query(mut self, source: &str) -> Result<Self> {
1135 let grammar = self
1136 .grammar_mut()
1137 .ok_or_else(|| anyhow!("cannot mutate grammar"))?;
1138 let query = Query::new(&grammar.ts_language, source)?;
1139 let mut item_capture_ix = None;
1140 let mut name_capture_ix = None;
1141 let mut context_capture_ix = None;
1142 let mut collapse_capture_ix = None;
1143 let mut keep_capture_ix = None;
1144 get_capture_indices(
1145 &query,
1146 &mut [
1147 ("item", &mut item_capture_ix),
1148 ("name", &mut name_capture_ix),
1149 ("context", &mut context_capture_ix),
1150 ("keep", &mut keep_capture_ix),
1151 ("collapse", &mut collapse_capture_ix),
1152 ],
1153 );
1154 if let Some(item_capture_ix) = item_capture_ix {
1155 grammar.embedding_config = Some(EmbeddingConfig {
1156 query,
1157 item_capture_ix,
1158 name_capture_ix,
1159 context_capture_ix,
1160 collapse_capture_ix,
1161 keep_capture_ix,
1162 });
1163 }
1164 Ok(self)
1165 }
1166
1167 pub fn with_brackets_query(mut self, source: &str) -> Result<Self> {
1168 let grammar = self
1169 .grammar_mut()
1170 .ok_or_else(|| anyhow!("cannot mutate grammar"))?;
1171 let query = Query::new(&grammar.ts_language, source)?;
1172 let mut open_capture_ix = None;
1173 let mut close_capture_ix = None;
1174 get_capture_indices(
1175 &query,
1176 &mut [
1177 ("open", &mut open_capture_ix),
1178 ("close", &mut close_capture_ix),
1179 ],
1180 );
1181 if let Some((open_capture_ix, close_capture_ix)) = open_capture_ix.zip(close_capture_ix) {
1182 grammar.brackets_config = Some(BracketConfig {
1183 query,
1184 open_capture_ix,
1185 close_capture_ix,
1186 });
1187 }
1188 Ok(self)
1189 }
1190
1191 pub fn with_indents_query(mut self, source: &str) -> Result<Self> {
1192 let grammar = self
1193 .grammar_mut()
1194 .ok_or_else(|| anyhow!("cannot mutate grammar"))?;
1195 let query = Query::new(&grammar.ts_language, source)?;
1196 let mut indent_capture_ix = None;
1197 let mut start_capture_ix = None;
1198 let mut end_capture_ix = None;
1199 let mut outdent_capture_ix = None;
1200 get_capture_indices(
1201 &query,
1202 &mut [
1203 ("indent", &mut indent_capture_ix),
1204 ("start", &mut start_capture_ix),
1205 ("end", &mut end_capture_ix),
1206 ("outdent", &mut outdent_capture_ix),
1207 ],
1208 );
1209 if let Some(indent_capture_ix) = indent_capture_ix {
1210 grammar.indents_config = Some(IndentConfig {
1211 query,
1212 indent_capture_ix,
1213 start_capture_ix,
1214 end_capture_ix,
1215 outdent_capture_ix,
1216 });
1217 }
1218 Ok(self)
1219 }
1220
1221 pub fn with_injection_query(mut self, source: &str) -> Result<Self> {
1222 let grammar = self
1223 .grammar_mut()
1224 .ok_or_else(|| anyhow!("cannot mutate grammar"))?;
1225 let query = Query::new(&grammar.ts_language, source)?;
1226 let mut language_capture_ix = None;
1227 let mut content_capture_ix = None;
1228 get_capture_indices(
1229 &query,
1230 &mut [
1231 ("language", &mut language_capture_ix),
1232 ("content", &mut content_capture_ix),
1233 ],
1234 );
1235 let patterns = (0..query.pattern_count())
1236 .map(|ix| {
1237 let mut config = InjectionPatternConfig::default();
1238 for setting in query.property_settings(ix) {
1239 match setting.key.as_ref() {
1240 "language" => {
1241 config.language.clone_from(&setting.value);
1242 }
1243 "combined" => {
1244 config.combined = true;
1245 }
1246 _ => {}
1247 }
1248 }
1249 config
1250 })
1251 .collect();
1252 if let Some(content_capture_ix) = content_capture_ix {
1253 grammar.injection_config = Some(InjectionConfig {
1254 query,
1255 language_capture_ix,
1256 content_capture_ix,
1257 patterns,
1258 });
1259 }
1260 Ok(self)
1261 }
1262
1263 pub fn with_override_query(mut self, source: &str) -> anyhow::Result<Self> {
1264 let query = {
1265 let grammar = self
1266 .grammar
1267 .as_ref()
1268 .ok_or_else(|| anyhow!("no grammar for language"))?;
1269 Query::new(&grammar.ts_language, source)?
1270 };
1271
1272 let mut override_configs_by_id = HashMap::default();
1273 for (ix, name) in query.capture_names().iter().enumerate() {
1274 if !name.starts_with('_') {
1275 let value = self.config.overrides.remove(*name).unwrap_or_default();
1276 for server_name in &value.opt_into_language_servers {
1277 if !self
1278 .config
1279 .scope_opt_in_language_servers
1280 .contains(server_name)
1281 {
1282 util::debug_panic!("Server {server_name:?} has been opted-in by scope {name:?} but has not been marked as an opt-in server");
1283 }
1284 }
1285
1286 override_configs_by_id.insert(ix as u32, (name.to_string(), value));
1287 }
1288 }
1289
1290 if !self.config.overrides.is_empty() {
1291 let keys = self.config.overrides.keys().collect::<Vec<_>>();
1292 Err(anyhow!(
1293 "language {:?} has overrides in config not in query: {keys:?}",
1294 self.config.name
1295 ))?;
1296 }
1297
1298 for disabled_scope_name in self
1299 .config
1300 .brackets
1301 .disabled_scopes_by_bracket_ix
1302 .iter()
1303 .flatten()
1304 {
1305 if !override_configs_by_id
1306 .values()
1307 .any(|(scope_name, _)| scope_name == disabled_scope_name)
1308 {
1309 Err(anyhow!(
1310 "language {:?} has overrides in config not in query: {disabled_scope_name:?}",
1311 self.config.name
1312 ))?;
1313 }
1314 }
1315
1316 for (name, override_config) in override_configs_by_id.values_mut() {
1317 override_config.disabled_bracket_ixs = self
1318 .config
1319 .brackets
1320 .disabled_scopes_by_bracket_ix
1321 .iter()
1322 .enumerate()
1323 .filter_map(|(ix, disabled_scope_names)| {
1324 if disabled_scope_names.contains(name) {
1325 Some(ix as u16)
1326 } else {
1327 None
1328 }
1329 })
1330 .collect();
1331 }
1332
1333 self.config.brackets.disabled_scopes_by_bracket_ix.clear();
1334
1335 let grammar = self
1336 .grammar_mut()
1337 .ok_or_else(|| anyhow!("cannot mutate grammar"))?;
1338 grammar.override_config = Some(OverrideConfig {
1339 query,
1340 values: override_configs_by_id,
1341 });
1342 Ok(self)
1343 }
1344
1345 pub fn with_redaction_query(mut self, source: &str) -> anyhow::Result<Self> {
1346 let grammar = self
1347 .grammar_mut()
1348 .ok_or_else(|| anyhow!("cannot mutate grammar"))?;
1349
1350 let query = Query::new(&grammar.ts_language, source)?;
1351 let mut redaction_capture_ix = None;
1352 get_capture_indices(&query, &mut [("redact", &mut redaction_capture_ix)]);
1353
1354 if let Some(redaction_capture_ix) = redaction_capture_ix {
1355 grammar.redactions_config = Some(RedactionConfig {
1356 query,
1357 redaction_capture_ix,
1358 });
1359 }
1360
1361 Ok(self)
1362 }
1363
1364 fn grammar_mut(&mut self) -> Option<&mut Grammar> {
1365 Arc::get_mut(self.grammar.as_mut()?)
1366 }
1367
1368 pub fn name(&self) -> LanguageName {
1369 self.config.name.clone()
1370 }
1371
1372 pub fn code_fence_block_name(&self) -> Arc<str> {
1373 self.config
1374 .code_fence_block_name
1375 .clone()
1376 .unwrap_or_else(|| self.config.name.0.to_lowercase().into())
1377 }
1378
1379 pub fn context_provider(&self) -> Option<Arc<dyn ContextProvider>> {
1380 self.context_provider.clone()
1381 }
1382
1383 pub fn highlight_text<'a>(
1384 self: &'a Arc<Self>,
1385 text: &'a Rope,
1386 range: Range<usize>,
1387 ) -> Vec<(Range<usize>, HighlightId)> {
1388 let mut result = Vec::new();
1389 if let Some(grammar) = &self.grammar {
1390 let tree = grammar.parse_text(text, None);
1391 let captures =
1392 SyntaxSnapshot::single_tree_captures(range.clone(), text, &tree, self, |grammar| {
1393 grammar.highlights_query.as_ref()
1394 });
1395 let highlight_maps = vec![grammar.highlight_map()];
1396 let mut offset = 0;
1397 for chunk in
1398 BufferChunks::new(text, range, Some((captures, highlight_maps)), false, None)
1399 {
1400 let end_offset = offset + chunk.text.len();
1401 if let Some(highlight_id) = chunk.syntax_highlight_id {
1402 if !highlight_id.is_default() {
1403 result.push((offset..end_offset, highlight_id));
1404 }
1405 }
1406 offset = end_offset;
1407 }
1408 }
1409 result
1410 }
1411
1412 pub fn path_suffixes(&self) -> &[String] {
1413 &self.config.matcher.path_suffixes
1414 }
1415
1416 pub fn should_autoclose_before(&self, c: char) -> bool {
1417 c.is_whitespace() || self.config.autoclose_before.contains(c)
1418 }
1419
1420 pub fn set_theme(&self, theme: &SyntaxTheme) {
1421 if let Some(grammar) = self.grammar.as_ref() {
1422 if let Some(highlights_query) = &grammar.highlights_query {
1423 *grammar.highlight_map.lock() =
1424 HighlightMap::new(highlights_query.capture_names(), theme);
1425 }
1426 }
1427 }
1428
1429 pub fn grammar(&self) -> Option<&Arc<Grammar>> {
1430 self.grammar.as_ref()
1431 }
1432
1433 pub fn default_scope(self: &Arc<Self>) -> LanguageScope {
1434 LanguageScope {
1435 language: self.clone(),
1436 override_id: None,
1437 }
1438 }
1439
1440 pub fn lsp_id(&self) -> String {
1441 self.config.name.lsp_id()
1442 }
1443
1444 pub fn prettier_parser_name(&self) -> Option<&str> {
1445 self.config.prettier_parser_name.as_deref()
1446 }
1447}
1448
1449impl LanguageScope {
1450 pub fn path_suffixes(&self) -> &[String] {
1451 &self.language.path_suffixes()
1452 }
1453
1454 pub fn language_name(&self) -> LanguageName {
1455 self.language.config.name.clone()
1456 }
1457
1458 pub fn collapsed_placeholder(&self) -> &str {
1459 self.language.config.collapsed_placeholder.as_ref()
1460 }
1461
1462 /// Returns line prefix that is inserted in e.g. line continuations or
1463 /// in `toggle comments` action.
1464 pub fn line_comment_prefixes(&self) -> &[Arc<str>] {
1465 Override::as_option(
1466 self.config_override().map(|o| &o.line_comments),
1467 Some(&self.language.config.line_comments),
1468 )
1469 .map_or(&[] as &[_], |e| e.as_slice())
1470 }
1471
1472 pub fn block_comment_delimiters(&self) -> Option<(&Arc<str>, &Arc<str>)> {
1473 Override::as_option(
1474 self.config_override().map(|o| &o.block_comment),
1475 self.language.config.block_comment.as_ref(),
1476 )
1477 .map(|e| (&e.0, &e.1))
1478 }
1479
1480 /// Returns a list of language-specific word characters.
1481 ///
1482 /// By default, Zed treats alphanumeric characters (and '_') as word characters for
1483 /// the purpose of actions like 'move to next word end` or whole-word search.
1484 /// It additionally accounts for language's additional word characters.
1485 pub fn word_characters(&self) -> Option<&HashSet<char>> {
1486 Override::as_option(
1487 self.config_override().map(|o| &o.word_characters),
1488 Some(&self.language.config.word_characters),
1489 )
1490 }
1491
1492 /// Returns a list of bracket pairs for a given language with an additional
1493 /// piece of information about whether the particular bracket pair is currently active for a given language.
1494 pub fn brackets(&self) -> impl Iterator<Item = (&BracketPair, bool)> {
1495 let mut disabled_ids = self
1496 .config_override()
1497 .map_or(&[] as _, |o| o.disabled_bracket_ixs.as_slice());
1498 self.language
1499 .config
1500 .brackets
1501 .pairs
1502 .iter()
1503 .enumerate()
1504 .map(move |(ix, bracket)| {
1505 let mut is_enabled = true;
1506 if let Some(next_disabled_ix) = disabled_ids.first() {
1507 if ix == *next_disabled_ix as usize {
1508 disabled_ids = &disabled_ids[1..];
1509 is_enabled = false;
1510 }
1511 }
1512 (bracket, is_enabled)
1513 })
1514 }
1515
1516 pub fn should_autoclose_before(&self, c: char) -> bool {
1517 c.is_whitespace() || self.language.config.autoclose_before.contains(c)
1518 }
1519
1520 pub fn language_allowed(&self, name: &LanguageServerName) -> bool {
1521 let config = &self.language.config;
1522 let opt_in_servers = &config.scope_opt_in_language_servers;
1523 if opt_in_servers.iter().any(|o| *o == *name) {
1524 if let Some(over) = self.config_override() {
1525 over.opt_into_language_servers.iter().any(|o| *o == *name)
1526 } else {
1527 false
1528 }
1529 } else {
1530 true
1531 }
1532 }
1533
1534 pub fn override_name(&self) -> Option<&str> {
1535 let id = self.override_id?;
1536 let grammar = self.language.grammar.as_ref()?;
1537 let override_config = grammar.override_config.as_ref()?;
1538 override_config.values.get(&id).map(|e| e.0.as_str())
1539 }
1540
1541 fn config_override(&self) -> Option<&LanguageConfigOverride> {
1542 let id = self.override_id?;
1543 let grammar = self.language.grammar.as_ref()?;
1544 let override_config = grammar.override_config.as_ref()?;
1545 override_config.values.get(&id).map(|e| &e.1)
1546 }
1547}
1548
1549impl Hash for Language {
1550 fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
1551 self.id.hash(state)
1552 }
1553}
1554
1555impl PartialEq for Language {
1556 fn eq(&self, other: &Self) -> bool {
1557 self.id.eq(&other.id)
1558 }
1559}
1560
1561impl Eq for Language {}
1562
1563impl Debug for Language {
1564 fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
1565 f.debug_struct("Language")
1566 .field("name", &self.config.name)
1567 .finish()
1568 }
1569}
1570
1571impl Grammar {
1572 pub fn id(&self) -> GrammarId {
1573 self.id
1574 }
1575
1576 fn parse_text(&self, text: &Rope, old_tree: Option<Tree>) -> Tree {
1577 with_parser(|parser| {
1578 parser
1579 .set_language(&self.ts_language)
1580 .expect("incompatible grammar");
1581 let mut chunks = text.chunks_in_range(0..text.len());
1582 parser
1583 .parse_with(
1584 &mut move |offset, _| {
1585 chunks.seek(offset);
1586 chunks.next().unwrap_or("").as_bytes()
1587 },
1588 old_tree.as_ref(),
1589 )
1590 .unwrap()
1591 })
1592 }
1593
1594 pub fn highlight_map(&self) -> HighlightMap {
1595 self.highlight_map.lock().clone()
1596 }
1597
1598 pub fn highlight_id_for_name(&self, name: &str) -> Option<HighlightId> {
1599 let capture_id = self
1600 .highlights_query
1601 .as_ref()?
1602 .capture_index_for_name(name)?;
1603 Some(self.highlight_map.lock().get(capture_id))
1604 }
1605}
1606
1607impl CodeLabel {
1608 pub fn plain(text: String, filter_text: Option<&str>) -> Self {
1609 let mut result = Self {
1610 runs: Vec::new(),
1611 filter_range: 0..text.len(),
1612 text,
1613 };
1614 if let Some(filter_text) = filter_text {
1615 if let Some(ix) = result.text.find(filter_text) {
1616 result.filter_range = ix..ix + filter_text.len();
1617 }
1618 }
1619 result
1620 }
1621
1622 pub fn push_str(&mut self, text: &str, highlight: Option<HighlightId>) {
1623 let start_ix = self.text.len();
1624 self.text.push_str(text);
1625 let end_ix = self.text.len();
1626 if let Some(highlight) = highlight {
1627 self.runs.push((start_ix..end_ix, highlight));
1628 }
1629 }
1630
1631 pub fn text(&self) -> &str {
1632 self.text.as_str()
1633 }
1634}
1635
1636impl From<String> for CodeLabel {
1637 fn from(value: String) -> Self {
1638 Self::plain(value, None)
1639 }
1640}
1641
1642impl From<&str> for CodeLabel {
1643 fn from(value: &str) -> Self {
1644 Self::plain(value.to_string(), None)
1645 }
1646}
1647
1648impl Ord for LanguageMatcher {
1649 fn cmp(&self, other: &Self) -> std::cmp::Ordering {
1650 self.path_suffixes.cmp(&other.path_suffixes).then_with(|| {
1651 self.first_line_pattern
1652 .as_ref()
1653 .map(Regex::as_str)
1654 .cmp(&other.first_line_pattern.as_ref().map(Regex::as_str))
1655 })
1656 }
1657}
1658
1659impl PartialOrd for LanguageMatcher {
1660 fn partial_cmp(&self, other: &Self) -> Option<std::cmp::Ordering> {
1661 Some(self.cmp(other))
1662 }
1663}
1664
1665impl Eq for LanguageMatcher {}
1666
1667impl PartialEq for LanguageMatcher {
1668 fn eq(&self, other: &Self) -> bool {
1669 self.path_suffixes == other.path_suffixes
1670 && self.first_line_pattern.as_ref().map(Regex::as_str)
1671 == other.first_line_pattern.as_ref().map(Regex::as_str)
1672 }
1673}
1674
1675#[cfg(any(test, feature = "test-support"))]
1676impl Default for FakeLspAdapter {
1677 fn default() -> Self {
1678 Self {
1679 name: "the-fake-language-server",
1680 capabilities: lsp::LanguageServer::full_capabilities(),
1681 initializer: None,
1682 disk_based_diagnostics_progress_token: None,
1683 initialization_options: None,
1684 disk_based_diagnostics_sources: Vec::new(),
1685 prettier_plugins: Vec::new(),
1686 language_server_binary: LanguageServerBinary {
1687 path: "/the/fake/lsp/path".into(),
1688 arguments: vec![],
1689 env: Default::default(),
1690 },
1691 }
1692 }
1693}
1694
1695#[cfg(any(test, feature = "test-support"))]
1696#[async_trait(?Send)]
1697impl LspAdapter for FakeLspAdapter {
1698 fn name(&self) -> LanguageServerName {
1699 LanguageServerName(self.name.into())
1700 }
1701
1702 async fn check_if_user_installed(
1703 &self,
1704 _: &dyn LspAdapterDelegate,
1705 _: &AsyncAppContext,
1706 ) -> Option<LanguageServerBinary> {
1707 Some(self.language_server_binary.clone())
1708 }
1709
1710 fn get_language_server_command<'a>(
1711 self: Arc<Self>,
1712 _: Option<Arc<Path>>,
1713 _: Arc<dyn LspAdapterDelegate>,
1714 _: futures::lock::MutexGuard<'a, Option<LanguageServerBinary>>,
1715 _: &'a mut AsyncAppContext,
1716 ) -> Pin<Box<dyn 'a + Future<Output = Result<LanguageServerBinary>>>> {
1717 async move { Ok(self.language_server_binary.clone()) }.boxed_local()
1718 }
1719
1720 async fn fetch_latest_server_version(
1721 &self,
1722 _: &dyn LspAdapterDelegate,
1723 ) -> Result<Box<dyn 'static + Send + Any>> {
1724 unreachable!();
1725 }
1726
1727 async fn fetch_server_binary(
1728 &self,
1729 _: Box<dyn 'static + Send + Any>,
1730 _: PathBuf,
1731 _: &dyn LspAdapterDelegate,
1732 ) -> Result<LanguageServerBinary> {
1733 unreachable!();
1734 }
1735
1736 async fn cached_server_binary(
1737 &self,
1738 _: PathBuf,
1739 _: &dyn LspAdapterDelegate,
1740 ) -> Option<LanguageServerBinary> {
1741 unreachable!();
1742 }
1743
1744 async fn installation_test_binary(&self, _: PathBuf) -> Option<LanguageServerBinary> {
1745 unreachable!();
1746 }
1747
1748 fn process_diagnostics(&self, _: &mut lsp::PublishDiagnosticsParams) {}
1749
1750 fn disk_based_diagnostic_sources(&self) -> Vec<String> {
1751 self.disk_based_diagnostics_sources.clone()
1752 }
1753
1754 fn disk_based_diagnostics_progress_token(&self) -> Option<String> {
1755 self.disk_based_diagnostics_progress_token.clone()
1756 }
1757
1758 async fn initialization_options(
1759 self: Arc<Self>,
1760 _: &Arc<dyn LspAdapterDelegate>,
1761 ) -> Result<Option<Value>> {
1762 Ok(self.initialization_options.clone())
1763 }
1764}
1765
1766fn get_capture_indices(query: &Query, captures: &mut [(&str, &mut Option<u32>)]) {
1767 for (ix, name) in query.capture_names().iter().enumerate() {
1768 for (capture_name, index) in captures.iter_mut() {
1769 if capture_name == name {
1770 **index = Some(ix as u32);
1771 break;
1772 }
1773 }
1774 }
1775}
1776
1777pub fn point_to_lsp(point: PointUtf16) -> lsp::Position {
1778 lsp::Position::new(point.row, point.column)
1779}
1780
1781pub fn point_from_lsp(point: lsp::Position) -> Unclipped<PointUtf16> {
1782 Unclipped(PointUtf16::new(point.line, point.character))
1783}
1784
1785pub fn range_to_lsp(range: Range<PointUtf16>) -> lsp::Range {
1786 lsp::Range {
1787 start: point_to_lsp(range.start),
1788 end: point_to_lsp(range.end),
1789 }
1790}
1791
1792pub fn range_from_lsp(range: lsp::Range) -> Range<Unclipped<PointUtf16>> {
1793 let mut start = point_from_lsp(range.start);
1794 let mut end = point_from_lsp(range.end);
1795 if start > end {
1796 mem::swap(&mut start, &mut end);
1797 }
1798 start..end
1799}
1800
1801#[cfg(test)]
1802mod tests {
1803 use super::*;
1804 use gpui::TestAppContext;
1805
1806 #[gpui::test(iterations = 10)]
1807 async fn test_language_loading(cx: &mut TestAppContext) {
1808 let languages = LanguageRegistry::test(cx.executor());
1809 let languages = Arc::new(languages);
1810 languages.register_native_grammars([
1811 ("json", tree_sitter_json::LANGUAGE),
1812 ("rust", tree_sitter_rust::LANGUAGE),
1813 ]);
1814 languages.register_test_language(LanguageConfig {
1815 name: "JSON".into(),
1816 grammar: Some("json".into()),
1817 matcher: LanguageMatcher {
1818 path_suffixes: vec!["json".into()],
1819 ..Default::default()
1820 },
1821 ..Default::default()
1822 });
1823 languages.register_test_language(LanguageConfig {
1824 name: "Rust".into(),
1825 grammar: Some("rust".into()),
1826 matcher: LanguageMatcher {
1827 path_suffixes: vec!["rs".into()],
1828 ..Default::default()
1829 },
1830 ..Default::default()
1831 });
1832 assert_eq!(
1833 languages.language_names(),
1834 &[
1835 "JSON".to_string(),
1836 "Plain Text".to_string(),
1837 "Rust".to_string(),
1838 ]
1839 );
1840
1841 let rust1 = languages.language_for_name("Rust");
1842 let rust2 = languages.language_for_name("Rust");
1843
1844 // Ensure language is still listed even if it's being loaded.
1845 assert_eq!(
1846 languages.language_names(),
1847 &[
1848 "JSON".to_string(),
1849 "Plain Text".to_string(),
1850 "Rust".to_string(),
1851 ]
1852 );
1853
1854 let (rust1, rust2) = futures::join!(rust1, rust2);
1855 assert!(Arc::ptr_eq(&rust1.unwrap(), &rust2.unwrap()));
1856
1857 // Ensure language is still listed even after loading it.
1858 assert_eq!(
1859 languages.language_names(),
1860 &[
1861 "JSON".to_string(),
1862 "Plain Text".to_string(),
1863 "Rust".to_string(),
1864 ]
1865 );
1866
1867 // Loading an unknown language returns an error.
1868 assert!(languages.language_for_name("Unknown").await.is_err());
1869 }
1870}