Yehowshua/ZIm
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Use SyntaxMap in Buffer

Browse source
This commit is contained in:
Max Brunsfeld 2022-08-23 17:09:13 -07:00
parent 9113c94371
commit ced45cbb0a
6 changed files with 830 additions and 532 deletions
Download patch file Download diff file Expand all files Collapse all files

616
crates/language/src/buffer.rs
View file

@ -6,13 +6,15 @@ pub use crate::{
use crate::{ use crate::{
diagnostic_set::{DiagnosticEntry, DiagnosticGroup}, diagnostic_set::{DiagnosticEntry, DiagnosticGroup},
outline::OutlineItem, outline::OutlineItem,
syntax_map::{
SyntaxMap, SyntaxMapCapture, SyntaxMapCaptures, SyntaxSnapshot, ToTreeSitterPoint,
},
CodeLabel, Outline, CodeLabel, Outline,
}; };
use anyhow::{anyhow, Result}; use anyhow::{anyhow, Result};
use clock::ReplicaId; use clock::ReplicaId;
use futures::FutureExt as _; use futures::FutureExt as _;
use gpui::{fonts::HighlightStyle, AppContext, Entity, ModelContext, MutableAppContext, Task}; use gpui::{fonts::HighlightStyle, AppContext, Entity, ModelContext, MutableAppContext, Task};
use lazy_static::lazy_static;
use parking_lot::Mutex; use parking_lot::Mutex;
use settings::Settings; use settings::Settings;
use similar::{ChangeTag, TextDiff}; use similar::{ChangeTag, TextDiff};
@ -25,7 +27,7 @@ use std::{
future::Future, future::Future,
iter::{self, Iterator, Peekable}, iter::{self, Iterator, Peekable},
mem, mem,
ops::{Deref, DerefMut, Range}, ops::{Deref, Range},
path::{Path, PathBuf}, path::{Path, PathBuf},
str, str,
sync::Arc, sync::Arc,
@ -36,7 +38,6 @@ use sum_tree::TreeMap;
use text::operation_queue::OperationQueue; use text::operation_queue::OperationQueue;
pub use text::{Buffer as TextBuffer, BufferSnapshot as TextBufferSnapshot, Operation as _, *}; pub use text::{Buffer as TextBuffer, BufferSnapshot as TextBufferSnapshot, Operation as _, *};
use theme::SyntaxTheme; use theme::SyntaxTheme;
use tree_sitter::{InputEdit, QueryCursor, Tree};
use util::TryFutureExt as _; use util::TryFutureExt as _;
#[cfg(any(test, feature = "test-support"))] #[cfg(any(test, feature = "test-support"))]
@ -44,10 +45,6 @@ pub use {tree_sitter_rust, tree_sitter_typescript};
pub use lsp::DiagnosticSeverity; pub use lsp::DiagnosticSeverity;
lazy_static! {
static ref QUERY_CURSORS: Mutex<Vec<QueryCursor>> = Default::default();
}
pub struct Buffer { pub struct Buffer {
text: TextBuffer, text: TextBuffer,
file: Option<Arc<dyn File>>, file: Option<Arc<dyn File>>,
@ -60,7 +57,7 @@ pub struct Buffer {
autoindent_requests: Vec<Arc<AutoindentRequest>>, autoindent_requests: Vec<Arc<AutoindentRequest>>,
pending_autoindent: Option<Task<()>>, pending_autoindent: Option<Task<()>>,
sync_parse_timeout: Duration, sync_parse_timeout: Duration,
syntax_tree: Mutex<Option<SyntaxTree>>, syntax_map: Mutex<SyntaxMap>,
parsing_in_background: bool, parsing_in_background: bool,
parse_count: usize, parse_count: usize,
diagnostics: DiagnosticSet, diagnostics: DiagnosticSet,
@ -75,7 +72,7 @@ pub struct Buffer {
pub struct BufferSnapshot { pub struct BufferSnapshot {
text: text::BufferSnapshot, text: text::BufferSnapshot,
tree: Option<Tree>, syntax: SyntaxSnapshot,
file: Option<Arc<dyn File>>, file: Option<Arc<dyn File>>,
diagnostics: DiagnosticSet, diagnostics: DiagnosticSet,
diagnostics_update_count: usize, diagnostics_update_count: usize,
@ -221,14 +218,6 @@ pub trait LocalFile: File {
); );
} }
pub(crate) struct QueryCursorHandle(Option<QueryCursor>);
#[derive(Clone)]
struct SyntaxTree {
tree: Tree,
version: clock::Global,
}
#[derive(Clone, Debug)] #[derive(Clone, Debug)]
pub enum AutoindentMode { pub enum AutoindentMode {
/// Indent each line of inserted text. /// Indent each line of inserted text.
@ -268,14 +257,11 @@ struct IndentSuggestion {
delta: Ordering, delta: Ordering,
} }
pub(crate) struct TextProvider<'a>(pub(crate) &'a Rope);
struct BufferChunkHighlights<'a> { struct BufferChunkHighlights<'a> {
captures: tree_sitter::QueryCaptures<'a, 'a, TextProvider<'a>>, captures: SyntaxMapCaptures<'a>,
next_capture: Option<(tree_sitter::QueryMatch<'a, 'a>, usize)>, next_capture: Option<SyntaxMapCapture<'a>>,
stack: Vec<(usize, HighlightId)>, stack: Vec<(usize, HighlightId)>,
highlight_map: HighlightMap, highlight_maps: Vec<HighlightMap>,
_query_cursor: QueryCursorHandle,
} }
pub struct BufferChunks<'a> { pub struct BufferChunks<'a> {
@ -456,7 +442,7 @@ impl Buffer {
was_dirty_before_starting_transaction: None, was_dirty_before_starting_transaction: None,
text: buffer, text: buffer,
file, file,
syntax_tree: Mutex::new(None), syntax_map: Mutex::new(SyntaxMap::new()),
parsing_in_background: false, parsing_in_background: false,
parse_count: 0, parse_count: 0,
sync_parse_timeout: Duration::from_millis(1), sync_parse_timeout: Duration::from_millis(1),
@ -477,7 +463,7 @@ impl Buffer {
pub fn snapshot(&self) -> BufferSnapshot { pub fn snapshot(&self) -> BufferSnapshot {
BufferSnapshot { BufferSnapshot {
text: self.text.snapshot(), text: self.text.snapshot(),
tree: self.syntax_tree(), syntax: self.syntax_map(),
file: self.file.clone(), file: self.file.clone(),
remote_selections: self.remote_selections.clone(), remote_selections: self.remote_selections.clone(),
diagnostics: self.diagnostics.clone(), diagnostics: self.diagnostics.clone(),
@ -533,11 +519,17 @@ impl Buffer {
} }
pub fn set_language(&mut self, language: Option<Arc<Language>>, cx: &mut ModelContext<Self>) { pub fn set_language(&mut self, language: Option<Arc<Language>>, cx: &mut ModelContext<Self>) {
*self.syntax_tree.lock() = None; self.syntax_map.lock().clear();
self.language = language; self.language = language;
self.reparse(cx); self.reparse(cx);
} }
pub fn set_language_registry(&mut self, language_registry: Arc<LanguageRegistry>) {
self.syntax_map
.lock()
.set_language_registry(language_registry);
}
pub fn did_save( pub fn did_save(
&mut self, &mut self,
version: clock::Global, version: clock::Global,
@ -682,13 +674,10 @@ impl Buffer {
self.file_update_count self.file_update_count
} }
pub(crate) fn syntax_tree(&self) -> Option<Tree> { pub(crate) fn syntax_map(&self) -> SyntaxSnapshot {
if let Some(syntax_tree) = self.syntax_tree.lock().as_mut() { let mut syntax_map = self.syntax_map.lock();
self.interpolate_tree(syntax_tree); syntax_map.interpolate(&self.text_snapshot());
Some(syntax_tree.tree.clone()) syntax_map.snapshot()
} else {
None
}
} }
#[cfg(any(test, feature = "test-support"))] #[cfg(any(test, feature = "test-support"))]
@ -706,35 +695,49 @@ impl Buffer {
return false; return false;
} }
if let Some(grammar) = self.grammar().cloned() { if let Some(language) = self.language.clone() {
let old_tree = self.syntax_tree(); let text = self.text_snapshot();
let text = self.as_rope().clone();
let parsed_version = self.version(); let parsed_version = self.version();
let mut syntax_map;
let language_registry;
let syntax_map_version;
{
let mut map = self.syntax_map.lock();
map.interpolate(&text);
language_registry = map.language_registry();
syntax_map = map.snapshot();
syntax_map_version = map.parsed_version();
}
let parse_task = cx.background().spawn({ let parse_task = cx.background().spawn({
let grammar = grammar.clone(); let language = language.clone();
async move { grammar.parse_text(&text, old_tree) } async move {
syntax_map.reparse(&syntax_map_version, &text, language_registry, language);
syntax_map
}
}); });
match cx match cx
.background() .background()
.block_with_timeout(self.sync_parse_timeout, parse_task) .block_with_timeout(self.sync_parse_timeout, parse_task)
{ {
Ok(new_tree) => { Ok(new_syntax_map) => {
self.did_finish_parsing(new_tree, parsed_version, cx); self.did_finish_parsing(new_syntax_map, parsed_version, cx);
return true; return true;
} }
Err(parse_task) => { Err(parse_task) => {
self.parsing_in_background = true; self.parsing_in_background = true;
cx.spawn(move |this, mut cx| async move { cx.spawn(move |this, mut cx| async move {
let new_tree = parse_task.await; let new_syntax_map = parse_task.await;
this.update(&mut cx, move |this, cx| { this.update(&mut cx, move |this, cx| {
let grammar_changed = this let grammar_changed =
.grammar() this.language.as_ref().map_or(true, |current_language| {
.map_or(true, |curr_grammar| !Arc::ptr_eq(&grammar, curr_grammar)); !Arc::ptr_eq(&language, current_language)
});
let parse_again = let parse_again =
this.version.changed_since(&parsed_version) || grammar_changed; this.version.changed_since(&parsed_version) || grammar_changed;
this.parsing_in_background = false; this.parsing_in_background = false;
this.did_finish_parsing(new_tree, parsed_version, cx); this.did_finish_parsing(new_syntax_map, parsed_version, cx);
if parse_again && this.reparse(cx) {} if parse_again && this.reparse(cx) {}
}); });
@ -746,30 +749,14 @@ impl Buffer {
false false
} }
fn interpolate_tree(&self, tree: &mut SyntaxTree) {
for edit in self.edits_since::<(usize, Point)>(&tree.version) {
let (bytes, lines) = edit.flatten();
tree.tree.edit(&InputEdit {
start_byte: bytes.new.start,
old_end_byte: bytes.new.start + bytes.old.len(),
new_end_byte: bytes.new.end,
start_position: lines.new.start.to_ts_point(),
old_end_position: (lines.new.start + (lines.old.end - lines.old.start))
.to_ts_point(),
new_end_position: lines.new.end.to_ts_point(),
});
}
tree.version = self.version();
}
fn did_finish_parsing( fn did_finish_parsing(
&mut self, &mut self,
tree: Tree, syntax_map: SyntaxSnapshot,
version: clock::Global, version: clock::Global,
cx: &mut ModelContext<Self>, cx: &mut ModelContext<Self>,
) { ) {
self.parse_count += 1; self.parse_count += 1;
*self.syntax_tree.lock() = Some(SyntaxTree { tree, version }); self.syntax_map.lock().did_parse(syntax_map, version);
self.request_autoindent(cx); self.request_autoindent(cx);
cx.emit(Event::Reparsed); cx.emit(Event::Reparsed);
cx.notify(); cx.notify();
@ -808,10 +795,7 @@ impl Buffer {
fn compute_autoindents(&self) -> Option<impl Future<Output = BTreeMap<u32, IndentSize>>> { fn compute_autoindents(&self) -> Option<impl Future<Output = BTreeMap<u32, IndentSize>>> {
let max_rows_between_yields = 100; let max_rows_between_yields = 100;
let snapshot = self.snapshot(); let snapshot = self.snapshot();
if snapshot.language.is_none() if snapshot.syntax.is_empty() || self.autoindent_requests.is_empty() {
|| snapshot.tree.is_none()
|| self.autoindent_requests.is_empty()
{
return None; return None;
} }
@ -1310,10 +1294,6 @@ impl Buffer {
cx.notify(); cx.notify();
} }
fn grammar(&self) -> Option<&Arc<Grammar>> {
self.language.as_ref().and_then(|l| l.grammar.as_ref())
}
pub fn apply_ops<I: IntoIterator<Item = Operation>>( pub fn apply_ops<I: IntoIterator<Item = Operation>>(
&mut self, &mut self,
ops: I, ops: I,
@ -1654,32 +1634,30 @@ impl BufferSnapshot {
let prev_non_blank_row = self.prev_non_blank_row(row_range.start); let prev_non_blank_row = self.prev_non_blank_row(row_range.start);
// Find the suggested indentation ranges based on the syntax tree. // Find the suggested indentation ranges based on the syntax tree.
let indents_query = grammar.indents_query.as_ref()?; let start = Point::new(prev_non_blank_row.unwrap_or(row_range.start), 0);
let mut query_cursor = QueryCursorHandle::new(); let end = Point::new(row_range.end, 0);
let indent_capture_ix = indents_query.capture_index_for_name("indent"); let range = (start..end).to_offset(&self.text);
let end_capture_ix = indents_query.capture_index_for_name("end"); let mut matches = self.syntax.matches(range, &self.text, |grammar| {
query_cursor.set_point_range( Some(&grammar.indents_config.as_ref()?.query)
Point::new(prev_non_blank_row.unwrap_or(row_range.start), 0).to_ts_point() });
..Point::new(row_range.end, 0).to_ts_point(),
);
let mut indent_ranges = Vec::<Range<Point>>::new(); let mut indent_ranges = Vec::<Range<Point>>::new();
for mat in query_cursor.matches( while let Some(mat) = matches.peek() {
indents_query,
self.tree.as_ref()?.root_node(),
TextProvider(self.as_rope()),
) {
let mut start: Option<Point> = None; let mut start: Option<Point> = None;
let mut end: Option<Point> = None; let mut end: Option<Point> = None;
for capture in mat.captures {
if Some(capture.index) == indent_capture_ix { if let Some(config) = &grammar.indents_config {
start.get_or_insert(Point::from_ts_point(capture.node.start_position())); for capture in mat.captures {
end.get_or_insert(Point::from_ts_point(capture.node.end_position())); if capture.index == config.indent_capture_ix {
} else if Some(capture.index) == end_capture_ix { start.get_or_insert(Point::from_ts_point(capture.node.start_position()));
end = Some(Point::from_ts_point(capture.node.start_position())); end.get_or_insert(Point::from_ts_point(capture.node.end_position()));
} else if Some(capture.index) == config.end_capture_ix {
end = Some(Point::from_ts_point(capture.node.start_position()));
}
} }
} }
matches.advance();
if let Some((start, end)) = start.zip(end) { if let Some((start, end)) = start.zip(end) {
if start.row == end.row { if start.row == end.row {
continue; continue;
@ -1811,10 +1789,18 @@ impl BufferSnapshot {
pub fn chunks<T: ToOffset>(&self, range: Range<T>, language_aware: bool) -> BufferChunks { pub fn chunks<T: ToOffset>(&self, range: Range<T>, language_aware: bool) -> BufferChunks {
let range = range.start.to_offset(self)..range.end.to_offset(self); let range = range.start.to_offset(self)..range.end.to_offset(self);
let mut tree = None; let mut syntax = None;
let mut diagnostic_endpoints = Vec::new(); let mut diagnostic_endpoints = Vec::new();
if language_aware { if language_aware {
tree = self.tree.as_ref(); let captures = self.syntax.captures(range.clone(), &self.text, |grammar| {
grammar.highlights_query.as_ref()
});
let highlight_maps = captures
.grammars()
.into_iter()
.map(|grammar| grammar.highlight_map())
.collect();
syntax = Some((captures, highlight_maps));
for entry in self.diagnostics_in_range::<_, usize>(range.clone(), false) { for entry in self.diagnostics_in_range::<_, usize>(range.clone(), false) {
diagnostic_endpoints.push(DiagnosticEndpoint { diagnostic_endpoints.push(DiagnosticEndpoint {
offset: entry.range.start, offset: entry.range.start,
@ -1833,13 +1819,7 @@ impl BufferSnapshot {
.sort_unstable_by_key(|endpoint| (endpoint.offset, !endpoint.is_start)); .sort_unstable_by_key(|endpoint| (endpoint.offset, !endpoint.is_start));
} }
BufferChunks::new( BufferChunks::new(self.text.as_rope(), range, syntax, diagnostic_endpoints)
self.text.as_rope(),
range,
tree,
self.grammar(),
diagnostic_endpoints,
)
} }
pub fn for_each_line(&self, range: Range<Point>, mut callback: impl FnMut(u32, &str)) { pub fn for_each_line(&self, range: Range<Point>, mut callback: impl FnMut(u32, &str)) {
@ -1865,12 +1845,6 @@ impl BufferSnapshot {
self.language.as_ref() self.language.as_ref()
} }
fn grammar(&self) -> Option<&Arc<Grammar>> {
self.language
.as_ref()
.and_then(|language| language.grammar.as_ref())
}
pub fn surrounding_word<T: ToOffset>(&self, start: T) -> (Range<usize>, Option<CharKind>) { pub fn surrounding_word<T: ToOffset>(&self, start: T) -> (Range<usize>, Option<CharKind>) {
let mut start = start.to_offset(self); let mut start = start.to_offset(self);
let mut end = start; let mut end = start;
@ -1901,61 +1875,71 @@ impl BufferSnapshot {
} }
pub fn range_for_syntax_ancestor<T: ToOffset>(&self, range: Range<T>) -> Option<Range<usize>> { pub fn range_for_syntax_ancestor<T: ToOffset>(&self, range: Range<T>) -> Option<Range<usize>> {
let tree = self.tree.as_ref()?;
let range = range.start.to_offset(self)..range.end.to_offset(self); let range = range.start.to_offset(self)..range.end.to_offset(self);
let mut cursor = tree.root_node().walk(); let mut result: Option<Range<usize>> = None;
'outer: for (_, _, node) in self.syntax.layers_for_range(range.clone(), &self.text) {
let mut cursor = node.walk();
// Descend to the first leaf that touches the start of the range, // Descend to the first leaf that touches the start of the range,
// and if the range is non-empty, extends beyond the start. // and if the range is non-empty, extends beyond the start.
while cursor.goto_first_child_for_byte(range.start).is_some() { while cursor.goto_first_child_for_byte(range.start).is_some() {
if !range.is_empty() && cursor.node().end_byte() == range.start { if !range.is_empty() && cursor.node().end_byte() == range.start {
cursor.goto_next_sibling(); cursor.goto_next_sibling();
}
} }
}
// Ascend to the smallest ancestor that strictly contains the range. // Ascend to the smallest ancestor that strictly contains the range.
loop { loop {
let node_range = cursor.node().byte_range(); let node_range = cursor.node().byte_range();
if node_range.start <= range.start if node_range.start <= range.start
&& node_range.end >= range.end && node_range.end >= range.end
&& node_range.len() > range.len() && node_range.len() > range.len()
{ {
break; break;
} }
if !cursor.goto_parent() {
break;
}
}
let left_node = cursor.node();
// For an empty range, try to find another node immediately to the right of the range.
if left_node.end_byte() == range.start {
let mut right_node = None;
while !cursor.goto_next_sibling() {
if !cursor.goto_parent() { if !cursor.goto_parent() {
break; continue 'outer;
} }
} }
while cursor.node().start_byte() == range.start { let left_node = cursor.node();
right_node = Some(cursor.node()); let mut layer_result = left_node.byte_range();
if !cursor.goto_first_child() {
break; // For an empty range, try to find another node immediately to the right of the range.
if left_node.end_byte() == range.start {
let mut right_node = None;
while !cursor.goto_next_sibling() {
if !cursor.goto_parent() {
break;
}
}
while cursor.node().start_byte() == range.start {
right_node = Some(cursor.node());
if !cursor.goto_first_child() {
break;
}
}
// If there is a candidate node on both sides of the (empty) range, then
// decide between the two by favoring a named node over an anonymous token.
// If both nodes are the same in that regard, favor the right one.
if let Some(right_node) = right_node {
if right_node.is_named() || !left_node.is_named() {
layer_result = right_node.byte_range();
}
} }
} }
// If there is a candidate node on both sides of the (empty) range, then if let Some(previous_result) = &result {
// decide between the two by favoring a named node over an anonymous token. if previous_result.len() < layer_result.len() {
// If both nodes are the same in that regard, favor the right one. continue;
if let Some(right_node) = right_node {
if right_node.is_named() || !left_node.is_named() {
return Some(right_node.byte_range());
} }
} }
result = Some(layer_result);
} }
Some(left_node.byte_range()) result
} }
pub fn outline(&self, theme: Option<&SyntaxTheme>) -> Option<Outline<Anchor>> { pub fn outline(&self, theme: Option<&SyntaxTheme>) -> Option<Outline<Anchor>> {
@ -1985,109 +1969,107 @@ impl BufferSnapshot {
range: Range<usize>, range: Range<usize>,
theme: Option<&SyntaxTheme>, theme: Option<&SyntaxTheme>,
) -> Option<Vec<OutlineItem<Anchor>>> { ) -> Option<Vec<OutlineItem<Anchor>>> {
let tree = self.tree.as_ref()?; let mut matches = self.syntax.matches(range.clone(), &self.text, |grammar| {
let grammar = self grammar.outline_config.as_ref().map(|c| &c.query)
.language });
.as_ref() let configs = matches
.and_then(|language| language.grammar.as_ref())?; .grammars()
.iter()
let outline_query = grammar.outline_query.as_ref()?; .map(|g| g.outline_config.as_ref().unwrap())
let mut cursor = QueryCursorHandle::new(); .collect::<Vec<_>>();
cursor.set_byte_range(range.clone());
let matches = cursor.matches(
outline_query,
tree.root_node(),
TextProvider(self.as_rope()),
);
let mut chunks = self.chunks(0..self.len(), true); let mut chunks = self.chunks(0..self.len(), true);
let item_capture_ix = outline_query.capture_index_for_name("item")?;
let name_capture_ix = outline_query.capture_index_for_name("name")?;
let context_capture_ix = outline_query
.capture_index_for_name("context")
.unwrap_or(u32::MAX);
let mut stack = Vec::<Range<usize>>::new(); let mut stack = Vec::<Range<usize>>::new();
let items = matches let mut items = Vec::new();
.filter_map(|mat| { while let Some(mat) = matches.peek() {
let item_node = mat.nodes_for_capture_index(item_capture_ix).next()?; let config = &configs[mat.grammar_index];
let item_range = item_node.start_byte()..item_node.end_byte(); let item_node = mat.captures.iter().find_map(|cap| {
if item_range.end < range.start || item_range.start > range.end { if cap.index == config.item_capture_ix {
return None; Some(cap.node)
} else {
None
} }
let mut text = String::new(); })?;
let mut name_ranges = Vec::new();
let mut highlight_ranges = Vec::new();
for capture in mat.captures { let item_range = item_node.byte_range();
let node_is_name; if item_range.end < range.start || item_range.start > range.end {
if capture.index == name_capture_ix { matches.advance();
node_is_name = true; continue;
} else if capture.index == context_capture_ix { }
node_is_name = false;
// TODO - move later, after processing captures
let mut text = String::new();
let mut name_ranges = Vec::new();
let mut highlight_ranges = Vec::new();
for capture in mat.captures {
let node_is_name;
if capture.index == config.name_capture_ix {
node_is_name = true;
} else if Some(capture.index) == config.context_capture_ix {
node_is_name = false;
} else {
continue;
}
let range = capture.node.start_byte()..capture.node.end_byte();
if !text.is_empty() {
text.push(' ');
}
if node_is_name {
let mut start = text.len();
let end = start + range.len();
// When multiple names are captured, then the matcheable text
// includes the whitespace in between the names.
if !name_ranges.is_empty() {
start -= 1;
}
name_ranges.push(start..end);
}
let mut offset = range.start;
chunks.seek(offset);
for mut chunk in chunks.by_ref() {
if chunk.text.len() > range.end - offset {
chunk.text = &chunk.text[0..(range.end - offset)];
offset = range.end;
} else { } else {
continue; offset += chunk.text.len();
} }
let style = chunk
let range = capture.node.start_byte()..capture.node.end_byte(); .syntax_highlight_id
if !text.is_empty() { .zip(theme)
text.push(' '); .and_then(|(highlight, theme)| highlight.style(theme));
if let Some(style) = style {
let start = text.len();
let end = start + chunk.text.len();
highlight_ranges.push((start..end, style));
} }
if node_is_name { text.push_str(chunk.text);
let mut start = text.len(); if offset >= range.end {
let end = start + range.len(); break;
// When multiple names are captured, then the matcheable text
// includes the whitespace in between the names.
if !name_ranges.is_empty() {
start -= 1;
}
name_ranges.push(start..end);
}
let mut offset = range.start;
chunks.seek(offset);
for mut chunk in chunks.by_ref() {
if chunk.text.len() > range.end - offset {
chunk.text = &chunk.text[0..(range.end - offset)];
offset = range.end;
} else {
offset += chunk.text.len();
}
let style = chunk
.syntax_highlight_id
.zip(theme)
.and_then(|(highlight, theme)| highlight.style(theme));
if let Some(style) = style {
let start = text.len();
let end = start + chunk.text.len();
highlight_ranges.push((start..end, style));
}
text.push_str(chunk.text);
if offset >= range.end {
break;
}
} }
} }
}
while stack.last().map_or(false, |prev_range| { matches.advance();
prev_range.start > item_range.start || prev_range.end < item_range.end while stack.last().map_or(false, |prev_range| {
}) { prev_range.start > item_range.start || prev_range.end < item_range.end
stack.pop(); }) {
} stack.pop();
stack.push(item_range.clone()); }
stack.push(item_range.clone());
Some(OutlineItem { items.push(OutlineItem {
depth: stack.len() - 1, depth: stack.len() - 1,
range: self.anchor_after(item_range.start)..self.anchor_before(item_range.end), range: self.anchor_after(item_range.start)..self.anchor_before(item_range.end),
text, text,
highlight_ranges, highlight_ranges,
name_ranges, name_ranges,
})
}) })
.collect::<Vec<_>>(); }
Some(items) Some(items)
} }
@ -2095,28 +2077,48 @@ impl BufferSnapshot {
&self, &self,
range: Range<T>, range: Range<T>,
) -> Option<(Range<usize>, Range<usize>)> { ) -> Option<(Range<usize>, Range<usize>)> {
let (grammar, tree) = self.grammar().zip(self.tree.as_ref())?;
let brackets_query = grammar.brackets_query.as_ref()?;
let open_capture_ix = brackets_query.capture_index_for_name("open")?;
let close_capture_ix = brackets_query.capture_index_for_name("close")?;
// Find bracket pairs that *inclusively* contain the given range. // Find bracket pairs that *inclusively* contain the given range.
let range = range.start.to_offset(self).saturating_sub(1)..range.end.to_offset(self) + 1; let range = range.start.to_offset(self).saturating_sub(1)..range.end.to_offset(self) + 1;
let mut cursor = QueryCursorHandle::new(); let mut matches = self.syntax.matches(range, &self.text, |grammar| {
let matches = cursor.set_byte_range(range).matches( grammar.brackets_config.as_ref().map(|c| &c.query)
brackets_query, });
tree.root_node(), let configs = matches
TextProvider(self.as_rope()), .grammars()
); .iter()
.map(|grammar| grammar.brackets_config.as_ref().unwrap())
.collect::<Vec<_>>();
// Get the ranges of the innermost pair of brackets. // Get the ranges of the innermost pair of brackets.
matches let mut result: Option<(Range<usize>, Range<usize>)> = None;
.filter_map(|mat| { while let Some(mat) = matches.peek() {
let open = mat.nodes_for_capture_index(open_capture_ix).next()?; let mut open = None;
let close = mat.nodes_for_capture_index(close_capture_ix).next()?; let mut close = None;
Some((open.byte_range(), close.byte_range())) let config = &configs[mat.grammar_index];
}) for capture in mat.captures {
.min_by_key(|(open_range, close_range)| close_range.end - open_range.start) if capture.index == config.open_capture_ix {
open = Some(capture.node.byte_range());
} else if capture.index == config.close_capture_ix {
close = Some(capture.node.byte_range());
}
}
matches.advance();
if let Some((open, close)) = open.zip(close) {
let len = close.end - open.start;
if let Some((existing_open, existing_close)) = &result {
let existing_len = existing_close.end - existing_open.start;
if len > existing_len {
continue;
}
}
result = Some((open, close));
}
}
result
} }
#[allow(clippy::type_complexity)] #[allow(clippy::type_complexity)]
@ -2228,7 +2230,7 @@ impl Clone for BufferSnapshot {
fn clone(&self) -> Self { fn clone(&self) -> Self {
Self { Self {
text: self.text.clone(), text: self.text.clone(),
tree: self.tree.clone(), syntax: self.syntax.clone(),
file: self.file.clone(), file: self.file.clone(),
remote_selections: self.remote_selections.clone(), remote_selections: self.remote_selections.clone(),
diagnostics: self.diagnostics.clone(), diagnostics: self.diagnostics.clone(),
@ -2249,56 +2251,23 @@ impl Deref for BufferSnapshot {
} }
} }
impl<'a> tree_sitter::TextProvider<'a> for TextProvider<'a> {
type I = ByteChunks<'a>;
fn text(&mut self, node: tree_sitter::Node) -> Self::I {
ByteChunks(self.0.chunks_in_range(node.byte_range()))
}
}
pub(crate) struct ByteChunks<'a>(rope::Chunks<'a>);
impl<'a> Iterator for ByteChunks<'a> {
type Item = &'a [u8];
fn next(&mut self) -> Option<Self::Item> {
self.0.next().map(str::as_bytes)
}
}
unsafe impl<'a> Send for BufferChunks<'a> {} unsafe impl<'a> Send for BufferChunks<'a> {}
impl<'a> BufferChunks<'a> { impl<'a> BufferChunks<'a> {
pub(crate) fn new( pub(crate) fn new(
text: &'a Rope, text: &'a Rope,
range: Range<usize>, range: Range<usize>,
tree: Option<&'a Tree>, syntax: Option<(SyntaxMapCaptures<'a>, Vec<HighlightMap>)>,
grammar: Option<&'a Arc<Grammar>>,
diagnostic_endpoints: Vec<DiagnosticEndpoint>, diagnostic_endpoints: Vec<DiagnosticEndpoint>,
) -> Self { ) -> Self {
let mut highlights = None; let mut highlights = None;
if let Some((grammar, tree)) = grammar.zip(tree) { if let Some((captures, highlight_maps)) = syntax {
if let Some(highlights_query) = grammar.highlights_query.as_ref() { highlights = Some(BufferChunkHighlights {
let mut query_cursor = QueryCursorHandle::new(); captures,
next_capture: None,
// TODO - add a Tree-sitter API to remove the need for this. stack: Default::default(),
let cursor = unsafe { highlight_maps,
std::mem::transmute::<_, &'static mut QueryCursor>(query_cursor.deref_mut()) })
};
let captures = cursor.set_byte_range(range.clone()).captures(
highlights_query,
tree.root_node(),
TextProvider(text),
);
highlights = Some(BufferChunkHighlights {
captures,
next_capture: None,
stack: Default::default(),
highlight_map: grammar.highlight_map(),
_query_cursor: query_cursor,
})
}
} }
let diagnostic_endpoints = diagnostic_endpoints.into_iter().peekable(); let diagnostic_endpoints = diagnostic_endpoints.into_iter().peekable();
@ -2324,14 +2293,13 @@ impl<'a> BufferChunks<'a> {
highlights highlights
.stack .stack
.retain(|(end_offset, _)| *end_offset > offset); .retain(|(end_offset, _)| *end_offset > offset);
if let Some((mat, capture_ix)) = &highlights.next_capture { if let Some(capture) = &highlights.next_capture {
let capture = mat.captures[*capture_ix as usize];
if offset >= capture.node.start_byte() { if offset >= capture.node.start_byte() {
let next_capture_end = capture.node.end_byte(); let next_capture_end = capture.node.end_byte();
if offset < next_capture_end { if offset < next_capture_end {
highlights.stack.push(( highlights.stack.push((
next_capture_end, next_capture_end,
highlights.highlight_map.get(capture.index), highlights.highlight_maps[capture.grammar_index].get(capture.index),
)); ));
} }
highlights.next_capture.take(); highlights.next_capture.take();
@ -2407,13 +2375,13 @@ impl<'a> Iterator for BufferChunks<'a> {
highlights.next_capture = highlights.captures.next(); highlights.next_capture = highlights.captures.next();
} }
while let Some((mat, capture_ix)) = highlights.next_capture.as_ref() { while let Some(capture) = highlights.next_capture.as_ref() {
let capture = mat.captures[*capture_ix as usize];
if self.range.start < capture.node.start_byte() { if self.range.start < capture.node.start_byte() {
next_capture_start = capture.node.start_byte(); next_capture_start = capture.node.start_byte();
break; break;
} else { } else {
let highlight_id = highlights.highlight_map.get(capture.index); let highlight_id =
highlights.highlight_maps[capture.grammar_index].get(capture.index);
highlights highlights
.stack .stack
.push((capture.node.end_byte(), highlight_id)); .push((capture.node.end_byte(), highlight_id));
@ -2465,52 +2433,6 @@ impl<'a> Iterator for BufferChunks<'a> {
} }
} }
impl QueryCursorHandle {
pub(crate) fn new() -> Self {
let mut cursor = QUERY_CURSORS.lock().pop().unwrap_or_else(QueryCursor::new);
cursor.set_match_limit(64);
QueryCursorHandle(Some(cursor))
}
}
impl Deref for QueryCursorHandle {
type Target = QueryCursor;
fn deref(&self) -> &Self::Target {
self.0.as_ref().unwrap()
}
}
impl DerefMut for QueryCursorHandle {
fn deref_mut(&mut self) -> &mut Self::Target {
self.0.as_mut().unwrap()
}
}
impl Drop for QueryCursorHandle {
fn drop(&mut self) {
let mut cursor = self.0.take().unwrap();
cursor.set_byte_range(0..usize::MAX);
cursor.set_point_range(Point::zero().to_ts_point()..Point::MAX.to_ts_point());
QUERY_CURSORS.lock().push(cursor)
}
}
pub(crate) trait ToTreeSitterPoint {
fn to_ts_point(self) -> tree_sitter::Point;
fn from_ts_point(point: tree_sitter::Point) -> Self;
}
impl ToTreeSitterPoint for Point {
fn to_ts_point(self) -> tree_sitter::Point {
tree_sitter::Point::new(self.row as usize, self.column as usize)
}
fn from_ts_point(point: tree_sitter::Point) -> Self {
Point::new(point.row as u32, point.column as u32)
}
}
impl operation_queue::Operation for Operation { impl operation_queue::Operation for Operation {
fn lamport_timestamp(&self) -> clock::Lamport { fn lamport_timestamp(&self) -> clock::Lamport {
match self { match self {

144
crates/language/src/language.rs
View file

@ -30,8 +30,12 @@ use std::{
ops::Range, ops::Range,
path::{Path, PathBuf}, path::{Path, PathBuf},
str, str,
sync::Arc, sync::{
atomic::{AtomicUsize, Ordering::SeqCst},
Arc,
},
}; };
use syntax_map::SyntaxSnapshot;
use theme::{SyntaxTheme, Theme}; use theme::{SyntaxTheme, Theme};
use tree_sitter::{self, Query}; use tree_sitter::{self, Query};
use util::ResultExt; use util::ResultExt;
@ -50,6 +54,7 @@ thread_local! {
} }
lazy_static! { lazy_static! {
pub static ref NEXT_GRAMMAR_ID: AtomicUsize = Default::default();
pub static ref PLAIN_TEXT: Arc<Language> = Arc::new(Language::new( pub static ref PLAIN_TEXT: Arc<Language> = Arc::new(Language::new(
LanguageConfig { LanguageConfig {
name: "Plain Text".into(), name: "Plain Text".into(),
@ -286,15 +291,29 @@ pub struct Language {
} }
pub struct Grammar { pub struct Grammar {
id: usize,
pub(crate) ts_language: tree_sitter::Language, pub(crate) ts_language: tree_sitter::Language,
pub(crate) highlights_query: Option<Query>, pub(crate) highlights_query: Option<Query>,
pub(crate) brackets_query: Option<Query>, pub(crate) brackets_config: Option<BracketConfig>,
pub(crate) indents_query: Option<Query>, pub(crate) indents_config: Option<IndentConfig>,
pub(crate) outline_query: Option<Query>, pub(crate) outline_config: Option<OutlineConfig>,
pub(crate) injection_config: Option<InjectionConfig>, pub(crate) injection_config: Option<InjectionConfig>,
pub(crate) highlight_map: Mutex<HighlightMap>, pub(crate) highlight_map: Mutex<HighlightMap>,
} }
struct IndentConfig {
query: Query,
indent_capture_ix: u32,
end_capture_ix: Option<u32>,
}
struct OutlineConfig {
query: Query,
item_capture_ix: u32,
name_capture_ix: u32,
context_capture_ix: Option<u32>,
}
struct InjectionConfig { struct InjectionConfig {
query: Query, query: Query,
content_capture_ix: u32, content_capture_ix: u32,
@ -302,6 +321,12 @@ struct InjectionConfig {
languages_by_pattern_ix: Vec<Option<Box<str>>>, languages_by_pattern_ix: Vec<Option<Box<str>>>,
} }
struct BracketConfig {
query: Query,
open_capture_ix: u32,
close_capture_ix: u32,
}
#[derive(Clone)] #[derive(Clone)]
pub enum LanguageServerBinaryStatus { pub enum LanguageServerBinaryStatus {
CheckingForUpdate, CheckingForUpdate,
@ -499,6 +524,13 @@ impl LanguageRegistry {
} }
} }
#[cfg(any(test, feature = "test-support"))]
impl Default for LanguageRegistry {
fn default() -> Self {
Self::test()
}
}
async fn get_server_binary_path( async fn get_server_binary_path(
adapter: Arc<CachedLspAdapter>, adapter: Arc<CachedLspAdapter>,
language: Arc<Language>, language: Arc<Language>,
@ -576,10 +608,11 @@ impl Language {
config, config,
grammar: ts_language.map(|ts_language| { grammar: ts_language.map(|ts_language| {
Arc::new(Grammar { Arc::new(Grammar {
id: NEXT_GRAMMAR_ID.fetch_add(1, SeqCst),
highlights_query: None, highlights_query: None,
brackets_query: None, brackets_config: None,
indents_query: None, outline_config: None,
outline_query: None, indents_config: None,
injection_config: None, injection_config: None,
ts_language, ts_language,
highlight_map: Default::default(), highlight_map: Default::default(),
@ -604,19 +637,70 @@ impl Language {
pub fn with_brackets_query(mut self, source: &str) -> Result<Self> { pub fn with_brackets_query(mut self, source: &str) -> Result<Self> {
let grammar = self.grammar_mut(); let grammar = self.grammar_mut();
grammar.brackets_query = Some(Query::new(grammar.ts_language, source)?); let query = Query::new(grammar.ts_language, source)?;
let mut open_capture_ix = None;
let mut close_capture_ix = None;
get_capture_indices(
&query,
&mut [
("open", &mut open_capture_ix),
("close", &mut close_capture_ix),
],
);
if let Some((open_capture_ix, close_capture_ix)) = open_capture_ix.zip(close_capture_ix) {
grammar.brackets_config = Some(BracketConfig {
query,
open_capture_ix,
close_capture_ix,
});
}
Ok(self) Ok(self)
} }
pub fn with_indents_query(mut self, source: &str) -> Result<Self> { pub fn with_indents_query(mut self, source: &str) -> Result<Self> {
let grammar = self.grammar_mut(); let grammar = self.grammar_mut();
grammar.indents_query = Some(Query::new(grammar.ts_language, source)?); let query = Query::new(grammar.ts_language, source)?;
let mut indent_capture_ix = None;
let mut end_capture_ix = None;
get_capture_indices(
&query,
&mut [
("indent", &mut indent_capture_ix),
("end", &mut end_capture_ix),
],
);
if let Some(indent_capture_ix) = indent_capture_ix {
grammar.indents_config = Some(IndentConfig {
query,
indent_capture_ix,
end_capture_ix,
});
}
Ok(self) Ok(self)
} }
pub fn with_outline_query(mut self, source: &str) -> Result<Self> { pub fn with_outline_query(mut self, source: &str) -> Result<Self> {
let grammar = self.grammar_mut(); let grammar = self.grammar_mut();
grammar.outline_query = Some(Query::new(grammar.ts_language, source)?); let query = Query::new(grammar.ts_language, source)?;
let mut item_capture_ix = None;
let mut name_capture_ix = None;
let mut context_capture_ix = None;
get_capture_indices(
&query,
&mut [
("item", &mut item_capture_ix),
("name", &mut name_capture_ix),
("context", &mut context_capture_ix),
],
);
if let Some((item_capture_ix, name_capture_ix)) = item_capture_ix.zip(name_capture_ix) {
grammar.outline_config = Some(OutlineConfig {
query,
item_capture_ix,
name_capture_ix,
context_capture_ix,
});
}
Ok(self) Ok(self)
} }
@ -625,13 +709,13 @@ impl Language {
let query = Query::new(grammar.ts_language, source)?; let query = Query::new(grammar.ts_language, source)?;
let mut language_capture_ix = None; let mut language_capture_ix = None;
let mut content_capture_ix = None; let mut content_capture_ix = None;
for (ix, name) in query.capture_names().iter().enumerate() { get_capture_indices(
*match name.as_str() { &query,
"language" => &mut language_capture_ix, &mut [
"content" => &mut content_capture_ix, ("language", &mut language_capture_ix),
_ => continue, ("content", &mut content_capture_ix),
} = Some(ix as u32); ],
} );
let languages_by_pattern_ix = (0..query.pattern_count()) let languages_by_pattern_ix = (0..query.pattern_count())
.map(|ix| { .map(|ix| {
query.property_settings(ix).iter().find_map(|setting| { query.property_settings(ix).iter().find_map(|setting| {
@ -729,9 +813,16 @@ impl Language {
let mut result = Vec::new(); let mut result = Vec::new();
if let Some(grammar) = &self.grammar { if let Some(grammar) = &self.grammar {
let tree = grammar.parse_text(text, None); let tree = grammar.parse_text(text, None);
let captures = SyntaxSnapshot::single_tree_captures(
range.clone(),
text,
&tree,
grammar,
|grammar| grammar.highlights_query.as_ref(),
);
let highlight_maps = vec![grammar.highlight_map()];
let mut offset = 0; let mut offset = 0;
for chunk in BufferChunks::new(text, range, Some(&tree), self.grammar.as_ref(), vec![]) for chunk in BufferChunks::new(text, range, Some((captures, highlight_maps)), vec![]) {
{
let end_offset = offset + chunk.text.len(); let end_offset = offset + chunk.text.len();
if let Some(highlight_id) = chunk.syntax_highlight_id { if let Some(highlight_id) = chunk.syntax_highlight_id {
if !highlight_id.is_default() { if !highlight_id.is_default() {
@ -771,6 +862,10 @@ impl Language {
} }
impl Grammar { impl Grammar {
pub fn id(&self) -> usize {
self.id
}
fn parse_text(&self, text: &Rope, old_tree: Option<Tree>) -> Tree { fn parse_text(&self, text: &Rope, old_tree: Option<Tree>) -> Tree {
PARSER.with(|parser| { PARSER.with(|parser| {
let mut parser = parser.borrow_mut(); let mut parser = parser.borrow_mut();
@ -870,6 +965,17 @@ impl LspAdapter for Arc<FakeLspAdapter> {
} }
} }
fn get_capture_indices(query: &Query, captures: &mut [(&str, &mut Option<u32>)]) {
for (ix, name) in query.capture_names().iter().enumerate() {
for (capture_name, index) in captures.iter_mut() {
if capture_name == name {
**index = Some(ix as u32);
break;
}
}
}
}
pub fn point_to_lsp(point: PointUtf16) -> lsp::Position { pub fn point_to_lsp(point: PointUtf16) -> lsp::Position {
lsp::Position::new(point.row, point.column) lsp::Position::new(point.row, point.column)
} }

591
crates/language/src/syntax_map.rs
View file

@ -1,26 +1,28 @@
use crate::{ use crate::{Grammar, InjectionConfig, Language, LanguageRegistry};
Grammar, InjectionConfig, Language, LanguageRegistry, QueryCursorHandle, TextProvider, use lazy_static::lazy_static;
ToTreeSitterPoint, use parking_lot::Mutex;
};
use std::{ use std::{
borrow::Cow, borrow::Cow,
cell::RefCell, cell::RefCell,
cmp::{Ordering, Reverse}, cmp::{Ordering, Reverse},
collections::BinaryHeap, collections::BinaryHeap,
iter::Peekable, ops::{Deref, DerefMut, Range},
ops::{DerefMut, Range},
sync::Arc, sync::Arc,
}; };
use sum_tree::{Bias, SeekTarget, SumTree}; use sum_tree::{Bias, SeekTarget, SumTree};
use text::{Anchor, BufferSnapshot, OffsetRangeExt, Point, Rope, ToOffset, ToPoint}; use text::{rope, Anchor, BufferSnapshot, OffsetRangeExt, Point, Rope, ToOffset, ToPoint};
use tree_sitter::{ use tree_sitter::{
Node, Parser, Query, QueryCapture, QueryCaptures, QueryCursor, QueryMatch, QueryMatches, Tree, Node, Parser, Query, QueryCapture, QueryCaptures, QueryCursor, QueryMatches, Tree,
}; };
thread_local! { thread_local! {
static PARSER: RefCell<Parser> = RefCell::new(Parser::new()); static PARSER: RefCell<Parser> = RefCell::new(Parser::new());
} }
lazy_static! {
static ref QUERY_CURSORS: Mutex<Vec<QueryCursor>> = Default::default();
}
#[derive(Default)] #[derive(Default)]
pub struct SyntaxMap { pub struct SyntaxMap {
parsed_version: clock::Global, parsed_version: clock::Global,
@ -34,39 +36,51 @@ pub struct SyntaxSnapshot {
layers: SumTree<SyntaxLayer>, layers: SumTree<SyntaxLayer>,
} }
#[derive(Default)]
pub struct SyntaxMapCaptures<'a> { pub struct SyntaxMapCaptures<'a> {
layers: Vec<SyntaxMapCapturesLayer<'a>>, layers: Vec<SyntaxMapCapturesLayer<'a>>,
active_layer_count: usize,
grammars: Vec<&'a Grammar>,
} }
#[derive(Default)]
pub struct SyntaxMapMatches<'a> { pub struct SyntaxMapMatches<'a> {
layers: Vec<SyntaxMapMatchesLayer<'a>>, layers: Vec<SyntaxMapMatchesLayer<'a>>,
active_layer_count: usize,
grammars: Vec<&'a Grammar>,
} }
#[derive(Debug)]
pub struct SyntaxMapCapture<'a> { pub struct SyntaxMapCapture<'a> {
pub grammar: &'a Grammar,
pub depth: usize, pub depth: usize,
pub node: Node<'a>, pub node: Node<'a>,
pub index: u32, pub index: u32,
pub grammar_index: usize,
} }
#[derive(Debug)]
pub struct SyntaxMapMatch<'a> { pub struct SyntaxMapMatch<'a> {
pub grammar: &'a Grammar,
pub depth: usize, pub depth: usize,
pub pattern_index: usize, pub pattern_index: usize,
pub captures: &'a [QueryCapture<'a>], pub captures: &'a [QueryCapture<'a>],
pub grammar_index: usize,
} }
struct SyntaxMapCapturesLayer<'a> { struct SyntaxMapCapturesLayer<'a> {
depth: usize, depth: usize,
captures: Peekable<QueryCaptures<'a, 'a, TextProvider<'a>>>, captures: QueryCaptures<'a, 'a, TextProvider<'a>>,
grammar: &'a Grammar, next_capture: Option<QueryCapture<'a>>,
grammar_index: usize,
_query_cursor: QueryCursorHandle, _query_cursor: QueryCursorHandle,
} }
struct SyntaxMapMatchesLayer<'a> { struct SyntaxMapMatchesLayer<'a> {
depth: usize, depth: usize,
matches: Peekable<QueryMatches<'a, 'a, TextProvider<'a>>>, next_pattern_index: usize,
grammar: &'a Grammar, next_captures: Vec<QueryCapture<'a>>,
has_next: bool,
matches: QueryMatches<'a, 'a, TextProvider<'a>>,
grammar_index: usize,
_query_cursor: QueryCursorHandle, _query_cursor: QueryCursorHandle,
} }
@ -80,6 +94,7 @@ struct SyntaxLayer {
#[derive(Debug, Clone)] #[derive(Debug, Clone)]
struct SyntaxLayerSummary { struct SyntaxLayerSummary {
min_depth: usize,
max_depth: usize, max_depth: usize,
range: Range<Anchor>, range: Range<Anchor>,
last_layer_range: Range<Anchor>, last_layer_range: Range<Anchor>,
@ -110,6 +125,12 @@ struct ChangedRegion {
#[derive(Default)] #[derive(Default)]
struct ChangeRegionSet(Vec<ChangedRegion>); struct ChangeRegionSet(Vec<ChangedRegion>);
struct TextProvider<'a>(&'a Rope);
struct ByteChunks<'a>(rope::Chunks<'a>);
struct QueryCursorHandle(Option<QueryCursor>);
impl SyntaxMap { impl SyntaxMap {
pub fn new() -> Self { pub fn new() -> Self {
Self::default() Self::default()
@ -123,11 +144,20 @@ impl SyntaxMap {
self.snapshot.clone() self.snapshot.clone()
} }
pub fn language_registry(&self) -> Option<Arc<LanguageRegistry>> {
self.language_registry.clone()
}
pub fn parsed_version(&self) -> clock::Global {
self.parsed_version.clone()
}
pub fn interpolate(&mut self, text: &BufferSnapshot) { pub fn interpolate(&mut self, text: &BufferSnapshot) {
self.snapshot.interpolate(&self.interpolated_version, text); self.snapshot.interpolate(&self.interpolated_version, text);
self.interpolated_version = text.version.clone(); self.interpolated_version = text.version.clone();
} }
#[cfg(test)]
pub fn reparse(&mut self, language: Arc<Language>, text: &BufferSnapshot) { pub fn reparse(&mut self, language: Arc<Language>, text: &BufferSnapshot) {
if !self.interpolated_version.observed_all(&text.version) { if !self.interpolated_version.observed_all(&text.version) {
self.interpolate(text); self.interpolate(text);
@ -141,9 +171,22 @@ impl SyntaxMap {
); );
self.parsed_version = text.version.clone(); self.parsed_version = text.version.clone();
} }
pub fn did_parse(&mut self, snapshot: SyntaxSnapshot, version: clock::Global) {
self.parsed_version = version;
self.snapshot = snapshot;
}
pub fn clear(&mut self) {
self.snapshot = SyntaxSnapshot::default();
}
} }
impl SyntaxSnapshot { impl SyntaxSnapshot {
pub fn is_empty(&self) -> bool {
self.layers.is_empty()
}
pub fn interpolate(&mut self, from_version: &clock::Global, text: &BufferSnapshot) { pub fn interpolate(&mut self, from_version: &clock::Global, text: &BufferSnapshot) {
let edits = text let edits = text
.edits_since::<(usize, Point)>(&from_version) .edits_since::<(usize, Point)>(&from_version)
@ -429,117 +472,52 @@ impl SyntaxSnapshot {
self.layers = layers; self.layers = layers;
} }
pub fn single_tree_captures<'a>(
range: Range<usize>,
text: &'a Rope,
tree: &'a Tree,
grammar: &'a Grammar,
query: fn(&Grammar) -> Option<&Query>,
) -> SyntaxMapCaptures<'a> {
SyntaxMapCaptures::new(
range.clone(),
text,
[(grammar, 0, tree.root_node())].into_iter(),
query,
)
}
pub fn captures<'a>( pub fn captures<'a>(
&'a self, &'a self,
range: Range<usize>, range: Range<usize>,
buffer: &'a BufferSnapshot, buffer: &'a BufferSnapshot,
query: impl Fn(&Grammar) -> Option<&Query>, query: fn(&Grammar) -> Option<&Query>,
) -> SyntaxMapCaptures { ) -> SyntaxMapCaptures {
let mut result = SyntaxMapCaptures { layers: Vec::new() }; SyntaxMapCaptures::new(
for (grammar, depth, node) in self.layers_for_range(range.clone(), buffer) { range.clone(),
let query = if let Some(query) = query(grammar) { buffer.as_rope(),
query self.layers_for_range(range, buffer).into_iter(),
} else { query,
continue; )
};
let mut query_cursor = QueryCursorHandle::new();
// TODO - add a Tree-sitter API to remove the need for this.
let cursor = unsafe {
std::mem::transmute::<_, &'static mut QueryCursor>(query_cursor.deref_mut())
};
cursor.set_byte_range(range.clone());
let captures = cursor.captures(query, node, TextProvider(buffer.as_rope()));
let mut layer = SyntaxMapCapturesLayer {
depth,
grammar,
captures: captures.peekable(),
_query_cursor: query_cursor,
};
if let Some(key) = layer.sort_key() {
let mut ix = 0;
while let Some(next_layer) = result.layers.get_mut(ix) {
if let Some(next_key) = next_layer.sort_key() {
if key > next_key {
ix += 1;
continue;
}
}
break;
}
result.layers.insert(ix, layer);
}
}
result
} }
pub fn matches<'a>( pub fn matches<'a>(
&'a self, &'a self,
range: Range<usize>, range: Range<usize>,
buffer: &'a BufferSnapshot, buffer: &'a BufferSnapshot,
query: impl Fn(&Grammar) -> Option<&Query>, query: fn(&Grammar) -> Option<&Query>,
) -> SyntaxMapMatches { ) -> SyntaxMapMatches {
let mut result = SyntaxMapMatches { layers: Vec::new() }; SyntaxMapMatches::new(
for (grammar, depth, node) in self.layers_for_range(range.clone(), buffer) { range.clone(),
let query = if let Some(query) = query(grammar) { buffer.as_rope(),
query self.layers_for_range(range, buffer).into_iter(),
} else { query,
continue; )
};
let mut query_cursor = QueryCursorHandle::new();
// TODO - add a Tree-sitter API to remove the need for this.
let cursor = unsafe {
std::mem::transmute::<_, &'static mut QueryCursor>(query_cursor.deref_mut())
};
cursor.set_byte_range(range.clone());
let matches = cursor.matches(query, node, TextProvider(buffer.as_rope()));
let mut layer = SyntaxMapMatchesLayer {
depth,
grammar,
matches: matches.peekable(),
_query_cursor: query_cursor,
};
if let Some(key) = layer.sort_key() {
let mut ix = 0;
while let Some(next_layer) = result.layers.get_mut(ix) {
if let Some(next_key) = next_layer.sort_key() {
if key > next_key {
ix += 1;
continue;
}
}
break;
}
result.layers.insert(ix, layer);
}
}
result
} }
pub fn layers(&self, buffer: &BufferSnapshot) -> Vec<(&Grammar, Node)> { #[cfg(test)]
self.layers pub fn layers(&self, buffer: &BufferSnapshot) -> Vec<(&Grammar, usize, Node)> {
.iter() self.layers_for_range(0..buffer.len(), buffer)
.filter_map(|layer| {
if let Some(grammar) = &layer.language.grammar {
Some((
grammar.as_ref(),
layer.tree.root_node_with_offset(
layer.range.start.to_offset(buffer),
layer.range.start.to_point(buffer).to_ts_point(),
),
))
} else {
None
}
})
.collect()
} }
pub fn layers_for_range<'a, T: ToOffset>( pub fn layers_for_range<'a, T: ToOffset>(
@ -551,9 +529,13 @@ impl SyntaxSnapshot {
let end = buffer.anchor_after(range.end.to_offset(buffer)); let end = buffer.anchor_after(range.end.to_offset(buffer));
let mut cursor = self.layers.filter::<_, ()>(|summary| { let mut cursor = self.layers.filter::<_, ()>(|summary| {
let is_before_start = summary.range.end.cmp(&start, buffer).is_lt(); if summary.max_depth > summary.min_depth {
let is_after_end = summary.range.start.cmp(&end, buffer).is_gt(); true
!is_before_start && !is_after_end } else {
let is_before_start = summary.range.end.cmp(&start, buffer).is_lt();
let is_after_end = summary.range.start.cmp(&end, buffer).is_gt();
!is_before_start && !is_after_end
}
}); });
let mut result = Vec::new(); let mut result = Vec::new();
@ -576,57 +558,274 @@ impl SyntaxSnapshot {
} }
} }
impl<'a> Iterator for SyntaxMapCaptures<'a> { impl<'a> SyntaxMapCaptures<'a> {
type Item = SyntaxMapCapture<'a>; fn new(
range: Range<usize>,
text: &'a Rope,
layers: impl Iterator<Item = (&'a Grammar, usize, Node<'a>)>,
query: fn(&Grammar) -> Option<&Query>,
) -> Self {
let mut result = Self {
layers: Vec::new(),
grammars: Vec::new(),
active_layer_count: 0,
};
for (grammar, depth, node) in layers {
let query = if let Some(query) = query(grammar) {
query
} else {
continue;
};
fn next(&mut self) -> Option<Self::Item> { let mut query_cursor = QueryCursorHandle::new();
let layer = self.layers.first_mut()?;
let (mat, ix) = layer.captures.next()?;
let capture = mat.captures[ix as usize]; // TODO - add a Tree-sitter API to remove the need for this.
let grammar = layer.grammar; let cursor = unsafe {
let depth = layer.depth; std::mem::transmute::<_, &'static mut QueryCursor>(query_cursor.deref_mut())
};
if let Some(key) = layer.sort_key() { cursor.set_byte_range(range.clone());
let mut i = 1; let captures = cursor.captures(query, node, TextProvider(text));
while let Some(later_layer) = self.layers.get_mut(i) { let grammar_index = result
if let Some(later_key) = later_layer.sort_key() { .grammars
if key > later_key { .iter()
i += 1; .position(|g| g.id == grammar.id())
continue; .unwrap_or_else(|| {
} result.grammars.push(grammar);
} result.grammars.len() - 1
break; });
let mut layer = SyntaxMapCapturesLayer {
depth,
grammar_index,
next_capture: None,
captures,
_query_cursor: query_cursor,
};
layer.advance();
if layer.next_capture.is_some() {
let key = layer.sort_key();
let ix = match result.layers[..result.active_layer_count]
.binary_search_by_key(&key, |layer| layer.sort_key())
{
Ok(ix) | Err(ix) => ix,
};
result.layers.insert(ix, layer);
result.active_layer_count += 1;
} else {
result.layers.push(layer);
} }
if i > 1 {
self.layers[0..i].rotate_left(1);
}
} else {
self.layers.remove(0);
} }
result
}
pub fn grammars(&self) -> &[&'a Grammar] {
&self.grammars
}
pub fn peek(&self) -> Option<SyntaxMapCapture<'a>> {
let layer = self.layers[..self.active_layer_count].first()?;
let capture = layer.next_capture?;
Some(SyntaxMapCapture { Some(SyntaxMapCapture {
grammar, depth: layer.depth,
depth, grammar_index: layer.grammar_index,
node: capture.node,
index: capture.index, index: capture.index,
node: capture.node,
}) })
} }
pub fn advance(&mut self) -> bool {
let layer = if let Some(layer) = self.layers[..self.active_layer_count].first_mut() {
layer
} else {
return false;
};
layer.advance();
if layer.next_capture.is_some() {
let key = layer.sort_key();
let i = 1 + self.layers[1..self.active_layer_count]
.iter()
.position(|later_layer| key < later_layer.sort_key())
.unwrap_or(self.active_layer_count - 1);
self.layers[0..i].rotate_left(1);
} else {
self.layers[0..self.active_layer_count].rotate_left(1);
self.active_layer_count -= 1;
}
true
}
pub fn set_byte_range(&mut self, range: Range<usize>) {
for layer in &mut self.layers {
layer.captures.set_byte_range(range.clone());
if let Some(capture) = &layer.next_capture {
if capture.node.end_byte() > range.start {
continue;
}
}
layer.advance();
}
self.layers.sort_unstable_by_key(|layer| layer.sort_key());
self.active_layer_count = self
.layers
.iter()
.position(|layer| layer.next_capture.is_none())
.unwrap_or(self.layers.len());
}
}
impl<'a> SyntaxMapMatches<'a> {
fn new(
range: Range<usize>,
text: &'a Rope,
layers: impl Iterator<Item = (&'a Grammar, usize, Node<'a>)>,
query: fn(&Grammar) -> Option<&Query>,
) -> Self {
let mut result = Self::default();
for (grammar, depth, node) in layers {
let query = if let Some(query) = query(grammar) {
query
} else {
continue;
};
let mut query_cursor = QueryCursorHandle::new();
// TODO - add a Tree-sitter API to remove the need for this.
let cursor = unsafe {
std::mem::transmute::<_, &'static mut QueryCursor>(query_cursor.deref_mut())
};
cursor.set_byte_range(range.clone());
let matches = cursor.matches(query, node, TextProvider(text));
let grammar_index = result
.grammars
.iter()
.position(|g| g.id == grammar.id())
.unwrap_or_else(|| {
result.grammars.push(grammar);
result.grammars.len() - 1
});
let mut layer = SyntaxMapMatchesLayer {
depth,
grammar_index,
matches,
next_pattern_index: 0,
next_captures: Vec::new(),
has_next: false,
_query_cursor: query_cursor,
};
layer.advance();
if layer.has_next {
let key = layer.sort_key();
let ix = match result.layers[..result.active_layer_count]
.binary_search_by_key(&key, |layer| layer.sort_key())
{
Ok(ix) | Err(ix) => ix,
};
result.layers.insert(ix, layer);
result.active_layer_count += 1;
} else {
result.layers.push(layer);
}
}
result
}
pub fn grammars(&self) -> &[&'a Grammar] {
&self.grammars
}
pub fn peek(&self) -> Option<SyntaxMapMatch> {
let layer = self.layers.first()?;
if !layer.has_next {
return None;
}
Some(SyntaxMapMatch {
depth: layer.depth,
grammar_index: layer.grammar_index,
pattern_index: layer.next_pattern_index,
captures: &layer.next_captures,
})
}
pub fn advance(&mut self) -> bool {
let layer = if let Some(layer) = self.layers.first_mut() {
layer
} else {
return false;
};
layer.advance();
if layer.has_next {
let key = layer.sort_key();
let i = 1 + self.layers[1..self.active_layer_count]
.iter()
.position(|later_layer| key < later_layer.sort_key())
.unwrap_or(self.active_layer_count - 1);
self.layers[0..i].rotate_left(1);
} else {
self.layers[0..self.active_layer_count].rotate_left(1);
self.active_layer_count -= 1;
}
true
}
} }
impl<'a> SyntaxMapCapturesLayer<'a> { impl<'a> SyntaxMapCapturesLayer<'a> {
fn sort_key(&mut self) -> Option<(usize, Reverse<usize>, usize)> { fn advance(&mut self) {
let (mat, ix) = self.captures.peek()?; self.next_capture = self.captures.next().map(|(mat, ix)| mat.captures[ix]);
let range = &mat.captures[*ix].node.byte_range(); }
Some((range.start, Reverse(range.end), self.depth))
fn sort_key(&self) -> (usize, Reverse<usize>, usize) {
if let Some(capture) = &self.next_capture {
let range = capture.node.byte_range();
(range.start, Reverse(range.end), self.depth)
} else {
(usize::MAX, Reverse(0), usize::MAX)
}
} }
} }
impl<'a> SyntaxMapMatchesLayer<'a> { impl<'a> SyntaxMapMatchesLayer<'a> {
fn sort_key(&mut self) -> Option<(usize, Reverse<usize>, usize)> { fn advance(&mut self) {
let mat = self.matches.peek()?; if let Some(mat) = self.matches.next() {
let range = mat.captures.first()?.node.start_byte()..mat.captures.last()?.node.end_byte(); self.next_captures.clear();
Some((range.start, Reverse(range.end), self.depth)) self.next_captures.extend_from_slice(&mat.captures);
self.next_pattern_index = mat.pattern_index;
self.has_next = true;
} else {
self.has_next = false;
}
}
fn sort_key(&self) -> (usize, Reverse<usize>, usize) {
if self.has_next {
let captures = &self.next_captures;
if let Some((first, last)) = captures.first().zip(captures.last()) {
return (
first.node.start_byte(),
Reverse(last.node.end_byte()),
self.depth,
);
}
}
(usize::MAX, Reverse(0), usize::MAX)
}
}
impl<'a> Iterator for SyntaxMapCaptures<'a> {
type Item = SyntaxMapCapture<'a>;
fn next(&mut self) -> Option<Self::Item> {
let result = self.peek();
self.advance();
result
} }
} }
@ -864,6 +1063,7 @@ impl Default for SyntaxLayerSummary {
fn default() -> Self { fn default() -> Self {
Self { Self {
max_depth: 0, max_depth: 0,
min_depth: 0,
range: Anchor::MAX..Anchor::MIN, range: Anchor::MAX..Anchor::MIN,
last_layer_range: Anchor::MIN..Anchor::MAX, last_layer_range: Anchor::MIN..Anchor::MAX,
} }
@ -875,7 +1075,8 @@ impl sum_tree::Summary for SyntaxLayerSummary {
fn add_summary(&mut self, other: &Self, buffer: &Self::Context) { fn add_summary(&mut self, other: &Self, buffer: &Self::Context) {
if other.max_depth > self.max_depth { if other.max_depth > self.max_depth {
*self = other.clone(); self.max_depth = other.max_depth;
self.range = other.range.clone();
} else { } else {
if other.range.start.cmp(&self.range.start, buffer).is_lt() { if other.range.start.cmp(&self.range.start, buffer).is_lt() {
self.range.start = other.range.start; self.range.start = other.range.start;
@ -883,8 +1084,8 @@ impl sum_tree::Summary for SyntaxLayerSummary {
if other.range.end.cmp(&self.range.end, buffer).is_gt() { if other.range.end.cmp(&self.range.end, buffer).is_gt() {
self.range.end = other.range.end; self.range.end = other.range.end;
} }
self.last_layer_range = other.last_layer_range.clone();
} }
self.last_layer_range = other.last_layer_range.clone();
} }
} }
@ -927,6 +1128,7 @@ impl sum_tree::Item for SyntaxLayer {
fn summary(&self) -> Self::Summary { fn summary(&self) -> Self::Summary {
SyntaxLayerSummary { SyntaxLayerSummary {
min_depth: self.depth,
max_depth: self.depth, max_depth: self.depth,
range: self.range.clone(), range: self.range.clone(),
last_layer_range: self.range.clone(), last_layer_range: self.range.clone(),
@ -944,12 +1146,73 @@ impl std::fmt::Debug for SyntaxLayer {
} }
} }
impl<'a> tree_sitter::TextProvider<'a> for TextProvider<'a> {
type I = ByteChunks<'a>;
fn text(&mut self, node: tree_sitter::Node) -> Self::I {
ByteChunks(self.0.chunks_in_range(node.byte_range()))
}
}
impl<'a> Iterator for ByteChunks<'a> {
type Item = &'a [u8];
fn next(&mut self) -> Option<Self::Item> {
self.0.next().map(str::as_bytes)
}
}
impl QueryCursorHandle {
pub(crate) fn new() -> Self {
let mut cursor = QUERY_CURSORS.lock().pop().unwrap_or_else(QueryCursor::new);
cursor.set_match_limit(64);
QueryCursorHandle(Some(cursor))
}
}
impl Deref for QueryCursorHandle {
type Target = QueryCursor;
fn deref(&self) -> &Self::Target {
self.0.as_ref().unwrap()
}
}
impl DerefMut for QueryCursorHandle {
fn deref_mut(&mut self) -> &mut Self::Target {
self.0.as_mut().unwrap()
}
}
impl Drop for QueryCursorHandle {
fn drop(&mut self) {
let mut cursor = self.0.take().unwrap();
cursor.set_byte_range(0..usize::MAX);
cursor.set_point_range(Point::zero().to_ts_point()..Point::MAX.to_ts_point());
QUERY_CURSORS.lock().push(cursor)
}
}
pub(crate) trait ToTreeSitterPoint {
fn to_ts_point(self) -> tree_sitter::Point;
fn from_ts_point(point: tree_sitter::Point) -> Self;
}
impl ToTreeSitterPoint for Point {
fn to_ts_point(self) -> tree_sitter::Point {
tree_sitter::Point::new(self.row as usize, self.column as usize)
}
fn from_ts_point(point: tree_sitter::Point) -> Self {
Point::new(point.row as u32, point.column as u32)
}
}
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::*; use super::*;
use crate::LanguageConfig; use crate::LanguageConfig;
use text::{Buffer, Point}; use text::{Buffer, Point};
use tree_sitter::Query;
use unindent::Unindent as _; use unindent::Unindent as _;
use util::test::marked_text_ranges; use util::test::marked_text_ranges;
@ -1298,13 +1561,13 @@ mod tests {
mutated_layers.into_iter().zip(reference_layers.into_iter()) mutated_layers.into_iter().zip(reference_layers.into_iter())
{ {
assert_eq!( assert_eq!(
edited_layer.1.to_sexp(), edited_layer.2.to_sexp(),
reference_layer.1.to_sexp(), reference_layer.2.to_sexp(),
"different layer at step {i}" "different layer at step {i}"
); );
assert_eq!( assert_eq!(
edited_layer.1.range(), edited_layer.2.range(),
reference_layer.1.range(), reference_layer.2.range(),
"different layer at step {i}" "different layer at step {i}"
); );
} }
@ -1377,16 +1640,16 @@ mod tests {
marked_string: &str, marked_string: &str,
) { ) {
let mut actual_ranges = Vec::<Range<usize>>::new(); let mut actual_ranges = Vec::<Range<usize>>::new();
for capture in syntax_map.captures(0..buffer.len(), buffer, |grammar| { let captures = syntax_map.captures(0..buffer.len(), buffer, |grammar| {
grammar.highlights_query.as_ref() grammar.highlights_query.as_ref()
}) { });
let name = &capture let queries = captures
.grammar .grammars()
.highlights_query .iter()
.as_ref() .map(|grammar| grammar.highlights_query.as_ref().unwrap())
.unwrap() .collect::<Vec<_>>();
.capture_names()[capture.index as usize]; for capture in captures {
dbg!(capture.node, capture.index, name); let name = &queries[capture.grammar_index].capture_names()[capture.index as usize];
if highlight_query_capture_names.contains(&name.as_str()) { if highlight_query_capture_names.contains(&name.as_str()) {
actual_ranges.push(capture.node.byte_range()); actual_ranges.push(capture.node.byte_range());
} }

4
crates/language/src/tests.rs
View file

@ -1407,7 +1407,9 @@ fn json_lang() -> Language {
fn get_tree_sexp(buffer: &ModelHandle<Buffer>, cx: &gpui::TestAppContext) -> String { fn get_tree_sexp(buffer: &ModelHandle<Buffer>, cx: &gpui::TestAppContext) -> String {
buffer.read_with(cx, |buffer, _| { buffer.read_with(cx, |buffer, _| {
buffer.syntax_tree().unwrap().root_node().to_sexp() let syntax_map = buffer.syntax_map();
let layers = syntax_map.layers(buffer.as_text_snapshot());
layers[0].2.to_sexp()
}) })
} }

1
crates/project/src/project.rs
View file

@ -2056,6 +2056,7 @@ impl Project {
let full_path = buffer.read(cx).file()?.full_path(cx); let full_path = buffer.read(cx).file()?.full_path(cx);
let language = self.languages.select_language(&full_path)?; let language = self.languages.select_language(&full_path)?;
buffer.update(cx, |buffer, cx| { buffer.update(cx, |buffer, cx| {
buffer.set_language_registry(self.languages.clone());
buffer.set_language(Some(language.clone()), cx); buffer.set_language(Some(language.clone()), cx);
}); });

6
crates/zed/src/languages/rust/injections.scm
View file

@ -1,3 +1,7 @@
(macro_invocation (macro_invocation
(token_tree) @content) (token_tree) @content
(#set! "language" "rust"))
(macro_rule
(token_tree) @content
(#set! "language" "rust")) (#set! "language" "rust"))