tmp

Cargo.lock

@@ -1311,6 +1311,7 @@ dependencies = [
  "hashbrown 0.14.5",
  "helix-loader",
  "helix-stdx",
+ "helix-syntax",
  "imara-diff",
  "indoc",
  "log",
@@ -1425,6 +1426,27 @@ dependencies = [
  "windows-sys 0.52.0",
 ]
 
+[[package]]
+name = "helix-syntax"
+version = "24.7.0"
+dependencies = [
+ "ahash",
+ "arc-swap",
+ "bitflags 2.6.0",
+ "cc",
+ "hashbrown 0.14.5",
+ "helix-stdx",
+ "libloading",
+ "log",
+ "once_cell",
+ "regex",
+ "regex-cursor",
+ "ropey",
+ "slotmap",
+ "thiserror",
+ "tree-sitter",
+]
+
 [[package]]
 name = "helix-term"
 version = "24.7.0"

Cargo.toml

@@ -12,6 +12,7 @@ members = [
   "helix-vcs",
   "helix-parsec",
   "helix-stdx",
+  "helix-syntax",
   "xtask",
 ]
 

helix-core/Cargo.toml

@@ -17,6 +17,7 @@ integration = []
 
 [dependencies]
 helix-stdx = { path = "../helix-stdx" }
+helix-syntax = { path = "../helix-syntax" }
 helix-loader = { path = "../helix-loader" }
 
 ropey = { version = "1.6.1", default-features = false, features = ["simd"] }

helix-core/src/indent.rs

@@ -738,18 +738,9 @@ fn init_indent_query<'a, 'b>(
             .map(|prec| prec.byte_range().end - 1..byte_pos + 1)
             .unwrap_or(byte_pos..byte_pos + 1);
 
-        crate::syntax::PARSER.with(|ts_parser| {
-            let mut ts_parser = ts_parser.borrow_mut();
-            let mut cursor = ts_parser.cursors.pop().unwrap_or_default();
-            let query_result = query_indents(
-                query,
-                syntax,
-                &mut cursor,
-                text,
-                query_range,
-                new_line_byte_pos,
-            );
-            ts_parser.cursors.push(cursor);
+        crate::syntax::with_cursor(|cursor| {
+            let query_result =
+                query_indents(query, syntax, cursor, text, query_range, new_line_byte_pos);
             (query_result, deepest_preceding)
         })
     };

helix-core/tests/indent.rs

@@ -1,4 +1,3 @@
-use arc_swap::ArcSwap;
 use helix_core::{
     indent::{indent_level_for_line, treesitter_indent_for_pos, IndentStyle},
     syntax::{Configuration, Loader},
@@ -6,7 +5,7 @@ use helix_core::{
 };
 use helix_stdx::rope::RopeSliceExt;
 use ropey::Rope;
-use std::{ops::Range, path::PathBuf, process::Command, sync::Arc};
+use std::{ops::Range, path::PathBuf, process::Command};
 
 #[test]
 fn test_treesitter_indent_rust() {
@@ -200,12 +199,7 @@ fn test_treesitter_indent(
     let indent_style = IndentStyle::from_str(&language_config.indent.as_ref().unwrap().unit);
     let highlight_config = language_config.highlight_config(&[]).unwrap();
     let text = doc.slice(..);
-    let syntax = Syntax::new(
-        text,
-        highlight_config,
-        Arc::new(ArcSwap::from_pointee(loader)),
-    )
-    .unwrap();
+    let syntax = Syntax::new(text, highlight_config, |_| None).unwrap();
     let indent_query = language_config.indent_query().unwrap();
 
     for i in 0..doc.len_lines() {

helix-syntax/Cargo.toml

@@ -0,0 +1,34 @@
+[package]
+name = "helix-syntax"
+description = "Helix syntax highlighting "
+include = ["src/**/*", "README.md"]
+version.workspace = true
+authors.workspace = true
+edition.workspace = true
+license.workspace = true
+rust-version.workspace = true
+categories.workspace = true
+repository.workspace = true
+homepage.workspace = true
+
+[features]
+
+[dependencies]
+helix-stdx = { path = "../helix-stdx" }
+
+ropey = { version = "1.6.1", default-features = false, features = ["simd"] }
+slotmap = "1.0"
+tree-sitter.workspace = true
+once_cell = "1.19"
+arc-swap = "1"
+regex = "1"
+bitflags = "2.4"
+ahash = "0.8.9"
+hashbrown = { version = "0.14.3", features = ["raw"] }
+log = "0.4"
+regex-cursor = "0.1.4"
+libloading = "0.8.3"
+thiserror = "1.0.59"
+
+[build-dependencies]
+cc = "1.0.95"

helix-syntax/build.rs

@@ -0,0 +1,28 @@
+use std::path::PathBuf;
+use std::{env, fs};
+
+fn main() {
+    if env::var_os("DISABLED_TS_BUILD").is_some() {
+        return;
+    }
+    let mut config = cc::Build::new();
+
+    let manifest_path = PathBuf::from(env::var_os("CARGO_MANIFEST_DIR").unwrap());
+    let include_path = manifest_path.join("../vendor/tree-sitter/include");
+    let src_path = manifest_path.join("../vendor/tree-sitter/src");
+    for entry in fs::read_dir(&src_path).unwrap() {
+        let entry = entry.unwrap();
+        let path = src_path.join(entry.file_name());
+        println!("cargo:rerun-if-changed={}", path.to_str().unwrap());
+    }
+
+    config
+        .flag_if_supported("-std=c11")
+        .flag_if_supported("-fvisibility=hidden")
+        .flag_if_supported("-Wshadow")
+        .flag_if_supported("-Wno-unused-parameter")
+        .include(&src_path)
+        .include(&include_path)
+        .file(src_path.join("lib.c"))
+        .compile("tree-sitter");
+}

helix-syntax/src/config.rs

@@ -0,0 +1,346 @@
+use std::borrow::Cow;
+use std::path::Path;
+use std::sync::Arc;
+
+use crate::tree_sitter::query::{Capture, Pattern, QueryStr, UserPredicate};
+use crate::tree_sitter::{query, Grammar, Query, QueryMatch, SyntaxTreeNode};
+use arc_swap::ArcSwap;
+use helix_stdx::rope::{self, RopeSliceExt};
+use once_cell::sync::Lazy;
+use regex::Regex;
+use ropey::RopeSlice;
+
+use crate::byte_range_to_str;
+use crate::highlighter::Highlight;
+
+/// Contains the data needed to highlight code written in a particular language.
+///
+/// This struct is immutable and can be shared between threads.
+#[derive(Debug)]
+pub struct HighlightConfiguration {
+    pub grammar: Grammar,
+    pub query: Query,
+    pub(crate) injections_query: Query,
+    pub(crate) combined_injections_patterns: Vec<Pattern>,
+    first_highlights_pattern: Pattern,
+    pub(crate) highlight_indices: ArcSwap<Vec<Highlight>>,
+    pub(crate) non_local_variable_patterns: Vec<bool>,
+    pub(crate) injection_content_capture: Option<Capture>,
+    pub(crate) injection_language_capture: Option<Capture>,
+    pub(crate) injection_filename_capture: Option<Capture>,
+    pub(crate) injection_shebang_capture: Option<Capture>,
+    pub(crate) local_scope_capture: Option<Capture>,
+    pub(crate) local_def_capture: Option<Capture>,
+    pub(crate) local_def_value_capture: Option<Capture>,
+    pub(crate) local_ref_capture: Option<Capture>,
+}
+
+impl HighlightConfiguration {
+    /// Creates a `HighlightConfiguration` for a given `Grammar` and set of highlighting
+    /// queries.
+    ///
+    /// # Parameters
+    ///
+    /// * `language`  - The Tree-sitter `Grammar` that should be used for parsing.
+    /// * `highlights_query` - A string containing tree patterns for syntax highlighting. This
+    ///   should be non-empty, otherwise no syntax highlights will be added.
+    /// * `injections_query` -  A string containing tree patterns for injecting other languages
+    ///   into the document. This can be empty if no injections are desired.
+    /// * `locals_query` - A string containing tree patterns for tracking local variable
+    ///   definitions and references. This can be empty if local variable tracking is not needed.
+    ///
+    /// Returns a `HighlightConfiguration` that can then be used with the `highlight` method.
+    pub fn new(
+        grammar: Grammar,
+        path: impl AsRef<Path>,
+        highlights_query: &str,
+        injection_query: &str,
+        locals_query: &str,
+    ) -> Result<Self, query::ParseError> {
+        // Concatenate the query strings, keeping track of the start offset of each section.
+        let mut query_source = String::new();
+        query_source.push_str(locals_query);
+        let highlights_query_offset = query_source.len();
+        query_source.push_str(highlights_query);
+
+        let mut non_local_variable_patterns = Vec::with_capacity(32);
+        // Construct a single query by concatenating the three query strings, but record the
+        // range of pattern indices that belong to each individual string.
+        let query = Query::new(grammar, &query_source, path, |pattern, predicate| {
+            match predicate {
+                UserPredicate::IsPropertySet {
+                    negate: true,
+                    key: "local",
+                    val: None,
+                } => {
+                    if non_local_variable_patterns.len() < pattern.idx() {
+                        non_local_variable_patterns.resize(pattern.idx(), false)
+                    }
+                    non_local_variable_patterns[pattern.idx()] = true;
+                }
+                predicate => {
+                    return Err(format!("unsupported predicate {predicate}").into());
+                }
+            }
+            Ok(())
+        })?;
+
+        let mut combined_injections_patterns = Vec::new();
+        let injections_query = Query::new(grammar, injection_query, path, |pattern, predicate| {
+            match predicate {
+                UserPredicate::SetProperty {
+                    key: "injection.combined",
+                    val: None,
+                } => combined_injections_patterns.push(pattern),
+                predicate => {
+                    return Err(format!("unsupported predicate {predicate}").into());
+                }
+            }
+            Ok(())
+        })?;
+
+        let first_highlights_pattern = query
+            .patterns()
+            .find(|pattern| query.start_byte_for_pattern(*pattern) >= highlights_query_offset)
+            .unwrap_or(Pattern::SENTINEL);
+
+        let injection_content_capture = query.get_capture("injection.content");
+        let injection_language_capture = query.get_capture("injection.language");
+        let injection_filename_capture = query.get_capture("injection.filename");
+        let injection_shebang_capture = query.get_capture("injection.shebang");
+        let local_def_capture = query.get_capture("local.definition");
+        let local_def_value_capture = query.get_capture("local.definition-value");
+        let local_ref_capture = query.get_capture("local.reference");
+        let local_scope_capture = query.get_capture("local.scope");
+
+        let highlight_indices =
+            ArcSwap::from_pointee(vec![Highlight::NONE; query.num_captures() as usize]);
+        Ok(Self {
+            grammar,
+            query,
+            injections_query,
+            combined_injections_patterns,
+            first_highlights_pattern,
+            highlight_indices,
+            non_local_variable_patterns,
+            injection_content_capture,
+            injection_language_capture,
+            injection_filename_capture,
+            injection_shebang_capture,
+            local_scope_capture,
+            local_def_capture,
+            local_def_value_capture,
+            local_ref_capture,
+        })
+    }
+
+    /// Set the list of recognized highlight names.
+    ///
+    /// Tree-sitter syntax-highlighting queries specify highlights in the form of dot-separated
+    /// highlight names like `punctuation.bracket` and `function.method.builtin`. Consumers of
+    /// these queries can choose to recognize highlights with different levels of specificity.
+    /// For example, the string `function.builtin` will match against `function.builtin.constructor`
+    /// but will not match `function.method.builtin` and `function.method`.
+    ///
+    /// When highlighting, results are returned as `Highlight` values, which contain the index
+    /// of the matched highlight this list of highlight names.
+    pub fn configure(&self, recognized_names: &[String]) {
+        let mut capture_parts = Vec::new();
+        let indices: Vec<_> = self
+            .query
+            .captures()
+            .map(move |(_, capture_name)| {
+                capture_parts.clear();
+                capture_parts.extend(capture_name.split('.'));
+
+                let mut best_index = u32::MAX;
+                let mut best_match_len = 0;
+                for (i, recognized_name) in recognized_names.iter().enumerate() {
+                    let mut len = 0;
+                    let mut matches = true;
+                    for (i, part) in recognized_name.split('.').enumerate() {
+                        match capture_parts.get(i) {
+                            Some(capture_part) if *capture_part == part => len += 1,
+                            _ => {
+                                matches = false;
+                                break;
+                            }
+                        }
+                    }
+                    if matches && len > best_match_len {
+                        best_index = i as u32;
+                        best_match_len = len;
+                    }
+                }
+                Highlight(best_index)
+            })
+            .collect();
+
+        self.highlight_indices.store(Arc::new(indices));
+    }
+
+    fn injection_pair<'a>(
+        &self,
+        query_match: &QueryMatch<'a, 'a>,
+        source: RopeSlice<'a>,
+    ) -> (
+        Option<InjectionLanguageMarker<'a>>,
+        Option<SyntaxTreeNode<'a>>,
+    ) {
+        let mut injection_capture = None;
+        let mut content_node = None;
+
+        for matched_node in query_match.matched_nodes() {
+            let capture = Some(matched_node.capture);
+            if capture == self.injection_language_capture {
+                let name = byte_range_to_str(matched_node.syntax_node.byte_range(), source);
+                injection_capture = Some(InjectionLanguageMarker::Name(name));
+            } else if capture == self.injection_filename_capture {
+                let name = byte_range_to_str(matched_node.syntax_node.byte_range(), source);
+                let path = Path::new(name.as_ref()).to_path_buf();
+                injection_capture = Some(InjectionLanguageMarker::Filename(path.into()));
+            } else if capture == self.injection_shebang_capture {
+                let node_slice = source.byte_slice(matched_node.syntax_node.byte_range());
+
+                // some languages allow space and newlines before the actual string content
+                // so a shebang could be on either the first or second line
+                let lines = if let Ok(end) = node_slice.try_line_to_byte(2) {
+                    node_slice.byte_slice(..end)
+                } else {
+                    node_slice
+                };
+
+                injection_capture = SHEBANG_REGEX
+                    .captures_iter(lines.regex_input())
+                    .map(|cap| {
+                        let cap = lines.byte_slice(cap.get_group(1).unwrap().range());
+                        InjectionLanguageMarker::Shebang(cap.into())
+                    })
+                    .next()
+            } else if capture == self.injection_content_capture {
+                content_node = Some(matched_node.syntax_node.clone());
+            }
+        }
+        (injection_capture, content_node)
+    }
+
+    pub(super) fn injection_for_match<'a>(
+        &self,
+        query: &'a Query,
+        query_match: &QueryMatch<'a, 'a>,
+        source: RopeSlice<'a>,
+    ) -> (
+        Option<InjectionLanguageMarker<'a>>,
+        Option<SyntaxTreeNode<'a>>,
+        IncludedChildren,
+    ) {
+        let (mut injection_capture, content_node) = self.injection_pair(query_match, source);
+
+        let mut included_children = IncludedChildren::default();
+        for prop in query.property_settings(query_match.pattern_index) {
+            match prop.key.as_ref() {
+                // In addition to specifying the language name via the text of a
+                // captured node, it can also be hard-coded via a `#set!` predicate
+                // that sets the injection.language key.
+                "injection.language" if injection_capture.is_none() => {
+                    injection_capture = prop
+                        .value
+                        .as_ref()
+                        .map(|s| InjectionLanguageMarker::Name(s.as_ref().into()));
+                }
+
+                // By default, injections do not include the *children* of an
+                // `injection.content` node - only the ranges that belong to the
+                // node itself. This can be changed using a `#set!` predicate that
+                // sets the `injection.include-children` key.
+                "injection.include-children" => included_children = IncludedChildren::All,
+
+                // Some queries might only exclude named children but include unnamed
+                // children in their `injection.content` node. This can be enabled using
+                // a `#set!` predicate that sets the `injection.include-unnamed-children` key.
+                "injection.include-unnamed-children" => {
+                    included_children = IncludedChildren::Unnamed
+                }
+                _ => {}
+            }
+        }
+
+        (injection_capture, content_node, included_children)
+    }
+
+    // pub fn load_query(
+    //     &self,
+    //     language: &str,
+    //     filename: &str,
+    //     read_query_text: impl FnMut(&str, &str) -> String,
+    // ) -> Result<Option<Query>, QueryError> {
+    //     let query_text = read_query(language, filename, read_query_text);
+    //     if query_text.is_empty() {
+    //         return Ok(None);
+    //     }
+
+    //     Query::new(&self.grammar, &query_text, ).map(Some)
+    // }
+}
+
+/// reads a query by invoking `read_query_text`, handeles any `inherits` directives
+pub fn read_query(
+    language: &str,
+    filename: &str,
+    mut read_query_text: impl FnMut(&str, &str) -> String,
+) -> String {
+    fn read_query_impl(
+        language: &str,
+        filename: &str,
+        read_query_text: &mut impl FnMut(&str, &str) -> String,
+    ) -> String {
+        static INHERITS_REGEX: Lazy<Regex> =
+            Lazy::new(|| Regex::new(r";+\s*inherits\s*:?\s*([a-z_,()-]+)\s*").unwrap());
+
+        let query = read_query_text(language, filename);
+
+        // replaces all "; inherits <language>(,<language>)*" with the queries of the given language(s)
+        INHERITS_REGEX
+            .replace_all(&query, |captures: &regex::Captures| {
+                captures[1]
+                    .split(',')
+                    .map(|language| {
+                        format!(
+                            "\n{}\n",
+                            read_query_impl(language, filename, &mut *read_query_text)
+                        )
+                    })
+                    .collect::<String>()
+            })
+            .to_string()
+    }
+    read_query_impl(language, filename, &mut read_query_text)
+}
+
+const SHEBANG: &str = r"#!\s*(?:\S*[/\\](?:env\s+(?:\-\S+\s+)*)?)?([^\s\.\d]+)";
+static SHEBANG_REGEX: Lazy<rope::Regex> = Lazy::new(|| rope::Regex::new(SHEBANG).unwrap());
+
+struct InjectionSettings {
+    include_children: IncludedChildren,
+    language: Option<QueryStr>,
+}
+
+#[derive(Debug, Clone)]
+pub enum InjectionLanguageMarker<'a> {
+    Name(Cow<'a, str>),
+    Filename(Cow<'a, Path>),
+    Shebang(String),
+}
+
+#[derive(Clone)]
+enum IncludedChildren {
+    None,
+    All,
+    Unnamed,
+}
+
+impl Default for IncludedChildren {
+    fn default() -> Self {
+        Self::None
+    }
+}

helix-syntax/src/highlighter.rs

@@ -0,0 +1,438 @@
+pub use super::highlighter2::*;
+
+// use std::borrow::Cow;
+// use std::cell::RefCell;
+// use std::sync::atomic::{self, AtomicUsize};
+// use std::{fmt, iter, mem, ops};
+
+// use ropey::RopeSlice;
+// use tree_sitter::{QueryCaptures, QueryCursor, Tree};
+
+// use crate::{byte_range_to_str, Error, HighlightConfiguration, Syntax, TREE_SITTER_MATCH_LIMIT};
+
+// const CANCELLATION_CHECK_INTERVAL: usize = 100;
+
+// /// Indicates which highlight should be applied to a region of source code.
+// #[derive(Copy, Clone, Debug, PartialEq, Eq)]
+// pub struct Highlight(pub usize);
+
+// /// Represents a single step in rendering a syntax-highlighted document.
+// #[derive(Copy, Clone, Debug)]
+// pub enum HighlightEvent {
+//     Source { start: usize, end: usize },
+//     HighlightStart(Highlight),
+//     HighlightEnd,
+// }
+
+// #[derive(Debug)]
+// struct LocalDef<'a> {
+//     name: Cow<'a, str>,
+//     value_range: ops::Range<usize>,
+//     highlight: Option<Highlight>,
+// }
+
+// #[derive(Debug)]
+// struct LocalScope<'a> {
+//     inherits: bool,
+//     range: ops::Range<usize>,
+//     local_defs: Vec<LocalDef<'a>>,
+// }
+
+// #[derive(Debug)]
+// struct HighlightIter<'a> {
+//     source: RopeSlice<'a>,
+//     byte_offset: usize,
+//     cancellation_flag: Option<&'a AtomicUsize>,
+//     layers: Vec<HighlightIterLayer<'a>>,
+//     iter_count: usize,
+//     next_event: Option<HighlightEvent>,
+//     last_highlight_range: Option<(usize, usize, u32)>,
+// }
+
+// struct HighlightIterLayer<'a> {
+//     _tree: Option<Tree>,
+//     cursor: QueryCursor,
+//     captures: RefCell<iter::Peekable<QueryCaptures<'a, 'a, RopeProvider<'a>, &'a [u8]>>>,
+//     config: &'a HighlightConfiguration,
+//     highlight_end_stack: Vec<usize>,
+//     scope_stack: Vec<LocalScope<'a>>,
+//     depth: u32,
+// }
+
+// impl<'a> fmt::Debug for HighlightIterLayer<'a> {
+//     fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+//         f.debug_struct("HighlightIterLayer").finish()
+//     }
+// }
+
+// impl<'a> HighlightIterLayer<'a> {
+//     // First, sort scope boundaries by their byte offset in the document. At a
+//     // given position, emit scope endings before scope beginnings. Finally, emit
+//     // scope boundaries from deeper layers first.
+//     fn sort_key(&self) -> Option<(usize, bool, isize)> {
+//         let depth = -(self.depth as isize);
+//         let next_start = self
+//             .captures
+//             .borrow_mut()
+//             .peek()
+//             .map(|(m, i)| m.captures[*i].node.start_byte());
+//         let next_end = self.highlight_end_stack.last().cloned();
+//         match (next_start, next_end) {
+//             (Some(start), Some(end)) => {
+//                 if start < end {
+//                     Some((start, true, depth))
+//                 } else {
+//                     Some((end, false, depth))
+//                 }
+//             }
+//             (Some(i), None) => Some((i, true, depth)),
+//             (None, Some(j)) => Some((j, false, depth)),
+//             _ => None,
+//         }
+//     }
+// }
+
+// impl<'a> HighlightIter<'a> {
+//     fn emit_event(
+//         &mut self,
+//         offset: usize,
+//         event: Option<HighlightEvent>,
+//     ) -> Option<Result<HighlightEvent, Error>> {
+//         let result;
+//         if self.byte_offset < offset {
+//             result = Some(Ok(HighlightEvent::Source {
+//                 start: self.byte_offset,
+//                 end: offset,
+//             }));
+//             self.byte_offset = offset;
+//             self.next_event = event;
+//         } else {
+//             result = event.map(Ok);
+//         }
+//         self.sort_layers();
+//         result
+//     }
+
+//     fn sort_layers(&mut self) {
+//         while !self.layers.is_empty() {
+//             if let Some(sort_key) = self.layers[0].sort_key() {
+//                 let mut i = 0;
+//                 while i + 1 < self.layers.len() {
+//                     if let Some(next_offset) = self.layers[i + 1].sort_key() {
+//                         if next_offset < sort_key {
+//                             i += 1;
+//                             continue;
+//                         }
+//                     } else {
+//                         let layer = self.layers.remove(i + 1);
+//                         PARSER.with(|ts_parser| {
+//                             let highlighter = &mut ts_parser.borrow_mut();
+//                             highlighter.cursors.push(layer.cursor);
+//                         });
+//                     }
+//                     break;
+//                 }
+//                 if i > 0 {
+//                     self.layers[0..(i + 1)].rotate_left(1);
+//                 }
+//                 break;
+//             } else {
+//                 let layer = self.layers.remove(0);
+//                 PARSER.with(|ts_parser| {
+//                     let highlighter = &mut ts_parser.borrow_mut();
+//                     highlighter.cursors.push(layer.cursor);
+//                 });
+//             }
+//         }
+//     }
+// }
+
+// impl<'a> Iterator for HighlightIter<'a> {
+//     type Item = Result<HighlightEvent, Error>;
+
+//     fn next(&mut self) -> Option<Self::Item> {
+//         'main: loop {
+//             // If we've already determined the next highlight boundary, just return it.
+//             if let Some(e) = self.next_event.take() {
+//                 return Some(Ok(e));
+//             }
+
+//             // Periodically check for cancellation, returning `Cancelled` error if the
+//             // cancellation flag was flipped.
+//             if let Some(cancellation_flag) = self.cancellation_flag {
+//                 self.iter_count += 1;
+//                 if self.iter_count >= CANCELLATION_CHECK_INTERVAL {
+//                     self.iter_count = 0;
+//                     if cancellation_flag.load(atomic::Ordering::Relaxed) != 0 {
+//                         return Some(Err(Error::Cancelled));
+//                     }
+//                 }
+//             }
+
+//             // If none of the layers have any more highlight boundaries, terminate.
+//             if self.layers.is_empty() {
+//                 let len = self.source.len_bytes();
+//                 return if self.byte_offset < len {
+//                     let result = Some(Ok(HighlightEvent::Source {
+//                         start: self.byte_offset,
+//                         end: len,
+//                     }));
+//                     self.byte_offset = len;
+//                     result
+//                 } else {
+//                     None
+//                 };
+//             }
+
+//             // Get the next capture from whichever layer has the earliest highlight boundary.
+//             let range;
+//             let layer = &mut self.layers[0];
+//             let captures = layer.captures.get_mut();
+//             if let Some((next_match, capture_index)) = captures.peek() {
+//                 let next_capture = next_match.captures[*capture_index];
+//                 range = next_capture.node.byte_range();
+
+//                 // If any previous highlight ends before this node starts, then before
+//                 // processing this capture, emit the source code up until the end of the
+//                 // previous highlight, and an end event for that highlight.
+//                 if let Some(end_byte) = layer.highlight_end_stack.last().cloned() {
+//                     if end_byte <= range.start {
+//                         layer.highlight_end_stack.pop();
+//                         return self.emit_event(end_byte, Some(HighlightEvent::HighlightEnd));
+//                     }
+//                 }
+//             }
+//             // If there are no more captures, then emit any remaining highlight end events.
+//             // And if there are none of those, then just advance to the end of the document.
+//             else if let Some(end_byte) = layer.highlight_end_stack.last().cloned() {
+//                 layer.highlight_end_stack.pop();
+//                 return self.emit_event(end_byte, Some(HighlightEvent::HighlightEnd));
+//             } else {
+//                 return self.emit_event(self.source.len_bytes(), None);
+//             };
+
+//             let (mut match_, capture_index) = captures.next().unwrap();
+//             let mut capture = match_.captures[capture_index];
+
+//             // Remove from the local scope stack any local scopes that have already ended.
+//             while range.start > layer.scope_stack.last().unwrap().range.end {
+//                 layer.scope_stack.pop();
+//             }
+
+//             // If this capture is for tracking local variables, then process the
+//             // local variable info.
+//             let mut reference_highlight = None;
+//             let mut definition_highlight = None;
+//             while match_.pattern_index < layer.config.highlights_pattern_index {
+//                 // If the node represents a local scope, push a new local scope onto
+//                 // the scope stack.
+//                 if Some(capture.index) == layer.config.local_scope_capture_index {
+//                     definition_highlight = None;
+//                     let mut scope = LocalScope {
+//                         inherits: true,
+//                         range: range.clone(),
+//                         local_defs: Vec::new(),
+//                     };
+//                     for prop in layer.config.query.property_settings(match_.pattern_index) {
+//                         if let "local.scope-inherits" = prop.key.as_ref() {
+//                             scope.inherits =
+//                                 prop.value.as_ref().map_or(true, |r| r.as_ref() == "true");
+//                         }
+//                     }
+//                     layer.scope_stack.push(scope);
+//                 }
+//                 // If the node represents a definition, add a new definition to the
+//                 // local scope at the top of the scope stack.
+//                 else if Some(capture.index) == layer.config.local_def_capture_index {
+//                     reference_highlight = None;
+//                     let scope = layer.scope_stack.last_mut().unwrap();
+
+//                     let mut value_range = 0..0;
+//                     for capture in match_.captures {
+//                         if Some(capture.index) == layer.config.local_def_value_capture_index {
+//                             value_range = capture.node.byte_range();
+//                         }
+//                     }
+
+//                     let name = byte_range_to_str(range.clone(), self.source);
+//                     scope.local_defs.push(LocalDef {
+//                         name,
+//                         value_range,
+//                         highlight: None,
+//                     });
+//                     definition_highlight = scope.local_defs.last_mut().map(|s| &mut s.highlight);
+//                 }
+//                 // If the node represents a reference, then try to find the corresponding
+//                 // definition in the scope stack.
+//                 else if Some(capture.index) == layer.config.local_ref_capture_index
+//                     && definition_highlight.is_none()
+//                 {
+//                     definition_highlight = None;
+//                     let name = byte_range_to_str(range.clone(), self.source);
+//                     for scope in layer.scope_stack.iter().rev() {
+//                         if let Some(highlight) = scope.local_defs.iter().rev().find_map(|def| {
+//                             if def.name == name && range.start >= def.value_range.end {
+//                                 Some(def.highlight)
+//                             } else {
+//                                 None
+//                             }
+//                         }) {
+//                             reference_highlight = highlight;
+//                             break;
+//                         }
+//                         if !scope.inherits {
+//                             break;
+//                         }
+//                     }
+//                 }
+
+//                 // Continue processing any additional matches for the same node.
+//                 if let Some((next_match, next_capture_index)) = captures.peek() {
+//                     let next_capture = next_match.captures[*next_capture_index];
+//                     if next_capture.node == capture.node {
+//                         capture = next_capture;
+//                         match_ = captures.next().unwrap().0;
+//                         continue;
+//                     }
+//                 }
+
+//                 self.sort_layers();
+//                 continue 'main;
+//             }
+
+//             // Otherwise, this capture must represent a highlight.
+//             // If this exact range has already been highlighted by an earlier pattern, or by
+//             // a different layer, then skip over this one.
+//             if let Some((last_start, last_end, last_depth)) = self.last_highlight_range {
+//                 if range.start == last_start && range.end == last_end && layer.depth < last_depth {
+//                     self.sort_layers();
+//                     continue 'main;
+//                 }
+//             }
+
+//             // If the current node was found to be a local variable, then skip over any
+//             // highlighting patterns that are disabled for local variables.
+//             if definition_highlight.is_some() || reference_highlight.is_some() {
+//                 while layer.config.non_local_variable_patterns[match_.pattern_index] {
+//                     match_.remove();
+//                     if let Some((next_match, next_capture_index)) = captures.peek() {
+//                         let next_capture = next_match.captures[*next_capture_index];
+//                         if next_capture.node == capture.node {
+//                             capture = next_capture;
+//                             match_ = captures.next().unwrap().0;
+//                             continue;
+//                         }
+//                     }
+
+//                     self.sort_layers();
+//                     continue 'main;
+//                 }
+//             }
+
+//             // Once a highlighting pattern is found for the current node, skip over
+//             // any later highlighting patterns that also match this node. Captures
+//             // for a given node are ordered by pattern index, so these subsequent
+//             // captures are guaranteed to be for highlighting, not injections or
+//             // local variables.
+//             while let Some((next_match, next_capture_index)) = captures.peek() {
+//                 let next_capture = next_match.captures[*next_capture_index];
+//                 if next_capture.node == capture.node {
+//                     captures.next();
+//                 } else {
+//                     break;
+//                 }
+//             }
+
+//             let current_highlight = layer.config.highlight_indices.load()[capture.index as usize];
+
+//             // If this node represents a local definition, then store the current
+//             // highlight value on the local scope entry representing this node.
+//             if let Some(definition_highlight) = definition_highlight {
+//                 *definition_highlight = current_highlight;
+//             }
+
+//             // Emit a scope start event and push the node's end position to the stack.
+//             if let Some(highlight) = reference_highlight.or(current_highlight) {
+//                 self.last_highlight_range = Some((range.start, range.end, layer.depth));
+//                 layer.highlight_end_stack.push(range.end);
+//                 return self
+//                     .emit_event(range.start, Some(HighlightEvent::HighlightStart(highlight)));
+//             }
+
+//             self.sort_layers();
+//         }
+//     }
+// }
+
+// impl Syntax {
+//     /// Iterate over the highlighted regions for a given slice of source code.
+//     pub fn highlight_iter<'a>(
+//         &'a self,
+//         source: RopeSlice<'a>,
+//         range: Option<std::ops::Range<usize>>,
+//         cancellation_flag: Option<&'a AtomicUsize>,
+//     ) -> impl Iterator<Item = Result<HighlightEvent, Error>> + 'a {
+//         let mut layers = self
+//             .layers
+//             .iter()
+//             .filter_map(|(_, layer)| {
+//                 // TODO: if range doesn't overlap layer range, skip it
+
+//                 // Reuse a cursor from the pool if available.
+//                 let mut cursor = PARSER.with(|ts_parser| {
+//                     let highlighter = &mut ts_parser.borrow_mut();
+//                     highlighter.cursors.pop().unwrap_or_else(QueryCursor::new)
+//                 });
+
+//                 // The `captures` iterator borrows the `Tree` and the `QueryCursor`, which
+//                 // prevents them from being moved. But both of these values are really just
+//                 // pointers, so it's actually ok to move them.
+//                 let cursor_ref =
+//                     unsafe { mem::transmute::<_, &'static mut QueryCursor>(&mut cursor) };
+
+//                 // if reusing cursors & no range this resets to whole range
+//                 cursor_ref.set_byte_range(range.clone().unwrap_or(0..usize::MAX));
+//                 cursor_ref.set_match_limit(TREE_SITTER_MATCH_LIMIT);
+
+//                 let mut captures = cursor_ref
+//                     .captures(
+//                         &layer.config.query,
+//                         layer.tree().root_node(),
+//                         RopeProvider(source),
+//                     )
+//                     .peekable();
+
+//                 // If there's no captures, skip the layer
+//                 captures.peek()?;
+
+//                 Some(HighlightIterLayer {
+//                     highlight_end_stack: Vec::new(),
+//                     scope_stack: vec![LocalScope {
+//                         inherits: false,
+//                         range: 0..usize::MAX,
+//                         local_defs: Vec::new(),
+//                     }],
+//                     cursor,
+//                     _tree: None,
+//                     captures: RefCell::new(captures),
+//                     config: layer.config.as_ref(), // TODO: just reuse `layer`
+//                     depth: layer.depth,            // TODO: just reuse `layer`
+//                 })
+//             })
+//             .collect::<Vec<_>>();
+
+//         layers.sort_unstable_by_key(|layer| layer.sort_key());
+
+//         let mut result = HighlightIter {
+//             source,
+//             byte_offset: range.map_or(0, |r| r.start),
+//             cancellation_flag,
+//             iter_count: 0,
+//             layers,
+//             next_event: None,
+//             last_highlight_range: None,
+//         };
+//         result.sort_layers();
+//         result
+//     }
+// }

helix-syntax/src/highlighter2.rs

@@ -0,0 +1,206 @@
+use std::borrow::Cow;
+use std::iter::{self, Peekable};
+use std::mem::{replace, take};
+use std::slice;
+
+use hashbrown::HashMap;
+
+use crate::query_iter::{MatchedNode, QueryIter, QueryIterEvent};
+use crate::{Injection, LayerId, Range, Syntax};
+
+/// Indicates which highlight should be applied to a region of source code.
+#[derive(Copy, Clone, Debug, PartialEq, Eq)]
+pub struct Highlight(pub u32);
+impl Highlight{
+    pub(crate) const NONE = Highlight(u32::MAX);
+}
+
+#[derive(Debug)]
+struct LocalDef<'a> {
+    name: Cow<'a, str>,
+    value_range: Range,
+    highlight: Option<Highlight>,
+}
+
+#[derive(Debug)]
+struct LocalScope<'a> {
+    inherits: bool,
+    range: Range,
+    local_defs: Vec<LocalDef<'a>>,
+}
+
+#[derive(Debug)]
+struct HighlightedNode {
+    end: u32,
+    highlight: Highlight,
+}
+
+#[derive(Debug, Default)]
+struct LayerData<'a> {
+    parent_highlights: usize,
+    dormant_highlights: Vec<HighlightedNode>,
+    scope_stack: Vec<LocalDef<'a>>,
+}
+
+struct HighlighterConfig<'a> {
+    new_precedance: bool,
+    highlight_indices: &'a [Highlight],
+}
+
+pub struct Highligther<'a> {
+    query: QueryIter<'a, LayerData<'a>>,
+    next_query_event: Option<QueryIterEvent<LayerData<'a>>>,
+    active_highlights: Vec<HighlightedNode>,
+    next_highlight_end: u32,
+    next_highlight_start: u32,
+    config: HighlighterConfig<'a>,
+}
+
+pub struct HighlightList<'a>(slice::Iter<'a, HighlightedNode>);
+
+impl<'a> Iterator for HighlightList<'a> {
+    type Item = Highlight;
+
+    fn next(&mut self) -> Option<Highlight> {
+        self.0.next().map(|node| node.highlight)
+    }
+}
+
+pub enum HighlighEvent<'a> {
+    RefreshHiglights(HighlightList<'a>),
+    PushHighlights(HighlightList<'a>),
+}
+
+impl<'a> Highligther<'a> {
+    pub fn active_highlights(&self) -> HighlightList<'_> {
+        HighlightList(self.active_highlights.iter())
+    }
+
+    pub fn next_event_offset(&self) -> u32 {
+        self.next_highlight_start.min(self.next_highlight_end)
+    }
+
+    pub fn advance(&mut self) -> HighlighEvent<'_> {
+        let mut refresh = false;
+        let prev_stack_size = self.active_highlights.len();
+
+        let pos = self.next_event_offset();
+        if self.next_highlight_end == pos {
+            self.process_injection_ends();
+            self.process_higlight_end();
+            refresh = true;
+        }
+
+        let mut first_highlight = true;
+        while self.next_highlight_start == pos {
+            let Some(query_event) = self.adance_query_iter() else {
+                break;
+            };
+            match query_event {
+                QueryIterEvent::EnterInjection(_) => self.enter_injection(),
+                QueryIterEvent::Match(node) => self.start_highlight(node, &mut first_highlight),
+                QueryIterEvent::ExitInjection { injection, state } => {
+                    // state is returned if the layer is finifhed, if it isn't we have
+                    // a combined injection and need to deactive its highlights
+                    if state.is_none() {
+                        self.deactive_layer(injection.layer);
+                        refresh = true;
+                    }
+                }
+            }
+        }
+        self.next_highlight_end = self
+            .active_highlights
+            .last()
+            .map_or(u32::MAX, |node| node.end);
+
+        if refresh {
+            HighlighEvent::RefreshHiglights(HighlightList(self.active_highlights.iter()))
+        } else {
+            HighlighEvent::PushHighlights(HighlightList(
+                self.active_highlights[prev_stack_size..].iter(),
+            ))
+        }
+    }
+
+    fn adance_query_iter(&mut self) -> Option<QueryIterEvent<LayerData<'a>>> {
+        let event = replace(&mut self.next_query_event, self.query.next());
+        self.next_highlight_start = self
+            .next_query_event
+            .as_ref()
+            .map_or(u32::MAX, |event| event.start());
+        event
+    }
+
+    fn process_higlight_end(&mut self) {
+        let i = self
+            .active_highlights
+            .iter()
+            .rposition(|highlight| highlight.end != self.next_highlight_end)
+            .unwrap();
+        self.active_highlights.truncate(i);
+    }
+
+    /// processes injections that end at the same position as highlights first.
+    fn process_injection_ends(&mut self) {
+        while self.next_highlight_end == self.next_highlight_start {
+            match self.next_query_event.as_ref() {
+                Some(QueryIterEvent::ExitInjection { injection, state }) => {
+                    if state.is_none() {
+                        self.deactive_layer(injection.layer);
+                    }
+                }
+                Some(QueryIterEvent::Match(matched_node)) if matched_node.byte_range.is_empty() => {
+                }
+                _ => break,
+            }
+        }
+    }
+
+    fn enter_injection(&mut self) {
+        self.query.current_layer_state().parent_highlights = self.active_highlights.len();
+    }
+
+    fn deactive_layer(&mut self, layer: LayerId) {
+        let LayerData {
+            parent_highlights,
+            ref mut dormant_highlights,
+            ..
+        } = *self.query.layer_state(layer);
+        let i = self.active_highlights[parent_highlights..]
+            .iter()
+            .rposition(|highlight| highlight.end != self.next_highlight_end)
+            .unwrap();
+        self.active_highlights.truncate(parent_highlights + i);
+        dormant_highlights.extend(self.active_highlights.drain(parent_highlights..))
+    }
+
+    fn start_highlight(&mut self, node: MatchedNode, first_highlight: &mut bool) {
+        if node.byte_range.is_empty() {
+            return;
+        }
+
+        // if there are multiple matches for the exact same node
+        // only use one of the (the last with new/nvim precedance)
+        if !*first_highlight
+            && self.active_highlights.last().map_or(false, |prev_node| {
+                prev_node.end == node.byte_range.end as u32
+            })
+        {
+            if self.config.new_precedance {
+                self.active_highlights.pop();
+            } else {
+                return;
+            }
+        }
+        let highlight = self.config.highlight_indices[node.capture.idx()];
+        if highlight.0 == u32::MAX {
+            return;
+        }
+        self.active_highlights.push(HighlightedNode {
+            end: node.byte_range.end as u32,
+            highlight,
+        });
+        *first_highlight = false;
+    }
+}

helix-syntax/src/lib.rs

@@ -0,0 +1,250 @@
+use ::ropey::RopeSlice;
+use slotmap::{new_key_type, HopSlotMap};
+
+use std::borrow::Cow;
+use std::hash::{Hash, Hasher};
+use std::path::Path;
+use std::str;
+use std::sync::Arc;
+
+use crate::parse::LayerUpdateFlags;
+
+pub use crate::config::{read_query, HighlightConfiguration};
+use crate::tree_sitter::{SyntaxTree, SyntaxTreeNode};
+pub use pretty_print::pretty_print_tree;
+pub use tree_cursor::TreeCursor;
+
+mod config;
+pub mod highlighter;
+pub mod highlighter2;
+mod parse;
+mod pretty_print;
+mod query_iter;
+pub mod text_object;
+mod tree_cursor;
+pub mod tree_sitter;
+
+new_key_type! {
+    /// The default slot map key type.
+    pub struct LayerId;
+}
+
+/// The maximum number of in-progress matches a TS cursor can consider at once.
+/// This is set to a constant in order to avoid performance problems for medium to large files. Set with `set_match_limit`.
+/// Using such a limit means that we lose valid captures, so there is fundamentally a tradeoff here.
+///
+///
+/// Old tree sitter versions used a limit of 32 by default until this limit was removed in version `0.19.5` (must now be set manually).
+/// However, this causes performance issues for medium to large files.
+/// In helix, this problem caused treesitter motions to take multiple seconds to complete in medium-sized rust files (3k loc).
+///
+///
+/// Neovim also encountered this problem and reintroduced this limit after it was removed upstream
+/// (see <https://github.com/neovim/neovim/issues/14897> and <https://github.com/neovim/neovim/pull/14915>).
+/// The number used here is fundamentally a tradeoff between breaking some obscure edge cases and performance.
+///
+///
+/// Neovim chose 64 for this value somewhat arbitrarily (<https://github.com/neovim/neovim/pull/18397>).
+/// 64 is too low for some languages though. In particular, it breaks some highlighting for record fields in Erlang record definitions.
+/// This number can be increased if new syntax highlight breakages are found, as long as the performance penalty is not too high.
+pub const TREE_SITTER_MATCH_LIMIT: u32 = 256;
+
+// TODO(perf): replace std::ops::Range<usize> with helix_stdx::Range<u32> once added
+type Range = std::ops::Range<usize>;
+
+/// The Tree siitter syntax tree for a single language.
+
+/// This is really multipe nested  different syntax trees due to tree sitter
+/// injections. A single syntax tree/parser is called layer. Each layer
+/// is parsed as a single "file" by tree sitter. There can be multiple layers
+/// for the same language. A layer corresponds to one of three things:
+/// * the root layer
+/// * a singular injection limited to a single node in it's parent layer
+/// * Multiple injections (multiple disjoint nodes in parent layer) that are
+/// parsed as tough they are a single uninterrupted file.
+///
+/// An injection always refer to a single node into which another layer is
+/// injected. As injections only correspond to syntax tree nodes injections in
+/// the same layer do not intersect. However, the syntax tree in a an injected
+/// layer can have nodes that intersect with nodes from the parent layer. For
+/// example:
+/// ```
+/// layer2: | Sibling A |      Sibling B (layer3)     | Sibling C |
+/// layer1: | Sibling A (layer2) | Sibling B | Sibling C (layer2) |
+/// ````
+/// In this case Sibling B really spans across a "GAP" in layer2. While the syntax
+/// node can not be split up by tree sitter directly, we can treat Sibling B as two
+/// seperate injections. That is done while parsing/running the query capture. As
+/// a result the injections from a tree. Note that such other queries must account for
+/// such multi injection nodes.
+#[derive(Debug)]
+pub struct Syntax {
+    layers: HopSlotMap<LayerId, LanguageLayer>,
+    root: LayerId,
+}
+
+impl Syntax {
+    pub fn new(
+        source: RopeSlice,
+        config: Arc<HighlightConfiguration>,
+        injection_callback: impl Fn(&InjectionLanguageMarker) -> Option<Arc<HighlightConfiguration>>,
+    ) -> Option<Self> {
+        let root_layer = LanguageLayer {
+            parse_tree: None,
+            config,
+            flags: LayerUpdateFlags::empty(),
+            ranges: vec![tree_sitter::Range {
+                start_byte: 0,
+                end_byte: u32::MAX,
+                start_point: tree_sitter::Point { row: 0, col: 0 },
+                end_point: tree_sitter::Point {
+                    row: u32::MAX,
+                    col: u32::MAX,
+                },
+            }]
+            .into_boxed_slice(),
+            injections: Box::new([]),
+            parent: None,
+        };
+
+        // track scope_descriptor: a Vec of scopes for item in tree
+
+        let mut layers = HopSlotMap::default();
+        let root = layers.insert(root_layer);
+
+        let mut syntax = Self { root, layers };
+
+        let res = syntax.update(source, Vec::new(), injection_callback);
+
+        if res.is_err() {
+            log::error!("TS parser failed, disabling TS for the current buffer: {res:?}");
+            return None;
+        }
+        Some(syntax)
+    }
+
+    pub fn tree(&self) -> &SyntaxTree {
+        self.layers[self.root].tree()
+    }
+
+    pub fn tree_for_byte_range(&self, start: usize, end: usize) -> &SyntaxTree {
+        let layer = self.layer_for_byte_range(start, end);
+        self.layers[layer].tree()
+    }
+
+    pub fn named_descendant_for_byte_range(
+        &self,
+        start: usize,
+        end: usize,
+    ) -> Option<SyntaxTreeNode<'_>> {
+        self.tree_for_byte_range(start, end)
+            .root_node()
+            .named_descendant_for_byte_range(start, end)
+    }
+
+    pub fn descendant_for_byte_range(
+        &self,
+        start: usize,
+        end: usize,
+    ) -> Option<SyntaxTreeNode<'_>> {
+        self.tree_for_byte_range(start, end)
+            .root_node()
+            .descendant_for_byte_range(start, end)
+    }
+
+    pub fn layer_for_byte_range(&self, start: usize, end: usize) -> LayerId {
+        let mut cursor = self.root;
+        loop {
+            let layer = &self.layers[cursor];
+            let Some(start_injection) = layer.injection_at_byte_idx(start) else {
+                break;
+            };
+            let Some(end_injection) = layer.injection_at_byte_idx(end) else {
+                break;
+            };
+            if start_injection.layer == end_injection.layer {
+                cursor = start_injection.layer;
+            } else {
+                break;
+            }
+        }
+        cursor
+    }
+
+    pub fn walk(&self) -> TreeCursor<'_> {
+        TreeCursor::new(&self.layers, self.root)
+    }
+}
+
+#[derive(Debug, Clone)]
+pub(crate) struct Injection {
+    pub byte_range: Range,
+    pub layer: LayerId,
+}
+
+#[derive(Debug)]
+pub struct LanguageLayer {
+    pub config: Arc<HighlightConfiguration>,
+    parse_tree: Option<SyntaxTree>,
+    ranges: Box<[tree_sitter::Range]>,
+    /// a list of **sorted** non-overlapping injection ranges. Note that
+    /// injection ranges are not relative to the start of this layer but the
+    /// start of the root layer
+    injections: Box<[Injection]>,
+    /// internal flags used during parsing to track incremental invalidation
+    flags: LayerUpdateFlags,
+    parent: Option<LayerId>,
+}
+
+/// This PartialEq implementation only checks if that
+/// two layers are theoretically identical (meaning they highlight the same text range with the same language).
+/// It does not check whether the layers have the same internal treesitter
+/// state.
+impl PartialEq for LanguageLayer {
+    fn eq(&self, other: &Self) -> bool {
+        self.parent == other.parent
+            && self.config.grammar == other.config.grammar
+            && self.ranges == other.ranges
+    }
+}
+
+/// Hash implementation belongs to PartialEq implementation above.
+/// See its documentation for details.
+impl Hash for LanguageLayer {
+    fn hash<H: Hasher>(&self, state: &mut H) {
+        self.parent.hash(state);
+        self.config.grammar.hash(state);
+        self.ranges.hash(state);
+    }
+}
+
+impl LanguageLayer {
+    pub fn tree(&self) -> &SyntaxTree {
+        // TODO: no unwrap
+        self.parse_tree.as_ref().unwrap()
+    }
+
+    /// Returns the injection range **within this layers** that contains `idx`.
+    /// This function will not descend into nested injections
+    pub(crate) fn injection_at_byte_idx(&self, idx: usize) -> Option<&Injection> {
+        let i = self
+            .injections
+            .partition_point(|range| range.byte_range.start < idx);
+        self.injections
+            .get(i)
+            .filter(|injection| injection.byte_range.end > idx)
+    }
+}
+
+/// Represents the reason why syntax highlighting failed.
+#[derive(Debug, PartialEq, Eq)]
+pub enum Error {
+    Cancelled,
+    InvalidLanguage,
+    InvalidRanges,
+    Unknown,
+}
+
+fn byte_range_to_str(range: std::ops::Range<usize>, source: RopeSlice) -> Cow<str> {
+    Cow::from(source.byte_slice(range))
+}

helix-syntax/src/merge.rs

@@ -0,0 +1,135 @@
+use crate::highlighter::{Highlight, HighlightEvent};
+
+pub struct Merge<I> {
+    iter: I,
+    spans: Box<dyn Iterator<Item = (usize, std::ops::Range<usize>)>>,
+
+    next_event: Option<HighlightEvent>,
+    next_span: Option<(usize, std::ops::Range<usize>)>,
+
+    queue: Vec<HighlightEvent>,
+}
+
+/// Merge a list of spans into the highlight event stream.
+pub fn merge<I: Iterator<Item = HighlightEvent>>(
+    iter: I,
+    spans: Vec<(usize, std::ops::Range<usize>)>,
+) -> Merge<I> {
+    let spans = Box::new(spans.into_iter());
+    let mut merge = Merge {
+        iter,
+        spans,
+        next_event: None,
+        next_span: None,
+        queue: Vec::new(),
+    };
+    merge.next_event = merge.iter.next();
+    merge.next_span = merge.spans.next();
+    merge
+}
+
+impl<I: Iterator<Item = HighlightEvent>> Iterator for Merge<I> {
+    type Item = HighlightEvent;
+    fn next(&mut self) -> Option<Self::Item> {
+        use HighlightEvent::*;
+        if let Some(event) = self.queue.pop() {
+            return Some(event);
+        }
+
+        loop {
+            match (self.next_event, &self.next_span) {
+                // this happens when range is partially or fully offscreen
+                (Some(Source { start, .. }), Some((span, range))) if start > range.start => {
+                    if start > range.end {
+                        self.next_span = self.spans.next();
+                    } else {
+                        self.next_span = Some((*span, start..range.end));
+                    };
+                }
+                _ => break,
+            }
+        }
+
+        match (self.next_event, &self.next_span) {
+            (Some(HighlightStart(i)), _) => {
+                self.next_event = self.iter.next();
+                Some(HighlightStart(i))
+            }
+            (Some(HighlightEnd), _) => {
+                self.next_event = self.iter.next();
+                Some(HighlightEnd)
+            }
+            (Some(Source { start, end }), Some((_, range))) if start < range.start => {
+                let intersect = range.start.min(end);
+                let event = Source {
+                    start,
+                    end: intersect,
+                };
+
+                if end == intersect {
+                    // the event is complete
+                    self.next_event = self.iter.next();
+                } else {
+                    // subslice the event
+                    self.next_event = Some(Source {
+                        start: intersect,
+                        end,
+                    });
+                };
+
+                Some(event)
+            }
+            (Some(Source { start, end }), Some((span, range))) if start == range.start => {
+                let intersect = range.end.min(end);
+                let event = HighlightStart(Highlight(*span));
+
+                // enqueue in reverse order
+                self.queue.push(HighlightEnd);
+                self.queue.push(Source {
+                    start,
+                    end: intersect,
+                });
+
+                if end == intersect {
+                    // the event is complete
+                    self.next_event = self.iter.next();
+                } else {
+                    // subslice the event
+                    self.next_event = Some(Source {
+                        start: intersect,
+                        end,
+                    });
+                };
+
+                if intersect == range.end {
+                    self.next_span = self.spans.next();
+                } else {
+                    self.next_span = Some((*span, intersect..range.end));
+                }
+
+                Some(event)
+            }
+            (Some(event), None) => {
+                self.next_event = self.iter.next();
+                Some(event)
+            }
+            // Can happen if cursor at EOF and/or diagnostic reaches past the end.
+            // We need to actually emit events for the cursor-at-EOF situation,
+            // even though the range is past the end of the text.  This needs to be
+            // handled appropriately by the drawing code by not assuming that
+            // all `Source` events point to valid indices in the rope.
+            (None, Some((span, range))) => {
+                let event = HighlightStart(Highlight(*span));
+                self.queue.push(HighlightEnd);
+                self.queue.push(Source {
+                    start: range.start,
+                    end: range.end,
+                });
+                self.next_span = self.spans.next();
+                Some(event)
+            }
+            (None, None) => None,
+            e => unreachable!("{:?}", e),
+        }
+    }
+}

helix-syntax/src/parse.rs

@@ -0,0 +1,429 @@
+// use std::collections::VecDeque;
+// use std::mem::replace;
+// use std::sync::Arc;
+
+// use ahash::RandomState;
+use bitflags::bitflags;
+// use hashbrown::raw::RawTable;
+// use ropey::RopeSlice;
+// use tree_sitter::{Node, Parser, Point, QueryCursor, Range};
+
+// use crate::ropey::RopeProvider;
+// use crate::{
+//     Error, HighlightConfiguration, IncludedChildren, InjectionLanguageMarker, LanguageLayer,
+//     Syntax, PARSER, TREE_SITTER_MATCH_LIMIT,
+// };
+
+bitflags! {
+    /// Flags that track the status of a layer
+    /// in the `Sytaxn::update` function
+    #[derive(Debug)]
+    pub(crate) struct LayerUpdateFlags : u32{
+        const MODIFIED = 0b001;
+        const MOVED = 0b010;
+        const TOUCHED = 0b100;
+    }
+}
+
+// impl Syntax {
+//     pub fn update(
+//         &mut self,
+//         source: RopeSlice,
+//         edits: Vec<tree_sitter::InputEdit>,
+//         injection_callback: impl Fn(&InjectionLanguageMarker) -> Option<Arc<HighlightConfiguration>>,
+//     ) -> Result<(), Error> {
+//         let mut queue = VecDeque::new();
+//         queue.push_back(self.root);
+
+//         // This table allows inverse indexing of `layers`.
+//         // That is by hashing a `Layer` you can find
+//         // the `LayerId` of an existing equivalent `Layer` in `layers`.
+//         //
+//         // It is used to determine if a new layer exists for an injection
+//         // or if an existing layer needs to be updated.
+//         let mut layers_table = RawTable::with_capacity(self.layers.len());
+//         let layers_hasher = RandomState::new();
+//         // Use the edits to update all layers markers
+//         fn point_add(a: Point, b: Point) -> Point {
+//             if b.row > 0 {
+//                 Point::new(a.row.saturating_add(b.row), b.column)
+//             } else {
+//                 Point::new(0, a.column.saturating_add(b.column))
+//             }
+//         }
+//         fn point_sub(a: Point, b: Point) -> Point {
+//             if a.row > b.row {
+//                 Point::new(a.row.saturating_sub(b.row), a.column)
+//             } else {
+//                 Point::new(0, a.column.saturating_sub(b.column))
+//             }
+//         }
+
+//         for (layer_id, layer) in self.layers.iter_mut() {
+//             // The root layer always covers the whole range (0..usize::MAX)
+//             if layer.depth == 0 {
+//                 layer.flags = LayerUpdateFlags::MODIFIED;
+//                 continue;
+//             }
+
+//             if !edits.is_empty() {
+//                 for range in &mut layer.ranges {
+//                     // Roughly based on https://github.com/tree-sitter/tree-sitter/blob/ddeaa0c7f534268b35b4f6cb39b52df082754413/lib/src/subtree.c#L691-L720
+//                     for edit in edits.iter().rev() {
+//                         let is_pure_insertion = edit.old_end_byte == edit.start_byte;
+
+//                         // if edit is after range, skip
+//                         if edit.start_byte > range.end_byte {
+//                             // TODO: || (is_noop && edit.start_byte == range.end_byte)
+//                             continue;
+//                         }
+
+//                         // if edit is before range, shift entire range by len
+//                         if edit.old_end_byte < range.start_byte {
+//                             range.start_byte =
+//                                 edit.new_end_byte + (range.start_byte - edit.old_end_byte);
+//                             range.start_point = point_add(
+//                                 edit.new_end_position,
+//                                 point_sub(range.start_point, edit.old_end_position),
+//                             );
+
+//                             range.end_byte = edit
+//                                 .new_end_byte
+//                                 .saturating_add(range.end_byte - edit.old_end_byte);
+//                             range.end_point = point_add(
+//                                 edit.new_end_position,
+//                                 point_sub(range.end_point, edit.old_end_position),
+//                             );
+
+//                             layer.flags |= LayerUpdateFlags::MOVED;
+//                         }
+//                         // if the edit starts in the space before and extends into the range
+//                         else if edit.start_byte < range.start_byte {
+//                             range.start_byte = edit.new_end_byte;
+//                             range.start_point = edit.new_end_position;
+
+//                             range.end_byte = range
+//                                 .end_byte
+//                                 .saturating_sub(edit.old_end_byte)
+//                                 .saturating_add(edit.new_end_byte);
+//                             range.end_point = point_add(
+//                                 edit.new_end_position,
+//                                 point_sub(range.end_point, edit.old_end_position),
+//                             );
+//                             layer.flags = LayerUpdateFlags::MODIFIED;
+//                         }
+//                         // If the edit is an insertion at the start of the tree, shift
+//                         else if edit.start_byte == range.start_byte && is_pure_insertion {
+//                             range.start_byte = edit.new_end_byte;
+//                             range.start_point = edit.new_end_position;
+//                             layer.flags |= LayerUpdateFlags::MOVED;
+//                         } else {
+//                             range.end_byte = range
+//                                 .end_byte
+//                                 .saturating_sub(edit.old_end_byte)
+//                                 .saturating_add(edit.new_end_byte);
+//                             range.end_point = point_add(
+//                                 edit.new_end_position,
+//                                 point_sub(range.end_point, edit.old_end_position),
+//                             );
+//                             layer.flags = LayerUpdateFlags::MODIFIED;
+//                         }
+//                     }
+//                 }
+//             }
+
+//             let hash = layers_hasher.hash_one(layer);
+//             // Safety: insert_no_grow is unsafe because it assumes that the table
+//             // has enough capacity to hold additional elements.
+//             // This is always the case as we reserved enough capacity above.
+//             unsafe { layers_table.insert_no_grow(hash, layer_id) };
+//         }
+
+//         PARSER.with(|ts_parser| {
+//             let ts_parser = &mut ts_parser.borrow_mut();
+//             ts_parser.parser.set_timeout_micros(1000 * 500); // half a second is pretty generours
+//             let mut cursor = ts_parser.cursors.pop().unwrap_or_else(QueryCursor::new);
+//             // TODO: might need to set cursor range
+//             cursor.set_byte_range(0..usize::MAX);
+//             cursor.set_match_limit(TREE_SITTER_MATCH_LIMIT);
+
+//             let source_slice = source.slice(..);
+
+//             while let Some(layer_id) = queue.pop_front() {
+//                 let layer = &mut self.layers[layer_id];
+
+//                 // Mark the layer as touched
+//                 layer.flags |= LayerUpdateFlags::TOUCHED;
+
+//                 // If a tree already exists, notify it of changes.
+//                 if let Some(tree) = &mut layer.parse_tree {
+//                     if layer
+//                         .flags
+//                         .intersects(LayerUpdateFlags::MODIFIED | LayerUpdateFlags::MOVED)
+//                     {
+//                         for edit in edits.iter().rev() {
+//                             // Apply the edits in reverse.
+//                             // If we applied them in order then edit 1 would disrupt the positioning of edit 2.
+//                             tree.edit(edit);
+//                         }
+//                     }
+
+//                     if layer.flags.contains(LayerUpdateFlags::MODIFIED) {
+//                         // Re-parse the tree.
+//                         layer.parse(&mut ts_parser.parser, source)?;
+//                     }
+//                 } else {
+//                     // always parse if this layer has never been parsed before
+//                     layer.parse(&mut ts_parser.parser, source)?;
+//                 }
+
+//                 // Switch to an immutable borrow.
+//                 let layer = &self.layers[layer_id];
+
+//                 // Process injections.
+//                 let matches = cursor.matches(
+//                     &layer.config.injections_query,
+//                     layer.tree().root_node(),
+//                     RopeProvider(source_slice),
+//                 );
+//                 let mut combined_injections = vec![
+//                     (None, Vec::new(), IncludedChildren::default());
+//                     layer.config.combined_injections_patterns.len()
+//                 ];
+//                 let mut injections = Vec::new();
+//                 let mut last_injection_end = 0;
+//                 for mat in matches {
+//                     let (injection_capture, content_node, included_children) = layer
+//                         .config
+//                         .injection_for_match(&layer.config.injections_query, &mat, source_slice);
+
+//                     // in case this is a combined injection save it for more processing later
+//                     if let Some(combined_injection_idx) = layer
+//                         .config
+//                         .combined_injections_patterns
+//                         .iter()
+//                         .position(|&pattern| pattern == mat.pattern_index)
+//                     {
+//                         let entry = &mut combined_injections[combined_injection_idx];
+//                         if injection_capture.is_some() {
+//                             entry.0 = injection_capture;
+//                         }
+//                         if let Some(content_node) = content_node {
+//                             if content_node.start_byte() >= last_injection_end {
+//                                 entry.1.push(content_node);
+//                                 last_injection_end = content_node.end_byte();
+//                             }
+//                         }
+//                         entry.2 = included_children;
+//                         continue;
+//                     }
+
+//                     // Explicitly remove this match so that none of its other captures will remain
+//                     // in the stream of captures.
+//                     mat.remove();
+
+//                     // If a language is found with the given name, then add a new language layer
+//                     // to the highlighted document.
+//                     if let (Some(injection_capture), Some(content_node)) =
+//                         (injection_capture, content_node)
+//                     {
+//                         if let Some(config) = (injection_callback)(&injection_capture) {
+//                             let ranges =
+//                                 intersect_ranges(&layer.ranges, &[content_node], included_children);
+
+//                             if !ranges.is_empty() {
+//                                 if content_node.start_byte() < last_injection_end {
+//                                     continue;
+//                                 }
+//                                 last_injection_end = content_node.end_byte();
+//                                 injections.push((config, ranges));
+//                             }
+//                         }
+//                     }
+//                 }
+
+//                 for (lang_name, content_nodes, included_children) in combined_injections {
+//                     if let (Some(lang_name), false) = (lang_name, content_nodes.is_empty()) {
+//                         if let Some(config) = (injection_callback)(&lang_name) {
+//                             let ranges =
+//                                 intersect_ranges(&layer.ranges, &content_nodes, included_children);
+//                             if !ranges.is_empty() {
+//                                 injections.push((config, ranges));
+//                             }
+//                         }
+//                     }
+//                 }
+
+//                 let depth = layer.depth + 1;
+//                 // TODO: can't inline this since matches borrows self.layers
+//                 for (config, ranges) in injections {
+//                     let parent = Some(layer_id);
+//                     let new_layer = LanguageLayer {
+//                         parse_tree: None,
+//                         config,
+//                         depth,
+//                         ranges,
+//                         flags: LayerUpdateFlags::empty(),
+//                         parent: None,
+//                     };
+
+//                     // Find an identical existing layer
+//                     let layer = layers_table
+//                         .get(layers_hasher.hash_one(&new_layer), |&it| {
+//                             self.layers[it] == new_layer
+//                         })
+//                         .copied();
+
+//                     // ...or insert a new one.
+//                     let layer_id = layer.unwrap_or_else(|| self.layers.insert(new_layer));
+//                     self.layers[layer_id].parent = parent;
+
+//                     queue.push_back(layer_id);
+//                 }
+
+//                 // TODO: pre-process local scopes at this time, rather than highlight?
+//                 // would solve problems with locals not working across boundaries
+//             }
+
+//             // Return the cursor back in the pool.
+//             ts_parser.cursors.push(cursor);
+
+//             // Reset all `LayerUpdateFlags` and remove all untouched layers
+//             self.layers.retain(|_, layer| {
+//                 replace(&mut layer.flags, LayerUpdateFlags::empty())
+//                     .contains(LayerUpdateFlags::TOUCHED)
+//             });
+
+//             Ok(())
+//         })
+//     }
+// }
+
+// /// Compute the ranges that should be included when parsing an injection.
+// /// This takes into account three things:
+// /// * `parent_ranges` - The ranges must all fall within the *current* layer's ranges.
+// /// * `nodes` - Every injection takes place within a set of nodes. The injection ranges
+// ///   are the ranges of those nodes.
+// /// * `includes_children` - For some injections, the content nodes' children should be
+// ///   excluded from the nested document, so that only the content nodes' *own* content
+// ///   is reparsed. For other injections, the content nodes' entire ranges should be
+// ///   reparsed, including the ranges of their children.
+// fn intersect_ranges(
+//     parent_ranges: &[Range],
+//     nodes: &[Node],
+//     included_children: IncludedChildren,
+// ) -> Vec<Range> {
+//     let mut cursor = nodes[0].walk();
+//     let mut result = Vec::new();
+//     let mut parent_range_iter = parent_ranges.iter();
+//     let mut parent_range = parent_range_iter
+//         .next()
+//         .expect("Layers should only be constructed with non-empty ranges vectors");
+//     for node in nodes.iter() {
+//         let mut preceding_range = Range {
+//             start_byte: 0,
+//             start_point: Point::new(0, 0),
+//             end_byte: node.start_byte(),
+//             end_point: node.start_position(),
+//         };
+//         let following_range = Range {
+//             start_byte: node.end_byte(),
+//             start_point: node.end_position(),
+//             end_byte: usize::MAX,
+//             end_point: Point::new(usize::MAX, usize::MAX),
+//         };
+
+//         for excluded_range in node
+//             .children(&mut cursor)
+//             .filter_map(|child| match included_children {
+//                 IncludedChildren::None => Some(child.range()),
+//                 IncludedChildren::All => None,
+//                 IncludedChildren::Unnamed => {
+//                     if child.is_named() {
+//                         Some(child.range())
+//                     } else {
+//                         None
+//                     }
+//                 }
+//             })
+//             .chain([following_range].iter().cloned())
+//         {
+//             let mut range = Range {
+//                 start_byte: preceding_range.end_byte,
+//                 start_point: preceding_range.end_point,
+//                 end_byte: excluded_range.start_byte,
+//                 end_point: excluded_range.start_point,
+//             };
+//             preceding_range = excluded_range;
+
+//             if range.end_byte < parent_range.start_byte {
+//                 continue;
+//             }
+
+//             while parent_range.start_byte <= range.end_byte {
+//                 if parent_range.end_byte > range.start_byte {
+//                     if range.start_byte < parent_range.start_byte {
+//                         range.start_byte = parent_range.start_byte;
+//                         range.start_point = parent_range.start_point;
+//                     }
+
+//                     if parent_range.end_byte < range.end_byte {
+//                         if range.start_byte < parent_range.end_byte {
+//                             result.push(Range {
+//                                 start_byte: range.start_byte,
+//                                 start_point: range.start_point,
+//                                 end_byte: parent_range.end_byte,
+//                                 end_point: parent_range.end_point,
+//                             });
+//                         }
+//                         range.start_byte = parent_range.end_byte;
+//                         range.start_point = parent_range.end_point;
+//                     } else {
+//                         if range.start_byte < range.end_byte {
+//                             result.push(range);
+//                         }
+//                         break;
+//                     }
+//                 }
+
+//                 if let Some(next_range) = parent_range_iter.next() {
+//                     parent_range = next_range;
+//                 } else {
+//                     return result;
+//                 }
+//             }
+//         }
+//     }
+//     result
+// }
+
+// impl LanguageLayer {
+//     fn parse(&mut self, parser: &mut Parser, source: RopeSlice) -> Result<(), Error> {
+//         parser
+//             .set_included_ranges(&self.ranges)
+//             .map_err(|_| Error::InvalidRanges)?;
+
+//         parser
+//             .set_language(&self.config.language)
+//             .map_err(|_| Error::InvalidLanguage)?;
+
+//         // unsafe { syntax.parser.set_cancellation_flag(cancellation_flag) };
+//         let tree = parser
+//             .parse_with(
+//                 &mut |byte, _| {
+//                     if byte <= source.len_bytes() {
+//                         let (chunk, start_byte, _, _) = source.chunk_at_byte(byte);
+//                         &chunk.as_bytes()[byte - start_byte..]
+//                     } else {
+//                         // out of range
+//                         &[]
+//                     }
+//                 },
+//                 self.parse_tree.as_ref(),
+//             )
+//             .ok_or(Error::Cancelled)?;
+//         // unsafe { ts_parser.parser.set_cancellation_flag(None) };
+//         self.parse_tree = Some(tree);
+//         Ok(())
+//     }
+// }

helix-syntax/src/pretty_print.rs

@@ -0,0 +1,65 @@
+use std::fmt;
+
+use tree_sitter::{Node, TreeCursor};
+
+pub fn pretty_print_tree<W: fmt::Write>(fmt: &mut W, node: Node) -> fmt::Result {
+    if node.child_count() == 0 {
+        if node_is_visible(&node) {
+            write!(fmt, "({})", node.kind())
+        } else {
+            write!(fmt, "\"{}\"", node.kind())
+        }
+    } else {
+        pretty_print_tree_impl(fmt, &mut node.walk(), 0)
+    }
+}
+
+fn pretty_print_tree_impl<W: fmt::Write>(
+    fmt: &mut W,
+    cursor: &mut TreeCursor,
+    depth: usize,
+) -> fmt::Result {
+    let node = cursor.node();
+    let visible = node_is_visible(&node);
+
+    if visible {
+        let indentation_columns = depth * 2;
+        write!(fmt, "{:indentation_columns$}", "")?;
+
+        if let Some(field_name) = cursor.field_name() {
+            write!(fmt, "{}: ", field_name)?;
+        }
+
+        write!(fmt, "({}", node.kind())?;
+    }
+
+    // Handle children.
+    if cursor.goto_first_child() {
+        loop {
+            if node_is_visible(&cursor.node()) {
+                fmt.write_char('\n')?;
+            }
+
+            pretty_print_tree_impl(fmt, cursor, depth + 1)?;
+
+            if !cursor.goto_next_sibling() {
+                break;
+            }
+        }
+
+        let moved = cursor.goto_parent();
+        // The parent of the first child must exist, and must be `node`.
+        debug_assert!(moved);
+        debug_assert!(cursor.node() == node);
+    }
+
+    if visible {
+        fmt.write_char(')')?;
+    }
+
+    Ok(())
+}
+
+fn node_is_visible(node: &Node) -> bool {
+    node.is_missing() || (node.is_named() && node.language().node_kind_is_visible(node.kind_id()))
+}

helix-syntax/src/query_iter.rs

@@ -0,0 +1,236 @@
+use core::slice;
+use std::iter::Peekable;
+use std::mem::replace;
+
+use hashbrown::HashMap;
+use ropey::RopeSlice;
+
+use crate::tree_sitter::{
+    Capture, InactiveQueryCursor, Query, QueryCursor, RopeTsInput, SyntaxTreeNode,
+};
+use crate::{Injection, LayerId, Range, Syntax};
+
+#[derive(Clone)]
+pub struct MatchedNode {
+    pub capture: Capture,
+    pub byte_range: Range,
+}
+
+struct LayerQueryIter<'a> {
+    cursor: QueryCursor<'a, 'a, RopeTsInput<'a>>,
+    peeked: Option<MatchedNode>,
+}
+
+impl<'a> LayerQueryIter<'a> {
+    fn peek(&mut self) -> Option<&MatchedNode> {
+        if self.peeked.is_none() {
+            let (query_match, node_idx) = self.cursor.next_matched_node()?;
+            let matched_node = query_match.matched_node(node_idx);
+            self.peeked = Some(MatchedNode {
+                capture: matched_node.capture,
+                byte_range: matched_node.syntax_node.byte_range(),
+            });
+        }
+        self.peeked.as_ref()
+    }
+
+    fn consume(&mut self) -> MatchedNode {
+        self.peeked.take().unwrap()
+    }
+}
+
+struct ActiveLayer<'a, S> {
+    state: S,
+    query_iter: LayerQueryIter<'a>,
+    injections: Peekable<slice::Iter<'a, Injection>>,
+}
+
+// data only needed when entering and exiting injections
+// seperate struck to keep the QueryIter reasonably small
+struct QueryIterLayerManager<'a, S> {
+    query: &'a Query,
+    node: SyntaxTreeNode<'a>,
+    src: RopeSlice<'a>,
+    syntax: &'a Syntax,
+    active_layers: HashMap<LayerId, Box<ActiveLayer<'a, S>>>,
+    active_injections: Vec<Injection>,
+}
+
+impl<'a, S: Default> QueryIterLayerManager<'a, S> {
+    fn init_layer(&mut self, injection: Injection) -> Box<ActiveLayer<'a, S>> {
+        self.active_layers
+            .remove(&injection.layer)
+            .unwrap_or_else(|| {
+                let layer = &self.syntax.layers[injection.layer];
+                let injection_start = layer
+                    .injections
+                    .partition_point(|child| child.byte_range.start < injection.byte_range.start);
+                let cursor = InactiveQueryCursor::new().execute_query(
+                    self.query,
+                    &self.node,
+                    RopeTsInput::new(self.src),
+                );
+                Box::new(ActiveLayer {
+                    state: S::default(),
+                    query_iter: LayerQueryIter {
+                        cursor,
+                        peeked: None,
+                    },
+                    injections: layer.injections[injection_start..].iter().peekable(),
+                })
+            })
+    }
+}
+
+pub struct QueryIter<'a, LayerState: Default = ()> {
+    layer_manager: Box<QueryIterLayerManager<'a, LayerState>>,
+    current_layer: Box<ActiveLayer<'a, LayerState>>,
+    current_injection: Injection,
+}
+
+impl<'a, LayerState: Default> QueryIter<'a, LayerState> {
+    pub fn new(syntax: &'a Syntax, src: RopeSlice<'a>, query: &'a Query) -> Self {
+        Self::at(syntax, src, query, syntax.tree().root_node(), syntax.root)
+    }
+
+    pub fn at(
+        syntax: &'a Syntax,
+        src: RopeSlice<'a>,
+        query: &'a Query,
+        node: SyntaxTreeNode<'a>,
+        layer: LayerId,
+    ) -> Self {
+        // create fake injection for query root
+        let injection = Injection {
+            byte_range: node.byte_range(),
+            layer,
+        };
+        let mut layer_manager = Box::new(QueryIterLayerManager {
+            query,
+            node,
+            src,
+            syntax,
+            // TODO: reuse allocations with an allocation pool
+            active_layers: HashMap::with_capacity(8),
+            active_injections: Vec::with_capacity(8),
+        });
+        Self {
+            current_layer: layer_manager.init_layer(injection),
+            current_injection: injection,
+            layer_manager,
+        }
+    }
+
+    pub fn current_layer_state(&mut self) -> &mut LayerState {
+        &mut self.current_layer.state
+    }
+
+    pub fn layer_state(&mut self, layer: LayerId) -> &mut LayerState {
+        if layer == self.current_injection.layer {
+            self.current_layer_state()
+        } else {
+            &mut self
+                .layer_manager
+                .active_layers
+                .get_mut(&layer)
+                .unwrap()
+                .state
+        }
+    }
+
+    fn enter_injection(&mut self, injection: Injection) {
+        let active_layer = self.layer_manager.init_layer(injection);
+        let old_injection = replace(&mut self.current_injection, injection);
+        let old_layer = replace(&mut self.current_layer, active_layer);
+        self.layer_manager
+            .active_layers
+            .insert(old_injection.layer, old_layer);
+        self.layer_manager.active_injections.push(old_injection);
+    }
+
+    fn exit_injection(&mut self) -> Option<(Injection, Option<LayerState>)> {
+        let injection = replace(
+            &mut self.current_injection,
+            self.layer_manager.active_injections.pop()?,
+        );
+        let layer = replace(
+            &mut self.current_layer,
+            self.layer_manager
+                .active_layers
+                .remove(&self.current_injection.layer)?,
+        );
+        let layer_unfinished = layer.query_iter.peeked.is_some();
+        if layer_unfinished {
+            self.layer_manager
+                .active_layers
+                .insert(injection.layer, layer)
+                .unwrap();
+            Some((injection, None))
+        } else {
+            Some((injection, Some(layer.state)))
+        }
+    }
+}
+
+impl<'a, S: Default> Iterator for QueryIter<'a, S> {
+    type Item = QueryIterEvent<S>;
+
+    fn next(&mut self) -> Option<QueryIterEvent<S>> {
+        loop {
+            let next_injection = self.current_layer.injections.peek().filter(|injection| {
+                injection.byte_range.start < self.current_injection.byte_range.end
+            });
+            let next_match = self.current_layer.query_iter.peek().filter(|matched_node| {
+                matched_node.byte_range.start < self.current_injection.byte_range.end
+            });
+
+            match (next_match, next_injection) {
+                (None, None) => {
+                    return self.exit_injection().map(|(injection, state)| {
+                        QueryIterEvent::ExitInjection { injection, state }
+                    });
+                }
+                (Some(_), None) => {
+                    // consume match
+                    let matched_node = self.current_layer.query_iter.consume();
+                    return Some(QueryIterEvent::Match(matched_node));
+                }
+                (Some(matched_node), Some(injection))
+                    if matched_node.byte_range.start <= injection.byte_range.end =>
+                {
+                    // consume match
+                    let matched_node = self.current_layer.query_iter.consume();
+                    // ignore nodes that are overlapped by the injection
+                    if matched_node.byte_range.start <= injection.byte_range.start {
+                        return Some(QueryIterEvent::Match(matched_node));
+                    }
+                }
+                (Some(_), Some(_)) | (None, Some(_)) => {
+                    // consume injection
+                    let injection = self.current_layer.injections.next().unwrap();
+                    self.enter_injection(injection.clone());
+                    return Some(QueryIterEvent::EnterInjection(injection.clone()));
+                }
+            }
+        }
+    }
+}
+
+pub enum QueryIterEvent<State = ()> {
+    EnterInjection(Injection),
+    Match(MatchedNode),
+    ExitInjection {
+        injection: Injection,
+        state: Option<State>,
+    },
+}
+
+impl<S> QueryIterEvent<S> {
+    pub fn start(&self) -> u32 {
+        match self {
+            QueryIterEvent::EnterInjection(injection) => injection.byte_range.start as u32,
+            QueryIterEvent::Match(mat) => mat.byte_range.start as u32,
+            QueryIterEvent::ExitInjection { injection, .. } => injection.byte_range.start as u32,
+        }
+    }
+}

helix-syntax/src/ropey.rs

@@ -0,0 +1 @@
+

helix-syntax/src/text_object.rs

@@ -0,0 +1,93 @@
+// TODO: rework using query iter
+
+use std::iter;
+
+use ropey::RopeSlice;
+
+use crate::tree_sitter::{InactiveQueryCursor, Query, RopeTsInput, SyntaxTreeNode};
+use crate::TREE_SITTER_MATCH_LIMIT;
+
+#[derive(Debug)]
+pub enum CapturedNode<'a> {
+    Single(SyntaxTreeNode<'a>),
+    /// Guaranteed to be not empty
+    Grouped(Vec<SyntaxTreeNode<'a>>),
+}
+
+impl<'a> CapturedNode<'a> {
+    pub fn start_byte(&self) -> usize {
+        match self {
+            Self::Single(n) => n.start_byte(),
+            Self::Grouped(ns) => ns[0].start_byte(),
+        }
+    }
+
+    pub fn end_byte(&self) -> usize {
+        match self {
+            Self::Single(n) => n.end_byte(),
+            Self::Grouped(ns) => ns.last().unwrap().end_byte(),
+        }
+    }
+}
+
+#[derive(Debug)]
+pub struct TextObjectQuery {
+    pub query: Query,
+}
+
+impl TextObjectQuery {
+    /// Run the query on the given node and return sub nodes which match given
+    /// capture ("function.inside", "class.around", etc).
+    ///
+    /// Captures may contain multiple nodes by using quantifiers (+, *, etc),
+    /// and support for this is partial and could use improvement.
+    ///
+    /// ```query
+    /// (comment)+ @capture
+    ///
+    /// ; OR
+    /// (
+    ///   (comment)*
+    ///   .
+    ///   (function)
+    /// ) @capture
+    /// ```
+    pub fn capture_nodes<'a>(
+        &'a self,
+        capture_name: &str,
+        node: SyntaxTreeNode<'a>,
+        slice: RopeSlice<'a>,
+        cursor: InactiveQueryCursor,
+    ) -> Option<impl Iterator<Item = CapturedNode<'a>>> {
+        self.capture_nodes_any(&[capture_name], node, slice, cursor)
+    }
+
+    /// Find the first capture that exists out of all given `capture_names`
+    /// and return sub nodes that match this capture.
+    pub fn capture_nodes_any<'a>(
+        &'a self,
+        capture_names: &[&str],
+        node: SyntaxTreeNode<'a>,
+        slice: RopeSlice<'a>,
+        mut cursor: InactiveQueryCursor,
+    ) -> Option<impl Iterator<Item = CapturedNode<'a>>> {
+        let capture = capture_names
+            .iter()
+            .find_map(|cap| self.query.get_capture(cap))?;
+
+        cursor.set_match_limit(TREE_SITTER_MATCH_LIMIT);
+        let mut cursor = cursor.execute_query(&self.query, &node, RopeTsInput::new(slice));
+        let capture_node = iter::from_fn(move || {
+            let (mat, _) = cursor.next_matched_node()?;
+            Some(mat.nodes_for_capture(capture).cloned().collect())
+        })
+        .filter_map(move |nodes: Vec<_>| {
+            if nodes.len() > 1 {
+                Some(CapturedNode::Grouped(nodes))
+            } else {
+                nodes.into_iter().map(CapturedNode::Single).next()
+            }
+        });
+        Some(capture_node)
+    }
+}

helix-syntax/src/tree_cursor.rs

@@ -0,0 +1,264 @@
+use std::{cmp::Reverse, ops::Range};
+
+use super::{LanguageLayer, LayerId};
+
+use slotmap::HopSlotMap;
+use tree_sitter::Node;
+
+/// The byte range of an injection layer.
+///
+/// Injection ranges may overlap, but all overlapping parts are subsets of their parent ranges.
+/// This allows us to sort the ranges ahead of time in order to efficiently find a range that
+/// contains a point with maximum depth.
+#[derive(Debug)]
+struct InjectionRange {
+    start: usize,
+    end: usize,
+    layer_id: LayerId,
+    depth: u32,
+}
+
+pub struct TreeCursor<'a> {
+    layers: &'a HopSlotMap<LayerId, LanguageLayer>,
+    root: LayerId,
+    current: LayerId,
+    injection_ranges: Vec<InjectionRange>,
+    // TODO: Ideally this would be a `tree_sitter::TreeCursor<'a>` but
+    // that returns very surprising results in testing.
+    cursor: Node<'a>,
+}
+
+impl<'a> TreeCursor<'a> {
+    pub(super) fn new(layers: &'a HopSlotMap<LayerId, LanguageLayer>, root: LayerId) -> Self {
+        let mut injection_ranges = Vec::new();
+
+        for (layer_id, layer) in layers.iter() {
+            // Skip the root layer
+            if layer.parent.is_none() {
+                continue;
+            }
+            for byte_range in layer.ranges.iter() {
+                let range = InjectionRange {
+                    start: byte_range.start_byte,
+                    end: byte_range.end_byte,
+                    layer_id,
+                    depth: layer.depth,
+                };
+                injection_ranges.push(range);
+            }
+        }
+
+        injection_ranges.sort_unstable_by_key(|range| (range.end, Reverse(range.depth)));
+
+        let cursor = layers[root].tree().root_node();
+
+        Self {
+            layers,
+            root,
+            current: root,
+            injection_ranges,
+            cursor,
+        }
+    }
+
+    pub fn node(&self) -> Node<'a> {
+        self.cursor
+    }
+
+    pub fn goto_parent(&mut self) -> bool {
+        if let Some(parent) = self.node().parent() {
+            self.cursor = parent;
+            return true;
+        }
+
+        // If we are already on the root layer, we cannot ascend.
+        if self.current == self.root {
+            return false;
+        }
+
+        // Ascend to the parent layer.
+        let range = self.node().byte_range();
+        let parent_id = self.layers[self.current]
+            .parent
+            .expect("non-root layers have a parent");
+        self.current = parent_id;
+        let root = self.layers[self.current].tree().root_node();
+        self.cursor = root
+            .descendant_for_byte_range(range.start, range.end)
+            .unwrap_or(root);
+
+        true
+    }
+
+    pub fn goto_parent_with<P>(&mut self, predicate: P) -> bool
+    where
+        P: Fn(&Node) -> bool,
+    {
+        while self.goto_parent() {
+            if predicate(&self.node()) {
+                return true;
+            }
+        }
+
+        false
+    }
+
+    /// Finds the injection layer that has exactly the same range as the given `range`.
+    fn layer_id_of_byte_range(&self, search_range: Range<usize>) -> Option<LayerId> {
+        let start_idx = self
+            .injection_ranges
+            .partition_point(|range| range.end < search_range.end);
+
+        self.injection_ranges[start_idx..]
+            .iter()
+            .take_while(|range| range.end == search_range.end)
+            .find_map(|range| (range.start == search_range.start).then_some(range.layer_id))
+    }
+
+    fn goto_first_child_impl(&mut self, named: bool) -> bool {
+        // Check if the current node's range is an exact injection layer range.
+        if let Some(layer_id) = self
+            .layer_id_of_byte_range(self.node().byte_range())
+            .filter(|&layer_id| layer_id != self.current)
+        {
+            // Switch to the child layer.
+            self.current = layer_id;
+            self.cursor = self.layers[self.current].tree().root_node();
+            return true;
+        }
+
+        let child = if named {
+            self.cursor.named_child(0)
+        } else {
+            self.cursor.child(0)
+        };
+
+        if let Some(child) = child {
+            // Otherwise descend in the current tree.
+            self.cursor = child;
+            true
+        } else {
+            false
+        }
+    }
+
+    pub fn goto_first_child(&mut self) -> bool {
+        self.goto_first_child_impl(false)
+    }
+
+    pub fn goto_first_named_child(&mut self) -> bool {
+        self.goto_first_child_impl(true)
+    }
+
+    fn goto_next_sibling_impl(&mut self, named: bool) -> bool {
+        let sibling = if named {
+            self.cursor.next_named_sibling()
+        } else {
+            self.cursor.next_sibling()
+        };
+
+        if let Some(sibling) = sibling {
+            self.cursor = sibling;
+            true
+        } else {
+            false
+        }
+    }
+
+    pub fn goto_next_sibling(&mut self) -> bool {
+        self.goto_next_sibling_impl(false)
+    }
+
+    pub fn goto_next_named_sibling(&mut self) -> bool {
+        self.goto_next_sibling_impl(true)
+    }
+
+    fn goto_prev_sibling_impl(&mut self, named: bool) -> bool {
+        let sibling = if named {
+            self.cursor.prev_named_sibling()
+        } else {
+            self.cursor.prev_sibling()
+        };
+
+        if let Some(sibling) = sibling {
+            self.cursor = sibling;
+            true
+        } else {
+            false
+        }
+    }
+
+    pub fn goto_prev_sibling(&mut self) -> bool {
+        self.goto_prev_sibling_impl(false)
+    }
+
+    pub fn goto_prev_named_sibling(&mut self) -> bool {
+        self.goto_prev_sibling_impl(true)
+    }
+
+    /// Finds the injection layer that contains the given start-end range.
+    fn layer_id_containing_byte_range(&self, start: usize, end: usize) -> LayerId {
+        let start_idx = self
+            .injection_ranges
+            .partition_point(|range| range.end < end);
+
+        self.injection_ranges[start_idx..]
+            .iter()
+            .take_while(|range| range.start < end)
+            .find_map(|range| (range.start <= start).then_some(range.layer_id))
+            .unwrap_or(self.root)
+    }
+
+    pub fn reset_to_byte_range(&mut self, start: usize, end: usize) {
+        self.current = self.layer_id_containing_byte_range(start, end);
+        let root = self.layers[self.current].tree().root_node();
+        self.cursor = root.descendant_for_byte_range(start, end).unwrap_or(root);
+    }
+
+    /// Returns an iterator over the children of the node the TreeCursor is on
+    /// at the time this is called.
+    pub fn children(&'a mut self) -> ChildIter {
+        let parent = self.node();
+
+        ChildIter {
+            cursor: self,
+            parent,
+            named: false,
+        }
+    }
+
+    /// Returns an iterator over the named children of the node the TreeCursor is on
+    /// at the time this is called.
+    pub fn named_children(&'a mut self) -> ChildIter {
+        let parent = self.node();
+
+        ChildIter {
+            cursor: self,
+            parent,
+            named: true,
+        }
+    }
+}
+
+pub struct ChildIter<'n> {
+    cursor: &'n mut TreeCursor<'n>,
+    parent: Node<'n>,
+    named: bool,
+}
+
+impl<'n> Iterator for ChildIter<'n> {
+    type Item = Node<'n>;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        // first iteration, just visit the first child
+        if self.cursor.node() == self.parent {
+            self.cursor
+                .goto_first_child_impl(self.named)
+                .then(|| self.cursor.node())
+        } else {
+            self.cursor
+                .goto_next_sibling_impl(self.named)
+                .then(|| self.cursor.node())
+        }
+    }
+}

helix-syntax/src/tree_sitter.rs

@@ -0,0 +1,45 @@
+mod grammar;
+mod parser;
+pub mod query;
+mod query_cursor;
+mod query_match;
+mod ropey;
+mod syntax_tree;
+mod syntax_tree_node;
+
+use std::ops;
+
+pub use grammar::Grammar;
+pub use parser::{Parser, ParserInputRaw};
+pub use query::{Capture, Pattern, Query, QueryStr};
+pub use query_cursor::{InactiveQueryCursor, MatchedNode, MatchedNodeIdx, QueryCursor, QueryMatch};
+pub use ropey::RopeTsInput;
+pub use syntax_tree::{InputEdit, SyntaxTree};
+pub use syntax_tree_node::SyntaxTreeNode;
+
+#[repr(C)]
+#[derive(Debug, Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub struct Point {
+    pub row: u32,
+    pub col: u32,
+}
+
+#[repr(C)]
+#[derive(Debug, Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub struct Range {
+    pub start_point: Point,
+    pub end_point: Point,
+    pub start_byte: u32,
+    pub end_byte: u32,
+}
+
+pub trait TsInput {
+    type Cursor: regex_cursor::Cursor;
+    fn cursor_at(&mut self, offset: usize) -> &mut Self::Cursor;
+    fn eq(&mut self, range1: ops::Range<usize>, range2: ops::Range<usize>) -> bool;
+}
+
+pub trait IntoTsInput {
+    type TsInput: TsInput;
+    fn into_ts_input(self) -> Self::TsInput;
+}

helix-syntax/src/tree_sitter/grammar.rs

@@ -0,0 +1,107 @@
+use std::fmt;
+use std::path::{Path, PathBuf};
+use std::ptr::NonNull;
+
+use libloading::{Library, Symbol};
+
+/// supported TS versions, WARNING: update when updating vendored c sources
+pub const MIN_COMPATIBLE_ABI_VERSION: u32 = 13;
+pub const ABI_VERSION: u32 = 14;
+
+// opaque pointer
+enum GrammarData {}
+
+#[repr(transparent)]
+#[derive(Clone, Copy, PartialEq, Eq, Hash)]
+pub struct Grammar {
+    ptr: NonNull<GrammarData>,
+}
+
+unsafe impl Send for Grammar {}
+unsafe impl Sync for Grammar {}
+
+impl std::fmt::Debug for Grammar {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("Grammar").finish_non_exhaustive()
+    }
+}
+
+impl Grammar {
+    /// Loads a shared library containg a tree sitter grammar with name `name`
+    // from `library_path`.
+    ///
+    /// # Safety
+    ///
+    /// `library_path` must be a valid tree sitter grammar
+    pub unsafe fn new(name: &str, library_path: &Path) -> Result<Grammar, Error> {
+        let library = unsafe {
+            Library::new(library_path).map_err(|err| Error::DlOpen {
+                err,
+                path: library_path.to_owned(),
+            })?
+        };
+        let language_fn_name = format!("tree_sitter_{}", name.replace('-', "_"));
+        let grammar = unsafe {
+            let language_fn: Symbol<unsafe extern "C" fn() -> NonNull<GrammarData>> = library
+                .get(language_fn_name.as_bytes())
+                .map_err(|err| Error::DlSym {
+                    err,
+                    symbol: name.to_owned(),
+                })?;
+            Grammar { ptr: language_fn() }
+        };
+        let version = grammar.version();
+        if (MIN_COMPATIBLE_ABI_VERSION..=ABI_VERSION).contains(&version) {
+            std::mem::forget(library);
+            Ok(grammar)
+        } else {
+            Err(Error::IncompatibleVersion { version })
+        }
+    }
+    pub fn version(self) -> u32 {
+        unsafe { ts_language_version(self) }
+    }
+}
+
+#[derive(thiserror::Error, Debug)]
+pub enum Error {
+    #[error("Error opening dynamic library {path:?}")]
+    DlOpen {
+        #[source]
+        err: libloading::Error,
+        path: PathBuf,
+    },
+    #[error("Failed to load symbol {symbol}")]
+    DlSym {
+        #[source]
+        err: libloading::Error,
+        symbol: String,
+    },
+    #[error("Tried to load grammar with incompatible ABI {version}.")]
+    IncompatibleVersion { version: u32 },
+}
+
+/// An error that occurred when trying to assign an incompatible [`Grammar`] to
+/// a [`Parser`].
+#[derive(Debug, PartialEq, Eq)]
+pub struct IncompatibleGrammarError {
+    version: u32,
+}
+
+impl fmt::Display for IncompatibleGrammarError {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(
+            f,
+            "Tried to load grammar with incompatible ABI {}.",
+            self.version,
+        )
+    }
+}
+impl std::error::Error for IncompatibleGrammarError {}
+
+extern "C" {
+    /// Get the ABI version number for this language. This version number
+    /// is used to ensure that languages were generated by a compatible version of
+    /// Tree-sitter. See also [`ts_parser_set_language`].
+    pub fn ts_language_version(grammar: Grammar) -> u32;
+}

helix-syntax/src/tree_sitter/parser.rs

@@ -0,0 +1,200 @@
+use std::os::raw::c_void;
+use std::panic::{catch_unwind, AssertUnwindSafe};
+use std::ptr::NonNull;
+use std::{fmt, ptr};
+
+use regex_cursor::Cursor;
+
+use crate::tree_sitter::syntax_tree::{SyntaxTree, SyntaxTreeData};
+use crate::tree_sitter::{Grammar, IntoTsInput, Point, Range, TsInput};
+
+// opaque data
+enum ParserData {}
+
+/// A stateful object that this is used to produce a [`Tree`] based on some
+/// source code.
+pub struct Parser {
+    ptr: NonNull<ParserData>,
+}
+
+impl Parser {
+    /// Create a new parser.
+    #[must_use]
+    pub fn new() -> Parser {
+        Parser {
+            ptr: unsafe { ts_parser_new() },
+        }
+    }
+
+    /// Set the language that the parser should use for parsing.
+    pub fn set_language(&mut self, grammar: Grammar) {
+        unsafe { ts_parser_set_language(self.ptr, grammar) };
+    }
+
+    /// Set the ranges of text that the parser should include when parsing. By default, the parser
+    /// will always include entire documents. This function allows you to parse only a *portion*
+    /// of a document but still return a syntax tree whose ranges match up with the document as a
+    /// whole. You can also pass multiple disjoint ranges.
+    ///
+    /// `ranges` must be non-overlapping and sorted.
+    pub fn set_included_ranges(&mut self, ranges: &[Range]) -> Result<(), InvalidRangesErrror> {
+        // TODO: save some memory by only storing byte ranges and converting them to TS ranges in an
+        // internal buffer here. Points are not used by TS. Alternatively we can path the TS C code
+        // to accept a simple pair (struct with two fields) of byte positions here instead of a full
+        // tree sitter range
+        let success = unsafe {
+            ts_parser_set_included_ranges(self.ptr, ranges.as_ptr(), ranges.len() as u32)
+        };
+        if success {
+            Ok(())
+        } else {
+            Err(InvalidRangesErrror)
+        }
+    }
+
+    #[must_use]
+    pub fn parse<I: TsInput>(
+        &mut self,
+        input: impl IntoTsInput<TsInput = I>,
+        old_tree: Option<&SyntaxTree>,
+    ) -> Option<SyntaxTree> {
+        let mut input = input.into_ts_input();
+        unsafe extern "C" fn read<C: TsInput>(
+            payload: NonNull<c_void>,
+            byte_index: u32,
+            _position: Point,
+            bytes_read: *mut u32,
+        ) -> *const u8 {
+            let cursor = catch_unwind(AssertUnwindSafe(move || {
+                let input: &mut C = payload.cast().as_mut();
+                let cursor = input.cursor_at(byte_index as usize);
+                let slice = cursor.chunk();
+                (slice.as_ptr(), slice.len().try_into().unwrap())
+            }));
+            match cursor {
+                Ok((ptr, len)) => {
+                    *bytes_read = len;
+                    ptr
+                }
+                Err(_) => {
+                    *bytes_read = 0;
+                    ptr::null()
+                }
+            }
+        }
+        let input = ParserInputRaw {
+            payload: NonNull::from(&mut input).cast(),
+            read: read::<I>,
+            // utf8
+            encoding: 0,
+        };
+        unsafe {
+            let old_tree = old_tree.map(|tree| tree.as_raw());
+            let new_tree = ts_parser_parse(self.ptr, old_tree, input);
+            new_tree.map(|raw| SyntaxTree::from_raw(raw))
+        }
+    }
+}
+
+impl Default for Parser {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+unsafe impl Sync for Parser {}
+unsafe impl Send for Parser {}
+impl Drop for Parser {
+    fn drop(&mut self) {
+        unsafe { ts_parser_delete(self.ptr) }
+    }
+}
+
+/// An error that occurred when trying to assign an incompatible [`Grammar`] to
+/// a [`Parser`].
+#[derive(Debug, PartialEq, Eq)]
+pub struct InvalidRangesErrror;
+
+impl fmt::Display for InvalidRangesErrror {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "include ranges are overlap or are not sorted",)
+    }
+}
+impl std::error::Error for InvalidRangesErrror {}
+
+type TreeSitterReadFn = unsafe extern "C" fn(
+    payload: NonNull<c_void>,
+    byte_index: u32,
+    position: Point,
+    bytes_read: *mut u32,
+) -> *const u8;
+
+#[repr(C)]
+#[derive(Debug)]
+pub struct ParserInputRaw {
+    pub payload: NonNull<c_void>,
+    pub read: TreeSitterReadFn,
+    pub encoding: u32,
+}
+
+extern "C" {
+    /// Create a new parser
+    fn ts_parser_new() -> NonNull<ParserData>;
+    /// Delete the parser, freeing all of the memory that it used.
+    fn ts_parser_delete(parser: NonNull<ParserData>);
+    /// Set the language that the parser should use for parsing. Returns a boolean indicating
+    /// whether or not the language was successfully assigned. True means assignment
+    /// succeeded. False means there was a version mismatch: the language was generated with
+    /// an incompatible version of the Tree-sitter CLI. Check the language's version using
+    /// [`ts_language_version`] and compare it to this library's [`TREE_SITTER_LANGUAGE_VERSION`]
+    /// and [`TREE_SITTER_MIN_COMPATIBLE_LANGUAGE_VERSION`] constants.
+    fn ts_parser_set_language(parser: NonNull<ParserData>, language: Grammar) -> bool;
+    /// Set the ranges of text that the parser should include when parsing. By default, the parser
+    /// will always include entire documents. This function allows you to parse only a *portion*
+    /// of a document but still return a syntax tree whose ranges match up with the document as a
+    /// whole. You can also pass multiple disjoint ranges. The second and third parameters specify
+    /// the location and length of an array of ranges. The parser does *not* take ownership of
+    /// these ranges; it copies the data, so it doesn't matter how these ranges are allocated.
+    /// If `count` is zero, then the entire document will be parsed. Otherwise, the given ranges
+    /// must be ordered from earliest to latest in the document, and they must not overlap. That
+    /// is, the following must hold for all: `i < count - 1`: `ranges[i].end_byte <= ranges[i +
+    /// 1].start_byte` If this requirement is not satisfied, the operation will fail, the ranges
+    /// will not be assigned, and this function will return `false`. On success, this function
+    /// returns `true`
+    fn ts_parser_set_included_ranges(
+        parser: NonNull<ParserData>,
+        ranges: *const Range,
+        count: u32,
+    ) -> bool;
+
+    /// Use the parser to parse some source code and create a syntax tree. If you are parsing this
+    /// document for the first time, pass `NULL` for the `old_tree` parameter. Otherwise, if you
+    /// have already parsed an earlier version of this document and the document has since been
+    /// edited, pass the previous syntax tree so that the unchanged parts of it can be reused.
+    /// This will save time and memory. For this to work correctly, you must have already edited
+    /// the old syntax tree using the [`ts_tree_edit`] function in a way that exactly matches
+    /// the source code changes. The [`TSInput`] parameter lets you specify how to read the text.
+    /// It has the following three fields: 1. [`read`]: A function to retrieve a chunk of text
+    /// at a given byte offset and (row, column) position. The function should return a pointer
+    /// to the text and write its length to the [`bytes_read`] pointer. The parser does not
+    /// take ownership of this buffer; it just borrows it until it has finished reading it. The
+    /// function should write a zero value to the [`bytes_read`] pointer to indicate the end of the
+    /// document. 2. [`payload`]: An arbitrary pointer that will be passed to each invocation of
+    /// the [`read`] function. 3. [`encoding`]: An indication of how the text is encoded. Either
+    /// `TSInputEncodingUTF8` or `TSInputEncodingUTF16`. This function returns a syntax tree
+    /// on success, and `NULL` on failure. There are three possible reasons for failure: 1. The
+    /// parser does not have a language assigned. Check for this using the [`ts_parser_language`]
+    /// function. 2. Parsing was cancelled due to a timeout that was set by an earlier call to the
+    /// [`ts_parser_set_timeout_micros`] function. You can resume parsing from where the parser
+    /// left out by calling [`ts_parser_parse`] again with the same arguments. Or you can start
+    /// parsing from scratch by first calling [`ts_parser_reset`]. 3. Parsing was cancelled using
+    /// a cancellation flag that was set by an earlier call to [`ts_parser_set_cancellation_flag`].
+    /// You can resume parsing from where the parser left out by calling [`ts_parser_parse`] again
+    /// with the same arguments. [`read`]: TSInput::read [`payload`]: TSInput::payload [`encoding`]:
+    /// TSInput::encoding [`bytes_read`]: TSInput::read
+    fn ts_parser_parse(
+        parser: NonNull<ParserData>,
+        old_tree: Option<NonNull<SyntaxTreeData>>,
+        input: ParserInputRaw,
+    ) -> Option<NonNull<SyntaxTreeData>>;
+}

helix-syntax/src/tree_sitter/query.rs

@@ -0,0 +1,451 @@
+use std::fmt::{self, Display};
+use std::ops::Range;
+use std::path::{Path, PathBuf};
+use std::ptr::NonNull;
+use std::{slice, str};
+
+use crate::tree_sitter::query::predicate::{InvalidPredicateError, Predicate, TextPredicate};
+use crate::tree_sitter::Grammar;
+
+mod predicate;
+mod property;
+
+pub enum UserPredicate<'a> {
+    IsPropertySet {
+        negate: bool,
+        key: &'a str,
+        val: Option<&'a str>,
+    },
+    SetProperty {
+        key: &'a str,
+        val: Option<&'a str>,
+    },
+    Other(Predicate<'a>),
+}
+
+impl Display for UserPredicate<'_> {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match *self {
+            UserPredicate::IsPropertySet { negate, key, val } => {
+                let predicate = if negate { "is-not?" } else { "is?" };
+                write!(f, " ({predicate} {key} {})", val.unwrap_or(""))
+            }
+            UserPredicate::SetProperty { key, val } => {
+                write!(f, "(set! {key} {})", val.unwrap_or(""))
+            }
+            UserPredicate::Other(ref predicate) => {
+                write!(f, "{}", predicate.name())
+            }
+        }
+    }
+}
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+pub struct Pattern(pub(crate) u32);
+
+impl Pattern {
+    pub const SENTINEL: Pattern = Pattern(u32::MAX);
+    pub fn idx(&self) -> usize {
+        self.0 as usize
+    }
+}
+
+pub enum QueryData {}
+
+#[derive(Debug)]
+pub(super) struct PatternData {
+    text_predicates: Range<u32>,
+}
+
+#[derive(Debug)]
+pub struct Query {
+    pub(crate) raw: NonNull<QueryData>,
+    num_captures: u32,
+    num_strings: u32,
+    text_predicates: Vec<TextPredicate>,
+    patterns: Box<[PatternData]>,
+}
+
+impl Query {
+    /// Create a new query from a string containing one or more S-expression
+    /// patterns.
+    ///
+    /// The query is associated with a particular grammar, and can only be run
+    /// on syntax nodes parsed with that grammar. References to Queries can be
+    /// shared between multiple threads.
+    pub fn new(
+        grammar: Grammar,
+        source: &str,
+        path: impl AsRef<Path>,
+        mut custom_predicate: impl FnMut(Pattern, UserPredicate) -> Result<(), InvalidPredicateError>,
+    ) -> Result<Self, ParseError> {
+        assert!(
+            source.len() <= i32::MAX as usize,
+            "TreeSitter queries must be smaller then 2 GiB (is {})",
+            source.len() as f64 / 1024.0 / 1024.0 / 1024.0
+        );
+        let mut error_offset = 0u32;
+        let mut error_kind = RawQueryError::None;
+        let bytes = source.as_bytes();
+
+        // Compile the query.
+        let ptr = unsafe {
+            ts_query_new(
+                grammar,
+                bytes.as_ptr(),
+                bytes.len() as u32,
+                &mut error_offset,
+                &mut error_kind,
+            )
+        };
+
+        let Some(raw) = ptr else {
+            let offset = error_offset as usize;
+            let error_word = || {
+                source[offset..]
+                    .chars()
+                    .take_while(|&c| c.is_alphanumeric() || matches!(c, '_' | '-'))
+                    .collect()
+            };
+            let err = match error_kind {
+                RawQueryError::NodeType => {
+                    let node: String = error_word();
+                    ParseError::InvalidNodeType {
+                        location: ParserErrorLocation::new(
+                            source,
+                            path.as_ref(),
+                            offset,
+                            node.chars().count(),
+                        ),
+                        node,
+                    }
+                }
+                RawQueryError::Field => {
+                    let field = error_word();
+                    ParseError::InvalidFieldName {
+                        location: ParserErrorLocation::new(
+                            source,
+                            path.as_ref(),
+                            offset,
+                            field.chars().count(),
+                        ),
+                        field,
+                    }
+                }
+                RawQueryError::Capture => {
+                    let capture = error_word();
+                    ParseError::InvalidCaptureName {
+                        location: ParserErrorLocation::new(
+                            source,
+                            path.as_ref(),
+                            offset,
+                            capture.chars().count(),
+                        ),
+                        capture,
+                    }
+                }
+                RawQueryError::Syntax => ParseError::SyntaxError(ParserErrorLocation::new(
+                    source,
+                    path.as_ref(),
+                    offset,
+                    0,
+                )),
+                RawQueryError::Structure => ParseError::ImpossiblePattern(
+                    ParserErrorLocation::new(source, path.as_ref(), offset, 0),
+                ),
+                RawQueryError::None => {
+                    unreachable!("tree-sitter returned a null pointer but did not set an error")
+                }
+                RawQueryError::Language => unreachable!("should be handled at grammar load"),
+            };
+            return Err(err);
+        };
+
+        // I am not going to bother with safety comments here, all of these are
+        // safe as long as TS is not buggy because raw is a properly constructed query
+        let num_captures = unsafe { ts_query_capture_count(raw) };
+        let num_strings = unsafe { ts_query_string_count(raw) };
+        let num_patterns = unsafe { ts_query_pattern_count(raw) };
+
+        let mut query = Query {
+            raw,
+            num_captures,
+            num_strings,
+            text_predicates: Vec::new(),
+            patterns: Box::default(),
+        };
+        let patterns: Result<_, ParseError> = (0..num_patterns)
+            .map(|pattern| {
+                query
+                    .parse_pattern_predicates(Pattern(pattern), &mut custom_predicate)
+                    .map_err(|err| ParseError::InvalidPredicate {
+                        message: err.msg.into(),
+                        location: ParserErrorLocation::new(
+                            source,
+                            path.as_ref(),
+                            unsafe { ts_query_start_byte_for_pattern(query.raw, pattern) as usize },
+                            0,
+                        ),
+                    })
+            })
+            .collect();
+        query.patterns = patterns?;
+        Ok(query)
+    }
+
+    #[inline]
+    fn get_string(&self, str: QueryStr) -> &str {
+        let value_id = str.0;
+        // need an assertions because the ts c api does not do bounds check
+        assert!(value_id <= self.num_captures, "invalid value index");
+        unsafe {
+            let mut len = 0;
+            let ptr = ts_query_string_value_for_id(self.raw, value_id, &mut len);
+            let data = slice::from_raw_parts(ptr, len as usize);
+            // safety: we only allow passing valid str(ings) as arguments to query::new
+            // name is always a substring of that. Treesitter does proper utf8 segmentation
+            // so any substrings it produces are codepoint aligned and therefore valid utf8
+            str::from_utf8_unchecked(data)
+        }
+    }
+
+    #[inline]
+    pub fn capture_name(&self, capture_idx: Capture) -> &str {
+        let capture_idx = capture_idx.0;
+        // need an assertions because the ts c api does not do bounds check
+        assert!(capture_idx <= self.num_captures, "invalid capture index");
+        let mut length = 0;
+        unsafe {
+            let ptr = ts_query_capture_name_for_id(self.raw, capture_idx, &mut length);
+            let name = slice::from_raw_parts(ptr, length as usize);
+            // safety: we only allow passing valid str(ings) as arguments to query::new
+            // name is always a substring of that. Treesitter does proper utf8 segmentation
+            // so any substrings it produces are codepoint aligned and therefore valid utf8
+            str::from_utf8_unchecked(name)
+        }
+    }
+
+    #[inline]
+    pub fn captures(&self) -> impl ExactSizeIterator<Item = (Capture, &str)> {
+        (0..self.num_captures).map(|cap| (Capture(cap), self.capture_name(Capture(cap))))
+    }
+
+    #[inline]
+    pub fn num_captures(&self) -> u32 {
+        self.num_captures
+    }
+
+    #[inline]
+    pub fn get_capture(&self, capture_name: &str) -> Option<Capture> {
+        for capture in 0..self.num_captures {
+            if capture_name == self.capture_name(Capture(capture)) {
+                return Some(Capture(capture));
+            }
+        }
+        None
+    }
+
+    pub(crate) fn pattern_text_predicates(&self, pattern_idx: u16) -> &[TextPredicate] {
+        let range = self.patterns[pattern_idx as usize].text_predicates.clone();
+        &self.text_predicates[range.start as usize..range.end as usize]
+    }
+
+    /// Get the byte offset where the given pattern starts in the query's
+    /// source.
+    #[doc(alias = "ts_query_start_byte_for_pattern")]
+    #[must_use]
+    pub fn start_byte_for_pattern(&self, pattern: Pattern) -> usize {
+        assert!(
+            pattern.0 < self.text_predicates.len() as u32,
+            "Pattern index is {pattern_index} but the pattern count is {}",
+            self.text_predicates.len(),
+        );
+        unsafe { ts_query_start_byte_for_pattern(self.raw, pattern.0) as usize }
+    }
+
+    /// Get the number of patterns in the query.
+    #[must_use]
+    pub fn pattern_count(&self) -> usize {
+        unsafe { ts_query_pattern_count(self.raw) as usize }
+    }
+    /// Get the number of patterns in the query.
+    #[must_use]
+    pub fn patterns(&self) -> impl ExactSizeIterator<Item = Pattern> {
+        (0..self.pattern_count() as u32).map(Pattern)
+    }
+}
+
+impl Drop for Query {
+    fn drop(&mut self) {
+        unsafe { ts_query_delete(self.raw) }
+    }
+}
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+#[repr(transparent)]
+pub struct Capture(u32);
+
+impl Capture {
+    pub fn name(self, query: &Query) -> &str {
+        query.capture_name(self)
+    }
+    pub fn idx(self) -> usize {
+        self.0 as usize
+    }
+}
+
+/// A reference to a string stroed in a query
+#[derive(Clone, Copy, Debug)]
+pub struct QueryStr(u32);
+
+impl QueryStr {
+    pub fn get(self, query: &Query) -> &str {
+        query.get_string(self)
+    }
+}
+
+#[derive(Debug, PartialEq, Eq)]
+pub struct ParserErrorLocation {
+    pub path: PathBuf,
+    /// at which line the error occured
+    pub line: usize,
+    /// at which codepoints/columns the errors starts in the line
+    pub column: usize,
+    /// how many codepoints/columns the error takes up
+    pub len: usize,
+    line_content: String,
+}
+
+impl ParserErrorLocation {
+    pub fn new(source: &str, path: &Path, offset: usize, len: usize) -> ParserErrorLocation {
+        let (line, line_content) = source[..offset]
+            .split('\n')
+            .map(|line| line.strip_suffix('\r').unwrap_or(line))
+            .enumerate()
+            .last()
+            .unwrap_or((0, ""));
+        let column = line_content.chars().count();
+        ParserErrorLocation {
+            path: path.to_owned(),
+            line,
+            column,
+            len,
+            line_content: line_content.to_owned(),
+        }
+    }
+}
+
+impl Display for ParserErrorLocation {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        writeln!(
+            f,
+            "  --> {}:{}:{}",
+            self.path.display(),
+            self.line,
+            self.column
+        )?;
+        let line = self.line.to_string();
+        let prefix = format!(" {:width$} |", "", width = line.len());
+        writeln!(f, "{prefix}")?;
+        writeln!(f, " {line} | {}", self.line_content)?;
+        writeln!(
+            f,
+            "{prefix}{:width$}{:^<len$}",
+            "",
+            "^",
+            width = self.column,
+            len = self.len
+        )?;
+        writeln!(f, "{prefix}")
+    }
+}
+
+#[derive(thiserror::Error, Debug, PartialEq, Eq)]
+pub enum ParseError {
+    #[error("unexpected EOF")]
+    UnexpectedEof,
+    #[error("invalid query syntax\n{0}")]
+    SyntaxError(ParserErrorLocation),
+    #[error("invalid node type {node:?}\n{location}")]
+    InvalidNodeType {
+        node: String,
+        location: ParserErrorLocation,
+    },
+    #[error("invalid field name {field:?}\n{location}")]
+    InvalidFieldName {
+        field: String,
+        location: ParserErrorLocation,
+    },
+    #[error("invalid capture name {capture:?}\n{location}")]
+    InvalidCaptureName {
+        capture: String,
+        location: ParserErrorLocation,
+    },
+    #[error("{message}\n{location}")]
+    InvalidPredicate {
+        message: String,
+        location: ParserErrorLocation,
+    },
+    #[error("invalid predicate\n{0}")]
+    ImpossiblePattern(ParserErrorLocation),
+}
+
+#[repr(C)]
+enum RawQueryError {
+    None = 0,
+    Syntax = 1,
+    NodeType = 2,
+    Field = 3,
+    Capture = 4,
+    Structure = 5,
+    Language = 6,
+}
+
+extern "C" {
+    /// Create a new query from a string containing one or more S-expression
+    /// patterns. The query is associated with a particular language, and can
+    /// only be run on syntax nodes parsed with that language. If all of the
+    /// given patterns are valid, this returns a [`TSQuery`]. If a pattern is
+    /// invalid, this returns `NULL`, and provides two pieces of information
+    /// about the problem: 1. The byte offset of the error is written to
+    /// the `error_offset` parameter. 2. The type of error is written to the
+    /// `error_type` parameter.
+    fn ts_query_new(
+        grammar: Grammar,
+        source: *const u8,
+        source_len: u32,
+        error_offset: &mut u32,
+        error_type: &mut RawQueryError,
+    ) -> Option<NonNull<QueryData>>;
+
+    /// Delete a query, freeing all of the memory that it used.
+    fn ts_query_delete(query: NonNull<QueryData>);
+
+    /// Get the number of patterns, captures, or string literals in the query.
+    fn ts_query_pattern_count(query: NonNull<QueryData>) -> u32;
+    fn ts_query_capture_count(query: NonNull<QueryData>) -> u32;
+    fn ts_query_string_count(query: NonNull<QueryData>) -> u32;
+
+    /// Get the byte offset where the given pattern starts in the query's
+    /// source. This can be useful when combining queries by concatenating their
+    /// source code strings.
+    fn ts_query_start_byte_for_pattern(query: NonNull<QueryData>, pattern_index: u32) -> u32;
+
+    // fn ts_query_is_pattern_rooted(query: NonNull<QueryData>, pattern_index: u32) -> bool;
+    // fn ts_query_is_pattern_non_local(query: NonNull<QueryData>, pattern_index: u32) -> bool;
+    // fn ts_query_is_pattern_guaranteed_at_step(query: NonNull<QueryData>, byte_offset: u32) -> bool;
+    /// Get the name and length of one of the query's captures, or one of the
+    /// query's string literals. Each capture and string is associated with a
+    /// numeric id based on the order that it appeared in the query's source.
+    fn ts_query_capture_name_for_id(
+        query: NonNull<QueryData>,
+        index: u32,
+        length: &mut u32,
+    ) -> *const u8;
+
+    fn ts_query_string_value_for_id(
+        self_: NonNull<QueryData>,
+        index: u32,
+        length: &mut u32,
+    ) -> *const u8;
+}

helix-syntax/src/tree_sitter/query/predicate.rs

@@ -0,0 +1,448 @@
+use std::error::Error;
+use std::iter::zip;
+use std::ops::Range;
+use std::ptr::NonNull;
+use std::{fmt, slice};
+
+use crate::tree_sitter::query::property::QueryProperty;
+use crate::tree_sitter::query::{
+    Capture, Pattern, PatternData, Query, QueryData, QueryStr, UserPredicate,
+};
+use crate::tree_sitter::query_cursor::MatchedNode;
+use crate::tree_sitter::TsInput;
+
+use regex_cursor::engines::meta::Regex;
+use regex_cursor::Cursor;
+
+macro_rules! bail {
+    ($($args:tt)*) => {{
+        return Err(InvalidPredicateError {msg: format!($($args)*).into() })
+    }}
+}
+
+macro_rules! ensure {
+    ($cond: expr, $($args:tt)*) => {{
+        if !$cond {
+            return Err(InvalidPredicateError { msg: format!($($args)*).into() })
+        }
+    }}
+}
+
+#[derive(Debug)]
+pub(super) enum TextPredicateKind {
+    EqString(QueryStr),
+    EqCapture(Capture),
+    MatchString(Regex),
+    AnyString(Box<[QueryStr]>),
+}
+
+#[derive(Debug)]
+pub(crate) struct TextPredicate {
+    capture: Capture,
+    kind: TextPredicateKind,
+    negated: bool,
+    match_all: bool,
+}
+
+fn input_matches_str<I: TsInput>(str: &str, range: Range<usize>, input: &mut I) -> bool {
+    if str.len() != range.len() {
+        return false;
+    }
+    let mut str = str.as_bytes();
+    let cursor = input.cursor_at(range.start);
+    let start_in_chunk = range.start - cursor.offset();
+    if range.end - cursor.offset() <= cursor.chunk().len() {
+        // hotpath
+        return &cursor.chunk()[start_in_chunk..range.end - cursor.offset()] == str;
+    }
+    if cursor.chunk()[start_in_chunk..] != str[..cursor.chunk().len() - start_in_chunk] {
+        return false;
+    }
+    str = &str[..cursor.chunk().len() - start_in_chunk];
+    while cursor.advance() {
+        if str.len() <= cursor.chunk().len() {
+            return &cursor.chunk()[..range.end - cursor.offset()] == str;
+        }
+        if &str[..cursor.chunk().len()] != cursor.chunk() {
+            return false;
+        }
+        str = &str[cursor.chunk().len()..]
+    }
+    // buggy cursor/invalid range
+    false
+}
+
+fn inputs_match<I: TsInput>(str: &str, range: Range<usize>, input: &mut I) -> bool {
+    if str.len() != range.len() {
+        return false;
+    }
+    let mut str = str.as_bytes();
+    let cursor = input.cursor_at(range.start);
+    let start_in_chunk = range.start - cursor.offset();
+    if range.end - cursor.offset() <= cursor.chunk().len() {
+        // hotpath
+        return &cursor.chunk()[start_in_chunk..range.end - cursor.offset()] == str;
+    }
+    if cursor.chunk()[start_in_chunk..] != str[..cursor.chunk().len() - start_in_chunk] {
+        return false;
+    }
+    str = &str[..cursor.chunk().len() - start_in_chunk];
+    while cursor.advance() {
+        if str.len() <= cursor.chunk().len() {
+            return &cursor.chunk()[..range.end - cursor.offset()] == str;
+        }
+        if &str[..cursor.chunk().len()] != cursor.chunk() {
+            return false;
+        }
+        str = &str[cursor.chunk().len()..]
+    }
+    // buggy cursor/invalid range
+    false
+}
+
+impl TextPredicate {
+    /// handlers match_all and negated
+    fn satisfied_helper(&self, mut nodes: impl Iterator<Item = bool>) -> bool {
+        if self.match_all {
+            nodes.all(|matched| matched != self.negated)
+        } else {
+            nodes.any(|matched| matched != self.negated)
+        }
+    }
+
+    pub fn satsified<I: TsInput>(
+        &self,
+        input: &mut I,
+        matched_nodes: &[MatchedNode],
+        query: &Query,
+    ) -> bool {
+        let mut capture_nodes = matched_nodes
+            .iter()
+            .filter(|matched_node| matched_node.capture == self.capture);
+        match self.kind {
+            TextPredicateKind::EqString(str) => self.satisfied_helper(capture_nodes.map(|node| {
+                let range = node.syntax_node.byte_range();
+                input_matches_str(query.get_string(str), range.clone(), input)
+            })),
+            TextPredicateKind::EqCapture(other_capture) => {
+                let mut other_nodes = matched_nodes
+                    .iter()
+                    .filter(|matched_node| matched_node.capture == other_capture);
+
+                let res = self.satisfied_helper(zip(&mut capture_nodes, &mut other_nodes).map(
+                    |(node1, node2)| {
+                        let range1 = node1.syntax_node.byte_range();
+                        let range2 = node2.syntax_node.byte_range();
+                        input.eq(range1, range2)
+                    },
+                ));
+                let consumed_all = capture_nodes.next().is_none() && other_nodes.next().is_none();
+                res && (!self.match_all || consumed_all)
+            }
+            TextPredicateKind::MatchString(ref regex) => {
+                self.satisfied_helper(capture_nodes.map(|node| {
+                    let range = node.syntax_node.byte_range();
+                    let input = regex_cursor::Input::new(input.cursor_at(range.start)).range(range);
+                    regex.is_match(input)
+                }))
+            }
+            TextPredicateKind::AnyString(ref strings) => {
+                let strings = strings.iter().map(|&str| query.get_string(str));
+                self.satisfied_helper(capture_nodes.map(|node| {
+                    let range = node.syntax_node.byte_range();
+                    strings
+                        .clone()
+                        .filter(|str| str.len() == range.len())
+                        .any(|str| input_matches_str(str, range.clone(), input))
+                }))
+            }
+        }
+    }
+}
+
+impl Query {
+    pub(super) fn parse_pattern_predicates(
+        &mut self,
+        pattern: Pattern,
+        mut custom_predicate: impl FnMut(Pattern, UserPredicate) -> Result<(), InvalidPredicateError>,
+    ) -> Result<PatternData, InvalidPredicateError> {
+        let text_predicate_start = self.text_predicates.len() as u32;
+
+        let predicate_steps = unsafe {
+            let mut len = 0u32;
+            let raw_predicates = ts_query_predicates_for_pattern(self.raw, pattern.0, &mut len);
+            (len != 0)
+                .then(|| slice::from_raw_parts(raw_predicates, len as usize))
+                .unwrap_or_default()
+        };
+        let predicates = predicate_steps
+            .split(|step| step.kind == PredicateStepKind::Done)
+            .filter(|predicate| !predicate.is_empty());
+
+        for predicate in predicates {
+            let predicate = unsafe { Predicate::new(self, predicate)? };
+
+            match predicate.name() {
+                "eq?" | "not-eq?" | "any-eq?" | "any-not-eq?" => {
+                    predicate.check_arg_count(2)?;
+                    let capture_idx = predicate.capture_arg(0)?;
+                    let arg2 = predicate.arg(1);
+
+                    let negated = matches!(predicate.name(), "not-eq?" | "not-any-eq?");
+                    let match_all = matches!(predicate.name(), "eq?" | "not-eq?");
+                    let kind = match arg2 {
+                        PredicateArg::Capture(capture) => TextPredicateKind::EqCapture(capture),
+                        PredicateArg::String(str) => TextPredicateKind::EqString(str),
+                    };
+                    self.text_predicates.push(TextPredicate {
+                        capture: capture_idx,
+                        kind,
+                        negated,
+                        match_all,
+                    });
+                }
+
+                "match?" | "not-match?" | "any-match?" | "any-not-match?" => {
+                    predicate.check_arg_count(2)?;
+                    let capture_idx = predicate.capture_arg(0)?;
+                    let regex = predicate.query_str_arg(1)?.get(self);
+
+                    let negated = matches!(predicate.name(), "not-match?" | "any-not-match?");
+                    let match_all = matches!(predicate.name(), "match?" | "not-match?");
+                    let regex = match Regex::new(regex) {
+                        Ok(regex) => regex,
+                        Err(err) => bail!("invalid regex '{regex}', {err}"),
+                    };
+                    self.text_predicates.push(TextPredicate {
+                        capture: capture_idx,
+                        kind: TextPredicateKind::MatchString(regex),
+                        negated,
+                        match_all,
+                    });
+                }
+
+                "set!" => {
+                    let property = QueryProperty::parse(&predicate)?;
+                    custom_predicate(
+                        pattern,
+                        UserPredicate::SetProperty {
+                            key: property.key.get(&self),
+                            val: property.val.map(|val| val.get(&self)),
+                        },
+                    )?
+                }
+                "is-not?" | "is?" => {
+                    let property = QueryProperty::parse(&predicate)?;
+                    custom_predicate(
+                        pattern,
+                        UserPredicate::IsPropertySet {
+                            negate: predicate.name() == "is-not?",
+                            key: property.key.get(&self),
+                            val: property.val.map(|val| val.get(&self)),
+                        },
+                    )?
+                }
+
+                "any-of?" | "not-any-of?" => {
+                    predicate.check_min_arg_count(1)?;
+                    let capture = predicate.capture_arg(0)?;
+                    let negated = predicate.name() == "not-any-of?";
+                    let values: Result<_, InvalidPredicateError> = (1..predicate.num_args())
+                        .map(|i| predicate.query_str_arg(i))
+                        .collect();
+                    self.text_predicates.push(TextPredicate {
+                        capture,
+                        kind: TextPredicateKind::AnyString(values?),
+                        negated,
+                        match_all: false,
+                    });
+                }
+
+                // is and is-not are better handeled as custom predicates since interpreting is context dependent
+                // "is?" => property_predicates.push((QueryProperty::parse(&predicate), false)),
+                // "is-not?" => property_predicates.push((QueryProperty::parse(&predicate), true)),
+                _ => custom_predicate(pattern, UserPredicate::Other(predicate))?,
+            }
+        }
+        Ok(PatternData {
+            text_predicates: text_predicate_start..self.text_predicates.len() as u32,
+        })
+    }
+}
+
+pub enum PredicateArg {
+    Capture(Capture),
+    String(QueryStr),
+}
+
+pub struct Predicate<'a> {
+    pub name: QueryStr,
+    args: &'a [PredicateStep],
+    query: &'a Query,
+}
+
+impl<'a> Predicate<'a> {
+    unsafe fn new(
+        query: &'a Query,
+        predicate: &'a [PredicateStep],
+    ) -> Result<Predicate<'a>, InvalidPredicateError> {
+        ensure!(
+            predicate[0].kind == PredicateStepKind::String,
+            "expected predicate to start with a function name. Got @{}.",
+            Capture(predicate[0].value_id).name(query)
+        );
+        let operator_name = QueryStr(predicate[0].value_id);
+        Ok(Predicate {
+            name: operator_name,
+            args: &predicate[1..],
+            query,
+        })
+    }
+
+    pub fn name(&self) -> &str {
+        self.name.get(self.query)
+    }
+
+    pub fn check_arg_count(&self, n: usize) -> Result<(), InvalidPredicateError> {
+        ensure!(
+            self.args.len() == n,
+            "expected {n} arguments for #{}, got {}",
+            self.name(),
+            self.args.len()
+        );
+        Ok(())
+    }
+
+    pub fn check_min_arg_count(&self, n: usize) -> Result<(), InvalidPredicateError> {
+        ensure!(
+            n <= self.args.len(),
+            "expected at least {n} arguments for #{}, got {}",
+            self.name(),
+            self.args.len()
+        );
+        Ok(())
+    }
+
+    pub fn check_max_arg_count(&self, n: usize) -> Result<(), InvalidPredicateError> {
+        ensure!(
+            self.args.len() <= n,
+            "expected at most {n} arguments for #{}, got {}",
+            self.name(),
+            self.args.len()
+        );
+        Ok(())
+    }
+
+    pub fn query_str_arg(&self, i: usize) -> Result<QueryStr, InvalidPredicateError> {
+        match self.arg(i) {
+            PredicateArg::String(str) => Ok(str),
+            PredicateArg::Capture(capture) => bail!(
+                "{i}. argument to #{} must be a literal, got capture @{:?}",
+                self.name(),
+                capture.name(self.query)
+            ),
+        }
+    }
+
+    pub fn str_arg(&self, i: usize) -> Result<&str, InvalidPredicateError> {
+        Ok(self.query_str_arg(i)?.get(self.query))
+    }
+
+    pub fn num_args(&self) -> usize {
+        self.args.len()
+    }
+
+    pub fn capture_arg(&self, i: usize) -> Result<Capture, InvalidPredicateError> {
+        match self.arg(i) {
+            PredicateArg::Capture(capture) => Ok(capture),
+            PredicateArg::String(str) => bail!(
+                "{i}. argument to #{} expected a capture, got literal {:?}",
+                self.name(),
+                str.get(self.query)
+            ),
+        }
+    }
+
+    pub fn arg(&self, i: usize) -> PredicateArg {
+        self.args[i].try_into().unwrap()
+    }
+
+    pub fn args(&self) -> impl Iterator<Item = PredicateArg> + '_ {
+        self.args.iter().map(|&arg| arg.try_into().unwrap())
+    }
+}
+
+#[derive(Debug)]
+pub struct InvalidPredicateError {
+    pub(super) msg: Box<str>,
+}
+
+impl From<String> for InvalidPredicateError {
+    fn from(value: String) -> Self {
+        InvalidPredicateError {
+            msg: value.into_boxed_str(),
+        }
+    }
+}
+
+impl<'a> From<&'a str> for InvalidPredicateError {
+    fn from(value: &'a str) -> Self {
+        InvalidPredicateError { msg: value.into() }
+    }
+}
+
+impl fmt::Display for InvalidPredicateError {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        f.write_str(&self.msg)
+    }
+}
+
+impl Error for InvalidPredicateError {}
+
+#[repr(C)]
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+enum PredicateStepKind {
+    Done = 0,
+    Capture = 1,
+    String = 2,
+}
+
+#[repr(C)]
+#[derive(Debug, Clone, Copy)]
+struct PredicateStep {
+    kind: PredicateStepKind,
+    value_id: u32,
+}
+
+impl TryFrom<PredicateStep> for PredicateArg {
+    type Error = ();
+
+    fn try_from(step: PredicateStep) -> Result<Self, Self::Error> {
+        match step.kind {
+            PredicateStepKind::String => Ok(PredicateArg::String(QueryStr(step.value_id))),
+            PredicateStepKind::Capture => Ok(PredicateArg::Capture(Capture(step.value_id))),
+            PredicateStepKind::Done => Err(()),
+        }
+    }
+}
+
+extern "C" {
+    /// Get all of the predicates for the given pattern in the query. The
+    /// predicates are represented as a single array of steps. There are three
+    /// types of steps in this array, which correspond to the three legal values
+    /// for the `type` field: - `TSQueryPredicateStepTypeCapture` - Steps with
+    /// this type represent names    of captures. Their `value_id` can be used
+    /// with the   [`ts_query_capture_name_for_id`] function to obtain the name
+    /// of the capture. - `TSQueryPredicateStepTypeString` - Steps with this
+    /// type represent literal    strings. Their `value_id` can be used with the
+    /// [`ts_query_string_value_for_id`] function to obtain their string value.
+    /// - `TSQueryPredicateStepTypeDone` - Steps with this type are *sentinels*
+    /// that represent the end of an individual predicate. If a pattern has two
+    /// predicates, then there will be two steps with this `type` in the array.
+    fn ts_query_predicates_for_pattern(
+        query: NonNull<QueryData>,
+        pattern_index: u32,
+        step_count: &mut u32,
+    ) -> *const PredicateStep;
+
+}

helix-syntax/src/tree_sitter/query/property.rs

@@ -0,0 +1,20 @@
+use crate::tree_sitter::query::predicate::{InvalidPredicateError, Predicate};
+use crate::tree_sitter::query::QueryStr;
+
+#[derive(Debug)]
+pub struct QueryProperty {
+    pub key: QueryStr,
+    pub val: Option<QueryStr>,
+}
+
+impl QueryProperty {
+    pub fn parse(predicate: &Predicate) -> Result<Self, InvalidPredicateError> {
+        predicate.check_min_arg_count(1)?;
+        predicate.check_max_arg_count(2)?;
+        let key = predicate.query_str_arg(0)?;
+        let val = (predicate.num_args() == 1)
+            .then(|| predicate.query_str_arg(1))
+            .transpose()?;
+        Ok(QueryProperty { key, val })
+    }
+}

helix-syntax/src/tree_sitter/query_cursor.rs

@@ -0,0 +1,343 @@
+use core::slice;
+use std::cell::UnsafeCell;
+use std::marker::PhantomData;
+use std::mem::replace;
+use std::ops::Range;
+use std::ptr::{self, NonNull};
+
+use crate::tree_sitter::query::{Capture, Pattern, Query, QueryData};
+use crate::tree_sitter::syntax_tree_node::SyntaxTreeNodeRaw;
+use crate::tree_sitter::{SyntaxTree, SyntaxTreeNode, TsInput};
+
+enum QueryCursorData {}
+
+thread_local! {
+    static CURSOR_CACHE: UnsafeCell<Vec<InactiveQueryCursor>> = UnsafeCell::new(Vec::with_capacity(8));
+}
+
+/// SAFETY: must not call itself recuresively
+unsafe fn with_cache<T>(f: impl FnOnce(&mut Vec<InactiveQueryCursor>) -> T) -> T {
+    CURSOR_CACHE.with(|cache| f(&mut *cache.get()))
+}
+
+pub struct QueryCursor<'a, 'tree, I: TsInput> {
+    query: &'a Query,
+    ptr: *mut QueryCursorData,
+    tree: PhantomData<&'tree SyntaxTree>,
+    input: I,
+}
+
+impl<'tree, I: TsInput> QueryCursor<'_, 'tree, I> {
+    pub fn next_match(&mut self) -> Option<QueryMatch<'_, 'tree>> {
+        let mut query_match = TSQueryMatch {
+            id: 0,
+            pattern_index: 0,
+            capture_count: 0,
+            captures: ptr::null(),
+        };
+        loop {
+            let success = unsafe { ts_query_cursor_next_match(self.ptr, &mut query_match) };
+            if !success {
+                return None;
+            }
+            let matched_nodes = unsafe {
+                slice::from_raw_parts(
+                    query_match.captures.cast(),
+                    query_match.capture_count as usize,
+                )
+            };
+            let satisfies_predicates = self
+                .query
+                .pattern_text_predicates(query_match.pattern_index)
+                .iter()
+                .all(|predicate| predicate.satsified(&mut self.input, matched_nodes, self.query));
+            if satisfies_predicates {
+                let res = QueryMatch {
+                    id: query_match.id,
+                    pattern: Pattern(query_match.pattern_index as u32),
+                    matched_nodes,
+                    query_cursor: unsafe { &mut *self.ptr },
+                    _tree: PhantomData,
+                };
+                return Some(res);
+            }
+        }
+    }
+
+    pub fn next_matched_node(&mut self) -> Option<(QueryMatch<'_, 'tree>, MatchedNodeIdx)> {
+        let mut query_match = TSQueryMatch {
+            id: 0,
+            pattern_index: 0,
+            capture_count: 0,
+            captures: ptr::null(),
+        };
+        let mut capture_idx = 0;
+        loop {
+            let success = unsafe {
+                ts_query_cursor_next_capture(self.ptr, &mut query_match, &mut capture_idx)
+            };
+            if !success {
+                return None;
+            }
+            let matched_nodes = unsafe {
+                slice::from_raw_parts(
+                    query_match.captures.cast(),
+                    query_match.capture_count as usize,
+                )
+            };
+            let satisfies_predicates = self
+                .query
+                .pattern_text_predicates(query_match.pattern_index)
+                .iter()
+                .all(|predicate| predicate.satsified(&mut self.input, matched_nodes, self.query));
+            if satisfies_predicates {
+                let res = QueryMatch {
+                    id: query_match.id,
+                    pattern: Pattern(query_match.pattern_index as u32),
+                    matched_nodes,
+                    query_cursor: unsafe { &mut *self.ptr },
+                    _tree: PhantomData,
+                };
+                return Some((res, capture_idx));
+            } else {
+                unsafe {
+                    ts_query_cursor_remove_match(self.ptr, query_match.id);
+                }
+            }
+        }
+    }
+
+    pub fn set_byte_range(&mut self, range: Range<usize>) {
+        unsafe {
+            ts_query_cursor_set_byte_range(self.ptr, range.start as u32, range.end as u32);
+        }
+    }
+
+    pub fn reuse(mut self) -> InactiveQueryCursor {
+        let ptr = replace(&mut self.ptr, ptr::null_mut());
+        InactiveQueryCursor {
+            ptr: unsafe { NonNull::new_unchecked(ptr) },
+        }
+    }
+}
+
+impl<I: TsInput> Drop for QueryCursor<'_, '_, I> {
+    fn drop(&mut self) {
+        // we allow moving the cursor data out so we need the null check here
+        // would be cleaner with a subtype but doesn't really matter at the end of the day
+        if let Some(ptr) = NonNull::new(self.ptr) {
+            unsafe { with_cache(|cache| cache.push(InactiveQueryCursor { ptr })) }
+        }
+    }
+}
+
+/// A query cursor that is not actively associated with a query
+pub struct InactiveQueryCursor {
+    ptr: NonNull<QueryCursorData>,
+}
+
+impl InactiveQueryCursor {
+    pub fn new() -> Self {
+        unsafe {
+            with_cache(|cache| {
+                cache.pop().unwrap_or_else(|| InactiveQueryCursor {
+                    ptr: NonNull::new_unchecked(ts_query_cursor_new()),
+                })
+            })
+        }
+    }
+
+    /// Return the maximum number of in-progress matches for this cursor.
+    #[doc(alias = "ts_query_cursor_match_limit")]
+    #[must_use]
+    pub fn match_limit(&self) -> u32 {
+        unsafe { ts_query_cursor_match_limit(self.ptr.as_ptr()) }
+    }
+
+    /// Set the maximum number of in-progress matches for this cursor.  The
+    /// limit must be > 0 and <= 65536.
+    #[doc(alias = "ts_query_cursor_set_match_limit")]
+    pub fn set_match_limit(&mut self, limit: u32) {
+        unsafe {
+            ts_query_cursor_set_match_limit(self.ptr.as_ptr(), limit);
+        }
+    }
+
+    /// Check if, on its last execution, this cursor exceeded its maximum number
+    /// of in-progress matches.
+    #[doc(alias = "ts_query_cursor_did_exceed_match_limit")]
+    #[must_use]
+    pub fn did_exceed_match_limit(&self) -> bool {
+        unsafe { ts_query_cursor_did_exceed_match_limit(self.ptr.as_ptr()) }
+    }
+
+    pub fn set_byte_range(&mut self, range: Range<usize>) {
+        unsafe {
+            ts_query_cursor_set_byte_range(self.ptr.as_ptr(), range.start as u32, range.end as u32);
+        }
+    }
+
+    pub fn execute_query<'a, 'tree, I: TsInput>(
+        self,
+        query: &'a Query,
+        node: &SyntaxTreeNode<'tree>,
+        input: I,
+    ) -> QueryCursor<'a, 'tree, I> {
+        let ptr = self.ptr.as_ptr();
+        unsafe { ts_query_cursor_exec(self.ptr.as_ptr(), query.raw.as_ref(), node.as_raw()) };
+        QueryCursor {
+            query,
+            ptr,
+            tree: PhantomData,
+            input,
+        }
+    }
+}
+
+impl Drop for InactiveQueryCursor {
+    fn drop(&mut self) {
+        unsafe { ts_query_cursor_delete(self.ptr.as_ptr()) }
+    }
+}
+
+pub type MatchedNodeIdx = u32;
+
+#[repr(C)]
+#[derive(Clone)]
+pub struct MatchedNode<'tree> {
+    pub syntax_node: SyntaxTreeNode<'tree>,
+    pub capture: Capture,
+}
+
+pub struct QueryMatch<'cursor, 'tree> {
+    id: u32,
+    pattern: Pattern,
+    matched_nodes: &'cursor [MatchedNode<'tree>],
+    query_cursor: &'cursor mut QueryCursorData,
+    _tree: PhantomData<&'tree super::SyntaxTree>,
+}
+
+impl<'tree> QueryMatch<'_, 'tree> {
+    pub fn matched_nodes(&self) -> impl Iterator<Item = &MatchedNode<'tree>> {
+        self.matched_nodes.iter()
+    }
+
+    pub fn nodes_for_capture(
+        &self,
+        capture: Capture,
+    ) -> impl Iterator<Item = &SyntaxTreeNode<'tree>> {
+        self.matched_nodes
+            .iter()
+            .filter(move |mat| mat.capture == capture)
+            .map(|mat| &mat.syntax_node)
+    }
+
+    pub fn matched_node(&self, i: MatchedNodeIdx) -> &MatchedNode {
+        &self.matched_nodes[i as usize]
+    }
+
+    #[must_use]
+    pub const fn id(&self) -> u32 {
+        self.id
+    }
+
+    #[must_use]
+    pub const fn pattern(&self) -> Pattern {
+        self.pattern
+    }
+
+    #[doc(alias = "ts_query_cursor_remove_match")]
+    /// removes this match from the cursor so that further captures
+    /// from its cursor so that future captures that belong to this match
+    /// are no longer returned by capture iterators
+    pub fn remove(self) {
+        unsafe {
+            ts_query_cursor_remove_match(self.query_cursor, self.id);
+        }
+    }
+}
+
+#[repr(C)]
+#[derive(Debug)]
+struct TSQueryCapture {
+    node: SyntaxTreeNodeRaw,
+    index: u32,
+}
+
+#[repr(C)]
+#[derive(Debug)]
+struct TSQueryMatch {
+    id: u32,
+    pattern_index: u16,
+    capture_count: u16,
+    captures: *const TSQueryCapture,
+}
+
+extern "C" {
+    /// Advance to the next capture of the currently running query.
+    /// If there is a capture, write its match to `*match` and its index within
+    /// the matche's capture list to `*capture_index`. Otherwise, return `false`.
+    fn ts_query_cursor_next_capture(
+        self_: *mut QueryCursorData,
+        match_: &mut TSQueryMatch,
+        capture_index: &mut u32,
+    ) -> bool;
+
+    /// Advance to the next match of the currently running query.
+    ///
+    /// If there is a match, write it to `*match` and return `true`.
+    /// Otherwise, return `false`.
+    pub fn ts_query_cursor_next_match(
+        self_: *mut QueryCursorData,
+        match_: &mut TSQueryMatch,
+    ) -> bool;
+    fn ts_query_cursor_remove_match(self_: *mut QueryCursorData, match_id: u32);
+    /// Delete a query cursor, freeing all of the memory that it used
+    pub fn ts_query_cursor_delete(self_: *mut QueryCursorData);
+    /// Create a new cursor for executing a given query.
+    /// The cursor stores the state that is needed to iteratively search
+    /// for matches. To use the query cursor, first call [`ts_query_cursor_exec`]
+    /// to start running a given query on a given syntax node. Then, there are
+    /// two options for consuming the results of the query:
+    /// 1. Repeatedly call [`ts_query_cursor_next_match`] to iterate over all of the
+    ///    *matches* in the order that they were found. Each match contains the
+    ///    index of the pattern that matched, and an array of captures. Because
+    ///    multiple patterns can match the same set of nodes, one match may contain
+    ///    captures that appear *before* some of the captures from a previous match.
+    /// 2. Repeatedly call [`ts_query_cursor_next_capture`] to iterate over all of the
+    ///    individual *captures* in the order that they appear. This is useful if
+    ///    don't care about which pattern matched, and just want a single ordered
+    ///    sequence of captures.
+    /// If you don't care about consuming all of the results, you can stop calling
+    /// [`ts_query_cursor_next_match`] or [`ts_query_cursor_next_capture`] at any point.
+    ///  You can then start executing another query on another node by calling
+    ///  [`ts_query_cursor_exec`] again."]
+    pub fn ts_query_cursor_new() -> *mut QueryCursorData;
+
+    /// Start running a given query on a given node.
+    pub fn ts_query_cursor_exec(
+        self_: *mut QueryCursorData,
+        query: &QueryData,
+        node: SyntaxTreeNodeRaw,
+    );
+    /// Manage the maximum number of in-progress matches allowed by this query
+    /// cursor.
+    ///
+    /// Query cursors have an optional maximum capacity for storing lists of
+    /// in-progress captures. If this capacity is exceeded, then the
+    /// earliest-starting match will silently be dropped to make room for further
+    /// matches. This maximum capacity is optional — by default, query cursors allow
+    /// any number of pending matches, dynamically allocating new space for them as
+    /// needed as the query is executed.
+    pub fn ts_query_cursor_did_exceed_match_limit(self_: *const QueryCursorData) -> bool;
+    pub fn ts_query_cursor_match_limit(self_: *const QueryCursorData) -> u32;
+    pub fn ts_query_cursor_set_match_limit(self_: *mut QueryCursorData, limit: u32);
+    /// Set the range of bytes or (row, column) positions in which the query
+    /// will be executed.
+    pub fn ts_query_cursor_set_byte_range(
+        self_: *mut QueryCursorData,
+        start_byte: u32,
+        end_byte: u32,
+    );
+
+}

helix-syntax/src/tree_sitter/query_match.rs

@@ -0,0 +1 @@
+pub struct QueryMatch {}

helix-syntax/src/tree_sitter/ropey.rs

@@ -0,0 +1,54 @@
+use std::ops;
+
+use regex_cursor::{Cursor, RopeyCursor};
+use ropey::RopeSlice;
+
+use crate::tree_sitter::{IntoTsInput, TsInput};
+
+pub struct RopeTsInput<'a> {
+    src: RopeSlice<'a>,
+    cursor: regex_cursor::RopeyCursor<'a>,
+}
+
+impl<'a> RopeTsInput<'a> {
+    pub fn new(src: RopeSlice<'a>) -> Self {
+        RopeTsInput {
+            src,
+            cursor: regex_cursor::RopeyCursor::new(src),
+        }
+    }
+}
+
+impl<'a> IntoTsInput for RopeSlice<'a> {
+    type TsInput = RopeTsInput<'a>;
+
+    fn into_ts_input(self) -> Self::TsInput {
+        RopeTsInput {
+            src: self,
+            cursor: RopeyCursor::new(self),
+        }
+    }
+}
+
+impl<'a> TsInput for RopeTsInput<'a> {
+    type Cursor = RopeyCursor<'a>;
+    fn cursor_at(&mut self, offset: usize) -> &mut RopeyCursor<'a> {
+        // this cursor is optimized for contigous reads which are by far the most common during parsing
+        // very far jumps (like injections at the other end of the document) are handelde
+        // by starting a new cursor (new chunks iterator)
+        if offset < self.cursor.offset() || self.cursor.offset() - offset > 4906 {
+            self.cursor = regex_cursor::RopeyCursor::at(self.src, offset);
+        } else {
+            while self.cursor.offset() + self.cursor.chunk().len() >= offset {
+                if !self.cursor.advance() {
+                    break;
+                }
+            }
+        }
+        &mut self.cursor
+    }
+
+    fn eq(&mut self, range1: ops::Range<usize>, range2: ops::Range<usize>) -> bool {
+        self.src.byte_slice(range1) == self.src.byte_slice(range2)
+    }
+}

helix-syntax/src/tree_sitter/syntax_tree.rs

@@ -0,0 +1,80 @@
+use std::fmt;
+use std::ptr::NonNull;
+
+use crate::tree_sitter::syntax_tree_node::{SyntaxTreeNode, SyntaxTreeNodeRaw};
+use crate::tree_sitter::Point;
+
+// opaque pointers
+pub(super) enum SyntaxTreeData {}
+
+pub struct SyntaxTree {
+    ptr: NonNull<SyntaxTreeData>,
+}
+
+impl SyntaxTree {
+    pub(super) unsafe fn from_raw(raw: NonNull<SyntaxTreeData>) -> SyntaxTree {
+        SyntaxTree { ptr: raw }
+    }
+
+    pub(super) fn as_raw(&self) -> NonNull<SyntaxTreeData> {
+        self.ptr
+    }
+
+    pub fn root_node(&self) -> SyntaxTreeNode<'_> {
+        unsafe { SyntaxTreeNode::from_raw(ts_tree_root_node(self.ptr)).unwrap() }
+    }
+
+    pub fn edit(&mut self, edit: &InputEdit) {
+        unsafe { ts_tree_edit(self.ptr, edit) }
+    }
+}
+
+impl fmt::Debug for SyntaxTree {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "{{Tree {:?}}}", self.root_node())
+    }
+}
+
+impl Drop for SyntaxTree {
+    fn drop(&mut self) {
+        unsafe { ts_tree_delete(self.ptr) }
+    }
+}
+
+impl Clone for SyntaxTree {
+    fn clone(&self) -> Self {
+        unsafe {
+            SyntaxTree {
+                ptr: ts_tree_copy(self.ptr),
+            }
+        }
+    }
+}
+
+#[repr(C)]
+#[derive(Debug, Copy, Clone)]
+pub struct InputEdit {
+    pub start_byte: u32,
+    pub old_end_byte: u32,
+    pub new_end_byte: u32,
+    pub start_point: Point,
+    pub old_end_point: Point,
+    pub new_end_point: Point,
+}
+
+extern "C" {
+    /// Create a shallow copy of the syntax tree. This is very fast. You need to
+    /// copy a syntax tree in order to use it on more than one thread at a time,
+    /// as syntax trees are not thread safe.
+    fn ts_tree_copy(self_: NonNull<SyntaxTreeData>) -> NonNull<SyntaxTreeData>;
+    /// Delete the syntax tree, freeing all of the memory that it used.
+    fn ts_tree_delete(self_: NonNull<SyntaxTreeData>);
+    /// Get the root node of the syntax tree.
+    fn ts_tree_root_node<'tree>(self_: NonNull<SyntaxTreeData>) -> SyntaxTreeNodeRaw;
+    /// Edit the syntax tree to keep it in sync with source code that has been
+    /// edited.
+    ///
+    /// You must describe the edit both in terms of byte offsets and in terms of
+    /// row/column coordinates.
+    fn ts_tree_edit(self_: NonNull<SyntaxTreeData>, edit: &InputEdit);
+}

helix-syntax/src/tree_sitter/syntax_tree_node.rs

@@ -0,0 +1,292 @@
+use std::ffi::c_void;
+use std::marker::PhantomData;
+use std::ops::Range;
+use std::ptr::NonNull;
+
+use crate::tree_sitter::syntax_tree::SyntaxTree;
+use crate::tree_sitter::Grammar;
+
+#[repr(C)]
+#[derive(Debug, Clone, Copy)]
+pub(super) struct SyntaxTreeNodeRaw {
+    context: [u32; 4],
+    id: *const c_void,
+    tree: *const c_void,
+}
+
+impl From<SyntaxTreeNode<'_>> for SyntaxTreeNodeRaw {
+    fn from(node: SyntaxTreeNode) -> SyntaxTreeNodeRaw {
+        SyntaxTreeNodeRaw {
+            context: node.context,
+            id: node.id.as_ptr(),
+            tree: node.tree.as_ptr(),
+        }
+    }
+}
+
+#[derive(Debug, Clone)]
+#[repr(C)]
+pub struct SyntaxTreeNode<'tree> {
+    context: [u32; 4],
+    id: NonNull<c_void>,
+    tree: NonNull<c_void>,
+    _phantom: PhantomData<&'tree SyntaxTree>,
+}
+
+impl<'tree> SyntaxTreeNode<'tree> {
+    #[inline]
+    pub(super) unsafe fn from_raw(raw: SyntaxTreeNodeRaw) -> Option<Self> {
+        Some(SyntaxTreeNode {
+            context: raw.context,
+            id: NonNull::new(raw.id as *mut _)?,
+            tree: unsafe { NonNull::new_unchecked(raw.tree as *mut _) },
+            _phantom: PhantomData,
+        })
+    }
+
+    #[inline]
+    pub(crate) fn as_raw(&self) -> SyntaxTreeNodeRaw {
+        SyntaxTreeNodeRaw {
+            context: self.context,
+            id: self.id.as_ptr(),
+            tree: self.tree.as_ptr(),
+        }
+    }
+
+    /// Get this node's type as a numerical id.
+    #[inline]
+    pub fn kind_id(&self) -> u16 {
+        unsafe { ts_node_symbol(self.as_raw()) }
+    }
+
+    /// Get the [`Language`] that was used to parse this node's syntax tree.
+    #[inline]
+    pub fn grammar(&self) -> Grammar {
+        unsafe { ts_node_language(self.as_raw()) }
+    }
+
+    /// Check if this node is *named*.
+    ///
+    /// Named nodes correspond to named rules in the grammar, whereas
+    /// *anonymous* nodes correspond to string literals in the grammar.
+    #[inline]
+    pub fn is_named(&self) -> bool {
+        unsafe { ts_node_is_named(self.as_raw()) }
+    }
+
+    /// Check if this node is *missing*.
+    ///
+    /// Missing nodes are inserted by the parser in order to recover from
+    /// certain kinds of syntax errors.
+    #[inline]
+    pub fn is_missing(&self) -> bool {
+        unsafe { ts_node_is_missing(self.as_raw()) }
+    }
+    /// Get the byte offsets where this node starts.
+    #[inline]
+    pub fn start_byte(&self) -> usize {
+        unsafe { ts_node_start_byte(self.as_raw()) as usize }
+    }
+
+    /// Get the byte offsets where this node end.
+    #[inline]
+    pub fn end_byte(&self) -> usize {
+        unsafe { ts_node_end_byte(self.as_raw()) as usize }
+    }
+
+    /// Get the byte range of source code that this node represents.
+    // TODO: use helix_stdx::Range once available
+    #[inline]
+    pub fn byte_range(&self) -> Range<usize> {
+        self.start_byte()..self.end_byte()
+    }
+
+    /// Get the node's child at the given index, where zero represents the first
+    /// child.
+    ///
+    /// This method is fairly fast, but its cost is technically log(i), so if
+    /// you might be iterating over a long list of children, you should use
+    /// [`SyntaxTreeNode::children`] instead.
+    #[inline]
+    pub fn child(&self, i: usize) -> Option<SyntaxTreeNode<'tree>> {
+        unsafe { SyntaxTreeNode::from_raw(ts_node_child(self.as_raw(), i as u32)) }
+    }
+
+    /// Get this node's number of children.
+    #[inline]
+    pub fn child_count(&self) -> usize {
+        unsafe { ts_node_child_count(self.as_raw()) as usize }
+    }
+
+    /// Get this node's *named* child at the given index.
+    ///
+    /// See also [`SyntaxTreeNode::is_named`].
+    /// This method is fairly fast, but its cost is technically log(i), so if
+    /// you might be iterating over a long list of children, you should use
+    /// [`SyntaxTreeNode::named_children`] instead.
+    #[inline]
+    pub fn named_child(&self, i: usize) -> Option<SyntaxTreeNode<'tree>> {
+        unsafe { SyntaxTreeNode::from_raw(ts_node_named_child(self.as_raw(), i as u32)) }
+    }
+
+    /// Get this node's number of *named* children.
+    ///
+    /// See also [`SyntaxTreeNode::is_named`].
+    #[inline]
+    pub fn named_child_count(&self) -> usize {
+        unsafe { ts_node_named_child_count(self.as_raw()) as usize }
+    }
+
+    #[inline]
+    unsafe fn map(
+        &self,
+        f: unsafe extern "C" fn(SyntaxTreeNodeRaw) -> SyntaxTreeNodeRaw,
+    ) -> Option<SyntaxTreeNode<'tree>> {
+        SyntaxTreeNode::from_raw(f(self.as_raw()))
+    }
+
+    /// Get this node's immediate parent.
+    #[inline]
+    pub fn parent(&self) -> Option<Self> {
+        unsafe { self.map(ts_node_parent) }
+    }
+
+    /// Get this node's next sibling.
+    #[inline]
+    pub fn next_sibling(&self) -> Option<Self> {
+        unsafe { self.map(ts_node_next_sibling) }
+    }
+
+    /// Get this node's previous sibling.
+    #[inline]
+    pub fn prev_sibling(&self) -> Option<Self> {
+        unsafe { self.map(ts_node_prev_sibling) }
+    }
+
+    /// Get this node's next named sibling.
+    #[inline]
+    pub fn next_named_sibling(&self) -> Option<Self> {
+        unsafe { self.map(ts_node_next_named_sibling) }
+    }
+
+    /// Get this node's previous named sibling.
+    #[inline]
+    pub fn prev_named_sibling(&self) -> Option<Self> {
+        unsafe { self.map(ts_node_prev_named_sibling) }
+    }
+
+    /// Get the smallest node within this node that spans the given range.
+    #[inline]
+    pub fn descendant_for_byte_range(&self, start: usize, end: usize) -> Option<Self> {
+        unsafe {
+            Self::from_raw(ts_node_descendant_for_byte_range(
+                self.as_raw(),
+                start as u32,
+                end as u32,
+            ))
+        }
+    }
+
+    /// Get the smallest named node within this node that spans the given range.
+    #[inline]
+    pub fn named_descendant_for_byte_range(&self, start: usize, end: usize) -> Option<Self> {
+        unsafe {
+            Self::from_raw(ts_node_named_descendant_for_byte_range(
+                self.as_raw(),
+                start as u32,
+                end as u32,
+            ))
+        }
+    }
+    // /// Iterate over this node's children.
+    // ///
+    // /// A [`TreeCursor`] is used to retrieve the children efficiently. Obtain
+    // /// a [`TreeCursor`] by calling [`Tree::walk`] or [`SyntaxTreeNode::walk`]. To avoid
+    // /// unnecessary allocations, you should reuse the same cursor for
+    // /// subsequent calls to this method.
+    // ///
+    // /// If you're walking the tree recursively, you may want to use the
+    // /// [`TreeCursor`] APIs directly instead.
+    // pub fn children<'cursor>(
+    //     &self,
+    //     cursor: &'cursor mut TreeCursor<'tree>,
+    // ) -> impl ExactSizeIterator<Item = SyntaxTreeNode<'tree>> + 'cursor {
+    //     cursor.reset(self.to_raw());
+    //     cursor.goto_first_child();
+    //     (0..self.child_count()).map(move |_| {
+    //         let result = cursor.node();
+    //         cursor.goto_next_sibling();
+    //         result
+    //     })
+    // }
+}
+
+unsafe impl Send for SyntaxTreeNode<'_> {}
+unsafe impl Sync for SyntaxTreeNode<'_> {}
+
+extern "C" {
+    /// Get the node's type as a numerical id.
+    fn ts_node_symbol(node: SyntaxTreeNodeRaw) -> u16;
+
+    /// Get the node's language.
+    fn ts_node_language(node: SyntaxTreeNodeRaw) -> Grammar;
+
+    /// Check if the node is *named*. Named nodes correspond to named rules in
+    /// the grammar, whereas *anonymous* nodes correspond to string literals in
+    /// the grammar
+    fn ts_node_is_named(node: SyntaxTreeNodeRaw) -> bool;
+
+    /// Check if the node is *missing*. Missing nodes are inserted by the parser
+    /// in order to recover from certain kinds of syntax errors
+    fn ts_node_is_missing(node: SyntaxTreeNodeRaw) -> bool;
+
+    /// Get the node's immediate parent
+    fn ts_node_parent(node: SyntaxTreeNodeRaw) -> SyntaxTreeNodeRaw;
+
+    /// Get the node's child at the given index, where zero represents the first
+    /// child
+    fn ts_node_child(node: SyntaxTreeNodeRaw, child_index: u32) -> SyntaxTreeNodeRaw;
+
+    /// Get the node's number of children
+    fn ts_node_child_count(node: SyntaxTreeNodeRaw) -> u32;
+
+    /// Get the node's *named* child at the given index. See also
+    /// [`ts_node_is_named`]
+    fn ts_node_named_child(node: SyntaxTreeNodeRaw, child_index: u32) -> SyntaxTreeNodeRaw;
+
+    /// Get the node's number of *named* children. See also [`ts_node_is_named`]
+    fn ts_node_named_child_count(node: SyntaxTreeNodeRaw) -> u32;
+
+    /// Get the node's next sibling
+    fn ts_node_next_sibling(node: SyntaxTreeNodeRaw) -> SyntaxTreeNodeRaw;
+
+    fn ts_node_prev_sibling(node: SyntaxTreeNodeRaw) -> SyntaxTreeNodeRaw;
+
+    /// Get the node's next *named* sibling
+    fn ts_node_next_named_sibling(node: SyntaxTreeNodeRaw) -> SyntaxTreeNodeRaw;
+
+    fn ts_node_prev_named_sibling(node: SyntaxTreeNodeRaw) -> SyntaxTreeNodeRaw;
+
+    /// Get the smallest node within this node that spans the given range of
+    /// bytes or (row, column) positions
+    fn ts_node_descendant_for_byte_range(
+        node: SyntaxTreeNodeRaw,
+
+        start: u32,
+        end: u32,
+    ) -> SyntaxTreeNodeRaw;
+
+    /// Get the smallest named node within this node that spans the given range
+    /// of bytes or (row, column) positions
+    fn ts_node_named_descendant_for_byte_range(
+        node: SyntaxTreeNodeRaw,
+        start: u32,
+        end: u32,
+    ) -> SyntaxTreeNodeRaw;
+
+    /// Get the node's start byte.
+    fn ts_node_start_byte(self_: SyntaxTreeNodeRaw) -> u32;
+
+    /// Get the node's end byte.
+    fn ts_node_end_byte(node: SyntaxTreeNodeRaw) -> u32;
+}

helix-term/src/ui/markdown.rs

@@ -54,7 +54,14 @@ pub fn highlighted_code_block<'a>(
             language.into(),
         ))
         .and_then(|config| config.highlight_config(theme.scopes()))
-        .and_then(|config| Syntax::new(ropeslice, config, Arc::clone(&config_loader)));
+        .and_then(|config| {
+            Syntax::new(ropeslice, config, |injection| {
+                config_loader
+                    .load()
+                    .language_configuration_for_injection_string(injection)
+                    .and_then(|config| config.get_highlight_config())
+            })
+        });
 
     let syntax = match syntax {
         Some(s) => s,

helix-term/src/ui/picker/handlers.rs

@@ -82,7 +82,12 @@ impl<T: 'static + Send + Sync, D: 'static + Send + Sync> AsyncHook
                 let Some(syntax) = language_config
                     .highlight_config(&loader.load().scopes())
                     .and_then(|highlight_config| {
-                        helix_core::Syntax::new(text.slice(..), highlight_config, loader)
+                        helix_core::Syntax::new(text.slice(..), highlight_config, |injection| {
+                            loader
+                                .load()
+                                .language_configuration_for_injection_string(injection)
+                                .and_then(|config| config.get_highlight_config())
+                        })
                     })
                 else {
                     log::info!("highlighting picker item failed");

helix-view/src/document.rs

@@ -7,7 +7,7 @@ use helix_core::auto_pairs::AutoPairs;
 use helix_core::chars::char_is_word;
 use helix_core::doc_formatter::TextFormat;
 use helix_core::encoding::Encoding;
-use helix_core::syntax::{Highlight, LanguageServerFeature};
+use helix_core::syntax::{generate_edits, Highlight, LanguageServerFeature};
 use helix_core::text_annotations::{InlineAnnotation, Overlay};
 use helix_lsp::util::lsp_pos_to_pos;
 use helix_stdx::faccess::{copy_metadata, readonly};
@@ -156,6 +156,7 @@ pub struct Document {
     pub syntax: Option<Syntax>,
     /// Corresponding language scope name. Usually `source.<lang>`.
     pub language: Option<Arc<LanguageConfiguration>>,
+    loader: Option<Arc<ArcSwap<helix_core::syntax::Loader>>>,
 
     /// Pending changes since last history commit.
     changes: ChangeSet,
@@ -678,6 +679,7 @@ impl Document {
             focused_at: std::time::Instant::now(),
             readonly: false,
             jump_labels: HashMap::new(),
+            loader: None,
         }
     }
 
@@ -1131,9 +1133,15 @@ impl Document {
             if let Some(highlight_config) =
                 language_config.highlight_config(&(*loader).load().scopes())
             {
-                self.syntax = Syntax::new(self.text.slice(..), highlight_config, loader);
+                let loader_ = loader.load_full();
+                self.syntax = Syntax::new(self.text.slice(..), highlight_config, |injection| {
+                    loader_
+                        .language_configuration_for_injection_string(injection)
+                        .and_then(|config| config.get_highlight_config())
+                });
             }
 
+            self.loader = Some(loader);
             self.language = Some(language_config);
         } else {
             self.syntax = None;
@@ -1275,11 +1283,16 @@ impl Document {
 
             // update tree-sitter syntax tree
             if let Some(syntax) = &mut self.syntax {
+                let loader = self.loader.as_ref().unwrap().load_full();
                 // TODO: no unwrap
                 let res = syntax.update(
-                    old_doc.slice(..),
                     self.text.slice(..),
-                    transaction.changes(),
+                    generate_edits(old_doc.slice(..), transaction.changes()),
+                    |injection| {
+                        loader
+                            .language_configuration_for_injection_string(injection)
+                            .and_then(|config| config.get_highlight_config())
+                    },
                 );
                 if res.is_err() {
                     log::error!("TS parser failed, disabling TS for the current buffer: {res:?}");

vendor/tree-sitter/LICENSE

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2018-2024 Max Brunsfeld
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

vendor/tree-sitter/src/alloc.c

@@ -0,0 +1,48 @@
+#include "alloc.h"
+#include "tree_sitter/api.h"
+#include <stdlib.h>
+
+static void *ts_malloc_default(size_t size) {
+  void *result = malloc(size);
+  if (size > 0 && !result) {
+    fprintf(stderr, "tree-sitter failed to allocate %zu bytes", size);
+    abort();
+  }
+  return result;
+}
+
+static void *ts_calloc_default(size_t count, size_t size) {
+  void *result = calloc(count, size);
+  if (count > 0 && !result) {
+    fprintf(stderr, "tree-sitter failed to allocate %zu bytes", count * size);
+    abort();
+  }
+  return result;
+}
+
+static void *ts_realloc_default(void *buffer, size_t size) {
+  void *result = realloc(buffer, size);
+  if (size > 0 && !result) {
+    fprintf(stderr, "tree-sitter failed to reallocate %zu bytes", size);
+    abort();
+  }
+  return result;
+}
+
+// Allow clients to override allocation functions dynamically
+TS_PUBLIC void *(*ts_current_malloc)(size_t) = ts_malloc_default;
+TS_PUBLIC void *(*ts_current_calloc)(size_t, size_t) = ts_calloc_default;
+TS_PUBLIC void *(*ts_current_realloc)(void *, size_t) = ts_realloc_default;
+TS_PUBLIC void (*ts_current_free)(void *) = free;
+
+void ts_set_allocator(
+  void *(*new_malloc)(size_t size),
+  void *(*new_calloc)(size_t count, size_t size),
+  void *(*new_realloc)(void *ptr, size_t size),
+  void (*new_free)(void *ptr)
+) {
+  ts_current_malloc = new_malloc ? new_malloc : ts_malloc_default;
+  ts_current_calloc = new_calloc ? new_calloc : ts_calloc_default;
+  ts_current_realloc = new_realloc ? new_realloc : ts_realloc_default;
+  ts_current_free = new_free ? new_free : free;
+}

vendor/tree-sitter/src/alloc.h

@@ -0,0 +1,41 @@
+#ifndef TREE_SITTER_ALLOC_H_
+#define TREE_SITTER_ALLOC_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#if defined(TREE_SITTER_HIDDEN_SYMBOLS) || defined(_WIN32)
+#define TS_PUBLIC
+#else
+#define TS_PUBLIC __attribute__((visibility("default")))
+#endif
+
+TS_PUBLIC extern void *(*ts_current_malloc)(size_t);
+TS_PUBLIC extern void *(*ts_current_calloc)(size_t, size_t);
+TS_PUBLIC extern void *(*ts_current_realloc)(void *, size_t);
+TS_PUBLIC extern void (*ts_current_free)(void *);
+
+// Allow clients to override allocation functions
+#ifndef ts_malloc
+#define ts_malloc  ts_current_malloc
+#endif
+#ifndef ts_calloc
+#define ts_calloc  ts_current_calloc
+#endif
+#ifndef ts_realloc
+#define ts_realloc ts_current_realloc
+#endif
+#ifndef ts_free
+#define ts_free    ts_current_free
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif // TREE_SITTER_ALLOC_H_

vendor/tree-sitter/src/array.h

@@ -0,0 +1,290 @@
+#ifndef TREE_SITTER_ARRAY_H_
+#define TREE_SITTER_ARRAY_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "./alloc.h"
+
+#include <assert.h>
+#include <stdbool.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+#ifdef _MSC_VER
+#pragma warning(disable : 4101)
+#elif defined(__GNUC__) || defined(__clang__)
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wunused-variable"
+#endif
+
+#define Array(T)       \
+  struct {             \
+    T *contents;       \
+    uint32_t size;     \
+    uint32_t capacity; \
+  }
+
+/// Initialize an array.
+#define array_init(self) \
+  ((self)->size = 0, (self)->capacity = 0, (self)->contents = NULL)
+
+/// Create an empty array.
+#define array_new() \
+  { NULL, 0, 0 }
+
+/// Get a pointer to the element at a given `index` in the array.
+#define array_get(self, _index) \
+  (assert((uint32_t)(_index) < (self)->size), &(self)->contents[_index])
+
+/// Get a pointer to the first element in the array.
+#define array_front(self) array_get(self, 0)
+
+/// Get a pointer to the last element in the array.
+#define array_back(self) array_get(self, (self)->size - 1)
+
+/// Clear the array, setting its size to zero. Note that this does not free any
+/// memory allocated for the array's contents.
+#define array_clear(self) ((self)->size = 0)
+
+/// Reserve `new_capacity` elements of space in the array. If `new_capacity` is
+/// less than the array's current capacity, this function has no effect.
+#define array_reserve(self, new_capacity) \
+  _array__reserve((Array *)(self), array_elem_size(self), new_capacity)
+
+/// Free any memory allocated for this array. Note that this does not free any
+/// memory allocated for the array's contents.
+#define array_delete(self) _array__delete((Array *)(self))
+
+/// Push a new `element` onto the end of the array.
+#define array_push(self, element)                            \
+  (_array__grow((Array *)(self), 1, array_elem_size(self)), \
+   (self)->contents[(self)->size++] = (element))
+
+/// Increase the array's size by `count` elements.
+/// New elements are zero-initialized.
+#define array_grow_by(self, count) \
+  do { \
+    if ((count) == 0) break; \
+    _array__grow((Array *)(self), count, array_elem_size(self)); \
+    memset((self)->contents + (self)->size, 0, (count) * array_elem_size(self)); \
+    (self)->size += (count); \
+  } while (0)
+
+/// Append all elements from one array to the end of another.
+#define array_push_all(self, other)                                       \
+  array_extend((self), (other)->size, (other)->contents)
+
+/// Append `count` elements to the end of the array, reading their values from the
+/// `contents` pointer.
+#define array_extend(self, count, contents)                    \
+  _array__splice(                                               \
+    (Array *)(self), array_elem_size(self), (self)->size, \
+    0, count,  contents                                        \
+  )
+
+/// Remove `old_count` elements from the array starting at the given `index`. At
+/// the same index, insert `new_count` new elements, reading their values from the
+/// `new_contents` pointer.
+#define array_splice(self, _index, old_count, new_count, new_contents)  \
+  _array__splice(                                                       \
+    (Array *)(self), array_elem_size(self), _index,                \
+    old_count, new_count, new_contents                                 \
+  )
+
+/// Insert one `element` into the array at the given `index`.
+#define array_insert(self, _index, element) \
+  _array__splice((Array *)(self), array_elem_size(self), _index, 0, 1, &(element))
+
+/// Remove one element from the array at the given `index`.
+#define array_erase(self, _index) \
+  _array__erase((Array *)(self), array_elem_size(self), _index)
+
+/// Pop the last element off the array, returning the element by value.
+#define array_pop(self) ((self)->contents[--(self)->size])
+
+/// Assign the contents of one array to another, reallocating if necessary.
+#define array_assign(self, other) \
+  _array__assign((Array *)(self), (const Array *)(other), array_elem_size(self))
+
+/// Swap one array with another
+#define array_swap(self, other) \
+  _array__swap((Array *)(self), (Array *)(other))
+
+/// Get the size of the array contents
+#define array_elem_size(self) (sizeof *(self)->contents)
+
+/// Search a sorted array for a given `needle` value, using the given `compare`
+/// callback to determine the order.
+///
+/// If an existing element is found to be equal to `needle`, then the `index`
+/// out-parameter is set to the existing value's index, and the `exists`
+/// out-parameter is set to true. Otherwise, `index` is set to an index where
+/// `needle` should be inserted in order to preserve the sorting, and `exists`
+/// is set to false.
+#define array_search_sorted_with(self, compare, needle, _index, _exists) \
+  _array__search_sorted(self, 0, compare, , needle, _index, _exists)
+
+/// Search a sorted array for a given `needle` value, using integer comparisons
+/// of a given struct field (specified with a leading dot) to determine the order.
+///
+/// See also `array_search_sorted_with`.
+#define array_search_sorted_by(self, field, needle, _index, _exists) \
+  _array__search_sorted(self, 0, _compare_int, field, needle, _index, _exists)
+
+/// Insert a given `value` into a sorted array, using the given `compare`
+/// callback to determine the order.
+#define array_insert_sorted_with(self, compare, value) \
+  do { \
+    unsigned _index, _exists; \
+    array_search_sorted_with(self, compare, &(value), &_index, &_exists); \
+    if (!_exists) array_insert(self, _index, value); \
+  } while (0)
+
+/// Insert a given `value` into a sorted array, using integer comparisons of
+/// a given struct field (specified with a leading dot) to determine the order.
+///
+/// See also `array_search_sorted_by`.
+#define array_insert_sorted_by(self, field, value) \
+  do { \
+    unsigned _index, _exists; \
+    array_search_sorted_by(self, field, (value) field, &_index, &_exists); \
+    if (!_exists) array_insert(self, _index, value); \
+  } while (0)
+
+// Private
+
+typedef Array(void) Array;
+
+/// This is not what you're looking for, see `array_delete`.
+static inline void _array__delete(Array *self) {
+  if (self->contents) {
+    ts_free(self->contents);
+    self->contents = NULL;
+    self->size = 0;
+    self->capacity = 0;
+  }
+}
+
+/// This is not what you're looking for, see `array_erase`.
+static inline void _array__erase(Array *self, size_t element_size,
+                                uint32_t index) {
+  assert(index < self->size);
+  char *contents = (char *)self->contents;
+  memmove(contents + index * element_size, contents + (index + 1) * element_size,
+          (self->size - index - 1) * element_size);
+  self->size--;
+}
+
+/// This is not what you're looking for, see `array_reserve`.
+static inline void _array__reserve(Array *self, size_t element_size, uint32_t new_capacity) {
+  if (new_capacity > self->capacity) {
+    if (self->contents) {
+      self->contents = ts_realloc(self->contents, new_capacity * element_size);
+    } else {
+      self->contents = ts_malloc(new_capacity * element_size);
+    }
+    self->capacity = new_capacity;
+  }
+}
+
+/// This is not what you're looking for, see `array_assign`.
+static inline void _array__assign(Array *self, const Array *other, size_t element_size) {
+  _array__reserve(self, element_size, other->size);
+  self->size = other->size;
+  memcpy(self->contents, other->contents, self->size * element_size);
+}
+
+/// This is not what you're looking for, see `array_swap`.
+static inline void _array__swap(Array *self, Array *other) {
+  Array swap = *other;
+  *other = *self;
+  *self = swap;
+}
+
+/// This is not what you're looking for, see `array_push` or `array_grow_by`.
+static inline void _array__grow(Array *self, uint32_t count, size_t element_size) {
+  uint32_t new_size = self->size + count;
+  if (new_size > self->capacity) {
+    uint32_t new_capacity = self->capacity * 2;
+    if (new_capacity < 8) new_capacity = 8;
+    if (new_capacity < new_size) new_capacity = new_size;
+    _array__reserve(self, element_size, new_capacity);
+  }
+}
+
+/// This is not what you're looking for, see `array_splice`.
+static inline void _array__splice(Array *self, size_t element_size,
+                                 uint32_t index, uint32_t old_count,
+                                 uint32_t new_count, const void *elements) {
+  uint32_t new_size = self->size + new_count - old_count;
+  uint32_t old_end = index + old_count;
+  uint32_t new_end = index + new_count;
+  assert(old_end <= self->size);
+
+  _array__reserve(self, element_size, new_size);
+
+  char *contents = (char *)self->contents;
+  if (self->size > old_end) {
+    memmove(
+      contents + new_end * element_size,
+      contents + old_end * element_size,
+      (self->size - old_end) * element_size
+    );
+  }
+  if (new_count > 0) {
+    if (elements) {
+      memcpy(
+        (contents + index * element_size),
+        elements,
+        new_count * element_size
+      );
+    } else {
+      memset(
+        (contents + index * element_size),
+        0,
+        new_count * element_size
+      );
+    }
+  }
+  self->size += new_count - old_count;
+}
+
+/// A binary search routine, based on Rust's `std::slice::binary_search_by`.
+/// This is not what you're looking for, see `array_search_sorted_with` or `array_search_sorted_by`.
+#define _array__search_sorted(self, start, compare, suffix, needle, _index, _exists) \
+  do { \
+    *(_index) = start; \
+    *(_exists) = false; \
+    uint32_t size = (self)->size - *(_index); \
+    if (size == 0) break; \
+    int comparison; \
+    while (size > 1) { \
+      uint32_t half_size = size / 2; \
+      uint32_t mid_index = *(_index) + half_size; \
+      comparison = compare(&((self)->contents[mid_index] suffix), (needle)); \
+      if (comparison <= 0) *(_index) = mid_index; \
+      size -= half_size; \
+    } \
+    comparison = compare(&((self)->contents[*(_index)] suffix), (needle)); \
+    if (comparison == 0) *(_exists) = true; \
+    else if (comparison < 0) *(_index) += 1; \
+  } while (0)
+
+/// Helper macro for the `_sorted_by` routines below. This takes the left (existing)
+/// parameter by reference in order to work with the generic sorting function above.
+#define _compare_int(a, b) ((int)*(a) - (int)(b))
+
+#ifdef _MSC_VER
+#pragma warning(default : 4101)
+#elif defined(__GNUC__) || defined(__clang__)
+#pragma GCC diagnostic pop
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TREE_SITTER_ARRAY_H_

vendor/tree-sitter/src/atomic.h

@@ -0,0 +1,68 @@
+#ifndef TREE_SITTER_ATOMIC_H_
+#define TREE_SITTER_ATOMIC_H_
+
+#include <stddef.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+#ifdef __TINYC__
+
+static inline size_t atomic_load(const volatile size_t *p) {
+  return *p;
+}
+
+static inline uint32_t atomic_inc(volatile uint32_t *p) {
+  *p += 1;
+  return *p;
+}
+
+static inline uint32_t atomic_dec(volatile uint32_t *p) {
+  *p-= 1;
+  return *p;
+}
+
+#elif defined(_WIN32)
+
+#include <windows.h>
+
+static inline size_t atomic_load(const volatile size_t *p) {
+  return *p;
+}
+
+static inline uint32_t atomic_inc(volatile uint32_t *p) {
+  return InterlockedIncrement((long volatile *)p);
+}
+
+static inline uint32_t atomic_dec(volatile uint32_t *p) {
+  return InterlockedDecrement((long volatile *)p);
+}
+
+#else
+
+static inline size_t atomic_load(const volatile size_t *p) {
+#ifdef __ATOMIC_RELAXED
+  return __atomic_load_n(p, __ATOMIC_RELAXED);
+#else
+  return __sync_fetch_and_add((volatile size_t *)p, 0);
+#endif
+}
+
+static inline uint32_t atomic_inc(volatile uint32_t *p) {
+  #ifdef __ATOMIC_RELAXED
+    return __atomic_add_fetch(p, 1U, __ATOMIC_SEQ_CST);
+  #else
+    return __sync_add_and_fetch(p, 1U);
+  #endif
+}
+
+static inline uint32_t atomic_dec(volatile uint32_t *p) {
+  #ifdef __ATOMIC_RELAXED
+    return __atomic_sub_fetch(p, 1U, __ATOMIC_SEQ_CST);
+  #else
+    return __sync_sub_and_fetch(p, 1U);
+  #endif
+}
+
+#endif
+
+#endif  // TREE_SITTER_ATOMIC_H_

vendor/tree-sitter/src/clock.h

@@ -0,0 +1,146 @@
+#ifndef TREE_SITTER_CLOCK_H_
+#define TREE_SITTER_CLOCK_H_
+
+#include <stdbool.h>
+#include <stdint.h>
+
+typedef uint64_t TSDuration;
+
+#ifdef _WIN32
+
+// Windows:
+// * Represent a time as a performance counter value.
+// * Represent a duration as a number of performance counter ticks.
+
+#include <windows.h>
+typedef uint64_t TSClock;
+
+static inline TSDuration duration_from_micros(uint64_t micros) {
+  LARGE_INTEGER frequency;
+  QueryPerformanceFrequency(&frequency);
+  return micros * (uint64_t)frequency.QuadPart / 1000000;
+}
+
+static inline uint64_t duration_to_micros(TSDuration self) {
+  LARGE_INTEGER frequency;
+  QueryPerformanceFrequency(&frequency);
+  return self * 1000000 / (uint64_t)frequency.QuadPart;
+}
+
+static inline TSClock clock_null(void) {
+  return 0;
+}
+
+static inline TSClock clock_now(void) {
+  LARGE_INTEGER result;
+  QueryPerformanceCounter(&result);
+  return (uint64_t)result.QuadPart;
+}
+
+static inline TSClock clock_after(TSClock base, TSDuration duration) {
+  return base + duration;
+}
+
+static inline bool clock_is_null(TSClock self) {
+  return !self;
+}
+
+static inline bool clock_is_gt(TSClock self, TSClock other) {
+  return self > other;
+}
+
+#elif defined(CLOCK_MONOTONIC) && !defined(__APPLE__)
+
+// POSIX with monotonic clock support (Linux)
+// * Represent a time as a monotonic (seconds, nanoseconds) pair.
+// * Represent a duration as a number of microseconds.
+//
+// On these platforms, parse timeouts will correspond accurately to
+// real time, regardless of what other processes are running.
+
+#include <time.h>
+typedef struct timespec TSClock;
+
+static inline TSDuration duration_from_micros(uint64_t micros) {
+  return micros;
+}
+
+static inline uint64_t duration_to_micros(TSDuration self) {
+  return self;
+}
+
+static inline TSClock clock_now(void) {
+  TSClock result;
+  clock_gettime(CLOCK_MONOTONIC, &result);
+  return result;
+}
+
+static inline TSClock clock_null(void) {
+  return (TSClock) {0, 0};
+}
+
+static inline TSClock clock_after(TSClock base, TSDuration duration) {
+  TSClock result = base;
+  result.tv_sec += duration / 1000000;
+  result.tv_nsec += (duration % 1000000) * 1000;
+  if (result.tv_nsec >= 1000000000) {
+    result.tv_nsec -= 1000000000;
+    ++(result.tv_sec);
+  }
+  return result;
+}
+
+static inline bool clock_is_null(TSClock self) {
+  return !self.tv_sec;
+}
+
+static inline bool clock_is_gt(TSClock self, TSClock other) {
+  if (self.tv_sec > other.tv_sec) return true;
+  if (self.tv_sec < other.tv_sec) return false;
+  return self.tv_nsec > other.tv_nsec;
+}
+
+#else
+
+// macOS or POSIX without monotonic clock support
+// * Represent a time as a process clock value.
+// * Represent a duration as a number of process clock ticks.
+//
+// On these platforms, parse timeouts may be affected by other processes,
+// which is not ideal, but is better than using a non-monotonic time API
+// like `gettimeofday`.
+
+#include <time.h>
+typedef uint64_t TSClock;
+
+static inline TSDuration duration_from_micros(uint64_t micros) {
+  return micros * (uint64_t)CLOCKS_PER_SEC / 1000000;
+}
+
+static inline uint64_t duration_to_micros(TSDuration self) {
+  return self * 1000000 / (uint64_t)CLOCKS_PER_SEC;
+}
+
+static inline TSClock clock_null(void) {
+  return 0;
+}
+
+static inline TSClock clock_now(void) {
+  return (uint64_t)clock();
+}
+
+static inline TSClock clock_after(TSClock base, TSDuration duration) {
+  return base + duration;
+}
+
+static inline bool clock_is_null(TSClock self) {
+  return !self;
+}
+
+static inline bool clock_is_gt(TSClock self, TSClock other) {
+  return self > other;
+}
+
+#endif
+
+#endif  // TREE_SITTER_CLOCK_H_

vendor/tree-sitter/src/error_costs.h

@@ -0,0 +1,11 @@
+#ifndef TREE_SITTER_ERROR_COSTS_H_
+#define TREE_SITTER_ERROR_COSTS_H_
+
+#define ERROR_STATE 0
+#define ERROR_COST_PER_RECOVERY 500
+#define ERROR_COST_PER_MISSING_TREE 110
+#define ERROR_COST_PER_SKIPPED_TREE 100
+#define ERROR_COST_PER_SKIPPED_LINE 30
+#define ERROR_COST_PER_SKIPPED_CHAR 1
+
+#endif

vendor/tree-sitter/src/get_changed_ranges.c

@@ -0,0 +1,501 @@
+#include "./get_changed_ranges.h"
+#include "./subtree.h"
+#include "./language.h"
+#include "./error_costs.h"
+#include "./tree_cursor.h"
+#include <assert.h>
+
+// #define DEBUG_GET_CHANGED_RANGES
+
+static void ts_range_array_add(
+  TSRangeArray *self,
+  Length start,
+  Length end
+) {
+  if (self->size > 0) {
+    TSRange *last_range = array_back(self);
+    if (start.bytes <= last_range->end_byte) {
+      last_range->end_byte = end.bytes;
+      last_range->end_point = end.extent;
+      return;
+    }
+  }
+
+  if (start.bytes < end.bytes) {
+    TSRange range = { start.extent, end.extent, start.bytes, end.bytes };
+    array_push(self, range);
+  }
+}
+
+bool ts_range_array_intersects(
+  const TSRangeArray *self,
+  unsigned start_index,
+  uint32_t start_byte,
+  uint32_t end_byte
+) {
+  for (unsigned i = start_index; i < self->size; i++) {
+    TSRange *range = &self->contents[i];
+    if (range->end_byte > start_byte) {
+      if (range->start_byte >= end_byte) break;
+      return true;
+    }
+  }
+  return false;
+}
+
+void ts_range_array_get_changed_ranges(
+  const TSRange *old_ranges, unsigned old_range_count,
+  const TSRange *new_ranges, unsigned new_range_count,
+  TSRangeArray *differences
+) {
+  unsigned new_index = 0;
+  unsigned old_index = 0;
+  Length current_position = length_zero();
+  bool in_old_range = false;
+  bool in_new_range = false;
+
+  while (old_index < old_range_count || new_index < new_range_count) {
+    const TSRange *old_range = &old_ranges[old_index];
+    const TSRange *new_range = &new_ranges[new_index];
+
+    Length next_old_position;
+    if (in_old_range) {
+      next_old_position = (Length) {old_range->end_byte, old_range->end_point};
+    } else if (old_index < old_range_count) {
+      next_old_position = (Length) {old_range->start_byte, old_range->start_point};
+    } else {
+      next_old_position = LENGTH_MAX;
+    }
+
+    Length next_new_position;
+    if (in_new_range) {
+      next_new_position = (Length) {new_range->end_byte, new_range->end_point};
+    } else if (new_index < new_range_count) {
+      next_new_position = (Length) {new_range->start_byte, new_range->start_point};
+    } else {
+      next_new_position = LENGTH_MAX;
+    }
+
+    if (next_old_position.bytes < next_new_position.bytes) {
+      if (in_old_range != in_new_range) {
+        ts_range_array_add(differences, current_position, next_old_position);
+      }
+      if (in_old_range) old_index++;
+      current_position = next_old_position;
+      in_old_range = !in_old_range;
+    } else if (next_new_position.bytes < next_old_position.bytes) {
+      if (in_old_range != in_new_range) {
+        ts_range_array_add(differences, current_position, next_new_position);
+      }
+      if (in_new_range) new_index++;
+      current_position = next_new_position;
+      in_new_range = !in_new_range;
+    } else {
+      if (in_old_range != in_new_range) {
+        ts_range_array_add(differences, current_position, next_new_position);
+      }
+      if (in_old_range) old_index++;
+      if (in_new_range) new_index++;
+      in_old_range = !in_old_range;
+      in_new_range = !in_new_range;
+      current_position = next_new_position;
+    }
+  }
+}
+
+typedef struct {
+  TreeCursor cursor;
+  const TSLanguage *language;
+  unsigned visible_depth;
+  bool in_padding;
+} Iterator;
+
+static Iterator iterator_new(
+  TreeCursor *cursor,
+  const Subtree *tree,
+  const TSLanguage *language
+) {
+  array_clear(&cursor->stack);
+  array_push(&cursor->stack, ((TreeCursorEntry) {
+    .subtree = tree,
+    .position = length_zero(),
+    .child_index = 0,
+    .structural_child_index = 0,
+  }));
+  return (Iterator) {
+    .cursor = *cursor,
+    .language = language,
+    .visible_depth = 1,
+    .in_padding = false,
+  };
+}
+
+static bool iterator_done(Iterator *self) {
+  return self->cursor.stack.size == 0;
+}
+
+static Length iterator_start_position(Iterator *self) {
+  TreeCursorEntry entry = *array_back(&self->cursor.stack);
+  if (self->in_padding) {
+    return entry.position;
+  } else {
+    return length_add(entry.position, ts_subtree_padding(*entry.subtree));
+  }
+}
+
+static Length iterator_end_position(Iterator *self) {
+  TreeCursorEntry entry = *array_back(&self->cursor.stack);
+  Length result = length_add(entry.position, ts_subtree_padding(*entry.subtree));
+  if (self->in_padding) {
+    return result;
+  } else {
+    return length_add(result, ts_subtree_size(*entry.subtree));
+  }
+}
+
+static bool iterator_tree_is_visible(const Iterator *self) {
+  TreeCursorEntry entry = *array_back(&self->cursor.stack);
+  if (ts_subtree_visible(*entry.subtree)) return true;
+  if (self->cursor.stack.size > 1) {
+    Subtree parent = *self->cursor.stack.contents[self->cursor.stack.size - 2].subtree;
+    return ts_language_alias_at(
+      self->language,
+      parent.ptr->production_id,
+      entry.structural_child_index
+    ) != 0;
+  }
+  return false;
+}
+
+static void iterator_get_visible_state(
+  const Iterator *self,
+  Subtree *tree,
+  TSSymbol *alias_symbol,
+  uint32_t *start_byte
+) {
+  uint32_t i = self->cursor.stack.size - 1;
+
+  if (self->in_padding) {
+    if (i == 0) return;
+    i--;
+  }
+
+  for (; i + 1 > 0; i--) {
+    TreeCursorEntry entry = self->cursor.stack.contents[i];
+
+    if (i > 0) {
+      const Subtree *parent = self->cursor.stack.contents[i - 1].subtree;
+      *alias_symbol = ts_language_alias_at(
+        self->language,
+        parent->ptr->production_id,
+        entry.structural_child_index
+      );
+    }
+
+    if (ts_subtree_visible(*entry.subtree) || *alias_symbol) {
+      *tree = *entry.subtree;
+      *start_byte = entry.position.bytes;
+      break;
+    }
+  }
+}
+
+static void iterator_ascend(Iterator *self) {
+  if (iterator_done(self)) return;
+  if (iterator_tree_is_visible(self) && !self->in_padding) self->visible_depth--;
+  if (array_back(&self->cursor.stack)->child_index > 0) self->in_padding = false;
+  self->cursor.stack.size--;
+}
+
+static bool iterator_descend(Iterator *self, uint32_t goal_position) {
+  if (self->in_padding) return false;
+
+  bool did_descend = false;
+  do {
+    did_descend = false;
+    TreeCursorEntry entry = *array_back(&self->cursor.stack);
+    Length position = entry.position;
+    uint32_t structural_child_index = 0;
+    for (uint32_t i = 0, n = ts_subtree_child_count(*entry.subtree); i < n; i++) {
+      const Subtree *child = &ts_subtree_children(*entry.subtree)[i];
+      Length child_left = length_add(position, ts_subtree_padding(*child));
+      Length child_right = length_add(child_left, ts_subtree_size(*child));
+
+      if (child_right.bytes > goal_position) {
+        array_push(&self->cursor.stack, ((TreeCursorEntry) {
+          .subtree = child,
+          .position = position,
+          .child_index = i,
+          .structural_child_index = structural_child_index,
+        }));
+
+        if (iterator_tree_is_visible(self)) {
+          if (child_left.bytes > goal_position) {
+            self->in_padding = true;
+          } else {
+            self->visible_depth++;
+          }
+          return true;
+        }
+
+        did_descend = true;
+        break;
+      }
+
+      position = child_right;
+      if (!ts_subtree_extra(*child)) structural_child_index++;
+    }
+  } while (did_descend);
+
+  return false;
+}
+
+static void iterator_advance(Iterator *self) {
+  if (self->in_padding) {
+    self->in_padding = false;
+    if (iterator_tree_is_visible(self)) {
+      self->visible_depth++;
+    } else {
+      iterator_descend(self, 0);
+    }
+    return;
+  }
+
+  for (;;) {
+    if (iterator_tree_is_visible(self)) self->visible_depth--;
+    TreeCursorEntry entry = array_pop(&self->cursor.stack);
+    if (iterator_done(self)) return;
+
+    const Subtree *parent = array_back(&self->cursor.stack)->subtree;
+    uint32_t child_index = entry.child_index + 1;
+    if (ts_subtree_child_count(*parent) > child_index) {
+      Length position = length_add(entry.position, ts_subtree_total_size(*entry.subtree));
+      uint32_t structural_child_index = entry.structural_child_index;
+      if (!ts_subtree_extra(*entry.subtree)) structural_child_index++;
+      const Subtree *next_child = &ts_subtree_children(*parent)[child_index];
+
+      array_push(&self->cursor.stack, ((TreeCursorEntry) {
+        .subtree = next_child,
+        .position = position,
+        .child_index = child_index,
+        .structural_child_index = structural_child_index,
+      }));
+
+      if (iterator_tree_is_visible(self)) {
+        if (ts_subtree_padding(*next_child).bytes > 0) {
+          self->in_padding = true;
+        } else {
+          self->visible_depth++;
+        }
+      } else {
+        iterator_descend(self, 0);
+      }
+      break;
+    }
+  }
+}
+
+typedef enum {
+  IteratorDiffers,
+  IteratorMayDiffer,
+  IteratorMatches,
+} IteratorComparison;
+
+static IteratorComparison iterator_compare(
+  const Iterator *old_iter,
+  const Iterator *new_iter
+) {
+  Subtree old_tree = NULL_SUBTREE;
+  Subtree new_tree = NULL_SUBTREE;
+  uint32_t old_start = 0;
+  uint32_t new_start = 0;
+  TSSymbol old_alias_symbol = 0;
+  TSSymbol new_alias_symbol = 0;
+  iterator_get_visible_state(old_iter, &old_tree, &old_alias_symbol, &old_start);
+  iterator_get_visible_state(new_iter, &new_tree, &new_alias_symbol, &new_start);
+
+  if (!old_tree.ptr && !new_tree.ptr) return IteratorMatches;
+  if (!old_tree.ptr || !new_tree.ptr) return IteratorDiffers;
+
+  if (
+    old_alias_symbol == new_alias_symbol &&
+    ts_subtree_symbol(old_tree) == ts_subtree_symbol(new_tree)
+  ) {
+    if (old_start == new_start &&
+        !ts_subtree_has_changes(old_tree) &&
+        ts_subtree_symbol(old_tree) != ts_builtin_sym_error &&
+        ts_subtree_size(old_tree).bytes == ts_subtree_size(new_tree).bytes &&
+        ts_subtree_parse_state(old_tree) != TS_TREE_STATE_NONE &&
+        ts_subtree_parse_state(new_tree) != TS_TREE_STATE_NONE &&
+        (ts_subtree_parse_state(old_tree) == ERROR_STATE) ==
+        (ts_subtree_parse_state(new_tree) == ERROR_STATE)) {
+      return IteratorMatches;
+    } else {
+      return IteratorMayDiffer;
+    }
+  }
+
+  return IteratorDiffers;
+}
+
+#ifdef DEBUG_GET_CHANGED_RANGES
+static inline void iterator_print_state(Iterator *self) {
+  TreeCursorEntry entry = *array_back(&self->cursor.stack);
+  TSPoint start = iterator_start_position(self).extent;
+  TSPoint end = iterator_end_position(self).extent;
+  const char *name = ts_language_symbol_name(self->language, ts_subtree_symbol(*entry.subtree));
+  printf(
+    "(%-25s %s\t depth:%u [%u, %u] - [%u, %u])",
+    name, self->in_padding ? "(p)" : "   ",
+    self->visible_depth,
+    start.row + 1, start.column,
+    end.row + 1, end.column
+  );
+}
+#endif
+
+unsigned ts_subtree_get_changed_ranges(
+  const Subtree *old_tree, const Subtree *new_tree,
+  TreeCursor *cursor1, TreeCursor *cursor2,
+  const TSLanguage *language,
+  const TSRangeArray *included_range_differences,
+  TSRange **ranges
+) {
+  TSRangeArray results = array_new();
+
+  Iterator old_iter = iterator_new(cursor1, old_tree, language);
+  Iterator new_iter = iterator_new(cursor2, new_tree, language);
+
+  unsigned included_range_difference_index = 0;
+
+  Length position = iterator_start_position(&old_iter);
+  Length next_position = iterator_start_position(&new_iter);
+  if (position.bytes < next_position.bytes) {
+    ts_range_array_add(&results, position, next_position);
+    position = next_position;
+  } else if (position.bytes > next_position.bytes) {
+    ts_range_array_add(&results, next_position, position);
+    next_position = position;
+  }
+
+  do {
+    #ifdef DEBUG_GET_CHANGED_RANGES
+    printf("At [%-2u, %-2u] Compare ", position.extent.row + 1, position.extent.column);
+    iterator_print_state(&old_iter);
+    printf("\tvs\t");
+    iterator_print_state(&new_iter);
+    puts("");
+    #endif
+
+    // Compare the old and new subtrees.
+    IteratorComparison comparison = iterator_compare(&old_iter, &new_iter);
+
+    // Even if the two subtrees appear to be identical, they could differ
+    // internally if they contain a range of text that was previously
+    // excluded from the parse, and is now included, or vice-versa.
+    if (comparison == IteratorMatches && ts_range_array_intersects(
+      included_range_differences,
+      included_range_difference_index,
+      position.bytes,
+      iterator_end_position(&old_iter).bytes
+    )) {
+      comparison = IteratorMayDiffer;
+    }
+
+    bool is_changed = false;
+    switch (comparison) {
+      // If the subtrees are definitely identical, move to the end
+      // of both subtrees.
+      case IteratorMatches:
+        next_position = iterator_end_position(&old_iter);
+        break;
+
+      // If the subtrees might differ internally, descend into both
+      // subtrees, finding the first child that spans the current position.
+      case IteratorMayDiffer:
+        if (iterator_descend(&old_iter, position.bytes)) {
+          if (!iterator_descend(&new_iter, position.bytes)) {
+            is_changed = true;
+            next_position = iterator_end_position(&old_iter);
+          }
+        } else if (iterator_descend(&new_iter, position.bytes)) {
+          is_changed = true;
+          next_position = iterator_end_position(&new_iter);
+        } else {
+          next_position = length_min(
+            iterator_end_position(&old_iter),
+            iterator_end_position(&new_iter)
+          );
+        }
+        break;
+
+      // If the subtrees are different, record a change and then move
+      // to the end of both subtrees.
+      case IteratorDiffers:
+        is_changed = true;
+        next_position = length_min(
+          iterator_end_position(&old_iter),
+          iterator_end_position(&new_iter)
+        );
+        break;
+    }
+
+    // Ensure that both iterators are caught up to the current position.
+    while (
+      !iterator_done(&old_iter) &&
+      iterator_end_position(&old_iter).bytes <= next_position.bytes
+    ) iterator_advance(&old_iter);
+    while (
+      !iterator_done(&new_iter) &&
+      iterator_end_position(&new_iter).bytes <= next_position.bytes
+    ) iterator_advance(&new_iter);
+
+    // Ensure that both iterators are at the same depth in the tree.
+    while (old_iter.visible_depth > new_iter.visible_depth) {
+      iterator_ascend(&old_iter);
+    }
+    while (new_iter.visible_depth > old_iter.visible_depth) {
+      iterator_ascend(&new_iter);
+    }
+
+    if (is_changed) {
+      #ifdef DEBUG_GET_CHANGED_RANGES
+      printf(
+        "  change: [[%u, %u] - [%u, %u]]\n",
+        position.extent.row + 1, position.extent.column,
+        next_position.extent.row + 1, next_position.extent.column
+      );
+      #endif
+
+      ts_range_array_add(&results, position, next_position);
+    }
+
+    position = next_position;
+
+    // Keep track of the current position in the included range differences
+    // array in order to avoid scanning the entire array on each iteration.
+    while (included_range_difference_index < included_range_differences->size) {
+      const TSRange *range = &included_range_differences->contents[
+        included_range_difference_index
+      ];
+      if (range->end_byte <= position.bytes) {
+        included_range_difference_index++;
+      } else {
+        break;
+      }
+    }
+  } while (!iterator_done(&old_iter) && !iterator_done(&new_iter));
+
+  Length old_size = ts_subtree_total_size(*old_tree);
+  Length new_size = ts_subtree_total_size(*new_tree);
+  if (old_size.bytes < new_size.bytes) {
+    ts_range_array_add(&results, old_size, new_size);
+  } else if (new_size.bytes < old_size.bytes) {
+    ts_range_array_add(&results, new_size, old_size);
+  }
+
+  *cursor1 = old_iter.cursor;
+  *cursor2 = new_iter.cursor;
+  *ranges = results.contents;
+  return results.size;
+}

vendor/tree-sitter/src/get_changed_ranges.h

@@ -0,0 +1,36 @@
+#ifndef TREE_SITTER_GET_CHANGED_RANGES_H_
+#define TREE_SITTER_GET_CHANGED_RANGES_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "./tree_cursor.h"
+#include "./subtree.h"
+
+typedef Array(TSRange) TSRangeArray;
+
+void ts_range_array_get_changed_ranges(
+  const TSRange *old_ranges, unsigned old_range_count,
+  const TSRange *new_ranges, unsigned new_range_count,
+  TSRangeArray *differences
+);
+
+bool ts_range_array_intersects(
+  const TSRangeArray *self, unsigned start_index,
+  uint32_t start_byte, uint32_t end_byte
+);
+
+unsigned ts_subtree_get_changed_ranges(
+  const Subtree *old_tree, const Subtree *new_tree,
+  TreeCursor *cursor1, TreeCursor *cursor2,
+  const TSLanguage *language,
+  const TSRangeArray *included_range_differences,
+  TSRange **ranges
+);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TREE_SITTER_GET_CHANGED_RANGES_H_

vendor/tree-sitter/src/host.h

@@ -0,0 +1,21 @@
+
+// Determine endian and pointer size based on known defines.
+// TS_BIG_ENDIAN and TS_PTR_SIZE can be set as -D compiler arguments
+// to override this.
+
+#if !defined(TS_BIG_ENDIAN)
+#if (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__) \
+  || (defined( __APPLE_CC__) && (defined(__ppc__) || defined(__ppc64__)))
+#define TS_BIG_ENDIAN 1
+#else
+#define TS_BIG_ENDIAN 0
+#endif
+#endif
+
+#if !defined(TS_PTR_SIZE)
+#if UINTPTR_MAX == 0xFFFFFFFF
+#define TS_PTR_SIZE 32
+#else
+#define TS_PTR_SIZE 64
+#endif
+#endif

vendor/tree-sitter/src/language.c

@@ -0,0 +1,221 @@
+#include "./language.h"
+#include "./wasm_store.h"
+#include "tree_sitter/api.h"
+#include <string.h>
+
+const TSLanguage *ts_language_copy(const TSLanguage *self) {
+  if (self && ts_language_is_wasm(self)) {
+    ts_wasm_language_retain(self);
+  }
+  return self;
+}
+
+void ts_language_delete(const TSLanguage *self) {
+  if (self && ts_language_is_wasm(self)) {
+    ts_wasm_language_release(self);
+  }
+}
+
+uint32_t ts_language_symbol_count(const TSLanguage *self) {
+  return self->symbol_count + self->alias_count;
+}
+
+uint32_t ts_language_state_count(const TSLanguage *self) {
+  return self->state_count;
+}
+
+uint32_t ts_language_version(const TSLanguage *self) {
+  return self->version;
+}
+
+uint32_t ts_language_field_count(const TSLanguage *self) {
+  return self->field_count;
+}
+
+void ts_language_table_entry(
+  const TSLanguage *self,
+  TSStateId state,
+  TSSymbol symbol,
+  TableEntry *result
+) {
+  if (symbol == ts_builtin_sym_error || symbol == ts_builtin_sym_error_repeat) {
+    result->action_count = 0;
+    result->is_reusable = false;
+    result->actions = NULL;
+  } else {
+    assert(symbol < self->token_count);
+    uint32_t action_index = ts_language_lookup(self, state, symbol);
+    const TSParseActionEntry *entry = &self->parse_actions[action_index];
+    result->action_count = entry->entry.count;
+    result->is_reusable = entry->entry.reusable;
+    result->actions = (const TSParseAction *)(entry + 1);
+  }
+}
+
+TSSymbolMetadata ts_language_symbol_metadata(
+  const TSLanguage *self,
+  TSSymbol symbol
+) {
+  if (symbol == ts_builtin_sym_error)  {
+    return (TSSymbolMetadata) {.visible = true, .named = true};
+  } else if (symbol == ts_builtin_sym_error_repeat) {
+    return (TSSymbolMetadata) {.visible = false, .named = false};
+  } else {
+    return self->symbol_metadata[symbol];
+  }
+}
+
+TSSymbol ts_language_public_symbol(
+  const TSLanguage *self,
+  TSSymbol symbol
+) {
+  if (symbol == ts_builtin_sym_error) return symbol;
+  return self->public_symbol_map[symbol];
+}
+
+TSStateId ts_language_next_state(
+  const TSLanguage *self,
+  TSStateId state,
+  TSSymbol symbol
+) {
+  if (symbol == ts_builtin_sym_error || symbol == ts_builtin_sym_error_repeat) {
+    return 0;
+  } else if (symbol < self->token_count) {
+    uint32_t count;
+    const TSParseAction *actions = ts_language_actions(self, state, symbol, &count);
+    if (count > 0) {
+      TSParseAction action = actions[count - 1];
+      if (action.type == TSParseActionTypeShift) {
+        return action.shift.extra ? state : action.shift.state;
+      }
+    }
+    return 0;
+  } else {
+    return ts_language_lookup(self, state, symbol);
+  }
+}
+
+const char *ts_language_symbol_name(
+  const TSLanguage *self,
+  TSSymbol symbol
+) {
+  if (symbol == ts_builtin_sym_error) {
+    return "ERROR";
+  } else if (symbol == ts_builtin_sym_error_repeat) {
+    return "_ERROR";
+  } else if (symbol < ts_language_symbol_count(self)) {
+    return self->symbol_names[symbol];
+  } else {
+    return NULL;
+  }
+}
+
+TSSymbol ts_language_symbol_for_name(
+  const TSLanguage *self,
+  const char *string,
+  uint32_t length,
+  bool is_named
+) {
+  if (!strncmp(string, "ERROR", length)) return ts_builtin_sym_error;
+  uint16_t count = (uint16_t)ts_language_symbol_count(self);
+  for (TSSymbol i = 0; i < count; i++) {
+    TSSymbolMetadata metadata = ts_language_symbol_metadata(self, i);
+    if ((!metadata.visible && !metadata.supertype) || metadata.named != is_named) continue;
+    const char *symbol_name = self->symbol_names[i];
+    if (!strncmp(symbol_name, string, length) && !symbol_name[length]) {
+      return self->public_symbol_map[i];
+    }
+  }
+  return 0;
+}
+
+TSSymbolType ts_language_symbol_type(
+  const TSLanguage *self,
+  TSSymbol symbol
+) {
+  TSSymbolMetadata metadata = ts_language_symbol_metadata(self, symbol);
+  if (metadata.named && metadata.visible) {
+    return TSSymbolTypeRegular;
+  } else if (metadata.visible) {
+    return TSSymbolTypeAnonymous;
+  } else {
+    return TSSymbolTypeAuxiliary;
+  }
+}
+
+const char *ts_language_field_name_for_id(
+  const TSLanguage *self,
+  TSFieldId id
+) {
+  uint32_t count = ts_language_field_count(self);
+  if (count && id <= count) {
+    return self->field_names[id];
+  } else {
+    return NULL;
+  }
+}
+
+TSFieldId ts_language_field_id_for_name(
+  const TSLanguage *self,
+  const char *name,
+  uint32_t name_length
+) {
+  uint16_t count = (uint16_t)ts_language_field_count(self);
+  for (TSSymbol i = 1; i < count + 1; i++) {
+    switch (strncmp(name, self->field_names[i], name_length)) {
+      case 0:
+        if (self->field_names[i][name_length] == 0) return i;
+        break;
+      case -1:
+        return 0;
+      default:
+        break;
+    }
+  }
+  return 0;
+}
+
+TSLookaheadIterator *ts_lookahead_iterator_new(const TSLanguage *self, TSStateId state) {
+  if (state >= self->state_count) return NULL;
+  LookaheadIterator *iterator = ts_malloc(sizeof(LookaheadIterator));
+  *iterator = ts_language_lookaheads(self, state);
+  return (TSLookaheadIterator *)iterator;
+}
+
+void ts_lookahead_iterator_delete(TSLookaheadIterator *self) {
+  ts_free(self);
+}
+
+bool ts_lookahead_iterator_reset_state(TSLookaheadIterator * self, TSStateId state) {
+  LookaheadIterator *iterator = (LookaheadIterator *)self;
+  if (state >= iterator->language->state_count) return false;
+  *iterator = ts_language_lookaheads(iterator->language, state);
+  return true;
+}
+
+const TSLanguage *ts_lookahead_iterator_language(const TSLookaheadIterator *self) {
+  const LookaheadIterator *iterator = (const LookaheadIterator *)self;
+  return iterator->language;
+}
+
+bool ts_lookahead_iterator_reset(TSLookaheadIterator *self, const TSLanguage *language, TSStateId state) {
+  if (state >= language->state_count) return false;
+  LookaheadIterator *iterator = (LookaheadIterator *)self;
+  *iterator = ts_language_lookaheads(language, state);
+  return true;
+}
+
+bool ts_lookahead_iterator_next(TSLookaheadIterator *self) {
+  LookaheadIterator *iterator = (LookaheadIterator *)self;
+  return ts_lookahead_iterator__next(iterator);
+}
+
+TSSymbol ts_lookahead_iterator_current_symbol(const TSLookaheadIterator *self) {
+  const LookaheadIterator *iterator = (const LookaheadIterator *)self;
+  return iterator->symbol;
+}
+
+const char *ts_lookahead_iterator_current_symbol_name(const TSLookaheadIterator *self) {
+  const LookaheadIterator *iterator = (const LookaheadIterator *)self;
+  return ts_language_symbol_name(iterator->language, iterator->symbol);
+}

vendor/tree-sitter/src/language.h

@@ -0,0 +1,299 @@
+#ifndef TREE_SITTER_LANGUAGE_H_
+#define TREE_SITTER_LANGUAGE_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "./subtree.h"
+#include "./parser.h"
+
+#define ts_builtin_sym_error_repeat (ts_builtin_sym_error - 1)
+
+#define LANGUAGE_VERSION_WITH_PRIMARY_STATES 14
+#define LANGUAGE_VERSION_USABLE_VIA_WASM 13
+
+typedef struct {
+  const TSParseAction *actions;
+  uint32_t action_count;
+  bool is_reusable;
+} TableEntry;
+
+typedef struct {
+  const TSLanguage *language;
+  const uint16_t *data;
+  const uint16_t *group_end;
+  TSStateId state;
+  uint16_t table_value;
+  uint16_t section_index;
+  uint16_t group_count;
+  bool is_small_state;
+
+  const TSParseAction *actions;
+  TSSymbol symbol;
+  TSStateId next_state;
+  uint16_t action_count;
+} LookaheadIterator;
+
+void ts_language_table_entry(const TSLanguage *, TSStateId, TSSymbol, TableEntry *);
+
+TSSymbolMetadata ts_language_symbol_metadata(const TSLanguage *, TSSymbol);
+
+TSSymbol ts_language_public_symbol(const TSLanguage *, TSSymbol);
+
+TSStateId ts_language_next_state(const TSLanguage *self, TSStateId state, TSSymbol symbol);
+
+static inline bool ts_language_is_symbol_external(const TSLanguage *self, TSSymbol symbol) {
+  return 0 < symbol && symbol < self->external_token_count + 1;
+}
+
+static inline const TSParseAction *ts_language_actions(
+  const TSLanguage *self,
+  TSStateId state,
+  TSSymbol symbol,
+  uint32_t *count
+) {
+  TableEntry entry;
+  ts_language_table_entry(self, state, symbol, &entry);
+  *count = entry.action_count;
+  return entry.actions;
+}
+
+static inline bool ts_language_has_reduce_action(
+  const TSLanguage *self,
+  TSStateId state,
+  TSSymbol symbol
+) {
+  TableEntry entry;
+  ts_language_table_entry(self, state, symbol, &entry);
+  return entry.action_count > 0 && entry.actions[0].type == TSParseActionTypeReduce;
+}
+
+// Lookup the table value for a given symbol and state.
+//
+// For non-terminal symbols, the table value represents a successor state.
+// For terminal symbols, it represents an index in the actions table.
+// For 'large' parse states, this is a direct lookup. For 'small' parse
+// states, this requires searching through the symbol groups to find
+// the given symbol.
+static inline uint16_t ts_language_lookup(
+  const TSLanguage *self,
+  TSStateId state,
+  TSSymbol symbol
+) {
+  if (state >= self->large_state_count) {
+    uint32_t index = self->small_parse_table_map[state - self->large_state_count];
+    const uint16_t *data = &self->small_parse_table[index];
+    uint16_t group_count = *(data++);
+    for (unsigned i = 0; i < group_count; i++) {
+      uint16_t section_value = *(data++);
+      uint16_t symbol_count = *(data++);
+      for (unsigned j = 0; j < symbol_count; j++) {
+        if (*(data++) == symbol) return section_value;
+      }
+    }
+    return 0;
+  } else {
+    return self->parse_table[state * self->symbol_count + symbol];
+  }
+}
+
+static inline bool ts_language_has_actions(
+  const TSLanguage *self,
+  TSStateId state,
+  TSSymbol symbol
+) {
+  return ts_language_lookup(self, state, symbol) != 0;
+}
+
+// Iterate over all of the symbols that are valid in the given state.
+//
+// For 'large' parse states, this just requires iterating through
+// all possible symbols and checking the parse table for each one.
+// For 'small' parse states, this exploits the structure of the
+// table to only visit the valid symbols.
+static inline LookaheadIterator ts_language_lookaheads(
+  const TSLanguage *self,
+  TSStateId state
+) {
+  bool is_small_state = state >= self->large_state_count;
+  const uint16_t *data;
+  const uint16_t *group_end = NULL;
+  uint16_t group_count = 0;
+  if (is_small_state) {
+    uint32_t index = self->small_parse_table_map[state - self->large_state_count];
+    data = &self->small_parse_table[index];
+    group_end = data + 1;
+    group_count = *data;
+  } else {
+    data = &self->parse_table[state * self->symbol_count] - 1;
+  }
+  return (LookaheadIterator) {
+    .language = self,
+    .data = data,
+    .group_end = group_end,
+    .group_count = group_count,
+    .is_small_state = is_small_state,
+    .symbol = UINT16_MAX,
+    .next_state = 0,
+  };
+}
+
+static inline bool ts_lookahead_iterator__next(LookaheadIterator *self) {
+  // For small parse states, valid symbols are listed explicitly,
+  // grouped by their value. There's no need to look up the actions
+  // again until moving to the next group.
+  if (self->is_small_state) {
+    self->data++;
+    if (self->data == self->group_end) {
+      if (self->group_count == 0) return false;
+      self->group_count--;
+      self->table_value = *(self->data++);
+      unsigned symbol_count = *(self->data++);
+      self->group_end = self->data + symbol_count;
+      self->symbol = *self->data;
+    } else {
+      self->symbol = *self->data;
+      return true;
+    }
+  }
+
+  // For large parse states, iterate through every symbol until one
+  // is found that has valid actions.
+  else {
+    do {
+      self->data++;
+      self->symbol++;
+      if (self->symbol >= self->language->symbol_count) return false;
+      self->table_value = *self->data;
+    } while (!self->table_value);
+  }
+
+  // Depending on if the symbols is terminal or non-terminal, the table value either
+  // represents a list of actions or a successor state.
+  if (self->symbol < self->language->token_count) {
+    const TSParseActionEntry *entry = &self->language->parse_actions[self->table_value];
+    self->action_count = entry->entry.count;
+    self->actions = (const TSParseAction *)(entry + 1);
+    self->next_state = 0;
+  } else {
+    self->action_count = 0;
+    self->next_state = self->table_value;
+  }
+  return true;
+}
+
+// Whether the state is a "primary state". If this returns false, it indicates that there exists
+// another state that behaves identically to this one with respect to query analysis.
+static inline bool ts_language_state_is_primary(
+  const TSLanguage *self,
+  TSStateId state
+) {
+  if (self->version >= LANGUAGE_VERSION_WITH_PRIMARY_STATES) {
+    return state == self->primary_state_ids[state];
+  } else {
+    return true;
+  }
+}
+
+static inline const bool *ts_language_enabled_external_tokens(
+  const TSLanguage *self,
+  unsigned external_scanner_state
+) {
+  if (external_scanner_state == 0) {
+    return NULL;
+  } else {
+    return self->external_scanner.states + self->external_token_count * external_scanner_state;
+  }
+}
+
+static inline const TSSymbol *ts_language_alias_sequence(
+  const TSLanguage *self,
+  uint32_t production_id
+) {
+  return production_id ?
+    &self->alias_sequences[production_id * self->max_alias_sequence_length] :
+    NULL;
+}
+
+static inline TSSymbol ts_language_alias_at(
+  const TSLanguage *self,
+  uint32_t production_id,
+  uint32_t child_index
+) {
+  return production_id ?
+    self->alias_sequences[production_id * self->max_alias_sequence_length + child_index] :
+    0;
+}
+
+static inline void ts_language_field_map(
+  const TSLanguage *self,
+  uint32_t production_id,
+  const TSFieldMapEntry **start,
+  const TSFieldMapEntry **end
+) {
+  if (self->field_count == 0) {
+    *start = NULL;
+    *end = NULL;
+    return;
+  }
+
+  TSFieldMapSlice slice = self->field_map_slices[production_id];
+  *start = &self->field_map_entries[slice.index];
+  *end = &self->field_map_entries[slice.index] + slice.length;
+}
+
+static inline void ts_language_aliases_for_symbol(
+  const TSLanguage *self,
+  TSSymbol original_symbol,
+  const TSSymbol **start,
+  const TSSymbol **end
+) {
+  *start = &self->public_symbol_map[original_symbol];
+  *end = *start + 1;
+
+  unsigned idx = 0;
+  for (;;) {
+    TSSymbol symbol = self->alias_map[idx++];
+    if (symbol == 0 || symbol > original_symbol) break;
+    uint16_t count = self->alias_map[idx++];
+    if (symbol == original_symbol) {
+      *start = &self->alias_map[idx];
+      *end = &self->alias_map[idx + count];
+      break;
+    }
+    idx += count;
+  }
+}
+
+static inline void ts_language_write_symbol_as_dot_string(
+  const TSLanguage *self,
+  FILE *f,
+  TSSymbol symbol
+) {
+  const char *name = ts_language_symbol_name(self, symbol);
+  for (const char *chr = name; *chr; chr++) {
+    switch (*chr) {
+      case '"':
+      case '\\':
+        fputc('\\', f);
+        fputc(*chr, f);
+        break;
+      case '\n':
+        fputs("\\n", f);
+        break;
+      case '\t':
+        fputs("\\t", f);
+        break;
+      default:
+        fputc(*chr, f);
+        break;
+    }
+  }
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TREE_SITTER_LANGUAGE_H_

vendor/tree-sitter/src/length.h

@@ -0,0 +1,52 @@
+#ifndef TREE_SITTER_LENGTH_H_
+#define TREE_SITTER_LENGTH_H_
+
+#include <stdlib.h>
+#include <stdbool.h>
+#include "./point.h"
+#include "tree_sitter/api.h"
+
+typedef struct {
+  uint32_t bytes;
+  TSPoint extent;
+} Length;
+
+static const Length LENGTH_UNDEFINED = {0, {0, 1}};
+static const Length LENGTH_MAX = {UINT32_MAX, {UINT32_MAX, UINT32_MAX}};
+
+static inline bool length_is_undefined(Length length) {
+  return length.bytes == 0 && length.extent.column != 0;
+}
+
+static inline Length length_min(Length len1, Length len2) {
+  return (len1.bytes < len2.bytes) ? len1 : len2;
+}
+
+static inline Length length_add(Length len1, Length len2) {
+  Length result;
+  result.bytes = len1.bytes + len2.bytes;
+  result.extent = point_add(len1.extent, len2.extent);
+  return result;
+}
+
+static inline Length length_sub(Length len1, Length len2) {
+  Length result;
+  result.bytes = len1.bytes - len2.bytes;
+  result.extent = point_sub(len1.extent, len2.extent);
+  return result;
+}
+
+static inline Length length_zero(void) {
+  Length result = {0, {0, 0}};
+  return result;
+}
+
+static inline Length length_saturating_sub(Length len1, Length len2) {
+  if (len1.bytes > len2.bytes) {
+    return length_sub(len1, len2);
+  } else {
+    return length_zero();
+  }
+}
+
+#endif

vendor/tree-sitter/src/lexer.c

@@ -0,0 +1,419 @@
+#include <stdio.h>
+#include "./lexer.h"
+#include "./subtree.h"
+#include "./length.h"
+#include "./unicode.h"
+
+#define LOG(message, character)              \
+  if (self->logger.log) {                    \
+    snprintf(                                \
+      self->debug_buffer,                    \
+      TREE_SITTER_SERIALIZATION_BUFFER_SIZE, \
+      32 <= character && character < 127 ?   \
+        message " character:'%c'" :          \
+        message " character:%d",             \
+      character                              \
+    );                                       \
+    self->logger.log(                        \
+      self->logger.payload,                  \
+      TSLogTypeLex,                          \
+      self->debug_buffer                     \
+    );                                       \
+  }
+
+static const int32_t BYTE_ORDER_MARK = 0xFEFF;
+
+static const TSRange DEFAULT_RANGE = {
+  .start_point = {
+    .row = 0,
+    .column = 0,
+  },
+  .end_point = {
+    .row = UINT32_MAX,
+    .column = UINT32_MAX,
+  },
+  .start_byte = 0,
+  .end_byte = UINT32_MAX
+};
+
+// Check if the lexer has reached EOF. This state is stored
+// by setting the lexer's `current_included_range_index` such that
+// it has consumed all of its available ranges.
+static bool ts_lexer__eof(const TSLexer *_self) {
+  Lexer *self = (Lexer *)_self;
+  return self->current_included_range_index == self->included_range_count;
+}
+
+// Clear the currently stored chunk of source code, because the lexer's
+// position has changed.
+static void ts_lexer__clear_chunk(Lexer *self) {
+  self->chunk = NULL;
+  self->chunk_size = 0;
+  self->chunk_start = 0;
+}
+
+// Call the lexer's input callback to obtain a new chunk of source code
+// for the current position.
+static void ts_lexer__get_chunk(Lexer *self) {
+  self->chunk_start = self->current_position.bytes;
+  self->chunk = self->input.read(
+    self->input.payload,
+    self->current_position.bytes,
+    self->current_position.extent,
+    &self->chunk_size
+  );
+  if (!self->chunk_size) {
+    self->current_included_range_index = self->included_range_count;
+    self->chunk = NULL;
+  }
+}
+
+// Decode the next unicode character in the current chunk of source code.
+// This assumes that the lexer has already retrieved a chunk of source
+// code that spans the current position.
+static void ts_lexer__get_lookahead(Lexer *self) {
+  uint32_t position_in_chunk = self->current_position.bytes - self->chunk_start;
+  uint32_t size = self->chunk_size - position_in_chunk;
+
+  if (size == 0) {
+    self->lookahead_size = 1;
+    self->data.lookahead = '\0';
+    return;
+  }
+
+  const uint8_t *chunk = (const uint8_t *)self->chunk + position_in_chunk;
+  UnicodeDecodeFunction decode = self->input.encoding == TSInputEncodingUTF8
+    ? ts_decode_utf8
+    : ts_decode_utf16;
+
+  self->lookahead_size = decode(chunk, size, &self->data.lookahead);
+
+  // If this chunk ended in the middle of a multi-byte character,
+  // try again with a fresh chunk.
+  if (self->data.lookahead == TS_DECODE_ERROR && size < 4) {
+    ts_lexer__get_chunk(self);
+    chunk = (const uint8_t *)self->chunk;
+    size = self->chunk_size;
+    self->lookahead_size = decode(chunk, size, &self->data.lookahead);
+  }
+
+  if (self->data.lookahead == TS_DECODE_ERROR) {
+    self->lookahead_size = 1;
+  }
+}
+
+static void ts_lexer_goto(Lexer *self, Length position) {
+  self->current_position = position;
+
+  // Move to the first valid position at or after the given position.
+  bool found_included_range = false;
+  for (unsigned i = 0; i < self->included_range_count; i++) {
+    TSRange *included_range = &self->included_ranges[i];
+    if (
+      included_range->end_byte > self->current_position.bytes &&
+      included_range->end_byte > included_range->start_byte
+    ) {
+      if (included_range->start_byte >= self->current_position.bytes) {
+        self->current_position = (Length) {
+          .bytes = included_range->start_byte,
+          .extent = included_range->start_point,
+        };
+      }
+
+      self->current_included_range_index = i;
+      found_included_range = true;
+      break;
+    }
+  }
+
+  if (found_included_range) {
+    // If the current position is outside of the current chunk of text,
+    // then clear out the current chunk of text.
+    if (self->chunk && (
+      self->current_position.bytes < self->chunk_start ||
+      self->current_position.bytes >= self->chunk_start + self->chunk_size
+    )) {
+      ts_lexer__clear_chunk(self);
+    }
+
+    self->lookahead_size = 0;
+    self->data.lookahead = '\0';
+  }
+
+  // If the given position is beyond any of included ranges, move to the EOF
+  // state - past the end of the included ranges.
+  else {
+    self->current_included_range_index = self->included_range_count;
+    TSRange *last_included_range = &self->included_ranges[self->included_range_count - 1];
+    self->current_position = (Length) {
+      .bytes = last_included_range->end_byte,
+      .extent = last_included_range->end_point,
+    };
+    ts_lexer__clear_chunk(self);
+    self->lookahead_size = 1;
+    self->data.lookahead = '\0';
+  }
+}
+
+// Intended to be called only from functions that control logging.
+static void ts_lexer__do_advance(Lexer *self, bool skip) {
+  if (self->lookahead_size) {
+    self->current_position.bytes += self->lookahead_size;
+    if (self->data.lookahead == '\n') {
+      self->current_position.extent.row++;
+      self->current_position.extent.column = 0;
+    } else {
+      self->current_position.extent.column += self->lookahead_size;
+    }
+  }
+
+  const TSRange *current_range = &self->included_ranges[self->current_included_range_index];
+  while (
+    self->current_position.bytes >= current_range->end_byte ||
+    current_range->end_byte == current_range->start_byte
+  ) {
+    if (self->current_included_range_index < self->included_range_count) {
+      self->current_included_range_index++;
+    }
+    if (self->current_included_range_index < self->included_range_count) {
+      current_range++;
+      self->current_position = (Length) {
+        current_range->start_byte,
+        current_range->start_point,
+      };
+    } else {
+      current_range = NULL;
+      break;
+    }
+  }
+
+  if (skip) self->token_start_position = self->current_position;
+
+  if (current_range) {
+    if (
+      self->current_position.bytes < self->chunk_start ||
+      self->current_position.bytes >= self->chunk_start + self->chunk_size
+    ) {
+      ts_lexer__get_chunk(self);
+    }
+    ts_lexer__get_lookahead(self);
+  } else {
+    ts_lexer__clear_chunk(self);
+    self->data.lookahead = '\0';
+    self->lookahead_size = 1;
+  }
+}
+
+// Advance to the next character in the source code, retrieving a new
+// chunk of source code if needed.
+static void ts_lexer__advance(TSLexer *_self, bool skip) {
+  Lexer *self = (Lexer *)_self;
+  if (!self->chunk) return;
+
+  if (skip) {
+    LOG("skip", self->data.lookahead)
+  } else {
+    LOG("consume", self->data.lookahead)
+  }
+
+  ts_lexer__do_advance(self, skip);
+}
+
+// Mark that a token match has completed. This can be called multiple
+// times if a longer match is found later.
+static void ts_lexer__mark_end(TSLexer *_self) {
+  Lexer *self = (Lexer *)_self;
+  if (!ts_lexer__eof(&self->data)) {
+    // If the lexer is right at the beginning of included range,
+    // then the token should be considered to end at the *end* of the
+    // previous included range, rather than here.
+    TSRange *current_included_range = &self->included_ranges[
+      self->current_included_range_index
+    ];
+    if (
+      self->current_included_range_index > 0 &&
+      self->current_position.bytes == current_included_range->start_byte
+    ) {
+      TSRange *previous_included_range = current_included_range - 1;
+      self->token_end_position = (Length) {
+        previous_included_range->end_byte,
+        previous_included_range->end_point,
+      };
+      return;
+    }
+  }
+  self->token_end_position = self->current_position;
+}
+
+static uint32_t ts_lexer__get_column(TSLexer *_self) {
+  Lexer *self = (Lexer *)_self;
+
+  uint32_t goal_byte = self->current_position.bytes;
+
+  self->did_get_column = true;
+  self->current_position.bytes -= self->current_position.extent.column;
+  self->current_position.extent.column = 0;
+
+  if (self->current_position.bytes < self->chunk_start) {
+    ts_lexer__get_chunk(self);
+  }
+
+  uint32_t result = 0;
+  if (!ts_lexer__eof(_self)) {
+    ts_lexer__get_lookahead(self);
+    while (self->current_position.bytes < goal_byte && self->chunk) {
+      result++;
+      ts_lexer__do_advance(self, false);
+      if (ts_lexer__eof(_self)) break;
+    }
+  }
+
+  return result;
+}
+
+// Is the lexer at a boundary between two disjoint included ranges of
+// source code? This is exposed as an API because some languages' external
+// scanners need to perform custom actions at these boundaries.
+static bool ts_lexer__is_at_included_range_start(const TSLexer *_self) {
+  const Lexer *self = (const Lexer *)_self;
+  if (self->current_included_range_index < self->included_range_count) {
+    TSRange *current_range = &self->included_ranges[self->current_included_range_index];
+    return self->current_position.bytes == current_range->start_byte;
+  } else {
+    return false;
+  }
+}
+
+void ts_lexer_init(Lexer *self) {
+  *self = (Lexer) {
+    .data = {
+      // The lexer's methods are stored as struct fields so that generated
+      // parsers can call them without needing to be linked against this
+      // library.
+      .advance = ts_lexer__advance,
+      .mark_end = ts_lexer__mark_end,
+      .get_column = ts_lexer__get_column,
+      .is_at_included_range_start = ts_lexer__is_at_included_range_start,
+      .eof = ts_lexer__eof,
+      .lookahead = 0,
+      .result_symbol = 0,
+    },
+    .chunk = NULL,
+    .chunk_size = 0,
+    .chunk_start = 0,
+    .current_position = {0, {0, 0}},
+    .logger = {
+      .payload = NULL,
+      .log = NULL
+    },
+    .included_ranges = NULL,
+    .included_range_count = 0,
+    .current_included_range_index = 0,
+  };
+  ts_lexer_set_included_ranges(self, NULL, 0);
+}
+
+void ts_lexer_delete(Lexer *self) {
+  ts_free(self->included_ranges);
+}
+
+void ts_lexer_set_input(Lexer *self, TSInput input) {
+  self->input = input;
+  ts_lexer__clear_chunk(self);
+  ts_lexer_goto(self, self->current_position);
+}
+
+// Move the lexer to the given position. This doesn't do any work
+// if the parser is already at the given position.
+void ts_lexer_reset(Lexer *self, Length position) {
+  if (position.bytes != self->current_position.bytes) {
+    ts_lexer_goto(self, position);
+  }
+}
+
+void ts_lexer_start(Lexer *self) {
+  self->token_start_position = self->current_position;
+  self->token_end_position = LENGTH_UNDEFINED;
+  self->data.result_symbol = 0;
+  self->did_get_column = false;
+  if (!ts_lexer__eof(&self->data)) {
+    if (!self->chunk_size) ts_lexer__get_chunk(self);
+    if (!self->lookahead_size) ts_lexer__get_lookahead(self);
+    if (
+      self->current_position.bytes == 0 &&
+      self->data.lookahead == BYTE_ORDER_MARK
+    ) ts_lexer__advance(&self->data, true);
+  }
+}
+
+void ts_lexer_finish(Lexer *self, uint32_t *lookahead_end_byte) {
+  if (length_is_undefined(self->token_end_position)) {
+    ts_lexer__mark_end(&self->data);
+  }
+
+  // If the token ended at an included range boundary, then its end position
+  // will have been reset to the end of the preceding range. Reset the start
+  // position to match.
+  if (self->token_end_position.bytes < self->token_start_position.bytes) {
+    self->token_start_position = self->token_end_position;
+  }
+
+  uint32_t current_lookahead_end_byte = self->current_position.bytes + 1;
+
+  // In order to determine that a byte sequence is invalid UTF8 or UTF16,
+  // the character decoding algorithm may have looked at the following byte.
+  // Therefore, the next byte *after* the current (invalid) character
+  // affects the interpretation of the current character.
+  if (self->data.lookahead == TS_DECODE_ERROR) {
+    current_lookahead_end_byte++;
+  }
+
+  if (current_lookahead_end_byte > *lookahead_end_byte) {
+    *lookahead_end_byte = current_lookahead_end_byte;
+  }
+}
+
+void ts_lexer_advance_to_end(Lexer *self) {
+  while (self->chunk) {
+    ts_lexer__advance(&self->data, false);
+  }
+}
+
+void ts_lexer_mark_end(Lexer *self) {
+  ts_lexer__mark_end(&self->data);
+}
+
+bool ts_lexer_set_included_ranges(
+  Lexer *self,
+  const TSRange *ranges,
+  uint32_t count
+) {
+  if (count == 0 || !ranges) {
+    ranges = &DEFAULT_RANGE;
+    count = 1;
+  } else {
+    uint32_t previous_byte = 0;
+    for (unsigned i = 0; i < count; i++) {
+      const TSRange *range = &ranges[i];
+      if (
+        range->start_byte < previous_byte ||
+        range->end_byte < range->start_byte
+      ) return false;
+      previous_byte = range->end_byte;
+    }
+  }
+
+  size_t size = count * sizeof(TSRange);
+  self->included_ranges = ts_realloc(self->included_ranges, size);
+  memcpy(self->included_ranges, ranges, size);
+  self->included_range_count = count;
+  ts_lexer_goto(self, self->current_position);
+  return true;
+}
+
+TSRange *ts_lexer_included_ranges(const Lexer *self, uint32_t *count) {
+  *count = self->included_range_count;
+  return self->included_ranges;
+}
+
+#undef LOG

vendor/tree-sitter/src/lexer.h

@@ -0,0 +1,49 @@
+#ifndef TREE_SITTER_LEXER_H_
+#define TREE_SITTER_LEXER_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "./length.h"
+#include "./subtree.h"
+#include "tree_sitter/api.h"
+#include "./parser.h"
+
+typedef struct {
+  TSLexer data;
+  Length current_position;
+  Length token_start_position;
+  Length token_end_position;
+
+  TSRange *included_ranges;
+  const char *chunk;
+  TSInput input;
+  TSLogger logger;
+
+  uint32_t included_range_count;
+  uint32_t current_included_range_index;
+  uint32_t chunk_start;
+  uint32_t chunk_size;
+  uint32_t lookahead_size;
+  bool did_get_column;
+
+  char debug_buffer[TREE_SITTER_SERIALIZATION_BUFFER_SIZE];
+} Lexer;
+
+void ts_lexer_init(Lexer *);
+void ts_lexer_delete(Lexer *);
+void ts_lexer_set_input(Lexer *, TSInput);
+void ts_lexer_reset(Lexer *, Length);
+void ts_lexer_start(Lexer *);
+void ts_lexer_finish(Lexer *, uint32_t *);
+void ts_lexer_advance_to_end(Lexer *);
+void ts_lexer_mark_end(Lexer *);
+bool ts_lexer_set_included_ranges(Lexer *self, const TSRange *ranges, uint32_t count);
+TSRange *ts_lexer_included_ranges(const Lexer *self, uint32_t *count);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TREE_SITTER_LEXER_H_

vendor/tree-sitter/src/lib.c

@@ -0,0 +1,14 @@
+#define _POSIX_C_SOURCE 200112L
+
+#include "./alloc.c"
+#include "./get_changed_ranges.c"
+#include "./language.c"
+#include "./lexer.c"
+#include "./node.c"
+#include "./parser.c"
+#include "./query.c"
+#include "./stack.c"
+#include "./subtree.c"
+#include "./tree_cursor.c"
+#include "./tree.c"
+#include "./wasm_store.c"

vendor/tree-sitter/src/node.c

@@ -0,0 +1,774 @@
+#include <stdbool.h>
+#include "./subtree.h"
+#include "./tree.h"
+#include "./language.h"
+
+typedef struct {
+  Subtree parent;
+  const TSTree *tree;
+  Length position;
+  uint32_t child_index;
+  uint32_t structural_child_index;
+  const TSSymbol *alias_sequence;
+} NodeChildIterator;
+
+// TSNode - constructors
+
+TSNode ts_node_new(
+  const TSTree *tree,
+  const Subtree *subtree,
+  Length position,
+  TSSymbol alias
+) {
+  return (TSNode) {
+    {position.bytes, position.extent.row, position.extent.column, alias},
+    subtree,
+    tree,
+  };
+}
+
+static inline TSNode ts_node__null(void) {
+  return ts_node_new(NULL, NULL, length_zero(), 0);
+}
+
+// TSNode - accessors
+
+uint32_t ts_node_start_byte(TSNode self) {
+  return self.context[0];
+}
+
+TSPoint ts_node_start_point(TSNode self) {
+  return (TSPoint) {self.context[1], self.context[2]};
+}
+
+static inline uint32_t ts_node__alias(const TSNode *self) {
+  return self->context[3];
+}
+
+static inline Subtree ts_node__subtree(TSNode self) {
+  return *(const Subtree *)self.id;
+}
+
+// NodeChildIterator
+
+static inline NodeChildIterator ts_node_iterate_children(const TSNode *node) {
+  Subtree subtree = ts_node__subtree(*node);
+  if (ts_subtree_child_count(subtree) == 0) {
+    return (NodeChildIterator) {NULL_SUBTREE, node->tree, length_zero(), 0, 0, NULL};
+  }
+  const TSSymbol *alias_sequence = ts_language_alias_sequence(
+    node->tree->language,
+    subtree.ptr->production_id
+  );
+  return (NodeChildIterator) {
+    .tree = node->tree,
+    .parent = subtree,
+    .position = {ts_node_start_byte(*node), ts_node_start_point(*node)},
+    .child_index = 0,
+    .structural_child_index = 0,
+    .alias_sequence = alias_sequence,
+  };
+}
+
+static inline bool ts_node_child_iterator_done(NodeChildIterator *self) {
+  return self->child_index == self->parent.ptr->child_count;
+}
+
+static inline bool ts_node_child_iterator_next(
+  NodeChildIterator *self,
+  TSNode *result
+) {
+  if (!self->parent.ptr || ts_node_child_iterator_done(self)) return false;
+  const Subtree *child = &ts_subtree_children(self->parent)[self->child_index];
+  TSSymbol alias_symbol = 0;
+  if (!ts_subtree_extra(*child)) {
+    if (self->alias_sequence) {
+      alias_symbol = self->alias_sequence[self->structural_child_index];
+    }
+    self->structural_child_index++;
+  }
+  if (self->child_index > 0) {
+    self->position = length_add(self->position, ts_subtree_padding(*child));
+  }
+  *result = ts_node_new(
+    self->tree,
+    child,
+    self->position,
+    alias_symbol
+  );
+  self->position = length_add(self->position, ts_subtree_size(*child));
+  self->child_index++;
+  return true;
+}
+
+// TSNode - private
+
+static inline bool ts_node__is_relevant(TSNode self, bool include_anonymous) {
+  Subtree tree = ts_node__subtree(self);
+  if (include_anonymous) {
+    return ts_subtree_visible(tree) || ts_node__alias(&self);
+  } else {
+    TSSymbol alias = ts_node__alias(&self);
+    if (alias) {
+      return ts_language_symbol_metadata(self.tree->language, alias).named;
+    } else {
+      return ts_subtree_visible(tree) && ts_subtree_named(tree);
+    }
+  }
+}
+
+static inline uint32_t ts_node__relevant_child_count(
+  TSNode self,
+  bool include_anonymous
+) {
+  Subtree tree = ts_node__subtree(self);
+  if (ts_subtree_child_count(tree) > 0) {
+    if (include_anonymous) {
+      return tree.ptr->visible_child_count;
+    } else {
+      return tree.ptr->named_child_count;
+    }
+  } else {
+    return 0;
+  }
+}
+
+static inline TSNode ts_node__child(
+  TSNode self,
+  uint32_t child_index,
+  bool include_anonymous
+) {
+  TSNode result = self;
+  bool did_descend = true;
+
+  while (did_descend) {
+    did_descend = false;
+
+    TSNode child;
+    uint32_t index = 0;
+    NodeChildIterator iterator = ts_node_iterate_children(&result);
+    while (ts_node_child_iterator_next(&iterator, &child)) {
+      if (ts_node__is_relevant(child, include_anonymous)) {
+        if (index == child_index) {
+          return child;
+        }
+        index++;
+      } else {
+        uint32_t grandchild_index = child_index - index;
+        uint32_t grandchild_count = ts_node__relevant_child_count(child, include_anonymous);
+        if (grandchild_index < grandchild_count) {
+          did_descend = true;
+          result = child;
+          child_index = grandchild_index;
+          break;
+        }
+        index += grandchild_count;
+      }
+    }
+  }
+
+  return ts_node__null();
+}
+
+static bool ts_subtree_has_trailing_empty_descendant(
+  Subtree self,
+  Subtree other
+) {
+  for (unsigned i = ts_subtree_child_count(self) - 1; i + 1 > 0; i--) {
+    Subtree child = ts_subtree_children(self)[i];
+    if (ts_subtree_total_bytes(child) > 0) break;
+    if (child.ptr == other.ptr || ts_subtree_has_trailing_empty_descendant(child, other)) {
+      return true;
+    }
+  }
+  return false;
+}
+
+static inline TSNode ts_node__prev_sibling(TSNode self, bool include_anonymous) {
+  Subtree self_subtree = ts_node__subtree(self);
+  bool self_is_empty = ts_subtree_total_bytes(self_subtree) == 0;
+  uint32_t target_end_byte = ts_node_end_byte(self);
+
+  TSNode node = ts_node_parent(self);
+  TSNode earlier_node = ts_node__null();
+  bool earlier_node_is_relevant = false;
+
+  while (!ts_node_is_null(node)) {
+    TSNode earlier_child = ts_node__null();
+    bool earlier_child_is_relevant = false;
+    bool found_child_containing_target = false;
+
+    TSNode child;
+    NodeChildIterator iterator = ts_node_iterate_children(&node);
+    while (ts_node_child_iterator_next(&iterator, &child)) {
+      if (child.id == self.id) break;
+      if (iterator.position.bytes > target_end_byte) {
+        found_child_containing_target = true;
+        break;
+      }
+
+      if (iterator.position.bytes == target_end_byte &&
+          (!self_is_empty ||
+           ts_subtree_has_trailing_empty_descendant(ts_node__subtree(child), self_subtree))) {
+        found_child_containing_target = true;
+        break;
+      }
+
+      if (ts_node__is_relevant(child, include_anonymous)) {
+        earlier_child = child;
+        earlier_child_is_relevant = true;
+      } else if (ts_node__relevant_child_count(child, include_anonymous) > 0) {
+        earlier_child = child;
+        earlier_child_is_relevant = false;
+      }
+    }
+
+    if (found_child_containing_target) {
+      if (!ts_node_is_null(earlier_child)) {
+        earlier_node = earlier_child;
+        earlier_node_is_relevant = earlier_child_is_relevant;
+      }
+      node = child;
+    } else if (earlier_child_is_relevant) {
+      return earlier_child;
+    } else if (!ts_node_is_null(earlier_child)) {
+      node = earlier_child;
+    } else if (earlier_node_is_relevant) {
+      return earlier_node;
+    } else {
+      node = earlier_node;
+      earlier_node = ts_node__null();
+      earlier_node_is_relevant = false;
+    }
+  }
+
+  return ts_node__null();
+}
+
+static inline TSNode ts_node__next_sibling(TSNode self, bool include_anonymous) {
+  uint32_t target_end_byte = ts_node_end_byte(self);
+
+  TSNode node = ts_node_parent(self);
+  TSNode later_node = ts_node__null();
+  bool later_node_is_relevant = false;
+
+  while (!ts_node_is_null(node)) {
+    TSNode later_child = ts_node__null();
+    bool later_child_is_relevant = false;
+    TSNode child_containing_target = ts_node__null();
+
+    TSNode child;
+    NodeChildIterator iterator = ts_node_iterate_children(&node);
+    while (ts_node_child_iterator_next(&iterator, &child)) {
+      if (iterator.position.bytes < target_end_byte) continue;
+      if (ts_node_start_byte(child) <= ts_node_start_byte(self)) {
+        if (ts_node__subtree(child).ptr != ts_node__subtree(self).ptr) {
+          child_containing_target = child;
+        }
+      } else if (ts_node__is_relevant(child, include_anonymous)) {
+        later_child = child;
+        later_child_is_relevant = true;
+        break;
+      } else if (ts_node__relevant_child_count(child, include_anonymous) > 0) {
+        later_child = child;
+        later_child_is_relevant = false;
+        break;
+      }
+    }
+
+    if (!ts_node_is_null(child_containing_target)) {
+      if (!ts_node_is_null(later_child)) {
+        later_node = later_child;
+        later_node_is_relevant = later_child_is_relevant;
+      }
+      node = child_containing_target;
+    } else if (later_child_is_relevant) {
+      return later_child;
+    } else if (!ts_node_is_null(later_child)) {
+      node = later_child;
+    } else if (later_node_is_relevant) {
+      return later_node;
+    } else {
+      node = later_node;
+    }
+  }
+
+  return ts_node__null();
+}
+
+static inline TSNode ts_node__first_child_for_byte(
+  TSNode self,
+  uint32_t goal,
+  bool include_anonymous
+) {
+  TSNode node = self;
+  bool did_descend = true;
+
+  while (did_descend) {
+    did_descend = false;
+
+    TSNode child;
+    NodeChildIterator iterator = ts_node_iterate_children(&node);
+    while (ts_node_child_iterator_next(&iterator, &child)) {
+      if (ts_node_end_byte(child) > goal) {
+        if (ts_node__is_relevant(child, include_anonymous)) {
+          return child;
+        } else if (ts_node_child_count(child) > 0) {
+          did_descend = true;
+          node = child;
+          break;
+        }
+      }
+    }
+  }
+
+  return ts_node__null();
+}
+
+static inline TSNode ts_node__descendant_for_byte_range(
+  TSNode self,
+  uint32_t range_start,
+  uint32_t range_end,
+  bool include_anonymous
+) {
+  TSNode node = self;
+  TSNode last_visible_node = self;
+
+  bool did_descend = true;
+  while (did_descend) {
+    did_descend = false;
+
+    TSNode child;
+    NodeChildIterator iterator = ts_node_iterate_children(&node);
+    while (ts_node_child_iterator_next(&iterator, &child)) {
+      uint32_t node_end = iterator.position.bytes;
+
+      // The end of this node must extend far enough forward to touch
+      // the end of the range and exceed the start of the range.
+      if (node_end < range_end) continue;
+      if (node_end <= range_start) continue;
+
+      // The start of this node must extend far enough backward to
+      // touch the start of the range.
+      if (range_start < ts_node_start_byte(child)) break;
+
+      node = child;
+      if (ts_node__is_relevant(node, include_anonymous)) {
+        last_visible_node = node;
+      }
+      did_descend = true;
+      break;
+    }
+  }
+
+  return last_visible_node;
+}
+
+static inline TSNode ts_node__descendant_for_point_range(
+  TSNode self,
+  TSPoint range_start,
+  TSPoint range_end,
+  bool include_anonymous
+) {
+  TSNode node = self;
+  TSNode last_visible_node = self;
+
+  bool did_descend = true;
+  while (did_descend) {
+    did_descend = false;
+
+    TSNode child;
+    NodeChildIterator iterator = ts_node_iterate_children(&node);
+    while (ts_node_child_iterator_next(&iterator, &child)) {
+      TSPoint node_end = iterator.position.extent;
+
+      // The end of this node must extend far enough forward to touch
+      // the end of the range and exceed the start of the range.
+      if (point_lt(node_end, range_end)) continue;
+      if (point_lte(node_end, range_start)) continue;
+
+      // The start of this node must extend far enough backward to
+      // touch the start of the range.
+      if (point_lt(range_start, ts_node_start_point(child))) break;
+
+      node = child;
+      if (ts_node__is_relevant(node, include_anonymous)) {
+        last_visible_node = node;
+      }
+      did_descend = true;
+      break;
+    }
+  }
+
+  return last_visible_node;
+}
+
+// TSNode - public
+
+uint32_t ts_node_end_byte(TSNode self) {
+  return ts_node_start_byte(self) + ts_subtree_size(ts_node__subtree(self)).bytes;
+}
+
+TSPoint ts_node_end_point(TSNode self) {
+  return point_add(ts_node_start_point(self), ts_subtree_size(ts_node__subtree(self)).extent);
+}
+
+TSSymbol ts_node_symbol(TSNode self) {
+  TSSymbol symbol = ts_node__alias(&self);
+  if (!symbol) symbol = ts_subtree_symbol(ts_node__subtree(self));
+  return ts_language_public_symbol(self.tree->language, symbol);
+}
+
+const char *ts_node_type(TSNode self) {
+  TSSymbol symbol = ts_node__alias(&self);
+  if (!symbol) symbol = ts_subtree_symbol(ts_node__subtree(self));
+  return ts_language_symbol_name(self.tree->language, symbol);
+}
+
+const TSLanguage *ts_node_language(TSNode self) {
+  return self.tree->language;
+}
+
+TSSymbol ts_node_grammar_symbol(TSNode self) {
+  return ts_subtree_symbol(ts_node__subtree(self));
+}
+
+const char *ts_node_grammar_type(TSNode self) {
+  TSSymbol symbol = ts_subtree_symbol(ts_node__subtree(self));
+  return ts_language_symbol_name(self.tree->language, symbol);
+}
+
+char *ts_node_string(TSNode self) {
+  TSSymbol alias_symbol = ts_node__alias(&self);
+  return ts_subtree_string(
+    ts_node__subtree(self),
+    alias_symbol,
+    ts_language_symbol_metadata(self.tree->language, alias_symbol).visible,
+    self.tree->language,
+    false
+  );
+}
+
+bool ts_node_eq(TSNode self, TSNode other) {
+  return self.tree == other.tree && self.id == other.id;
+}
+
+bool ts_node_is_null(TSNode self) {
+  return self.id == 0;
+}
+
+bool ts_node_is_extra(TSNode self) {
+  return ts_subtree_extra(ts_node__subtree(self));
+}
+
+bool ts_node_is_named(TSNode self) {
+  TSSymbol alias = ts_node__alias(&self);
+  return alias
+    ? ts_language_symbol_metadata(self.tree->language, alias).named
+    : ts_subtree_named(ts_node__subtree(self));
+}
+
+bool ts_node_is_missing(TSNode self) {
+  return ts_subtree_missing(ts_node__subtree(self));
+}
+
+bool ts_node_has_changes(TSNode self) {
+  return ts_subtree_has_changes(ts_node__subtree(self));
+}
+
+bool ts_node_has_error(TSNode self) {
+  return ts_subtree_error_cost(ts_node__subtree(self)) > 0;
+}
+
+bool ts_node_is_error(TSNode self) {
+  TSSymbol symbol = ts_node_symbol(self);
+  return symbol == ts_builtin_sym_error;
+}
+
+uint32_t ts_node_descendant_count(TSNode self) {
+  return ts_subtree_visible_descendant_count(ts_node__subtree(self)) + 1;
+}
+
+TSStateId ts_node_parse_state(TSNode self) {
+  return ts_subtree_parse_state(ts_node__subtree(self));
+}
+
+TSStateId ts_node_next_parse_state(TSNode self) {
+  const TSLanguage *language = self.tree->language;
+  uint16_t state = ts_node_parse_state(self);
+  if (state == TS_TREE_STATE_NONE) {
+    return TS_TREE_STATE_NONE;
+  }
+  uint16_t symbol = ts_node_grammar_symbol(self);
+  return ts_language_next_state(language, state, symbol);
+}
+
+TSNode ts_node_parent(TSNode self) {
+  TSNode node = ts_tree_root_node(self.tree);
+  uint32_t end_byte = ts_node_end_byte(self);
+  if (node.id == self.id) return ts_node__null();
+
+  TSNode last_visible_node = node;
+  bool did_descend = true;
+  while (did_descend) {
+    did_descend = false;
+
+    TSNode child;
+    NodeChildIterator iterator = ts_node_iterate_children(&node);
+    while (ts_node_child_iterator_next(&iterator, &child)) {
+      if (
+        ts_node_start_byte(child) > ts_node_start_byte(self) ||
+        child.id == self.id
+      ) break;
+      if (iterator.position.bytes >= end_byte && ts_node_child_count(child) > 0) {
+        node = child;
+        if (ts_node__is_relevant(child, true)) {
+          last_visible_node = node;
+        }
+        did_descend = true;
+        break;
+      }
+    }
+  }
+
+  return last_visible_node;
+}
+
+TSNode ts_node_child(TSNode self, uint32_t child_index) {
+  return ts_node__child(self, child_index, true);
+}
+
+TSNode ts_node_named_child(TSNode self, uint32_t child_index) {
+  return ts_node__child(self, child_index, false);
+}
+
+TSNode ts_node_child_by_field_id(TSNode self, TSFieldId field_id) {
+recur:
+  if (!field_id || ts_node_child_count(self) == 0) return ts_node__null();
+
+  const TSFieldMapEntry *field_map, *field_map_end;
+  ts_language_field_map(
+    self.tree->language,
+    ts_node__subtree(self).ptr->production_id,
+    &field_map,
+    &field_map_end
+  );
+  if (field_map == field_map_end) return ts_node__null();
+
+  // The field mappings are sorted by their field id. Scan all
+  // the mappings to find the ones for the given field id.
+  while (field_map->field_id < field_id) {
+    field_map++;
+    if (field_map == field_map_end) return ts_node__null();
+  }
+  while (field_map_end[-1].field_id > field_id) {
+    field_map_end--;
+    if (field_map == field_map_end) return ts_node__null();
+  }
+
+  TSNode child;
+  NodeChildIterator iterator = ts_node_iterate_children(&self);
+  while (ts_node_child_iterator_next(&iterator, &child)) {
+    if (!ts_subtree_extra(ts_node__subtree(child))) {
+      uint32_t index = iterator.structural_child_index - 1;
+      if (index < field_map->child_index) continue;
+
+      // Hidden nodes' fields are "inherited" by their visible parent.
+      if (field_map->inherited) {
+
+        // If this is the *last* possible child node for this field,
+        // then perform a tail call to avoid recursion.
+        if (field_map + 1 == field_map_end) {
+          self = child;
+          goto recur;
+        }
+
+        // Otherwise, descend into this child, but if it doesn't contain
+        // the field, continue searching subsequent children.
+        else {
+          TSNode result = ts_node_child_by_field_id(child, field_id);
+          if (result.id) return result;
+          field_map++;
+          if (field_map == field_map_end) return ts_node__null();
+        }
+      }
+
+      else if (ts_node__is_relevant(child, true)) {
+        return child;
+      }
+
+      // If the field refers to a hidden node with visible children,
+      // return the first visible child.
+      else if (ts_node_child_count(child) > 0 ) {
+        return ts_node_child(child, 0);
+      }
+
+      // Otherwise, continue searching subsequent children.
+      else {
+        field_map++;
+        if (field_map == field_map_end) return ts_node__null();
+      }
+    }
+  }
+
+  return ts_node__null();
+}
+
+static inline const char *ts_node__field_name_from_language(TSNode self, uint32_t structural_child_index) {
+    const TSFieldMapEntry *field_map, *field_map_end;
+    ts_language_field_map(
+      self.tree->language,
+      ts_node__subtree(self).ptr->production_id,
+      &field_map,
+      &field_map_end
+    );
+    for (; field_map != field_map_end; field_map++) {
+      if (!field_map->inherited && field_map->child_index == structural_child_index) {
+        return self.tree->language->field_names[field_map->field_id];
+      }
+    }
+    return NULL;
+}
+
+const char *ts_node_field_name_for_child(TSNode self, uint32_t child_index) {
+  TSNode result = self;
+  bool did_descend = true;
+  const char *inherited_field_name = NULL;
+
+  while (did_descend) {
+    did_descend = false;
+
+    TSNode child;
+    uint32_t index = 0;
+    NodeChildIterator iterator = ts_node_iterate_children(&result);
+    while (ts_node_child_iterator_next(&iterator, &child)) {
+      if (ts_node__is_relevant(child, true)) {
+        if (index == child_index) {
+          const char *field_name = ts_node__field_name_from_language(result, iterator.structural_child_index - 1);
+          if (field_name) return field_name;
+          return inherited_field_name;
+        }
+        index++;
+      } else {
+        uint32_t grandchild_index = child_index - index;
+        uint32_t grandchild_count = ts_node__relevant_child_count(child, true);
+        if (grandchild_index < grandchild_count) {
+          const char *field_name = ts_node__field_name_from_language(result, iterator.structural_child_index - 1);
+          if (field_name) inherited_field_name = field_name;
+
+          did_descend = true;
+          result = child;
+          child_index = grandchild_index;
+          break;
+        }
+        index += grandchild_count;
+      }
+    }
+  }
+
+  return NULL;
+}
+
+TSNode ts_node_child_by_field_name(
+  TSNode self,
+  const char *name,
+  uint32_t name_length
+) {
+  TSFieldId field_id = ts_language_field_id_for_name(
+    self.tree->language,
+    name,
+    name_length
+  );
+  return ts_node_child_by_field_id(self, field_id);
+}
+
+uint32_t ts_node_child_count(TSNode self) {
+  Subtree tree = ts_node__subtree(self);
+  if (ts_subtree_child_count(tree) > 0) {
+    return tree.ptr->visible_child_count;
+  } else {
+    return 0;
+  }
+}
+
+uint32_t ts_node_named_child_count(TSNode self) {
+  Subtree tree = ts_node__subtree(self);
+  if (ts_subtree_child_count(tree) > 0) {
+    return tree.ptr->named_child_count;
+  } else {
+    return 0;
+  }
+}
+
+TSNode ts_node_next_sibling(TSNode self) {
+  return ts_node__next_sibling(self, true);
+}
+
+TSNode ts_node_next_named_sibling(TSNode self) {
+  return ts_node__next_sibling(self, false);
+}
+
+TSNode ts_node_prev_sibling(TSNode self) {
+  return ts_node__prev_sibling(self, true);
+}
+
+TSNode ts_node_prev_named_sibling(TSNode self) {
+  return ts_node__prev_sibling(self, false);
+}
+
+TSNode ts_node_first_child_for_byte(TSNode self, uint32_t byte) {
+  return ts_node__first_child_for_byte(self, byte, true);
+}
+
+TSNode ts_node_first_named_child_for_byte(TSNode self, uint32_t byte) {
+  return ts_node__first_child_for_byte(self, byte, false);
+}
+
+TSNode ts_node_descendant_for_byte_range(
+  TSNode self,
+  uint32_t start,
+  uint32_t end
+) {
+  return ts_node__descendant_for_byte_range(self, start, end, true);
+}
+
+TSNode ts_node_named_descendant_for_byte_range(
+  TSNode self,
+  uint32_t start,
+  uint32_t end
+) {
+  return ts_node__descendant_for_byte_range(self, start, end, false);
+}
+
+TSNode ts_node_descendant_for_point_range(
+  TSNode self,
+  TSPoint start,
+  TSPoint end
+) {
+  return ts_node__descendant_for_point_range(self, start, end, true);
+}
+
+TSNode ts_node_named_descendant_for_point_range(
+  TSNode self,
+  TSPoint start,
+  TSPoint end
+) {
+  return ts_node__descendant_for_point_range(self, start, end, false);
+}
+
+void ts_node_edit(TSNode *self, const TSInputEdit *edit) {
+  uint32_t start_byte = ts_node_start_byte(*self);
+  TSPoint start_point = ts_node_start_point(*self);
+
+  if (start_byte >= edit->old_end_byte) {
+    start_byte = edit->new_end_byte + (start_byte - edit->old_end_byte);
+    start_point = point_add(edit->new_end_point, point_sub(start_point, edit->old_end_point));
+  } else if (start_byte > edit->start_byte) {
+    start_byte = edit->new_end_byte;
+    start_point = edit->new_end_point;
+  }
+
+  self->context[0] = start_byte;
+  self->context[1] = start_point.row;
+  self->context[2] = start_point.column;
+}

vendor/tree-sitter/src/parser.h

@@ -0,0 +1,265 @@
+#ifndef TREE_SITTER_PARSER_H_
+#define TREE_SITTER_PARSER_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <stdbool.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+#define ts_builtin_sym_error ((TSSymbol)-1)
+#define ts_builtin_sym_end 0
+#define TREE_SITTER_SERIALIZATION_BUFFER_SIZE 1024
+
+#ifndef TREE_SITTER_API_H_
+typedef uint16_t TSStateId;
+typedef uint16_t TSSymbol;
+typedef uint16_t TSFieldId;
+typedef struct TSLanguage TSLanguage;
+#endif
+
+typedef struct {
+  TSFieldId field_id;
+  uint8_t child_index;
+  bool inherited;
+} TSFieldMapEntry;
+
+typedef struct {
+  uint16_t index;
+  uint16_t length;
+} TSFieldMapSlice;
+
+typedef struct {
+  bool visible;
+  bool named;
+  bool supertype;
+} TSSymbolMetadata;
+
+typedef struct TSLexer TSLexer;
+
+struct TSLexer {
+  int32_t lookahead;
+  TSSymbol result_symbol;
+  void (*advance)(TSLexer *, bool);
+  void (*mark_end)(TSLexer *);
+  uint32_t (*get_column)(TSLexer *);
+  bool (*is_at_included_range_start)(const TSLexer *);
+  bool (*eof)(const TSLexer *);
+};
+
+typedef enum {
+  TSParseActionTypeShift,
+  TSParseActionTypeReduce,
+  TSParseActionTypeAccept,
+  TSParseActionTypeRecover,
+} TSParseActionType;
+
+typedef union {
+  struct {
+    uint8_t type;
+    TSStateId state;
+    bool extra;
+    bool repetition;
+  } shift;
+  struct {
+    uint8_t type;
+    uint8_t child_count;
+    TSSymbol symbol;
+    int16_t dynamic_precedence;
+    uint16_t production_id;
+  } reduce;
+  uint8_t type;
+} TSParseAction;
+
+typedef struct {
+  uint16_t lex_state;
+  uint16_t external_lex_state;
+} TSLexMode;
+
+typedef union {
+  TSParseAction action;
+  struct {
+    uint8_t count;
+    bool reusable;
+  } entry;
+} TSParseActionEntry;
+
+typedef struct {
+  int32_t start;
+  int32_t end;
+} TSCharacterRange;
+
+struct TSLanguage {
+  uint32_t version;
+  uint32_t symbol_count;
+  uint32_t alias_count;
+  uint32_t token_count;
+  uint32_t external_token_count;
+  uint32_t state_count;
+  uint32_t large_state_count;
+  uint32_t production_id_count;
+  uint32_t field_count;
+  uint16_t max_alias_sequence_length;
+  const uint16_t *parse_table;
+  const uint16_t *small_parse_table;
+  const uint32_t *small_parse_table_map;
+  const TSParseActionEntry *parse_actions;
+  const char * const *symbol_names;
+  const char * const *field_names;
+  const TSFieldMapSlice *field_map_slices;
+  const TSFieldMapEntry *field_map_entries;
+  const TSSymbolMetadata *symbol_metadata;
+  const TSSymbol *public_symbol_map;
+  const uint16_t *alias_map;
+  const TSSymbol *alias_sequences;
+  const TSLexMode *lex_modes;
+  bool (*lex_fn)(TSLexer *, TSStateId);
+  bool (*keyword_lex_fn)(TSLexer *, TSStateId);
+  TSSymbol keyword_capture_token;
+  struct {
+    const bool *states;
+    const TSSymbol *symbol_map;
+    void *(*create)(void);
+    void (*destroy)(void *);
+    bool (*scan)(void *, TSLexer *, const bool *symbol_whitelist);
+    unsigned (*serialize)(void *, char *);
+    void (*deserialize)(void *, const char *, unsigned);
+  } external_scanner;
+  const TSStateId *primary_state_ids;
+};
+
+static inline bool set_contains(TSCharacterRange *ranges, uint32_t len, int32_t lookahead) {
+  uint32_t index = 0;
+  uint32_t size = len - index;
+  while (size > 1) {
+    uint32_t half_size = size / 2;
+    uint32_t mid_index = index + half_size;
+    TSCharacterRange *range = &ranges[mid_index];
+    if (lookahead >= range->start && lookahead <= range->end) {
+      return true;
+    } else if (lookahead > range->end) {
+      index = mid_index;
+    }
+    size -= half_size;
+  }
+  TSCharacterRange *range = &ranges[index];
+  return (lookahead >= range->start && lookahead <= range->end);
+}
+
+/*
+ *  Lexer Macros
+ */
+
+#ifdef _MSC_VER
+#define UNUSED __pragma(warning(suppress : 4101))
+#else
+#define UNUSED __attribute__((unused))
+#endif
+
+#define START_LEXER()           \
+  bool result = false;          \
+  bool skip = false;            \
+  UNUSED                        \
+  bool eof = false;             \
+  int32_t lookahead;            \
+  goto start;                   \
+  next_state:                   \
+  lexer->advance(lexer, skip);  \
+  start:                        \
+  skip = false;                 \
+  lookahead = lexer->lookahead;
+
+#define ADVANCE(state_value) \
+  {                          \
+    state = state_value;     \
+    goto next_state;         \
+  }
+
+#define ADVANCE_MAP(...)                                              \
+  {                                                                   \
+    static const uint16_t map[] = { __VA_ARGS__ };                    \
+    for (uint32_t i = 0; i < sizeof(map) / sizeof(map[0]); i += 2) {  \
+      if (map[i] == lookahead) {                                      \
+        state = map[i + 1];                                           \
+        goto next_state;                                              \
+      }                                                               \
+    }                                                                 \
+  }
+
+#define SKIP(state_value) \
+  {                       \
+    skip = true;          \
+    state = state_value;  \
+    goto next_state;      \
+  }
+
+#define ACCEPT_TOKEN(symbol_value)     \
+  result = true;                       \
+  lexer->result_symbol = symbol_value; \
+  lexer->mark_end(lexer);
+
+#define END_STATE() return result;
+
+/*
+ *  Parse Table Macros
+ */
+
+#define SMALL_STATE(id) ((id) - LARGE_STATE_COUNT)
+
+#define STATE(id) id
+
+#define ACTIONS(id) id
+
+#define SHIFT(state_value)            \
+  {{                                  \
+    .shift = {                        \
+      .type = TSParseActionTypeShift, \
+      .state = (state_value)          \
+    }                                 \
+  }}
+
+#define SHIFT_REPEAT(state_value)     \
+  {{                                  \
+    .shift = {                        \
+      .type = TSParseActionTypeShift, \
+      .state = (state_value),         \
+      .repetition = true              \
+    }                                 \
+  }}
+
+#define SHIFT_EXTRA()                 \
+  {{                                  \
+    .shift = {                        \
+      .type = TSParseActionTypeShift, \
+      .extra = true                   \
+    }                                 \
+  }}
+
+#define REDUCE(symbol_name, children, precedence, prod_id) \
+  {{                                                       \
+    .reduce = {                                            \
+      .type = TSParseActionTypeReduce,                     \
+      .symbol = symbol_name,                               \
+      .child_count = children,                             \
+      .dynamic_precedence = precedence,                    \
+      .production_id = prod_id                             \
+    },                                                     \
+  }}
+
+#define RECOVER()                    \
+  {{                                 \
+    .type = TSParseActionTypeRecover \
+  }}
+
+#define ACCEPT_INPUT()              \
+  {{                                \
+    .type = TSParseActionTypeAccept \
+  }}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TREE_SITTER_PARSER_H_

vendor/tree-sitter/src/point.h

@@ -0,0 +1,62 @@
+#ifndef TREE_SITTER_POINT_H_
+#define TREE_SITTER_POINT_H_
+
+#include "tree_sitter/api.h"
+
+#define POINT_ZERO ((TSPoint) {0, 0})
+#define POINT_MAX ((TSPoint) {UINT32_MAX, UINT32_MAX})
+
+static inline TSPoint point__new(unsigned row, unsigned column) {
+  TSPoint result = {row, column};
+  return result;
+}
+
+static inline TSPoint point_add(TSPoint a, TSPoint b) {
+  if (b.row > 0)
+    return point__new(a.row + b.row, b.column);
+  else
+    return point__new(a.row, a.column + b.column);
+}
+
+static inline TSPoint point_sub(TSPoint a, TSPoint b) {
+  if (a.row > b.row)
+    return point__new(a.row - b.row, a.column);
+  else
+    return point__new(0, a.column - b.column);
+}
+
+static inline bool point_lte(TSPoint a, TSPoint b) {
+  return (a.row < b.row) || (a.row == b.row && a.column <= b.column);
+}
+
+static inline bool point_lt(TSPoint a, TSPoint b) {
+  return (a.row < b.row) || (a.row == b.row && a.column < b.column);
+}
+
+static inline bool point_gt(TSPoint a, TSPoint b) {
+  return (a.row > b.row) || (a.row == b.row && a.column > b.column);
+}
+
+static inline bool point_gte(TSPoint a, TSPoint b) {
+  return (a.row > b.row) || (a.row == b.row && a.column >= b.column);
+}
+
+static inline bool point_eq(TSPoint a, TSPoint b) {
+  return a.row == b.row && a.column == b.column;
+}
+
+static inline TSPoint point_min(TSPoint a, TSPoint b) {
+  if (a.row < b.row || (a.row == b.row && a.column < b.column))
+    return a;
+  else
+    return b;
+}
+
+static inline TSPoint point_max(TSPoint a, TSPoint b) {
+  if (a.row > b.row || (a.row == b.row && a.column > b.column))
+    return a;
+  else
+    return b;
+}
+
+#endif

vendor/tree-sitter/src/reduce_action.h

@@ -0,0 +1,34 @@
+#ifndef TREE_SITTER_REDUCE_ACTION_H_
+#define TREE_SITTER_REDUCE_ACTION_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "./array.h"
+#include "tree_sitter/api.h"
+
+typedef struct {
+  uint32_t count;
+  TSSymbol symbol;
+  int dynamic_precedence;
+  unsigned short production_id;
+} ReduceAction;
+
+typedef Array(ReduceAction) ReduceActionSet;
+
+static inline void ts_reduce_action_set_add(ReduceActionSet *self,
+                                            ReduceAction new_action) {
+  for (uint32_t i = 0; i < self->size; i++) {
+    ReduceAction action = self->contents[i];
+    if (action.symbol == new_action.symbol && action.count == new_action.count)
+      return;
+  }
+  array_push(self, new_action);
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TREE_SITTER_REDUCE_ACTION_H_

vendor/tree-sitter/src/reusable_node.h

@@ -0,0 +1,95 @@
+#include "./subtree.h"
+
+typedef struct {
+  Subtree tree;
+  uint32_t child_index;
+  uint32_t byte_offset;
+} StackEntry;
+
+typedef struct {
+  Array(StackEntry) stack;
+  Subtree last_external_token;
+} ReusableNode;
+
+static inline ReusableNode reusable_node_new(void) {
+  return (ReusableNode) {array_new(), NULL_SUBTREE};
+}
+
+static inline void reusable_node_clear(ReusableNode *self) {
+  array_clear(&self->stack);
+  self->last_external_token = NULL_SUBTREE;
+}
+
+static inline Subtree reusable_node_tree(ReusableNode *self) {
+  return self->stack.size > 0
+    ? self->stack.contents[self->stack.size - 1].tree
+    : NULL_SUBTREE;
+}
+
+static inline uint32_t reusable_node_byte_offset(ReusableNode *self) {
+  return self->stack.size > 0
+    ? self->stack.contents[self->stack.size - 1].byte_offset
+    : UINT32_MAX;
+}
+
+static inline void reusable_node_delete(ReusableNode *self) {
+  array_delete(&self->stack);
+}
+
+static inline void reusable_node_advance(ReusableNode *self) {
+  StackEntry last_entry = *array_back(&self->stack);
+  uint32_t byte_offset = last_entry.byte_offset + ts_subtree_total_bytes(last_entry.tree);
+  if (ts_subtree_has_external_tokens(last_entry.tree)) {
+    self->last_external_token = ts_subtree_last_external_token(last_entry.tree);
+  }
+
+  Subtree tree;
+  uint32_t next_index;
+  do {
+    StackEntry popped_entry = array_pop(&self->stack);
+    next_index = popped_entry.child_index + 1;
+    if (self->stack.size == 0) return;
+    tree = array_back(&self->stack)->tree;
+  } while (ts_subtree_child_count(tree) <= next_index);
+
+  array_push(&self->stack, ((StackEntry) {
+    .tree = ts_subtree_children(tree)[next_index],
+    .child_index = next_index,
+    .byte_offset = byte_offset,
+  }));
+}
+
+static inline bool reusable_node_descend(ReusableNode *self) {
+  StackEntry last_entry = *array_back(&self->stack);
+  if (ts_subtree_child_count(last_entry.tree) > 0) {
+    array_push(&self->stack, ((StackEntry) {
+      .tree = ts_subtree_children(last_entry.tree)[0],
+      .child_index = 0,
+      .byte_offset = last_entry.byte_offset,
+    }));
+    return true;
+  } else {
+    return false;
+  }
+}
+
+static inline void reusable_node_advance_past_leaf(ReusableNode *self) {
+  while (reusable_node_descend(self)) {}
+  reusable_node_advance(self);
+}
+
+static inline void reusable_node_reset(ReusableNode *self, Subtree tree) {
+  reusable_node_clear(self);
+  array_push(&self->stack, ((StackEntry) {
+    .tree = tree,
+    .child_index = 0,
+    .byte_offset = 0,
+  }));
+
+  // Never reuse the root node, because it has a non-standard internal structure
+  // due to transformations that are applied when it is accepted: adding the EOF
+  // child and any extra children.
+  if (!reusable_node_descend(self)) {
+    reusable_node_clear(self);
+  }
+}

vendor/tree-sitter/src/stack.c

@@ -0,0 +1,899 @@
+#include "./alloc.h"
+#include "./language.h"
+#include "./subtree.h"
+#include "./array.h"
+#include "./stack.h"
+#include "./length.h"
+#include <assert.h>
+#include <inttypes.h>
+#include <stdio.h>
+
+#define MAX_LINK_COUNT 8
+#define MAX_NODE_POOL_SIZE 50
+#define MAX_ITERATOR_COUNT 64
+
+#if defined _WIN32 && !defined __GNUC__
+#define forceinline __forceinline
+#else
+#define forceinline static inline __attribute__((always_inline))
+#endif
+
+typedef struct StackNode StackNode;
+
+typedef struct {
+  StackNode *node;
+  Subtree subtree;
+  bool is_pending;
+} StackLink;
+
+struct StackNode {
+  TSStateId state;
+  Length position;
+  StackLink links[MAX_LINK_COUNT];
+  short unsigned int link_count;
+  uint32_t ref_count;
+  unsigned error_cost;
+  unsigned node_count;
+  int dynamic_precedence;
+};
+
+typedef struct {
+  StackNode *node;
+  SubtreeArray subtrees;
+  uint32_t subtree_count;
+  bool is_pending;
+} StackIterator;
+
+typedef Array(StackNode *) StackNodeArray;
+
+typedef enum {
+  StackStatusActive,
+  StackStatusPaused,
+  StackStatusHalted,
+} StackStatus;
+
+typedef struct {
+  StackNode *node;
+  StackSummary *summary;
+  unsigned node_count_at_last_error;
+  Subtree last_external_token;
+  Subtree lookahead_when_paused;
+  StackStatus status;
+} StackHead;
+
+struct Stack {
+  Array(StackHead) heads;
+  StackSliceArray slices;
+  Array(StackIterator) iterators;
+  StackNodeArray node_pool;
+  StackNode *base_node;
+  SubtreePool *subtree_pool;
+};
+
+typedef unsigned StackAction;
+enum {
+  StackActionNone,
+  StackActionStop = 1,
+  StackActionPop = 2,
+};
+
+typedef StackAction (*StackCallback)(void *, const StackIterator *);
+
+static void stack_node_retain(StackNode *self) {
+  if (!self)
+    return;
+  assert(self->ref_count > 0);
+  self->ref_count++;
+  assert(self->ref_count != 0);
+}
+
+static void stack_node_release(
+  StackNode *self,
+  StackNodeArray *pool,
+  SubtreePool *subtree_pool
+) {
+recur:
+  assert(self->ref_count != 0);
+  self->ref_count--;
+  if (self->ref_count > 0) return;
+
+  StackNode *first_predecessor = NULL;
+  if (self->link_count > 0) {
+    for (unsigned i = self->link_count - 1; i > 0; i--) {
+      StackLink link = self->links[i];
+      if (link.subtree.ptr) ts_subtree_release(subtree_pool, link.subtree);
+      stack_node_release(link.node, pool, subtree_pool);
+    }
+    StackLink link = self->links[0];
+    if (link.subtree.ptr) ts_subtree_release(subtree_pool, link.subtree);
+    first_predecessor = self->links[0].node;
+  }
+
+  if (pool->size < MAX_NODE_POOL_SIZE) {
+    array_push(pool, self);
+  } else {
+    ts_free(self);
+  }
+
+  if (first_predecessor) {
+    self = first_predecessor;
+    goto recur;
+  }
+}
+
+/// Get the number of nodes in the subtree, for the purpose of measuring
+/// how much progress has been made by a given version of the stack.
+static uint32_t stack__subtree_node_count(Subtree subtree) {
+  uint32_t count = ts_subtree_visible_descendant_count(subtree);
+  if (ts_subtree_visible(subtree)) count++;
+
+  // Count intermediate error nodes even though they are not visible,
+  // because a stack version's node count is used to check whether it
+  // has made any progress since the last time it encountered an error.
+  if (ts_subtree_symbol(subtree) == ts_builtin_sym_error_repeat) count++;
+
+  return count;
+}
+
+static StackNode *stack_node_new(
+  StackNode *previous_node,
+  Subtree subtree,
+  bool is_pending,
+  TSStateId state,
+  StackNodeArray *pool
+) {
+  StackNode *node = pool->size > 0
+    ? array_pop(pool)
+    : ts_malloc(sizeof(StackNode));
+  *node = (StackNode) {
+    .ref_count = 1,
+    .link_count = 0,
+    .state = state
+  };
+
+  if (previous_node) {
+    node->link_count = 1;
+    node->links[0] = (StackLink) {
+      .node = previous_node,
+      .subtree = subtree,
+      .is_pending = is_pending,
+    };
+
+    node->position = previous_node->position;
+    node->error_cost = previous_node->error_cost;
+    node->dynamic_precedence = previous_node->dynamic_precedence;
+    node->node_count = previous_node->node_count;
+
+    if (subtree.ptr) {
+      node->error_cost += ts_subtree_error_cost(subtree);
+      node->position = length_add(node->position, ts_subtree_total_size(subtree));
+      node->node_count += stack__subtree_node_count(subtree);
+      node->dynamic_precedence += ts_subtree_dynamic_precedence(subtree);
+    }
+  } else {
+    node->position = length_zero();
+    node->error_cost = 0;
+  }
+
+  return node;
+}
+
+static bool stack__subtree_is_equivalent(Subtree left, Subtree right) {
+  if (left.ptr == right.ptr) return true;
+  if (!left.ptr || !right.ptr) return false;
+
+  // Symbols must match
+  if (ts_subtree_symbol(left) != ts_subtree_symbol(right)) return false;
+
+  // If both have errors, don't bother keeping both.
+  if (ts_subtree_error_cost(left) > 0 && ts_subtree_error_cost(right) > 0) return true;
+
+  return (
+    ts_subtree_padding(left).bytes == ts_subtree_padding(right).bytes &&
+    ts_subtree_size(left).bytes == ts_subtree_size(right).bytes &&
+    ts_subtree_child_count(left) == ts_subtree_child_count(right) &&
+    ts_subtree_extra(left) == ts_subtree_extra(right) &&
+    ts_subtree_external_scanner_state_eq(left, right)
+  );
+}
+
+static void stack_node_add_link(
+  StackNode *self,
+  StackLink link,
+  SubtreePool *subtree_pool
+) {
+  if (link.node == self) return;
+
+  for (int i = 0; i < self->link_count; i++) {
+    StackLink *existing_link = &self->links[i];
+    if (stack__subtree_is_equivalent(existing_link->subtree, link.subtree)) {
+      // In general, we preserve ambiguities until they are removed from the stack
+      // during a pop operation where multiple paths lead to the same node. But in
+      // the special case where two links directly connect the same pair of nodes,
+      // we can safely remove the ambiguity ahead of time without changing behavior.
+      if (existing_link->node == link.node) {
+        if (
+          ts_subtree_dynamic_precedence(link.subtree) >
+          ts_subtree_dynamic_precedence(existing_link->subtree)
+        ) {
+          ts_subtree_retain(link.subtree);
+          ts_subtree_release(subtree_pool, existing_link->subtree);
+          existing_link->subtree = link.subtree;
+          self->dynamic_precedence =
+            link.node->dynamic_precedence + ts_subtree_dynamic_precedence(link.subtree);
+        }
+        return;
+      }
+
+      // If the previous nodes are mergeable, merge them recursively.
+      if (
+        existing_link->node->state == link.node->state &&
+        existing_link->node->position.bytes == link.node->position.bytes &&
+        existing_link->node->error_cost == link.node->error_cost
+      ) {
+        for (int j = 0; j < link.node->link_count; j++) {
+          stack_node_add_link(existing_link->node, link.node->links[j], subtree_pool);
+        }
+        int32_t dynamic_precedence = link.node->dynamic_precedence;
+        if (link.subtree.ptr) {
+          dynamic_precedence += ts_subtree_dynamic_precedence(link.subtree);
+        }
+        if (dynamic_precedence > self->dynamic_precedence) {
+          self->dynamic_precedence = dynamic_precedence;
+        }
+        return;
+      }
+    }
+  }
+
+  if (self->link_count == MAX_LINK_COUNT) return;
+
+  stack_node_retain(link.node);
+  unsigned node_count = link.node->node_count;
+  int dynamic_precedence = link.node->dynamic_precedence;
+  self->links[self->link_count++] = link;
+
+  if (link.subtree.ptr) {
+    ts_subtree_retain(link.subtree);
+    node_count += stack__subtree_node_count(link.subtree);
+    dynamic_precedence += ts_subtree_dynamic_precedence(link.subtree);
+  }
+
+  if (node_count > self->node_count) self->node_count = node_count;
+  if (dynamic_precedence > self->dynamic_precedence) self->dynamic_precedence = dynamic_precedence;
+}
+
+static void stack_head_delete(
+  StackHead *self,
+  StackNodeArray *pool,
+  SubtreePool *subtree_pool
+) {
+  if (self->node) {
+    if (self->last_external_token.ptr) {
+      ts_subtree_release(subtree_pool, self->last_external_token);
+    }
+    if (self->lookahead_when_paused.ptr) {
+      ts_subtree_release(subtree_pool, self->lookahead_when_paused);
+    }
+    if (self->summary) {
+      array_delete(self->summary);
+      ts_free(self->summary);
+    }
+    stack_node_release(self->node, pool, subtree_pool);
+  }
+}
+
+static StackVersion ts_stack__add_version(
+  Stack *self,
+  StackVersion original_version,
+  StackNode *node
+) {
+  StackHead head = {
+    .node = node,
+    .node_count_at_last_error = self->heads.contents[original_version].node_count_at_last_error,
+    .last_external_token = self->heads.contents[original_version].last_external_token,
+    .status = StackStatusActive,
+    .lookahead_when_paused = NULL_SUBTREE,
+  };
+  array_push(&self->heads, head);
+  stack_node_retain(node);
+  if (head.last_external_token.ptr) ts_subtree_retain(head.last_external_token);
+  return (StackVersion)(self->heads.size - 1);
+}
+
+static void ts_stack__add_slice(
+  Stack *self,
+  StackVersion original_version,
+  StackNode *node,
+  SubtreeArray *subtrees
+) {
+  for (uint32_t i = self->slices.size - 1; i + 1 > 0; i--) {
+    StackVersion version = self->slices.contents[i].version;
+    if (self->heads.contents[version].node == node) {
+      StackSlice slice = {*subtrees, version};
+      array_insert(&self->slices, i + 1, slice);
+      return;
+    }
+  }
+
+  StackVersion version = ts_stack__add_version(self, original_version, node);
+  StackSlice slice = { *subtrees, version };
+  array_push(&self->slices, slice);
+}
+
+static StackSliceArray stack__iter(
+  Stack *self,
+  StackVersion version,
+  StackCallback callback,
+  void *payload,
+  int goal_subtree_count
+) {
+  array_clear(&self->slices);
+  array_clear(&self->iterators);
+
+  StackHead *head = array_get(&self->heads, version);
+  StackIterator new_iterator = {
+    .node = head->node,
+    .subtrees = array_new(),
+    .subtree_count = 0,
+    .is_pending = true,
+  };
+
+  bool include_subtrees = false;
+  if (goal_subtree_count >= 0) {
+    include_subtrees = true;
+    array_reserve(&new_iterator.subtrees, (uint32_t)ts_subtree_alloc_size(goal_subtree_count) / sizeof(Subtree));
+  }
+
+  array_push(&self->iterators, new_iterator);
+
+  while (self->iterators.size > 0) {
+    for (uint32_t i = 0, size = self->iterators.size; i < size; i++) {
+      StackIterator *iterator = &self->iterators.contents[i];
+      StackNode *node = iterator->node;
+
+      StackAction action = callback(payload, iterator);
+      bool should_pop = action & StackActionPop;
+      bool should_stop = action & StackActionStop || node->link_count == 0;
+
+      if (should_pop) {
+        SubtreeArray subtrees = iterator->subtrees;
+        if (!should_stop) {
+          ts_subtree_array_copy(subtrees, &subtrees);
+        }
+        ts_subtree_array_reverse(&subtrees);
+        ts_stack__add_slice(
+          self,
+          version,
+          node,
+          &subtrees
+        );
+      }
+
+      if (should_stop) {
+        if (!should_pop) {
+          ts_subtree_array_delete(self->subtree_pool, &iterator->subtrees);
+        }
+        array_erase(&self->iterators, i);
+        i--, size--;
+        continue;
+      }
+
+      for (uint32_t j = 1; j <= node->link_count; j++) {
+        StackIterator *next_iterator;
+        StackLink link;
+        if (j == node->link_count) {
+          link = node->links[0];
+          next_iterator = &self->iterators.contents[i];
+        } else {
+          if (self->iterators.size >= MAX_ITERATOR_COUNT) continue;
+          link = node->links[j];
+          StackIterator current_iterator = self->iterators.contents[i];
+          array_push(&self->iterators, current_iterator);
+          next_iterator = array_back(&self->iterators);
+          ts_subtree_array_copy(next_iterator->subtrees, &next_iterator->subtrees);
+        }
+
+        next_iterator->node = link.node;
+        if (link.subtree.ptr) {
+          if (include_subtrees) {
+            array_push(&next_iterator->subtrees, link.subtree);
+            ts_subtree_retain(link.subtree);
+          }
+
+          if (!ts_subtree_extra(link.subtree)) {
+            next_iterator->subtree_count++;
+            if (!link.is_pending) {
+              next_iterator->is_pending = false;
+            }
+          }
+        } else {
+          next_iterator->subtree_count++;
+          next_iterator->is_pending = false;
+        }
+      }
+    }
+  }
+
+  return self->slices;
+}
+
+Stack *ts_stack_new(SubtreePool *subtree_pool) {
+  Stack *self = ts_calloc(1, sizeof(Stack));
+
+  array_init(&self->heads);
+  array_init(&self->slices);
+  array_init(&self->iterators);
+  array_init(&self->node_pool);
+  array_reserve(&self->heads, 4);
+  array_reserve(&self->slices, 4);
+  array_reserve(&self->iterators, 4);
+  array_reserve(&self->node_pool, MAX_NODE_POOL_SIZE);
+
+  self->subtree_pool = subtree_pool;
+  self->base_node = stack_node_new(NULL, NULL_SUBTREE, false, 1, &self->node_pool);
+  ts_stack_clear(self);
+
+  return self;
+}
+
+void ts_stack_delete(Stack *self) {
+  if (self->slices.contents)
+    array_delete(&self->slices);
+  if (self->iterators.contents)
+    array_delete(&self->iterators);
+  stack_node_release(self->base_node, &self->node_pool, self->subtree_pool);
+  for (uint32_t i = 0; i < self->heads.size; i++) {
+    stack_head_delete(&self->heads.contents[i], &self->node_pool, self->subtree_pool);
+  }
+  array_clear(&self->heads);
+  if (self->node_pool.contents) {
+    for (uint32_t i = 0; i < self->node_pool.size; i++)
+      ts_free(self->node_pool.contents[i]);
+    array_delete(&self->node_pool);
+  }
+  array_delete(&self->heads);
+  ts_free(self);
+}
+
+uint32_t ts_stack_version_count(const Stack *self) {
+  return self->heads.size;
+}
+
+TSStateId ts_stack_state(const Stack *self, StackVersion version) {
+  return array_get(&self->heads, version)->node->state;
+}
+
+Length ts_stack_position(const Stack *self, StackVersion version) {
+  return array_get(&self->heads, version)->node->position;
+}
+
+Subtree ts_stack_last_external_token(const Stack *self, StackVersion version) {
+  return array_get(&self->heads, version)->last_external_token;
+}
+
+void ts_stack_set_last_external_token(Stack *self, StackVersion version, Subtree token) {
+  StackHead *head = array_get(&self->heads, version);
+  if (token.ptr) ts_subtree_retain(token);
+  if (head->last_external_token.ptr) ts_subtree_release(self->subtree_pool, head->last_external_token);
+  head->last_external_token = token;
+}
+
+unsigned ts_stack_error_cost(const Stack *self, StackVersion version) {
+  StackHead *head = array_get(&self->heads, version);
+  unsigned result = head->node->error_cost;
+  if (
+    head->status == StackStatusPaused ||
+    (head->node->state == ERROR_STATE && !head->node->links[0].subtree.ptr)) {
+    result += ERROR_COST_PER_RECOVERY;
+  }
+  return result;
+}
+
+unsigned ts_stack_node_count_since_error(const Stack *self, StackVersion version) {
+  StackHead *head = array_get(&self->heads, version);
+  if (head->node->node_count < head->node_count_at_last_error) {
+    head->node_count_at_last_error = head->node->node_count;
+  }
+  return head->node->node_count - head->node_count_at_last_error;
+}
+
+void ts_stack_push(
+  Stack *self,
+  StackVersion version,
+  Subtree subtree,
+  bool pending,
+  TSStateId state
+) {
+  StackHead *head = array_get(&self->heads, version);
+  StackNode *new_node = stack_node_new(head->node, subtree, pending, state, &self->node_pool);
+  if (!subtree.ptr) head->node_count_at_last_error = new_node->node_count;
+  head->node = new_node;
+}
+
+forceinline StackAction pop_count_callback(void *payload, const StackIterator *iterator) {
+  unsigned *goal_subtree_count = payload;
+  if (iterator->subtree_count == *goal_subtree_count) {
+    return StackActionPop | StackActionStop;
+  } else {
+    return StackActionNone;
+  }
+}
+
+StackSliceArray ts_stack_pop_count(Stack *self, StackVersion version, uint32_t count) {
+  return stack__iter(self, version, pop_count_callback, &count, (int)count);
+}
+
+forceinline StackAction pop_pending_callback(void *payload, const StackIterator *iterator) {
+  (void)payload;
+  if (iterator->subtree_count >= 1) {
+    if (iterator->is_pending) {
+      return StackActionPop | StackActionStop;
+    } else {
+      return StackActionStop;
+    }
+  } else {
+    return StackActionNone;
+  }
+}
+
+StackSliceArray ts_stack_pop_pending(Stack *self, StackVersion version) {
+  StackSliceArray pop = stack__iter(self, version, pop_pending_callback, NULL, 0);
+  if (pop.size > 0) {
+    ts_stack_renumber_version(self, pop.contents[0].version, version);
+    pop.contents[0].version = version;
+  }
+  return pop;
+}
+
+forceinline StackAction pop_error_callback(void *payload, const StackIterator *iterator) {
+  if (iterator->subtrees.size > 0) {
+    bool *found_error = payload;
+    if (!*found_error && ts_subtree_is_error(iterator->subtrees.contents[0])) {
+      *found_error = true;
+      return StackActionPop | StackActionStop;
+    } else {
+      return StackActionStop;
+    }
+  } else {
+    return StackActionNone;
+  }
+}
+
+SubtreeArray ts_stack_pop_error(Stack *self, StackVersion version) {
+  StackNode *node = array_get(&self->heads, version)->node;
+  for (unsigned i = 0; i < node->link_count; i++) {
+    if (node->links[i].subtree.ptr && ts_subtree_is_error(node->links[i].subtree)) {
+      bool found_error = false;
+      StackSliceArray pop = stack__iter(self, version, pop_error_callback, &found_error, 1);
+      if (pop.size > 0) {
+        assert(pop.size == 1);
+        ts_stack_renumber_version(self, pop.contents[0].version, version);
+        return pop.contents[0].subtrees;
+      }
+      break;
+    }
+  }
+  return (SubtreeArray) {.size = 0};
+}
+
+forceinline StackAction pop_all_callback(void *payload, const StackIterator *iterator) {
+  (void)payload;
+  return iterator->node->link_count == 0 ? StackActionPop : StackActionNone;
+}
+
+StackSliceArray ts_stack_pop_all(Stack *self, StackVersion version) {
+  return stack__iter(self, version, pop_all_callback, NULL, 0);
+}
+
+typedef struct {
+  StackSummary *summary;
+  unsigned max_depth;
+} SummarizeStackSession;
+
+forceinline StackAction summarize_stack_callback(void *payload, const StackIterator *iterator) {
+  SummarizeStackSession *session = payload;
+  TSStateId state = iterator->node->state;
+  unsigned depth = iterator->subtree_count;
+  if (depth > session->max_depth) return StackActionStop;
+  for (unsigned i = session->summary->size - 1; i + 1 > 0; i--) {
+    StackSummaryEntry entry = session->summary->contents[i];
+    if (entry.depth < depth) break;
+    if (entry.depth == depth && entry.state == state) return StackActionNone;
+  }
+  array_push(session->summary, ((StackSummaryEntry) {
+    .position = iterator->node->position,
+    .depth = depth,
+    .state = state,
+  }));
+  return StackActionNone;
+}
+
+void ts_stack_record_summary(Stack *self, StackVersion version, unsigned max_depth) {
+  SummarizeStackSession session = {
+    .summary = ts_malloc(sizeof(StackSummary)),
+    .max_depth = max_depth
+  };
+  array_init(session.summary);
+  stack__iter(self, version, summarize_stack_callback, &session, -1);
+  StackHead *head = &self->heads.contents[version];
+  if (head->summary) {
+    array_delete(head->summary);
+    ts_free(head->summary);
+  }
+  head->summary = session.summary;
+}
+
+StackSummary *ts_stack_get_summary(Stack *self, StackVersion version) {
+  return array_get(&self->heads, version)->summary;
+}
+
+int ts_stack_dynamic_precedence(Stack *self, StackVersion version) {
+  return array_get(&self->heads, version)->node->dynamic_precedence;
+}
+
+bool ts_stack_has_advanced_since_error(const Stack *self, StackVersion version) {
+  const StackHead *head = array_get(&self->heads, version);
+  const StackNode *node = head->node;
+  if (node->error_cost == 0) return true;
+  while (node) {
+    if (node->link_count > 0) {
+      Subtree subtree = node->links[0].subtree;
+      if (subtree.ptr) {
+        if (ts_subtree_total_bytes(subtree) > 0) {
+          return true;
+        } else if (
+          node->node_count > head->node_count_at_last_error &&
+          ts_subtree_error_cost(subtree) == 0
+        ) {
+          node = node->links[0].node;
+          continue;
+        }
+      }
+    }
+    break;
+  }
+  return false;
+}
+
+void ts_stack_remove_version(Stack *self, StackVersion version) {
+  stack_head_delete(array_get(&self->heads, version), &self->node_pool, self->subtree_pool);
+  array_erase(&self->heads, version);
+}
+
+void ts_stack_renumber_version(Stack *self, StackVersion v1, StackVersion v2) {
+  if (v1 == v2) return;
+  assert(v2 < v1);
+  assert((uint32_t)v1 < self->heads.size);
+  StackHead *source_head = &self->heads.contents[v1];
+  StackHead *target_head = &self->heads.contents[v2];
+  if (target_head->summary && !source_head->summary) {
+    source_head->summary = target_head->summary;
+    target_head->summary = NULL;
+  }
+  stack_head_delete(target_head, &self->node_pool, self->subtree_pool);
+  *target_head = *source_head;
+  array_erase(&self->heads, v1);
+}
+
+void ts_stack_swap_versions(Stack *self, StackVersion v1, StackVersion v2) {
+  StackHead temporary_head = self->heads.contents[v1];
+  self->heads.contents[v1] = self->heads.contents[v2];
+  self->heads.contents[v2] = temporary_head;
+}
+
+StackVersion ts_stack_copy_version(Stack *self, StackVersion version) {
+  assert(version < self->heads.size);
+  array_push(&self->heads, self->heads.contents[version]);
+  StackHead *head = array_back(&self->heads);
+  stack_node_retain(head->node);
+  if (head->last_external_token.ptr) ts_subtree_retain(head->last_external_token);
+  head->summary = NULL;
+  return self->heads.size - 1;
+}
+
+bool ts_stack_merge(Stack *self, StackVersion version1, StackVersion version2) {
+  if (!ts_stack_can_merge(self, version1, version2)) return false;
+  StackHead *head1 = &self->heads.contents[version1];
+  StackHead *head2 = &self->heads.contents[version2];
+  for (uint32_t i = 0; i < head2->node->link_count; i++) {
+    stack_node_add_link(head1->node, head2->node->links[i], self->subtree_pool);
+  }
+  if (head1->node->state == ERROR_STATE) {
+    head1->node_count_at_last_error = head1->node->node_count;
+  }
+  ts_stack_remove_version(self, version2);
+  return true;
+}
+
+bool ts_stack_can_merge(Stack *self, StackVersion version1, StackVersion version2) {
+  StackHead *head1 = &self->heads.contents[version1];
+  StackHead *head2 = &self->heads.contents[version2];
+  return
+    head1->status == StackStatusActive &&
+    head2->status == StackStatusActive &&
+    head1->node->state == head2->node->state &&
+    head1->node->position.bytes == head2->node->position.bytes &&
+    head1->node->error_cost == head2->node->error_cost &&
+    ts_subtree_external_scanner_state_eq(head1->last_external_token, head2->last_external_token);
+}
+
+void ts_stack_halt(Stack *self, StackVersion version) {
+  array_get(&self->heads, version)->status = StackStatusHalted;
+}
+
+void ts_stack_pause(Stack *self, StackVersion version, Subtree lookahead) {
+  StackHead *head = array_get(&self->heads, version);
+  head->status = StackStatusPaused;
+  head->lookahead_when_paused = lookahead;
+  head->node_count_at_last_error = head->node->node_count;
+}
+
+bool ts_stack_is_active(const Stack *self, StackVersion version) {
+  return array_get(&self->heads, version)->status == StackStatusActive;
+}
+
+bool ts_stack_is_halted(const Stack *self, StackVersion version) {
+  return array_get(&self->heads, version)->status == StackStatusHalted;
+}
+
+bool ts_stack_is_paused(const Stack *self, StackVersion version) {
+  return array_get(&self->heads, version)->status == StackStatusPaused;
+}
+
+Subtree ts_stack_resume(Stack *self, StackVersion version) {
+  StackHead *head = array_get(&self->heads, version);
+  assert(head->status == StackStatusPaused);
+  Subtree result = head->lookahead_when_paused;
+  head->status = StackStatusActive;
+  head->lookahead_when_paused = NULL_SUBTREE;
+  return result;
+}
+
+void ts_stack_clear(Stack *self) {
+  stack_node_retain(self->base_node);
+  for (uint32_t i = 0; i < self->heads.size; i++) {
+    stack_head_delete(&self->heads.contents[i], &self->node_pool, self->subtree_pool);
+  }
+  array_clear(&self->heads);
+  array_push(&self->heads, ((StackHead) {
+    .node = self->base_node,
+    .status = StackStatusActive,
+    .last_external_token = NULL_SUBTREE,
+    .lookahead_when_paused = NULL_SUBTREE,
+  }));
+}
+
+bool ts_stack_print_dot_graph(Stack *self, const TSLanguage *language, FILE *f) {
+  array_reserve(&self->iterators, 32);
+  if (!f) f = stderr;
+
+  fprintf(f, "digraph stack {\n");
+  fprintf(f, "rankdir=\"RL\";\n");
+  fprintf(f, "edge [arrowhead=none]\n");
+
+  Array(StackNode *) visited_nodes = array_new();
+
+  array_clear(&self->iterators);
+  for (uint32_t i = 0; i < self->heads.size; i++) {
+    StackHead *head = &self->heads.contents[i];
+    if (head->status == StackStatusHalted) continue;
+
+    fprintf(f, "node_head_%u [shape=none, label=\"\"]\n", i);
+    fprintf(f, "node_head_%u -> node_%p [", i, (void *)head->node);
+
+    if (head->status == StackStatusPaused) {
+      fprintf(f, "color=red ");
+    }
+    fprintf(f,
+      "label=%u, fontcolor=blue, weight=10000, labeltooltip=\"node_count: %u\nerror_cost: %u",
+      i,
+      ts_stack_node_count_since_error(self, i),
+      ts_stack_error_cost(self, i)
+    );
+
+    if (head->summary) {
+      fprintf(f, "\nsummary:");
+      for (uint32_t j = 0; j < head->summary->size; j++) fprintf(f, " %u", head->summary->contents[j].state);
+    }
+
+    if (head->last_external_token.ptr) {
+      const ExternalScannerState *state = &head->last_external_token.ptr->external_scanner_state;
+      const char *data = ts_external_scanner_state_data(state);
+      fprintf(f, "\nexternal_scanner_state:");
+      for (uint32_t j = 0; j < state->length; j++) fprintf(f, " %2X", data[j]);
+    }
+
+    fprintf(f, "\"]\n");
+    array_push(&self->iterators, ((StackIterator) {
+      .node = head->node
+    }));
+  }
+
+  bool all_iterators_done = false;
+  while (!all_iterators_done) {
+    all_iterators_done = true;
+
+    for (uint32_t i = 0; i < self->iterators.size; i++) {
+      StackIterator iterator = self->iterators.contents[i];
+      StackNode *node = iterator.node;
+
+      for (uint32_t j = 0; j < visited_nodes.size; j++) {
+        if (visited_nodes.contents[j] == node) {
+          node = NULL;
+          break;
+        }
+      }
+
+      if (!node) continue;
+      all_iterators_done = false;
+
+      fprintf(f, "node_%p [", (void *)node);
+      if (node->state == ERROR_STATE) {
+        fprintf(f, "label=\"?\"");
+      } else if (
+        node->link_count == 1 &&
+        node->links[0].subtree.ptr &&
+        ts_subtree_extra(node->links[0].subtree)
+      ) {
+        fprintf(f, "shape=point margin=0 label=\"\"");
+      } else {
+        fprintf(f, "label=\"%d\"", node->state);
+      }
+
+      fprintf(
+        f,
+        " tooltip=\"position: %u,%u\nnode_count:%u\nerror_cost: %u\ndynamic_precedence: %d\"];\n",
+        node->position.extent.row + 1,
+        node->position.extent.column,
+        node->node_count,
+        node->error_cost,
+        node->dynamic_precedence
+      );
+
+      for (int j = 0; j < node->link_count; j++) {
+        StackLink link = node->links[j];
+        fprintf(f, "node_%p -> node_%p [", (void *)node, (void *)link.node);
+        if (link.is_pending) fprintf(f, "style=dashed ");
+        if (link.subtree.ptr && ts_subtree_extra(link.subtree)) fprintf(f, "fontcolor=gray ");
+
+        if (!link.subtree.ptr) {
+          fprintf(f, "color=red");
+        } else {
+          fprintf(f, "label=\"");
+          bool quoted = ts_subtree_visible(link.subtree) && !ts_subtree_named(link.subtree);
+          if (quoted) fprintf(f, "'");
+          ts_language_write_symbol_as_dot_string(language, f, ts_subtree_symbol(link.subtree));
+          if (quoted) fprintf(f, "'");
+          fprintf(f, "\"");
+          fprintf(
+            f,
+            "labeltooltip=\"error_cost: %u\ndynamic_precedence: %" PRId32 "\"",
+            ts_subtree_error_cost(link.subtree),
+            ts_subtree_dynamic_precedence(link.subtree)
+          );
+        }
+
+        fprintf(f, "];\n");
+
+        StackIterator *next_iterator;
+        if (j == 0) {
+          next_iterator = &self->iterators.contents[i];
+        } else {
+          array_push(&self->iterators, iterator);
+          next_iterator = array_back(&self->iterators);
+        }
+        next_iterator->node = link.node;
+      }
+
+      array_push(&visited_nodes, node);
+    }
+  }
+
+  fprintf(f, "}\n");
+
+  array_delete(&visited_nodes);
+  return true;
+}
+
+#undef forceinline

vendor/tree-sitter/src/stack.h

@@ -0,0 +1,133 @@
+#ifndef TREE_SITTER_PARSE_STACK_H_
+#define TREE_SITTER_PARSE_STACK_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "./array.h"
+#include "./subtree.h"
+#include "./error_costs.h"
+#include <stdio.h>
+
+typedef struct Stack Stack;
+
+typedef unsigned StackVersion;
+#define STACK_VERSION_NONE ((StackVersion)-1)
+
+typedef struct {
+  SubtreeArray subtrees;
+  StackVersion version;
+} StackSlice;
+typedef Array(StackSlice) StackSliceArray;
+
+typedef struct {
+  Length position;
+  unsigned depth;
+  TSStateId state;
+} StackSummaryEntry;
+typedef Array(StackSummaryEntry) StackSummary;
+
+// Create a stack.
+Stack *ts_stack_new(SubtreePool *);
+
+// Release the memory reserved for a given stack.
+void ts_stack_delete(Stack *);
+
+// Get the stack's current number of versions.
+uint32_t ts_stack_version_count(const Stack *);
+
+// Get the state at the top of the given version of the stack. If the stack is
+// empty, this returns the initial state, 0.
+TSStateId ts_stack_state(const Stack *, StackVersion);
+
+// Get the last external token associated with a given version of the stack.
+Subtree ts_stack_last_external_token(const Stack *, StackVersion);
+
+// Set the last external token associated with a given version of the stack.
+void ts_stack_set_last_external_token(Stack *, StackVersion, Subtree );
+
+// Get the position of the given version of the stack within the document.
+Length ts_stack_position(const Stack *, StackVersion);
+
+// Push a tree and state onto the given version of the stack.
+//
+// This transfers ownership of the tree to the Stack. Callers that
+// need to retain ownership of the tree for their own purposes should
+// first retain the tree.
+void ts_stack_push(Stack *, StackVersion, Subtree , bool, TSStateId);
+
+// Pop the given number of entries from the given version of the stack. This
+// operation can increase the number of stack versions by revealing multiple
+// versions which had previously been merged. It returns an array that
+// specifies the index of each revealed version and the trees that were
+// removed from that version.
+StackSliceArray ts_stack_pop_count(Stack *, StackVersion, uint32_t count);
+
+// Remove an error at the top of the given version of the stack.
+SubtreeArray ts_stack_pop_error(Stack *, StackVersion);
+
+// Remove any pending trees from the top of the given version of the stack.
+StackSliceArray ts_stack_pop_pending(Stack *, StackVersion);
+
+// Remove any all trees from the given version of the stack.
+StackSliceArray ts_stack_pop_all(Stack *, StackVersion);
+
+// Get the maximum number of tree nodes reachable from this version of the stack
+// since the last error was detected.
+unsigned ts_stack_node_count_since_error(const Stack *, StackVersion);
+
+int ts_stack_dynamic_precedence(Stack *, StackVersion);
+
+bool ts_stack_has_advanced_since_error(const Stack *, StackVersion);
+
+// Compute a summary of all the parse states near the top of the given
+// version of the stack and store the summary for later retrieval.
+void ts_stack_record_summary(Stack *, StackVersion, unsigned max_depth);
+
+// Retrieve a summary of all the parse states near the top of the
+// given version of the stack.
+StackSummary *ts_stack_get_summary(Stack *, StackVersion);
+
+// Get the total cost of all errors on the given version of the stack.
+unsigned ts_stack_error_cost(const Stack *, StackVersion version);
+
+// Merge the given two stack versions if possible, returning true
+// if they were successfully merged and false otherwise.
+bool ts_stack_merge(Stack *, StackVersion, StackVersion);
+
+// Determine whether the given two stack versions can be merged.
+bool ts_stack_can_merge(Stack *, StackVersion, StackVersion);
+
+Subtree ts_stack_resume(Stack *, StackVersion);
+
+void ts_stack_pause(Stack *, StackVersion, Subtree);
+
+void ts_stack_halt(Stack *, StackVersion);
+
+bool ts_stack_is_active(const Stack *, StackVersion);
+
+bool ts_stack_is_paused(const Stack *, StackVersion);
+
+bool ts_stack_is_halted(const Stack *, StackVersion);
+
+void ts_stack_renumber_version(Stack *, StackVersion, StackVersion);
+
+void ts_stack_swap_versions(Stack *, StackVersion, StackVersion);
+
+StackVersion ts_stack_copy_version(Stack *, StackVersion);
+
+// Remove the given version from the stack.
+void ts_stack_remove_version(Stack *, StackVersion);
+
+void ts_stack_clear(Stack *);
+
+bool ts_stack_print_dot_graph(Stack *, const TSLanguage *, FILE *);
+
+typedef void (*StackIterateCallback)(void *, TSStateId, uint32_t);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TREE_SITTER_PARSE_STACK_H_

vendor/tree-sitter/src/subtree.h

@@ -0,0 +1,382 @@
+#ifndef TREE_SITTER_SUBTREE_H_
+#define TREE_SITTER_SUBTREE_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <limits.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include "./length.h"
+#include "./array.h"
+#include "./error_costs.h"
+#include "./host.h"
+#include "tree_sitter/api.h"
+#include "./parser.h"
+
+#define TS_TREE_STATE_NONE USHRT_MAX
+#define NULL_SUBTREE ((Subtree) {.ptr = NULL})
+
+// The serialized state of an external scanner.
+//
+// Every time an external token subtree is created after a call to an
+// external scanner, the scanner's `serialize` function is called to
+// retrieve a serialized copy of its state. The bytes are then copied
+// onto the subtree itself so that the scanner's state can later be
+// restored using its `deserialize` function.
+//
+// Small byte arrays are stored inline, and long ones are allocated
+// separately on the heap.
+typedef struct {
+  union {
+    char *long_data;
+    char short_data[24];
+  };
+  uint32_t length;
+} ExternalScannerState;
+
+// A compact representation of a subtree.
+//
+// This representation is used for small leaf nodes that are not
+// errors, and were not created by an external scanner.
+//
+// The idea behind the layout of this struct is that the `is_inline`
+// bit will fall exactly into the same location as the least significant
+// bit of the pointer in `Subtree` or `MutableSubtree`, respectively.
+// Because of alignment, for any valid pointer this will be 0, giving
+// us the opportunity to make use of this bit to signify whether to use
+// the pointer or the inline struct.
+typedef struct SubtreeInlineData SubtreeInlineData;
+
+#define SUBTREE_BITS    \
+  bool visible : 1;     \
+  bool named : 1;       \
+  bool extra : 1;       \
+  bool has_changes : 1; \
+  bool is_missing : 1;  \
+  bool is_keyword : 1;
+
+#define SUBTREE_SIZE           \
+  uint8_t padding_columns;     \
+  uint8_t padding_rows : 4;    \
+  uint8_t lookahead_bytes : 4; \
+  uint8_t padding_bytes;       \
+  uint8_t size_bytes;
+
+#if TS_BIG_ENDIAN
+#if TS_PTR_SIZE == 32
+
+struct SubtreeInlineData {
+  uint16_t parse_state;
+  uint8_t symbol;
+  SUBTREE_BITS
+  bool unused : 1;
+  bool is_inline : 1;
+  SUBTREE_SIZE
+};
+
+#else
+
+struct SubtreeInlineData {
+  SUBTREE_SIZE
+  uint16_t parse_state;
+  uint8_t symbol;
+  SUBTREE_BITS
+  bool unused : 1;
+  bool is_inline : 1;
+};
+
+#endif
+#else
+
+struct SubtreeInlineData {
+  bool is_inline : 1;
+  SUBTREE_BITS
+  uint8_t symbol;
+  uint16_t parse_state;
+  SUBTREE_SIZE
+};
+
+#endif
+
+#undef SUBTREE_BITS
+#undef SUBTREE_SIZE
+
+// A heap-allocated representation of a subtree.
+//
+// This representation is used for parent nodes, external tokens,
+// errors, and other leaf nodes whose data is too large to fit into
+// the inline representation.
+typedef struct {
+  volatile uint32_t ref_count;
+  Length padding;
+  Length size;
+  uint32_t lookahead_bytes;
+  uint32_t error_cost;
+  uint32_t child_count;
+  TSSymbol symbol;
+  TSStateId parse_state;
+
+  bool visible : 1;
+  bool named : 1;
+  bool extra : 1;
+  bool fragile_left : 1;
+  bool fragile_right : 1;
+  bool has_changes : 1;
+  bool has_external_tokens : 1;
+  bool has_external_scanner_state_change : 1;
+  bool depends_on_column: 1;
+  bool is_missing : 1;
+  bool is_keyword : 1;
+
+  union {
+    // Non-terminal subtrees (`child_count > 0`)
+    struct {
+      uint32_t visible_child_count;
+      uint32_t named_child_count;
+      uint32_t visible_descendant_count;
+      int32_t dynamic_precedence;
+      uint16_t repeat_depth;
+      uint16_t production_id;
+      struct {
+        TSSymbol symbol;
+        TSStateId parse_state;
+      } first_leaf;
+    };
+
+    // External terminal subtrees (`child_count == 0 && has_external_tokens`)
+    ExternalScannerState external_scanner_state;
+
+    // Error terminal subtrees (`child_count == 0 && symbol == ts_builtin_sym_error`)
+    int32_t lookahead_char;
+  };
+} SubtreeHeapData;
+
+// The fundamental building block of a syntax tree.
+typedef union {
+  SubtreeInlineData data;
+  const SubtreeHeapData *ptr;
+} Subtree;
+
+// Like Subtree, but mutable.
+typedef union {
+  SubtreeInlineData data;
+  SubtreeHeapData *ptr;
+} MutableSubtree;
+
+typedef Array(Subtree) SubtreeArray;
+typedef Array(MutableSubtree) MutableSubtreeArray;
+
+typedef struct {
+  MutableSubtreeArray free_trees;
+  MutableSubtreeArray tree_stack;
+} SubtreePool;
+
+void ts_external_scanner_state_init(ExternalScannerState *, const char *, unsigned);
+const char *ts_external_scanner_state_data(const ExternalScannerState *);
+bool ts_external_scanner_state_eq(const ExternalScannerState *self, const char *, unsigned);
+void ts_external_scanner_state_delete(ExternalScannerState *self);
+
+void ts_subtree_array_copy(SubtreeArray, SubtreeArray *);
+void ts_subtree_array_clear(SubtreePool *, SubtreeArray *);
+void ts_subtree_array_delete(SubtreePool *, SubtreeArray *);
+void ts_subtree_array_remove_trailing_extras(SubtreeArray *, SubtreeArray *);
+void ts_subtree_array_reverse(SubtreeArray *);
+
+SubtreePool ts_subtree_pool_new(uint32_t capacity);
+void ts_subtree_pool_delete(SubtreePool *);
+
+Subtree ts_subtree_new_leaf(
+  SubtreePool *, TSSymbol, Length, Length, uint32_t,
+  TSStateId, bool, bool, bool, const TSLanguage *
+);
+Subtree ts_subtree_new_error(
+  SubtreePool *, int32_t, Length, Length, uint32_t, TSStateId, const TSLanguage *
+);
+MutableSubtree ts_subtree_new_node(TSSymbol, SubtreeArray *, unsigned, const TSLanguage *);
+Subtree ts_subtree_new_error_node(SubtreeArray *, bool, const TSLanguage *);
+Subtree ts_subtree_new_missing_leaf(SubtreePool *, TSSymbol, Length, uint32_t, const TSLanguage *);
+MutableSubtree ts_subtree_make_mut(SubtreePool *, Subtree);
+void ts_subtree_retain(Subtree);
+void ts_subtree_release(SubtreePool *, Subtree);
+int ts_subtree_compare(Subtree, Subtree, SubtreePool *);
+void ts_subtree_set_symbol(MutableSubtree *, TSSymbol, const TSLanguage *);
+void ts_subtree_summarize(MutableSubtree, const Subtree *, uint32_t, const TSLanguage *);
+void ts_subtree_summarize_children(MutableSubtree, const TSLanguage *);
+void ts_subtree_balance(Subtree, SubtreePool *, const TSLanguage *);
+Subtree ts_subtree_edit(Subtree, const TSInputEdit *edit, SubtreePool *);
+char *ts_subtree_string(Subtree, TSSymbol, bool, const TSLanguage *, bool include_all);
+void ts_subtree_print_dot_graph(Subtree, const TSLanguage *, FILE *);
+Subtree ts_subtree_last_external_token(Subtree);
+const ExternalScannerState *ts_subtree_external_scanner_state(Subtree self);
+bool ts_subtree_external_scanner_state_eq(Subtree, Subtree);
+
+#define SUBTREE_GET(self, name) ((self).data.is_inline ? (self).data.name : (self).ptr->name)
+
+static inline TSSymbol ts_subtree_symbol(Subtree self) { return SUBTREE_GET(self, symbol); }
+static inline bool ts_subtree_visible(Subtree self) { return SUBTREE_GET(self, visible); }
+static inline bool ts_subtree_named(Subtree self) { return SUBTREE_GET(self, named); }
+static inline bool ts_subtree_extra(Subtree self) { return SUBTREE_GET(self, extra); }
+static inline bool ts_subtree_has_changes(Subtree self) { return SUBTREE_GET(self, has_changes); }
+static inline bool ts_subtree_missing(Subtree self) { return SUBTREE_GET(self, is_missing); }
+static inline bool ts_subtree_is_keyword(Subtree self) { return SUBTREE_GET(self, is_keyword); }
+static inline TSStateId ts_subtree_parse_state(Subtree self) { return SUBTREE_GET(self, parse_state); }
+static inline uint32_t ts_subtree_lookahead_bytes(Subtree self) { return SUBTREE_GET(self, lookahead_bytes); }
+
+#undef SUBTREE_GET
+
+// Get the size needed to store a heap-allocated subtree with the given
+// number of children.
+static inline size_t ts_subtree_alloc_size(uint32_t child_count) {
+  return child_count * sizeof(Subtree) + sizeof(SubtreeHeapData);
+}
+
+// Get a subtree's children, which are allocated immediately before the
+// tree's own heap data.
+#define ts_subtree_children(self) \
+  ((self).data.is_inline ? NULL : (Subtree *)((self).ptr) - (self).ptr->child_count)
+
+static inline void ts_subtree_set_extra(MutableSubtree *self, bool is_extra) {
+  if (self->data.is_inline) {
+    self->data.extra = is_extra;
+  } else {
+    self->ptr->extra = is_extra;
+  }
+}
+
+static inline TSSymbol ts_subtree_leaf_symbol(Subtree self) {
+  if (self.data.is_inline) return self.data.symbol;
+  if (self.ptr->child_count == 0) return self.ptr->symbol;
+  return self.ptr->first_leaf.symbol;
+}
+
+static inline TSStateId ts_subtree_leaf_parse_state(Subtree self) {
+  if (self.data.is_inline) return self.data.parse_state;
+  if (self.ptr->child_count == 0) return self.ptr->parse_state;
+  return self.ptr->first_leaf.parse_state;
+}
+
+static inline Length ts_subtree_padding(Subtree self) {
+  if (self.data.is_inline) {
+    Length result = {self.data.padding_bytes, {self.data.padding_rows, self.data.padding_columns}};
+    return result;
+  } else {
+    return self.ptr->padding;
+  }
+}
+
+static inline Length ts_subtree_size(Subtree self) {
+  if (self.data.is_inline) {
+    Length result = {self.data.size_bytes, {0, self.data.size_bytes}};
+    return result;
+  } else {
+    return self.ptr->size;
+  }
+}
+
+static inline Length ts_subtree_total_size(Subtree self) {
+  return length_add(ts_subtree_padding(self), ts_subtree_size(self));
+}
+
+static inline uint32_t ts_subtree_total_bytes(Subtree self) {
+  return ts_subtree_total_size(self).bytes;
+}
+
+static inline uint32_t ts_subtree_child_count(Subtree self) {
+  return self.data.is_inline ? 0 : self.ptr->child_count;
+}
+
+static inline uint32_t ts_subtree_repeat_depth(Subtree self) {
+  return self.data.is_inline ? 0 : self.ptr->repeat_depth;
+}
+
+static inline uint32_t ts_subtree_is_repetition(Subtree self) {
+  return self.data.is_inline
+    ? 0
+    : !self.ptr->named && !self.ptr->visible && self.ptr->child_count != 0;
+}
+
+static inline uint32_t ts_subtree_visible_descendant_count(Subtree self) {
+  return (self.data.is_inline || self.ptr->child_count == 0)
+    ? 0
+    : self.ptr->visible_descendant_count;
+}
+
+static inline uint32_t ts_subtree_visible_child_count(Subtree self) {
+  if (ts_subtree_child_count(self) > 0) {
+    return self.ptr->visible_child_count;
+  } else {
+    return 0;
+  }
+}
+
+static inline uint32_t ts_subtree_error_cost(Subtree self) {
+  if (ts_subtree_missing(self)) {
+    return ERROR_COST_PER_MISSING_TREE + ERROR_COST_PER_RECOVERY;
+  } else {
+    return self.data.is_inline ? 0 : self.ptr->error_cost;
+  }
+}
+
+static inline int32_t ts_subtree_dynamic_precedence(Subtree self) {
+  return (self.data.is_inline || self.ptr->child_count == 0) ? 0 : self.ptr->dynamic_precedence;
+}
+
+static inline uint16_t ts_subtree_production_id(Subtree self) {
+  if (ts_subtree_child_count(self) > 0) {
+    return self.ptr->production_id;
+  } else {
+    return 0;
+  }
+}
+
+static inline bool ts_subtree_fragile_left(Subtree self) {
+  return self.data.is_inline ? false : self.ptr->fragile_left;
+}
+
+static inline bool ts_subtree_fragile_right(Subtree self) {
+  return self.data.is_inline ? false : self.ptr->fragile_right;
+}
+
+static inline bool ts_subtree_has_external_tokens(Subtree self) {
+  return self.data.is_inline ? false : self.ptr->has_external_tokens;
+}
+
+static inline bool ts_subtree_has_external_scanner_state_change(Subtree self) {
+  return self.data.is_inline ? false : self.ptr->has_external_scanner_state_change;
+}
+
+static inline bool ts_subtree_depends_on_column(Subtree self) {
+  return self.data.is_inline ? false : self.ptr->depends_on_column;
+}
+
+static inline bool ts_subtree_is_fragile(Subtree self) {
+  return self.data.is_inline ? false : (self.ptr->fragile_left || self.ptr->fragile_right);
+}
+
+static inline bool ts_subtree_is_error(Subtree self) {
+  return ts_subtree_symbol(self) == ts_builtin_sym_error;
+}
+
+static inline bool ts_subtree_is_eof(Subtree self) {
+  return ts_subtree_symbol(self) == ts_builtin_sym_end;
+}
+
+static inline Subtree ts_subtree_from_mut(MutableSubtree self) {
+  Subtree result;
+  result.data = self.data;
+  return result;
+}
+
+static inline MutableSubtree ts_subtree_to_mut_unsafe(Subtree self) {
+  MutableSubtree result;
+  result.data = self.data;
+  return result;
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TREE_SITTER_SUBTREE_H_

vendor/tree-sitter/src/tree.c

@@ -0,0 +1,165 @@
+#define _POSIX_C_SOURCE 200112L
+
+#include "tree_sitter/api.h"
+#include "./array.h"
+#include "./get_changed_ranges.h"
+#include "./length.h"
+#include "./subtree.h"
+#include "./tree_cursor.h"
+#include "./tree.h"
+
+TSTree *ts_tree_new(
+  Subtree root, const TSLanguage *language,
+  const TSRange *included_ranges, unsigned included_range_count
+) {
+  TSTree *result = ts_malloc(sizeof(TSTree));
+  result->root = root;
+  result->language = ts_language_copy(language);
+  result->included_ranges = ts_calloc(included_range_count, sizeof(TSRange));
+  memcpy(result->included_ranges, included_ranges, included_range_count * sizeof(TSRange));
+  result->included_range_count = included_range_count;
+  return result;
+}
+
+TSTree *ts_tree_copy(const TSTree *self) {
+  ts_subtree_retain(self->root);
+  return ts_tree_new(self->root, self->language, self->included_ranges, self->included_range_count);
+}
+
+void ts_tree_delete(TSTree *self) {
+  if (!self) return;
+
+  SubtreePool pool = ts_subtree_pool_new(0);
+  ts_subtree_release(&pool, self->root);
+  ts_subtree_pool_delete(&pool);
+  ts_language_delete(self->language);
+  ts_free(self->included_ranges);
+  ts_free(self);
+}
+
+TSNode ts_tree_root_node(const TSTree *self) {
+  return ts_node_new(self, &self->root, ts_subtree_padding(self->root), 0);
+}
+
+TSNode ts_tree_root_node_with_offset(
+  const TSTree *self,
+  uint32_t offset_bytes,
+  TSPoint offset_extent
+) {
+  Length offset = {offset_bytes, offset_extent};
+  return ts_node_new(self, &self->root, length_add(offset, ts_subtree_padding(self->root)), 0);
+}
+
+const TSLanguage *ts_tree_language(const TSTree *self) {
+  return self->language;
+}
+
+void ts_tree_edit(TSTree *self, const TSInputEdit *edit) {
+  for (unsigned i = 0; i < self->included_range_count; i++) {
+    TSRange *range = &self->included_ranges[i];
+    if (range->end_byte >= edit->old_end_byte) {
+      if (range->end_byte != UINT32_MAX) {
+        range->end_byte = edit->new_end_byte + (range->end_byte - edit->old_end_byte);
+        range->end_point = point_add(
+          edit->new_end_point,
+          point_sub(range->end_point, edit->old_end_point)
+        );
+        if (range->end_byte < edit->new_end_byte) {
+          range->end_byte = UINT32_MAX;
+          range->end_point = POINT_MAX;
+        }
+      }
+    } else if (range->end_byte > edit->start_byte) {
+      range->end_byte = edit->start_byte;
+      range->end_point = edit->start_point;
+    }
+    if (range->start_byte >= edit->old_end_byte) {
+      range->start_byte = edit->new_end_byte + (range->start_byte - edit->old_end_byte);
+      range->start_point = point_add(
+        edit->new_end_point,
+        point_sub(range->start_point, edit->old_end_point)
+      );
+      if (range->start_byte < edit->new_end_byte) {
+        range->start_byte = UINT32_MAX;
+        range->start_point = POINT_MAX;
+      }
+    } else if (range->start_byte > edit->start_byte) {
+      range->start_byte = edit->start_byte;
+      range->start_point = edit->start_point;
+    }
+  }
+
+  SubtreePool pool = ts_subtree_pool_new(0);
+  self->root = ts_subtree_edit(self->root, edit, &pool);
+  ts_subtree_pool_delete(&pool);
+}
+
+TSRange *ts_tree_included_ranges(const TSTree *self, uint32_t *length) {
+  *length = self->included_range_count;
+  TSRange *ranges = ts_calloc(self->included_range_count, sizeof(TSRange));
+  memcpy(ranges, self->included_ranges, self->included_range_count * sizeof(TSRange));
+  return ranges;
+}
+
+TSRange *ts_tree_get_changed_ranges(const TSTree *old_tree, const TSTree *new_tree, uint32_t *length) {
+  TreeCursor cursor1 = {NULL, array_new(), 0};
+  TreeCursor cursor2 = {NULL, array_new(), 0};
+  ts_tree_cursor_init(&cursor1, ts_tree_root_node(old_tree));
+  ts_tree_cursor_init(&cursor2, ts_tree_root_node(new_tree));
+
+  TSRangeArray included_range_differences = array_new();
+  ts_range_array_get_changed_ranges(
+    old_tree->included_ranges, old_tree->included_range_count,
+    new_tree->included_ranges, new_tree->included_range_count,
+    &included_range_differences
+  );
+
+  TSRange *result;
+  *length = ts_subtree_get_changed_ranges(
+    &old_tree->root, &new_tree->root, &cursor1, &cursor2,
+    old_tree->language, &included_range_differences, &result
+  );
+
+  array_delete(&included_range_differences);
+  array_delete(&cursor1.stack);
+  array_delete(&cursor2.stack);
+  return result;
+}
+
+#ifdef _WIN32
+
+#include <io.h>
+#include <windows.h>
+
+int _ts_dup(HANDLE handle) {
+  HANDLE dup_handle;
+  if (!DuplicateHandle(
+    GetCurrentProcess(), handle,
+    GetCurrentProcess(), &dup_handle,
+    0, FALSE, DUPLICATE_SAME_ACCESS
+  )) return -1;
+
+  return _open_osfhandle((intptr_t)dup_handle, 0);
+}
+
+void ts_tree_print_dot_graph(const TSTree *self, int fd) {
+  FILE *file = _fdopen(_ts_dup((HANDLE)_get_osfhandle(fd)), "a");
+  ts_subtree_print_dot_graph(self->root, self->language, file);
+  fclose(file);
+}
+
+#else
+
+#include <unistd.h>
+
+int _ts_dup(int file_descriptor) {
+  return dup(file_descriptor);
+}
+
+void ts_tree_print_dot_graph(const TSTree *self, int file_descriptor) {
+  FILE *file = fdopen(_ts_dup(file_descriptor), "a");
+  ts_subtree_print_dot_graph(self->root, self->language, file);
+  fclose(file);
+}
+
+#endif

vendor/tree-sitter/src/tree.h

@@ -0,0 +1,31 @@
+#ifndef TREE_SITTER_TREE_H_
+#define TREE_SITTER_TREE_H_
+
+#include "./subtree.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct {
+  const Subtree *child;
+  const Subtree *parent;
+  Length position;
+  TSSymbol alias_symbol;
+} ParentCacheEntry;
+
+struct TSTree {
+  Subtree root;
+  const TSLanguage *language;
+  TSRange *included_ranges;
+  unsigned included_range_count;
+};
+
+TSTree *ts_tree_new(Subtree root, const TSLanguage *language, const TSRange *, unsigned);
+TSNode ts_node_new(const TSTree *, const Subtree *, Length, TSSymbol);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TREE_SITTER_TREE_H_

vendor/tree-sitter/src/tree_cursor.c

@@ -0,0 +1,714 @@
+#include "tree_sitter/api.h"
+#include "./alloc.h"
+#include "./tree_cursor.h"
+#include "./language.h"
+#include "./tree.h"
+
+typedef struct {
+  Subtree parent;
+  const TSTree *tree;
+  Length position;
+  uint32_t child_index;
+  uint32_t structural_child_index;
+  uint32_t descendant_index;
+  const TSSymbol *alias_sequence;
+} CursorChildIterator;
+
+// CursorChildIterator
+
+static inline bool ts_tree_cursor_is_entry_visible(const TreeCursor *self, uint32_t index) {
+  TreeCursorEntry *entry = &self->stack.contents[index];
+  if (index == 0 || ts_subtree_visible(*entry->subtree)) {
+    return true;
+  } else if (!ts_subtree_extra(*entry->subtree)) {
+    TreeCursorEntry *parent_entry = &self->stack.contents[index - 1];
+    return ts_language_alias_at(
+      self->tree->language,
+      parent_entry->subtree->ptr->production_id,
+      entry->structural_child_index
+    );
+  } else {
+    return false;
+  }
+}
+
+static inline CursorChildIterator ts_tree_cursor_iterate_children(const TreeCursor *self) {
+  TreeCursorEntry *last_entry = array_back(&self->stack);
+  if (ts_subtree_child_count(*last_entry->subtree) == 0) {
+    return (CursorChildIterator) {NULL_SUBTREE, self->tree, length_zero(), 0, 0, 0, NULL};
+  }
+  const TSSymbol *alias_sequence = ts_language_alias_sequence(
+    self->tree->language,
+    last_entry->subtree->ptr->production_id
+  );
+
+  uint32_t descendant_index = last_entry->descendant_index;
+  if (ts_tree_cursor_is_entry_visible(self, self->stack.size - 1)) {
+    descendant_index += 1;
+  }
+
+  return (CursorChildIterator) {
+    .tree = self->tree,
+    .parent = *last_entry->subtree,
+    .position = last_entry->position,
+    .child_index = 0,
+    .structural_child_index = 0,
+    .descendant_index = descendant_index,
+    .alias_sequence = alias_sequence,
+  };
+}
+
+static inline bool ts_tree_cursor_child_iterator_next(
+  CursorChildIterator *self,
+  TreeCursorEntry *result,
+  bool *visible
+) {
+  if (!self->parent.ptr || self->child_index == self->parent.ptr->child_count) return false;
+  const Subtree *child = &ts_subtree_children(self->parent)[self->child_index];
+  *result = (TreeCursorEntry) {
+    .subtree = child,
+    .position = self->position,
+    .child_index = self->child_index,
+    .structural_child_index = self->structural_child_index,
+    .descendant_index = self->descendant_index,
+  };
+  *visible = ts_subtree_visible(*child);
+  bool extra = ts_subtree_extra(*child);
+  if (!extra) {
+    if (self->alias_sequence) {
+      *visible |= self->alias_sequence[self->structural_child_index];
+    }
+    self->structural_child_index++;
+  }
+
+  self->descendant_index += ts_subtree_visible_descendant_count(*child);
+  if (*visible) {
+    self->descendant_index += 1;
+  }
+
+  self->position = length_add(self->position, ts_subtree_size(*child));
+  self->child_index++;
+
+  if (self->child_index < self->parent.ptr->child_count) {
+    Subtree next_child = ts_subtree_children(self->parent)[self->child_index];
+    self->position = length_add(self->position, ts_subtree_padding(next_child));
+  }
+
+  return true;
+}
+
+// Return a position that, when `b` is added to it, yields `a`. This
+// can only be computed if `b` has zero rows. Otherwise, this function
+// returns `LENGTH_UNDEFINED`, and the caller needs to recompute
+// the position some other way.
+static inline Length length_backtrack(Length a, Length b) {
+  if (length_is_undefined(a) || b.extent.row != 0) {
+    return LENGTH_UNDEFINED;
+  }
+
+  Length result;
+  result.bytes = a.bytes - b.bytes;
+  result.extent.row = a.extent.row;
+  result.extent.column = a.extent.column - b.extent.column;
+  return result;
+}
+
+static inline bool ts_tree_cursor_child_iterator_previous(
+  CursorChildIterator *self,
+  TreeCursorEntry *result,
+  bool *visible
+) {
+  // this is mostly a reverse `ts_tree_cursor_child_iterator_next` taking into
+  // account unsigned underflow
+  if (!self->parent.ptr || (int8_t)self->child_index == -1) return false;
+  const Subtree *child = &ts_subtree_children(self->parent)[self->child_index];
+  *result = (TreeCursorEntry) {
+    .subtree = child,
+    .position = self->position,
+    .child_index = self->child_index,
+    .structural_child_index = self->structural_child_index,
+  };
+  *visible = ts_subtree_visible(*child);
+  bool extra = ts_subtree_extra(*child);
+  if (!extra && self->alias_sequence) {
+    *visible |= self->alias_sequence[self->structural_child_index];
+    self->structural_child_index--;
+  }
+
+  self->position = length_backtrack(self->position, ts_subtree_padding(*child));
+  self->child_index--;
+
+  // unsigned can underflow so compare it to child_count
+  if (self->child_index < self->parent.ptr->child_count) {
+    Subtree previous_child = ts_subtree_children(self->parent)[self->child_index];
+    Length size = ts_subtree_size(previous_child);
+    self->position = length_backtrack(self->position, size);
+  }
+
+  return true;
+}
+
+// TSTreeCursor - lifecycle
+
+TSTreeCursor ts_tree_cursor_new(TSNode node) {
+  TSTreeCursor self = {NULL, NULL, {0, 0, 0}};
+  ts_tree_cursor_init((TreeCursor *)&self, node);
+  return self;
+}
+
+void ts_tree_cursor_reset(TSTreeCursor *_self, TSNode node) {
+  ts_tree_cursor_init((TreeCursor *)_self, node);
+}
+
+void ts_tree_cursor_init(TreeCursor *self, TSNode node) {
+  self->tree = node.tree;
+  self->root_alias_symbol = node.context[3];
+  array_clear(&self->stack);
+  array_push(&self->stack, ((TreeCursorEntry) {
+    .subtree = (const Subtree *)node.id,
+    .position = {
+      ts_node_start_byte(node),
+      ts_node_start_point(node)
+    },
+    .child_index = 0,
+    .structural_child_index = 0,
+    .descendant_index = 0,
+  }));
+}
+
+void ts_tree_cursor_delete(TSTreeCursor *_self) {
+  TreeCursor *self = (TreeCursor *)_self;
+  array_delete(&self->stack);
+}
+
+// TSTreeCursor - walking the tree
+
+TreeCursorStep ts_tree_cursor_goto_first_child_internal(TSTreeCursor *_self) {
+  TreeCursor *self = (TreeCursor *)_self;
+  bool visible;
+  TreeCursorEntry entry;
+  CursorChildIterator iterator = ts_tree_cursor_iterate_children(self);
+  while (ts_tree_cursor_child_iterator_next(&iterator, &entry, &visible)) {
+    if (visible) {
+      array_push(&self->stack, entry);
+      return TreeCursorStepVisible;
+    }
+    if (ts_subtree_visible_child_count(*entry.subtree) > 0) {
+      array_push(&self->stack, entry);
+      return TreeCursorStepHidden;
+    }
+  }
+  return TreeCursorStepNone;
+}
+
+bool ts_tree_cursor_goto_first_child(TSTreeCursor *self) {
+  for (;;) {
+    switch (ts_tree_cursor_goto_first_child_internal(self)) {
+      case TreeCursorStepHidden:
+        continue;
+      case TreeCursorStepVisible:
+        return true;
+      default:
+        return false;
+    }
+  }
+  return false;
+}
+
+TreeCursorStep ts_tree_cursor_goto_last_child_internal(TSTreeCursor *_self) {
+  TreeCursor *self = (TreeCursor *)_self;
+  bool visible;
+  TreeCursorEntry entry;
+  CursorChildIterator iterator = ts_tree_cursor_iterate_children(self);
+  if (!iterator.parent.ptr || iterator.parent.ptr->child_count == 0) return TreeCursorStepNone;
+
+  TreeCursorEntry last_entry = {0};
+  TreeCursorStep last_step = TreeCursorStepNone;
+  while (ts_tree_cursor_child_iterator_next(&iterator, &entry, &visible)) {
+    if (visible) {
+      last_entry = entry;
+      last_step = TreeCursorStepVisible;
+    }
+    else if (ts_subtree_visible_child_count(*entry.subtree) > 0) {
+      last_entry = entry;
+      last_step = TreeCursorStepHidden;
+    }
+  }
+  if (last_entry.subtree) {
+    array_push(&self->stack, last_entry);
+    return last_step;
+  }
+
+  return TreeCursorStepNone;
+}
+
+bool ts_tree_cursor_goto_last_child(TSTreeCursor *self) {
+  for (;;) {
+    switch (ts_tree_cursor_goto_last_child_internal(self)) {
+      case TreeCursorStepHidden:
+        continue;
+      case TreeCursorStepVisible:
+        return true;
+      default:
+        return false;
+    }
+  }
+  return false;
+}
+
+static inline int64_t ts_tree_cursor_goto_first_child_for_byte_and_point(
+  TSTreeCursor *_self,
+  uint32_t goal_byte,
+  TSPoint goal_point
+) {
+  TreeCursor *self = (TreeCursor *)_self;
+  uint32_t initial_size = self->stack.size;
+  uint32_t visible_child_index = 0;
+
+  bool did_descend;
+  do {
+    did_descend = false;
+
+    bool visible;
+    TreeCursorEntry entry;
+    CursorChildIterator iterator = ts_tree_cursor_iterate_children(self);
+    while (ts_tree_cursor_child_iterator_next(&iterator, &entry, &visible)) {
+      Length entry_end = length_add(entry.position, ts_subtree_size(*entry.subtree));
+      bool at_goal = entry_end.bytes >= goal_byte && point_gte(entry_end.extent, goal_point);
+      uint32_t visible_child_count = ts_subtree_visible_child_count(*entry.subtree);
+      if (at_goal) {
+        if (visible) {
+          array_push(&self->stack, entry);
+          return visible_child_index;
+        }
+        if (visible_child_count > 0) {
+          array_push(&self->stack, entry);
+          did_descend = true;
+          break;
+        }
+      } else if (visible) {
+        visible_child_index++;
+      } else {
+        visible_child_index += visible_child_count;
+      }
+    }
+  } while (did_descend);
+
+  self->stack.size = initial_size;
+  return -1;
+}
+
+int64_t ts_tree_cursor_goto_first_child_for_byte(TSTreeCursor *self, uint32_t goal_byte) {
+  return ts_tree_cursor_goto_first_child_for_byte_and_point(self, goal_byte, POINT_ZERO);
+}
+
+int64_t ts_tree_cursor_goto_first_child_for_point(TSTreeCursor *self, TSPoint goal_point) {
+  return ts_tree_cursor_goto_first_child_for_byte_and_point(self, 0, goal_point);
+}
+
+TreeCursorStep ts_tree_cursor_goto_sibling_internal(
+    TSTreeCursor *_self,
+    bool (*advance)(CursorChildIterator *, TreeCursorEntry *, bool *)) {
+  TreeCursor *self = (TreeCursor *)_self;
+  uint32_t initial_size = self->stack.size;
+
+  while (self->stack.size > 1) {
+    TreeCursorEntry entry = array_pop(&self->stack);
+    CursorChildIterator iterator = ts_tree_cursor_iterate_children(self);
+    iterator.child_index = entry.child_index;
+    iterator.structural_child_index = entry.structural_child_index;
+    iterator.position = entry.position;
+    iterator.descendant_index = entry.descendant_index;
+
+    bool visible = false;
+    advance(&iterator, &entry, &visible);
+    if (visible && self->stack.size + 1 < initial_size) break;
+
+    while (advance(&iterator, &entry, &visible)) {
+      if (visible) {
+        array_push(&self->stack, entry);
+        return TreeCursorStepVisible;
+      }
+
+      if (ts_subtree_visible_child_count(*entry.subtree)) {
+        array_push(&self->stack, entry);
+        return TreeCursorStepHidden;
+      }
+    }
+  }
+
+  self->stack.size = initial_size;
+  return TreeCursorStepNone;
+}
+
+TreeCursorStep ts_tree_cursor_goto_next_sibling_internal(TSTreeCursor *_self) {
+  return ts_tree_cursor_goto_sibling_internal(_self, ts_tree_cursor_child_iterator_next);
+}
+
+bool ts_tree_cursor_goto_next_sibling(TSTreeCursor *self) {
+  switch (ts_tree_cursor_goto_next_sibling_internal(self)) {
+    case TreeCursorStepHidden:
+      ts_tree_cursor_goto_first_child(self);
+      return true;
+    case TreeCursorStepVisible:
+      return true;
+    default:
+      return false;
+  }
+}
+
+TreeCursorStep ts_tree_cursor_goto_previous_sibling_internal(TSTreeCursor *_self) {
+  // since subtracting across row loses column information, we may have to
+  // restore it
+  TreeCursor *self = (TreeCursor *)_self;
+
+  // for that, save current position before traversing
+  TreeCursorStep step = ts_tree_cursor_goto_sibling_internal(
+      _self, ts_tree_cursor_child_iterator_previous);
+  if (step == TreeCursorStepNone)
+    return step;
+
+  // if length is already valid, there's no need to recompute it
+  if (!length_is_undefined(array_back(&self->stack)->position))
+    return step;
+
+  // restore position from the parent node
+  const TreeCursorEntry *parent = &self->stack.contents[self->stack.size - 2];
+  Length position = parent->position;
+  uint32_t child_index = array_back(&self->stack)->child_index;
+  const Subtree *children = ts_subtree_children((*(parent->subtree)));
+
+  if (child_index > 0) {
+    // skip first child padding since its position should match the position of the parent
+    position = length_add(position, ts_subtree_size(children[0]));
+    for (uint32_t i = 1; i < child_index; ++i) {
+      position = length_add(position, ts_subtree_total_size(children[i]));
+    }
+    position = length_add(position, ts_subtree_padding(children[child_index]));
+  }
+
+  array_back(&self->stack)->position = position;
+
+  return step;
+}
+
+bool ts_tree_cursor_goto_previous_sibling(TSTreeCursor *self) {
+  switch (ts_tree_cursor_goto_previous_sibling_internal(self)) {
+    case TreeCursorStepHidden:
+      ts_tree_cursor_goto_last_child(self);
+      return true;
+    case TreeCursorStepVisible:
+      return true;
+    default:
+      return false;
+  }
+}
+
+bool ts_tree_cursor_goto_parent(TSTreeCursor *_self) {
+  TreeCursor *self = (TreeCursor *)_self;
+  for (unsigned i = self->stack.size - 2; i + 1 > 0; i--) {
+    if (ts_tree_cursor_is_entry_visible(self, i)) {
+      self->stack.size = i + 1;
+      return true;
+    }
+  }
+  return false;
+}
+
+void ts_tree_cursor_goto_descendant(
+  TSTreeCursor *_self,
+  uint32_t goal_descendant_index
+) {
+  TreeCursor *self = (TreeCursor *)_self;
+
+  // Ascend to the lowest ancestor that contains the goal node.
+  for (;;) {
+    uint32_t i = self->stack.size - 1;
+    TreeCursorEntry *entry = &self->stack.contents[i];
+    uint32_t next_descendant_index =
+      entry->descendant_index +
+      (ts_tree_cursor_is_entry_visible(self, i) ? 1 : 0) +
+      ts_subtree_visible_descendant_count(*entry->subtree);
+    if (
+      (entry->descendant_index <= goal_descendant_index) &&
+      (next_descendant_index > goal_descendant_index)
+    ) {
+      break;
+    } else if (self->stack.size <= 1) {
+      return;
+    } else {
+      self->stack.size--;
+    }
+  }
+
+  // Descend to the goal node.
+  bool did_descend = true;
+  do {
+    did_descend = false;
+    bool visible;
+    TreeCursorEntry entry;
+    CursorChildIterator iterator = ts_tree_cursor_iterate_children(self);
+    if (iterator.descendant_index > goal_descendant_index) {
+      return;
+    }
+
+    while (ts_tree_cursor_child_iterator_next(&iterator, &entry, &visible)) {
+      if (iterator.descendant_index > goal_descendant_index) {
+        array_push(&self->stack, entry);
+        if (visible && entry.descendant_index == goal_descendant_index) {
+          return;
+        } else {
+          did_descend = true;
+          break;
+        }
+      }
+    }
+  } while (did_descend);
+}
+
+uint32_t ts_tree_cursor_current_descendant_index(const TSTreeCursor *_self) {
+  const TreeCursor *self = (const TreeCursor *)_self;
+  TreeCursorEntry *last_entry = array_back(&self->stack);
+  return last_entry->descendant_index;
+}
+
+TSNode ts_tree_cursor_current_node(const TSTreeCursor *_self) {
+  const TreeCursor *self = (const TreeCursor *)_self;
+  TreeCursorEntry *last_entry = array_back(&self->stack);
+  TSSymbol alias_symbol = self->root_alias_symbol;
+  if (self->stack.size > 1 && !ts_subtree_extra(*last_entry->subtree)) {
+    TreeCursorEntry *parent_entry = &self->stack.contents[self->stack.size - 2];
+    alias_symbol = ts_language_alias_at(
+      self->tree->language,
+      parent_entry->subtree->ptr->production_id,
+      last_entry->structural_child_index
+    );
+  }
+  return ts_node_new(
+    self->tree,
+    last_entry->subtree,
+    last_entry->position,
+    alias_symbol
+  );
+}
+
+// Private - Get various facts about the current node that are needed
+// when executing tree queries.
+void ts_tree_cursor_current_status(
+  const TSTreeCursor *_self,
+  TSFieldId *field_id,
+  bool *has_later_siblings,
+  bool *has_later_named_siblings,
+  bool *can_have_later_siblings_with_this_field,
+  TSSymbol *supertypes,
+  unsigned *supertype_count
+) {
+  const TreeCursor *self = (const TreeCursor *)_self;
+  unsigned max_supertypes = *supertype_count;
+  *field_id = 0;
+  *supertype_count = 0;
+  *has_later_siblings = false;
+  *has_later_named_siblings = false;
+  *can_have_later_siblings_with_this_field = false;
+
+  // Walk up the tree, visiting the current node and its invisible ancestors,
+  // because fields can refer to nodes through invisible *wrapper* nodes,
+  for (unsigned i = self->stack.size - 1; i > 0; i--) {
+    TreeCursorEntry *entry = &self->stack.contents[i];
+    TreeCursorEntry *parent_entry = &self->stack.contents[i - 1];
+
+    const TSSymbol *alias_sequence = ts_language_alias_sequence(
+      self->tree->language,
+      parent_entry->subtree->ptr->production_id
+    );
+
+    #define subtree_symbol(subtree, structural_child_index) \
+      ((                                                    \
+        !ts_subtree_extra(subtree) &&                       \
+        alias_sequence &&                                   \
+        alias_sequence[structural_child_index]              \
+      ) ?                                                   \
+        alias_sequence[structural_child_index] :            \
+        ts_subtree_symbol(subtree))
+
+    // Stop walking up when a visible ancestor is found.
+    TSSymbol entry_symbol = subtree_symbol(
+      *entry->subtree,
+      entry->structural_child_index
+    );
+    TSSymbolMetadata entry_metadata = ts_language_symbol_metadata(
+      self->tree->language,
+      entry_symbol
+    );
+    if (i != self->stack.size - 1 && entry_metadata.visible) break;
+
+    // Record any supertypes
+    if (entry_metadata.supertype && *supertype_count < max_supertypes) {
+      supertypes[*supertype_count] = entry_symbol;
+      (*supertype_count)++;
+    }
+
+    // Determine if the current node has later siblings.
+    if (!*has_later_siblings) {
+      unsigned sibling_count = parent_entry->subtree->ptr->child_count;
+      unsigned structural_child_index = entry->structural_child_index;
+      if (!ts_subtree_extra(*entry->subtree)) structural_child_index++;
+      for (unsigned j = entry->child_index + 1; j < sibling_count; j++) {
+        Subtree sibling = ts_subtree_children(*parent_entry->subtree)[j];
+        TSSymbolMetadata sibling_metadata = ts_language_symbol_metadata(
+          self->tree->language,
+          subtree_symbol(sibling, structural_child_index)
+        );
+        if (sibling_metadata.visible) {
+          *has_later_siblings = true;
+          if (*has_later_named_siblings) break;
+          if (sibling_metadata.named) {
+            *has_later_named_siblings = true;
+            break;
+          }
+        } else if (ts_subtree_visible_child_count(sibling) > 0) {
+          *has_later_siblings = true;
+          if (*has_later_named_siblings) break;
+          if (sibling.ptr->named_child_count > 0) {
+            *has_later_named_siblings = true;
+            break;
+          }
+        }
+        if (!ts_subtree_extra(sibling)) structural_child_index++;
+      }
+    }
+
+    #undef subtree_symbol
+
+    if (!ts_subtree_extra(*entry->subtree)) {
+      const TSFieldMapEntry *field_map, *field_map_end;
+      ts_language_field_map(
+        self->tree->language,
+        parent_entry->subtree->ptr->production_id,
+        &field_map, &field_map_end
+      );
+
+      // Look for a field name associated with the current node.
+      if (!*field_id) {
+        for (const TSFieldMapEntry *map = field_map; map < field_map_end; map++) {
+          if (!map->inherited && map->child_index == entry->structural_child_index) {
+            *field_id = map->field_id;
+            break;
+          }
+        }
+      }
+
+      // Determine if the current node can have later siblings with the same field name.
+      if (*field_id) {
+        for (const TSFieldMapEntry *map = field_map; map < field_map_end; map++) {
+          if (
+            map->field_id == *field_id &&
+            map->child_index > entry->structural_child_index
+          ) {
+            *can_have_later_siblings_with_this_field = true;
+            break;
+          }
+        }
+      }
+    }
+  }
+}
+
+uint32_t ts_tree_cursor_current_depth(const TSTreeCursor *_self) {
+  const TreeCursor *self = (const TreeCursor *)_self;
+  uint32_t depth = 0;
+  for (unsigned i = 1; i < self->stack.size; i++) {
+    if (ts_tree_cursor_is_entry_visible(self, i)) {
+      depth++;
+    }
+  }
+  return depth;
+}
+
+TSNode ts_tree_cursor_parent_node(const TSTreeCursor *_self) {
+  const TreeCursor *self = (const TreeCursor *)_self;
+  for (int i = (int)self->stack.size - 2; i >= 0; i--) {
+    TreeCursorEntry *entry = &self->stack.contents[i];
+    bool is_visible = true;
+    TSSymbol alias_symbol = 0;
+    if (i > 0) {
+      TreeCursorEntry *parent_entry = &self->stack.contents[i - 1];
+      alias_symbol = ts_language_alias_at(
+        self->tree->language,
+        parent_entry->subtree->ptr->production_id,
+        entry->structural_child_index
+      );
+      is_visible = (alias_symbol != 0) || ts_subtree_visible(*entry->subtree);
+    }
+    if (is_visible) {
+      return ts_node_new(
+        self->tree,
+        entry->subtree,
+        entry->position,
+        alias_symbol
+      );
+    }
+  }
+  return ts_node_new(NULL, NULL, length_zero(), 0);
+}
+
+TSFieldId ts_tree_cursor_current_field_id(const TSTreeCursor *_self) {
+  const TreeCursor *self = (const TreeCursor *)_self;
+
+  // Walk up the tree, visiting the current node and its invisible ancestors.
+  for (unsigned i = self->stack.size - 1; i > 0; i--) {
+    TreeCursorEntry *entry = &self->stack.contents[i];
+    TreeCursorEntry *parent_entry = &self->stack.contents[i - 1];
+
+    // Stop walking up when another visible node is found.
+    if (
+      i != self->stack.size - 1 &&
+      ts_tree_cursor_is_entry_visible(self, i)
+    ) break;
+
+    if (ts_subtree_extra(*entry->subtree)) break;
+
+    const TSFieldMapEntry *field_map, *field_map_end;
+    ts_language_field_map(
+      self->tree->language,
+      parent_entry->subtree->ptr->production_id,
+      &field_map, &field_map_end
+    );
+    for (const TSFieldMapEntry *map = field_map; map < field_map_end; map++) {
+      if (!map->inherited && map->child_index == entry->structural_child_index) {
+        return map->field_id;
+      }
+    }
+  }
+  return 0;
+}
+
+const char *ts_tree_cursor_current_field_name(const TSTreeCursor *_self) {
+  TSFieldId id = ts_tree_cursor_current_field_id(_self);
+  if (id) {
+    const TreeCursor *self = (const TreeCursor *)_self;
+    return self->tree->language->field_names[id];
+  } else {
+    return NULL;
+  }
+}
+
+TSTreeCursor ts_tree_cursor_copy(const TSTreeCursor *_cursor) {
+  const TreeCursor *cursor = (const TreeCursor *)_cursor;
+  TSTreeCursor res = {NULL, NULL, {0, 0}};
+  TreeCursor *copy = (TreeCursor *)&res;
+  copy->tree = cursor->tree;
+  copy->root_alias_symbol = cursor->root_alias_symbol;
+  array_init(&copy->stack);
+  array_push_all(&copy->stack, &cursor->stack);
+  return res;
+}
+
+void ts_tree_cursor_reset_to(TSTreeCursor *_dst, const TSTreeCursor *_src) {
+  const TreeCursor *cursor = (const TreeCursor *)_src;
+  TreeCursor *copy = (TreeCursor *)_dst;
+  copy->tree = cursor->tree;
+  copy->root_alias_symbol = cursor->root_alias_symbol;
+  array_clear(&copy->stack);
+  array_push_all(&copy->stack, &cursor->stack);
+}

vendor/tree-sitter/src/tree_cursor.h

@@ -0,0 +1,48 @@
+#ifndef TREE_SITTER_TREE_CURSOR_H_
+#define TREE_SITTER_TREE_CURSOR_H_
+
+#include "./subtree.h"
+
+typedef struct {
+  const Subtree *subtree;
+  Length position;
+  uint32_t child_index;
+  uint32_t structural_child_index;
+  uint32_t descendant_index;
+} TreeCursorEntry;
+
+typedef struct {
+  const TSTree *tree;
+  Array(TreeCursorEntry) stack;
+  TSSymbol root_alias_symbol;
+} TreeCursor;
+
+typedef enum {
+  TreeCursorStepNone,
+  TreeCursorStepHidden,
+  TreeCursorStepVisible,
+} TreeCursorStep;
+
+void ts_tree_cursor_init(TreeCursor *, TSNode);
+void ts_tree_cursor_current_status(
+  const TSTreeCursor *,
+  TSFieldId *,
+  bool *,
+  bool *,
+  bool *,
+  TSSymbol *,
+  unsigned *
+);
+
+TreeCursorStep ts_tree_cursor_goto_first_child_internal(TSTreeCursor *);
+TreeCursorStep ts_tree_cursor_goto_next_sibling_internal(TSTreeCursor *);
+
+static inline Subtree ts_tree_cursor_current_subtree(const TSTreeCursor *_self) {
+  const TreeCursor *self = (const TreeCursor *)_self;
+  TreeCursorEntry *last_entry = array_back(&self->stack);
+  return *last_entry->subtree;
+}
+
+TSNode ts_tree_cursor_parent_node(const TSTreeCursor *);
+
+#endif  // TREE_SITTER_TREE_CURSOR_H_

vendor/tree-sitter/src/unicode.h

@@ -0,0 +1,50 @@
+#ifndef TREE_SITTER_UNICODE_H_
+#define TREE_SITTER_UNICODE_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <limits.h>
+#include <stdint.h>
+
+#define U_EXPORT
+#define U_EXPORT2
+#include "unicode/utf8.h"
+#include "unicode/utf16.h"
+
+static const int32_t TS_DECODE_ERROR = U_SENTINEL;
+
+// These functions read one unicode code point from the given string,
+// returning the number of bytes consumed.
+typedef uint32_t (*UnicodeDecodeFunction)(
+  const uint8_t *string,
+  uint32_t length,
+  int32_t *code_point
+);
+
+static inline uint32_t ts_decode_utf8(
+  const uint8_t *string,
+  uint32_t length,
+  int32_t *code_point
+) {
+  uint32_t i = 0;
+  U8_NEXT(string, i, length, *code_point);
+  return i;
+}
+
+static inline uint32_t ts_decode_utf16(
+  const uint8_t *string,
+  uint32_t length,
+  int32_t *code_point
+) {
+  uint32_t i = 0;
+  U16_NEXT(((uint16_t *)string), i, length, *code_point);
+  return i * 2;
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TREE_SITTER_UNICODE_H_

vendor/tree-sitter/src/unicode/ICU_SHA

@@ -0,0 +1 @@
+552b01f61127d30d6589aa4bf99468224979b661

vendor/tree-sitter/src/unicode/LICENSE

@@ -0,0 +1,414 @@
+COPYRIGHT AND PERMISSION NOTICE (ICU 58 and later)
+
+Copyright © 1991-2019 Unicode, Inc. All rights reserved.
+Distributed under the Terms of Use in https://www.unicode.org/copyright.html.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of the Unicode data files and any associated documentation
+(the "Data Files") or Unicode software and any associated documentation
+(the "Software") to deal in the Data Files or Software
+without restriction, including without limitation the rights to use,
+copy, modify, merge, publish, distribute, and/or sell copies of
+the Data Files or Software, and to permit persons to whom the Data Files
+or Software are furnished to do so, provided that either
+(a) this copyright and permission notice appear with all copies
+of the Data Files or Software, or
+(b) this copyright and permission notice appear in associated
+Documentation.
+
+THE DATA FILES AND SOFTWARE ARE PROVIDED "AS IS", WITHOUT WARRANTY OF
+ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE
+WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+NONINFRINGEMENT OF THIRD PARTY RIGHTS.
+IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN THIS
+NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSEQUENTIAL
+DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE,
+DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
+TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
+PERFORMANCE OF THE DATA FILES OR SOFTWARE.
+
+Except as contained in this notice, the name of a copyright holder
+shall not be used in advertising or otherwise to promote the sale,
+use or other dealings in these Data Files or Software without prior
+written authorization of the copyright holder.
+
+---------------------
+
+Third-Party Software Licenses
+
+This section contains third-party software notices and/or additional
+terms for licensed third-party software components included within ICU
+libraries.
+
+1. ICU License - ICU 1.8.1 to ICU 57.1
+
+COPYRIGHT AND PERMISSION NOTICE
+
+Copyright (c) 1995-2016 International Business Machines Corporation and others
+All rights reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, and/or sell copies of the Software, and to permit persons
+to whom the Software is furnished to do so, provided that the above
+copyright notice(s) and this permission notice appear in all copies of
+the Software and that both the above copyright notice(s) and this
+permission notice appear in supporting documentation.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT
+OF THIRD PARTY RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR
+HOLDERS INCLUDED IN THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY
+SPECIAL INDIRECT OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER
+RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF
+CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
+CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+
+Except as contained in this notice, the name of a copyright holder
+shall not be used in advertising or otherwise to promote the sale, use
+or other dealings in this Software without prior written authorization
+of the copyright holder.
+
+All trademarks and registered trademarks mentioned herein are the
+property of their respective owners.
+
+2. Chinese/Japanese Word Break Dictionary Data (cjdict.txt)
+
+ #     The Google Chrome software developed by Google is licensed under
+ # the BSD license. Other software included in this distribution is
+ # provided under other licenses, as set forth below.
+ #
+ #  The BSD License
+ #  http://opensource.org/licenses/bsd-license.php
+ #  Copyright (C) 2006-2008, Google Inc.
+ #
+ #  All rights reserved.
+ #
+ #  Redistribution and use in source and binary forms, with or without
+ # modification, are permitted provided that the following conditions are met:
+ #
+ #  Redistributions of source code must retain the above copyright notice,
+ # this list of conditions and the following disclaimer.
+ #  Redistributions in binary form must reproduce the above
+ # copyright notice, this list of conditions and the following
+ # disclaimer in the documentation and/or other materials provided with
+ # the distribution.
+ #  Neither the name of  Google Inc. nor the names of its
+ # contributors may be used to endorse or promote products derived from
+ # this software without specific prior written permission.
+ #
+ #
+ #  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
+ # CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
+ # INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+ # MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ # DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ # LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ # CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ # SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
+ # BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+ # LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+ # NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+ # SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ #
+ #
+ #  The word list in cjdict.txt are generated by combining three word lists
+ # listed below with further processing for compound word breaking. The
+ # frequency is generated with an iterative training against Google web
+ # corpora.
+ #
+ #  * Libtabe (Chinese)
+ #    - https://sourceforge.net/project/?group_id=1519
+ #    - Its license terms and conditions are shown below.
+ #
+ #  * IPADIC (Japanese)
+ #    - http://chasen.aist-nara.ac.jp/chasen/distribution.html
+ #    - Its license terms and conditions are shown below.
+ #
+ #  ---------COPYING.libtabe ---- BEGIN--------------------
+ #
+ #  /*
+ #   * Copyright (c) 1999 TaBE Project.
+ #   * Copyright (c) 1999 Pai-Hsiang Hsiao.
+ #   * All rights reserved.
+ #   *
+ #   * Redistribution and use in source and binary forms, with or without
+ #   * modification, are permitted provided that the following conditions
+ #   * are met:
+ #   *
+ #   * . Redistributions of source code must retain the above copyright
+ #   *   notice, this list of conditions and the following disclaimer.
+ #   * . Redistributions in binary form must reproduce the above copyright
+ #   *   notice, this list of conditions and the following disclaimer in
+ #   *   the documentation and/or other materials provided with the
+ #   *   distribution.
+ #   * . Neither the name of the TaBE Project nor the names of its
+ #   *   contributors may be used to endorse or promote products derived
+ #   *   from this software without specific prior written permission.
+ #   *
+ #   * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ #   * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ #   * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ #   * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+ #   * REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+ #   * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+ #   * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ #   * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ #   * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ #   * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ #   * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ #   * OF THE POSSIBILITY OF SUCH DAMAGE.
+ #   */
+ #
+ #  /*
+ #   * Copyright (c) 1999 Computer Systems and Communication Lab,
+ #   *                    Institute of Information Science, Academia
+ #       *                    Sinica. All rights reserved.
+ #   *
+ #   * Redistribution and use in source and binary forms, with or without
+ #   * modification, are permitted provided that the following conditions
+ #   * are met:
+ #   *
+ #   * . Redistributions of source code must retain the above copyright
+ #   *   notice, this list of conditions and the following disclaimer.
+ #   * . Redistributions in binary form must reproduce the above copyright
+ #   *   notice, this list of conditions and the following disclaimer in
+ #   *   the documentation and/or other materials provided with the
+ #   *   distribution.
+ #   * . Neither the name of the Computer Systems and Communication Lab
+ #   *   nor the names of its contributors may be used to endorse or
+ #   *   promote products derived from this software without specific
+ #   *   prior written permission.
+ #   *
+ #   * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ #   * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ #   * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ #   * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+ #   * REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
+ #   * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+ #   * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ #   * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ #   * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ #   * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ #   * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ #   * OF THE POSSIBILITY OF SUCH DAMAGE.
+ #   */
+ #
+ #  Copyright 1996 Chih-Hao Tsai @ Beckman Institute,
+ #      University of Illinois
+ #  c-tsai4@uiuc.edu  http://casper.beckman.uiuc.edu/~c-tsai4
+ #
+ #  ---------------COPYING.libtabe-----END--------------------------------
+ #
+ #
+ #  ---------------COPYING.ipadic-----BEGIN-------------------------------
+ #
+ #  Copyright 2000, 2001, 2002, 2003 Nara Institute of Science
+ #  and Technology.  All Rights Reserved.
+ #
+ #  Use, reproduction, and distribution of this software is permitted.
+ #  Any copy of this software, whether in its original form or modified,
+ #  must include both the above copyright notice and the following
+ #  paragraphs.
+ #
+ #  Nara Institute of Science and Technology (NAIST),
+ #  the copyright holders, disclaims all warranties with regard to this
+ #  software, including all implied warranties of merchantability and
+ #  fitness, in no event shall NAIST be liable for
+ #  any special, indirect or consequential damages or any damages
+ #  whatsoever resulting from loss of use, data or profits, whether in an
+ #  action of contract, negligence or other tortuous action, arising out
+ #  of or in connection with the use or performance of this software.
+ #
+ #  A large portion of the dictionary entries
+ #  originate from ICOT Free Software.  The following conditions for ICOT
+ #  Free Software applies to the current dictionary as well.
+ #
+ #  Each User may also freely distribute the Program, whether in its
+ #  original form or modified, to any third party or parties, PROVIDED
+ #  that the provisions of Section 3 ("NO WARRANTY") will ALWAYS appear
+ #  on, or be attached to, the Program, which is distributed substantially
+ #  in the same form as set out herein and that such intended
+ #  distribution, if actually made, will neither violate or otherwise
+ #  contravene any of the laws and regulations of the countries having
+ #  jurisdiction over the User or the intended distribution itself.
+ #
+ #  NO WARRANTY
+ #
+ #  The program was produced on an experimental basis in the course of the
+ #  research and development conducted during the project and is provided
+ #  to users as so produced on an experimental basis.  Accordingly, the
+ #  program is provided without any warranty whatsoever, whether express,
+ #  implied, statutory or otherwise.  The term "warranty" used herein
+ #  includes, but is not limited to, any warranty of the quality,
+ #  performance, merchantability and fitness for a particular purpose of
+ #  the program and the nonexistence of any infringement or violation of
+ #  any right of any third party.
+ #
+ #  Each user of the program will agree and understand, and be deemed to
+ #  have agreed and understood, that there is no warranty whatsoever for
+ #  the program and, accordingly, the entire risk arising from or
+ #  otherwise connected with the program is assumed by the user.
+ #
+ #  Therefore, neither ICOT, the copyright holder, or any other
+ #  organization that participated in or was otherwise related to the
+ #  development of the program and their respective officials, directors,
+ #  officers and other employees shall be held liable for any and all
+ #  damages, including, without limitation, general, special, incidental
+ #  and consequential damages, arising out of or otherwise in connection
+ #  with the use or inability to use the program or any product, material
+ #  or result produced or otherwise obtained by using the program,
+ #  regardless of whether they have been advised of, or otherwise had
+ #  knowledge of, the possibility of such damages at any time during the
+ #  project or thereafter.  Each user will be deemed to have agreed to the
+ #  foregoing by his or her commencement of use of the program.  The term
+ #  "use" as used herein includes, but is not limited to, the use,
+ #  modification, copying and distribution of the program and the
+ #  production of secondary products from the program.
+ #
+ #  In the case where the program, whether in its original form or
+ #  modified, was distributed or delivered to or received by a user from
+ #  any person, organization or entity other than ICOT, unless it makes or
+ #  grants independently of ICOT any specific warranty to the user in
+ #  writing, such person, organization or entity, will also be exempted
+ #  from and not be held liable to the user for any such damages as noted
+ #  above as far as the program is concerned.
+ #
+ #  ---------------COPYING.ipadic-----END----------------------------------
+
+3. Lao Word Break Dictionary Data (laodict.txt)
+
+ #  Copyright (c) 2013 International Business Machines Corporation
+ #  and others. All Rights Reserved.
+ #
+ # Project: http://code.google.com/p/lao-dictionary/
+ # Dictionary: http://lao-dictionary.googlecode.com/git/Lao-Dictionary.txt
+ # License: http://lao-dictionary.googlecode.com/git/Lao-Dictionary-LICENSE.txt
+ #              (copied below)
+ #
+ #  This file is derived from the above dictionary, with slight
+ #  modifications.
+ #  ----------------------------------------------------------------------
+ #  Copyright (C) 2013 Brian Eugene Wilson, Robert Martin Campbell.
+ #  All rights reserved.
+ #
+ #  Redistribution and use in source and binary forms, with or without
+ #  modification,
+ #  are permitted provided that the following conditions are met:
+ #
+ #
+ # Redistributions of source code must retain the above copyright notice, this
+ #  list of conditions and the following disclaimer. Redistributions in
+ #  binary form must reproduce the above copyright notice, this list of
+ #  conditions and the following disclaimer in the documentation and/or
+ #  other materials provided with the distribution.
+ #
+ #
+ # THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ # "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ # LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ # FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+ # COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
+ # INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+ # (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ # SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ # HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ # STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ # ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ # OF THE POSSIBILITY OF SUCH DAMAGE.
+ #  --------------------------------------------------------------------------
+
+4. Burmese Word Break Dictionary Data (burmesedict.txt)
+
+ #  Copyright (c) 2014 International Business Machines Corporation
+ #  and others. All Rights Reserved.
+ #
+ #  This list is part of a project hosted at:
+ #    github.com/kanyawtech/myanmar-karen-word-lists
+ #
+ #  --------------------------------------------------------------------------
+ #  Copyright (c) 2013, LeRoy Benjamin Sharon
+ #  All rights reserved.
+ #
+ #  Redistribution and use in source and binary forms, with or without
+ #  modification, are permitted provided that the following conditions
+ #  are met: Redistributions of source code must retain the above
+ #  copyright notice, this list of conditions and the following
+ #  disclaimer.  Redistributions in binary form must reproduce the
+ #  above copyright notice, this list of conditions and the following
+ #  disclaimer in the documentation and/or other materials provided
+ #  with the distribution.
+ #
+ #    Neither the name Myanmar Karen Word Lists, nor the names of its
+ #    contributors may be used to endorse or promote products derived
+ #    from this software without specific prior written permission.
+ #
+ #  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
+ #  CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
+ #  INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+ #  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ #  DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS
+ #  BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+ #  EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
+ #  TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ #  DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
+ #  ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
+ #  TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF
+ #  THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ #  SUCH DAMAGE.
+ #  --------------------------------------------------------------------------
+
+5. Time Zone Database
+
+  ICU uses the public domain data and code derived from Time Zone
+Database for its time zone support. The ownership of the TZ database
+is explained in BCP 175: Procedure for Maintaining the Time Zone
+Database section 7.
+
+ # 7.  Database Ownership
+ #
+ #    The TZ database itself is not an IETF Contribution or an IETF
+ #    document.  Rather it is a pre-existing and regularly updated work
+ #    that is in the public domain, and is intended to remain in the
+ #    public domain.  Therefore, BCPs 78 [RFC5378] and 79 [RFC3979] do
+ #    not apply to the TZ Database or contributions that individuals make
+ #    to it.  Should any claims be made and substantiated against the TZ
+ #    Database, the organization that is providing the IANA
+ #    Considerations defined in this RFC, under the memorandum of
+ #    understanding with the IETF, currently ICANN, may act in accordance
+ #    with all competent court orders.  No ownership claims will be made
+ #    by ICANN or the IETF Trust on the database or the code.  Any person
+ #    making a contribution to the database or code waives all rights to
+ #    future claims in that contribution or in the TZ Database.
+
+6. Google double-conversion
+
+Copyright 2006-2011, the V8 project authors. All rights reserved.
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+    * Redistributions of source code must retain the above copyright
+      notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above
+      copyright notice, this list of conditions and the following
+      disclaimer in the documentation and/or other materials provided
+      with the distribution.
+    * Neither the name of Google Inc. nor the names of its
+      contributors may be used to endorse or promote products derived
+      from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

vendor/tree-sitter/src/unicode/README.md

@@ -0,0 +1,29 @@
+# ICU Parts
+
+This directory contains a small subset of files from the Unicode organization's [ICU repository](https://github.com/unicode-org/icu).
+
+### License
+
+The license for these files is contained in the `LICENSE` file within this directory.
+
+### Contents
+
+* Source files taken from the [`icu4c/source/common/unicode`](https://github.com/unicode-org/icu/tree/552b01f61127d30d6589aa4bf99468224979b661/icu4c/source/common/unicode) directory:
+  * `utf8.h`
+  * `utf16.h`
+  * `umachine.h`
+* Empty source files that are referenced by the above source files, but whose original contents in `libicu` are not needed:
+  * `ptypes.h`
+  * `urename.h`
+  * `utf.h`
+* `ICU_SHA` - File containing the Git SHA of the commit in the `icu` repository from which the files were obtained.
+* `LICENSE` - The license file from the [`icu4c`](https://github.com/unicode-org/icu/tree/552b01f61127d30d6589aa4bf99468224979b661/icu4c) directory of the `icu` repository.
+* `README.md` - This text file.
+
+### Updating ICU
+
+To incorporate changes from the upstream `icu` repository:
+
+* Update `ICU_SHA` with the new Git SHA.
+* Update `LICENSE` with the license text from the directory mentioned above.
+* Update `utf8.h`, `utf16.h`, and `umachine.h` with their new contents in the `icu` repository.

vendor/tree-sitter/src/unicode/ptypes.h

@@ -0,0 +1 @@
+// This file must exist in order for `utf8.h` and `utf16.h` to be used.

vendor/tree-sitter/src/unicode/umachine.h

@@ -0,0 +1,448 @@
+// © 2016 and later: Unicode, Inc. and others.
+// License & terms of use: http://www.unicode.org/copyright.html
+/*
+******************************************************************************
+*
+*   Copyright (C) 1999-2015, International Business Machines
+*   Corporation and others.  All Rights Reserved.
+*
+******************************************************************************
+*   file name:  umachine.h
+*   encoding:   UTF-8
+*   tab size:   8 (not used)
+*   indentation:4
+*
+*   created on: 1999sep13
+*   created by: Markus W. Scherer
+*
+*   This file defines basic types and constants for ICU to be
+*   platform-independent. umachine.h and utf.h are included into
+*   utypes.h to provide all the general definitions for ICU.
+*   All of these definitions used to be in utypes.h before
+*   the UTF-handling macros made this unmaintainable.
+*/
+
+#ifndef __UMACHINE_H__
+#define __UMACHINE_H__
+
+
+/**
+ * \file
+ * \brief Basic types and constants for UTF
+ *
+ * <h2> Basic types and constants for UTF </h2>
+ *   This file defines basic types and constants for utf.h to be
+ *   platform-independent. umachine.h and utf.h are included into
+ *   utypes.h to provide all the general definitions for ICU.
+ *   All of these definitions used to be in utypes.h before
+ *   the UTF-handling macros made this unmaintainable.
+ *
+ */
+/*==========================================================================*/
+/* Include platform-dependent definitions                                   */
+/* which are contained in the platform-specific file platform.h             */
+/*==========================================================================*/
+
+#include "unicode/ptypes.h" /* platform.h is included in ptypes.h */
+
+/*
+ * ANSI C headers:
+ * stddef.h defines wchar_t
+ */
+#include <stddef.h>
+
+/*==========================================================================*/
+/* For C wrappers, we use the symbol U_STABLE.                                */
+/* This works properly if the includer is C or C++.                         */
+/* Functions are declared   U_STABLE return-type U_EXPORT2 function-name()... */
+/*==========================================================================*/
+
+/**
+ * \def U_CFUNC
+ * This is used in a declaration of a library private ICU C function.
+ * @stable ICU 2.4
+ */
+
+/**
+ * \def U_CDECL_BEGIN
+ * This is used to begin a declaration of a library private ICU C API.
+ * @stable ICU 2.4
+ */
+
+/**
+ * \def U_CDECL_END
+ * This is used to end a declaration of a library private ICU C API
+ * @stable ICU 2.4
+ */
+
+#ifdef __cplusplus
+#   define U_CFUNC extern "C"
+#   define U_CDECL_BEGIN extern "C" {
+#   define U_CDECL_END   }
+#else
+#   define U_CFUNC extern
+#   define U_CDECL_BEGIN
+#   define U_CDECL_END
+#endif
+
+#ifndef U_ATTRIBUTE_DEPRECATED
+/**
+ * \def U_ATTRIBUTE_DEPRECATED
+ *  This is used for GCC specific attributes
+ * @internal
+ */
+#if U_GCC_MAJOR_MINOR >= 302
+#    define U_ATTRIBUTE_DEPRECATED __attribute__ ((deprecated))
+/**
+ * \def U_ATTRIBUTE_DEPRECATED
+ * This is used for Visual C++ specific attributes
+ * @internal
+ */
+#elif defined(_MSC_VER) && (_MSC_VER >= 1400)
+#    define U_ATTRIBUTE_DEPRECATED __declspec(deprecated)
+#else
+#    define U_ATTRIBUTE_DEPRECATED
+#endif
+#endif
+
+/** This is used to declare a function as a public ICU C API @stable ICU 2.0*/
+#define U_CAPI U_CFUNC U_EXPORT
+/** This is used to declare a function as a stable public ICU C API*/
+#define U_STABLE U_CAPI
+/** This is used to declare a function as a draft public ICU C API  */
+#define U_DRAFT  U_CAPI
+/** This is used to declare a function as a deprecated public ICU C API  */
+#define U_DEPRECATED U_CAPI U_ATTRIBUTE_DEPRECATED
+/** This is used to declare a function as an obsolete public ICU C API  */
+#define U_OBSOLETE U_CAPI
+/** This is used to declare a function as an internal ICU C API  */
+#define U_INTERNAL U_CAPI
+
+/**
+ * \def U_OVERRIDE
+ * Defined to the C++11 "override" keyword if available.
+ * Denotes a class or member which is an override of the base class.
+ * May result in an error if it applied to something not an override.
+ * @internal
+ */
+#ifndef U_OVERRIDE
+#define U_OVERRIDE override
+#endif
+
+/**
+ * \def U_FINAL
+ * Defined to the C++11 "final" keyword if available.
+ * Denotes a class or member which may not be overridden in subclasses.
+ * May result in an error if subclasses attempt to override.
+ * @internal
+ */
+#if !defined(U_FINAL) || defined(U_IN_DOXYGEN)
+#define U_FINAL final
+#endif
+
+// Before ICU 65, function-like, multi-statement ICU macros were just defined as
+// series of statements wrapped in { } blocks and the caller could choose to
+// either treat them as if they were actual functions and end the invocation
+// with a trailing ; creating an empty statement after the block or else omit
+// this trailing ; using the knowledge that the macro would expand to { }.
+//
+// But doing so doesn't work well with macros that look like functions and
+// compiler warnings about empty statements (ICU-20601) and ICU 65 therefore
+// switches to the standard solution of wrapping such macros in do { } while.
+//
+// This will however break existing code that depends on being able to invoke
+// these macros without a trailing ; so to be able to remain compatible with
+// such code the wrapper is itself defined as macros so that it's possible to
+// build ICU 65 and later with the old macro behaviour, like this:
+//
+// CPPFLAGS='-DUPRV_BLOCK_MACRO_BEGIN="" -DUPRV_BLOCK_MACRO_END=""'
+// runConfigureICU ...
+
+/**
+ * \def UPRV_BLOCK_MACRO_BEGIN
+ * Defined as the "do" keyword by default.
+ * @internal
+ */
+#ifndef UPRV_BLOCK_MACRO_BEGIN
+#define UPRV_BLOCK_MACRO_BEGIN do
+#endif
+
+/**
+ * \def UPRV_BLOCK_MACRO_END
+ * Defined as "while (FALSE)" by default.
+ * @internal
+ */
+#ifndef UPRV_BLOCK_MACRO_END
+#define UPRV_BLOCK_MACRO_END while (FALSE)
+#endif
+
+/*==========================================================================*/
+/* limits for int32_t etc., like in POSIX inttypes.h                        */
+/*==========================================================================*/
+
+#ifndef INT8_MIN
+/** The smallest value an 8 bit signed integer can hold @stable ICU 2.0 */
+#   define INT8_MIN        ((int8_t)(-128))
+#endif
+#ifndef INT16_MIN
+/** The smallest value a 16 bit signed integer can hold @stable ICU 2.0 */
+#   define INT16_MIN       ((int16_t)(-32767-1))
+#endif
+#ifndef INT32_MIN
+/** The smallest value a 32 bit signed integer can hold @stable ICU 2.0 */
+#   define INT32_MIN       ((int32_t)(-2147483647-1))
+#endif
+
+#ifndef INT8_MAX
+/** The largest value an 8 bit signed integer can hold @stable ICU 2.0 */
+#   define INT8_MAX        ((int8_t)(127))
+#endif
+#ifndef INT16_MAX
+/** The largest value a 16 bit signed integer can hold @stable ICU 2.0 */
+#   define INT16_MAX       ((int16_t)(32767))
+#endif
+#ifndef INT32_MAX
+/** The largest value a 32 bit signed integer can hold @stable ICU 2.0 */
+#   define INT32_MAX       ((int32_t)(2147483647))
+#endif
+
+#ifndef UINT8_MAX
+/** The largest value an 8 bit unsigned integer can hold @stable ICU 2.0 */
+#   define UINT8_MAX       ((uint8_t)(255U))
+#endif
+#ifndef UINT16_MAX
+/** The largest value a 16 bit unsigned integer can hold @stable ICU 2.0 */
+#   define UINT16_MAX      ((uint16_t)(65535U))
+#endif
+#ifndef UINT32_MAX
+/** The largest value a 32 bit unsigned integer can hold @stable ICU 2.0 */
+#   define UINT32_MAX      ((uint32_t)(4294967295U))
+#endif
+
+#if defined(U_INT64_T_UNAVAILABLE)
+# error int64_t is required for decimal format and rule-based number format.
+#else
+# ifndef INT64_C
+/**
+ * Provides a platform independent way to specify a signed 64-bit integer constant.
+ * note: may be wrong for some 64 bit platforms - ensure your compiler provides INT64_C
+ * @stable ICU 2.8
+ */
+#   define INT64_C(c) c ## LL
+# endif
+# ifndef UINT64_C
+/**
+ * Provides a platform independent way to specify an unsigned 64-bit integer constant.
+ * note: may be wrong for some 64 bit platforms - ensure your compiler provides UINT64_C
+ * @stable ICU 2.8
+ */
+#   define UINT64_C(c) c ## ULL
+# endif
+# ifndef U_INT64_MIN
+/** The smallest value a 64 bit signed integer can hold @stable ICU 2.8 */
+#     define U_INT64_MIN       ((int64_t)(INT64_C(-9223372036854775807)-1))
+# endif
+# ifndef U_INT64_MAX
+/** The largest value a 64 bit signed integer can hold @stable ICU 2.8 */
+#     define U_INT64_MAX       ((int64_t)(INT64_C(9223372036854775807)))
+# endif
+# ifndef U_UINT64_MAX
+/** The largest value a 64 bit unsigned integer can hold @stable ICU 2.8 */
+#     define U_UINT64_MAX      ((uint64_t)(UINT64_C(18446744073709551615)))
+# endif
+#endif
+
+/*==========================================================================*/
+/* Boolean data type                                                        */
+/*==========================================================================*/
+
+/** The ICU boolean type @stable ICU 2.0 */
+typedef int8_t UBool;
+
+#ifndef TRUE
+/** The TRUE value of a UBool @stable ICU 2.0 */
+#   define TRUE  1
+#endif
+#ifndef FALSE
+/** The FALSE value of a UBool @stable ICU 2.0 */
+#   define FALSE 0
+#endif
+
+
+/*==========================================================================*/
+/* Unicode data types                                                       */
+/*==========================================================================*/
+
+/* wchar_t-related definitions -------------------------------------------- */
+
+/*
+ * \def U_WCHAR_IS_UTF16
+ * Defined if wchar_t uses UTF-16.
+ *
+ * @stable ICU 2.0
+ */
+/*
+ * \def U_WCHAR_IS_UTF32
+ * Defined if wchar_t uses UTF-32.
+ *
+ * @stable ICU 2.0
+ */
+#if !defined(U_WCHAR_IS_UTF16) && !defined(U_WCHAR_IS_UTF32)
+#   ifdef __STDC_ISO_10646__
+#       if (U_SIZEOF_WCHAR_T==2)
+#           define U_WCHAR_IS_UTF16
+#       elif (U_SIZEOF_WCHAR_T==4)
+#           define  U_WCHAR_IS_UTF32
+#       endif
+#   elif defined __UCS2__
+#       if (U_PF_OS390 <= U_PLATFORM && U_PLATFORM <= U_PF_OS400) && (U_SIZEOF_WCHAR_T==2)
+#           define U_WCHAR_IS_UTF16
+#       endif
+#   elif defined(__UCS4__) || (U_PLATFORM == U_PF_OS400 && defined(__UTF32__))
+#       if (U_SIZEOF_WCHAR_T==4)
+#           define U_WCHAR_IS_UTF32
+#       endif
+#   elif U_PLATFORM_IS_DARWIN_BASED || (U_SIZEOF_WCHAR_T==4 && U_PLATFORM_IS_LINUX_BASED)
+#       define U_WCHAR_IS_UTF32
+#   elif U_PLATFORM_HAS_WIN32_API
+#       define U_WCHAR_IS_UTF16
+#   endif
+#endif
+
+/* UChar and UChar32 definitions -------------------------------------------- */
+
+/** Number of bytes in a UChar. @stable ICU 2.0 */
+#define U_SIZEOF_UCHAR 2
+
+/**
+ * \def U_CHAR16_IS_TYPEDEF
+ * If 1, then char16_t is a typedef and not a real type (yet)
+ * @internal
+ */
+#if (U_PLATFORM == U_PF_AIX) && defined(__cplusplus) &&(U_CPLUSPLUS_VERSION < 11)
+// for AIX, uchar.h needs to be included
+# include <uchar.h>
+# define U_CHAR16_IS_TYPEDEF 1
+#elif defined(_MSC_VER) && (_MSC_VER < 1900)
+// Versions of Visual Studio/MSVC below 2015 do not support char16_t as a real type,
+// and instead use a typedef.  https://msdn.microsoft.com/library/bb531344.aspx
+# define U_CHAR16_IS_TYPEDEF 1
+#else
+# define U_CHAR16_IS_TYPEDEF 0
+#endif
+
+
+/**
+ * \var UChar
+ *
+ * The base type for UTF-16 code units and pointers.
+ * Unsigned 16-bit integer.
+ * Starting with ICU 59, C++ API uses char16_t directly, while C API continues to use UChar.
+ *
+ * UChar is configurable by defining the macro UCHAR_TYPE
+ * on the preprocessor or compiler command line:
+ * -DUCHAR_TYPE=uint16_t or -DUCHAR_TYPE=wchar_t (if U_SIZEOF_WCHAR_T==2) etc.
+ * (The UCHAR_TYPE can also be \#defined earlier in this file, for outside the ICU library code.)
+ * This is for transitional use from application code that uses uint16_t or wchar_t for UTF-16.
+ *
+ * The default is UChar=char16_t.
+ *
+ * C++11 defines char16_t as bit-compatible with uint16_t, but as a distinct type.
+ *
+ * In C, char16_t is a simple typedef of uint_least16_t.
+ * ICU requires uint_least16_t=uint16_t for data memory mapping.
+ * On macOS, char16_t is not available because the uchar.h standard header is missing.
+ *
+ * @stable ICU 4.4
+ */
+
+#if 1
+    // #if 1 is normal. UChar defaults to char16_t in C++.
+    // For configuration testing of UChar=uint16_t temporarily change this to #if 0.
+    // The intltest Makefile #defines UCHAR_TYPE=char16_t,
+    // so we only #define it to uint16_t if it is undefined so far.
+#elif !defined(UCHAR_TYPE)
+#   define UCHAR_TYPE uint16_t
+#endif
+
+#if defined(U_COMBINED_IMPLEMENTATION) || defined(U_COMMON_IMPLEMENTATION) || \
+        defined(U_I18N_IMPLEMENTATION) || defined(U_IO_IMPLEMENTATION)
+    // Inside the ICU library code, never configurable.
+    typedef char16_t UChar;
+#elif defined(UCHAR_TYPE)
+    typedef UCHAR_TYPE UChar;
+#elif defined(__cplusplus)
+    typedef char16_t UChar;
+#else
+    typedef uint16_t UChar;
+#endif
+
+/**
+ * \var OldUChar
+ * Default ICU 58 definition of UChar.
+ * A base type for UTF-16 code units and pointers.
+ * Unsigned 16-bit integer.
+ *
+ * Define OldUChar to be wchar_t if that is 16 bits wide.
+ * If wchar_t is not 16 bits wide, then define UChar to be uint16_t.
+ *
+ * This makes the definition of OldUChar platform-dependent
+ * but allows direct string type compatibility with platforms with
+ * 16-bit wchar_t types.
+ *
+ * This is how UChar was defined in ICU 58, for transition convenience.
+ * Exception: ICU 58 UChar was defined to UCHAR_TYPE if that macro was defined.
+ * The current UChar responds to UCHAR_TYPE but OldUChar does not.
+ *
+ * @stable ICU 59
+ */
+#if U_SIZEOF_WCHAR_T==2
+    typedef wchar_t OldUChar;
+#elif defined(__CHAR16_TYPE__)
+    typedef __CHAR16_TYPE__ OldUChar;
+#else
+    typedef uint16_t OldUChar;
+#endif
+
+/**
+ * Define UChar32 as a type for single Unicode code points.
+ * UChar32 is a signed 32-bit integer (same as int32_t).
+ *
+ * The Unicode code point range is 0..0x10ffff.
+ * All other values (negative or >=0x110000) are illegal as Unicode code points.
+ * They may be used as sentinel values to indicate "done", "error"
+ * or similar non-code point conditions.
+ *
+ * Before ICU 2.4 (Jitterbug 2146), UChar32 was defined
+ * to be wchar_t if that is 32 bits wide (wchar_t may be signed or unsigned)
+ * or else to be uint32_t.
+ * That is, the definition of UChar32 was platform-dependent.
+ *
+ * @see U_SENTINEL
+ * @stable ICU 2.4
+ */
+typedef int32_t UChar32;
+
+/**
+ * This value is intended for sentinel values for APIs that
+ * (take or) return single code points (UChar32).
+ * It is outside of the Unicode code point range 0..0x10ffff.
+ *
+ * For example, a "done" or "error" value in a new API
+ * could be indicated with U_SENTINEL.
+ *
+ * ICU APIs designed before ICU 2.4 usually define service-specific "done"
+ * values, mostly 0xffff.
+ * Those may need to be distinguished from
+ * actual U+ffff text contents by calling functions like
+ * CharacterIterator::hasNext() or UnicodeString::length().
+ *
+ * @return -1
+ * @see UChar32
+ * @stable ICU 2.4
+ */
+#define U_SENTINEL (-1)
+
+#include "unicode/urename.h"
+
+#endif

vendor/tree-sitter/src/unicode/urename.h

@@ -0,0 +1 @@
+// This file must exist in order for `utf8.h` and `utf16.h` to be used.

vendor/tree-sitter/src/unicode/utf.h

@@ -0,0 +1 @@
+// This file must exist in order for `utf8.h` and `utf16.h` to be used.

vendor/tree-sitter/src/unicode/utf16.h

@@ -0,0 +1,733 @@
+// © 2016 and later: Unicode, Inc. and others.
+// License & terms of use: http://www.unicode.org/copyright.html
+/*
+*******************************************************************************
+*
+*   Copyright (C) 1999-2012, International Business Machines
+*   Corporation and others.  All Rights Reserved.
+*
+*******************************************************************************
+*   file name:  utf16.h
+*   encoding:   UTF-8
+*   tab size:   8 (not used)
+*   indentation:4
+*
+*   created on: 1999sep09
+*   created by: Markus W. Scherer
+*/
+
+/**
+ * \file
+ * \brief C API: 16-bit Unicode handling macros
+ * 
+ * This file defines macros to deal with 16-bit Unicode (UTF-16) code units and strings.
+ *
+ * For more information see utf.h and the ICU User Guide Strings chapter
+ * (http://userguide.icu-project.org/strings).
+ *
+ * <em>Usage:</em>
+ * ICU coding guidelines for if() statements should be followed when using these macros.
+ * Compound statements (curly braces {}) must be used  for if-else-while... 
+ * bodies and all macro statements should be terminated with semicolon.
+ */
+
+#ifndef __UTF16_H__
+#define __UTF16_H__
+
+#include "unicode/umachine.h"
+#ifndef __UTF_H__
+#   include "unicode/utf.h"
+#endif
+
+/* single-code point definitions -------------------------------------------- */
+
+/**
+ * Does this code unit alone encode a code point (BMP, not a surrogate)?
+ * @param c 16-bit code unit
+ * @return TRUE or FALSE
+ * @stable ICU 2.4
+ */
+#define U16_IS_SINGLE(c) !U_IS_SURROGATE(c)
+
+/**
+ * Is this code unit a lead surrogate (U+d800..U+dbff)?
+ * @param c 16-bit code unit
+ * @return TRUE or FALSE
+ * @stable ICU 2.4
+ */
+#define U16_IS_LEAD(c) (((c)&0xfffffc00)==0xd800)
+
+/**
+ * Is this code unit a trail surrogate (U+dc00..U+dfff)?
+ * @param c 16-bit code unit
+ * @return TRUE or FALSE
+ * @stable ICU 2.4
+ */
+#define U16_IS_TRAIL(c) (((c)&0xfffffc00)==0xdc00)
+
+/**
+ * Is this code unit a surrogate (U+d800..U+dfff)?
+ * @param c 16-bit code unit
+ * @return TRUE or FALSE
+ * @stable ICU 2.4
+ */
+#define U16_IS_SURROGATE(c) U_IS_SURROGATE(c)
+
+/**
+ * Assuming c is a surrogate code point (U16_IS_SURROGATE(c)),
+ * is it a lead surrogate?
+ * @param c 16-bit code unit
+ * @return TRUE or FALSE
+ * @stable ICU 2.4
+ */
+#define U16_IS_SURROGATE_LEAD(c) (((c)&0x400)==0)
+
+/**
+ * Assuming c is a surrogate code point (U16_IS_SURROGATE(c)),
+ * is it a trail surrogate?
+ * @param c 16-bit code unit
+ * @return TRUE or FALSE
+ * @stable ICU 4.2
+ */
+#define U16_IS_SURROGATE_TRAIL(c) (((c)&0x400)!=0)
+
+/**
+ * Helper constant for U16_GET_SUPPLEMENTARY.
+ * @internal
+ */
+#define U16_SURROGATE_OFFSET ((0xd800<<10UL)+0xdc00-0x10000)
+
+/**
+ * Get a supplementary code point value (U+10000..U+10ffff)
+ * from its lead and trail surrogates.
+ * The result is undefined if the input values are not
+ * lead and trail surrogates.
+ *
+ * @param lead lead surrogate (U+d800..U+dbff)
+ * @param trail trail surrogate (U+dc00..U+dfff)
+ * @return supplementary code point (U+10000..U+10ffff)
+ * @stable ICU 2.4
+ */
+#define U16_GET_SUPPLEMENTARY(lead, trail) \
+    (((UChar32)(lead)<<10UL)+(UChar32)(trail)-U16_SURROGATE_OFFSET)
+
+
+/**
+ * Get the lead surrogate (0xd800..0xdbff) for a
+ * supplementary code point (0x10000..0x10ffff).
+ * @param supplementary 32-bit code point (U+10000..U+10ffff)
+ * @return lead surrogate (U+d800..U+dbff) for supplementary
+ * @stable ICU 2.4
+ */
+#define U16_LEAD(supplementary) (UChar)(((supplementary)>>10)+0xd7c0)
+
+/**
+ * Get the trail surrogate (0xdc00..0xdfff) for a
+ * supplementary code point (0x10000..0x10ffff).
+ * @param supplementary 32-bit code point (U+10000..U+10ffff)
+ * @return trail surrogate (U+dc00..U+dfff) for supplementary
+ * @stable ICU 2.4
+ */
+#define U16_TRAIL(supplementary) (UChar)(((supplementary)&0x3ff)|0xdc00)
+
+/**
+ * How many 16-bit code units are used to encode this Unicode code point? (1 or 2)
+ * The result is not defined if c is not a Unicode code point (U+0000..U+10ffff).
+ * @param c 32-bit code point
+ * @return 1 or 2
+ * @stable ICU 2.4
+ */
+#define U16_LENGTH(c) ((uint32_t)(c)<=0xffff ? 1 : 2)
+
+/**
+ * The maximum number of 16-bit code units per Unicode code point (U+0000..U+10ffff).
+ * @return 2
+ * @stable ICU 2.4
+ */
+#define U16_MAX_LENGTH 2
+
+/**
+ * Get a code point from a string at a random-access offset,
+ * without changing the offset.
+ * "Unsafe" macro, assumes well-formed UTF-16.
+ *
+ * The offset may point to either the lead or trail surrogate unit
+ * for a supplementary code point, in which case the macro will read
+ * the adjacent matching surrogate as well.
+ * The result is undefined if the offset points to a single, unpaired surrogate.
+ * Iteration through a string is more efficient with U16_NEXT_UNSAFE or U16_NEXT.
+ *
+ * @param s const UChar * string
+ * @param i string offset
+ * @param c output UChar32 variable
+ * @see U16_GET
+ * @stable ICU 2.4
+ */
+#define U16_GET_UNSAFE(s, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(s)[i]; \
+    if(U16_IS_SURROGATE(c)) { \
+        if(U16_IS_SURROGATE_LEAD(c)) { \
+            (c)=U16_GET_SUPPLEMENTARY((c), (s)[(i)+1]); \
+        } else { \
+            (c)=U16_GET_SUPPLEMENTARY((s)[(i)-1], (c)); \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Get a code point from a string at a random-access offset,
+ * without changing the offset.
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * The offset may point to either the lead or trail surrogate unit
+ * for a supplementary code point, in which case the macro will read
+ * the adjacent matching surrogate as well.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * If the offset points to a single, unpaired surrogate, then
+ * c is set to that unpaired surrogate.
+ * Iteration through a string is more efficient with U16_NEXT_UNSAFE or U16_NEXT.
+ *
+ * @param s const UChar * string
+ * @param start starting string offset (usually 0)
+ * @param i string offset, must be start<=i<length
+ * @param length string length
+ * @param c output UChar32 variable
+ * @see U16_GET_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U16_GET(s, start, i, length, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(s)[i]; \
+    if(U16_IS_SURROGATE(c)) { \
+        uint16_t __c2; \
+        if(U16_IS_SURROGATE_LEAD(c)) { \
+            if((i)+1!=(length) && U16_IS_TRAIL(__c2=(s)[(i)+1])) { \
+                (c)=U16_GET_SUPPLEMENTARY((c), __c2); \
+            } \
+        } else { \
+            if((i)>(start) && U16_IS_LEAD(__c2=(s)[(i)-1])) { \
+                (c)=U16_GET_SUPPLEMENTARY(__c2, (c)); \
+            } \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Get a code point from a string at a random-access offset,
+ * without changing the offset.
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * The offset may point to either the lead or trail surrogate unit
+ * for a supplementary code point, in which case the macro will read
+ * the adjacent matching surrogate as well.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * If the offset points to a single, unpaired surrogate, then
+ * c is set to U+FFFD.
+ * Iteration through a string is more efficient with U16_NEXT_UNSAFE or U16_NEXT_OR_FFFD.
+ *
+ * @param s const UChar * string
+ * @param start starting string offset (usually 0)
+ * @param i string offset, must be start<=i<length
+ * @param length string length
+ * @param c output UChar32 variable
+ * @see U16_GET_UNSAFE
+ * @stable ICU 60
+ */
+#define U16_GET_OR_FFFD(s, start, i, length, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(s)[i]; \
+    if(U16_IS_SURROGATE(c)) { \
+        uint16_t __c2; \
+        if(U16_IS_SURROGATE_LEAD(c)) { \
+            if((i)+1!=(length) && U16_IS_TRAIL(__c2=(s)[(i)+1])) { \
+                (c)=U16_GET_SUPPLEMENTARY((c), __c2); \
+            } else { \
+                (c)=0xfffd; \
+            } \
+        } else { \
+            if((i)>(start) && U16_IS_LEAD(__c2=(s)[(i)-1])) { \
+                (c)=U16_GET_SUPPLEMENTARY(__c2, (c)); \
+            } else { \
+                (c)=0xfffd; \
+            } \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/* definitions with forward iteration --------------------------------------- */
+
+/**
+ * Get a code point from a string at a code point boundary offset,
+ * and advance the offset to the next code point boundary.
+ * (Post-incrementing forward iteration.)
+ * "Unsafe" macro, assumes well-formed UTF-16.
+ *
+ * The offset may point to the lead surrogate unit
+ * for a supplementary code point, in which case the macro will read
+ * the following trail surrogate as well.
+ * If the offset points to a trail surrogate, then that itself
+ * will be returned as the code point.
+ * The result is undefined if the offset points to a single, unpaired lead surrogate.
+ *
+ * @param s const UChar * string
+ * @param i string offset
+ * @param c output UChar32 variable
+ * @see U16_NEXT
+ * @stable ICU 2.4
+ */
+#define U16_NEXT_UNSAFE(s, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(s)[(i)++]; \
+    if(U16_IS_LEAD(c)) { \
+        (c)=U16_GET_SUPPLEMENTARY((c), (s)[(i)++]); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Get a code point from a string at a code point boundary offset,
+ * and advance the offset to the next code point boundary.
+ * (Post-incrementing forward iteration.)
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * The offset may point to the lead surrogate unit
+ * for a supplementary code point, in which case the macro will read
+ * the following trail surrogate as well.
+ * If the offset points to a trail surrogate or
+ * to a single, unpaired lead surrogate, then c is set to that unpaired surrogate.
+ *
+ * @param s const UChar * string
+ * @param i string offset, must be i<length
+ * @param length string length
+ * @param c output UChar32 variable
+ * @see U16_NEXT_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U16_NEXT(s, i, length, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(s)[(i)++]; \
+    if(U16_IS_LEAD(c)) { \
+        uint16_t __c2; \
+        if((i)!=(length) && U16_IS_TRAIL(__c2=(s)[(i)])) { \
+            ++(i); \
+            (c)=U16_GET_SUPPLEMENTARY((c), __c2); \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Get a code point from a string at a code point boundary offset,
+ * and advance the offset to the next code point boundary.
+ * (Post-incrementing forward iteration.)
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * The offset may point to the lead surrogate unit
+ * for a supplementary code point, in which case the macro will read
+ * the following trail surrogate as well.
+ * If the offset points to a trail surrogate or
+ * to a single, unpaired lead surrogate, then c is set to U+FFFD.
+ *
+ * @param s const UChar * string
+ * @param i string offset, must be i<length
+ * @param length string length
+ * @param c output UChar32 variable
+ * @see U16_NEXT_UNSAFE
+ * @stable ICU 60
+ */
+#define U16_NEXT_OR_FFFD(s, i, length, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(s)[(i)++]; \
+    if(U16_IS_SURROGATE(c)) { \
+        uint16_t __c2; \
+        if(U16_IS_SURROGATE_LEAD(c) && (i)!=(length) && U16_IS_TRAIL(__c2=(s)[(i)])) { \
+            ++(i); \
+            (c)=U16_GET_SUPPLEMENTARY((c), __c2); \
+        } else { \
+            (c)=0xfffd; \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Append a code point to a string, overwriting 1 or 2 code units.
+ * The offset points to the current end of the string contents
+ * and is advanced (post-increment).
+ * "Unsafe" macro, assumes a valid code point and sufficient space in the string.
+ * Otherwise, the result is undefined.
+ *
+ * @param s const UChar * string buffer
+ * @param i string offset
+ * @param c code point to append
+ * @see U16_APPEND
+ * @stable ICU 2.4
+ */
+#define U16_APPEND_UNSAFE(s, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    if((uint32_t)(c)<=0xffff) { \
+        (s)[(i)++]=(uint16_t)(c); \
+    } else { \
+        (s)[(i)++]=(uint16_t)(((c)>>10)+0xd7c0); \
+        (s)[(i)++]=(uint16_t)(((c)&0x3ff)|0xdc00); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Append a code point to a string, overwriting 1 or 2 code units.
+ * The offset points to the current end of the string contents
+ * and is advanced (post-increment).
+ * "Safe" macro, checks for a valid code point.
+ * If a surrogate pair is written, checks for sufficient space in the string.
+ * If the code point is not valid or a trail surrogate does not fit,
+ * then isError is set to TRUE.
+ *
+ * @param s const UChar * string buffer
+ * @param i string offset, must be i<capacity
+ * @param capacity size of the string buffer
+ * @param c code point to append
+ * @param isError output UBool set to TRUE if an error occurs, otherwise not modified
+ * @see U16_APPEND_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U16_APPEND(s, i, capacity, c, isError) UPRV_BLOCK_MACRO_BEGIN { \
+    if((uint32_t)(c)<=0xffff) { \
+        (s)[(i)++]=(uint16_t)(c); \
+    } else if((uint32_t)(c)<=0x10ffff && (i)+1<(capacity)) { \
+        (s)[(i)++]=(uint16_t)(((c)>>10)+0xd7c0); \
+        (s)[(i)++]=(uint16_t)(((c)&0x3ff)|0xdc00); \
+    } else /* c>0x10ffff or not enough space */ { \
+        (isError)=TRUE; \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Advance the string offset from one code point boundary to the next.
+ * (Post-incrementing iteration.)
+ * "Unsafe" macro, assumes well-formed UTF-16.
+ *
+ * @param s const UChar * string
+ * @param i string offset
+ * @see U16_FWD_1
+ * @stable ICU 2.4
+ */
+#define U16_FWD_1_UNSAFE(s, i) UPRV_BLOCK_MACRO_BEGIN { \
+    if(U16_IS_LEAD((s)[(i)++])) { \
+        ++(i); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Advance the string offset from one code point boundary to the next.
+ * (Post-incrementing iteration.)
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * @param s const UChar * string
+ * @param i string offset, must be i<length
+ * @param length string length
+ * @see U16_FWD_1_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U16_FWD_1(s, i, length) UPRV_BLOCK_MACRO_BEGIN { \
+    if(U16_IS_LEAD((s)[(i)++]) && (i)!=(length) && U16_IS_TRAIL((s)[i])) { \
+        ++(i); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Advance the string offset from one code point boundary to the n-th next one,
+ * i.e., move forward by n code points.
+ * (Post-incrementing iteration.)
+ * "Unsafe" macro, assumes well-formed UTF-16.
+ *
+ * @param s const UChar * string
+ * @param i string offset
+ * @param n number of code points to skip
+ * @see U16_FWD_N
+ * @stable ICU 2.4
+ */
+#define U16_FWD_N_UNSAFE(s, i, n) UPRV_BLOCK_MACRO_BEGIN { \
+    int32_t __N=(n); \
+    while(__N>0) { \
+        U16_FWD_1_UNSAFE(s, i); \
+        --__N; \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Advance the string offset from one code point boundary to the n-th next one,
+ * i.e., move forward by n code points.
+ * (Post-incrementing iteration.)
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * @param s const UChar * string
+ * @param i int32_t string offset, must be i<length
+ * @param length int32_t string length
+ * @param n number of code points to skip
+ * @see U16_FWD_N_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U16_FWD_N(s, i, length, n) UPRV_BLOCK_MACRO_BEGIN { \
+    int32_t __N=(n); \
+    while(__N>0 && ((i)<(length) || ((length)<0 && (s)[i]!=0))) { \
+        U16_FWD_1(s, i, length); \
+        --__N; \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Adjust a random-access offset to a code point boundary
+ * at the start of a code point.
+ * If the offset points to the trail surrogate of a surrogate pair,
+ * then the offset is decremented.
+ * Otherwise, it is not modified.
+ * "Unsafe" macro, assumes well-formed UTF-16.
+ *
+ * @param s const UChar * string
+ * @param i string offset
+ * @see U16_SET_CP_START
+ * @stable ICU 2.4
+ */
+#define U16_SET_CP_START_UNSAFE(s, i) UPRV_BLOCK_MACRO_BEGIN { \
+    if(U16_IS_TRAIL((s)[i])) { \
+        --(i); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Adjust a random-access offset to a code point boundary
+ * at the start of a code point.
+ * If the offset points to the trail surrogate of a surrogate pair,
+ * then the offset is decremented.
+ * Otherwise, it is not modified.
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * @param s const UChar * string
+ * @param start starting string offset (usually 0)
+ * @param i string offset, must be start<=i
+ * @see U16_SET_CP_START_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U16_SET_CP_START(s, start, i) UPRV_BLOCK_MACRO_BEGIN { \
+    if(U16_IS_TRAIL((s)[i]) && (i)>(start) && U16_IS_LEAD((s)[(i)-1])) { \
+        --(i); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/* definitions with backward iteration -------------------------------------- */
+
+/**
+ * Move the string offset from one code point boundary to the previous one
+ * and get the code point between them.
+ * (Pre-decrementing backward iteration.)
+ * "Unsafe" macro, assumes well-formed UTF-16.
+ *
+ * The input offset may be the same as the string length.
+ * If the offset is behind a trail surrogate unit
+ * for a supplementary code point, then the macro will read
+ * the preceding lead surrogate as well.
+ * If the offset is behind a lead surrogate, then that itself
+ * will be returned as the code point.
+ * The result is undefined if the offset is behind a single, unpaired trail surrogate.
+ *
+ * @param s const UChar * string
+ * @param i string offset
+ * @param c output UChar32 variable
+ * @see U16_PREV
+ * @stable ICU 2.4
+ */
+#define U16_PREV_UNSAFE(s, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(s)[--(i)]; \
+    if(U16_IS_TRAIL(c)) { \
+        (c)=U16_GET_SUPPLEMENTARY((s)[--(i)], (c)); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the previous one
+ * and get the code point between them.
+ * (Pre-decrementing backward iteration.)
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * The input offset may be the same as the string length.
+ * If the offset is behind a trail surrogate unit
+ * for a supplementary code point, then the macro will read
+ * the preceding lead surrogate as well.
+ * If the offset is behind a lead surrogate or behind a single, unpaired
+ * trail surrogate, then c is set to that unpaired surrogate.
+ *
+ * @param s const UChar * string
+ * @param start starting string offset (usually 0)
+ * @param i string offset, must be start<i
+ * @param c output UChar32 variable
+ * @see U16_PREV_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U16_PREV(s, start, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(s)[--(i)]; \
+    if(U16_IS_TRAIL(c)) { \
+        uint16_t __c2; \
+        if((i)>(start) && U16_IS_LEAD(__c2=(s)[(i)-1])) { \
+            --(i); \
+            (c)=U16_GET_SUPPLEMENTARY(__c2, (c)); \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the previous one
+ * and get the code point between them.
+ * (Pre-decrementing backward iteration.)
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * The input offset may be the same as the string length.
+ * If the offset is behind a trail surrogate unit
+ * for a supplementary code point, then the macro will read
+ * the preceding lead surrogate as well.
+ * If the offset is behind a lead surrogate or behind a single, unpaired
+ * trail surrogate, then c is set to U+FFFD.
+ *
+ * @param s const UChar * string
+ * @param start starting string offset (usually 0)
+ * @param i string offset, must be start<i
+ * @param c output UChar32 variable
+ * @see U16_PREV_UNSAFE
+ * @stable ICU 60
+ */
+#define U16_PREV_OR_FFFD(s, start, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(s)[--(i)]; \
+    if(U16_IS_SURROGATE(c)) { \
+        uint16_t __c2; \
+        if(U16_IS_SURROGATE_TRAIL(c) && (i)>(start) && U16_IS_LEAD(__c2=(s)[(i)-1])) { \
+            --(i); \
+            (c)=U16_GET_SUPPLEMENTARY(__c2, (c)); \
+        } else { \
+            (c)=0xfffd; \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the previous one.
+ * (Pre-decrementing backward iteration.)
+ * The input offset may be the same as the string length.
+ * "Unsafe" macro, assumes well-formed UTF-16.
+ *
+ * @param s const UChar * string
+ * @param i string offset
+ * @see U16_BACK_1
+ * @stable ICU 2.4
+ */
+#define U16_BACK_1_UNSAFE(s, i) UPRV_BLOCK_MACRO_BEGIN { \
+    if(U16_IS_TRAIL((s)[--(i)])) { \
+        --(i); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the previous one.
+ * (Pre-decrementing backward iteration.)
+ * The input offset may be the same as the string length.
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * @param s const UChar * string
+ * @param start starting string offset (usually 0)
+ * @param i string offset, must be start<i
+ * @see U16_BACK_1_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U16_BACK_1(s, start, i) UPRV_BLOCK_MACRO_BEGIN { \
+    if(U16_IS_TRAIL((s)[--(i)]) && (i)>(start) && U16_IS_LEAD((s)[(i)-1])) { \
+        --(i); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the n-th one before it,
+ * i.e., move backward by n code points.
+ * (Pre-decrementing backward iteration.)
+ * The input offset may be the same as the string length.
+ * "Unsafe" macro, assumes well-formed UTF-16.
+ *
+ * @param s const UChar * string
+ * @param i string offset
+ * @param n number of code points to skip
+ * @see U16_BACK_N
+ * @stable ICU 2.4
+ */
+#define U16_BACK_N_UNSAFE(s, i, n) UPRV_BLOCK_MACRO_BEGIN { \
+    int32_t __N=(n); \
+    while(__N>0) { \
+        U16_BACK_1_UNSAFE(s, i); \
+        --__N; \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the n-th one before it,
+ * i.e., move backward by n code points.
+ * (Pre-decrementing backward iteration.)
+ * The input offset may be the same as the string length.
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * @param s const UChar * string
+ * @param start start of string
+ * @param i string offset, must be start<i
+ * @param n number of code points to skip
+ * @see U16_BACK_N_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U16_BACK_N(s, start, i, n) UPRV_BLOCK_MACRO_BEGIN { \
+    int32_t __N=(n); \
+    while(__N>0 && (i)>(start)) { \
+        U16_BACK_1(s, start, i); \
+        --__N; \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Adjust a random-access offset to a code point boundary after a code point.
+ * If the offset is behind the lead surrogate of a surrogate pair,
+ * then the offset is incremented.
+ * Otherwise, it is not modified.
+ * The input offset may be the same as the string length.
+ * "Unsafe" macro, assumes well-formed UTF-16.
+ *
+ * @param s const UChar * string
+ * @param i string offset
+ * @see U16_SET_CP_LIMIT
+ * @stable ICU 2.4
+ */
+#define U16_SET_CP_LIMIT_UNSAFE(s, i) UPRV_BLOCK_MACRO_BEGIN { \
+    if(U16_IS_LEAD((s)[(i)-1])) { \
+        ++(i); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Adjust a random-access offset to a code point boundary after a code point.
+ * If the offset is behind the lead surrogate of a surrogate pair,
+ * then the offset is incremented.
+ * Otherwise, it is not modified.
+ * The input offset may be the same as the string length.
+ * "Safe" macro, handles unpaired surrogates and checks for string boundaries.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * @param s const UChar * string
+ * @param start int32_t starting string offset (usually 0)
+ * @param i int32_t string offset, start<=i<=length
+ * @param length int32_t string length
+ * @see U16_SET_CP_LIMIT_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U16_SET_CP_LIMIT(s, start, i, length) UPRV_BLOCK_MACRO_BEGIN { \
+    if((start)<(i) && ((i)<(length) || (length)<0) && U16_IS_LEAD((s)[(i)-1]) && U16_IS_TRAIL((s)[i])) { \
+        ++(i); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+#endif

vendor/tree-sitter/src/unicode/utf8.h

@@ -0,0 +1,881 @@
+// © 2016 and later: Unicode, Inc. and others.
+// License & terms of use: http://www.unicode.org/copyright.html
+/*
+*******************************************************************************
+*
+*   Copyright (C) 1999-2015, International Business Machines
+*   Corporation and others.  All Rights Reserved.
+*
+*******************************************************************************
+*   file name:  utf8.h
+*   encoding:   UTF-8
+*   tab size:   8 (not used)
+*   indentation:4
+*
+*   created on: 1999sep13
+*   created by: Markus W. Scherer
+*/
+
+/**
+ * \file
+ * \brief C API: 8-bit Unicode handling macros
+ *
+ * This file defines macros to deal with 8-bit Unicode (UTF-8) code units (bytes) and strings.
+ *
+ * For more information see utf.h and the ICU User Guide Strings chapter
+ * (http://userguide.icu-project.org/strings).
+ *
+ * <em>Usage:</em>
+ * ICU coding guidelines for if() statements should be followed when using these macros.
+ * Compound statements (curly braces {}) must be used  for if-else-while...
+ * bodies and all macro statements should be terminated with semicolon.
+ */
+
+#ifndef __UTF8_H__
+#define __UTF8_H__
+
+#include "unicode/umachine.h"
+#ifndef __UTF_H__
+#   include "unicode/utf.h"
+#endif
+
+/* internal definitions ----------------------------------------------------- */
+
+/**
+ * Counts the trail bytes for a UTF-8 lead byte.
+ * Returns 0 for 0..0xc1 as well as for 0xf5..0xff.
+ * leadByte might be evaluated multiple times.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is called by public macros in this file and thus must remain stable.
+ *
+ * @param leadByte The first byte of a UTF-8 sequence. Must be 0..0xff.
+ * @internal
+ */
+#define U8_COUNT_TRAIL_BYTES(leadByte) \
+    (U8_IS_LEAD(leadByte) ? \
+        ((uint8_t)(leadByte)>=0xe0)+((uint8_t)(leadByte)>=0xf0)+1 : 0)
+
+/**
+ * Counts the trail bytes for a UTF-8 lead byte of a valid UTF-8 sequence.
+ * Returns 0 for 0..0xc1. Undefined for 0xf5..0xff.
+ * leadByte might be evaluated multiple times.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is called by public macros in this file and thus must remain stable.
+ *
+ * @param leadByte The first byte of a UTF-8 sequence. Must be 0..0xff.
+ * @internal
+ */
+#define U8_COUNT_TRAIL_BYTES_UNSAFE(leadByte) \
+    (((uint8_t)(leadByte)>=0xc2)+((uint8_t)(leadByte)>=0xe0)+((uint8_t)(leadByte)>=0xf0))
+
+/**
+ * Mask a UTF-8 lead byte, leave only the lower bits that form part of the code point value.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is called by public macros in this file and thus must remain stable.
+ * @internal
+ */
+#define U8_MASK_LEAD_BYTE(leadByte, countTrailBytes) ((leadByte)&=(1<<(6-(countTrailBytes)))-1)
+
+/**
+ * Internal bit vector for 3-byte UTF-8 validity check, for use in U8_IS_VALID_LEAD3_AND_T1.
+ * Each bit indicates whether one lead byte + first trail byte pair starts a valid sequence.
+ * Lead byte E0..EF bits 3..0 are used as byte index,
+ * first trail byte bits 7..5 are used as bit index into that byte.
+ * @see U8_IS_VALID_LEAD3_AND_T1
+ * @internal
+ */
+#define U8_LEAD3_T1_BITS "\x20\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x30\x10\x30\x30"
+
+/**
+ * Internal 3-byte UTF-8 validity check.
+ * Non-zero if lead byte E0..EF and first trail byte 00..FF start a valid sequence.
+ * @internal
+ */
+#define U8_IS_VALID_LEAD3_AND_T1(lead, t1) (U8_LEAD3_T1_BITS[(lead)&0xf]&(1<<((uint8_t)(t1)>>5)))
+
+/**
+ * Internal bit vector for 4-byte UTF-8 validity check, for use in U8_IS_VALID_LEAD4_AND_T1.
+ * Each bit indicates whether one lead byte + first trail byte pair starts a valid sequence.
+ * First trail byte bits 7..4 are used as byte index,
+ * lead byte F0..F4 bits 2..0 are used as bit index into that byte.
+ * @see U8_IS_VALID_LEAD4_AND_T1
+ * @internal
+ */
+#define U8_LEAD4_T1_BITS "\x00\x00\x00\x00\x00\x00\x00\x00\x1E\x0F\x0F\x0F\x00\x00\x00\x00"
+
+/**
+ * Internal 4-byte UTF-8 validity check.
+ * Non-zero if lead byte F0..F4 and first trail byte 00..FF start a valid sequence.
+ * @internal
+ */
+#define U8_IS_VALID_LEAD4_AND_T1(lead, t1) (U8_LEAD4_T1_BITS[(uint8_t)(t1)>>4]&(1<<((lead)&7)))
+
+/**
+ * Function for handling "next code point" with error-checking.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
+ * file and thus must remain stable, and should not be hidden when other internal
+ * functions are hidden (otherwise public macros would fail to compile).
+ * @internal
+ */
+U_STABLE UChar32 U_EXPORT2
+utf8_nextCharSafeBody(const uint8_t *s, int32_t *pi, int32_t length, UChar32 c, UBool strict);
+
+/**
+ * Function for handling "append code point" with error-checking.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
+ * file and thus must remain stable, and should not be hidden when other internal
+ * functions are hidden (otherwise public macros would fail to compile).
+ * @internal
+ */
+U_STABLE int32_t U_EXPORT2
+utf8_appendCharSafeBody(uint8_t *s, int32_t i, int32_t length, UChar32 c, UBool *pIsError);
+
+/**
+ * Function for handling "previous code point" with error-checking.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
+ * file and thus must remain stable, and should not be hidden when other internal
+ * functions are hidden (otherwise public macros would fail to compile).
+ * @internal
+ */
+U_STABLE UChar32 U_EXPORT2
+utf8_prevCharSafeBody(const uint8_t *s, int32_t start, int32_t *pi, UChar32 c, UBool strict);
+
+/**
+ * Function for handling "skip backward one code point" with error-checking.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
+ * file and thus must remain stable, and should not be hidden when other internal
+ * functions are hidden (otherwise public macros would fail to compile).
+ * @internal
+ */
+U_STABLE int32_t U_EXPORT2
+utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
+
+/* single-code point definitions -------------------------------------------- */
+
+/**
+ * Does this code unit (byte) encode a code point by itself (US-ASCII 0..0x7f)?
+ * @param c 8-bit code unit (byte)
+ * @return TRUE or FALSE
+ * @stable ICU 2.4
+ */
+#define U8_IS_SINGLE(c) (((c)&0x80)==0)
+
+/**
+ * Is this code unit (byte) a UTF-8 lead byte? (0xC2..0xF4)
+ * @param c 8-bit code unit (byte)
+ * @return TRUE or FALSE
+ * @stable ICU 2.4
+ */
+#define U8_IS_LEAD(c) ((uint8_t)((c)-0xc2)<=0x32)
+// 0x32=0xf4-0xc2
+
+/**
+ * Is this code unit (byte) a UTF-8 trail byte? (0x80..0xBF)
+ * @param c 8-bit code unit (byte)
+ * @return TRUE or FALSE
+ * @stable ICU 2.4
+ */
+#define U8_IS_TRAIL(c) ((int8_t)(c)<-0x40)
+
+/**
+ * How many code units (bytes) are used for the UTF-8 encoding
+ * of this Unicode code point?
+ * @param c 32-bit code point
+ * @return 1..4, or 0 if c is a surrogate or not a Unicode code point
+ * @stable ICU 2.4
+ */
+#define U8_LENGTH(c) \
+    ((uint32_t)(c)<=0x7f ? 1 : \
+        ((uint32_t)(c)<=0x7ff ? 2 : \
+            ((uint32_t)(c)<=0xd7ff ? 3 : \
+                ((uint32_t)(c)<=0xdfff || (uint32_t)(c)>0x10ffff ? 0 : \
+                    ((uint32_t)(c)<=0xffff ? 3 : 4)\
+                ) \
+            ) \
+        ) \
+    )
+
+/**
+ * The maximum number of UTF-8 code units (bytes) per Unicode code point (U+0000..U+10ffff).
+ * @return 4
+ * @stable ICU 2.4
+ */
+#define U8_MAX_LENGTH 4
+
+/**
+ * Get a code point from a string at a random-access offset,
+ * without changing the offset.
+ * The offset may point to either the lead byte or one of the trail bytes
+ * for a code point, in which case the macro will read all of the bytes
+ * for the code point.
+ * The result is undefined if the offset points to an illegal UTF-8
+ * byte sequence.
+ * Iteration through a string is more efficient with U8_NEXT_UNSAFE or U8_NEXT.
+ *
+ * @param s const uint8_t * string
+ * @param i string offset
+ * @param c output UChar32 variable
+ * @see U8_GET
+ * @stable ICU 2.4
+ */
+#define U8_GET_UNSAFE(s, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    int32_t _u8_get_unsafe_index=(int32_t)(i); \
+    U8_SET_CP_START_UNSAFE(s, _u8_get_unsafe_index); \
+    U8_NEXT_UNSAFE(s, _u8_get_unsafe_index, c); \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Get a code point from a string at a random-access offset,
+ * without changing the offset.
+ * The offset may point to either the lead byte or one of the trail bytes
+ * for a code point, in which case the macro will read all of the bytes
+ * for the code point.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * If the offset points to an illegal UTF-8 byte sequence, then
+ * c is set to a negative value.
+ * Iteration through a string is more efficient with U8_NEXT_UNSAFE or U8_NEXT.
+ *
+ * @param s const uint8_t * string
+ * @param start int32_t starting string offset
+ * @param i int32_t string offset, must be start<=i<length
+ * @param length int32_t string length
+ * @param c output UChar32 variable, set to <0 in case of an error
+ * @see U8_GET_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U8_GET(s, start, i, length, c) UPRV_BLOCK_MACRO_BEGIN { \
+    int32_t _u8_get_index=(i); \
+    U8_SET_CP_START(s, start, _u8_get_index); \
+    U8_NEXT(s, _u8_get_index, length, c); \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Get a code point from a string at a random-access offset,
+ * without changing the offset.
+ * The offset may point to either the lead byte or one of the trail bytes
+ * for a code point, in which case the macro will read all of the bytes
+ * for the code point.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * If the offset points to an illegal UTF-8 byte sequence, then
+ * c is set to U+FFFD.
+ * Iteration through a string is more efficient with U8_NEXT_UNSAFE or U8_NEXT_OR_FFFD.
+ *
+ * This macro does not distinguish between a real U+FFFD in the text
+ * and U+FFFD returned for an ill-formed sequence.
+ * Use U8_GET() if that distinction is important.
+ *
+ * @param s const uint8_t * string
+ * @param start int32_t starting string offset
+ * @param i int32_t string offset, must be start<=i<length
+ * @param length int32_t string length
+ * @param c output UChar32 variable, set to U+FFFD in case of an error
+ * @see U8_GET
+ * @stable ICU 51
+ */
+#define U8_GET_OR_FFFD(s, start, i, length, c) UPRV_BLOCK_MACRO_BEGIN { \
+    int32_t _u8_get_index=(i); \
+    U8_SET_CP_START(s, start, _u8_get_index); \
+    U8_NEXT_OR_FFFD(s, _u8_get_index, length, c); \
+} UPRV_BLOCK_MACRO_END
+
+/* definitions with forward iteration --------------------------------------- */
+
+/**
+ * Get a code point from a string at a code point boundary offset,
+ * and advance the offset to the next code point boundary.
+ * (Post-incrementing forward iteration.)
+ * "Unsafe" macro, assumes well-formed UTF-8.
+ *
+ * The offset may point to the lead byte of a multi-byte sequence,
+ * in which case the macro will read the whole sequence.
+ * The result is undefined if the offset points to a trail byte
+ * or an illegal UTF-8 sequence.
+ *
+ * @param s const uint8_t * string
+ * @param i string offset
+ * @param c output UChar32 variable
+ * @see U8_NEXT
+ * @stable ICU 2.4
+ */
+#define U8_NEXT_UNSAFE(s, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(uint8_t)(s)[(i)++]; \
+    if(!U8_IS_SINGLE(c)) { \
+        if((c)<0xe0) { \
+            (c)=(((c)&0x1f)<<6)|((s)[(i)++]&0x3f); \
+        } else if((c)<0xf0) { \
+            /* no need for (c&0xf) because the upper bits are truncated after <<12 in the cast to (UChar) */ \
+            (c)=(UChar)(((c)<<12)|(((s)[i]&0x3f)<<6)|((s)[(i)+1]&0x3f)); \
+            (i)+=2; \
+        } else { \
+            (c)=(((c)&7)<<18)|(((s)[i]&0x3f)<<12)|(((s)[(i)+1]&0x3f)<<6)|((s)[(i)+2]&0x3f); \
+            (i)+=3; \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Get a code point from a string at a code point boundary offset,
+ * and advance the offset to the next code point boundary.
+ * (Post-incrementing forward iteration.)
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * The offset may point to the lead byte of a multi-byte sequence,
+ * in which case the macro will read the whole sequence.
+ * If the offset points to a trail byte or an illegal UTF-8 sequence, then
+ * c is set to a negative value.
+ *
+ * @param s const uint8_t * string
+ * @param i int32_t string offset, must be i<length
+ * @param length int32_t string length
+ * @param c output UChar32 variable, set to <0 in case of an error
+ * @see U8_NEXT_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U8_NEXT(s, i, length, c) U8_INTERNAL_NEXT_OR_SUB(s, i, length, c, U_SENTINEL)
+
+/**
+ * Get a code point from a string at a code point boundary offset,
+ * and advance the offset to the next code point boundary.
+ * (Post-incrementing forward iteration.)
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * The offset may point to the lead byte of a multi-byte sequence,
+ * in which case the macro will read the whole sequence.
+ * If the offset points to a trail byte or an illegal UTF-8 sequence, then
+ * c is set to U+FFFD.
+ *
+ * This macro does not distinguish between a real U+FFFD in the text
+ * and U+FFFD returned for an ill-formed sequence.
+ * Use U8_NEXT() if that distinction is important.
+ *
+ * @param s const uint8_t * string
+ * @param i int32_t string offset, must be i<length
+ * @param length int32_t string length
+ * @param c output UChar32 variable, set to U+FFFD in case of an error
+ * @see U8_NEXT
+ * @stable ICU 51
+ */
+#define U8_NEXT_OR_FFFD(s, i, length, c) U8_INTERNAL_NEXT_OR_SUB(s, i, length, c, 0xfffd)
+
+/** @internal */
+#define U8_INTERNAL_NEXT_OR_SUB(s, i, length, c, sub) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(uint8_t)(s)[(i)++]; \
+    if(!U8_IS_SINGLE(c)) { \
+        uint8_t __t = 0; \
+        if((i)!=(length) && \
+            /* fetch/validate/assemble all but last trail byte */ \
+            ((c)>=0xe0 ? \
+                ((c)<0xf0 ?  /* U+0800..U+FFFF except surrogates */ \
+                    U8_LEAD3_T1_BITS[(c)&=0xf]&(1<<((__t=(s)[i])>>5)) && \
+                    (__t&=0x3f, 1) \
+                :  /* U+10000..U+10FFFF */ \
+                    ((c)-=0xf0)<=4 && \
+                    U8_LEAD4_T1_BITS[(__t=(s)[i])>>4]&(1<<(c)) && \
+                    ((c)=((c)<<6)|(__t&0x3f), ++(i)!=(length)) && \
+                    (__t=(s)[i]-0x80)<=0x3f) && \
+                /* valid second-to-last trail byte */ \
+                ((c)=((c)<<6)|__t, ++(i)!=(length)) \
+            :  /* U+0080..U+07FF */ \
+                (c)>=0xc2 && ((c)&=0x1f, 1)) && \
+            /* last trail byte */ \
+            (__t=(s)[i]-0x80)<=0x3f && \
+            ((c)=((c)<<6)|__t, ++(i), 1)) { \
+        } else { \
+            (c)=(sub);  /* ill-formed*/ \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Append a code point to a string, overwriting 1 to 4 bytes.
+ * The offset points to the current end of the string contents
+ * and is advanced (post-increment).
+ * "Unsafe" macro, assumes a valid code point and sufficient space in the string.
+ * Otherwise, the result is undefined.
+ *
+ * @param s const uint8_t * string buffer
+ * @param i string offset
+ * @param c code point to append
+ * @see U8_APPEND
+ * @stable ICU 2.4
+ */
+#define U8_APPEND_UNSAFE(s, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    uint32_t __uc=(c); \
+    if(__uc<=0x7f) { \
+        (s)[(i)++]=(uint8_t)__uc; \
+    } else { \
+        if(__uc<=0x7ff) { \
+            (s)[(i)++]=(uint8_t)((__uc>>6)|0xc0); \
+        } else { \
+            if(__uc<=0xffff) { \
+                (s)[(i)++]=(uint8_t)((__uc>>12)|0xe0); \
+            } else { \
+                (s)[(i)++]=(uint8_t)((__uc>>18)|0xf0); \
+                (s)[(i)++]=(uint8_t)(((__uc>>12)&0x3f)|0x80); \
+            } \
+            (s)[(i)++]=(uint8_t)(((__uc>>6)&0x3f)|0x80); \
+        } \
+        (s)[(i)++]=(uint8_t)((__uc&0x3f)|0x80); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Append a code point to a string, overwriting 1 to 4 bytes.
+ * The offset points to the current end of the string contents
+ * and is advanced (post-increment).
+ * "Safe" macro, checks for a valid code point.
+ * If a non-ASCII code point is written, checks for sufficient space in the string.
+ * If the code point is not valid or trail bytes do not fit,
+ * then isError is set to TRUE.
+ *
+ * @param s const uint8_t * string buffer
+ * @param i int32_t string offset, must be i<capacity
+ * @param capacity int32_t size of the string buffer
+ * @param c UChar32 code point to append
+ * @param isError output UBool set to TRUE if an error occurs, otherwise not modified
+ * @see U8_APPEND_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U8_APPEND(s, i, capacity, c, isError) UPRV_BLOCK_MACRO_BEGIN { \
+    uint32_t __uc=(c); \
+    if(__uc<=0x7f) { \
+        (s)[(i)++]=(uint8_t)__uc; \
+    } else if(__uc<=0x7ff && (i)+1<(capacity)) { \
+        (s)[(i)++]=(uint8_t)((__uc>>6)|0xc0); \
+        (s)[(i)++]=(uint8_t)((__uc&0x3f)|0x80); \
+    } else if((__uc<=0xd7ff || (0xe000<=__uc && __uc<=0xffff)) && (i)+2<(capacity)) { \
+        (s)[(i)++]=(uint8_t)((__uc>>12)|0xe0); \
+        (s)[(i)++]=(uint8_t)(((__uc>>6)&0x3f)|0x80); \
+        (s)[(i)++]=(uint8_t)((__uc&0x3f)|0x80); \
+    } else if(0xffff<__uc && __uc<=0x10ffff && (i)+3<(capacity)) { \
+        (s)[(i)++]=(uint8_t)((__uc>>18)|0xf0); \
+        (s)[(i)++]=(uint8_t)(((__uc>>12)&0x3f)|0x80); \
+        (s)[(i)++]=(uint8_t)(((__uc>>6)&0x3f)|0x80); \
+        (s)[(i)++]=(uint8_t)((__uc&0x3f)|0x80); \
+    } else { \
+        (isError)=TRUE; \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Advance the string offset from one code point boundary to the next.
+ * (Post-incrementing iteration.)
+ * "Unsafe" macro, assumes well-formed UTF-8.
+ *
+ * @param s const uint8_t * string
+ * @param i string offset
+ * @see U8_FWD_1
+ * @stable ICU 2.4
+ */
+#define U8_FWD_1_UNSAFE(s, i) UPRV_BLOCK_MACRO_BEGIN { \
+    (i)+=1+U8_COUNT_TRAIL_BYTES_UNSAFE((s)[i]); \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Advance the string offset from one code point boundary to the next.
+ * (Post-incrementing iteration.)
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * @param s const uint8_t * string
+ * @param i int32_t string offset, must be i<length
+ * @param length int32_t string length
+ * @see U8_FWD_1_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U8_FWD_1(s, i, length) UPRV_BLOCK_MACRO_BEGIN { \
+    uint8_t __b=(s)[(i)++]; \
+    if(U8_IS_LEAD(__b) && (i)!=(length)) { \
+        uint8_t __t1=(s)[i]; \
+        if((0xe0<=__b && __b<0xf0)) { \
+            if(U8_IS_VALID_LEAD3_AND_T1(__b, __t1) && \
+                    ++(i)!=(length) && U8_IS_TRAIL((s)[i])) { \
+                ++(i); \
+            } \
+        } else if(__b<0xe0) { \
+            if(U8_IS_TRAIL(__t1)) { \
+                ++(i); \
+            } \
+        } else /* c>=0xf0 */ { \
+            if(U8_IS_VALID_LEAD4_AND_T1(__b, __t1) && \
+                    ++(i)!=(length) && U8_IS_TRAIL((s)[i]) && \
+                    ++(i)!=(length) && U8_IS_TRAIL((s)[i])) { \
+                ++(i); \
+            } \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Advance the string offset from one code point boundary to the n-th next one,
+ * i.e., move forward by n code points.
+ * (Post-incrementing iteration.)
+ * "Unsafe" macro, assumes well-formed UTF-8.
+ *
+ * @param s const uint8_t * string
+ * @param i string offset
+ * @param n number of code points to skip
+ * @see U8_FWD_N
+ * @stable ICU 2.4
+ */
+#define U8_FWD_N_UNSAFE(s, i, n) UPRV_BLOCK_MACRO_BEGIN { \
+    int32_t __N=(n); \
+    while(__N>0) { \
+        U8_FWD_1_UNSAFE(s, i); \
+        --__N; \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Advance the string offset from one code point boundary to the n-th next one,
+ * i.e., move forward by n code points.
+ * (Post-incrementing iteration.)
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * @param s const uint8_t * string
+ * @param i int32_t string offset, must be i<length
+ * @param length int32_t string length
+ * @param n number of code points to skip
+ * @see U8_FWD_N_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U8_FWD_N(s, i, length, n) UPRV_BLOCK_MACRO_BEGIN { \
+    int32_t __N=(n); \
+    while(__N>0 && ((i)<(length) || ((length)<0 && (s)[i]!=0))) { \
+        U8_FWD_1(s, i, length); \
+        --__N; \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Adjust a random-access offset to a code point boundary
+ * at the start of a code point.
+ * If the offset points to a UTF-8 trail byte,
+ * then the offset is moved backward to the corresponding lead byte.
+ * Otherwise, it is not modified.
+ * "Unsafe" macro, assumes well-formed UTF-8.
+ *
+ * @param s const uint8_t * string
+ * @param i string offset
+ * @see U8_SET_CP_START
+ * @stable ICU 2.4
+ */
+#define U8_SET_CP_START_UNSAFE(s, i) UPRV_BLOCK_MACRO_BEGIN { \
+    while(U8_IS_TRAIL((s)[i])) { --(i); } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Adjust a random-access offset to a code point boundary
+ * at the start of a code point.
+ * If the offset points to a UTF-8 trail byte,
+ * then the offset is moved backward to the corresponding lead byte.
+ * Otherwise, it is not modified.
+ *
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ * Unlike U8_TRUNCATE_IF_INCOMPLETE(), this macro always reads s[i].
+ *
+ * @param s const uint8_t * string
+ * @param start int32_t starting string offset (usually 0)
+ * @param i int32_t string offset, must be start<=i
+ * @see U8_SET_CP_START_UNSAFE
+ * @see U8_TRUNCATE_IF_INCOMPLETE
+ * @stable ICU 2.4
+ */
+#define U8_SET_CP_START(s, start, i) UPRV_BLOCK_MACRO_BEGIN { \
+    if(U8_IS_TRAIL((s)[(i)])) { \
+        (i)=utf8_back1SafeBody(s, start, (i)); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * If the string ends with a UTF-8 byte sequence that is valid so far
+ * but incomplete, then reduce the length of the string to end before
+ * the lead byte of that incomplete sequence.
+ * For example, if the string ends with E1 80, the length is reduced by 2.
+ *
+ * In all other cases (the string ends with a complete sequence, or it is not
+ * possible for any further trail byte to extend the trailing sequence)
+ * the length remains unchanged.
+ *
+ * Useful for processing text split across multiple buffers
+ * (save the incomplete sequence for later)
+ * and for optimizing iteration
+ * (check for string length only once per character).
+ *
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ * Unlike U8_SET_CP_START(), this macro never reads s[length].
+ *
+ * (In UTF-16, simply check for U16_IS_LEAD(last code unit).)
+ *
+ * @param s const uint8_t * string
+ * @param start int32_t starting string offset (usually 0)
+ * @param length int32_t string length (usually start<=length)
+ * @see U8_SET_CP_START
+ * @stable ICU 61
+ */
+#define U8_TRUNCATE_IF_INCOMPLETE(s, start, length) UPRV_BLOCK_MACRO_BEGIN { \
+    if((length)>(start)) { \
+        uint8_t __b1=s[(length)-1]; \
+        if(U8_IS_SINGLE(__b1)) { \
+            /* common ASCII character */ \
+        } else if(U8_IS_LEAD(__b1)) { \
+            --(length); \
+        } else if(U8_IS_TRAIL(__b1) && ((length)-2)>=(start)) { \
+            uint8_t __b2=s[(length)-2]; \
+            if(0xe0<=__b2 && __b2<=0xf4) { \
+                if(__b2<0xf0 ? U8_IS_VALID_LEAD3_AND_T1(__b2, __b1) : \
+                        U8_IS_VALID_LEAD4_AND_T1(__b2, __b1)) { \
+                    (length)-=2; \
+                } \
+            } else if(U8_IS_TRAIL(__b2) && ((length)-3)>=(start)) { \
+                uint8_t __b3=s[(length)-3]; \
+                if(0xf0<=__b3 && __b3<=0xf4 && U8_IS_VALID_LEAD4_AND_T1(__b3, __b2)) { \
+                    (length)-=3; \
+                } \
+            } \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/* definitions with backward iteration -------------------------------------- */
+
+/**
+ * Move the string offset from one code point boundary to the previous one
+ * and get the code point between them.
+ * (Pre-decrementing backward iteration.)
+ * "Unsafe" macro, assumes well-formed UTF-8.
+ *
+ * The input offset may be the same as the string length.
+ * If the offset is behind a multi-byte sequence, then the macro will read
+ * the whole sequence.
+ * If the offset is behind a lead byte, then that itself
+ * will be returned as the code point.
+ * The result is undefined if the offset is behind an illegal UTF-8 sequence.
+ *
+ * @param s const uint8_t * string
+ * @param i string offset
+ * @param c output UChar32 variable
+ * @see U8_PREV
+ * @stable ICU 2.4
+ */
+#define U8_PREV_UNSAFE(s, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(uint8_t)(s)[--(i)]; \
+    if(U8_IS_TRAIL(c)) { \
+        uint8_t __b, __count=1, __shift=6; \
+\
+        /* c is a trail byte */ \
+        (c)&=0x3f; \
+        for(;;) { \
+            __b=(s)[--(i)]; \
+            if(__b>=0xc0) { \
+                U8_MASK_LEAD_BYTE(__b, __count); \
+                (c)|=(UChar32)__b<<__shift; \
+                break; \
+            } else { \
+                (c)|=(UChar32)(__b&0x3f)<<__shift; \
+                ++__count; \
+                __shift+=6; \
+            } \
+        } \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the previous one
+ * and get the code point between them.
+ * (Pre-decrementing backward iteration.)
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ *
+ * The input offset may be the same as the string length.
+ * If the offset is behind a multi-byte sequence, then the macro will read
+ * the whole sequence.
+ * If the offset is behind a lead byte, then that itself
+ * will be returned as the code point.
+ * If the offset is behind an illegal UTF-8 sequence, then c is set to a negative value.
+ *
+ * @param s const uint8_t * string
+ * @param start int32_t starting string offset (usually 0)
+ * @param i int32_t string offset, must be start<i
+ * @param c output UChar32 variable, set to <0 in case of an error
+ * @see U8_PREV_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U8_PREV(s, start, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(uint8_t)(s)[--(i)]; \
+    if(!U8_IS_SINGLE(c)) { \
+        (c)=utf8_prevCharSafeBody((const uint8_t *)s, start, &(i), c, -1); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the previous one
+ * and get the code point between them.
+ * (Pre-decrementing backward iteration.)
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ *
+ * The input offset may be the same as the string length.
+ * If the offset is behind a multi-byte sequence, then the macro will read
+ * the whole sequence.
+ * If the offset is behind a lead byte, then that itself
+ * will be returned as the code point.
+ * If the offset is behind an illegal UTF-8 sequence, then c is set to U+FFFD.
+ *
+ * This macro does not distinguish between a real U+FFFD in the text
+ * and U+FFFD returned for an ill-formed sequence.
+ * Use U8_PREV() if that distinction is important.
+ *
+ * @param s const uint8_t * string
+ * @param start int32_t starting string offset (usually 0)
+ * @param i int32_t string offset, must be start<i
+ * @param c output UChar32 variable, set to U+FFFD in case of an error
+ * @see U8_PREV
+ * @stable ICU 51
+ */
+#define U8_PREV_OR_FFFD(s, start, i, c) UPRV_BLOCK_MACRO_BEGIN { \
+    (c)=(uint8_t)(s)[--(i)]; \
+    if(!U8_IS_SINGLE(c)) { \
+        (c)=utf8_prevCharSafeBody((const uint8_t *)s, start, &(i), c, -3); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the previous one.
+ * (Pre-decrementing backward iteration.)
+ * The input offset may be the same as the string length.
+ * "Unsafe" macro, assumes well-formed UTF-8.
+ *
+ * @param s const uint8_t * string
+ * @param i string offset
+ * @see U8_BACK_1
+ * @stable ICU 2.4
+ */
+#define U8_BACK_1_UNSAFE(s, i) UPRV_BLOCK_MACRO_BEGIN { \
+    while(U8_IS_TRAIL((s)[--(i)])) {} \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the previous one.
+ * (Pre-decrementing backward iteration.)
+ * The input offset may be the same as the string length.
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ *
+ * @param s const uint8_t * string
+ * @param start int32_t starting string offset (usually 0)
+ * @param i int32_t string offset, must be start<i
+ * @see U8_BACK_1_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U8_BACK_1(s, start, i) UPRV_BLOCK_MACRO_BEGIN { \
+    if(U8_IS_TRAIL((s)[--(i)])) { \
+        (i)=utf8_back1SafeBody(s, start, (i)); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the n-th one before it,
+ * i.e., move backward by n code points.
+ * (Pre-decrementing backward iteration.)
+ * The input offset may be the same as the string length.
+ * "Unsafe" macro, assumes well-formed UTF-8.
+ *
+ * @param s const uint8_t * string
+ * @param i string offset
+ * @param n number of code points to skip
+ * @see U8_BACK_N
+ * @stable ICU 2.4
+ */
+#define U8_BACK_N_UNSAFE(s, i, n) UPRV_BLOCK_MACRO_BEGIN { \
+    int32_t __N=(n); \
+    while(__N>0) { \
+        U8_BACK_1_UNSAFE(s, i); \
+        --__N; \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Move the string offset from one code point boundary to the n-th one before it,
+ * i.e., move backward by n code points.
+ * (Pre-decrementing backward iteration.)
+ * The input offset may be the same as the string length.
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ *
+ * @param s const uint8_t * string
+ * @param start int32_t index of the start of the string
+ * @param i int32_t string offset, must be start<i
+ * @param n number of code points to skip
+ * @see U8_BACK_N_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U8_BACK_N(s, start, i, n) UPRV_BLOCK_MACRO_BEGIN { \
+    int32_t __N=(n); \
+    while(__N>0 && (i)>(start)) { \
+        U8_BACK_1(s, start, i); \
+        --__N; \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Adjust a random-access offset to a code point boundary after a code point.
+ * If the offset is behind a partial multi-byte sequence,
+ * then the offset is incremented to behind the whole sequence.
+ * Otherwise, it is not modified.
+ * The input offset may be the same as the string length.
+ * "Unsafe" macro, assumes well-formed UTF-8.
+ *
+ * @param s const uint8_t * string
+ * @param i string offset
+ * @see U8_SET_CP_LIMIT
+ * @stable ICU 2.4
+ */
+#define U8_SET_CP_LIMIT_UNSAFE(s, i) UPRV_BLOCK_MACRO_BEGIN { \
+    U8_BACK_1_UNSAFE(s, i); \
+    U8_FWD_1_UNSAFE(s, i); \
+} UPRV_BLOCK_MACRO_END
+
+/**
+ * Adjust a random-access offset to a code point boundary after a code point.
+ * If the offset is behind a partial multi-byte sequence,
+ * then the offset is incremented to behind the whole sequence.
+ * Otherwise, it is not modified.
+ * The input offset may be the same as the string length.
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * @param s const uint8_t * string
+ * @param start int32_t starting string offset (usually 0)
+ * @param i int32_t string offset, must be start<=i<=length
+ * @param length int32_t string length
+ * @see U8_SET_CP_LIMIT_UNSAFE
+ * @stable ICU 2.4
+ */
+#define U8_SET_CP_LIMIT(s, start, i, length) UPRV_BLOCK_MACRO_BEGIN { \
+    if((start)<(i) && ((i)<(length) || (length)<0)) { \
+        U8_BACK_1(s, start, i); \
+        U8_FWD_1(s, i, length); \
+    } \
+} UPRV_BLOCK_MACRO_END
+
+#endif

vendor/tree-sitter/src/wasm/stdlib-symbols.txt

@@ -0,0 +1,24 @@
+"calloc",
+"free",
+"iswalnum",
+"iswalpha",
+"iswblank",
+"iswdigit",
+"iswlower",
+"iswspace",
+"iswupper",
+"iswxdigit",
+"malloc",
+"memchr",
+"memcmp",
+"memcpy",
+"memmove",
+"memset",
+"realloc",
+"strcmp",
+"strlen",
+"strncat",
+"strncmp",
+"strncpy",
+"towlower",
+"towupper",

vendor/tree-sitter/src/wasm/stdlib.c

@@ -0,0 +1,109 @@
+// This file implements a very simple allocator for external scanners running
+// in WASM. Allocation is just bumping a static pointer and growing the heap
+// as needed, and freeing is mostly a noop. But in the special case of freeing
+// the last-allocated pointer, we'll reuse that pointer again.
+
+#include <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <string.h>
+
+extern void tree_sitter_debug_message(const char *, size_t);
+
+#define PAGESIZE 0x10000
+#define MAX_HEAP_SIZE (4 * 1024 * 1024)
+
+typedef struct {
+  size_t size;
+  char data[0];
+} Region;
+
+static Region *heap_end = NULL;
+static Region *heap_start = NULL;
+static Region *next = NULL;
+
+// Get the region metadata for the given heap pointer.
+static inline Region *region_for_ptr(void *ptr) {
+  return ((Region *)ptr) - 1;
+}
+
+// Get the location of the next region after the given region,
+// if the given region had the given size.
+static inline Region *region_after(Region *self, size_t len) {
+  char *address = self->data + len;
+  char *aligned = (char *)((uintptr_t)(address + 3) & ~0x3);
+  return (Region *)aligned;
+}
+
+static void *get_heap_end() {
+  return (void *)(__builtin_wasm_memory_size(0) * PAGESIZE);
+}
+
+static int grow_heap(size_t size) {
+  size_t new_page_count = ((size - 1) / PAGESIZE) + 1;
+  return __builtin_wasm_memory_grow(0, new_page_count) != SIZE_MAX;
+}
+
+// Clear out the heap, and move it to the given address.
+void reset_heap(void *new_heap_start) {
+  heap_start = new_heap_start;
+  next = new_heap_start;
+  heap_end = get_heap_end();
+}
+
+void *malloc(size_t size) {
+  Region *region_end = region_after(next, size);
+
+  if (region_end > heap_end) {
+    if ((char *)region_end - (char *)heap_start > MAX_HEAP_SIZE) {
+      return NULL;
+    }
+    if (!grow_heap(size)) return NULL;
+    heap_end = get_heap_end();
+  }
+
+  void *result = &next->data;
+  next->size = size;
+  next = region_end;
+
+  return result;
+}
+
+void free(void *ptr) {
+  if (ptr == NULL) return;
+
+  Region *region = region_for_ptr(ptr);
+  Region *region_end = region_after(region, region->size);
+
+  // When freeing the last allocated pointer, re-use that
+  // pointer for the next allocation.
+  if (region_end == next) {
+    next = region;
+  }
+}
+
+void *calloc(size_t count, size_t size) {
+  void *result = malloc(count * size);
+  memset(result, 0, count * size);
+  return result;
+}
+
+void *realloc(void *ptr, size_t new_size) {
+  if (ptr == NULL) {
+    return malloc(new_size);
+  }
+
+  Region *region = region_for_ptr(ptr);
+  Region *region_end = region_after(region, region->size);
+
+  // When reallocating the last allocated region, return
+  // the same pointer, and skip copying the data.
+  if (region_end == next) {
+    next = region;
+    return malloc(new_size);
+  }
+
+  void *result = malloc(new_size);
+  memcpy(result, &region->data, region->size);
+  return result;
+}

vendor/tree-sitter/src/wasm_store.h

@@ -0,0 +1,31 @@
+#ifndef TREE_SITTER_WASM_H_
+#define TREE_SITTER_WASM_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "tree_sitter/api.h"
+#include "./parser.h"
+
+bool ts_wasm_store_start(TSWasmStore *, TSLexer *, const TSLanguage *);
+void ts_wasm_store_reset(TSWasmStore *);
+bool ts_wasm_store_has_error(const TSWasmStore *);
+
+bool ts_wasm_store_call_lex_main(TSWasmStore *, TSStateId);
+bool ts_wasm_store_call_lex_keyword(TSWasmStore *, TSStateId);
+
+uint32_t ts_wasm_store_call_scanner_create(TSWasmStore *);
+void ts_wasm_store_call_scanner_destroy(TSWasmStore *, uint32_t);
+bool ts_wasm_store_call_scanner_scan(TSWasmStore *, uint32_t, uint32_t);
+uint32_t ts_wasm_store_call_scanner_serialize(TSWasmStore *, uint32_t, char *);
+void ts_wasm_store_call_scanner_deserialize(TSWasmStore *, uint32_t, const char *, unsigned);
+
+void ts_wasm_language_retain(const TSLanguage *);
+void ts_wasm_language_release(const TSLanguage *);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif  // TREE_SITTER_WASM_H_

vendor/vendor_tree_sitter.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+
+REMOTE=https://github.com/tree-sitter/tree-sitter.git
+BRANCH=v0.22.5
+
+rm -rf tree-sitter
+rm -rf tmp
+git clone --depth 1 --branch $BRANCH $REMOTE tmp
+mkdir tree-sitter
+mv tmp/lib/src tree-sitter
+mv tmp/lib/include tree-sitter
+mv tmp/LICENSE tree-sitter
+rm -rf tmp