tree_sitter/

lib.rs

// code-gauge-allow-file fileNcss 852 -- core Parser, Language, Tree, Node, and TreeCursor bindings share private wrappers
#![cfg_attr(not(any(test, doctest)), doc = include_str!("./README.md"))]
#![cfg_attr(not(feature = "std"), no_std)]
#![cfg_attr(docsrs, feature(doc_cfg))]

pub mod ffi;
mod lookahead;
mod sexp;

pub use lookahead::LookaheadIterator;
#[doc(hidden)]
pub use sexp::format_sexp;

#[cfg(any(
    not(feature = "std"),
    all(target_arch = "wasm32", target_os = "unknown")
))]
extern crate alloc;
#[cfg(not(feature = "std"))]
use alloc::{
    boxed::Box,
    string::{String, ToString},
    vec::Vec,
};
use core::{
    ffi::{CStr, c_char, c_void},
    fmt, hash, iter,
    marker::PhantomData,
    num::NonZeroU16,
    ops::{ControlFlow, Deref},
    ptr::{self, NonNull},
    slice, str,
};
#[cfg(feature = "std")]
use std::error;
#[cfg(all(unix, feature = "std"))]
use std::os::fd::AsRawFd;
#[cfg(all(windows, feature = "std"))]
use std::os::windows::io::AsRawHandle;

use tree_sitter_language::LanguageFn;

#[cfg(feature = "wasm")]
mod wasm_language;
#[cfg(feature = "wasm")]
#[cfg_attr(docsrs, doc(cfg(feature = "wasm")))]
pub use wasm_language::*;

#[cfg(all(target_arch = "wasm32", target_os = "unknown"))]
mod wasm_allocator;

/// The latest ABI version that the current version of the library supports.
///
/// The CLI of this fork assigns it to each generated language. Its upper 16
/// bits (`0x5742`, "WB") tag the fork and its lower 16 bits count the fork's
/// ABI revisions, so that it never equals an ABI version of upstream
/// Tree-sitter: the library rejects languages generated by upstream's CLI, and
/// upstream's library rejects languages generated by this fork's CLI.
#[doc(alias = "TREE_SITTER_LANGUAGE_VERSION")]
pub const LANGUAGE_VERSION: usize = ffi::TREE_SITTER_LANGUAGE_VERSION as usize;

/// The earliest ABI version that is supported by the current version of the
/// library. Its parse tables are a subset of those of [`LANGUAGE_VERSION`], so
/// the library loads both.
#[doc(alias = "TREE_SITTER_MIN_COMPATIBLE_LANGUAGE_VERSION")]
pub const MIN_COMPATIBLE_LANGUAGE_VERSION: usize =
    ffi::TREE_SITTER_MIN_COMPATIBLE_LANGUAGE_VERSION as usize;

pub const PARSER_HEADER: &str = include_str!("../src/parser.h");

/// An opaque object that defines how to parse a particular language. The code
/// for each `Language` is generated by the Tree-sitter CLI.
#[doc(alias = "TSLanguage")]
#[derive(Debug, PartialEq, Eq, Hash)]
#[repr(transparent)]
pub struct Language(pub(crate) *const ffi::TSLanguage);

pub struct LanguageRef<'a>(
    pub(crate) *const ffi::TSLanguage,
    pub(crate) PhantomData<&'a ()>,
);

/// The metadata associated with a language.
///
/// Currently, this metadata can be used to check the [Semantic Version](https://semver.org/)
/// of the language.
#[doc(alias = "TSLanguageMetadata")]
pub struct LanguageMetadata {
    pub major_version: u8,
    pub minor_version: u8,
    pub patch_version: u8,
}

impl From<ffi::TSLanguageMetadata> for LanguageMetadata {
    fn from(val: ffi::TSLanguageMetadata) -> Self {
        Self {
            major_version: val.major_version,
            minor_version: val.minor_version,
            patch_version: val.patch_version,
        }
    }
}

/// A tree that represents the syntactic structure of a source code file.
#[doc(alias = "TSTree")]
pub struct Tree(NonNull<ffi::TSTree>);

/// A position in a multi-line text document, in terms of rows and columns.
///
/// Rows and columns are zero-based.
#[derive(Clone, Copy, Debug, Default, PartialEq, Eq, Hash, PartialOrd, Ord)]
pub struct Point {
    pub row: usize,
    pub column: usize,
}

/// A range of positions in a multi-line text document, both in terms of bytes
/// and of rows and columns.
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
pub struct Range {
    pub start_byte: usize,
    pub end_byte: usize,
    pub start_point: Point,
    pub end_point: Point,
}

/// A single node within a syntax [`Tree`].
#[doc(alias = "TSNode")]
#[derive(Clone, Copy)]
#[repr(transparent)]
pub struct Node<'tree>(ffi::TSNode, PhantomData<&'tree ()>);

/// A stateful object that this is used to produce a [`Tree`] based on some
/// source code.
#[doc(alias = "TSParser")]
pub struct Parser(NonNull<ffi::TSParser>);

/// A stateful object that is passed into a [`ParseProgressCallback`]
/// to pass in the current state of the parser.
pub struct ParseState(NonNull<ffi::TSParseState>);

impl ParseState {
    #[must_use]
    pub const fn current_byte_offset(&self) -> usize {
        unsafe { self.0.as_ref() }.current_byte_offset as usize
    }

    #[must_use]
    pub const fn has_error(&self) -> bool {
        unsafe { self.0.as_ref() }.has_error
    }
}

#[derive(Default)]
pub struct ParseOptions<'a> {
    pub progress_callback: Option<ParseProgressCallback<'a>>,
}

impl<'a> ParseOptions<'a> {
    #[must_use]
    pub fn new() -> Self {
        Self::default()
    }

    #[must_use]
    pub fn progress_callback<F: FnMut(&ParseState) -> ControlFlow<()>>(
        mut self,
        callback: &'a mut F,
    ) -> Self {
        self.progress_callback = Some(callback);
        self
    }

    /// Create a new `ParseOptions` with a shorter lifetime, borrowing from this one.
    ///
    /// This is useful when you need to reuse parse options multiple times, e.g., calling
    /// [`Parser::parse_with_options`] multiple times with the same options.
    #[must_use]
    pub fn reborrow(&mut self) -> ParseOptions {
        ParseOptions {
            progress_callback: match &mut self.progress_callback {
                Some(cb) => Some(*cb),
                None => None,
            },
        }
    }
}

/// A type of log message.
#[derive(Debug, PartialEq, Eq)]
pub enum LogType {
    Parse,
    Lex,
}

type FieldId = NonZeroU16;

/// A callback that receives log messages during parsing.
type Logger = Box<dyn FnMut(LogType, &str) + Send + 'static>;

/// A callback that receives log messages during parsing, with relaxed constraints.
type UnsafeLogger<'a> = Box<dyn FnMut(LogType, &str) + Send + 'a>;

/// A callback that receives the parse state during parsing.
type ParseProgressCallback<'a> = &'a mut dyn FnMut(&ParseState) -> ControlFlow<()>;

pub trait Decode {
    /// A callback that decodes the next code point from the input slice. It should return the code
    /// point, and how many bytes were decoded.
    fn decode(bytes: &[u8]) -> (i32, u32);
}

/// A stateful object for walking a syntax [`Tree`] efficiently.
#[doc(alias = "TSTreeCursor")]
pub struct TreeCursor<'tree>(ffi::TSTreeCursor, PhantomData<&'tree ()>);

/// An error that occurred when trying to assign a [`Language`] to a [`Parser`].
/// If the `wasm` feature is enabled, this can also indicate a failure to load
/// the Wasm store.
#[derive(Debug, PartialEq, Eq)]
pub enum LanguageError {
    Version(usize),
    NotParseable,
    #[cfg(feature = "wasm")]
    Wasm,
}

/// An error that occurred in [`Parser::set_included_ranges`].
#[derive(Debug, PartialEq, Eq)]
pub struct IncludedRangesError(pub usize);

impl Language {
    #[must_use]
    pub fn new(builder: LanguageFn) -> Self {
        Self(unsafe { builder.into_raw()().cast() })
    }

    /// Check whether this language can be assigned to a parser.
    ///
    /// When Tree-sitter is compiled to WebAssembly, languages obtained from a
    /// syntax tree can be used for parsing only within the same WebAssembly
    /// instance that created the tree. In other instances, such languages can
    /// still be used to inspect syntax trees.
    #[doc(alias = "ts_language_is_parseable")]
    #[must_use]
    pub fn is_parseable(&self) -> bool {
        unsafe { ffi::ts_language_is_parseable(self.0) }
    }

    /// Get the name of this language. This returns `None` for a language with
    /// another ABI version.
    #[doc(alias = "ts_language_name")]
    #[must_use]
    pub fn name(&self) -> Option<&str> {
        let ptr = unsafe { ffi::ts_language_name(self.0) };
        (!ptr.is_null()).then(|| unsafe { CStr::from_ptr(ptr) }.to_str().unwrap())
    }

    /// Get the ABI version number of this [`Language`].
    ///
    /// The library loads only languages whose number is from
    /// [`MIN_COMPATIBLE_LANGUAGE_VERSION`] to [`LANGUAGE_VERSION`], which the CLI
    /// of this fork assigns. Any other number comes from a parser
    /// that another CLI generated.
    #[doc(alias = "ts_language_abi_version")]
    #[must_use]
    pub fn abi_version(&self) -> usize {
        unsafe { ffi::ts_language_abi_version(self.0) as usize }
    }

    /// Get the metadata for this language. This information is generated by the
    /// CLI, and relies on the language author providing the correct metadata in
    /// the language's `tree-sitter.json` file. This returns `None` for a
    /// language with another ABI version.
    ///
    /// See also [`LanguageMetadata`].
    #[doc(alias = "ts_language_metadata")]
    #[must_use]
    pub fn metadata(&self) -> Option<LanguageMetadata> {
        unsafe {
            let ptr = ffi::ts_language_metadata(self.0);
            (!ptr.is_null()).then(|| (*ptr).into())
        }
    }

    /// Get the number of distinct node types in this language.
    #[doc(alias = "ts_language_symbol_count")]
    #[must_use]
    pub fn node_kind_count(&self) -> usize {
        unsafe { ffi::ts_language_symbol_count(self.0) as usize }
    }

    /// Get the number of valid states in this language.
    #[doc(alias = "ts_language_state_count")]
    #[must_use]
    pub fn parse_state_count(&self) -> usize {
        unsafe { ffi::ts_language_state_count(self.0) as usize }
    }

    /// Get a list of all supertype symbols for the language. The list is empty
    /// for a language with another ABI version.
    #[doc(alias = "ts_language_supertypes")]
    #[must_use]
    pub fn supertypes(&self) -> &[u16] {
        let mut length = 0u32;
        unsafe {
            let ptr = ffi::ts_language_supertypes(self.0, core::ptr::addr_of_mut!(length));
            if length == 0 {
                &[]
            } else {
                slice::from_raw_parts(ptr.cast_mut(), length as usize)
            }
        }
    }

    /// Get a list of all subtype symbols for a given supertype symbol. The list
    /// is empty for a language with another ABI version.
    #[doc(alias = "ts_language_supertype_map")]
    #[must_use]
    pub fn subtypes_for_supertype(&self, supertype: u16) -> &[u16] {
        unsafe {
            let mut length = 0u32;
            let ptr = ffi::ts_language_subtypes(self.0, supertype, core::ptr::addr_of_mut!(length));
            if length == 0 {
                &[]
            } else {
                slice::from_raw_parts(ptr.cast_mut(), length as usize)
            }
        }
    }

    /// Get the name of the node kind for the given numerical id.
    #[doc(alias = "ts_language_symbol_name")]
    #[must_use]
    pub fn node_kind_for_id(&self, id: u16) -> Option<&str> {
        let ptr = unsafe { ffi::ts_language_symbol_name(self.0, id) };
        (!ptr.is_null()).then(|| unsafe { CStr::from_ptr(ptr) }.to_str().unwrap())
    }

    /// Get the numeric id for the given node kind.
    #[doc(alias = "ts_language_symbol_for_name")]
    #[must_use]
    pub fn id_for_node_kind(&self, kind: &str, named: bool) -> u16 {
        unsafe {
            ffi::ts_language_symbol_for_name(
                self.0,
                kind.as_bytes().as_ptr().cast::<c_char>(),
                kind.len() as u32,
                named,
            )
        }
    }

    /// Check whether `id` can be used to index this language's symbol tables.
    ///
    /// `ts_symbol_metadata` holds one entry per grammar symbol plus one per alias,
    /// so [`Self::node_kind_count`] is an exact bound. `ERROR` and `_ERROR` sit
    /// at the top of the `u16` range and are handled by the C library before any
    /// table lookup.
    fn node_kind_id_is_valid(&self, id: u16) -> bool {
        (id as usize) < self.node_kind_count() || id >= u16::MAX - 1
    }

    /// Check if the node type for the given numerical id is named (as opposed
    /// to an anonymous node type).
    #[must_use]
    pub fn node_kind_is_named(&self, id: u16) -> bool {
        self.node_kind_id_is_valid(id)
            && unsafe { ffi::ts_language_symbol_type(self.0, id) == ffi::TSSymbolTypeRegular }
    }

    /// Check if the node type for the given numerical id is visible (as opposed
    /// to a hidden node type).
    #[must_use]
    pub fn node_kind_is_visible(&self, id: u16) -> bool {
        self.node_kind_id_is_valid(id)
            && unsafe { ffi::ts_language_symbol_type(self.0, id) <= ffi::TSSymbolTypeAnonymous }
    }

    /// Check if the node type for the given numerical id is a supertype.
    #[must_use]
    pub fn node_kind_is_supertype(&self, id: u16) -> bool {
        self.node_kind_id_is_valid(id)
            && unsafe { ffi::ts_language_symbol_type(self.0, id) == ffi::TSSymbolTypeSupertype }
    }

    /// Get the number of distinct field names in this language.
    #[doc(alias = "ts_language_field_count")]
    #[must_use]
    pub fn field_count(&self) -> usize {
        unsafe { ffi::ts_language_field_count(self.0) as usize }
    }

    /// Get the field name for the given numerical id.
    #[doc(alias = "ts_language_field_name_for_id")]
    #[must_use]
    pub fn field_name_for_id(&self, field_id: u16) -> Option<&str> {
        let ptr = unsafe { ffi::ts_language_field_name_for_id(self.0, field_id) };
        (!ptr.is_null()).then(|| unsafe { CStr::from_ptr(ptr) }.to_str().unwrap())
    }

    /// Get the numerical id for the given field name.
    #[doc(alias = "ts_language_field_id_for_name")]
    #[must_use]
    pub fn field_id_for_name(&self, field_name: impl AsRef<[u8]>) -> Option<FieldId> {
        let field_name = field_name.as_ref();
        let id = unsafe {
            ffi::ts_language_field_id_for_name(
                self.0,
                field_name.as_ptr().cast::<c_char>(),
                field_name.len() as u32,
            )
        };
        FieldId::new(id)
    }

    /// Get the next parse state. Combine this with
    /// [`lookahead_iterator`](Language::lookahead_iterator) to
    /// generate completion suggestions or valid symbols in error nodes.
    ///
    /// Example:
    /// ```ignore
    /// let state = language.next_state(node.parse_state(), node.grammar_id());
    /// ```
    #[doc(alias = "ts_language_next_state")]
    #[must_use]
    pub fn next_state(&self, state: u16, id: u16) -> u16 {
        unsafe { ffi::ts_language_next_state(self.0, state, id) }
    }

    /// Create a new lookahead iterator for this language and parse state.
    ///
    /// This returns `None` if state is invalid for this language.
    ///
    /// Iterating [`LookaheadIterator`] will yield valid symbols in the given
    /// parse state. A newly created iterator is not positioned on a symbol, so
    /// [`LookaheadIterator::current_symbol`] returns `None` until the first
    /// [`Iterator::next`] call.
    ///
    /// The iterator retains the language, so the language may be dropped while
    /// the iterator is still in use.
    ///
    /// Lookahead iterators can be useful to generate suggestions and improve
    /// syntax error diagnostics. To get symbols valid in an `ERROR` node, use the
    /// lookahead iterator on its first leaf node state. For a missing node, use
    /// the node's [`parse_state`](Node::parse_state). Keep in mind that lookahead
    /// symbols are valid in that parse state, but are not necessarily valid
    /// continuations in the context of the actual following token or guaranteed
    /// to be considered during error recovery.
    #[doc(alias = "ts_lookahead_iterator_new")]
    #[must_use]
    pub fn lookahead_iterator(&self, state: u16) -> Option<LookaheadIterator> {
        let ptr = unsafe { ffi::ts_lookahead_iterator_new(self.0, state) };
        (!ptr.is_null()).then(|| unsafe { LookaheadIterator::from_raw(ptr) })
    }
}

impl From<LanguageFn> for Language {
    fn from(value: LanguageFn) -> Self {
        Self::new(value)
    }
}

impl Clone for Language {
    fn clone(&self) -> Self {
        unsafe { Self(ffi::ts_language_copy(self.0)) }
    }
}

impl Drop for Language {
    fn drop(&mut self) {
        unsafe { ffi::ts_language_delete(self.0) }
    }
}

impl Deref for LanguageRef<'_> {
    type Target = Language;

    fn deref(&self) -> &Self::Target {
        unsafe { &*(core::ptr::addr_of!(self.0).cast::<Language>()) }
    }
}

impl Default for Parser {
    fn default() -> Self {
        Self::new()
    }
}

impl Parser {
    /// Create a new parser.
    #[doc(alias = "ts_parser_new")]
    #[must_use]
    pub fn new() -> Self {
        unsafe {
            let parser = ffi::ts_parser_new();
            Self(NonNull::new_unchecked(parser))
        }
    }

    /// Set the language that the parser should use for parsing.
    ///
    /// Returns a Result indicating whether or not the language was successfully
    /// assigned. Assignment fails if the language cannot be used for parsing,
    /// or if it was generated with an incompatible version of the Tree-sitter
    /// CLI. Check this using [`Language::is_parseable`] and
    /// [`Language::abi_version`].
    #[doc(alias = "ts_parser_set_language")]
    pub fn set_language(&mut self, language: &Language) -> Result<(), LanguageError> {
        let version = language.abi_version();
        if (MIN_COMPATIBLE_LANGUAGE_VERSION..=LANGUAGE_VERSION).contains(&version) {
            if !language.is_parseable() {
                return Err(LanguageError::NotParseable);
            }
            #[cfg_attr(
                not(feature = "wasm"),
                expect(unused_variables, reason = "only used when wasm feature is enabled")
            )]
            let success = unsafe { ffi::ts_parser_set_language(self.0.as_ptr(), language.0) };
            #[cfg(feature = "wasm")]
            if !success {
                return Err(LanguageError::Wasm);
            }
            Ok(())
        } else {
            Err(LanguageError::Version(version))
        }
    }

    /// Get the parser's current language.
    #[doc(alias = "ts_parser_language")]
    #[must_use]
    pub fn language(&self) -> Option<LanguageRef<'_>> {
        let ptr = unsafe { ffi::ts_parser_language(self.0.as_ptr()) };
        (!ptr.is_null()).then_some(LanguageRef(ptr, PhantomData))
    }

    /// Get the parser's current logger.
    #[doc(alias = "ts_parser_logger")]
    #[must_use]
    pub fn logger(&self) -> Option<&Logger> {
        let logger = unsafe { ffi::ts_parser_logger(self.0.as_ptr()) };
        unsafe { logger.payload.cast::<Logger>().as_ref() }
    }

    /// Set the logging callback that the parser should use during parsing.
    ///
    /// To log through a callback that borrows non-`'static` data, see
    /// [`set_logger_unchecked`](Parser::set_logger_unchecked).
    #[doc(alias = "ts_parser_set_logger")]
    pub fn set_logger(&mut self, logger: Option<Logger>) {
        // SAFETY: `Logger` is `Send + 'static`
        unsafe { self.set_logger_unchecked(logger) };
    }

    /// Set the logging callback, allowing the callback to borrow non-`'static`
    /// data. See [`set_logger`](Parser::set_logger).
    ///
    /// # Safety
    ///
    /// Any data borrowed by `logger` must remain valid until the logger is
    /// removed. Equivalently, the parser must not be used to parse once the
    /// borrowed data has gone out of scope.
    pub unsafe fn set_logger_unchecked(&mut self, logger: Option<UnsafeLogger<'_>>) {
        let prev_logger = unsafe { ffi::ts_parser_logger(self.0.as_ptr()) };
        if !prev_logger.payload.is_null() {
            drop(unsafe { Box::from_raw(prev_logger.payload.cast::<Logger>()) });
        }

        let c_logger = if let Some(logger) = logger {
            let container = Box::new(logger);

            unsafe extern "C" fn log(
                payload: *mut c_void,
                c_log_type: ffi::TSLogType,
                c_message: *const c_char,
            ) {
                unsafe {
                    let callback = payload.cast::<Logger>().as_mut().unwrap();
                    if let Ok(message) = CStr::from_ptr(c_message).to_str() {
                        let log_type = if c_log_type == ffi::TSLogTypeParse {
                            LogType::Parse
                        } else {
                            LogType::Lex
                        };
                        callback(log_type, message);
                    }
                }
            }

            let raw_container = Box::into_raw(container);

            ffi::TSLogger {
                payload: raw_container.cast::<c_void>(),
                log: Some(log),
            }
        } else {
            ffi::TSLogger {
                payload: ptr::null_mut(),
                log: None,
            }
        };

        unsafe { ffi::ts_parser_set_logger(self.0.as_ptr(), c_logger) };
    }

    /// Set the destination to which the parser should write debugging graphs
    /// during parsing. The graphs are formatted in the DOT language. You may
    /// want to pipe these graphs directly to a `dot(1)` process in order to
    /// generate SVG output.
    #[doc(alias = "ts_parser_print_dot_graphs")]
    #[cfg(not(target_os = "wasi"))]
    #[cfg(feature = "std")]
    #[cfg_attr(docsrs, doc(cfg(feature = "std")))]
    pub fn print_dot_graphs(
        &mut self,
        #[cfg(unix)] file: &impl AsRawFd,
        #[cfg(windows)] file: &impl AsRawHandle,
    ) {
        #[cfg(unix)]
        {
            let fd = file.as_raw_fd();
            unsafe {
                ffi::ts_parser_print_dot_graphs(self.0.as_ptr(), ffi::_ts_dup(fd));
            }
        }

        #[cfg(windows)]
        {
            let handle = file.as_raw_handle();
            unsafe {
                ffi::ts_parser_print_dot_graphs(self.0.as_ptr(), ffi::_ts_dup(handle));
            }
        }
    }

    /// Stop the parser from printing debugging graphs while parsing.
    #[doc(alias = "ts_parser_print_dot_graphs")]
    #[cfg(not(target_os = "wasi"))]
    #[cfg(feature = "std")]
    #[cfg_attr(docsrs, doc(cfg(feature = "std")))]
    pub fn stop_printing_dot_graphs(&mut self) {
        unsafe { ffi::ts_parser_print_dot_graphs(self.0.as_ptr(), -1) }
    }

    /// Parse a slice of UTF8 text.
    ///
    /// # Arguments:
    /// * `text` The UTF8-encoded text to parse.
    ///
    /// Returns a [`Tree`] if parsing succeeded, or `None` if:
    ///  * The parser has not yet had a language assigned with [`Parser::set_language`]
    #[doc(alias = "ts_parser_parse")]
    pub fn parse(&mut self, text: impl AsRef<[u8]>) -> Option<Tree> {
        let bytes = text.as_ref();
        if let Ok(len) = u32::try_from(bytes.len()) {
            return unsafe {
                self.parse_raw_bytes(
                    (bytes.as_ptr().cast::<c_char>(), len),
                    ffi::TSInputEncodingUTF8,
                )
            };
        }
        let len = bytes.len();
        self.parse_with_options(
            &mut |i, _| {
                if i < len {
                    &bytes[i..]
                } else {
                    Default::default()
                }
            },
            None,
        )
    }

    /// Parse text provided in chunks by a callback.
    ///
    /// # Arguments:
    /// * `callback` A function that takes a byte offset and position and returns a slice of
    ///   UTF8-encoded text starting at that byte offset and position. The slices can be of any
    ///   length. If the given position is at the end of the text, the callback should return an
    ///   empty slice.
    /// * `options` Options for parsing the text. This can be used to set a progress callback.
    pub fn parse_with_options<T: AsRef<[u8]>, F: FnMut(usize, Point) -> T>(
        &mut self,
        callback: &mut F,
        options: Option<ParseOptions>,
    ) -> Option<Tree> {
        self.parse_chunks::<0, u8, T, F>(callback, (options, ffi::TSInputEncodingUTF8, None))
    }

    /// Parse a slice of UTF16 little-endian text.
    ///
    /// # Arguments:
    /// * `input` The UTF16-encoded text to parse.
    pub fn parse_utf16_le(&mut self, input: impl AsRef<[u16]>) -> Option<Tree> {
        self.parse_utf16_slice::<true>(input.as_ref())
    }

    /// Parse UTF16 little-endian text provided in chunks by a callback.
    ///
    /// # Arguments:
    /// * `callback` A function that takes a code point offset and position and returns a slice of
    ///   UTF16-encoded text starting at that byte offset and position. The slices can be of any
    ///   length. If the given position is at the end of the text, the callback should return an
    ///   empty slice.
    /// * `options` Options for parsing the text. This can be used to set a progress callback.
    pub fn parse_utf16_le_with_options<T: AsRef<[u16]>, F: FnMut(usize, Point) -> T>(
        &mut self,
        callback: &mut F,
        options: Option<ParseOptions>,
    ) -> Option<Tree> {
        self.parse_utf16_with_options::<true, T, F>(callback, options)
    }

    /// Parse a slice of UTF16 big-endian text.
    ///
    /// # Arguments:
    /// * `input` The UTF16-encoded text to parse.
    pub fn parse_utf16_be(&mut self, input: impl AsRef<[u16]>) -> Option<Tree> {
        self.parse_utf16_slice::<false>(input.as_ref())
    }

    /// Parse UTF16 big-endian text provided in chunks by a callback.
    ///
    /// # Arguments:
    /// * `callback` A function that takes a code point offset and position and returns a slice of
    ///   UTF16-encoded text starting at that byte offset and position. The slices can be of any
    ///   length. If the given position is at the end of the text, the callback should return an
    ///   empty slice.
    /// * `options` Options for parsing the text. This can be used to set a progress callback.
    pub fn parse_utf16_be_with_options<T: AsRef<[u16]>, F: FnMut(usize, Point) -> T>(
        &mut self,
        callback: &mut F,
        options: Option<ParseOptions>,
    ) -> Option<Tree> {
        self.parse_utf16_with_options::<false, T, F>(callback, options)
    }

    /// Parse text provided in chunks by a callback using a custom encoding.
    /// This is useful for parsing text in encodings that are not UTF-8 or UTF-16.
    ///
    /// # Arguments:
    /// * `callback` A function that takes a byte offset and position and returns a slice of text
    ///   starting at that byte offset and position. The slices can be of any length. If the given
    ///   position is at the end of the text, the callback should return an empty slice.
    /// * `options` Options for parsing the text. This can be used to set a progress callback.
    ///
    /// Additionally, you must set the generic parameter [`D`] to a type that implements the
    /// [`Decode`] trait. This trait has a single method, [`decode`](Decode::decode), which takes a
    /// slice of bytes and returns a tuple of the code point and the number of bytes consumed.
    /// The `decode` method should return `-1` for the code point if decoding fails.
    pub fn parse_custom_encoding<D: Decode, T: AsRef<[u8]>, F: FnMut(usize, Point) -> T>(
        &mut self,
        callback: &mut F,
        options: Option<ParseOptions>,
    ) -> Option<Tree> {
        unsafe extern "C" fn decode_fn<D: Decode>(
            data: *const u8,
            len: u32,
            code_point: *mut i32,
        ) -> u32 {
            unsafe {
                let (c, len) = D::decode(core::slice::from_raw_parts(data, len as usize));
                if let Some(code_point) = code_point.as_mut() {
                    *code_point = c;
                }
                len
            }
        }

        self.parse_chunks::<0, u8, T, F>(
            callback,
            (options, ffi::TSInputEncodingCustom, Some(decode_fn::<D>)),
        )
    }

    fn parse_utf16_slice<const LE: bool>(&mut self, code_points: &[u16]) -> Option<Tree> {
        let encoding = if LE {
            ffi::TSInputEncodingUTF16LE
        } else {
            ffi::TSInputEncodingUTF16BE
        };
        if let Some(byte_len) = code_points
            .len()
            .checked_mul(2)
            .and_then(|n| u32::try_from(n).ok())
        {
            return unsafe {
                self.parse_raw_bytes((code_points.as_ptr().cast::<c_char>(), byte_len), encoding)
            };
        }
        let len = code_points.len();
        self.parse_utf16_with_options::<LE, _, _>(
            &mut |i, _| if i < len { &code_points[i..] } else { &[] },
            None,
        )
    }

    fn parse_utf16_with_options<const LE: bool, T: AsRef<[u16]>, F: FnMut(usize, Point) -> T>(
        &mut self,
        callback: &mut F,
        options: Option<ParseOptions>,
    ) -> Option<Tree> {
        let encoding = if LE {
            ffi::TSInputEncodingUTF16LE
        } else {
            ffi::TSInputEncodingUTF16BE
        };
        self.parse_chunks::<1, u16, T, F>(callback, (options, encoding, None))
    }

    fn parse_chunks<const UNIT_SHIFT: u32, U, T: AsRef<[U]>, F: FnMut(usize, Point) -> T>(
        &mut self,
        callback: &mut F,
        (options, encoding, decode): (
            Option<ParseOptions>,
            ffi::TSInputEncoding,
            ffi::TSDecodeFunction,
        ),
    ) -> Option<Tree> {
        type Payload<'a, F, T> = (&'a mut F, Option<T>);

        unsafe extern "C" fn read<
            const UNIT_SHIFT: u32,
            U,
            T: AsRef<[U]>,
            F: FnMut(usize, Point) -> T,
        >(
            payload: *mut c_void,
            byte_offset: u32,
            position: ffi::TSPoint,
            bytes_read: *mut u32,
        ) -> *const c_char {
            unsafe {
                let (callback, text) = payload.cast::<Payload<F, T>>().as_mut().unwrap();
                *text = Some(callback(
                    (byte_offset as usize) >> UNIT_SHIFT,
                    Point {
                        row: position.row as usize,
                        column: (position.column as usize) >> UNIT_SHIFT,
                    },
                ));
                let slice = text.as_ref().unwrap().as_ref();
                *bytes_read =
                    (slice.len().min((u32::MAX as usize) >> UNIT_SHIFT) << UNIT_SHIFT) as u32;
                slice.as_ptr().cast::<c_char>()
            }
        }

        // Hold the returned chunk in `payload` across the next C read so callbacks may return owned buffers.
        let mut payload: Payload<F, T> = (callback, None);
        let c_input = ffi::TSInput {
            payload: ptr::addr_of_mut!(payload).cast::<c_void>(),
            read: Some(read::<UNIT_SHIFT, U, T, F>),
            encoding,
            decode,
        };
        unsafe { self.parse_c_input(c_input, options) }
    }

    unsafe fn parse_raw_bytes(
        &mut self,
        (bytes, byte_len): (*const c_char, u32),
        encoding: ffi::TSInputEncoding,
    ) -> Option<Tree> {
        let c_new_tree = unsafe {
            ffi::ts_parser_parse_string_encoding(self.0.as_ptr(), bytes, byte_len, encoding)
        };
        NonNull::new(c_new_tree).map(Tree)
    }

    unsafe fn parse_c_input(
        &mut self,
        c_input: ffi::TSInput,
        options: Option<ParseOptions>,
    ) -> Option<Tree> {
        unsafe extern "C" fn progress(state: *mut ffi::TSParseState) -> bool {
            unsafe {
                let callback = (*state)
                    .payload
                    .cast::<ParseProgressCallback>()
                    .as_mut()
                    .unwrap();
                match callback(&ParseState::from_raw(state)) {
                    ControlFlow::Continue(()) => false,
                    ControlFlow::Break(()) => true,
                }
            }
        }

        let mut callback_ptr;
        let parse_options = if let Some(cb) = options.and_then(|o| o.progress_callback) {
            callback_ptr = cb;
            ffi::TSParseOptions {
                payload: core::ptr::addr_of_mut!(callback_ptr).cast::<c_void>(),
                progress_callback: Some(progress),
            }
        } else {
            ffi::TSParseOptions {
                payload: ptr::null_mut(),
                progress_callback: None,
            }
        };

        let c_new_tree =
            unsafe { ffi::ts_parser_parse_with_options(self.0.as_ptr(), c_input, parse_options) };
        NonNull::new(c_new_tree).map(Tree)
    }

    /// Instruct the parser to start the next parse from the beginning.
    ///
    /// If the parser previously failed because of a callback, then by default,
    /// it will resume where it left off on the next call to [`parse`](Parser::parse)
    /// or other parsing functions. If you don't want to resume, and instead intend to use
    /// this parser to parse some other document, you must call `reset` first.
    #[doc(alias = "ts_parser_reset")]
    pub fn reset(&mut self) {
        unsafe { ffi::ts_parser_reset(self.0.as_ptr()) }
    }

    /// 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.
    ///
    /// If `ranges` is empty, 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` < `length - 1`:
    /// ```text
    ///     ranges[i].end_byte <= ranges[i + 1].start_byte
    /// ```
    /// If this requirement is not satisfied, method will return
    /// [`IncludedRangesError`] error with an offset in the passed ranges
    /// slice pointing to a first incorrect range.
    #[doc(alias = "ts_parser_set_included_ranges")]
    pub fn set_included_ranges(&mut self, ranges: &[Range]) -> Result<(), IncludedRangesError> {
        let ts_ranges = ranges.iter().copied().map(Into::into).collect::<Vec<_>>();
        let result = unsafe {
            ffi::ts_parser_set_included_ranges(
                self.0.as_ptr(),
                ts_ranges.as_ptr(),
                ts_ranges.len() as u32,
            )
        };

        if result {
            Ok(())
        } else {
            let mut prev_end_byte = 0;
            for (i, range) in ranges.iter().enumerate() {
                if range.start_byte < prev_end_byte || range.end_byte < range.start_byte {
                    return Err(IncludedRangesError(i));
                }
                prev_end_byte = range.end_byte;
            }
            Err(IncludedRangesError(0))
        }
    }

    /// Get the ranges of text that the parser will include when parsing.
    #[doc(alias = "ts_parser_included_ranges")]
    #[must_use]
    pub fn included_ranges(&self) -> Vec<Range> {
        let mut count = 0u32;
        unsafe {
            let ptr =
                ffi::ts_parser_included_ranges(self.0.as_ptr(), core::ptr::addr_of_mut!(count));
            let ranges = slice::from_raw_parts(ptr, count as usize);
            ranges.iter().copied().map(Into::into).collect()
        }
    }
}

impl Drop for Parser {
    fn drop(&mut self) {
        #[cfg(feature = "std")]
        #[cfg(not(target_os = "wasi"))]
        {
            self.stop_printing_dot_graphs();
        }
        self.set_logger(None);
        unsafe { ffi::ts_parser_delete(self.0.as_ptr()) }
    }
}

#[cfg(windows)]
unsafe extern "C" {
    fn _open_osfhandle(osfhandle: isize, flags: core::ffi::c_int) -> core::ffi::c_int;
}

impl Tree {
    /// Get the root node of the syntax tree.
    #[doc(alias = "ts_tree_root_node")]
    #[must_use]
    pub fn root_node(&self) -> Node {
        Node::new(unsafe { ffi::ts_tree_root_node(self.0.as_ptr()) }).unwrap()
    }

    /// Get the root node of the syntax tree, but with its position shifted
    /// forward by the given offset.
    #[doc(alias = "ts_tree_root_node_with_offset")]
    #[must_use]
    pub fn root_node_with_offset(&self, offset_bytes: usize, offset_extent: Point) -> Node {
        Node::new(unsafe {
            ffi::ts_tree_root_node_with_offset(
                self.0.as_ptr(),
                offset_bytes as u32,
                offset_extent.into(),
            )
        })
        .unwrap()
    }

    /// Get the language that was used to parse the syntax tree.
    ///
    /// When Tree-sitter is compiled to WebAssembly, this returns the original
    /// language if the tree is being accessed from the same WebAssembly
    /// instance that created it. Otherwise, this returns a copy of the language
    /// that can be used to inspect the tree but cannot be assigned to a parser.
    #[doc(alias = "ts_tree_language")]
    #[must_use]
    pub fn language(&self) -> LanguageRef {
        LanguageRef(
            unsafe { ffi::ts_tree_language(self.0.as_ptr()) },
            PhantomData,
        )
    }

    /// Create a new [`TreeCursor`] starting from the root of the tree.
    #[must_use]
    pub fn walk(&self) -> TreeCursor {
        self.root_node().walk()
    }

    /// Get the included ranges that were used to parse the syntax tree.
    #[doc(alias = "ts_tree_included_ranges")]
    #[must_use]
    pub fn included_ranges(&self) -> Vec<Range> {
        let mut count = 0u32;
        unsafe {
            let ptr = ffi::ts_tree_included_ranges(self.0.as_ptr(), core::ptr::addr_of_mut!(count));
            let ranges = slice::from_raw_parts(ptr, count as usize);
            let result = ranges.iter().copied().map(Into::into).collect();
            ts_free(ptr.cast::<c_void>());
            result
        }
    }

    /// Print a graph of the tree to the given file descriptor.
    /// The graph is formatted in the DOT language. You may want to pipe this
    /// graph directly to a `dot(1)` process in order to generate SVG
    /// output.
    #[doc(alias = "ts_tree_print_dot_graph")]
    #[cfg(not(target_os = "wasi"))]
    #[cfg(feature = "std")]
    #[cfg_attr(docsrs, doc(cfg(feature = "std")))]
    pub fn print_dot_graph(
        &self,
        #[cfg(unix)] file: &impl AsRawFd,
        #[cfg(windows)] file: &impl AsRawHandle,
    ) {
        #[cfg(unix)]
        {
            let fd = file.as_raw_fd();
            unsafe { ffi::ts_tree_print_dot_graph(self.0.as_ptr(), fd) }
        }

        #[cfg(windows)]
        {
            let handle = file.as_raw_handle();
            let fd = unsafe { _open_osfhandle(handle as isize, 0) };
            unsafe { ffi::ts_tree_print_dot_graph(self.0.as_ptr(), fd) }
        }
    }
}

impl fmt::Debug for Tree {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        write!(f, "{{Tree {:?}}}", self.root_node())
    }
}

impl Drop for Tree {
    fn drop(&mut self) {
        unsafe { ffi::ts_tree_delete(self.0.as_ptr()) }
    }
}

impl Clone for Tree {
    fn clone(&self) -> Self {
        unsafe { Self(NonNull::new_unchecked(ffi::ts_tree_copy(self.0.as_ptr()))) }
    }
}

impl<'tree> Node<'tree> {
    fn new(node: ffi::TSNode) -> Option<Self> {
        (!node.id.is_null()).then_some(Node(node, PhantomData))
    }

    /// Get a numeric id for this node that is unique within its syntax tree.
    #[must_use]
    pub fn id(&self) -> usize {
        self.0.id as usize
    }

    /// Get this node's type as a numerical id.
    #[doc(alias = "ts_node_symbol")]
    #[must_use]
    pub fn kind_id(&self) -> u16 {
        unsafe { ffi::ts_node_symbol(self.0) }
    }

    /// Get the node's type as a numerical id as it appears in the grammar
    /// ignoring aliases.
    #[doc(alias = "ts_node_grammar_symbol")]
    #[must_use]
    pub fn grammar_id(&self) -> u16 {
        unsafe { ffi::ts_node_grammar_symbol(self.0) }
    }

    /// Get this node's type as a string.
    #[doc(alias = "ts_node_type")]
    #[must_use]
    pub fn kind(&self) -> &'tree str {
        let ptr = unsafe { ffi::ts_node_type(self.0) };
        assert!(!ptr.is_null());
        unsafe { CStr::from_ptr(ptr) }.to_str().unwrap()
    }

    /// Get this node's symbol name as it appears in the grammar ignoring
    /// aliases as a string.
    #[doc(alias = "ts_node_grammar_type")]
    #[must_use]
    pub fn grammar_name(&self) -> &'tree str {
        let ptr = unsafe { ffi::ts_node_grammar_type(self.0) };
        assert!(!ptr.is_null());
        unsafe { CStr::from_ptr(ptr) }.to_str().unwrap()
    }

    /// Get the [`Language`] that was used to parse this node's syntax tree.
    ///
    /// When Tree-sitter is compiled to WebAssembly, this returns the original
    /// language if the node is being accessed from the same WebAssembly
    /// instance that created its tree. Otherwise, this returns a copy of the
    /// language that can be used to inspect the tree but cannot be assigned to
    /// a parser.
    #[doc(alias = "ts_node_language")]
    #[must_use]
    pub fn language(&self) -> LanguageRef<'tree> {
        LanguageRef(unsafe { ffi::ts_node_language(self.0) }, PhantomData)
    }

    /// 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.
    #[doc(alias = "ts_node_is_named")]
    #[must_use]
    pub fn is_named(&self) -> bool {
        unsafe { ffi::ts_node_is_named(self.0) }
    }

    /// Check if this node is *extra*.
    ///
    /// Extra nodes represent things like comments, which are not required by the
    /// grammar, but can appear anywhere.
    #[doc(alias = "ts_node_is_extra")]
    #[must_use]
    pub fn is_extra(&self) -> bool {
        unsafe { ffi::ts_node_is_extra(self.0) }
    }

    /// Check if this node represents a syntax error or contains any syntax
    /// errors anywhere within it.
    #[doc(alias = "ts_node_has_error")]
    #[must_use]
    pub fn has_error(&self) -> bool {
        unsafe { ffi::ts_node_has_error(self.0) }
    }

    /// Check if this node represents a syntax error.
    ///
    /// Syntax errors represent parts of the code that could not be incorporated
    /// into a valid syntax tree.
    #[doc(alias = "ts_node_is_error")]
    #[must_use]
    pub fn is_error(&self) -> bool {
        unsafe { ffi::ts_node_is_error(self.0) }
    }

    /// Get the parse state immediately before this node.
    ///
    /// For a missing node, this is the state from the recovery path that was
    /// selected by the parser. It can be used with
    /// [`Language::lookahead_iterator`] to inspect the symbols that are valid in
    /// that state. This does not necessarily include every symbol that could be
    /// recovered by inserting a missing node, because the recovery process can
    /// consider multiple stack versions.
    #[doc(alias = "ts_node_parse_state")]
    #[must_use]
    pub fn parse_state(&self) -> u16 {
        unsafe { ffi::ts_node_parse_state(self.0) }
    }

    /// Get the parse state after this node.
    #[doc(alias = "ts_node_next_parse_state")]
    #[must_use]
    pub fn next_parse_state(&self) -> u16 {
        unsafe { ffi::ts_node_next_parse_state(self.0) }
    }

    /// Check if this node is *missing*.
    ///
    /// Missing nodes are inserted by the parser in order to recover from
    /// certain kinds of syntax errors.
    #[doc(alias = "ts_node_is_missing")]
    #[must_use]
    pub fn is_missing(&self) -> bool {
        unsafe { ffi::ts_node_is_missing(self.0) }
    }

    /// Get the byte offset where this node starts.
    #[doc(alias = "ts_node_start_byte")]
    #[must_use]
    pub fn start_byte(&self) -> usize {
        unsafe { ffi::ts_node_start_byte(self.0) as usize }
    }

    /// Get the byte offset where this node ends.
    #[doc(alias = "ts_node_end_byte")]
    #[must_use]
    pub fn end_byte(&self) -> usize {
        unsafe { ffi::ts_node_end_byte(self.0) as usize }
    }

    /// Get the byte range of source code that this node represents.
    #[must_use]
    pub fn byte_range(&self) -> core::ops::Range<usize> {
        self.start_byte()..self.end_byte()
    }

    /// Get the range of source code that this node represents, both in terms of
    /// raw bytes and of row/column coordinates.
    #[must_use]
    pub fn range(&self) -> Range {
        Range {
            start_byte: self.start_byte(),
            end_byte: self.end_byte(),
            start_point: self.start_position(),
            end_point: self.end_position(),
        }
    }

    /// Get this node's start position in terms of rows and columns.
    #[doc(alias = "ts_node_start_point")]
    #[must_use]
    pub fn start_position(&self) -> Point {
        let result = unsafe { ffi::ts_node_start_point(self.0) };
        result.into()
    }

    /// Get this node's end position in terms of rows and columns.
    #[doc(alias = "ts_node_end_point")]
    #[must_use]
    pub fn end_position(&self) -> Point {
        let result = unsafe { ffi::ts_node_end_point(self.0) };
        result.into()
    }

    /// 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
    /// [`Node::children`] instead.
    #[doc(alias = "ts_node_child")]
    #[must_use]
    pub fn child(&self, i: u32) -> Option<Self> {
        Self::new(unsafe { ffi::ts_node_child(self.0, i) })
    }

    /// Get this node's number of children.
    #[doc(alias = "ts_node_child_count")]
    #[must_use]
    pub fn child_count(&self) -> u32 {
        unsafe { ffi::ts_node_child_count(self.0) }
    }

    /// Get this node's *named* child at the given index.
    ///
    /// See also [`Node::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
    /// [`Node::named_children`] instead.
    #[doc(alias = "ts_node_named_child")]
    #[must_use]
    pub fn named_child(&self, i: u32) -> Option<Self> {
        Self::new(unsafe { ffi::ts_node_named_child(self.0, i) })
    }

    /// Get this node's number of *named* children.
    ///
    /// See also [`Node::is_named`].
    #[doc(alias = "ts_node_named_child_count")]
    #[must_use]
    pub fn named_child_count(&self) -> usize {
        unsafe { ffi::ts_node_named_child_count(self.0) as usize }
    }

    /// Get the first child with the given field name.
    ///
    /// If multiple children may have the same field name, access them using
    /// [`children_by_field_name`](Node::children_by_field_name)
    #[doc(alias = "ts_node_child_by_field_name")]
    #[must_use]
    pub fn child_by_field_name(&self, field_name: impl AsRef<[u8]>) -> Option<Self> {
        let field_name = field_name.as_ref();
        Self::new(unsafe {
            ffi::ts_node_child_by_field_name(
                self.0,
                field_name.as_ptr().cast::<c_char>(),
                field_name.len() as u32,
            )
        })
    }

    /// Get this node's child with the given numerical field id.
    ///
    /// See also [`child_by_field_name`](Node::child_by_field_name). You can
    /// convert a field name to an id using [`Language::field_id_for_name`].
    #[doc(alias = "ts_node_child_by_field_id")]
    #[must_use]
    pub fn child_by_field_id(&self, field_id: u16) -> Option<Self> {
        Self::new(unsafe { ffi::ts_node_child_by_field_id(self.0, field_id) })
    }

    /// Get the field name of this node's child at the given index.
    #[doc(alias = "ts_node_field_name_for_child")]
    #[must_use]
    pub fn field_name_for_child(&self, child_index: u32) -> Option<&'tree str> {
        unsafe {
            let ptr = ffi::ts_node_field_name_for_child(self.0, child_index);
            (!ptr.is_null()).then(|| CStr::from_ptr(ptr).to_str().unwrap())
        }
    }

    /// Get the field name of this node's named child at the given index.
    #[must_use]
    pub fn field_name_for_named_child(&self, named_child_index: u32) -> Option<&'tree str> {
        unsafe {
            let ptr = ffi::ts_node_field_name_for_named_child(self.0, named_child_index);
            (!ptr.is_null()).then(|| CStr::from_ptr(ptr).to_str().unwrap())
        }
    }

    /// Iterate over this node's children.
    ///
    /// A [`TreeCursor`] is used to retrieve the children efficiently. Obtain
    /// a [`TreeCursor`] by calling [`Tree::walk`] or [`Node::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 = Node<'tree>> + 'cursor {
        cursor.reset(*self);
        cursor.goto_first_child();
        (0..self.child_count()).map(move |_| {
            let result = cursor.node();
            cursor.goto_next_sibling();
            result
        })
    }

    /// Iterate over this node's named children.
    ///
    /// See also [`Node::children`].
    pub fn named_children<'cursor>(
        &self,
        cursor: &'cursor mut TreeCursor<'tree>,
    ) -> impl ExactSizeIterator<Item = Node<'tree>> + 'cursor {
        cursor.reset(*self);
        cursor.goto_first_child();
        (0..self.named_child_count()).map(move |_| {
            while !cursor.node().is_named() {
                if !cursor.goto_next_sibling() {
                    break;
                }
            }
            let result = cursor.node();
            cursor.goto_next_sibling();
            result
        })
    }

    /// Iterate over this node's children with a given field name.
    ///
    /// See also [`Node::children`].
    pub fn children_by_field_name<'cursor>(
        &self,
        field_name: &str,
        cursor: &'cursor mut TreeCursor<'tree>,
    ) -> impl Iterator<Item = Node<'tree>> + 'cursor {
        let field_id = self.language().field_id_for_name(field_name);
        let mut done = field_id.is_none();
        if !done {
            cursor.reset(*self);
            cursor.goto_first_child();
        }
        iter::from_fn(move || {
            if !done {
                while cursor.field_id() != field_id {
                    if !cursor.goto_next_sibling() {
                        return None;
                    }
                }
                let result = cursor.node();
                if !cursor.goto_next_sibling() {
                    done = true;
                }
                return Some(result);
            }
            None
        })
    }

    /// Iterate over this node's children with a given field id.
    ///
    /// See also [`Node::children_by_field_name`].
    pub fn children_by_field_id<'cursor>(
        &self,
        field_id: FieldId,
        cursor: &'cursor mut TreeCursor<'tree>,
    ) -> impl Iterator<Item = Node<'tree>> + 'cursor {
        cursor.reset(*self);
        cursor.goto_first_child();
        let mut done = false;
        iter::from_fn(move || {
            if !done {
                while cursor.field_id() != Some(field_id) {
                    if !cursor.goto_next_sibling() {
                        return None;
                    }
                }
                let result = cursor.node();
                if !cursor.goto_next_sibling() {
                    done = true;
                }
                return Some(result);
            }
            None
        })
    }

    /// Get this node's immediate parent.
    /// Prefer [`child_with_descendant`](Node::child_with_descendant)
    /// for iterating over this node's ancestors.
    #[doc(alias = "ts_node_parent")]
    #[must_use]
    pub fn parent(&self) -> Option<Self> {
        Self::new(unsafe { ffi::ts_node_parent(self.0) })
    }

    /// Get the node that contains `descendant`.
    ///
    /// Note that this can return `descendant` itself.
    #[doc(alias = "ts_node_child_with_descendant")]
    #[must_use]
    pub fn child_with_descendant(&self, descendant: Self) -> Option<Self> {
        Self::new(unsafe { ffi::ts_node_child_with_descendant(self.0, descendant.0) })
    }

    /// Get this node's next sibling.
    #[doc(alias = "ts_node_next_sibling")]
    #[must_use]
    pub fn next_sibling(&self) -> Option<Self> {
        Self::new(unsafe { ffi::ts_node_next_sibling(self.0) })
    }

    /// Get this node's previous sibling.
    #[doc(alias = "ts_node_prev_sibling")]
    #[must_use]
    pub fn prev_sibling(&self) -> Option<Self> {
        Self::new(unsafe { ffi::ts_node_prev_sibling(self.0) })
    }

    /// Get this node's next named sibling.
    #[doc(alias = "ts_node_next_named_sibling")]
    #[must_use]
    pub fn next_named_sibling(&self) -> Option<Self> {
        Self::new(unsafe { ffi::ts_node_next_named_sibling(self.0) })
    }

    /// Get this node's previous named sibling.
    #[doc(alias = "ts_node_prev_named_sibling")]
    #[must_use]
    pub fn prev_named_sibling(&self) -> Option<Self> {
        Self::new(unsafe { ffi::ts_node_prev_named_sibling(self.0) })
    }

    /// Get this node's first child that contains or starts after the given byte offset.
    #[doc(alias = "ts_node_first_child_for_byte")]
    #[must_use]
    pub fn first_child_for_byte(&self, byte: usize) -> Option<Self> {
        Self::new(unsafe { ffi::ts_node_first_child_for_byte(self.0, byte as u32) })
    }

    /// Get this node's first named child that contains or starts after the given byte offset.
    #[doc(alias = "ts_node_first_named_child_for_point")]
    #[must_use]
    pub fn first_named_child_for_byte(&self, byte: usize) -> Option<Self> {
        Self::new(unsafe { ffi::ts_node_first_named_child_for_byte(self.0, byte as u32) })
    }

    /// Get the node's number of descendants, including one for the node itself.
    #[doc(alias = "ts_node_descendant_count")]
    #[must_use]
    pub fn descendant_count(&self) -> usize {
        unsafe { ffi::ts_node_descendant_count(self.0) as usize }
    }

    /// Get the smallest node within this node that spans the given byte range.
    #[doc(alias = "ts_node_descendant_for_byte_range")]
    #[must_use]
    pub fn descendant_for_byte_range(&self, start: usize, end: usize) -> Option<Self> {
        Self::new(unsafe {
            ffi::ts_node_descendant_for_byte_range(self.0, start as u32, end as u32)
        })
    }

    /// Get the smallest named node within this node that spans the given byte range.
    #[doc(alias = "ts_node_named_descendant_for_byte_range")]
    #[must_use]
    pub fn named_descendant_for_byte_range(&self, start: usize, end: usize) -> Option<Self> {
        Self::new(unsafe {
            ffi::ts_node_named_descendant_for_byte_range(self.0, start as u32, end as u32)
        })
    }

    /// Get the smallest node within this node that spans the given point range.
    #[doc(alias = "ts_node_descendant_for_point_range")]
    #[must_use]
    pub fn descendant_for_point_range(&self, start: Point, end: Point) -> Option<Self> {
        Self::new(unsafe {
            ffi::ts_node_descendant_for_point_range(self.0, start.into(), end.into())
        })
    }

    /// Get the smallest named node within this node that spans the given point range.
    #[doc(alias = "ts_node_named_descendant_for_point_range")]
    #[must_use]
    pub fn named_descendant_for_point_range(&self, start: Point, end: Point) -> Option<Self> {
        Self::new(unsafe {
            ffi::ts_node_named_descendant_for_point_range(self.0, start.into(), end.into())
        })
    }

    /// Get an S-expression representing the node.
    #[doc(alias = "ts_node_string")]
    #[must_use]
    pub fn to_sexp(&self) -> String {
        let c_string = unsafe { ffi::ts_node_string(self.0) };
        let result = unsafe { CStr::from_ptr(c_string) }
            .to_str()
            .unwrap()
            .to_string();
        unsafe { ts_free(c_string.cast::<c_void>()) };
        result
    }

    pub fn utf8_text<'a>(&self, source: &'a [u8]) -> Result<&'a str, str::Utf8Error> {
        str::from_utf8(&source[self.start_byte()..self.end_byte()])
    }

    #[must_use]
    pub fn utf16_text<'a>(&self, source: &'a [u16]) -> &'a [u16] {
        &source[self.start_byte() / 2..self.end_byte() / 2]
    }

    /// Create a new [`TreeCursor`] starting from this node.
    ///
    /// Note that the given node is considered the root of the cursor,
    /// and the cursor cannot walk outside this node.
    #[doc(alias = "ts_tree_cursor_new")]
    #[must_use]
    pub fn walk(&self) -> TreeCursor<'tree> {
        TreeCursor(unsafe { ffi::ts_tree_cursor_new(self.0) }, PhantomData)
    }
}

impl PartialEq for Node<'_> {
    fn eq(&self, other: &Self) -> bool {
        core::ptr::eq(self.0.id, other.0.id)
    }
}

impl Eq for Node<'_> {}

impl hash::Hash for Node<'_> {
    fn hash<H: hash::Hasher>(&self, state: &mut H) {
        self.0.id.hash(state);
        self.0.context[0].hash(state);
        self.0.context[1].hash(state);
        self.0.context[2].hash(state);
        self.0.context[3].hash(state);
    }
}

impl fmt::Debug for Node<'_> {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        write!(
            f,
            "{{Node {} {} - {}}}",
            self.kind(),
            self.start_position(),
            self.end_position()
        )
    }
}

impl fmt::Display for Node<'_> {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        if !f.alternate() {
            return f.write_str(&self.to_sexp());
        }
        let c_string = unsafe { ts_node_string_pretty(self.0, f.width().unwrap_or(0)) };
        let result = unsafe { CStr::from_ptr(c_string) }
            .to_str()
            .unwrap()
            .to_string();
        unsafe { ts_free(c_string.cast::<c_void>()) };
        f.write_str(&result)
    }
}

impl<'tree> TreeCursor<'tree> {
    /// Get the tree cursor's current [`Node`].
    #[doc(alias = "ts_tree_cursor_current_node")]
    #[must_use]
    pub fn node(&self) -> Node<'tree> {
        Node(
            unsafe { ffi::ts_tree_cursor_current_node(&raw const self.0) },
            PhantomData,
        )
    }

    /// Get the numerical field id of this tree cursor's current node.
    ///
    /// See also [`field_name`](TreeCursor::field_name).
    #[doc(alias = "ts_tree_cursor_current_field_id")]
    #[must_use]
    pub fn field_id(&self) -> Option<FieldId> {
        let id = unsafe { ffi::ts_tree_cursor_current_field_id(&raw const self.0) };
        FieldId::new(id)
    }

    /// Get the field name of this tree cursor's current node.
    #[doc(alias = "ts_tree_cursor_current_field_name")]
    #[must_use]
    pub fn field_name(&self) -> Option<&'tree str> {
        unsafe {
            let ptr = ffi::ts_tree_cursor_current_field_name(&raw const self.0);
            (!ptr.is_null()).then(|| CStr::from_ptr(ptr).to_str().unwrap())
        }
    }

    /// Get the depth of the cursor's current node relative to the original
    /// node that the cursor was constructed with.
    #[doc(alias = "ts_tree_cursor_current_depth")]
    #[must_use]
    pub fn depth(&self) -> u32 {
        unsafe { ffi::ts_tree_cursor_current_depth(&raw const self.0) }
    }

    /// Get the index of the cursor's current node out of all of the
    /// descendants of the original node that the cursor was constructed with
    #[doc(alias = "ts_tree_cursor_current_descendant_index")]
    #[must_use]
    pub fn descendant_index(&self) -> usize {
        unsafe { ffi::ts_tree_cursor_current_descendant_index(&raw const self.0) as usize }
    }

    /// Move this cursor to the first child of its current node.
    ///
    /// This returns `true` if the cursor successfully moved, and returns
    /// `false` if there were no children.
    #[doc(alias = "ts_tree_cursor_goto_first_child")]
    pub fn goto_first_child(&mut self) -> bool {
        unsafe { ffi::ts_tree_cursor_goto_first_child(&raw mut self.0) }
    }

    /// Move this cursor to the last child of its current node.
    ///
    /// This returns `true` if the cursor successfully moved, and returns
    /// `false` if there were no children.
    ///
    /// Note that this function may be slower than
    /// [`goto_first_child`](TreeCursor::goto_first_child) because it needs to
    /// iterate through all the children to compute the child's position.
    #[doc(alias = "ts_tree_cursor_goto_last_child")]
    pub fn goto_last_child(&mut self) -> bool {
        unsafe { ffi::ts_tree_cursor_goto_last_child(&raw mut self.0) }
    }

    /// Move this cursor to the parent of its current node.
    ///
    /// This returns `true` if the cursor successfully moved, and returns
    /// `false` if there was no parent node (the cursor was already on the
    /// root node).
    ///
    /// Note that the node the cursor was constructed with is considered the root
    /// of the cursor, and the cursor cannot walk outside this node.
    #[doc(alias = "ts_tree_cursor_goto_parent")]
    pub fn goto_parent(&mut self) -> bool {
        unsafe { ffi::ts_tree_cursor_goto_parent(&raw mut self.0) }
    }

    /// Move this cursor to the next sibling of its current node.
    ///
    /// This returns `true` if the cursor successfully moved, and returns
    /// `false` if there was no next sibling node.
    ///
    /// Note that the node the cursor was constructed with is considered the root
    /// of the cursor, and the cursor cannot walk outside this node.
    #[doc(alias = "ts_tree_cursor_goto_next_sibling")]
    pub fn goto_next_sibling(&mut self) -> bool {
        unsafe { ffi::ts_tree_cursor_goto_next_sibling(&raw mut self.0) }
    }

    /// Move the cursor to the node that is the nth descendant of
    /// the original node that the cursor was constructed with, where
    /// zero represents the original node itself.
    #[doc(alias = "ts_tree_cursor_goto_descendant")]
    pub fn goto_descendant(&mut self, descendant_index: usize) {
        unsafe { ffi::ts_tree_cursor_goto_descendant(&raw mut self.0, descendant_index as u32) }
    }

    /// Move this cursor to the previous sibling of its current node.
    ///
    /// This returns `true` if the cursor successfully moved, and returns
    /// `false` if there was no previous sibling node.
    ///
    /// Note, that this function may be slower than
    /// [`goto_next_sibling`](TreeCursor::goto_next_sibling) due to how node
    /// positions are stored. In the worst case, this will need to iterate
    /// through all the children up to the previous sibling node to recalculate
    /// its position. Also note that the node the cursor was constructed with is
    /// considered the root of the cursor, and the cursor cannot walk outside this node.
    #[doc(alias = "ts_tree_cursor_goto_previous_sibling")]
    pub fn goto_previous_sibling(&mut self) -> bool {
        unsafe { ffi::ts_tree_cursor_goto_previous_sibling(&raw mut self.0) }
    }

    /// Move this cursor to the first child of its current node that contains or
    /// starts after the given byte offset.
    ///
    /// This returns the index of the child node if one was found, and returns
    /// `None` if no such child was found.
    #[doc(alias = "ts_tree_cursor_goto_first_child_for_byte")]
    pub fn goto_first_child_for_byte(&mut self, index: usize) -> Option<usize> {
        let result =
            unsafe { ffi::ts_tree_cursor_goto_first_child_for_byte(&raw mut self.0, index as u32) };
        result.try_into().ok()
    }

    /// Move this cursor to the first child of its current node that contains or
    /// starts after the given byte offset.
    ///
    /// This returns the index of the child node if one was found, and returns
    /// `None` if no such child was found.
    #[doc(alias = "ts_tree_cursor_goto_first_child_for_point")]
    pub fn goto_first_child_for_point(&mut self, point: Point) -> Option<usize> {
        let result = unsafe {
            ffi::ts_tree_cursor_goto_first_child_for_point(&raw mut self.0, point.into())
        };
        result.try_into().ok()
    }

    /// Re-initialize this tree cursor to start at the original node that the
    /// cursor was constructed with.
    #[doc(alias = "ts_tree_cursor_reset")]
    pub fn reset(&mut self, node: Node<'tree>) {
        unsafe { ffi::ts_tree_cursor_reset(&raw mut self.0, node.0) };
    }

    /// Re-initialize a tree cursor to the same position as another cursor.
    ///
    /// Unlike [`reset`](TreeCursor::reset), this will not lose parent
    /// information and allows reusing already created cursors.
    #[doc(alias = "ts_tree_cursor_reset_to")]
    pub fn reset_to(&mut self, cursor: &Self) {
        unsafe { ffi::ts_tree_cursor_reset_to(&raw mut self.0, &raw const cursor.0) };
    }
}

impl Clone for TreeCursor<'_> {
    fn clone(&self) -> Self {
        TreeCursor(
            unsafe { ffi::ts_tree_cursor_copy(&raw const self.0) },
            PhantomData,
        )
    }
}

impl Drop for TreeCursor<'_> {
    fn drop(&mut self) {
        unsafe { ffi::ts_tree_cursor_delete(&raw mut self.0) }
    }
}

impl Point {
    #[must_use]
    pub const fn new(row: usize, column: usize) -> Self {
        Self { row, column }
    }
}

impl fmt::Display for Point {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        write!(f, "({}, {})", self.row, self.column)
    }
}

impl From<Point> for ffi::TSPoint {
    fn from(val: Point) -> Self {
        Self {
            row: val.row as u32,
            column: val.column as u32,
        }
    }
}

impl From<ffi::TSPoint> for Point {
    fn from(point: ffi::TSPoint) -> Self {
        Self {
            row: point.row as usize,
            column: point.column as usize,
        }
    }
}

impl From<Range> for ffi::TSRange {
    fn from(val: Range) -> Self {
        Self {
            start_byte: val.start_byte as u32,
            end_byte: val.end_byte as u32,
            start_point: val.start_point.into(),
            end_point: val.end_point.into(),
        }
    }
}

impl From<ffi::TSRange> for Range {
    fn from(range: ffi::TSRange) -> Self {
        Self {
            start_byte: range.start_byte as usize,
            end_byte: range.end_byte as usize,
            start_point: range.start_point.into(),
            end_point: range.end_point.into(),
        }
    }
}

impl fmt::Display for IncludedRangesError {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        write!(f, "Incorrect range by index: {}", self.0)
    }
}

impl fmt::Display for LanguageError {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        match self {
            Self::Version(version) => {
                write!(
                    f,
                    "Incompatible language version {version:#x}. Expected {LANGUAGE_VERSION:#x}",
                )
            }
            Self::NotParseable => {
                write!(f, "Language cannot be used for parsing.")
            }
            #[cfg(feature = "wasm")]
            Self::Wasm => {
                write!(f, "Failed to load the Wasm store.")
            }
        }
    }
}

pub fn wasm_stdlib_symbols() -> impl Iterator<Item = &'static str> {
    const WASM_STDLIB_SYMBOLS: &str = include_str!(concat!(env!("OUT_DIR"), "/stdlib-symbols.txt"));

    WASM_STDLIB_SYMBOLS
        .lines()
        .map(|s| s.trim_matches(|c| c == '"' || c == ','))
}

unsafe extern "C" {
    fn ts_node_string_pretty(node: ffi::TSNode, initial_indent_level: usize) -> *mut c_char;
    static mut ts_current_free: unsafe extern "C" fn(ptr: *mut c_void);
}

/// Frees a pointer that was returned by a tree-sitter C API using whichever `free`
/// function is currently installed via [`set_allocator`] or [`ffi::ts_set_allocator`].
#[inline]
unsafe fn ts_free(ptr: *mut c_void) {
    let f = unsafe { core::ptr::addr_of!(ts_current_free).read() };
    unsafe { f(ptr) };
}

/// A complete set of memory allocation functions for the core C library.
#[derive(Copy, Clone)]
pub struct Allocator {
    pub malloc: unsafe extern "C" fn(size: usize) -> *mut c_void,
    pub calloc: unsafe extern "C" fn(nmemb: usize, size: usize) -> *mut c_void,
    pub realloc: unsafe extern "C" fn(ptr: *mut c_void, size: usize) -> *mut c_void,
    pub free: unsafe extern "C" fn(ptr: *mut c_void),
}

/// Replaces the memory allocation functions used by the core C library.
///
/// Pass `Some` to install an allocator, or `None` to restore libc defaults.
///
/// # Safety
///
/// All of the following must hold:
///
/// - The four functions must belong to a single allocator family.
///
/// - The functions must not return null for non-zero allocs.
///
/// - Returned pointers must satisfy the alignment that libc `malloc` provides.
///
/// - Call this before any other tree-sitter API call, and do not call it again
///   while live tree-sitter objects exist.
///
/// - This function is not thread-safe.
#[doc(alias = "ts_set_allocator")]
pub unsafe fn set_allocator(allocator: Option<Allocator>) {
    let (m, c, r, f) = match allocator {
        Some(a) => (
            Some(a.malloc),
            Some(a.calloc),
            Some(a.realloc),
            Some(a.free),
        ),
        None => (None, None, None, None),
    };
    unsafe { ffi::ts_set_allocator(m, c, r, f) };
}

#[cfg(feature = "std")]
#[cfg_attr(docsrs, doc(cfg(feature = "std")))]
impl error::Error for IncludedRangesError {}
#[cfg(feature = "std")]
#[cfg_attr(docsrs, doc(cfg(feature = "std")))]
impl error::Error for LanguageError {}

#[cfg(not(target_family = "wasm"))]
unsafe impl Send for Language {}
#[cfg(not(target_family = "wasm"))]
unsafe impl Sync for Language {}

unsafe impl Send for Node<'_> {}
unsafe impl Sync for Node<'_> {}

#[cfg(not(target_family = "wasm"))]
unsafe impl Send for Parser {}
#[cfg(not(target_family = "wasm"))]
unsafe impl Sync for Parser {}

unsafe impl Send for Tree {}
unsafe impl Sync for Tree {}

unsafe impl Send for TreeCursor<'_> {}
unsafe impl Sync for TreeCursor<'_> {}