mirror of https://github.com/helix-editor/helix
You cannot select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
1267 lines
39 KiB
C
1267 lines
39 KiB
C
#ifndef TREE_SITTER_API_H_
|
|
#define TREE_SITTER_API_H_
|
|
|
|
#ifndef TREE_SITTER_HIDE_SYMBOLS
|
|
#if defined(__GNUC__) || defined(__clang__)
|
|
#pragma GCC visibility push(default)
|
|
#endif
|
|
#endif
|
|
|
|
#ifdef __cplusplus
|
|
extern "C" {
|
|
#endif
|
|
|
|
#include <stdlib.h>
|
|
#include <stdint.h>
|
|
#include <stdbool.h>
|
|
|
|
/****************************/
|
|
/* Section - ABI Versioning */
|
|
/****************************/
|
|
|
|
/**
|
|
* The latest ABI version that is supported by the current version of the
|
|
* library. When Languages are generated by the Tree-sitter CLI, they are
|
|
* assigned an ABI version number that corresponds to the current CLI version.
|
|
* The Tree-sitter library is generally backwards-compatible with languages
|
|
* generated using older CLI versions, but is not forwards-compatible.
|
|
*/
|
|
#define TREE_SITTER_LANGUAGE_VERSION 14
|
|
|
|
/**
|
|
* The earliest ABI version that is supported by the current version of the
|
|
* library.
|
|
*/
|
|
#define TREE_SITTER_MIN_COMPATIBLE_LANGUAGE_VERSION 13
|
|
|
|
/*******************/
|
|
/* Section - Types */
|
|
/*******************/
|
|
|
|
typedef uint16_t TSStateId;
|
|
typedef uint16_t TSSymbol;
|
|
typedef uint16_t TSFieldId;
|
|
typedef struct TSLanguage TSLanguage;
|
|
typedef struct TSParser TSParser;
|
|
typedef struct TSTree TSTree;
|
|
typedef struct TSQuery TSQuery;
|
|
typedef struct TSQueryCursor TSQueryCursor;
|
|
typedef struct TSLookaheadIterator TSLookaheadIterator;
|
|
|
|
typedef enum TSInputEncoding {
|
|
TSInputEncodingUTF8,
|
|
TSInputEncodingUTF16,
|
|
} TSInputEncoding;
|
|
|
|
typedef enum TSSymbolType {
|
|
TSSymbolTypeRegular,
|
|
TSSymbolTypeAnonymous,
|
|
TSSymbolTypeAuxiliary,
|
|
} TSSymbolType;
|
|
|
|
typedef struct TSPoint {
|
|
uint32_t row;
|
|
uint32_t column;
|
|
} TSPoint;
|
|
|
|
typedef struct TSRange {
|
|
TSPoint start_point;
|
|
TSPoint end_point;
|
|
uint32_t start_byte;
|
|
uint32_t end_byte;
|
|
} TSRange;
|
|
|
|
typedef struct TSInput {
|
|
void *payload;
|
|
const char *(*read)(void *payload, uint32_t byte_index, TSPoint position, uint32_t *bytes_read);
|
|
TSInputEncoding encoding;
|
|
} TSInput;
|
|
|
|
typedef enum TSLogType {
|
|
TSLogTypeParse,
|
|
TSLogTypeLex,
|
|
} TSLogType;
|
|
|
|
typedef struct TSLogger {
|
|
void *payload;
|
|
void (*log)(void *payload, TSLogType log_type, const char *buffer);
|
|
} TSLogger;
|
|
|
|
typedef struct TSInputEdit {
|
|
uint32_t start_byte;
|
|
uint32_t old_end_byte;
|
|
uint32_t new_end_byte;
|
|
TSPoint start_point;
|
|
TSPoint old_end_point;
|
|
TSPoint new_end_point;
|
|
} TSInputEdit;
|
|
|
|
typedef struct TSNode {
|
|
uint32_t context[4];
|
|
const void *id;
|
|
const TSTree *tree;
|
|
} TSNode;
|
|
|
|
typedef struct TSTreeCursor {
|
|
const void *tree;
|
|
const void *id;
|
|
uint32_t context[3];
|
|
} TSTreeCursor;
|
|
|
|
typedef struct TSQueryCapture {
|
|
TSNode node;
|
|
uint32_t index;
|
|
} TSQueryCapture;
|
|
|
|
typedef enum TSQuantifier {
|
|
TSQuantifierZero = 0, // must match the array initialization value
|
|
TSQuantifierZeroOrOne,
|
|
TSQuantifierZeroOrMore,
|
|
TSQuantifierOne,
|
|
TSQuantifierOneOrMore,
|
|
} TSQuantifier;
|
|
|
|
typedef struct TSQueryMatch {
|
|
uint32_t id;
|
|
uint16_t pattern_index;
|
|
uint16_t capture_count;
|
|
const TSQueryCapture *captures;
|
|
} TSQueryMatch;
|
|
|
|
typedef enum TSQueryPredicateStepType {
|
|
TSQueryPredicateStepTypeDone,
|
|
TSQueryPredicateStepTypeCapture,
|
|
TSQueryPredicateStepTypeString,
|
|
} TSQueryPredicateStepType;
|
|
|
|
typedef struct TSQueryPredicateStep {
|
|
TSQueryPredicateStepType type;
|
|
uint32_t value_id;
|
|
} TSQueryPredicateStep;
|
|
|
|
typedef enum TSQueryError {
|
|
TSQueryErrorNone = 0,
|
|
TSQueryErrorSyntax,
|
|
TSQueryErrorNodeType,
|
|
TSQueryErrorField,
|
|
TSQueryErrorCapture,
|
|
TSQueryErrorStructure,
|
|
TSQueryErrorLanguage,
|
|
} TSQueryError;
|
|
|
|
/********************/
|
|
/* Section - Parser */
|
|
/********************/
|
|
|
|
/**
|
|
* Create a new parser.
|
|
*/
|
|
TSParser *ts_parser_new(void);
|
|
|
|
/**
|
|
* Delete the parser, freeing all of the memory that it used.
|
|
*/
|
|
void ts_parser_delete(TSParser *self);
|
|
|
|
/**
|
|
* Get the parser's current language.
|
|
*/
|
|
const TSLanguage *ts_parser_language(const TSParser *self);
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
bool ts_parser_set_language(TSParser *self, const TSLanguage *language);
|
|
|
|
/**
|
|
* 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`
|
|
*/
|
|
bool ts_parser_set_included_ranges(
|
|
TSParser *self,
|
|
const TSRange *ranges,
|
|
uint32_t count
|
|
);
|
|
|
|
/**
|
|
* Get the ranges of text that the parser will include when parsing.
|
|
*
|
|
* The returned pointer is owned by the parser. The caller should not free it
|
|
* or write to it. The length of the array will be written to the given
|
|
* `count` pointer.
|
|
*/
|
|
const TSRange *ts_parser_included_ranges(
|
|
const TSParser *self,
|
|
uint32_t *count
|
|
);
|
|
|
|
/**
|
|
* 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
|
|
*/
|
|
TSTree *ts_parser_parse(
|
|
TSParser *self,
|
|
const TSTree *old_tree,
|
|
TSInput input
|
|
);
|
|
|
|
/**
|
|
* Use the parser to parse some source code stored in one contiguous buffer.
|
|
* The first two parameters are the same as in the [`ts_parser_parse`] function
|
|
* above. The second two parameters indicate the location of the buffer and its
|
|
* length in bytes.
|
|
*/
|
|
TSTree *ts_parser_parse_string(
|
|
TSParser *self,
|
|
const TSTree *old_tree,
|
|
const char *string,
|
|
uint32_t length
|
|
);
|
|
|
|
/**
|
|
* Use the parser to parse some source code stored in one contiguous buffer with
|
|
* a given encoding. The first four parameters work the same as in the
|
|
* [`ts_parser_parse_string`] method above. The final parameter indicates whether
|
|
* the text is encoded as UTF8 or UTF16.
|
|
*/
|
|
TSTree *ts_parser_parse_string_encoding(
|
|
TSParser *self,
|
|
const TSTree *old_tree,
|
|
const char *string,
|
|
uint32_t length,
|
|
TSInputEncoding encoding
|
|
);
|
|
|
|
/**
|
|
* Instruct the parser to start the next parse from the beginning.
|
|
*
|
|
* If the parser previously failed because of a timeout or a cancellation, then
|
|
* by default, it will resume where it left off on the next call to
|
|
* [`ts_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 [`ts_parser_reset`] first.
|
|
*/
|
|
void ts_parser_reset(TSParser *self);
|
|
|
|
/**
|
|
* Set the maximum duration in microseconds that parsing should be allowed to
|
|
* take before halting.
|
|
*
|
|
* If parsing takes longer than this, it will halt early, returning NULL.
|
|
* See [`ts_parser_parse`] for more information.
|
|
*/
|
|
void ts_parser_set_timeout_micros(TSParser *self, uint64_t timeout_micros);
|
|
|
|
/**
|
|
* Get the duration in microseconds that parsing is allowed to take.
|
|
*/
|
|
uint64_t ts_parser_timeout_micros(const TSParser *self);
|
|
|
|
/**
|
|
* Set the parser's current cancellation flag pointer.
|
|
*
|
|
* If a non-null pointer is assigned, then the parser will periodically read
|
|
* from this pointer during parsing. If it reads a non-zero value, it will
|
|
* halt early, returning NULL. See [`ts_parser_parse`] for more information.
|
|
*/
|
|
void ts_parser_set_cancellation_flag(TSParser *self, const size_t *flag);
|
|
|
|
/**
|
|
* Get the parser's current cancellation flag pointer.
|
|
*/
|
|
const size_t *ts_parser_cancellation_flag(const TSParser *self);
|
|
|
|
/**
|
|
* Set the logger that a parser should use during parsing.
|
|
*
|
|
* The parser does not take ownership over the logger payload. If a logger was
|
|
* previously assigned, the caller is responsible for releasing any memory
|
|
* owned by the previous logger.
|
|
*/
|
|
void ts_parser_set_logger(TSParser *self, TSLogger logger);
|
|
|
|
/**
|
|
* Get the parser's current logger.
|
|
*/
|
|
TSLogger ts_parser_logger(const TSParser *self);
|
|
|
|
/**
|
|
* Set the file descriptor 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. You can turn off this logging by passing a negative number.
|
|
*/
|
|
void ts_parser_print_dot_graphs(TSParser *self, int fd);
|
|
|
|
/******************/
|
|
/* Section - Tree */
|
|
/******************/
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
TSTree *ts_tree_copy(const TSTree *self);
|
|
|
|
/**
|
|
* Delete the syntax tree, freeing all of the memory that it used.
|
|
*/
|
|
void ts_tree_delete(TSTree *self);
|
|
|
|
/**
|
|
* Get the root node of the syntax tree.
|
|
*/
|
|
TSNode ts_tree_root_node(const TSTree *self);
|
|
|
|
/**
|
|
* Get the root node of the syntax tree, but with its position
|
|
* shifted forward by the given offset.
|
|
*/
|
|
TSNode ts_tree_root_node_with_offset(
|
|
const TSTree *self,
|
|
uint32_t offset_bytes,
|
|
TSPoint offset_extent
|
|
);
|
|
|
|
/**
|
|
* Get the language that was used to parse the syntax tree.
|
|
*/
|
|
const TSLanguage *ts_tree_language(const TSTree *self);
|
|
|
|
/**
|
|
* Get the array of included ranges that was used to parse the syntax tree.
|
|
*
|
|
* The returned pointer must be freed by the caller.
|
|
*/
|
|
TSRange *ts_tree_included_ranges(const TSTree *self, uint32_t *length);
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
void ts_tree_edit(TSTree *self, const TSInputEdit *edit);
|
|
|
|
/**
|
|
* Compare an old edited syntax tree to a new syntax tree representing the same
|
|
* document, returning an array of ranges whose syntactic structure has changed.
|
|
*
|
|
* For this to work correctly, the old syntax tree must have been edited such
|
|
* that its ranges match up to the new tree. Generally, you'll want to call
|
|
* this function right after calling one of the [`ts_parser_parse`] functions.
|
|
* You need to pass the old tree that was passed to parse, as well as the new
|
|
* tree that was returned from that function.
|
|
*
|
|
* The returned array is allocated using `malloc` and the caller is responsible
|
|
* for freeing it using `free`. The length of the array will be written to the
|
|
* given `length` pointer.
|
|
*/
|
|
TSRange *ts_tree_get_changed_ranges(
|
|
const TSTree *old_tree,
|
|
const TSTree *new_tree,
|
|
uint32_t *length
|
|
);
|
|
|
|
/**
|
|
* Write a DOT graph describing the syntax tree to the given file.
|
|
*/
|
|
void ts_tree_print_dot_graph(const TSTree *self, int file_descriptor);
|
|
|
|
/******************/
|
|
/* Section - Node */
|
|
/******************/
|
|
|
|
/**
|
|
* Get the node's type as a null-terminated string.
|
|
*/
|
|
const char *ts_node_type(TSNode self);
|
|
|
|
/**
|
|
* Get the node's type as a numerical id.
|
|
*/
|
|
TSSymbol ts_node_symbol(TSNode self);
|
|
|
|
/**
|
|
* Get the node's language.
|
|
*/
|
|
const TSLanguage *ts_node_language(TSNode self);
|
|
|
|
/**
|
|
* Get the node's type as it appears in the grammar ignoring aliases as a
|
|
* null-terminated string.
|
|
*/
|
|
const char *ts_node_grammar_type(TSNode self);
|
|
|
|
/**
|
|
* Get the node's type as a numerical id as it appears in the grammar ignoring
|
|
* aliases. This should be used in [`ts_language_next_state`] instead of
|
|
* [`ts_node_symbol`].
|
|
*/
|
|
TSSymbol ts_node_grammar_symbol(TSNode self);
|
|
|
|
/**
|
|
* Get the node's start byte.
|
|
*/
|
|
uint32_t ts_node_start_byte(TSNode self);
|
|
|
|
/**
|
|
* Get the node's start position in terms of rows and columns.
|
|
*/
|
|
TSPoint ts_node_start_point(TSNode self);
|
|
|
|
/**
|
|
* Get the node's end byte.
|
|
*/
|
|
uint32_t ts_node_end_byte(TSNode self);
|
|
|
|
/**
|
|
* Get the node's end position in terms of rows and columns.
|
|
*/
|
|
TSPoint ts_node_end_point(TSNode self);
|
|
|
|
/**
|
|
* Get an S-expression representing the node as a string.
|
|
*
|
|
* This string is allocated with `malloc` and the caller is responsible for
|
|
* freeing it using `free`.
|
|
*/
|
|
char *ts_node_string(TSNode self);
|
|
|
|
/**
|
|
* Check if the node is null. Functions like [`ts_node_child`] and
|
|
* [`ts_node_next_sibling`] will return a null node to indicate that no such node
|
|
* was found.
|
|
*/
|
|
bool ts_node_is_null(TSNode self);
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
bool ts_node_is_named(TSNode self);
|
|
|
|
/**
|
|
* Check if the node is *missing*. Missing nodes are inserted by the parser in
|
|
* order to recover from certain kinds of syntax errors.
|
|
*/
|
|
bool ts_node_is_missing(TSNode self);
|
|
|
|
/**
|
|
* Check if the node is *extra*. Extra nodes represent things like comments,
|
|
* which are not required the grammar, but can appear anywhere.
|
|
*/
|
|
bool ts_node_is_extra(TSNode self);
|
|
|
|
/**
|
|
* Check if a syntax node has been edited.
|
|
*/
|
|
bool ts_node_has_changes(TSNode self);
|
|
|
|
/**
|
|
* Check if the node is a syntax error or contains any syntax errors.
|
|
*/
|
|
bool ts_node_has_error(TSNode self);
|
|
|
|
/**
|
|
* Check if the node is a syntax error.
|
|
*/
|
|
bool ts_node_is_error(TSNode self);
|
|
|
|
/**
|
|
* Get this node's parse state.
|
|
*/
|
|
TSStateId ts_node_parse_state(TSNode self);
|
|
|
|
/**
|
|
* Get the parse state after this node.
|
|
*/
|
|
TSStateId ts_node_next_parse_state(TSNode self);
|
|
|
|
/**
|
|
* Get the node's immediate parent.
|
|
*/
|
|
TSNode ts_node_parent(TSNode self);
|
|
|
|
/**
|
|
* Get the node's child at the given index, where zero represents the first
|
|
* child.
|
|
*/
|
|
TSNode ts_node_child(TSNode self, uint32_t child_index);
|
|
|
|
/**
|
|
* Get the field name for node's child at the given index, where zero represents
|
|
* the first child. Returns NULL, if no field is found.
|
|
*/
|
|
const char *ts_node_field_name_for_child(TSNode self, uint32_t child_index);
|
|
|
|
/**
|
|
* Get the node's number of children.
|
|
*/
|
|
uint32_t ts_node_child_count(TSNode self);
|
|
|
|
/**
|
|
* Get the node's *named* child at the given index.
|
|
*
|
|
* See also [`ts_node_is_named`].
|
|
*/
|
|
TSNode ts_node_named_child(TSNode self, uint32_t child_index);
|
|
|
|
/**
|
|
* Get the node's number of *named* children.
|
|
*
|
|
* See also [`ts_node_is_named`].
|
|
*/
|
|
uint32_t ts_node_named_child_count(TSNode self);
|
|
|
|
/**
|
|
* Get the node's child with the given field name.
|
|
*/
|
|
TSNode ts_node_child_by_field_name(
|
|
TSNode self,
|
|
const char *name,
|
|
uint32_t name_length
|
|
);
|
|
|
|
/**
|
|
* Get the node's child with the given numerical field id.
|
|
*
|
|
* You can convert a field name to an id using the
|
|
* [`ts_language_field_id_for_name`] function.
|
|
*/
|
|
TSNode ts_node_child_by_field_id(TSNode self, TSFieldId field_id);
|
|
|
|
/**
|
|
* Get the node's next / previous sibling.
|
|
*/
|
|
TSNode ts_node_next_sibling(TSNode self);
|
|
TSNode ts_node_prev_sibling(TSNode self);
|
|
|
|
/**
|
|
* Get the node's next / previous *named* sibling.
|
|
*/
|
|
TSNode ts_node_next_named_sibling(TSNode self);
|
|
TSNode ts_node_prev_named_sibling(TSNode self);
|
|
|
|
/**
|
|
* Get the node's first child that extends beyond the given byte offset.
|
|
*/
|
|
TSNode ts_node_first_child_for_byte(TSNode self, uint32_t byte);
|
|
|
|
/**
|
|
* Get the node's first named child that extends beyond the given byte offset.
|
|
*/
|
|
TSNode ts_node_first_named_child_for_byte(TSNode self, uint32_t byte);
|
|
|
|
/**
|
|
* Get the node's number of descendants, including one for the node itself.
|
|
*/
|
|
uint32_t ts_node_descendant_count(TSNode self);
|
|
|
|
/**
|
|
* Get the smallest node within this node that spans the given range of bytes
|
|
* or (row, column) positions.
|
|
*/
|
|
TSNode ts_node_descendant_for_byte_range(TSNode self, uint32_t start, uint32_t end);
|
|
TSNode ts_node_descendant_for_point_range(TSNode self, TSPoint start, TSPoint end);
|
|
|
|
/**
|
|
* Get the smallest named node within this node that spans the given range of
|
|
* bytes or (row, column) positions.
|
|
*/
|
|
TSNode ts_node_named_descendant_for_byte_range(TSNode self, uint32_t start, uint32_t end);
|
|
TSNode ts_node_named_descendant_for_point_range(TSNode self, TSPoint start, TSPoint end);
|
|
|
|
/**
|
|
* Edit the node to keep it in-sync with source code that has been edited.
|
|
*
|
|
* This function is only rarely needed. When you edit a syntax tree with the
|
|
* [`ts_tree_edit`] function, all of the nodes that you retrieve from the tree
|
|
* afterward will already reflect the edit. You only need to use [`ts_node_edit`]
|
|
* when you have a [`TSNode`] instance that you want to keep and continue to use
|
|
* after an edit.
|
|
*/
|
|
void ts_node_edit(TSNode *self, const TSInputEdit *edit);
|
|
|
|
/**
|
|
* Check if two nodes are identical.
|
|
*/
|
|
bool ts_node_eq(TSNode self, TSNode other);
|
|
|
|
/************************/
|
|
/* Section - TreeCursor */
|
|
/************************/
|
|
|
|
/**
|
|
* Create a new tree cursor starting from the given node.
|
|
*
|
|
* A tree cursor allows you to walk a syntax tree more efficiently than is
|
|
* possible using the [`TSNode`] functions. It is a mutable object that is always
|
|
* on a certain syntax node, and can be moved imperatively to different nodes.
|
|
*/
|
|
TSTreeCursor ts_tree_cursor_new(TSNode node);
|
|
|
|
/**
|
|
* Delete a tree cursor, freeing all of the memory that it used.
|
|
*/
|
|
void ts_tree_cursor_delete(TSTreeCursor *self);
|
|
|
|
/**
|
|
* Re-initialize a tree cursor to start at a different node.
|
|
*/
|
|
void ts_tree_cursor_reset(TSTreeCursor *self, TSNode node);
|
|
|
|
/**
|
|
* Re-initialize a tree cursor to the same position as another cursor.
|
|
*
|
|
* Unlike [`ts_tree_cursor_reset`], this will not lose parent information and
|
|
* allows reusing already created cursors.
|
|
*/
|
|
void ts_tree_cursor_reset_to(TSTreeCursor *dst, const TSTreeCursor *src);
|
|
|
|
/**
|
|
* Get the tree cursor's current node.
|
|
*/
|
|
TSNode ts_tree_cursor_current_node(const TSTreeCursor *self);
|
|
|
|
/**
|
|
* Get the field name of the tree cursor's current node.
|
|
*
|
|
* This returns `NULL` if the current node doesn't have a field.
|
|
* See also [`ts_node_child_by_field_name`].
|
|
*/
|
|
const char *ts_tree_cursor_current_field_name(const TSTreeCursor *self);
|
|
|
|
/**
|
|
* Get the field id of the tree cursor's current node.
|
|
*
|
|
* This returns zero if the current node doesn't have a field.
|
|
* See also [`ts_node_child_by_field_id`], [`ts_language_field_id_for_name`].
|
|
*/
|
|
TSFieldId ts_tree_cursor_current_field_id(const TSTreeCursor *self);
|
|
|
|
/**
|
|
* Move the 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).
|
|
*/
|
|
bool ts_tree_cursor_goto_parent(TSTreeCursor *self);
|
|
|
|
/**
|
|
* Move the 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.
|
|
*/
|
|
bool ts_tree_cursor_goto_next_sibling(TSTreeCursor *self);
|
|
|
|
/**
|
|
* Move the 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
|
|
* [`ts_tree_cursor_goto_next_sibling`] due to how node positions are stored. In
|
|
* the worst case, this will need to iterate through all the children upto the
|
|
* previous sibling node to recalculate its position.
|
|
*/
|
|
bool ts_tree_cursor_goto_previous_sibling(TSTreeCursor *self);
|
|
|
|
/**
|
|
* Move the 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.
|
|
*/
|
|
bool ts_tree_cursor_goto_first_child(TSTreeCursor *self);
|
|
|
|
/**
|
|
* Move the 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 [`ts_tree_cursor_goto_first_child`]
|
|
* because it needs to iterate through all the children to compute the child's
|
|
* position.
|
|
*/
|
|
bool ts_tree_cursor_goto_last_child(TSTreeCursor *self);
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
void ts_tree_cursor_goto_descendant(TSTreeCursor *self, uint32_t goal_descendant_index);
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
uint32_t ts_tree_cursor_current_descendant_index(const TSTreeCursor *self);
|
|
|
|
/**
|
|
* Get the depth of the cursor's current node relative to the original
|
|
* node that the cursor was constructed with.
|
|
*/
|
|
uint32_t ts_tree_cursor_current_depth(const TSTreeCursor *self);
|
|
|
|
/**
|
|
* Move the cursor to the first child of its current node that extends beyond
|
|
* the given byte offset or point.
|
|
*
|
|
* This returns the index of the child node if one was found, and returns -1
|
|
* if no such child was found.
|
|
*/
|
|
int64_t ts_tree_cursor_goto_first_child_for_byte(TSTreeCursor *self, uint32_t goal_byte);
|
|
int64_t ts_tree_cursor_goto_first_child_for_point(TSTreeCursor *self, TSPoint goal_point);
|
|
|
|
TSTreeCursor ts_tree_cursor_copy(const TSTreeCursor *cursor);
|
|
|
|
/*******************/
|
|
/* Section - Query */
|
|
/*******************/
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
TSQuery *ts_query_new(
|
|
const TSLanguage *language,
|
|
const char *source,
|
|
uint32_t source_len,
|
|
uint32_t *error_offset,
|
|
TSQueryError *error_type
|
|
);
|
|
|
|
/**
|
|
* Delete a query, freeing all of the memory that it used.
|
|
*/
|
|
void ts_query_delete(TSQuery *self);
|
|
|
|
/**
|
|
* Get the number of patterns, captures, or string literals in the query.
|
|
*/
|
|
uint32_t ts_query_pattern_count(const TSQuery *self);
|
|
uint32_t ts_query_capture_count(const TSQuery *self);
|
|
uint32_t ts_query_string_count(const TSQuery *self);
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
uint32_t ts_query_start_byte_for_pattern(const TSQuery *self, uint32_t pattern_index);
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
const TSQueryPredicateStep *ts_query_predicates_for_pattern(
|
|
const TSQuery *self,
|
|
uint32_t pattern_index,
|
|
uint32_t *step_count
|
|
);
|
|
|
|
/*
|
|
* Check if the given pattern in the query has a single root node.
|
|
*/
|
|
bool ts_query_is_pattern_rooted(const TSQuery *self, uint32_t pattern_index);
|
|
|
|
/*
|
|
* Check if the given pattern in the query is 'non local'.
|
|
*
|
|
* A non-local pattern has multiple root nodes and can match within a
|
|
* repeating sequence of nodes, as specified by the grammar. Non-local
|
|
* patterns disable certain optimizations that would otherwise be possible
|
|
* when executing a query on a specific range of a syntax tree.
|
|
*/
|
|
bool ts_query_is_pattern_non_local(const TSQuery *self, uint32_t pattern_index);
|
|
|
|
/*
|
|
* Check if a given pattern is guaranteed to match once a given step is reached.
|
|
* The step is specified by its byte offset in the query's source code.
|
|
*/
|
|
bool ts_query_is_pattern_guaranteed_at_step(const TSQuery *self, uint32_t byte_offset);
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
const char *ts_query_capture_name_for_id(
|
|
const TSQuery *self,
|
|
uint32_t index,
|
|
uint32_t *length
|
|
);
|
|
|
|
/**
|
|
* Get the quantifier of the query's captures. Each capture is * associated
|
|
* with a numeric id based on the order that it appeared in the query's source.
|
|
*/
|
|
TSQuantifier ts_query_capture_quantifier_for_id(
|
|
const TSQuery *self,
|
|
uint32_t pattern_index,
|
|
uint32_t capture_index
|
|
);
|
|
|
|
const char *ts_query_string_value_for_id(
|
|
const TSQuery *self,
|
|
uint32_t index,
|
|
uint32_t *length
|
|
);
|
|
|
|
/**
|
|
* Disable a certain capture within a query.
|
|
*
|
|
* This prevents the capture from being returned in matches, and also avoids
|
|
* any resource usage associated with recording the capture. Currently, there
|
|
* is no way to undo this.
|
|
*/
|
|
void ts_query_disable_capture(TSQuery *self, const char *name, uint32_t length);
|
|
|
|
/**
|
|
* Disable a certain pattern within a query.
|
|
*
|
|
* This prevents the pattern from matching and removes most of the overhead
|
|
* associated with the pattern. Currently, there is no way to undo this.
|
|
*/
|
|
void ts_query_disable_pattern(TSQuery *self, uint32_t pattern_index);
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
TSQueryCursor *ts_query_cursor_new(void);
|
|
|
|
/**
|
|
* Delete a query cursor, freeing all of the memory that it used.
|
|
*/
|
|
void ts_query_cursor_delete(TSQueryCursor *self);
|
|
|
|
/**
|
|
* Start running a given query on a given node.
|
|
*/
|
|
void ts_query_cursor_exec(TSQueryCursor *self, const TSQuery *query, TSNode node);
|
|
|
|
/**
|
|
* 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.
|
|
*/
|
|
bool ts_query_cursor_did_exceed_match_limit(const TSQueryCursor *self);
|
|
uint32_t ts_query_cursor_match_limit(const TSQueryCursor *self);
|
|
void ts_query_cursor_set_match_limit(TSQueryCursor *self, uint32_t limit);
|
|
|
|
/**
|
|
* Set the range of bytes or (row, column) positions in which the query
|
|
* will be executed.
|
|
*/
|
|
void ts_query_cursor_set_byte_range(TSQueryCursor *self, uint32_t start_byte, uint32_t end_byte);
|
|
void ts_query_cursor_set_point_range(TSQueryCursor *self, TSPoint start_point, TSPoint end_point);
|
|
|
|
/**
|
|
* 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`.
|
|
*/
|
|
bool ts_query_cursor_next_match(TSQueryCursor *self, TSQueryMatch *match);
|
|
void ts_query_cursor_remove_match(TSQueryCursor *self, uint32_t match_id);
|
|
|
|
/**
|
|
* 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`.
|
|
*/
|
|
bool ts_query_cursor_next_capture(
|
|
TSQueryCursor *self,
|
|
TSQueryMatch *match,
|
|
uint32_t *capture_index
|
|
);
|
|
|
|
/**
|
|
* Set the maximum start depth for a query cursor.
|
|
*
|
|
* This prevents cursors from exploring children nodes at a certain depth.
|
|
* Note if a pattern includes many children, then they will still be checked.
|
|
*
|
|
* The zero max start depth value can be used as a special behavior and
|
|
* it helps to destructure a subtree by staying on a node and using captures
|
|
* for interested parts. Note that the zero max start depth only limit a search
|
|
* depth for a pattern's root node but other nodes that are parts of the pattern
|
|
* may be searched at any depth what defined by the pattern structure.
|
|
*
|
|
* Set to `UINT32_MAX` to remove the maximum start depth.
|
|
*/
|
|
void ts_query_cursor_set_max_start_depth(TSQueryCursor *self, uint32_t max_start_depth);
|
|
|
|
/**********************/
|
|
/* Section - Language */
|
|
/**********************/
|
|
|
|
/**
|
|
* Get another reference to the given language.
|
|
*/
|
|
const TSLanguage *ts_language_copy(const TSLanguage *self);
|
|
|
|
/**
|
|
* Free any dynamically-allocated resources for this language, if
|
|
* this is the last reference.
|
|
*/
|
|
void ts_language_delete(const TSLanguage *self);
|
|
|
|
/**
|
|
* Get the number of distinct node types in the language.
|
|
*/
|
|
uint32_t ts_language_symbol_count(const TSLanguage *self);
|
|
|
|
/**
|
|
* Get the number of valid states in this language.
|
|
*/
|
|
uint32_t ts_language_state_count(const TSLanguage *self);
|
|
|
|
/**
|
|
* Get a node type string for the given numerical id.
|
|
*/
|
|
const char *ts_language_symbol_name(const TSLanguage *self, TSSymbol symbol);
|
|
|
|
/**
|
|
* Get the numerical id for the given node type string.
|
|
*/
|
|
TSSymbol ts_language_symbol_for_name(
|
|
const TSLanguage *self,
|
|
const char *string,
|
|
uint32_t length,
|
|
bool is_named
|
|
);
|
|
|
|
/**
|
|
* Get the number of distinct field names in the language.
|
|
*/
|
|
uint32_t ts_language_field_count(const TSLanguage *self);
|
|
|
|
/**
|
|
* Get the field name string for the given numerical id.
|
|
*/
|
|
const char *ts_language_field_name_for_id(const TSLanguage *self, TSFieldId id);
|
|
|
|
/**
|
|
* Get the numerical id for the given field name string.
|
|
*/
|
|
TSFieldId ts_language_field_id_for_name(const TSLanguage *self, const char *name, uint32_t name_length);
|
|
|
|
/**
|
|
* Check whether the given node type id belongs to named nodes, anonymous nodes,
|
|
* or a hidden nodes.
|
|
*
|
|
* See also [`ts_node_is_named`]. Hidden nodes are never returned from the API.
|
|
*/
|
|
TSSymbolType ts_language_symbol_type(const TSLanguage *self, TSSymbol symbol);
|
|
|
|
/**
|
|
* 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`].
|
|
*/
|
|
uint32_t ts_language_version(const TSLanguage *self);
|
|
|
|
/**
|
|
* Get the next parse state. Combine this with lookahead iterators to generate
|
|
* completion suggestions or valid symbols in error nodes. Use
|
|
* [`ts_node_grammar_symbol`] for valid symbols.
|
|
*/
|
|
TSStateId ts_language_next_state(const TSLanguage *self, TSStateId state, TSSymbol symbol);
|
|
|
|
/********************************/
|
|
/* Section - Lookahead Iterator */
|
|
/********************************/
|
|
|
|
/**
|
|
* Create a new lookahead iterator for the given language and parse state.
|
|
*
|
|
* This returns `NULL` if state is invalid for the language.
|
|
*
|
|
* Repeatedly using [`ts_lookahead_iterator_next`] and
|
|
* [`ts_lookahead_iterator_current_symbol`] will generate valid symbols in the
|
|
* given parse state. Newly created lookahead iterators will contain the `ERROR`
|
|
* symbol.
|
|
*
|
|
* 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 `MISSING` nodes, a lookahead
|
|
* iterator created on the previous non-extra leaf node may be appropriate.
|
|
*/
|
|
TSLookaheadIterator *ts_lookahead_iterator_new(const TSLanguage *self, TSStateId state);
|
|
|
|
/**
|
|
* Delete a lookahead iterator freeing all the memory used.
|
|
*/
|
|
void ts_lookahead_iterator_delete(TSLookaheadIterator *self);
|
|
|
|
/**
|
|
* Reset the lookahead iterator to another state.
|
|
*
|
|
* This returns `true` if the iterator was reset to the given state and `false`
|
|
* otherwise.
|
|
*/
|
|
bool ts_lookahead_iterator_reset_state(TSLookaheadIterator *self, TSStateId state);
|
|
|
|
/**
|
|
* Reset the lookahead iterator.
|
|
*
|
|
* This returns `true` if the language was set successfully and `false`
|
|
* otherwise.
|
|
*/
|
|
bool ts_lookahead_iterator_reset(TSLookaheadIterator *self, const TSLanguage *language, TSStateId state);
|
|
|
|
/**
|
|
* Get the current language of the lookahead iterator.
|
|
*/
|
|
const TSLanguage *ts_lookahead_iterator_language(const TSLookaheadIterator *self);
|
|
|
|
/**
|
|
* Advance the lookahead iterator to the next symbol.
|
|
*
|
|
* This returns `true` if there is a new symbol and `false` otherwise.
|
|
*/
|
|
bool ts_lookahead_iterator_next(TSLookaheadIterator *self);
|
|
|
|
/**
|
|
* Get the current symbol of the lookahead iterator;
|
|
*/
|
|
TSSymbol ts_lookahead_iterator_current_symbol(const TSLookaheadIterator *self);
|
|
|
|
/**
|
|
* Get the current symbol type of the lookahead iterator as a null terminated
|
|
* string.
|
|
*/
|
|
const char *ts_lookahead_iterator_current_symbol_name(const TSLookaheadIterator *self);
|
|
|
|
/*************************************/
|
|
/* Section - WebAssembly Integration */
|
|
/************************************/
|
|
|
|
typedef struct wasm_engine_t TSWasmEngine;
|
|
typedef struct TSWasmStore TSWasmStore;
|
|
|
|
typedef enum {
|
|
TSWasmErrorKindNone = 0,
|
|
TSWasmErrorKindParse,
|
|
TSWasmErrorKindCompile,
|
|
TSWasmErrorKindInstantiate,
|
|
TSWasmErrorKindAllocate,
|
|
} TSWasmErrorKind;
|
|
|
|
typedef struct {
|
|
TSWasmErrorKind kind;
|
|
char *message;
|
|
} TSWasmError;
|
|
|
|
/**
|
|
* Create a Wasm store.
|
|
*/
|
|
TSWasmStore *ts_wasm_store_new(
|
|
TSWasmEngine *engine,
|
|
TSWasmError *error
|
|
);
|
|
|
|
/**
|
|
* Free the memory associated with the given Wasm store.
|
|
*/
|
|
void ts_wasm_store_delete(TSWasmStore *);
|
|
|
|
/**
|
|
* Create a language from a buffer of Wasm. The resulting language behaves
|
|
* like any other Tree-sitter language, except that in order to use it with
|
|
* a parser, that parser must have a Wasm store. Note that the language
|
|
* can be used with any Wasm store, it doesn't need to be the same store that
|
|
* was used to originally load it.
|
|
*/
|
|
const TSLanguage *ts_wasm_store_load_language(
|
|
TSWasmStore *,
|
|
const char *name,
|
|
const char *wasm,
|
|
uint32_t wasm_len,
|
|
TSWasmError *error
|
|
);
|
|
|
|
/**
|
|
* Get the number of languages instantiated in the given wasm store.
|
|
*/
|
|
size_t ts_wasm_store_language_count(const TSWasmStore *);
|
|
|
|
/**
|
|
* Check if the language came from a Wasm module. If so, then in order to use
|
|
* this language with a Parser, that parser must have a Wasm store assigned.
|
|
*/
|
|
bool ts_language_is_wasm(const TSLanguage *);
|
|
|
|
/**
|
|
* Assign the given Wasm store to the parser. A parser must have a Wasm store
|
|
* in order to use Wasm languages.
|
|
*/
|
|
void ts_parser_set_wasm_store(TSParser *, TSWasmStore *);
|
|
|
|
/**
|
|
* Remove the parser's current Wasm store and return it. This returns NULL if
|
|
* the parser doesn't have a Wasm store.
|
|
*/
|
|
TSWasmStore *ts_parser_take_wasm_store(TSParser *);
|
|
|
|
/**********************************/
|
|
/* Section - Global Configuration */
|
|
/**********************************/
|
|
|
|
/**
|
|
* Set the allocation functions used by the library.
|
|
*
|
|
* By default, Tree-sitter uses the standard libc allocation functions,
|
|
* but aborts the process when an allocation fails. This function lets
|
|
* you supply alternative allocation functions at runtime.
|
|
*
|
|
* If you pass `NULL` for any parameter, Tree-sitter will switch back to
|
|
* its default implementation of that function.
|
|
*
|
|
* If you call this function after the library has already been used, then
|
|
* you must ensure that either:
|
|
* 1. All the existing objects have been freed.
|
|
* 2. The new allocator shares its state with the old one, so it is capable
|
|
* of freeing memory that was allocated by the old allocator.
|
|
*/
|
|
void ts_set_allocator(
|
|
void *(*new_malloc)(size_t),
|
|
void *(*new_calloc)(size_t, size_t),
|
|
void *(*new_realloc)(void *, size_t),
|
|
void (*new_free)(void *)
|
|
);
|
|
|
|
#ifdef __cplusplus
|
|
}
|
|
#endif
|
|
|
|
#ifndef TREE_SITTER_HIDE_SYMBOLS
|
|
#if defined(__GNUC__) || defined(__clang__)
|
|
#pragma GCC visibility pop
|
|
#endif
|
|
#endif
|
|
|
|
#endif // TREE_SITTER_API_H_
|