extension.c

#include "prism/extension.h"
 
#ifdef _WIN32
#include <ruby/win32.h>
#endif
 
#include <errno.h>
 
// NOTE: this file should contain only bindings. All non-trivial logic should be
// in libprism so it can be shared its the various callers.
 
VALUE rb_cPrism;
VALUE rb_cPrismNode;
VALUE rb_cPrismSource;
VALUE rb_cPrismToken;
VALUE rb_cPrismLocation;
 
VALUE rb_cPrismComment;
VALUE rb_cPrismInlineComment;
VALUE rb_cPrismEmbDocComment;
VALUE rb_cPrismMagicComment;
VALUE rb_cPrismParseError;
VALUE rb_cPrismParseWarning;
VALUE rb_cPrismResult;
VALUE rb_cPrismParseResult;
VALUE rb_cPrismLexResult;
VALUE rb_cPrismParseLexResult;
VALUE rb_cPrismStringQuery;
VALUE rb_cPrismScope;
VALUE rb_cPrismCurrentVersionError;
 
VALUE rb_cPrismDebugEncoding;
 
ID rb_id_option_command_line;
ID rb_id_option_encoding;
ID rb_id_option_filepath;
ID rb_id_option_freeze;
ID rb_id_option_frozen_string_literal;
ID rb_id_option_line;
ID rb_id_option_main_script;
ID rb_id_option_partial_script;
ID rb_id_option_scopes;
ID rb_id_option_version;
ID rb_id_source_for;
ID rb_id_forwarding_positionals;
ID rb_id_forwarding_keywords;
ID rb_id_forwarding_block;
ID rb_id_forwarding_all;
 
/******************************************************************************/
/* IO of Ruby code                                                            */
/******************************************************************************/
 
static const char *
check_string(VALUE value) {
    // Check if the value is a string. If it's not, then raise a type error.
    if (!RB_TYPE_P(value, T_STRING)) {
        rb_raise(rb_eTypeError, "wrong argument type %" PRIsVALUE " (expected String)", rb_obj_class(value));
    }
 
    // Otherwise, return the value as a C string.
    return RSTRING_PTR(value);
}
 
 
/******************************************************************************/
/* Building C options from Ruby options                                       */
/******************************************************************************/
 
static void
build_options_scopes(pm_options_t *options, VALUE scopes) {
    // Check if the value is an array. If it's not, then raise a type error.
    if (!RB_TYPE_P(scopes, T_ARRAY)) {
        rb_raise(rb_eTypeError, "wrong argument type %"PRIsVALUE" (expected Array)", rb_obj_class(scopes));
    }
 
    // Initialize the scopes array.
    size_t scopes_count = RARRAY_LEN(scopes);
    if (!pm_options_scopes_init(options, scopes_count)) {
        rb_raise(rb_eNoMemError, "failed to allocate memory");
    }
 
    // Iterate over the scopes and add them to the options.
    for (size_t scope_index = 0; scope_index < scopes_count; scope_index++) {
        VALUE scope = rb_ary_entry(scopes, scope_index);
 
        // The scope can be either an array or it can be a Prism::Scope object.
        // Parse out the correct values here from either.
        VALUE locals;
        uint8_t forwarding = PM_OPTIONS_SCOPE_FORWARDING_NONE;
 
        if (RB_TYPE_P(scope, T_ARRAY)) {
            locals = scope;
        } else if (rb_obj_is_kind_of(scope, rb_cPrismScope)) {
            locals = rb_ivar_get(scope, rb_intern("@locals"));
            if (!RB_TYPE_P(locals, T_ARRAY)) {
                rb_raise(rb_eTypeError, "wrong argument type %"PRIsVALUE" (expected Array)", rb_obj_class(locals));
            }
 
            VALUE names = rb_ivar_get(scope, rb_intern("@forwarding"));
            if (!RB_TYPE_P(names, T_ARRAY)) {
                rb_raise(rb_eTypeError, "wrong argument type %"PRIsVALUE" (expected Array)", rb_obj_class(names));
            }
 
            size_t names_count = RARRAY_LEN(names);
            for (size_t name_index = 0; name_index < names_count; name_index++) {
                VALUE name = rb_ary_entry(names, name_index);
 
                // Check that the name is a symbol. If it's not, then raise
                // a type error.
                if (!RB_TYPE_P(name, T_SYMBOL)) {
                    rb_raise(rb_eTypeError, "wrong argument type %"PRIsVALUE" (expected Symbol)", rb_obj_class(name));
                }
 
                ID id = SYM2ID(name);
                if (id == rb_id_forwarding_positionals) {
                    forwarding |= PM_OPTIONS_SCOPE_FORWARDING_POSITIONALS;
                } else if (id == rb_id_forwarding_keywords) {
                    forwarding |= PM_OPTIONS_SCOPE_FORWARDING_KEYWORDS;
                } else if (id == rb_id_forwarding_block) {
                    forwarding |= PM_OPTIONS_SCOPE_FORWARDING_BLOCK;
                } else if (id == rb_id_forwarding_all) {
                    forwarding |= PM_OPTIONS_SCOPE_FORWARDING_ALL;
                } else {
                    rb_raise(rb_eArgError, "invalid forwarding value: %" PRIsVALUE, name);
                }
            }
        } else {
            rb_raise(rb_eTypeError, "wrong argument type %"PRIsVALUE" (expected Array or Prism::Scope)", rb_obj_class(scope));
        }
 
        // Initialize the scope array.
        size_t locals_count = RARRAY_LEN(locals);
        pm_options_scope_t *options_scope = pm_options_scope_mut(options, scope_index);
        pm_options_scope_init(options_scope, locals_count);
 
        // Iterate over the locals and add them to the scope.
        for (size_t local_index = 0; local_index < locals_count; local_index++) {
            VALUE local = rb_ary_entry(locals, local_index);
 
            // Check that the local is a symbol. If it's not, then raise a
            // type error.
            if (!RB_TYPE_P(local, T_SYMBOL)) {
                rb_raise(rb_eTypeError, "wrong argument type %"PRIsVALUE" (expected Symbol)", rb_obj_class(local));
            }
 
            // Add the local to the scope.
            pm_string_t *scope_local = pm_options_scope_local_mut(options_scope, local_index);
            const char *name = rb_id2name(SYM2ID(local));
            pm_string_constant_init(scope_local, name, strlen(name));
        }
 
        // Now set the forwarding options.
        pm_options_scope_forwarding_set(options_scope, forwarding);
    }
}
 
static int
build_options_i(VALUE key, VALUE value, VALUE argument) {
    pm_options_t *options = (pm_options_t *) argument;
    ID key_id = SYM2ID(key);
 
    if (key_id == rb_id_option_filepath) {
        if (!NIL_P(value)) pm_options_filepath_set(options, check_string(value));
    } else if (key_id == rb_id_option_encoding) {
        if (!NIL_P(value)) {
            if (value == Qfalse) {
                pm_options_encoding_locked_set(options, true);
            } else {
                pm_options_encoding_set(options, rb_enc_name(rb_to_encoding(value)));
            }
        }
    } else if (key_id == rb_id_option_line) {
        if (!NIL_P(value)) pm_options_line_set(options, NUM2INT(value));
    } else if (key_id == rb_id_option_frozen_string_literal) {
        if (!NIL_P(value)) pm_options_frozen_string_literal_set(options, RTEST(value));
    } else if (key_id == rb_id_option_version) {
        if (!NIL_P(value)) {
            const char *version = check_string(value);
 
            if (RSTRING_LEN(value) == 7 && strncmp(version, "current", 7) == 0) {
                if (!pm_options_version_set(options, ruby_version, 3)) {
                    rb_exc_raise(rb_exc_new_cstr(rb_cPrismCurrentVersionError, ruby_version));
                }
            } else if (RSTRING_LEN(value) == 7 && strncmp(version, "nearest", 7) == 0) {
                if (!pm_options_version_set(options, ruby_version, 3)) {
                    // Prism doesn't know this specific version. Is it lower?
                    if (ruby_version[0] < '3' || (ruby_version[0] == '3' && ruby_version[2] < '3')) {
                        pm_options_version_set_lowest(options);
                    } else {
                        // Must be higher.
                        pm_options_version_set_highest(options);
                    }
                }
            } else if (!pm_options_version_set(options, version, RSTRING_LEN(value))) {
                rb_raise(rb_eArgError, "invalid version: %" PRIsVALUE, value);
            }
        }
    } else if (key_id == rb_id_option_scopes) {
        if (!NIL_P(value)) build_options_scopes(options, value);
    } else if (key_id == rb_id_option_command_line) {
        if (!NIL_P(value)) {
            const char *string = check_string(value);
            uint8_t command_line = 0;
 
            for (size_t index = 0; index < strlen(string); index++) {
                switch (string[index]) {
                    case 'a': command_line |= PM_OPTIONS_COMMAND_LINE_A; break;
                    case 'e': command_line |= PM_OPTIONS_COMMAND_LINE_E; break;
                    case 'l': command_line |= PM_OPTIONS_COMMAND_LINE_L; break;
                    case 'n': command_line |= PM_OPTIONS_COMMAND_LINE_N; break;
                    case 'p': command_line |= PM_OPTIONS_COMMAND_LINE_P; break;
                    case 'x': command_line |= PM_OPTIONS_COMMAND_LINE_X; break;
                    default: rb_raise(rb_eArgError, "invalid command line flag: '%c'", string[index]); break;
                }
            }
 
            pm_options_command_line_set(options, command_line);
        }
    } else if (key_id == rb_id_option_main_script) {
        if (!NIL_P(value)) pm_options_main_script_set(options, RTEST(value));
    } else if (key_id == rb_id_option_partial_script) {
        if (!NIL_P(value)) pm_options_partial_script_set(options, RTEST(value));
    } else if (key_id == rb_id_option_freeze) {
        if (!NIL_P(value)) pm_options_freeze_set(options, RTEST(value));
    } else {
        rb_raise(rb_eArgError, "unknown keyword: %" PRIsVALUE, key);
    }
 
    return ST_CONTINUE;
}
 
struct build_options_data {
    pm_options_t *options;
    VALUE keywords;
};
 
static VALUE
build_options(VALUE argument) {
    struct build_options_data *data = (struct build_options_data *) argument;
    rb_hash_foreach(data->keywords, build_options_i, (VALUE) data->options);
    return Qnil;
}
 
static void
extract_options(pm_options_t *options, VALUE filepath, VALUE keywords) {
    pm_options_line_set(options, 1); /* default */
 
    if (!NIL_P(keywords)) {
        struct build_options_data data = { .options = options, .keywords = keywords };
        struct build_options_data *argument = &data;
 
        int state = 0;
        rb_protect(build_options, (VALUE) argument, &state);
 
        if (state != 0) {
            pm_options_free(options);
            rb_jump_tag(state);
        }
    }
 
    if (!NIL_P(filepath)) {
        if (!RB_TYPE_P(filepath, T_STRING)) {
            pm_options_free(options);
            rb_raise(rb_eTypeError, "wrong argument type %"PRIsVALUE" (expected String)", rb_obj_class(filepath));
        }
 
        pm_options_filepath_set(options, RSTRING_PTR(filepath));
    }
}
 
static VALUE
string_options(int argc, VALUE *argv, pm_options_t *options) {
    VALUE string;
    VALUE keywords;
    rb_scan_args(argc, argv, "1:", &string, &keywords);
 
    if (!RB_TYPE_P(string, T_STRING)) {
        pm_options_free(options);
        rb_raise(rb_eTypeError, "wrong argument type %"PRIsVALUE" (expected String)", rb_obj_class(string));
    }
 
    extract_options(options, Qnil, keywords);
    return string;
}
 
static pm_source_t *
file_options(int argc, VALUE *argv, pm_options_t *options, VALUE *encoded_filepath) {
    VALUE filepath;
    VALUE keywords;
    rb_scan_args(argc, argv, "1:", &filepath, &keywords);
 
    if (!RB_TYPE_P(filepath, T_STRING)) {
        pm_options_free(options);
        rb_raise(rb_eTypeError, "wrong argument type %"PRIsVALUE" (expected String)", rb_obj_class(filepath));
    }
 
    *encoded_filepath = rb_str_encode_ospath(filepath);
    extract_options(options, *encoded_filepath, keywords);
 
    const char *source = (const char *) pm_string_source(pm_options_filepath(options));
    pm_source_init_result_t result;
    pm_source_t *pm_src = pm_source_file_new(source, &result);
 
    switch (result) {
        case PM_SOURCE_INIT_SUCCESS:
            break;
        case PM_SOURCE_INIT_ERROR_GENERIC: {
            pm_options_free(options);
 
#ifdef _WIN32
            int e = rb_w32_map_errno(GetLastError());
#else
            int e = errno;
#endif
 
            rb_syserr_fail(e, source);
            break;
        }
        case PM_SOURCE_INIT_ERROR_DIRECTORY:
            pm_options_free(options);
            rb_syserr_fail(EISDIR, source);
            break;
        default:
            pm_options_free(options);
            rb_raise(rb_eRuntimeError, "Unknown error (%d) initializing file: %s", result, source);
            break;
    }
 
    return pm_src;
}
 
#ifndef PRISM_EXCLUDE_SERIALIZATION
 
/******************************************************************************/
/* Serializing the AST                                                        */
/******************************************************************************/
 
static VALUE
dump_input(const uint8_t *input, size_t input_length, const pm_options_t *options) {
    pm_buffer_t *buffer = pm_buffer_new();
    if (!buffer) {
        rb_raise(rb_eNoMemError, "failed to allocate memory");
    }
 
    pm_arena_t *arena = pm_arena_new();
    pm_parser_t *parser = pm_parser_new(arena, input, input_length, options);
 
    pm_node_t *node = pm_parse(parser);
    pm_serialize(parser, node, buffer);
 
    VALUE result = rb_str_new(pm_buffer_value(buffer), pm_buffer_length(buffer));
    pm_buffer_free(buffer);
    pm_parser_free(parser);
    pm_arena_free(arena);
 
    return result;
}
 
static VALUE
dump(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
    VALUE string = string_options(argc, argv, options);
 
    const uint8_t *source = (const uint8_t *) RSTRING_PTR(string);
    size_t length = RSTRING_LEN(string);
 
#ifdef PRISM_BUILD_DEBUG
    char* dup = xmalloc(length);
    memcpy(dup, source, length);
    source = (const uint8_t *) dup;
#endif
 
    VALUE value = dump_input(source, length, options);
    if (pm_options_freeze(options)) rb_obj_freeze(value);
 
#ifdef PRISM_BUILD_DEBUG
#ifdef xfree_sized
    xfree_sized(dup, length);
#else
    xfree(dup);
#endif
#endif
 
    pm_options_free(options);
 
    return value;
}
 
static VALUE
dump_file(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
 
    VALUE encoded_filepath;
    pm_source_t *src = file_options(argc, argv, options, &encoded_filepath);
 
    VALUE value = dump_input(pm_source_source(src), pm_source_length(src), options);
    pm_source_free(src);
    pm_options_free(options);
 
    return value;
}
 
#endif
 
/******************************************************************************/
/* Extracting values for the parse result                                     */
/******************************************************************************/
 
static inline VALUE
rb_class_new_instance_freeze(int argc, const VALUE *argv, VALUE klass, bool freeze) {
    VALUE value = rb_class_new_instance(argc, argv, klass);
    if (freeze) rb_obj_freeze(value);
    return value;
}
 
static inline VALUE
parser_location(VALUE source, bool freeze, uint32_t start, uint32_t length) {
    VALUE argv[] = { source, LONG2FIX(start), LONG2FIX(length) };
    return rb_class_new_instance_freeze(3, argv, rb_cPrismLocation, freeze);
}
 
#define PARSER_LOCATION(source, freeze, location) \
    parser_location(source, freeze, location.start, location.length)
 
static inline VALUE
parser_comment(VALUE source, bool freeze, const pm_comment_t *comment) {
    VALUE argv[] = { PARSER_LOCATION(source, freeze, pm_comment_location(comment)) };
    VALUE type = (pm_comment_type(comment) == PM_COMMENT_EMBDOC) ? rb_cPrismEmbDocComment : rb_cPrismInlineComment;
    return rb_class_new_instance_freeze(1, argv, type, freeze);
}
 
typedef struct {
    VALUE comments;
    VALUE source;
    bool freeze;
} parser_comments_each_data_t;
 
static void
parser_comments_each(const pm_comment_t *comment, void *data) {
    parser_comments_each_data_t *each_data = (parser_comments_each_data_t *) data;
    VALUE value = parser_comment(each_data->source, each_data->freeze, comment);
    rb_ary_push(each_data->comments, value);
}
 
static VALUE
parser_comments(const pm_parser_t *parser, VALUE source, bool freeze) {
    VALUE comments = rb_ary_new_capa(pm_parser_comments_size(parser));
 
    parser_comments_each_data_t each_data = { comments, source, freeze };
    pm_parser_comments_each(parser, parser_comments_each, &each_data);
 
    if (freeze) rb_obj_freeze(comments);
    return comments;
}
 
static inline VALUE
parser_magic_comment(VALUE source, bool freeze, const pm_magic_comment_t *magic_comment) {
    pm_location_t key = pm_magic_comment_key(magic_comment);
    pm_location_t value = pm_magic_comment_value(magic_comment);
 
    VALUE key_loc = parser_location(source, freeze, key.start, key.length);
    VALUE value_loc = parser_location(source, freeze, value.start, value.length);
 
    VALUE argv[] = { key_loc, value_loc };
    return rb_class_new_instance_freeze(2, argv, rb_cPrismMagicComment, freeze);
}
 
typedef struct {
    VALUE magic_comments;
    VALUE source;
    bool freeze;
} parser_magic_comments_each_data_t;
 
static void
parser_magic_comments_each(const pm_magic_comment_t *magic_comment, void *data) {
    parser_magic_comments_each_data_t *each_data = (parser_magic_comments_each_data_t *) data;
    VALUE value = parser_magic_comment(each_data->source, each_data->freeze, magic_comment);
    rb_ary_push(each_data->magic_comments, value);
}
 
static VALUE
parser_magic_comments(const pm_parser_t *parser, VALUE source, bool freeze) {
    VALUE magic_comments = rb_ary_new_capa(pm_parser_magic_comments_size(parser));
 
    parser_magic_comments_each_data_t each_data = { magic_comments, source, freeze };
    pm_parser_magic_comments_each(parser, parser_magic_comments_each, &each_data);
 
    if (freeze) rb_obj_freeze(magic_comments);
    return magic_comments;
}
 
static VALUE
parser_data_loc(const pm_parser_t *parser, VALUE source, bool freeze) {
    const pm_location_t *data_loc = pm_parser_data_loc(parser);
 
    if (data_loc->length == 0) {
        return Qnil;
    } else {
        return parser_location(source, freeze, data_loc->start, data_loc->length);
    }
}
 
typedef struct {
    VALUE errors;
    rb_encoding *encoding;
    VALUE source;
    bool freeze;
} parser_errors_each_data_t;
 
static void
parser_errors_each(const pm_diagnostic_t *diagnostic, void *data) {
    parser_errors_each_data_t *each_data = (parser_errors_each_data_t *) data;
 
    VALUE type = ID2SYM(rb_intern(pm_diagnostic_type(diagnostic)));
    VALUE message = rb_obj_freeze(rb_enc_str_new_cstr(pm_diagnostic_message(diagnostic), each_data->encoding));
    VALUE location = PARSER_LOCATION(each_data->source, each_data->freeze, pm_diagnostic_location(diagnostic));
 
    pm_error_level_t error_level = pm_diagnostic_error_level(diagnostic);
    VALUE level = Qnil;
 
    switch (error_level) {
        case PM_ERROR_LEVEL_SYNTAX:
            level = ID2SYM(rb_intern("syntax"));
            break;
        case PM_ERROR_LEVEL_ARGUMENT:
            level = ID2SYM(rb_intern("argument"));
            break;
        case PM_ERROR_LEVEL_LOAD:
            level = ID2SYM(rb_intern("load"));
            break;
        default:
            rb_raise(rb_eRuntimeError, "Unknown level: %" PRIu8, error_level);
    }
 
    VALUE argv[] = { type, message, location, level };
    VALUE value = rb_class_new_instance_freeze(4, argv, rb_cPrismParseError, each_data->freeze);
    rb_ary_push(each_data->errors, value);
}
 
static VALUE
parser_errors(const pm_parser_t *parser, rb_encoding *encoding, VALUE source, bool freeze) {
    VALUE errors = rb_ary_new_capa(pm_parser_errors_size(parser));
 
    parser_errors_each_data_t each_data = { errors, encoding, source, freeze };
    pm_parser_errors_each(parser, parser_errors_each, &each_data);
 
    if (freeze) rb_obj_freeze(errors);
    return errors;
}
 
typedef struct {
    VALUE warnings;
    rb_encoding *encoding;
    VALUE source;
    bool freeze;
} parser_warnings_each_data_t;
 
static void
parser_warnings_each(const pm_diagnostic_t *diagnostic, void *data) {
    parser_warnings_each_data_t *each_data = (parser_warnings_each_data_t *) data;
 
    VALUE type = ID2SYM(rb_intern(pm_diagnostic_type(diagnostic)));
    VALUE message = rb_obj_freeze(rb_enc_str_new_cstr(pm_diagnostic_message(diagnostic), each_data->encoding));
    VALUE location = PARSER_LOCATION(each_data->source, each_data->freeze, pm_diagnostic_location(diagnostic));
 
    pm_warning_level_t warning_level = pm_diagnostic_warning_level(diagnostic);
    VALUE level = Qnil;
 
    switch (warning_level) {
        case PM_WARNING_LEVEL_DEFAULT:
            level = ID2SYM(rb_intern("default"));
            break;
        case PM_WARNING_LEVEL_VERBOSE:
            level = ID2SYM(rb_intern("verbose"));
            break;
        default:
            rb_raise(rb_eRuntimeError, "Unknown level: %" PRIu8, warning_level);
    }
 
    VALUE argv[] = { type, message, location, level };
    VALUE value = rb_class_new_instance_freeze(4, argv, rb_cPrismParseWarning, each_data->freeze);
    rb_ary_push(each_data->warnings, value);
}
 
static VALUE
parser_warnings(const pm_parser_t *parser, rb_encoding *encoding, VALUE source, bool freeze) {
    VALUE warnings = rb_ary_new_capa(pm_parser_warnings_size(parser));
 
    parser_warnings_each_data_t each_data = { warnings, encoding, source, freeze };
    pm_parser_warnings_each(parser, parser_warnings_each, &each_data);
 
    if (freeze) rb_obj_freeze(warnings);
    return warnings;
}
 
static VALUE
parse_result_create(VALUE class, const pm_parser_t *parser, VALUE value, rb_encoding *encoding, VALUE source, bool freeze) {
    VALUE result_argv[] = {
        value,
        parser_comments(parser, source, freeze),
        parser_magic_comments(parser, source, freeze),
        parser_data_loc(parser, source, freeze),
        parser_errors(parser, encoding, source, freeze),
        parser_warnings(parser, encoding, source, freeze),
        pm_parser_continuable(parser) ? Qtrue : Qfalse,
        source
    };
 
    return rb_class_new_instance_freeze(8, result_argv, class, freeze);
}
 
/******************************************************************************/
/* Lexing Ruby code                                                           */
/******************************************************************************/
 
typedef struct {
    VALUE source;
    VALUE tokens;
    rb_encoding *encoding;
    bool freeze;
} parse_lex_data_t;
 
static void
parse_lex_token(pm_parser_t *parser, pm_token_t *token, void *data) {
    parse_lex_data_t *parse_lex_data = (parse_lex_data_t *) data;
 
    VALUE value = pm_token_new(parser, token, parse_lex_data->encoding, parse_lex_data->source, parse_lex_data->freeze);
    VALUE yields = rb_assoc_new(value, INT2FIX(pm_parser_lex_state(parser)));
 
    if (parse_lex_data->freeze) {
        rb_obj_freeze(value);
        rb_obj_freeze(yields);
    }
 
    rb_ary_push(parse_lex_data->tokens, yields);
}
 
static void
parse_lex_encoding_changed_callback(pm_parser_t *parser) {
    parse_lex_data_t *parse_lex_data = (parse_lex_data_t *) pm_parser_lex_callback_data(parser);
    parse_lex_data->encoding = rb_enc_find(pm_parser_encoding_name(parser));
 
    // Since the encoding changed, we need to go back and change the encoding of
    // the tokens that were already lexed. This is only going to end up being
    // one or two tokens, since the encoding can only change at the top of the
    // file.
    VALUE tokens = parse_lex_data->tokens;
    VALUE next_tokens = rb_ary_new();
 
    for (long index = 0; index < RARRAY_LEN(tokens); index++) {
        VALUE yields = rb_ary_entry(tokens, index);
        VALUE token = rb_ary_entry(yields, 0);
 
        VALUE value = rb_ivar_get(token, rb_intern("@value"));
        VALUE next_value = rb_str_dup(value);
 
        rb_enc_associate(next_value, parse_lex_data->encoding);
        if (parse_lex_data->freeze) rb_obj_freeze(next_value);
 
        VALUE next_token_argv[] = {
            parse_lex_data->source,
            rb_ivar_get(token, rb_intern("@type")),
            next_value,
            rb_ivar_get(token, rb_intern("@location"))
        };
 
        VALUE next_token = rb_class_new_instance(4, next_token_argv, rb_cPrismToken);
        VALUE next_yields = rb_assoc_new(next_token, rb_ary_entry(yields, 1));
 
        if (parse_lex_data->freeze) {
            rb_obj_freeze(next_token);
            rb_obj_freeze(next_yields);
        }
 
        rb_ary_push(next_tokens, next_yields);
    }
 
    rb_ary_replace(parse_lex_data->tokens, next_tokens);
}
 
static VALUE
parse_lex_input(const uint8_t *input, size_t input_length, const pm_options_t *options, bool return_nodes) {
    pm_arena_t *arena = pm_arena_new();
    pm_parser_t *parser = pm_parser_new(arena, input, input_length, options);
    pm_parser_encoding_changed_callback_set(parser, parse_lex_encoding_changed_callback);
 
    VALUE source_string = rb_str_new((const char *) input, input_length);
    VALUE offsets = rb_ary_new_capa(pm_parser_line_offsets(parser)->size);
    VALUE source = rb_funcall(rb_cPrismSource, rb_id_source_for, 3, source_string, LONG2NUM(pm_parser_start_line(parser)), offsets);
 
    parse_lex_data_t parse_lex_data = {
        .source = source,
        .tokens = rb_ary_new(),
        .encoding = rb_utf8_encoding(),
        .freeze = pm_options_freeze(options),
    };
 
    parse_lex_data_t *data = &parse_lex_data;
    pm_parser_lex_callback_set(parser, parse_lex_token, data);
 
    pm_node_t *node = pm_parse(parser);
 
    /* Update the Source object with the correct encoding and line offsets,
     * which are only available after pm_parse() completes. */
    rb_encoding *encoding = rb_enc_find(pm_parser_encoding_name(parser));
    rb_enc_associate(source_string, encoding);
 
    const pm_line_offset_list_t *line_offsets = pm_parser_line_offsets(parser);
    for (size_t index = 0; index < line_offsets->size; index++) {
        rb_ary_store(offsets, (long) index, ULONG2NUM(line_offsets->offsets[index]));
    }
 
    if (pm_options_freeze(options)) {
        rb_obj_freeze(source_string);
        rb_obj_freeze(offsets);
        rb_obj_freeze(source);
        rb_obj_freeze(parse_lex_data.tokens);
    }
 
    VALUE result;
    if (return_nodes) {
        VALUE value = rb_ary_new_capa(2);
        rb_ary_push(value, pm_ast_new(parser, node, parse_lex_data.encoding, source, pm_options_freeze(options)));
        rb_ary_push(value, parse_lex_data.tokens);
        if (pm_options_freeze(options)) rb_obj_freeze(value);
        result = parse_result_create(rb_cPrismParseLexResult, parser, value, parse_lex_data.encoding, source, pm_options_freeze(options));
    } else {
        result = parse_result_create(rb_cPrismLexResult, parser, parse_lex_data.tokens, parse_lex_data.encoding, source, pm_options_freeze(options));
    }
 
    pm_parser_free(parser);
    pm_arena_free(arena);
 
    return result;
}
 
static VALUE
lex(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
    VALUE string = string_options(argc, argv, options);
 
    VALUE result = parse_lex_input((const uint8_t *) RSTRING_PTR(string), RSTRING_LEN(string), options, false);
    pm_options_free(options);
 
    return result;
}
 
static VALUE
lex_file(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
 
    VALUE encoded_filepath;
    pm_source_t *src = file_options(argc, argv, options, &encoded_filepath);
 
    VALUE value = parse_lex_input(pm_source_source(src), pm_source_length(src), options, false);
    pm_source_free(src);
    pm_options_free(options);
 
    return value;
}
 
/******************************************************************************/
/* Parsing Ruby code                                                          */
/******************************************************************************/
 
static VALUE
parse_input(const uint8_t *input, size_t input_length, const pm_options_t *options) {
    pm_arena_t *arena = pm_arena_new();
    pm_parser_t *parser = pm_parser_new(arena, input, input_length, options);
 
    pm_node_t *node = pm_parse(parser);
    rb_encoding *encoding = rb_enc_find(pm_parser_encoding_name(parser));
 
    bool freeze = pm_options_freeze(options);
    VALUE source = pm_source_new(parser, encoding, freeze);
    VALUE value = pm_ast_new(parser, node, encoding, source, freeze);
    VALUE result = parse_result_create(rb_cPrismParseResult, parser, value, encoding, source, freeze);
 
    if (freeze) {
        rb_obj_freeze(source);
    }
 
    pm_parser_free(parser);
    pm_arena_free(arena);
 
    return result;
}
 
static VALUE
parse(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
    VALUE string = string_options(argc, argv, options);
 
    const uint8_t *source = (const uint8_t *) RSTRING_PTR(string);
    size_t length = RSTRING_LEN(string);
 
#ifdef PRISM_BUILD_DEBUG
    char* dup = xmalloc(length);
    memcpy(dup, source, length);
    source = (const uint8_t *) dup;
#endif
 
    VALUE value = parse_input(source, length, options);
 
#ifdef PRISM_BUILD_DEBUG
#ifdef xfree_sized
    xfree_sized(dup, length);
#else
    xfree(dup);
#endif
#endif
 
    pm_options_free(options);
    return value;
}
 
static VALUE
parse_file(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
 
    VALUE encoded_filepath;
    pm_source_t *src = file_options(argc, argv, options, &encoded_filepath);
 
    VALUE value = parse_input(pm_source_source(src), pm_source_length(src), options);
    pm_source_free(src);
    pm_options_free(options);
 
    return value;
}
 
static void
profile_input(const uint8_t *input, size_t input_length, const pm_options_t *options) {
    pm_arena_t *arena = pm_arena_new();
    pm_parser_t *parser = pm_parser_new(arena, input, input_length, options);
 
    pm_parse(parser);
    pm_parser_free(parser);
    pm_arena_free(arena);
}
 
static VALUE
profile(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
    VALUE string = string_options(argc, argv, options);
 
    profile_input((const uint8_t *) RSTRING_PTR(string), RSTRING_LEN(string), options);
    pm_options_free(options);
 
    return Qnil;
}
 
static VALUE
profile_file(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
 
    VALUE encoded_filepath;
    pm_source_t *src = file_options(argc, argv, options, &encoded_filepath);
 
    profile_input(pm_source_source(src), pm_source_length(src), options);
    pm_source_free(src);
    pm_options_free(options);
 
    return Qnil;
}
 
static int
parse_stream_eof(void *stream) {
    if (rb_funcall((VALUE) stream, rb_intern("eof?"), 0)) {
        return 1;
    }
    return 0;
}
 
static char *
parse_stream_fgets(char *string, int size, void *stream) {
    RUBY_ASSERT(size > 0);
 
    VALUE line = rb_funcall((VALUE) stream, rb_intern("gets"), 1, INT2FIX(size - 1));
    if (NIL_P(line)) {
        return NULL;
    }
 
    const char *cstr = RSTRING_PTR(line);
    long length = RSTRING_LEN(line);
 
    memcpy(string, cstr, length);
    string[length] = '\0';
 
    return string;
}
 
static VALUE
parse_stream(int argc, VALUE *argv, VALUE self) {
    VALUE stream;
    VALUE keywords;
    rb_scan_args(argc, argv, "1:", &stream, &keywords);
 
    pm_options_t *options = pm_options_new();
    extract_options(options, Qnil, keywords);
 
    pm_source_t *src = pm_source_stream_new((void *) stream, parse_stream_fgets, parse_stream_eof);
    pm_arena_t *arena = pm_arena_new();
    pm_parser_t *parser;
 
    pm_node_t *node = pm_parse_stream(&parser, arena, src, options);
    rb_encoding *encoding = rb_enc_find(pm_parser_encoding_name(parser));
 
    VALUE source = pm_source_new(parser, encoding, pm_options_freeze(options));
    VALUE value = pm_ast_new(parser, node, encoding, source, pm_options_freeze(options));
    VALUE result = parse_result_create(rb_cPrismParseResult, parser, value, encoding, source, pm_options_freeze(options));
 
    pm_source_free(src);
    pm_parser_free(parser);
    pm_arena_free(arena);
    pm_options_free(options);
 
    return result;
}
 
static VALUE
parse_input_comments(const uint8_t *input, size_t input_length, const pm_options_t *options) {
    pm_arena_t *arena = pm_arena_new();
    pm_parser_t *parser = pm_parser_new(arena, input, input_length, options);
 
    pm_parse(parser);
    rb_encoding *encoding = rb_enc_find(pm_parser_encoding_name(parser));
 
    VALUE source = pm_source_new(parser, encoding, pm_options_freeze(options));
    VALUE comments = parser_comments(parser, source, pm_options_freeze(options));
 
    pm_parser_free(parser);
    pm_arena_free(arena);
 
    return comments;
}
 
static VALUE
parse_comments(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
    VALUE string = string_options(argc, argv, options);
 
    VALUE result = parse_input_comments((const uint8_t *) RSTRING_PTR(string), RSTRING_LEN(string), options);
    pm_options_free(options);
 
    return result;
}
 
static VALUE
parse_file_comments(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
 
    VALUE encoded_filepath;
    pm_source_t *src = file_options(argc, argv, options, &encoded_filepath);
 
    VALUE value = parse_input_comments(pm_source_source(src), pm_source_length(src), options);
    pm_source_free(src);
    pm_options_free(options);
 
    return value;
}
 
static VALUE
parse_lex(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
    VALUE string = string_options(argc, argv, options);
 
    VALUE value = parse_lex_input((const uint8_t *) RSTRING_PTR(string), RSTRING_LEN(string), options, true);
    pm_options_free(options);
 
    return value;
}
 
static VALUE
parse_lex_file(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
 
    VALUE encoded_filepath;
    pm_source_t *src = file_options(argc, argv, options, &encoded_filepath);
 
    VALUE value = parse_lex_input(pm_source_source(src), pm_source_length(src), options, true);
    pm_source_free(src);
    pm_options_free(options);
 
    return value;
}
 
static VALUE
parse_input_success_p(const uint8_t *input, size_t input_length, const pm_options_t *options) {
    pm_arena_t *arena = pm_arena_new();
    pm_parser_t *parser = pm_parser_new(arena, input, input_length, options);
 
    pm_parse(parser);
 
    VALUE result = pm_parser_errors_size(parser) == 0 ? Qtrue : Qfalse;
    pm_parser_free(parser);
    pm_arena_free(arena);
 
    return result;
}
 
static VALUE
parse_success_p(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
    VALUE string = string_options(argc, argv, options);
 
    VALUE result = parse_input_success_p((const uint8_t *) RSTRING_PTR(string), RSTRING_LEN(string), options);
    pm_options_free(options);
 
    return result;
}
 
static VALUE
parse_failure_p(int argc, VALUE *argv, VALUE self) {
    return RTEST(parse_success_p(argc, argv, self)) ? Qfalse : Qtrue;
}
 
static VALUE
parse_file_success_p(int argc, VALUE *argv, VALUE self) {
    pm_options_t *options = pm_options_new();
 
    VALUE encoded_filepath;
    pm_source_t *src = file_options(argc, argv, options, &encoded_filepath);
 
    VALUE result = parse_input_success_p(pm_source_source(src), pm_source_length(src), options);
    pm_source_free(src);
    pm_options_free(options);
 
    return result;
}
 
static VALUE
parse_file_failure_p(int argc, VALUE *argv, VALUE self) {
    return RTEST(parse_file_success_p(argc, argv, self)) ? Qfalse : Qtrue;
}
 
/******************************************************************************/
/* String query methods                                                       */
/******************************************************************************/
 
static VALUE
string_query(pm_string_query_t result) {
    switch (result) {
        case PM_STRING_QUERY_ERROR:
            rb_raise(rb_eArgError, "Invalid or non ascii-compatible encoding");
            return Qfalse;
        case PM_STRING_QUERY_FALSE:
            return Qfalse;
        case PM_STRING_QUERY_TRUE:
            return Qtrue;
    }
    return Qfalse;
}
 
static VALUE
string_query_local_p(VALUE self, VALUE string) {
    const uint8_t *source = (const uint8_t *) check_string(string);
    return string_query(pm_string_query_local(source, RSTRING_LEN(string), rb_enc_get(string)->name));
}
 
static VALUE
string_query_constant_p(VALUE self, VALUE string) {
    const uint8_t *source = (const uint8_t *) check_string(string);
    return string_query(pm_string_query_constant(source, RSTRING_LEN(string), rb_enc_get(string)->name));
}
 
static VALUE
string_query_method_name_p(VALUE self, VALUE string) {
    const uint8_t *source = (const uint8_t *) check_string(string);
    return string_query(pm_string_query_method_name(source, RSTRING_LEN(string), rb_enc_get(string)->name));
}
 
/******************************************************************************/
/* Initialization of the extension                                            */
/******************************************************************************/
 
RUBY_FUNC_EXPORTED void
Init_prism(void) {
    // Make sure that the prism library version matches the expected version.
    // Otherwise something was compiled incorrectly.
    if (strcmp(pm_version(), EXPECTED_PRISM_VERSION) != 0) {
        rb_raise(
            rb_eRuntimeError,
            "The prism library version (%s) does not match the expected version (%s)",
            pm_version(),
            EXPECTED_PRISM_VERSION
        );
    }
 
#ifdef HAVE_RB_EXT_RACTOR_SAFE
    // Mark this extension as Ractor-safe.
    rb_ext_ractor_safe(true);
#endif
 
    // Grab up references to all of the constants that we're going to need to
    // reference throughout this extension.
    rb_cPrism = rb_define_module("Prism");
    rb_cPrismNode = rb_define_class_under(rb_cPrism, "Node", rb_cObject);
    rb_cPrismSource = rb_define_class_under(rb_cPrism, "Source", rb_cObject);
    rb_cPrismToken = rb_define_class_under(rb_cPrism, "Token", rb_cObject);
    rb_cPrismLocation = rb_define_class_under(rb_cPrism, "Location", rb_cObject);
    rb_cPrismComment = rb_define_class_under(rb_cPrism, "Comment", rb_cObject);
    rb_cPrismInlineComment = rb_define_class_under(rb_cPrism, "InlineComment", rb_cPrismComment);
    rb_cPrismEmbDocComment = rb_define_class_under(rb_cPrism, "EmbDocComment", rb_cPrismComment);
    rb_cPrismMagicComment = rb_define_class_under(rb_cPrism, "MagicComment", rb_cObject);
    rb_cPrismParseError = rb_define_class_under(rb_cPrism, "ParseError", rb_cObject);
    rb_cPrismParseWarning = rb_define_class_under(rb_cPrism, "ParseWarning", rb_cObject);
    rb_cPrismResult = rb_define_class_under(rb_cPrism, "Result", rb_cObject);
    rb_cPrismParseResult = rb_define_class_under(rb_cPrism, "ParseResult", rb_cPrismResult);
    rb_cPrismLexResult = rb_define_class_under(rb_cPrism, "LexResult", rb_cPrismResult);
    rb_cPrismParseLexResult = rb_define_class_under(rb_cPrism, "ParseLexResult", rb_cPrismResult);
    rb_cPrismStringQuery = rb_define_class_under(rb_cPrism, "StringQuery", rb_cObject);
    rb_cPrismScope = rb_define_class_under(rb_cPrism, "Scope", rb_cObject);
 
    rb_cPrismCurrentVersionError = rb_const_get(rb_cPrism, rb_intern("CurrentVersionError"));
 
    // Intern all of the IDs eagerly that we support so that we don't have to do
    // it every time we parse.
    rb_id_option_command_line = rb_intern_const("command_line");
    rb_id_option_encoding = rb_intern_const("encoding");
    rb_id_option_filepath = rb_intern_const("filepath");
    rb_id_option_freeze = rb_intern_const("freeze");
    rb_id_option_frozen_string_literal = rb_intern_const("frozen_string_literal");
    rb_id_option_line = rb_intern_const("line");
    rb_id_option_main_script = rb_intern_const("main_script");
    rb_id_option_partial_script = rb_intern_const("partial_script");
    rb_id_option_scopes = rb_intern_const("scopes");
    rb_id_option_version = rb_intern_const("version");
    rb_id_source_for = rb_intern("for");
    rb_id_forwarding_positionals = rb_intern("*");
    rb_id_forwarding_keywords = rb_intern("**");
    rb_id_forwarding_block = rb_intern("&");
    rb_id_forwarding_all = rb_intern("...");
 
    rb_define_const(rb_cPrism, "VERSION", rb_str_freeze(rb_str_new_cstr(EXPECTED_PRISM_VERSION)));
 
    // First, the functions that have to do with lexing and parsing.
    rb_define_singleton_method(rb_cPrism, "lex", lex, -1);
    rb_define_singleton_method(rb_cPrism, "lex_file", lex_file, -1);
    rb_define_singleton_method(rb_cPrism, "parse", parse, -1);
    rb_define_singleton_method(rb_cPrism, "parse_file", parse_file, -1);
    rb_define_singleton_method(rb_cPrism, "profile", profile, -1);
    rb_define_singleton_method(rb_cPrism, "profile_file", profile_file, -1);
    rb_define_singleton_method(rb_cPrism, "parse_stream", parse_stream, -1);
    rb_define_singleton_method(rb_cPrism, "parse_comments", parse_comments, -1);
    rb_define_singleton_method(rb_cPrism, "parse_file_comments", parse_file_comments, -1);
    rb_define_singleton_method(rb_cPrism, "parse_lex", parse_lex, -1);
    rb_define_singleton_method(rb_cPrism, "parse_lex_file", parse_lex_file, -1);
    rb_define_singleton_method(rb_cPrism, "parse_success?", parse_success_p, -1);
    rb_define_singleton_method(rb_cPrism, "parse_failure?", parse_failure_p, -1);
    rb_define_singleton_method(rb_cPrism, "parse_file_success?", parse_file_success_p, -1);
    rb_define_singleton_method(rb_cPrism, "parse_file_failure?", parse_file_failure_p, -1);
 
#ifndef PRISM_EXCLUDE_SERIALIZATION
    rb_define_singleton_method(rb_cPrism, "dump", dump, -1);
    rb_define_singleton_method(rb_cPrism, "dump_file", dump_file, -1);
#endif
 
    rb_define_singleton_method(rb_cPrismStringQuery, "local?", string_query_local_p, 1);
    rb_define_singleton_method(rb_cPrismStringQuery, "constant?", string_query_constant_p, 1);
    rb_define_singleton_method(rb_cPrismStringQuery, "method_name?", string_query_method_name_p, 1);
 
    // Next, initialize the other APIs.
    Init_prism_api_node();
}