prism 0.15.1 → 0.16.0

data/src/prism.c

@@ -40,6 +40,7 @@ debug_context(pm_context_t context) {
         case PM_CONTEXT_BLOCK_BRACES: return "BLOCK_BRACES";
         case PM_CONTEXT_BLOCK_KEYWORDS: return "BLOCK_KEYWORDS";
         case PM_CONTEXT_FOR: return "FOR";
+        case PM_CONTEXT_FOR_INDEX: return "FOR_INDEX";
         case PM_CONTEXT_IF: return "IF";
         case PM_CONTEXT_MAIN: return "MAIN";
         case PM_CONTEXT_MODULE: return "MODULE";
@@ -80,14 +81,12 @@ debug_contexts(pm_parser_t *parser) {
 }
 
 PRISM_ATTRIBUTE_UNUSED static void
-debug_node(const char *message, pm_parser_t *parser, pm_node_t *node) {
-    pm_buffer_t buffer;
-    if (!pm_buffer_init(&buffer)) return;
-
-    pm_prettyprint(parser, node, &buffer);
+debug_node(const pm_parser_t *parser, const pm_node_t *node) {
+    pm_buffer_t output_buffer = { 0 };
+    pm_prettyprint(&output_buffer, parser, node);
 
-    fprintf(stderr, "%s\n%.*s\n", message, (int) buffer.length, buffer.value);
-    pm_buffer_free(&buffer);
+    fprintf(stderr, "%.*s", (int) output_buffer.length, output_buffer.value);
+    pm_buffer_free(&output_buffer);
 }
 
 PRISM_ATTRIBUTE_UNUSED static void
@@ -648,87 +647,6 @@ pm_arguments_validate_block(pm_parser_t *parser, pm_arguments_t *arguments, pm_b
     pm_parser_err_node(parser, (pm_node_t *) block, PM_ERR_ARGUMENT_UNEXPECTED_BLOCK);
 }
 
-/******************************************************************************/
-/* Scope node functions                                                       */
-/******************************************************************************/
-
-// Generate a scope node from the given node.
-void
-pm_scope_node_init(const pm_node_t *node, pm_scope_node_t *scope, pm_scope_node_t *previous, pm_parser_t *parser) {
-    scope->base.type = PM_SCOPE_NODE;
-    scope->base.location.start = node->location.start;
-    scope->base.location.end = node->location.end;
-
-    scope->previous = previous;
-    scope->parser = parser;
-    scope->ast_node = (pm_node_t *)node;
-    scope->parameters = NULL;
-    scope->body = NULL;
-    scope->constants = NULL;
-    if (previous) {
-        scope->constants = previous->constants;
-    }
-    scope->index_lookup_table = NULL;
-
-    pm_constant_id_list_init(&scope->locals);
-
-    switch (PM_NODE_TYPE(node)) {
-        case PM_BLOCK_NODE: {
-            pm_block_node_t *cast = (pm_block_node_t *) node;
-            if (cast->parameters) scope->parameters = cast->parameters->parameters;
-            scope->body = cast->body;
-            scope->locals = cast->locals;
-            break;
-        }
-        case PM_CLASS_NODE: {
-            pm_class_node_t *cast = (pm_class_node_t *) node;
-            scope->body = cast->body;
-            scope->locals = cast->locals;
-            break;
-        }
-        case PM_DEF_NODE: {
-            pm_def_node_t *cast = (pm_def_node_t *) node;
-            scope->parameters = cast->parameters;
-            scope->body = cast->body;
-            scope->locals = cast->locals;
-            break;
-        }
-        case PM_FOR_NODE: {
-            pm_for_node_t *cast = (pm_for_node_t *)node;
-            scope->body = (pm_node_t *)cast->statements;
-            break;
-        }
-        case PM_LAMBDA_NODE: {
-            pm_lambda_node_t *cast = (pm_lambda_node_t *) node;
-            if (cast->parameters) scope->parameters = cast->parameters->parameters;
-            scope->body = cast->body;
-            scope->locals = cast->locals;
-            break;
-        }
-        case PM_MODULE_NODE: {
-            pm_module_node_t *cast = (pm_module_node_t *) node;
-            scope->body = cast->body;
-            scope->locals = cast->locals;
-            break;
-        }
-        case PM_PROGRAM_NODE: {
-            pm_program_node_t *cast = (pm_program_node_t *) node;
-            scope->body = (pm_node_t *) cast->statements;
-            scope->locals = cast->locals;
-            break;
-        }
-        case PM_SINGLETON_CLASS_NODE: {
-            pm_singleton_class_node_t *cast = (pm_singleton_class_node_t *) node;
-            scope->body = cast->body;
-            scope->locals = cast->locals;
-            break;
-        }
-        default:
-            assert(false && "unreachable");
-            break;
-    }
-}
-
 /******************************************************************************/
 /* Node creation functions                                                    */
 /******************************************************************************/
@@ -765,11 +683,15 @@ parse_decimal_number(pm_parser_t *parser, const uint8_t *start, const uint8_t *e
     return (uint32_t) value;
 }
 
+// When you have an encoding flag on a regular expression, it takes precedence
+// over all of the previously set encoding flags. So we need to mask off any
+// previously set encoding flags before setting the new one.
+#define PM_REGULAR_EXPRESSION_ENCODING_MASK ~(PM_REGULAR_EXPRESSION_FLAGS_EUC_JP | PM_REGULAR_EXPRESSION_FLAGS_ASCII_8BIT | PM_REGULAR_EXPRESSION_FLAGS_WINDOWS_31J | PM_REGULAR_EXPRESSION_FLAGS_UTF_8)
+
 // Parse out the options for a regular expression.
 static inline pm_node_flags_t
 pm_regular_expression_flags_create(const pm_token_t *closing) {
     pm_node_flags_t flags = 0;
-    pm_node_flags_t mask = (uint16_t) 0xFF0F;
 
     if (closing->type == PM_TOKEN_REGEXP_END) {
         for (const uint8_t *flag = closing->start + 1; flag < closing->end; flag++) {
@@ -779,10 +701,10 @@ pm_regular_expression_flags_create(const pm_token_t *closing) {
                 case 'x': flags |= PM_REGULAR_EXPRESSION_FLAGS_EXTENDED; break;
                 case 'o': flags |= PM_REGULAR_EXPRESSION_FLAGS_ONCE; break;
 
-                case 'e': flags &= mask; flags |= PM_REGULAR_EXPRESSION_FLAGS_EUC_JP; break;
-                case 'n': flags &= mask; flags |= PM_REGULAR_EXPRESSION_FLAGS_ASCII_8BIT; break;
-                case 's': flags &= mask; flags |= PM_REGULAR_EXPRESSION_FLAGS_WINDOWS_31J; break;
-                case 'u': flags &= mask; flags |= PM_REGULAR_EXPRESSION_FLAGS_UTF_8; break;
+                case 'e': flags = (pm_node_flags_t) (((pm_node_flags_t) (flags & PM_REGULAR_EXPRESSION_ENCODING_MASK)) | PM_REGULAR_EXPRESSION_FLAGS_EUC_JP); break;
+                case 'n': flags = (pm_node_flags_t) (((pm_node_flags_t) (flags & PM_REGULAR_EXPRESSION_ENCODING_MASK)) | PM_REGULAR_EXPRESSION_FLAGS_ASCII_8BIT); break;
+                case 's': flags = (pm_node_flags_t) (((pm_node_flags_t) (flags & PM_REGULAR_EXPRESSION_ENCODING_MASK)) | PM_REGULAR_EXPRESSION_FLAGS_WINDOWS_31J); break;
+                case 'u': flags = (pm_node_flags_t) (((pm_node_flags_t) (flags & PM_REGULAR_EXPRESSION_ENCODING_MASK)) | PM_REGULAR_EXPRESSION_FLAGS_UTF_8); break;
 
                 default: assert(false && "unreachable");
             }
@@ -792,6 +714,8 @@ pm_regular_expression_flags_create(const pm_token_t *closing) {
     return flags;
 }
 
+#undef PM_REGULAR_EXPRESSION_ENCODING_MASK
+
 // Allocate and initialize a new StatementsNode node.
 static pm_statements_node_t *
 pm_statements_node_create(pm_parser_t *parser);
@@ -2664,10 +2588,10 @@ pm_hash_pattern_node_empty_create(pm_parser_t *parser, const pm_token_t *opening
             },
         },
         .constant = NULL,
-        .kwrest = NULL,
         .opening_loc = PM_LOCATION_TOKEN_VALUE(opening),
         .closing_loc = PM_LOCATION_TOKEN_VALUE(closing),
-        .assocs = PM_EMPTY_NODE_LIST
+        .elements = PM_EMPTY_NODE_LIST,
+        .rest = NULL
     };
 
     return node;
@@ -2675,27 +2599,44 @@ pm_hash_pattern_node_empty_create(pm_parser_t *parser, const pm_token_t *opening
 
 // Allocate and initialize a new hash pattern node.
 static pm_hash_pattern_node_t *
-pm_hash_pattern_node_node_list_create(pm_parser_t *parser, pm_node_list_t *assocs) {
+pm_hash_pattern_node_node_list_create(pm_parser_t *parser, pm_node_list_t *elements, pm_node_t *rest) {
     pm_hash_pattern_node_t *node = PM_ALLOC_NODE(parser, pm_hash_pattern_node_t);
 
+    const uint8_t *start;
+    const uint8_t *end;
+
+    if (elements->size > 0) {
+        if (rest) {
+            start = elements->nodes[0]->location.start;
+            end = rest->location.end;
+        } else {
+            start = elements->nodes[0]->location.start;
+            end = elements->nodes[elements->size - 1]->location.end;
+        }
+    } else {
+        assert(rest != NULL);
+        start = rest->location.start;
+        end = rest->location.end;
+    }
+
     *node = (pm_hash_pattern_node_t) {
         {
             .type = PM_HASH_PATTERN_NODE,
             .location = {
-                .start = assocs->nodes[0]->location.start,
-                .end = assocs->nodes[assocs->size - 1]->location.end
+                .start = start,
+                .end = end
             },
         },
         .constant = NULL,
-        .kwrest = NULL,
-        .assocs = PM_EMPTY_NODE_LIST,
+        .elements = PM_EMPTY_NODE_LIST,
+        .rest = rest,
         .opening_loc = PM_OPTIONAL_LOCATION_NOT_PROVIDED_VALUE,
         .closing_loc = PM_OPTIONAL_LOCATION_NOT_PROVIDED_VALUE
     };
 
-    for (size_t index = 0; index < assocs->size; index++) {
-        pm_node_t *assoc = assocs->nodes[index];
-        pm_node_list_append(&node->assocs, assoc);
+    for (size_t index = 0; index < elements->size; index++) {
+        pm_node_t *element = elements->nodes[index];
+        pm_node_list_append(&node->elements, element);
     }
 
     return node;
@@ -3692,7 +3633,9 @@ pm_multi_target_node_create(pm_parser_t *parser) {
             .type = PM_MULTI_TARGET_NODE,
             .location = { .start = NULL, .end = NULL }
         },
-        .targets = PM_EMPTY_NODE_LIST,
+        .lefts = PM_EMPTY_NODE_LIST,
+        .rest = NULL,
+        .rights = PM_EMPTY_NODE_LIST,
         .lparen_loc = PM_OPTIONAL_LOCATION_NOT_PROVIDED_VALUE,
         .rparen_loc = PM_OPTIONAL_LOCATION_NOT_PROVIDED_VALUE
     };
@@ -3702,8 +3645,19 @@ pm_multi_target_node_create(pm_parser_t *parser) {
 
 // Append a target to a MultiTargetNode node.
 static void
-pm_multi_target_node_targets_append(pm_multi_target_node_t *node, pm_node_t *target) {
-    pm_node_list_append(&node->targets, target);
+pm_multi_target_node_targets_append(pm_parser_t *parser, pm_multi_target_node_t *node, pm_node_t *target) {
+    if (PM_NODE_TYPE_P(target, PM_SPLAT_NODE)) {
+        if (node->rest == NULL) {
+            node->rest = target;
+        } else {
+            pm_parser_err_node(parser, target, PM_ERR_MULTI_ASSIGN_MULTI_SPLATS);
+            pm_node_list_append(&node->rights, target);
+        }
+    } else if (node->rest == NULL) {
+        pm_node_list_append(&node->lefts, target);
+    } else {
+        pm_node_list_append(&node->rights, target);
+    }
 
     if (node->base.location.start == NULL || (node->base.location.start > target->location.start)) {
         node->base.location.start = target->location.start;
@@ -3714,6 +3668,20 @@ pm_multi_target_node_targets_append(pm_multi_target_node_t *node, pm_node_t *tar
     }
 }
 
+// Set the opening of a MultiTargetNode node.
+static void
+pm_multi_target_node_opening_set(pm_multi_target_node_t *node, const pm_token_t *lparen) {
+    node->base.location.start = lparen->start;
+    node->lparen_loc = PM_LOCATION_TOKEN_VALUE(lparen);
+}
+
+// Set the closing of a MultiTargetNode node.
+static void
+pm_multi_target_node_closing_set(pm_multi_target_node_t *node, const pm_token_t *rparen) {
+    node->base.location.end = rparen->end;
+    node->rparen_loc = PM_LOCATION_TOKEN_VALUE(rparen);
+}
+
 // Allocate a new MultiWriteNode node.
 static pm_multi_write_node_t *
 pm_multi_write_node_create(pm_parser_t *parser, pm_multi_target_node_t *target, const pm_token_t *operator, pm_node_t *value) {
@@ -3727,7 +3695,9 @@ pm_multi_write_node_create(pm_parser_t *parser, pm_multi_target_node_t *target,
                 .end = value->location.end
             }
         },
-        .targets = target->targets,
+        .lefts = target->lefts,
+        .rest = target->rest,
+        .rights = target->rights,
         .lparen_loc = target->lparen_loc,
         .rparen_loc = target->rparen_loc,
         .operator_loc = PM_LOCATION_TOKEN_VALUE(operator),
@@ -4154,37 +4124,6 @@ pm_regular_expression_node_create(pm_parser_t *parser, const pm_token_t *opening
     return pm_regular_expression_node_create_unescaped(parser, opening, content, closing, &PM_EMPTY_STRING);
 }
 
-// Allocate a new RequiredDestructuredParameterNode node.
-static pm_required_destructured_parameter_node_t *
-pm_required_destructured_parameter_node_create(pm_parser_t *parser, const pm_token_t *opening) {
-    pm_required_destructured_parameter_node_t *node = PM_ALLOC_NODE(parser, pm_required_destructured_parameter_node_t);
-
-    *node = (pm_required_destructured_parameter_node_t) {
-        {
-            .type = PM_REQUIRED_DESTRUCTURED_PARAMETER_NODE,
-            .location = PM_LOCATION_TOKEN_VALUE(opening)
-        },
-        .opening_loc = PM_LOCATION_TOKEN_VALUE(opening),
-        .closing_loc = PM_OPTIONAL_LOCATION_NOT_PROVIDED_VALUE,
-        .parameters = PM_EMPTY_NODE_LIST
-    };
-
-    return node;
-}
-
-// Append a new parameter to the given RequiredDestructuredParameterNode node.
-static void
-pm_required_destructured_parameter_node_append_parameter(pm_required_destructured_parameter_node_t *node, pm_node_t *parameter) {
-    pm_node_list_append(&node->parameters, parameter);
-}
-
-// Set the closing token of the given RequiredDestructuredParameterNode node.
-static void
-pm_required_destructured_parameter_node_closing_set(pm_required_destructured_parameter_node_t *node, const pm_token_t *closing) {
-    node->closing_loc = PM_LOCATION_TOKEN_VALUE(closing);
-    node->base.location.end = closing->end;
-}
-
 // Allocate a new RequiredParameterNode node.
 static pm_required_parameter_node_t *
 pm_required_parameter_node_create(pm_parser_t *parser, const pm_token_t *token) {
@@ -5668,6 +5607,8 @@ context_terminator(pm_context_t context, pm_token_t *token) {
         case PM_CONTEXT_FOR:
         case PM_CONTEXT_ENSURE:
             return token->type == PM_TOKEN_KEYWORD_END;
+        case PM_CONTEXT_FOR_INDEX:
+            return token->type == PM_TOKEN_KEYWORD_IN;
         case PM_CONTEXT_CASE_WHEN:
             return token->type == PM_TOKEN_KEYWORD_WHEN || token->type == PM_TOKEN_KEYWORD_END || token->type == PM_TOKEN_KEYWORD_ELSE;
         case PM_CONTEXT_CASE_IN:
@@ -6103,16 +6044,21 @@ static pm_token_type_t
 lex_identifier(pm_parser_t *parser, bool previous_command_start) {
     // Lex as far as we can into the current identifier.
     size_t width;
-    while (parser->current.end < parser->end && (width = char_is_identifier(parser, parser->current.end)) > 0) {
-        parser->current.end += width;
+    const uint8_t *end = parser->end;
+    const uint8_t *current_start = parser->current.start;
+    const uint8_t *current_end = parser->current.end;
+
+    while (current_end < end && (width = char_is_identifier(parser, current_end)) > 0) {
+        current_end += width;
     }
+    parser->current.end = current_end;
 
     // Now cache the length of the identifier so that we can quickly compare it
     // against known keywords.
-    width = (size_t) (parser->current.end - parser->current.start);
+    width = (size_t) (current_end - current_start);
 
-    if (parser->current.end < parser->end) {
-        if (((parser->current.end + 1 >= parser->end) || (parser->current.end[1] != '=')) && (match(parser, '!') || match(parser, '?'))) {
+    if (current_end < end) {
+        if (((current_end + 1 >= end) || (current_end[1] != '=')) && (match(parser, '!') || match(parser, '?'))) {
             // First we'll attempt to extend the identifier by a ! or ?. Then we'll
             // check if we're returning the defined? keyword or just an identifier.
             width++;
@@ -6222,7 +6168,10 @@ lex_identifier(pm_parser_t *parser, bool previous_command_start) {
         }
     }
 
-    return parser->encoding.isupper_char(parser->current.start, parser->end - parser->current.start) ? PM_TOKEN_CONSTANT : PM_TOKEN_IDENTIFIER;
+    if (parser->encoding_changed) {
+        return parser->encoding.isupper_char(current_start, end - current_start) ? PM_TOKEN_CONSTANT : PM_TOKEN_IDENTIFIER;
+    }
+    return pm_encoding_utf_8_isupper_char(current_start, end - current_start) ? PM_TOKEN_CONSTANT : PM_TOKEN_IDENTIFIER;
 }
 
 // Returns true if the current token that the parser is considering is at the
@@ -6423,24 +6372,24 @@ escape_byte(uint8_t value, const uint8_t flags) {
 static inline void
 escape_write_unicode(pm_parser_t *parser, pm_buffer_t *buffer, const uint8_t *start, const uint8_t *end, uint32_t value) {
     if (value <= 0x7F) { // 0xxxxxxx
-        pm_buffer_append_u8(buffer, (uint8_t) value);
+        pm_buffer_append_byte(buffer, (uint8_t) value);
     } else if (value <= 0x7FF) { // 110xxxxx 10xxxxxx
-        pm_buffer_append_u8(buffer, (uint8_t) (0xC0 | (value >> 6)));
-        pm_buffer_append_u8(buffer, (uint8_t) (0x80 | (value & 0x3F)));
+        pm_buffer_append_byte(buffer, (uint8_t) (0xC0 | (value >> 6)));
+        pm_buffer_append_byte(buffer, (uint8_t) (0x80 | (value & 0x3F)));
     } else if (value <= 0xFFFF) { // 1110xxxx 10xxxxxx 10xxxxxx
-        pm_buffer_append_u8(buffer, (uint8_t) (0xE0 | (value >> 12)));
-        pm_buffer_append_u8(buffer, (uint8_t) (0x80 | ((value >> 6) & 0x3F)));
-        pm_buffer_append_u8(buffer, (uint8_t) (0x80 | (value & 0x3F)));
+        pm_buffer_append_byte(buffer, (uint8_t) (0xE0 | (value >> 12)));
+        pm_buffer_append_byte(buffer, (uint8_t) (0x80 | ((value >> 6) & 0x3F)));
+        pm_buffer_append_byte(buffer, (uint8_t) (0x80 | (value & 0x3F)));
     } else if (value <= 0x10FFFF) { // 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
-        pm_buffer_append_u8(buffer, (uint8_t) (0xF0 | (value >> 18)));
-        pm_buffer_append_u8(buffer, (uint8_t) (0x80 | ((value >> 12) & 0x3F)));
-        pm_buffer_append_u8(buffer, (uint8_t) (0x80 | ((value >> 6) & 0x3F)));
-        pm_buffer_append_u8(buffer, (uint8_t) (0x80 | (value & 0x3F)));
+        pm_buffer_append_byte(buffer, (uint8_t) (0xF0 | (value >> 18)));
+        pm_buffer_append_byte(buffer, (uint8_t) (0x80 | ((value >> 12) & 0x3F)));
+        pm_buffer_append_byte(buffer, (uint8_t) (0x80 | ((value >> 6) & 0x3F)));
+        pm_buffer_append_byte(buffer, (uint8_t) (0x80 | (value & 0x3F)));
     } else {
         pm_parser_err(parser, start, end, PM_ERR_ESCAPE_INVALID_UNICODE);
-        pm_buffer_append_u8(buffer, 0xEF);
-        pm_buffer_append_u8(buffer, 0xBF);
-        pm_buffer_append_u8(buffer, 0xBD);
+        pm_buffer_append_byte(buffer, 0xEF);
+        pm_buffer_append_byte(buffer, 0xBF);
+        pm_buffer_append_byte(buffer, 0xBD);
     }
 }
 
@@ -6466,18 +6415,18 @@ escape_write_byte(pm_buffer_t *buffer, uint8_t flags, uint8_t byte) {
         uint8_t byte2 = (uint8_t) (byte & 0xF);
 
         if (byte1 >= 0xA) {
-            pm_buffer_append_u8(buffer, (uint8_t) ((byte1 - 0xA) + 'A'));
+            pm_buffer_append_byte(buffer, (uint8_t) ((byte1 - 0xA) + 'A'));
         } else {
-            pm_buffer_append_u8(buffer, (uint8_t) (byte1 + '0'));
+            pm_buffer_append_byte(buffer, (uint8_t) (byte1 + '0'));
         }
 
         if (byte2 >= 0xA) {
-            pm_buffer_append_u8(buffer, (uint8_t) (byte2 - 0xA + 'A'));
+            pm_buffer_append_byte(buffer, (uint8_t) (byte2 - 0xA + 'A'));
         } else {
-            pm_buffer_append_u8(buffer, (uint8_t) (byte2 + '0'));
+            pm_buffer_append_byte(buffer, (uint8_t) (byte2 + '0'));
         }
     } else {
-        pm_buffer_append_u8(buffer, byte);
+        pm_buffer_append_byte(buffer, byte);
     }
 }
 
@@ -6487,57 +6436,57 @@ escape_read(pm_parser_t *parser, pm_buffer_t *buffer, uint8_t flags) {
     switch (peek(parser)) {
         case '\\': {
             parser->current.end++;
-            pm_buffer_append_u8(buffer, '\\');
+            pm_buffer_append_byte(buffer, '\\');
             return;
         }
         case '\'': {
             parser->current.end++;
-            pm_buffer_append_u8(buffer, '\'');
+            pm_buffer_append_byte(buffer, '\'');
             return;
         }
         case 'a': {
             parser->current.end++;
-            pm_buffer_append_u8(buffer, '\a');
+            pm_buffer_append_byte(buffer, '\a');
             return;
         }
         case 'b': {
             parser->current.end++;
-            pm_buffer_append_u8(buffer, '\b');
+            pm_buffer_append_byte(buffer, '\b');
             return;
         }
         case 'e': {
             parser->current.end++;
-            pm_buffer_append_u8(buffer, '\033');
+            pm_buffer_append_byte(buffer, '\033');
             return;
         }
         case 'f': {
             parser->current.end++;
-            pm_buffer_append_u8(buffer, '\f');
+            pm_buffer_append_byte(buffer, '\f');
             return;
         }
         case 'n': {
             parser->current.end++;
-            pm_buffer_append_u8(buffer, '\n');
+            pm_buffer_append_byte(buffer, '\n');
             return;
         }
         case 'r': {
             parser->current.end++;
-            pm_buffer_append_u8(buffer, '\r');
+            pm_buffer_append_byte(buffer, '\r');
             return;
         }
         case 's': {
             parser->current.end++;
-            pm_buffer_append_u8(buffer, ' ');
+            pm_buffer_append_byte(buffer, ' ');
             return;
         }
         case 't': {
             parser->current.end++;
-            pm_buffer_append_u8(buffer, '\t');
+            pm_buffer_append_byte(buffer, '\t');
             return;
         }
         case 'v': {
             parser->current.end++;
-            pm_buffer_append_u8(buffer, '\v');
+            pm_buffer_append_byte(buffer, '\v');
             return;
         }
         case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': {
@@ -6554,7 +6503,7 @@ escape_read(pm_parser_t *parser, pm_buffer_t *buffer, uint8_t flags) {
                 }
             }
 
-            pm_buffer_append_u8(buffer, value);
+            pm_buffer_append_byte(buffer, value);
             return;
         }
         case 'x': {
@@ -6576,7 +6525,7 @@ escape_read(pm_parser_t *parser, pm_buffer_t *buffer, uint8_t flags) {
                 if (flags & PM_ESCAPE_FLAG_REGEXP) {
                     pm_buffer_append_bytes(buffer, start, (size_t) (parser->current.end - start));
                 } else {
-                    pm_buffer_append_u8(buffer, value);
+                    pm_buffer_append_byte(buffer, value);
                 }
             } else {
                 pm_parser_err_current(parser, PM_ERR_ESCAPE_INVALID_HEXADECIMAL);
@@ -6770,14 +6719,14 @@ escape_read(pm_parser_t *parser, pm_buffer_t *buffer, uint8_t flags) {
         case '\r': {
             if (peek_offset(parser, 1) == '\n') {
                 parser->current.end += 2;
-                pm_buffer_append_u8(buffer, '\n');
+                pm_buffer_append_byte(buffer, '\n');
                 return;
             }
         }
         /* fallthrough */
         default: {
             if (parser->current.end < parser->end) {
-                pm_buffer_append_u8(buffer, *parser->current.end++);
+                pm_buffer_append_byte(buffer, *parser->current.end++);
             }
             return;
         }
@@ -7031,7 +6980,7 @@ typedef struct {
 // Push the given byte into the token buffer.
 static inline void
 pm_token_buffer_push(pm_token_buffer_t *token_buffer, uint8_t byte) {
-    pm_buffer_append_u8(&token_buffer->buffer, byte);
+    pm_buffer_append_byte(&token_buffer->buffer, byte);
 }
 
 // When we're about to return from lexing the current token and we know for sure
@@ -8328,7 +8277,7 @@ parser_lex(pm_parser_t *parser) {
 
             // If we haven't found an escape yet, then this buffer will be
             // unallocated since we can refer directly to the source string.
-            pm_token_buffer_t token_buffer = { 0 };
+            pm_token_buffer_t token_buffer = { { 0 }, 0 };
 
             while (breakpoint != NULL) {
                 // If we hit a null byte, skip directly past it.
@@ -8504,7 +8453,7 @@ parser_lex(pm_parser_t *parser) {
             // characters.
             const uint8_t *breakpoints = lex_mode->as.regexp.breakpoints;
             const uint8_t *breakpoint = pm_strpbrk(parser, parser->current.end, breakpoints, parser->end - parser->current.end);
-            pm_token_buffer_t token_buffer = { 0 };
+            pm_token_buffer_t token_buffer = { { 0 }, 0 };
 
             while (breakpoint != NULL) {
                 // If we hit a null byte, skip directly past it.
@@ -8693,7 +8642,7 @@ parser_lex(pm_parser_t *parser) {
 
             // If we haven't found an escape yet, then this buffer will be
             // unallocated since we can refer directly to the source string.
-            pm_token_buffer_t token_buffer = { 0 };
+            pm_token_buffer_t token_buffer = { { 0 }, 0 };
 
             while (breakpoint != NULL) {
                 // If we hit the incrementor, then we'll increment then nesting and
@@ -8954,7 +8903,7 @@ parser_lex(pm_parser_t *parser) {
             }
 
             const uint8_t *breakpoint = pm_strpbrk(parser, parser->current.end, breakpoints, parser->end - parser->current.end);
-            pm_token_buffer_t token_buffer = { 0 };
+            pm_token_buffer_t token_buffer = { { 0 }, 0 };
             bool was_escaped_newline = false;
 
             while (breakpoint != NULL) {
@@ -9572,10 +9521,7 @@ parse_target(pm_parser_t *parser, pm_node_t *target) {
                 splat->expression = parse_target(parser, splat->expression);
             }
 
-            pm_multi_target_node_t *multi_target = pm_multi_target_node_create(parser);
-            pm_multi_target_node_targets_append(multi_target, (pm_node_t *) splat);
-
-            return (pm_node_t *) multi_target;
+            return (pm_node_t *) splat;
         }
         case PM_CALL_NODE: {
             pm_call_node_t *call = (pm_call_node_t *) target;
@@ -9651,7 +9597,7 @@ parse_target(pm_parser_t *parser, pm_node_t *target) {
     }
 }
 
-// Parse a write targets and validate that it is in a valid position for
+// Parse a write target and validate that it is in a valid position for
 // assignment.
 static pm_node_t *
 parse_target_validate(pm_parser_t *parser, pm_node_t *target) {
@@ -9722,7 +9668,7 @@ parse_write(pm_parser_t *parser, pm_node_t *target, pm_token_t *operator, pm_nod
             }
 
             pm_multi_target_node_t *multi_target = pm_multi_target_node_create(parser);
-            pm_multi_target_node_targets_append(multi_target, (pm_node_t *) splat);
+            pm_multi_target_node_targets_append(parser, multi_target, (pm_node_t *) splat);
 
             return (pm_node_t *) pm_multi_write_node_create(parser, multi_target, operator, value);
         }
@@ -9838,7 +9784,7 @@ parse_targets(pm_parser_t *parser, pm_node_t *first_target, pm_binding_power_t b
     bool has_splat = PM_NODE_TYPE_P(first_target, PM_SPLAT_NODE);
 
     pm_multi_target_node_t *result = pm_multi_target_node_create(parser);
-    pm_multi_target_node_targets_append(result, parse_target(parser, first_target));
+    pm_multi_target_node_targets_append(parser, result, parse_target(parser, first_target));
 
     while (accept1(parser, PM_TOKEN_COMMA)) {
         if (accept1(parser, PM_TOKEN_USTAR)) {
@@ -9858,19 +9804,19 @@ parse_targets(pm_parser_t *parser, pm_node_t *first_target, pm_binding_power_t b
             }
 
             pm_node_t *splat = (pm_node_t *) pm_splat_node_create(parser, &star_operator, name);
-            pm_multi_target_node_targets_append(result, splat);
+            pm_multi_target_node_targets_append(parser, result, splat);
             has_splat = true;
         } else if (token_begins_expression_p(parser->current.type)) {
             pm_node_t *target = parse_expression(parser, binding_power, PM_ERR_EXPECT_EXPRESSION_AFTER_COMMA);
             target = parse_target(parser, target);
 
-            pm_multi_target_node_targets_append(result, target);
-        } else {
+            pm_multi_target_node_targets_append(parser, result, target);
+        } else if (!match1(parser, PM_TOKEN_EOF)) {
             // If we get here, then we have a trailing , in a multi target node.
             // We need to indicate this somehow in the tree, so we'll add an
             // anonymous splat.
             pm_node_t *splat = (pm_node_t *) pm_splat_node_create(parser, &parser->previous, NULL);
-            pm_multi_target_node_targets_append(result, splat);
+            pm_multi_target_node_targets_append(parser, result, splat);
             break;
         }
     }
@@ -9963,9 +9909,11 @@ parse_statements(pm_parser_t *parser, pm_context_t context) {
 }
 
 // Parse all of the elements of a hash.
-static void
+// Returns true if a double splat was found
+static bool
 parse_assocs(pm_parser_t *parser, pm_node_t *node) {
     assert(PM_NODE_TYPE_P(node, PM_HASH_NODE) || PM_NODE_TYPE_P(node, PM_KEYWORD_HASH_NODE));
+    bool contains_keyword_splat = false;
 
     while (true) {
         pm_node_t *element;
@@ -9983,6 +9931,7 @@ parse_assocs(pm_parser_t *parser, pm_node_t *node) {
                 }
 
                 element = (pm_node_t *) pm_assoc_splat_node_create(parser, value, &operator);
+                contains_keyword_splat = true;
                 break;
             }
             case PM_TOKEN_LABEL: {
@@ -10041,7 +9990,7 @@ parse_assocs(pm_parser_t *parser, pm_node_t *node) {
         }
 
         // If there's no comma after the element, then we're done.
-        if (!accept1(parser, PM_TOKEN_COMMA)) return;
+        if (!accept1(parser, PM_TOKEN_COMMA)) break;
 
         // If the next element starts with a label or a **, then we know we have
         // another element in the hash, so we'll continue parsing.
@@ -10052,8 +10001,9 @@ parse_assocs(pm_parser_t *parser, pm_node_t *node) {
         if (token_begins_expression_p(parser->current.type)) continue;
 
         // Otherwise by default we will exit out of this loop.
-        return;
+        break;
     }
+    return contains_keyword_splat;
 }
 
 // Append an argument to a list of arguments.
@@ -10101,12 +10051,16 @@ parse_arguments(pm_parser_t *parser, pm_arguments_t *arguments, bool accepts_for
                 pm_keyword_hash_node_t *hash = pm_keyword_hash_node_create(parser);
                 argument = (pm_node_t *) hash;
 
+                bool contains_keyword_splat = false;
                 if (!match7(parser, terminator, PM_TOKEN_NEWLINE, PM_TOKEN_SEMICOLON, PM_TOKEN_EOF, PM_TOKEN_BRACE_RIGHT, PM_TOKEN_KEYWORD_DO, PM_TOKEN_PARENTHESIS_RIGHT)) {
-                    parse_assocs(parser, (pm_node_t *) hash);
+                    contains_keyword_splat = parse_assocs(parser, (pm_node_t *) hash);
                 }
 
                 parsed_bare_hash = true;
                 parse_arguments_append(parser, arguments, argument);
+                if (contains_keyword_splat) {
+                    arguments->arguments->base.flags |= PM_ARGUMENTS_NODE_FLAGS_KEYWORD_SPLAT;
+                }
                 break;
             }
             case PM_TOKEN_UAMPERSAND: {
@@ -10180,6 +10134,7 @@ parse_arguments(pm_parser_t *parser, pm_arguments_t *arguments, bool accepts_for
                     argument = parse_expression(parser, PM_BINDING_POWER_DEFINED, PM_ERR_EXPECT_ARGUMENT);
                 }
 
+                bool contains_keyword_splat = false;
                 if (pm_symbol_node_label_p(argument) || accept1(parser, PM_TOKEN_EQUAL_GREATER)) {
                     if (parsed_bare_hash) {
                         pm_parser_err_previous(parser, PM_ERR_ARGUMENT_BARE_HASH);
@@ -10206,13 +10161,16 @@ parse_arguments(pm_parser_t *parser, pm_arguments_t *arguments, bool accepts_for
                         token_begins_expression_p(parser->current.type) ||
                         match2(parser, PM_TOKEN_USTAR_STAR, PM_TOKEN_LABEL)
                     )) {
-                        parse_assocs(parser, (pm_node_t *) bare_hash);
+                        contains_keyword_splat = parse_assocs(parser, (pm_node_t *) bare_hash);
                     }
 
                     parsed_bare_hash = true;
                 }
 
                 parse_arguments_append(parser, arguments, argument);
+                if (contains_keyword_splat) {
+                    arguments->arguments->base.flags |= PM_ARGUMENTS_NODE_FLAGS_KEYWORD_SPLAT;
+                }
                 break;
             }
         }
@@ -10248,34 +10206,27 @@ parse_arguments(pm_parser_t *parser, pm_arguments_t *arguments, bool accepts_for
 //     end
 //
 // It can recurse infinitely down, and splats are allowed to group arguments.
-static pm_required_destructured_parameter_node_t *
+static pm_multi_target_node_t *
 parse_required_destructured_parameter(pm_parser_t *parser) {
     expect1(parser, PM_TOKEN_PARENTHESIS_LEFT, PM_ERR_EXPECT_LPAREN_REQ_PARAMETER);
 
-    pm_token_t opening = parser->previous;
-    pm_required_destructured_parameter_node_t *node = pm_required_destructured_parameter_node_create(parser, &opening);
-    bool parsed_splat = false;
+    pm_multi_target_node_t *node = pm_multi_target_node_create(parser);
+    pm_multi_target_node_opening_set(node, &parser->previous);
 
     do {
         pm_node_t *param;
 
-        if (node->parameters.size > 0 && match1(parser, PM_TOKEN_PARENTHESIS_RIGHT)) {
-            if (parsed_splat) {
-                pm_parser_err_previous(parser, PM_ERR_ARGUMENT_SPLAT_AFTER_SPLAT);
-            }
-
+        // If we get here then we have a trailing comma. In this case we'll
+        // create an implicit splat node.
+        if (node->lefts.size > 0 && match1(parser, PM_TOKEN_PARENTHESIS_RIGHT)) {
             param = (pm_node_t *) pm_splat_node_create(parser, &parser->previous, NULL);
-            pm_required_destructured_parameter_node_append_parameter(node, param);
+            pm_multi_target_node_targets_append(parser, node, param);
             break;
         }
 
         if (match1(parser, PM_TOKEN_PARENTHESIS_LEFT)) {
             param = (pm_node_t *) parse_required_destructured_parameter(parser);
         } else if (accept1(parser, PM_TOKEN_USTAR)) {
-            if (parsed_splat) {
-                pm_parser_err_previous(parser, PM_ERR_ARGUMENT_SPLAT_AFTER_SPLAT);
-            }
-
             pm_token_t star = parser->previous;
             pm_node_t *value = NULL;
 
@@ -10287,7 +10238,6 @@ parse_required_destructured_parameter(pm_parser_t *parser) {
             }
 
             param = (pm_node_t *) pm_splat_node_create(parser, &star, value);
-            parsed_splat = true;
         } else {
             expect1(parser, PM_TOKEN_IDENTIFIER, PM_ERR_EXPECT_IDENT_REQ_PARAMETER);
             pm_token_t name = parser->previous;
@@ -10297,11 +10247,11 @@ parse_required_destructured_parameter(pm_parser_t *parser) {
             pm_parser_local_add_token(parser, &name);
         }
 
-        pm_required_destructured_parameter_node_append_parameter(node, param);
+        pm_multi_target_node_targets_append(parser, node, param);
     } while (accept1(parser, PM_TOKEN_COMMA));
 
     expect1(parser, PM_TOKEN_PARENTHESIS_RIGHT, PM_ERR_EXPECT_RPAREN_REQ_PARAMETER);
-    pm_required_destructured_parameter_node_closing_set(node, &parser->previous);
+    pm_multi_target_node_closing_set(node, &parser->previous);
 
     return node;
 }
@@ -11787,28 +11737,40 @@ parse_pattern_keyword_rest(pm_parser_t *parser) {
 // Parse a hash pattern.
 static pm_hash_pattern_node_t *
 parse_pattern_hash(pm_parser_t *parser, pm_node_t *first_assoc) {
-    if (PM_NODE_TYPE_P(first_assoc, PM_ASSOC_NODE)) {
-        if (!match7(parser, PM_TOKEN_COMMA, PM_TOKEN_KEYWORD_THEN, PM_TOKEN_BRACE_RIGHT, PM_TOKEN_BRACKET_RIGHT, PM_TOKEN_PARENTHESIS_RIGHT, PM_TOKEN_NEWLINE, PM_TOKEN_SEMICOLON)) {
-            // Here we have a value for the first assoc in the list, so we will parse it
-            // now and update the first assoc.
-            pm_node_t *value = parse_pattern(parser, false, PM_ERR_PATTERN_EXPRESSION_AFTER_KEY);
-
-            pm_assoc_node_t *assoc = (pm_assoc_node_t *) first_assoc;
-            assoc->base.location.end = value->location.end;
-            assoc->value = value;
-        } else {
-            pm_node_t *key = ((pm_assoc_node_t *) first_assoc)->key;
+    pm_node_list_t assocs = PM_EMPTY_NODE_LIST;
+    pm_node_t *rest = NULL;
 
-            if (PM_NODE_TYPE_P(key, PM_SYMBOL_NODE)) {
-                const pm_location_t *value_loc = &((pm_symbol_node_t *) key)->value_loc;
-                pm_parser_local_add_location(parser, value_loc->start, value_loc->end);
+    switch (PM_NODE_TYPE(first_assoc)) {
+        case PM_ASSOC_NODE: {
+            if (!match7(parser, PM_TOKEN_COMMA, PM_TOKEN_KEYWORD_THEN, PM_TOKEN_BRACE_RIGHT, PM_TOKEN_BRACKET_RIGHT, PM_TOKEN_PARENTHESIS_RIGHT, PM_TOKEN_NEWLINE, PM_TOKEN_SEMICOLON)) {
+                // Here we have a value for the first assoc in the list, so we will
+                // parse it now and update the first assoc.
+                pm_node_t *value = parse_pattern(parser, false, PM_ERR_PATTERN_EXPRESSION_AFTER_KEY);
+
+                pm_assoc_node_t *assoc = (pm_assoc_node_t *) first_assoc;
+                assoc->base.location.end = value->location.end;
+                assoc->value = value;
+            } else {
+                pm_node_t *key = ((pm_assoc_node_t *) first_assoc)->key;
+
+                if (PM_NODE_TYPE_P(key, PM_SYMBOL_NODE)) {
+                    const pm_location_t *value_loc = &((pm_symbol_node_t *) key)->value_loc;
+                    pm_parser_local_add_location(parser, value_loc->start, value_loc->end);
+                }
             }
+
+            pm_node_list_append(&assocs, first_assoc);
+            break;
         }
+        case PM_ASSOC_SPLAT_NODE:
+        case PM_NO_KEYWORDS_PARAMETER_NODE:
+            rest = first_assoc;
+            break;
+        default:
+            assert(false);
+            break;
     }
 
-    pm_node_list_t assocs = PM_EMPTY_NODE_LIST;
-    pm_node_list_append(&assocs, first_assoc);
-
     // If there are any other assocs, then we'll parse them now.
     while (accept1(parser, PM_TOKEN_COMMA)) {
         // Here we need to break to support trailing commas.
@@ -11839,7 +11801,7 @@ parse_pattern_hash(pm_parser_t *parser, pm_node_t *first_assoc) {
         pm_node_list_append(&assocs, assoc);
     }
 
-    pm_hash_pattern_node_t *node = pm_hash_pattern_node_node_list_create(parser, &assocs);
+    pm_hash_pattern_node_t *node = pm_hash_pattern_node_node_list_create(parser, &assocs, rest);
     free(assocs.nodes);
 
     return node;
@@ -11924,32 +11886,45 @@ parse_pattern_primitive(pm_parser_t *parser, pm_diagnostic_id_t diag_id) {
                 // pattern node.
                 node = pm_hash_pattern_node_empty_create(parser, &opening, &parser->previous);
             } else {
-                pm_node_t *key;
+                pm_node_t *first_assoc;
 
                 switch (parser->current.type) {
-                    case PM_TOKEN_LABEL:
+                    case PM_TOKEN_LABEL: {
                         parser_lex(parser);
-                        key = (pm_node_t *) pm_symbol_node_label_create(parser, &parser->previous);
+
+                        pm_symbol_node_t *key = pm_symbol_node_label_create(parser, &parser->previous);
+                        pm_token_t operator = not_provided(parser);
+
+                        first_assoc = (pm_node_t *) pm_assoc_node_create(parser, (pm_node_t *) key, &operator, NULL);
                         break;
+                    }
                     case PM_TOKEN_USTAR_STAR:
-                        key = parse_pattern_keyword_rest(parser);
+                        first_assoc = parse_pattern_keyword_rest(parser);
                         break;
-                    case PM_TOKEN_STRING_BEGIN:
-                        key = parse_expression(parser, PM_BINDING_POWER_MAX, PM_ERR_PATTERN_HASH_KEY);
+                    case PM_TOKEN_STRING_BEGIN: {
+                        pm_node_t *key = parse_expression(parser, PM_BINDING_POWER_MAX, PM_ERR_PATTERN_HASH_KEY);
+                        pm_token_t operator = not_provided(parser);
+
                         if (!pm_symbol_node_label_p(key)) {
                             pm_parser_err_node(parser, key, PM_ERR_PATTERN_HASH_KEY_LABEL);
                         }
 
+                        first_assoc = (pm_node_t *) pm_assoc_node_create(parser, key, &operator, NULL);
                         break;
-                    default:
+                    }
+                    default: {
                         parser_lex(parser);
                         pm_parser_err_previous(parser, PM_ERR_PATTERN_HASH_KEY);
-                        key = (pm_node_t *) pm_missing_node_create(parser, parser->previous.start, parser->previous.end);
+
+                        pm_missing_node_t *key = pm_missing_node_create(parser, parser->previous.start, parser->previous.end);
+                        pm_token_t operator = not_provided(parser);
+
+                        first_assoc = (pm_node_t *) pm_assoc_node_create(parser, (pm_node_t *) key, &operator, NULL);
                         break;
+                    }
                 }
 
-                pm_token_t operator = not_provided(parser);
-                node = parse_pattern_hash(parser, (pm_node_t *) pm_assoc_node_create(parser, key, &operator, NULL));
+                node = parse_pattern_hash(parser, first_assoc);
 
                 accept1(parser, PM_TOKEN_NEWLINE);
                 expect1(parser, PM_TOKEN_BRACE_RIGHT, PM_ERR_PATTERN_TERM_BRACE);
@@ -12577,16 +12552,17 @@ parse_expression_prefix(pm_parser_t *parser, pm_binding_power_t binding_power) {
                 parser_lex(parser);
                 pm_accepts_block_stack_pop(parser);
 
-                // If we have a single statement and are ending on a right
-                // parenthesis, then we need to check if this is possibly a
-                // multiple target node.
-                if (PM_NODE_TYPE_P(statement, PM_MULTI_TARGET_NODE)) {
+                if (PM_NODE_TYPE_P(statement, PM_MULTI_TARGET_NODE) || PM_NODE_TYPE_P(statement, PM_SPLAT_NODE)) {
+                    // If we have a single statement and are ending on a right
+                    // parenthesis, then we need to check if this is possibly a
+                    // multiple target node.
                     pm_multi_target_node_t *multi_target;
-                    if (((pm_multi_target_node_t *) statement)->lparen_loc.start == NULL) {
+
+                    if (PM_NODE_TYPE_P(statement, PM_MULTI_TARGET_NODE) && ((pm_multi_target_node_t *) statement)->lparen_loc.start == NULL) {
                         multi_target = (pm_multi_target_node_t *) statement;
                     } else {
                         multi_target = pm_multi_target_node_create(parser);
-                        pm_multi_target_node_targets_append(multi_target, statement);
+                        pm_multi_target_node_targets_append(parser, multi_target, statement);
                     }
 
                     pm_location_t lparen_loc = PM_LOCATION_TOKEN_VALUE(&opening);
@@ -12598,10 +12574,13 @@ parse_expression_prefix(pm_parser_t *parser, pm_binding_power_t binding_power) {
                     multi_target->base.location.end = rparen_loc.end;
 
                     if (match1(parser, PM_TOKEN_COMMA)) {
-                        return parse_targets_validate(parser, (pm_node_t *) multi_target, PM_BINDING_POWER_INDEX);
-                    } else {
-                        return parse_target_validate(parser, (pm_node_t *) multi_target);
+                        if (binding_power == PM_BINDING_POWER_STATEMENT) {
+                            return parse_targets_validate(parser, (pm_node_t *) multi_target, PM_BINDING_POWER_INDEX);
+                        }
+                        return (pm_node_t *) multi_target;
                     }
+
+                    return parse_target_validate(parser, (pm_node_t *) multi_target);
                 }
 
                 // If we have a single statement and are ending on a right parenthesis
@@ -13688,7 +13667,9 @@ parse_expression_prefix(pm_parser_t *parser, pm_binding_power_t binding_power) {
             parser_lex(parser);
             pm_token_t for_keyword = parser->previous;
             pm_node_t *index;
+
             pm_parser_scope_push_transparent(parser);
+            context_push(parser, PM_CONTEXT_FOR_INDEX);
 
             // First, parse out the first index expression.
             if (accept1(parser, PM_TOKEN_USTAR)) {
@@ -13714,6 +13695,7 @@ parse_expression_prefix(pm_parser_t *parser, pm_binding_power_t binding_power) {
                 index = parse_target(parser, index);
             }
 
+            context_pop(parser);
             pm_parser_scope_pop(parser);
             pm_do_loop_stack_push(parser, true);
 
@@ -14596,6 +14578,50 @@ parse_call_operator_write(pm_parser_t *parser, pm_call_node_t *call_node, const
     }
 }
 
+// Potentially change a =~ with a regular expression with named captures into a
+// match write node.
+static pm_node_t *
+parse_regular_expression_named_captures(pm_parser_t *parser, const pm_string_t *content, pm_call_node_t *call) {
+    pm_string_list_t named_captures;
+    pm_string_list_init(&named_captures);
+
+    pm_node_t *result;
+    if (pm_regexp_named_capture_group_names(pm_string_source(content), pm_string_length(content), &named_captures, parser->encoding_changed, &parser->encoding) && (named_captures.length > 0)) {
+        pm_match_write_node_t *match = pm_match_write_node_create(parser, call);
+
+        for (size_t index = 0; index < named_captures.length; index++) {
+            pm_string_t *name = &named_captures.strings[index];
+            pm_constant_id_t local;
+
+            if (content->type == PM_STRING_SHARED) {
+                // If the unescaped string is a slice of the source,
+                // then we can copy the names directly. The pointers
+                // will line up.
+                local = pm_parser_local_add_location(parser, name->source, name->source + name->length);
+            } else {
+                // Otherwise, the name is a slice of the malloc-ed
+                // owned string, in which case we need to copy it
+                // out into a new string.
+                size_t length = pm_string_length(name);
+
+                void *memory = malloc(length);
+                memcpy(memory, pm_string_source(name), length);
+
+                local = pm_parser_local_add_owned(parser, (const uint8_t *) memory, length);
+            }
+
+            pm_constant_id_list_append(&match->locals, local);
+        }
+
+        result = (pm_node_t *) match;
+    } else {
+        result = (pm_node_t *) call;
+    }
+
+    pm_string_list_free(&named_captures);
+    return result;
+}
+
 static inline pm_node_t *
 parse_expression_infix(pm_parser_t *parser, pm_node_t *node, pm_binding_power_t previous_binding_power, pm_binding_power_t binding_power) {
     pm_token_t token = parser->current;
@@ -14620,18 +14646,13 @@ parse_expression_infix(pm_parser_t *parser, pm_node_t *node, pm_binding_power_t
                     return parse_write(parser, node, &token, value);
                 }
                 case PM_SPLAT_NODE: {
-                    pm_splat_node_t *splat_node = (pm_splat_node_t *) node;
+                    pm_multi_target_node_t *multi_target = pm_multi_target_node_create(parser);
+                    pm_multi_target_node_targets_append(parser, multi_target, node);
 
-                    switch (PM_NODE_TYPE(splat_node->expression)) {
-                        case PM_CASE_WRITABLE:
-                            parser_lex(parser);
-                            pm_node_t *value = parse_assignment_value(parser, previous_binding_power, binding_power, PM_ERR_EXPECT_EXPRESSION_AFTER_EQUAL);
-                            return parse_write(parser, (pm_node_t *) splat_node, &token, value);
-                        default:
-                            break;
-                    }
+                    parser_lex(parser);
+                    pm_node_t *value = parse_assignment_value(parser, previous_binding_power, binding_power, PM_ERR_EXPECT_EXPRESSION_AFTER_EQUAL);
+                    return parse_write(parser, (pm_node_t *) multi_target, &token, value);
                 }
-                /* fallthrough */
                 default:
                     parser_lex(parser);
 
@@ -15026,42 +15047,51 @@ parse_expression_infix(pm_parser_t *parser, pm_node_t *node, pm_binding_power_t
             // If the receiver of this =~ is a regular expression node, then we
             // need to introduce local variables for it based on its named
             // capture groups.
-            if (PM_NODE_TYPE_P(node, PM_REGULAR_EXPRESSION_NODE)) {
-                pm_string_list_t named_captures;
-                pm_string_list_init(&named_captures);
-
-                const pm_string_t *unescaped = &((pm_regular_expression_node_t *) node)->unescaped;
-                if (pm_regexp_named_capture_group_names(pm_string_source(unescaped), pm_string_length(unescaped), &named_captures, parser->encoding_changed, &parser->encoding) && (named_captures.length > 0)) {
-                    pm_match_write_node_t *match = pm_match_write_node_create(parser, call);
-
-                    for (size_t index = 0; index < named_captures.length; index++) {
-                        pm_string_t *name = &named_captures.strings[index];
-                        pm_constant_id_t local;
-
-                        if (unescaped->type == PM_STRING_SHARED) {
-                            // If the unescaped string is a slice of the source,
-                            // then we can copy the names directly. The pointers
-                            // will line up.
-                            local = pm_parser_local_add_location(parser, name->source, name->source + name->length);
-                        } else {
-                            // Otherwise, the name is a slice of the malloc-ed
-                            // owned string, in which case we need to copy it
-                            // out into a new string.
-                            size_t length = pm_string_length(name);
+            if (PM_NODE_TYPE_P(node, PM_INTERPOLATED_REGULAR_EXPRESSION_NODE)) {
+                // It's possible to have an interpolated regular expression node
+                // that only contains strings. This is because it can be split
+                // up by a heredoc. In this case we need to concat the unescaped
+                // strings together and then parse them as a regular expression.
+                pm_node_list_t *parts = &((pm_interpolated_regular_expression_node_t *) node)->parts;
 
-                            void *memory = malloc(length);
-                            memcpy(memory, pm_string_source(name), length);
+                bool interpolated = false;
+                size_t total_length = 0;
 
-                            local = pm_parser_local_add_owned(parser, (const uint8_t *) memory, length);
-                        }
+                for (size_t index = 0; index < parts->size; index++) {
+                    pm_node_t *part = parts->nodes[index];
 
-                        pm_constant_id_list_append(&match->locals, local);
+                    if (PM_NODE_TYPE_P(part, PM_STRING_NODE)) {
+                        total_length += pm_string_length(&((pm_string_node_t *) part)->unescaped);
+                    } else {
+                        interpolated = true;
+                        break;
                     }
-
-                    result = (pm_node_t *) match;
                 }
 
-                pm_string_list_free(&named_captures);
+                if (!interpolated) {
+                    void *memory = malloc(total_length);
+                    if (!memory) abort();
+
+                    uint8_t *cursor = memory;
+                    for (size_t index = 0; index < parts->size; index++) {
+                        pm_string_t *unescaped = &((pm_string_node_t *) parts->nodes[index])->unescaped;
+                        size_t length = pm_string_length(unescaped);
+
+                        memcpy(cursor, pm_string_source(unescaped), length);
+                        cursor += length;
+                    }
+
+                    pm_string_t owned;
+                    pm_string_owned_init(&owned, (uint8_t *) memory, total_length);
+
+                    result = parse_regular_expression_named_captures(parser, &owned, call);
+                    pm_string_free(&owned);
+                }
+            } else if (PM_NODE_TYPE_P(node, PM_REGULAR_EXPRESSION_NODE)) {
+                // If we have a regular expression node, then we can just parse
+                // the named captures directly off the unescaped string.
+                const pm_string_t *content = &((pm_regular_expression_node_t *) node)->unescaped;
+                result = parse_regular_expression_named_captures(parser, content, call);
             }
 
             return result;
@@ -15667,16 +15697,20 @@ pm_parse(pm_parser_t *parser) {
     return parse_program(parser);
 }
 
+static inline void
+pm_serialize_header(pm_buffer_t *buffer) {
+    pm_buffer_append_string(buffer, "PRISM", 5);
+    pm_buffer_append_byte(buffer, PRISM_VERSION_MAJOR);
+    pm_buffer_append_byte(buffer, PRISM_VERSION_MINOR);
+    pm_buffer_append_byte(buffer, PRISM_VERSION_PATCH);
+    pm_buffer_append_byte(buffer, PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS ? 1 : 0);
+}
+
 PRISM_EXPORTED_FUNCTION void
 pm_serialize(pm_parser_t *parser, pm_node_t *node, pm_buffer_t *buffer) {
-    pm_buffer_append_str(buffer, "PRISM", 5);
-    pm_buffer_append_u8(buffer, PRISM_VERSION_MAJOR);
-    pm_buffer_append_u8(buffer, PRISM_VERSION_MINOR);
-    pm_buffer_append_u8(buffer, PRISM_VERSION_PATCH);
-    pm_buffer_append_u8(buffer, PRISM_SERIALIZE_ONLY_SEMANTICS_FIELDS ? 1 : 0);
-
+    pm_serialize_header(buffer);
     pm_serialize_content(parser, node, buffer);
-    pm_buffer_append_str(buffer, "\0", 1);
+    pm_buffer_append_string(buffer, "\0", 1);
 }
 
 // Parse and serialize the AST represented by the given source to the given
@@ -15688,7 +15722,26 @@ pm_parse_serialize(const uint8_t *source, size_t size, pm_buffer_t *buffer, cons
     if (metadata) pm_parser_metadata(&parser, metadata);
 
     pm_node_t *node = pm_parse(&parser);
-    pm_serialize(&parser, node, buffer);
+
+    pm_serialize_header(buffer);
+    pm_serialize_content(&parser, node, buffer);
+    pm_buffer_append_byte(buffer, '\0');
+
+    pm_node_destroy(&parser, node);
+    pm_parser_free(&parser);
+}
+
+// Parse and serialize the comments in the given source to the given buffer.
+PRISM_EXPORTED_FUNCTION void
+pm_parse_serialize_comments(const uint8_t *source, size_t size, pm_buffer_t *buffer, const char *metadata) {
+    pm_parser_t parser;
+    pm_parser_init(&parser, source, size, NULL);
+    if (metadata) pm_parser_metadata(&parser, metadata);
+
+    pm_node_t *node = pm_parse(&parser);
+    pm_serialize_header(buffer);
+    pm_serialize_encoding(&parser.encoding, buffer);
+    pm_serialize_comment_list(&parser, &parser.comment_list, buffer);
 
     pm_node_destroy(&parser, node);
     pm_parser_free(&parser);