LibJS: Parse RegExp literals at AST creation time, not execution time

The spec requires that invalid RegExp literals must cause a Syntax Error
before the JavaScript is executed. See:
https://tc39.es/ecma262/#sec-patterns-static-semantics-early-errors

This is explicitly tested in the RegExp/property-escapes test262 tests.
For example, see unsupported-property-Line_Break.js:

    $DONOTEVALUATE();
    /\p{Line_Break}/u;

That RegExp literal is invalid because Line_Break is not a supported
Unicode property. $DONOTEVALUATE() just throws an exception when it is
executed. The test expects that this file will fail to be parsed.

Note that RegExp patterns can still be parsed at execution time by way
of "new RegExp(...)".

Userland/Libraries/LibJS/AST.cpp

@@ -2020,7 +2020,9 @@ void RegExpLiteral::dump(int indent) const
 Value RegExpLiteral::execute(Interpreter& interpreter, GlobalObject& global_object) const
 {
     InterpreterNodeScope node_scope { interpreter, *this };
-    return regexp_create(global_object, js_string(interpreter.heap(), pattern()), js_string(interpreter.heap(), flags()));
+
+    Regex<ECMA262> regex(parsed_regex(), parsed_pattern(), parsed_flags());
+    return RegExpObject::create(global_object, move(regex), pattern(), flags());
 }
 
 void ArrayExpression::dump(int indent) const

Userland/Libraries/LibJS/AST.h

@@ -19,6 +19,7 @@
 #include <LibJS/Runtime/PropertyName.h>
 #include <LibJS/Runtime/Value.h>
 #include <LibJS/SourceRange.h>
+#include <LibRegex/Regex.h>
 
 namespace JS {
 
@@ -758,8 +759,11 @@ public:
 
 class RegExpLiteral final : public Literal {
 public:
-    explicit RegExpLiteral(SourceRange source_range, String pattern, String flags)
+    RegExpLiteral(SourceRange source_range, regex::Parser::Result parsed_regex, String parsed_pattern, regex::RegexOptions<ECMAScriptFlags> parsed_flags, String pattern, String flags)
         : Literal(source_range)
+        , m_parsed_regex(move(parsed_regex))
+        , m_parsed_pattern(move(parsed_pattern))
+        , m_parsed_flags(move(parsed_flags))
         , m_pattern(move(pattern))
         , m_flags(move(flags))
     {
@@ -769,10 +773,16 @@ public:
     virtual void dump(int indent) const override;
     virtual void generate_bytecode(Bytecode::Generator&) const override;
 
+    regex::Parser::Result const& parsed_regex() const { return m_parsed_regex; }
+    String const& parsed_pattern() const { return m_parsed_pattern; }
+    regex::RegexOptions<ECMAScriptFlags> const& parsed_flags() const { return m_parsed_flags; }
     String const& pattern() const { return m_pattern; }
     String const& flags() const { return m_flags; }
 
 private:
+    regex::Parser::Result m_parsed_regex;
+    String m_parsed_pattern;
+    regex::RegexOptions<ECMAScriptFlags> m_parsed_flags;
     String m_pattern;
     String m_flags;
 };

Userland/Libraries/LibJS/Parser.cpp

@@ -12,6 +12,8 @@
 #include <AK/ScopeGuard.h>
 #include <AK/StdLibExtras.h>
 #include <AK/TemporaryChange.h>
+#include <LibJS/Runtime/RegExpObject.h>
+#include <LibRegex/Regex.h>
 
 namespace JS {
 
@@ -848,21 +850,29 @@ NonnullRefPtr<RegExpLiteral> Parser::parse_regexp_literal()
     auto pattern = consume().value();
     // Remove leading and trailing slash.
     pattern = pattern.substring_view(1, pattern.length() - 2);
+
     auto flags = String::empty();
+    auto parsed_flags = RegExpObject::default_flags;
+
     if (match(TokenType::RegexFlags)) {
         auto flags_start = position();
         flags = consume().value();
-        HashTable<char> seen_flags;
-        for (size_t i = 0; i < flags.length(); ++i) {
-            auto flag = flags.substring_view(i, 1);
-            if (!flag.is_one_of("d", "g", "i", "m", "s", "u", "y"))
-                syntax_error(String::formatted("Invalid RegExp flag '{}'", flag), Position { flags_start.line, flags_start.column + i });
-            if (seen_flags.contains(*flag.characters_without_null_termination()))
-                syntax_error(String::formatted("Repeated RegExp flag '{}'", flag), Position { flags_start.line, flags_start.column + i });
-            seen_flags.set(*flag.characters_without_null_termination());
-        }
+
+        auto parsed_flags_or_error = regex_flags_from_string(flags);
+        if (parsed_flags_or_error.is_error())
+            syntax_error(parsed_flags_or_error.release_error(), flags_start);
+        else
+            parsed_flags = parsed_flags_or_error.release_value();
     }
-    return create_ast_node<RegExpLiteral>({ m_state.current_token.filename(), rule_start.position(), position() }, pattern, flags);
+
+    auto parsed_pattern = parse_regex_pattern(pattern, parsed_flags.has_flag_set(ECMAScriptFlags::Unicode));
+    auto parsed_regex = Regex<ECMA262>::parse_pattern(parsed_pattern, parsed_flags);
+
+    if (parsed_regex.error != regex::Error::NoError)
+        syntax_error(String::formatted("RegExp compile error: {}", Regex<ECMA262>(parsed_regex, parsed_pattern, parsed_flags).error_string()), rule_start.position());
+
+    SourceRange range { m_state.current_token.filename(), rule_start.position(), position() };
+    return create_ast_node<RegExpLiteral>(move(range), move(parsed_regex), move(parsed_pattern), move(parsed_flags), pattern.to_string(), move(flags));
 }
 
 NonnullRefPtr<Expression> Parser::parse_unary_prefixed_expression()

Userland/Libraries/LibJS/Runtime/RegExpObject.cpp

@@ -14,97 +14,108 @@
 
 namespace JS {
 
-static Flags options_from(GlobalObject& global_object, const String& flags)
+Result<regex::RegexOptions<ECMAScriptFlags>, String> regex_flags_from_string(StringView flags)
 {
-    auto& vm = global_object.vm();
     bool d = false, g = false, i = false, m = false, s = false, u = false, y = false;
-    Flags options {
-        // JS regexps are all 'global' by default as per our definition, but the "global" flag enables "stateful".
-        // FIXME: Enable 'BrowserExtended' only if in a browser context.
-        .effective_flags = { (regex::ECMAScriptFlags)regex::AllFlags::Global | (regex::ECMAScriptFlags)regex::AllFlags::SkipTrimEmptyMatches | regex::ECMAScriptFlags::BrowserExtended },
-        .declared_flags = {},
-    };
+    auto options = RegExpObject::default_flags;
 
     for (auto ch : flags) {
         switch (ch) {
         case 'd':
             if (d)
-                vm.throw_exception<SyntaxError>(global_object, ErrorType::RegExpObjectRepeatedFlag, ch);
+                return String::formatted(ErrorType::RegExpObjectRepeatedFlag.message(), ch);
             d = true;
             break;
         case 'g':
             if (g)
-                vm.throw_exception<SyntaxError>(global_object, ErrorType::RegExpObjectRepeatedFlag, ch);
+                return String::formatted(ErrorType::RegExpObjectRepeatedFlag.message(), ch);
             g = true;
-            options.effective_flags |= regex::ECMAScriptFlags::Global;
-            options.declared_flags |= regex::ECMAScriptFlags::Global;
+            options |= regex::ECMAScriptFlags::Global;
             break;
         case 'i':
             if (i)
-                vm.throw_exception<SyntaxError>(global_object, ErrorType::RegExpObjectRepeatedFlag, ch);
+                return String::formatted(ErrorType::RegExpObjectRepeatedFlag.message(), ch);
             i = true;
-            options.effective_flags |= regex::ECMAScriptFlags::Insensitive;
-            options.declared_flags |= regex::ECMAScriptFlags::Insensitive;
+            options |= regex::ECMAScriptFlags::Insensitive;
             break;
         case 'm':
             if (m)
-                vm.throw_exception<SyntaxError>(global_object, ErrorType::RegExpObjectRepeatedFlag, ch);
+                return String::formatted(ErrorType::RegExpObjectRepeatedFlag.message(), ch);
             m = true;
-            options.effective_flags |= regex::ECMAScriptFlags::Multiline;
-            options.declared_flags |= regex::ECMAScriptFlags::Multiline;
+            options |= regex::ECMAScriptFlags::Multiline;
             break;
         case 's':
             if (s)
-                vm.throw_exception<SyntaxError>(global_object, ErrorType::RegExpObjectRepeatedFlag, ch);
+                return String::formatted(ErrorType::RegExpObjectRepeatedFlag.message(), ch);
             s = true;
-            options.effective_flags |= regex::ECMAScriptFlags::SingleLine;
-            options.declared_flags |= regex::ECMAScriptFlags::SingleLine;
+            options |= regex::ECMAScriptFlags::SingleLine;
             break;
         case 'u':
             if (u)
-                vm.throw_exception<SyntaxError>(global_object, ErrorType::RegExpObjectRepeatedFlag, ch);
+                return String::formatted(ErrorType::RegExpObjectRepeatedFlag.message(), ch);
             u = true;
-            options.effective_flags |= regex::ECMAScriptFlags::Unicode;
-            options.declared_flags |= regex::ECMAScriptFlags::Unicode;
+            options |= regex::ECMAScriptFlags::Unicode;
             break;
         case 'y':
             if (y)
-                vm.throw_exception<SyntaxError>(global_object, ErrorType::RegExpObjectRepeatedFlag, ch);
+                return String::formatted(ErrorType::RegExpObjectRepeatedFlag.message(), ch);
             y = true;
             // Now for the more interesting flag, 'sticky' actually unsets 'global', part of which is the default.
-            options.effective_flags.reset_flag(regex::ECMAScriptFlags::Global);
+            options.reset_flag(regex::ECMAScriptFlags::Global);
             // "What's the difference between sticky and global, then", that's simple.
             // all the other flags imply 'global', and the "global" flag implies 'stateful';
             // however, the "sticky" flag does *not* imply 'global', only 'stateful'.
-            options.effective_flags |= (regex::ECMAScriptFlags)regex::AllFlags::Internal_Stateful;
-            options.effective_flags |= regex::ECMAScriptFlags::Sticky;
-            options.declared_flags |= regex::ECMAScriptFlags::Sticky;
+            options |= (regex::ECMAScriptFlags)regex::AllFlags::Internal_Stateful;
+            options |= regex::ECMAScriptFlags::Sticky;
             break;
         default:
-            vm.throw_exception<SyntaxError>(global_object, ErrorType::RegExpObjectBadFlag, ch);
-            return options;
+            return String::formatted(ErrorType::RegExpObjectBadFlag.message(), ch);
         }
     }
 
     return options;
 }
 
-RegExpObject* RegExpObject::create(GlobalObject& global_object, String original_pattern, String parsed_pattern, String flags)
+String parse_regex_pattern(StringView pattern, bool unicode)
 {
-    return global_object.heap().allocate<RegExpObject>(global_object, move(original_pattern), move(parsed_pattern), move(flags), *global_object.regexp_prototype());
+    auto utf16_pattern = AK::utf8_to_utf16(pattern);
+    Utf16View utf16_pattern_view { utf16_pattern };
+    StringBuilder builder;
+
+    // If the Unicode flag is set, append each code point to the pattern. Otherwise, append each
+    // code unit. But unlike the spec, multi-byte code units must be escaped for LibRegex to parse.
+    for (size_t i = 0; i < utf16_pattern_view.length_in_code_units();) {
+        if (unicode) {
+            auto code_point = code_point_at(utf16_pattern_view, i);
+            builder.append_code_point(code_point.code_point);
+            i += code_point.code_unit_count;
+            continue;
+        }
+
+        u16 code_unit = utf16_pattern_view.code_unit_at(i);
+        ++i;
+
+        if (code_unit > 0x7f)
+            builder.appendff("\\u{:04x}", code_unit);
+        else
+            builder.append_code_point(code_unit);
+    }
+
+    return builder.build();
 }
 
-RegExpObject::RegExpObject(String original_pattern, String parsed_pattern, String flags, Object& prototype)
-    : Object(prototype)
-    , m_original_pattern(move(original_pattern))
-    , m_parsed_pattern(move(parsed_pattern))
-    , m_flags(move(flags))
-    , m_active_flags(options_from(global_object(), m_flags))
-    , m_regex(m_parsed_pattern, m_active_flags.effective_flags)
+RegExpObject* RegExpObject::create(GlobalObject& global_object, Regex<ECMA262> regex, String pattern, String flags)
 {
-    if (m_regex.parser_result.error != regex::Error::NoError) {
-        vm().throw_exception<SyntaxError>(global_object(), ErrorType::RegExpCompileError, m_regex.error_string());
-    }
+    return global_object.heap().allocate<RegExpObject>(global_object, move(regex), move(pattern), move(flags), *global_object.regexp_prototype());
+}
+
+RegExpObject::RegExpObject(Regex<ECMA262> regex, String pattern, String flags, Object& prototype)
+    : Object(prototype)
+    , m_pattern(move(pattern))
+    , m_flags(move(flags))
+    , m_regex(move(regex))
+{
+    VERIFY(m_regex.parser_result.error == regex::Error::NoError);
 }
 
 RegExpObject::~RegExpObject()
@@ -115,7 +126,7 @@ void RegExpObject::initialize(GlobalObject& global_object)
 {
     auto& vm = this->vm();
     Object::initialize(global_object);
-    define_direct_property(vm.names.lastIndex, {}, Attribute::Writable);
+    define_direct_property(vm.names.lastIndex, Value(0), Attribute::Writable);
 }
 
 // 22.2.3.2.4 RegExpCreate ( P, F ), https://tc39.es/ecma262/#sec-regexpcreate
@@ -139,38 +150,27 @@ RegExpObject* regexp_create(GlobalObject& global_object, Value pattern, Value fl
         original_pattern = String::empty();
         parsed_pattern = String::empty();
     } else {
-        auto utf16_pattern = pattern.to_utf16_string(global_object);
+        original_pattern = pattern.to_string(global_object);
         if (vm.exception())
             return {};
 
-        Utf16View utf16_pattern_view { utf16_pattern };
         bool unicode = f.find('u').has_value();
-        StringBuilder builder;
-
-        // If the Unicode flag is set, append each code point to the pattern. Otherwise, append each
-        // code unit. But unlike the spec, multi-byte code units must be escaped for LibRegex to parse.
-        for (size_t i = 0; i < utf16_pattern_view.length_in_code_units();) {
-            if (unicode) {
-                auto code_point = code_point_at(utf16_pattern_view, i);
-                builder.append_code_point(code_point.code_point);
-                i += code_point.code_unit_count;
-                continue;
-            }
-
-            u16 code_unit = utf16_pattern_view.code_unit_at(i);
-            ++i;
-
-            if (code_unit > 0x7f)
-                builder.appendff("\\u{:04x}", code_unit);
-            else
-                builder.append_code_point(code_unit);
-        }
-
-        original_pattern = utf16_pattern_view.to_utf8(Utf16View::AllowInvalidCodeUnits::Yes);
-        parsed_pattern = builder.build();
+        parsed_pattern = parse_regex_pattern(original_pattern, unicode);
     }
 
-    auto* object = RegExpObject::create(global_object, move(original_pattern), move(parsed_pattern), move(f));
+    auto parsed_flags_or_error = regex_flags_from_string(f);
+    if (parsed_flags_or_error.is_error()) {
+        vm.throw_exception(global_object, SyntaxError::create(global_object, parsed_flags_or_error.release_error()));
+        return {};
+    }
+
+    Regex<ECMA262> regex(move(parsed_pattern), parsed_flags_or_error.release_value());
+    if (regex.parser_result.error != regex::Error::NoError) {
+        vm.throw_exception<SyntaxError>(global_object, ErrorType::RegExpCompileError, regex.error_string());
+        return {};
+    }
+
+    auto* object = RegExpObject::create(global_object, move(regex), move(original_pattern), move(f));
     object->set(vm.names.lastIndex, Value(0), Object::ShouldThrowExceptions::Yes);
     if (vm.exception())
         return {};

Userland/Libraries/LibJS/Runtime/RegExpObject.h

@@ -6,40 +6,40 @@
 
 #pragma once
 
+#include <AK/Result.h>
 #include <LibJS/AST.h>
 #include <LibJS/Runtime/Object.h>
 #include <LibRegex/Regex.h>
 
-struct Flags {
-    regex::RegexOptions<ECMAScriptFlags> effective_flags;
-    regex::RegexOptions<ECMAScriptFlags> declared_flags;
-};
-
 namespace JS {
 
 RegExpObject* regexp_create(GlobalObject&, Value pattern, Value flags);
 
+Result<regex::RegexOptions<ECMAScriptFlags>, String> regex_flags_from_string(StringView flags);
+String parse_regex_pattern(StringView pattern, bool unicode);
+
 class RegExpObject : public Object {
     JS_OBJECT(RegExpObject, Object);
 
 public:
-    static RegExpObject* create(GlobalObject&, String original_pattern, String parsed_pattern, String flags);
+    // JS regexps are all 'global' by default as per our definition, but the "global" flag enables "stateful".
+    // FIXME: Enable 'BrowserExtended' only if in a browser context.
+    static constexpr regex::RegexOptions<ECMAScriptFlags> default_flags { (regex::ECMAScriptFlags)regex::AllFlags::Global | (regex::ECMAScriptFlags)regex::AllFlags::SkipTrimEmptyMatches | regex::ECMAScriptFlags::BrowserExtended };
 
-    RegExpObject(String original_pattern, String parsed_pattern, String flags, Object& prototype);
+    static RegExpObject* create(GlobalObject&, Regex<ECMA262> regex, String pattern, String flags);
+
+    RegExpObject(Regex<ECMA262> regex, String pattern, String flags, Object& prototype);
     virtual void initialize(GlobalObject&) override;
     virtual ~RegExpObject() override;
 
-    const String& pattern() const { return m_original_pattern; }
+    const String& pattern() const { return m_pattern; }
     const String& flags() const { return m_flags; }
-    const regex::RegexOptions<ECMAScriptFlags>& declared_options() { return m_active_flags.declared_flags; }
     const Regex<ECMA262>& regex() { return m_regex; }
     const Regex<ECMA262>& regex() const { return m_regex; }
 
 private:
-    String m_original_pattern;
-    String m_parsed_pattern;
+    String m_pattern;
     String m_flags;
-    Flags m_active_flags;
     Regex<ECMA262> m_regex;
 };
 

Userland/Libraries/LibJS/Tests/builtins/RegExp/RegExp.js

@@ -52,3 +52,11 @@ test("regexp object as pattern parameter", () => {
     expect(RegExp(regex_like_object_with_flags, "").toString()).toBe("/foo/");
     expect(RegExp(regex_like_object_with_flags, "y").toString()).toBe("/foo/y");
 });
+
+test("regexp literals are re-useable", () => {
+    for (var i = 0; i < 2; ++i) {
+        const re = /test/;
+        expect(re.test("te")).toBeFalse();
+        expect(re.test("test")).toBeTrue();
+    }
+});