Add custom JavaScript regex engine (QuickJS libregexp port) (#2380)

Port of QuickJS libregexp to C# for patterns where .NET Regex diverges
from ECMAScript semantics. Three-tier architecture: direct .NET Regex
for simple patterns, Jint's converter for patterns needing rewriting,
and custom bytecode engine for ECMAScript-specific semantics.

97,638 test262 passed, 0 failures (+1,624 from baseline).

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: lahma

Directory.Packages.props

@@ -4,8 +4,8 @@
     <CentralPackageTransitivePinningEnabled>true</CentralPackageTransitivePinningEnabled>
   </PropertyGroup>
   <ItemGroup>
-    <PackageVersion Include="Acornima" Version="1.3.2" />
-    <PackageVersion Include="Acornima.Extras" Version="1.3.2" />
+    <PackageVersion Include="Acornima" Version="1.4.0" />
+    <PackageVersion Include="Acornima.Extras" Version="1.4.0" />
     <PackageVersion Include="BenchmarkDotNet" Version="0.15.8" />
     <PackageVersion Include="BenchmarkDotNet.TestAdapter" Version="0.13.12" />
     <PackageVersion Include="FluentAssertions" Version="[7.2.2]" />

Jint.Tests.Test262/Test262Harness.settings.json

@@ -7,8 +7,6 @@
   "ExcludedFeatures": [
     "import-defer",
     "import-text",
-    "regexp-unicode-property-escapes",
-    "regexp-v-flag",
     "source-phase-imports",
     "tail-call-optimization"
   ],
@@ -26,55 +24,6 @@
     // Currently quite impossible to detect if assignment target is CoverParenthesizedExpression
     "language/expressions/assignment/fn-name-lhs-cover.js",
 
-    // === REGEX EXCLUSIONS ===
-
-    // CharacterClassEscapes tests internally use v flag which is not supported
-    "built-ins/RegExp/CharacterClassEscapes/*.js",
-
-    // .NET regex engine: forward backreferences not supported in ECMAScript mode
-    "built-ins/RegExp/S15.10.2.11_A1_T5.js",
-    "built-ins/RegExp/S15.10.2.11_A1_T7.js",
-    "language/literals/regexp/named-groups/forward-reference.js",
-    "built-ins/RegExp/named-groups/non-unicode-references.js",
-    "built-ins/RegExp/named-groups/unicode-references.js",
-
-    // .NET regex engine: lookbehind backreference capture semantics differ from JS
-    "built-ins/RegExp/lookBehind/back-references-to-captures.js",
-    "built-ins/RegExp/lookBehind/mutual-recursive.js",
-
-    // .NET regex engine: nullable quantifier and lookahead capture group semantics differ
-    "built-ins/RegExp/nullable-quantifier.js",
-    "built-ins/RegExp/lookahead-quantifier-match-groups.js",
-
-    // .NET regex engine: scoped regexp modifiers don't fully match ECMAScript boundary and multiline semantics
-    "built-ins/RegExp/regexp-modifiers/add-ignoreCase-affects-slash-lower-b.js",
-    "built-ins/RegExp/regexp-modifiers/add-ignoreCase-affects-slash-lower-w.js",
-    "built-ins/RegExp/regexp-modifiers/add-ignoreCase-affects-slash-upper-b.js",
-    "built-ins/RegExp/regexp-modifiers/add-ignoreCase-affects-slash-upper-w.js",
-    "built-ins/RegExp/regexp-modifiers/add-multiline.js",
-    "built-ins/RegExp/regexp-modifiers/remove-multiline-does-not-affect-dotAll-flag.js",
-
-    // .NET regex engine: complex quantifier capture group tracking differs
-    "built-ins/RegExp/S15.10.2.5_A1_T4.js",
-    "built-ins/RegExp/prototype/exec/S15.10.6.2_A1_T6.js",
-
-    // .NET regex engine: Unicode case folding tables differ from ECMAScript CaseFolding.txt
-    "language/literals/regexp/u-case-mapping.js",
-    "built-ins/RegExp/unicode_full_case_folding.js",
-
-    // .NET regex engine: quantifier values exceeding int.MaxValue
-    "built-ins/RegExp/quantifier-integer-limit.js",
-
-    // .NET regex engine: quantified groups with alternation retain captures from previous iterations
-    // (JavaScript resets non-participating groups per iteration, .NET does not)
-    "built-ins/RegExp/named-groups/duplicate-names-exec.js",
-    "built-ins/RegExp/named-groups/duplicate-names-match.js",
-    "built-ins/RegExp/named-groups/duplicate-names-test.js",
-    "built-ins/RegExp/prototype/exec/duplicate-named-groups-properties.js",
-    "built-ins/RegExp/prototype/exec/duplicate-named-indices-groups-properties.js",
-    "built-ins/String/prototype/match/duplicate-named-groups-properties.js",
-    "built-ins/String/prototype/match/duplicate-named-indices-groups-properties.js",
-
     // requires investigation how to process complex function name evaluation for property
     "built-ins/Function/prototype/toString/method-computed-property-name.js",
 
@@ -355,13 +304,7 @@
     "intl402/Temporal/ZonedDateTime/prototype/with/leap-year-hebrew.js",
     "intl402/Temporal/ZonedDateTime/prototype/withCalendar/extreme-dates.js",
     "intl402/Temporal/ZonedDateTime/prototype/year/arithmetic-year.js",
-    "intl402/Temporal/ZonedDateTime/prototype/year/epoch-year.js",
-
-    // === ANNEX B EXCLUSIONS ===
-
-    // Acornima parser/RegExp limitation: malformed named groups in non-unicode mode
-    // Per B.1.2, non-unicode RegExp should accept some malformed named group syntax
-    "annexB/built-ins/RegExp/named-groups/non-unicode-malformed.js"
+    "intl402/Temporal/ZonedDateTime/prototype/year/epoch-year.js"
 
   ]
 }

Jint.Tests.Test262/Test262Test.cs

@@ -1,7 +1,6 @@
 #nullable enable
 
 using Jint.Native;
-using Jint.Native.Error;
 using Jint.Native.Object;
 using Jint.Runtime;
 using Jint.Runtime.Interop;

Jint.Tests/Runtime/ScriptModulePreparationTests.ScriptPreparation.cs

@@ -22,8 +22,12 @@ public void CanPreCompileRegex()
         var declaration = Assert.IsType<VariableDeclaration>(script.Program.Body[0]);
         var init = Assert.IsType<RegExpLiteral>(declaration.Declarations[0].Init);
 
-        init.Value.ToString().Should().Be("[cgt]");
-        (init.Value.Options & RegexOptions.Compiled).Should().Be(RegexOptions.Compiled);
+        // Regex is pre-compiled during preparation with Compiled flag
+        var regex = init.ParseResult.Regex;
+        regex.Should().NotBeNull();
+        (regex!.Options & RegexOptions.Compiled).Should().Be(RegexOptions.Compiled);
+
+        // Prepared script executes correctly
         new Engine().Evaluate(script).AsNumber().Should().Be(1);
     }
 

Jint/Engine.Ast.cs

@@ -29,6 +29,7 @@ public static Prepared<Script> PrepareScript(string code, string? source = null,
             Script preparedScript;
             var sourceOffset = options.ParsingOptions.SourceOffset;
             var padding = AcornimaExtensions.CreateSourceOffsetPadding(sourceOffset);
+
             if (padding.Length > 0)
             {
                 var paddedCode = padding + code;
@@ -38,6 +39,7 @@ public static Prepared<Script> PrepareScript(string code, string? source = null,
             {
                 preparedScript = parser.ParseScript(code, source, strict);
             }
+
             return new Prepared<Script>(preparedScript, parserOptions);
         }
         catch (Exception e)
@@ -64,6 +66,7 @@ public static Prepared<Module> PrepareModule(string code, string? source = null,
         try
         {
             var preparedModule = parser.ParseModule(code, source);
+
             return new Prepared<Module>(preparedModule, parserOptions);
         }
         catch (Exception e)

Jint/Engine.Defaults.cs

@@ -1,17 +1,93 @@
+using Jint.Native.RegExp;
+using Jint.Runtime.RegExp;
+
 namespace Jint;
 
 public partial class Engine
 {
     internal const bool FoldConstantsOnPrepareByDefault = true;
 
+    private static readonly TimeSpan DefaultPrepareRegexTimeout = TimeSpan.FromSeconds(10);
+
+    /// <summary>
+    /// OnRegExp callback that validates regex syntax at parse time without converting.
+    /// Conversion is done at runtime in RegExpInitialize with fallback to the custom engine.
+    /// </summary>
+    internal static readonly OnRegExpHandler ValidateRegExpHandler = static (in RegExpParsingContext ctx) =>
+    {
+        ctx.Validate();
+        return RegExpParseResult.ForSuccess();
+    };
+
+    /// <summary>
+    /// OnRegExp callback that validates and converts regex at parse time (interpreted .NET Regex).
+    /// Used during script/module preparation when CompileRegex is false.
+    /// </summary>
+    internal static readonly OnRegExpHandler ConvertRegExpHandler = static (in RegExpParsingContext ctx) =>
+    {
+        ctx.Validate();
+        // Skip pre-compilation for patterns that need the custom QuickJS engine at runtime
+        if (!RegExpConstructor.NeedCustomEngine(ctx.Pattern, ctx.Flags)
+            && JsRegExpConverter.TryConvert(ctx.Pattern, ctx.Flags, DefaultPrepareRegexTimeout, out var regex, out var groupCount))
+        {
+            return RegExpParseResult.ForSuccess(regex, groupCount);
+        }
+
+        return RegExpParseResult.ForSuccess();
+    };
+
+    /// <summary>
+    /// OnRegExp callback that validates and converts regex at parse time (compiled .NET Regex).
+    /// Used during script/module preparation when CompileRegex is true (default for preparation).
+    /// </summary>
+    internal static readonly OnRegExpHandler CompileRegExpHandler = static (in RegExpParsingContext ctx) =>
+    {
+        ctx.Validate();
+        // Skip pre-compilation for patterns that need the custom QuickJS engine at runtime
+        if (!RegExpConstructor.NeedCustomEngine(ctx.Pattern, ctx.Flags)
+            && JsRegExpConverter.TryConvert(ctx.Pattern, ctx.Flags, DefaultPrepareRegexTimeout, out var regex, out var groupCount, compiled: true))
+        {
+            return RegExpParseResult.ForSuccess(regex, groupCount);
+        }
+
+        return RegExpParseResult.ForSuccess();
+    };
+
+    /// <summary>
+    /// Creates an OnRegExp handler with a caller-specified timeout (interpreted .NET Regex).
+    /// </summary>
+    internal static OnRegExpHandler CreateConvertRegExpHandler(TimeSpan timeout) => (in RegExpParsingContext ctx) =>
+    {
+        ctx.Validate();
+        if (!RegExpConstructor.NeedCustomEngine(ctx.Pattern, ctx.Flags)
+            && JsRegExpConverter.TryConvert(ctx.Pattern, ctx.Flags, timeout, out var regex, out var groupCount))
+        {
+            return RegExpParseResult.ForSuccess(regex, groupCount);
+        }
+
+        return RegExpParseResult.ForSuccess();
+    };
+
+    /// <summary>
+    /// Creates an OnRegExp handler with a caller-specified timeout (compiled .NET Regex).
+    /// </summary>
+    internal static OnRegExpHandler CreateCompileRegExpHandler(TimeSpan timeout) => (in RegExpParsingContext ctx) =>
+    {
+        ctx.Validate();
+        if (!RegExpConstructor.NeedCustomEngine(ctx.Pattern, ctx.Flags)
+            && JsRegExpConverter.TryConvert(ctx.Pattern, ctx.Flags, timeout, out var regex, out var groupCount, compiled: true))
+        {
+            return RegExpParseResult.ForSuccess(regex, groupCount);
+        }
+
+        return RegExpParseResult.ForSuccess();
+    };
+
     internal static readonly ParserOptions BaseParserOptions = ParserOptions.Default with
     {
-        EcmaVersion = EcmaVersion.ES2023,
-        ExperimentalESFeatures = ExperimentalESFeatures.ImportAttributes
-                                 | ExperimentalESFeatures.RegExpDuplicateNamedCapturingGroups
-                                 | ExperimentalESFeatures.RegExpModifiers
-                                 | ExperimentalESFeatures.ExplicitResourceManagement
-                                 | ExperimentalESFeatures.Decorators,
+        EcmaVersion = EcmaVersion.ES2026,
+        ExperimentalESFeatures = ExperimentalESFeatures.Decorators,
+        OnRegExp = ValidateRegExpHandler,
         Tolerant = false,
     };
 }

Jint/Native/Iterator/IteratorInstance.cs

@@ -227,7 +227,7 @@ public override bool TryIteratorStep(out ObjectInstance nextItem)
                 if (macthStr == "")
                 {
                     var thisIndex = TypeConverter.ToLength(_iteratingRegExp.Get(JsRegExp.PropertyLastIndex));
-                    var nextIndex = thisIndex + 1;
+                    var nextIndex = RegExpPrototype.AdvanceStringIndex(_s, thisIndex, _unicode);
                     _iteratingRegExp.Set(JsRegExp.PropertyLastIndex, nextIndex, true);
                 }
             }

Jint/Native/JsRegExp.cs

@@ -3,6 +3,7 @@
 using Jint.Native.RegExp;
 using Jint.Runtime;
 using Jint.Runtime.Descriptors;
+using Jint.Runtime.RegExp;
 
 namespace Jint.Native;
 
@@ -24,6 +25,12 @@ public JsRegExp(Engine engine)
     public Regex Value { get; set; } = null!;
     public string Source { get; set; }
 
+    /// <summary>
+    /// Custom regex engine used when .NET Regex cannot handle the pattern.
+    /// When set, this takes priority over <see cref="Value"/>.
+    /// </summary>
+    internal JintRegExpEngine? CustomEngine { get; set; }
+
     public string Flags
     {
         get => _flags;
@@ -37,6 +44,7 @@ public string Flags
             IgnoreCase = false;
             Multiline = false;
             Sticky = false;
+            Unicode = false;
             FullUnicode = false;
             UnicodeSets = false;
             foreach (var c in _flags)
@@ -62,29 +70,51 @@ public string Flags
                         Sticky = true;
                         break;
                     case 'u':
+                        Unicode = true;
                         FullUnicode = true;
                         break;
                     case 'v':
                         UnicodeSets = true;
+                        FullUnicode = true; // v-flag implies unicode semantics
                         break;
                 }
             }
         }
     }
 
-    public RegExpParseResult ParseResult { get; set; }
+    /// <summary>
+    /// JS group count (including group 0 for full match) from Jint's converter.
+    /// 0 means not set — fall back to .NET Regex's own group count.
+    /// </summary>
+    internal int ConvertedGroupCount { get; set; }
+
+    /// <summary>
+    /// Provides backward-compatible access to the regex parse result.
+    /// </summary>
+    [Obsolete("The ParseResult property is no longer populated. Use Value to access the .NET Regex instance directly.")]
+    public RegExpParseResult ParseResult
+    {
+        get => UsesDotNetEngine && Value is not null ? RegExpParseResult.ForSuccess(Value) : default;
+        set { /* no-op for backward compatibility */ }
+    }
 
     public bool DotAll { get; private set; }
     public bool Global { get; private set; }
     public bool Indices { get; private set; }
     public bool IgnoreCase { get; private set; }
     public bool Multiline { get; private set; }
     public bool Sticky { get; private set; }
+    /// <summary>Whether the 'u' flag was explicitly set (for the unicode accessor).</summary>
+    public bool Unicode { get; private set; }
+    /// <summary>Whether unicode semantics apply (true for both 'u' and 'v' flags).</summary>
     public bool FullUnicode { get; private set; }
     public bool UnicodeSets { get; private set; }
 
     internal bool HasDefaultRegExpExec => Properties == null && Prototype is RegExpPrototype { HasDefaultExec: true };
 
+    /// <summary>Whether this regex uses the .NET Regex engine (not the custom engine).</summary>
+    internal bool UsesDotNetEngine => CustomEngine is null;
+
     public override PropertyDescriptor GetOwnProperty(JsValue property)
     {
         if (PropertyLastIndex.Equals(property))