diff --git a/StgSharp.RegularAnalysis/TextRegex/RegexAstNode.cs b/StgSharp.RegularAnalysis/Text/RegexAstNode.cs
similarity index 99%
rename from StgSharp.RegularAnalysis/TextRegex/RegexAstNode.cs
rename to StgSharp.RegularAnalysis/Text/RegexAstNode.cs
index 9b6de712..b5238042 100644
--- a/StgSharp.RegularAnalysis/TextRegex/RegexAstNode.cs
+++ b/StgSharp.RegularAnalysis/Text/RegexAstNode.cs
@@ -28,7 +28,7 @@
 using StgSharp.RegularAnalysis.Abstraction;
 using System.Runtime.CompilerServices;
 
-namespace StgSharp.RegularAnalysis.TextRegex
+namespace StgSharp.RegularAnalysis.Text
 {
     internal class RegexAstNode : ISyntaxNode<RegexAstNode, RegexElementLabel>
     {
diff --git a/StgSharp.RegularAnalysis/TextRegex/RegexElementLabel.cs b/StgSharp.RegularAnalysis/Text/RegexElementLabel.cs
similarity index 97%
rename from StgSharp.RegularAnalysis/TextRegex/RegexElementLabel.cs
rename to StgSharp.RegularAnalysis/Text/RegexElementLabel.cs
index dec57691..96e3b61b 100644
--- a/StgSharp.RegularAnalysis/TextRegex/RegexElementLabel.cs
+++ b/StgSharp.RegularAnalysis/Text/RegexElementLabel.cs
@@ -31,7 +31,7 @@ using System.Linq;
 using System.Text;
 using System.Threading.Tasks;
 
-namespace StgSharp.RegularAnalysis.TextRegex
+namespace StgSharp.RegularAnalysis.Text
 {
     [Flags]
     public enum RegexElementLabel
diff --git a/StgSharp.RegularAnalysis/TextRegex/RegexIR.cs b/StgSharp.RegularAnalysis/Text/RegexHIR.cs
similarity index 81%
rename from StgSharp.RegularAnalysis/TextRegex/RegexIR.cs
rename to StgSharp.RegularAnalysis/Text/RegexHIR.cs
index 46498c96..11e44c35 100644
--- a/StgSharp.RegularAnalysis/TextRegex/RegexIR.cs
+++ b/StgSharp.RegularAnalysis/Text/RegexHIR.cs
@@ -32,25 +32,24 @@ using System.Runtime.CompilerServices;
 using System.Text;
 using System.Threading.Tasks;
 
-namespace StgSharp.RegularAnalysis.TextRegex
+namespace StgSharp.RegularAnalysis.Text
 {
     internal enum RegexIRCommand
     {
 
-        Search_Seq = 0,
-        Search_Set = 1,
-        SearchChar = 2,
-        ConfigFail = 3,
-        Condition = 4,
-        Try = 5,
-        Count = 6,
-        Count_Seq = 7,
-        Count_Set = 8,
-        Pop = 10
+        Match_Seq = 0,
+        Match_Set = 1,
+        Condition = 3,
+        Try = 4,
+        Count = 5,
+        Count_Seq = 6,
+        Count_Set = 7,
+        Pop = 8,
+        Pack = 9,
 
     }
 
-    internal abstract class RegexIR
+    internal abstract class RegexHIR
     {
 
         internal abstract RegexIRCommand Command { get; }
@@ -64,70 +63,46 @@ namespace StgSharp.RegularAnalysis.TextRegex
     /// <summary>
     ///   Searches for a certain string pattern or charset from current position.
     /// </summary>
-    internal sealed class RegexMatchSeqIR(
+    internal sealed class RegexMatchSeqHIR(
                           string pattern
-    ) : RegexIR
+    ) : RegexHIR
     {
 
         public string Pattern { get; } = pattern;
 
-        internal override RegexIRCommand Command => RegexIRCommand.Search_Seq;
+        internal override RegexIRCommand Command => RegexIRCommand.Match_Seq;
 
         internal override string AsCommand(
                                  int lineNum
         )
         {
-            return $"[{lineNum}]\t SEARCH_SEQ \"{(string.IsNullOrEmpty(Pattern) ? "<EPSILON>" : Pattern)}\"";
+            return $"[{lineNum}]\t MATCH_SEQ \"{(string.IsNullOrEmpty(Pattern) ? "<EPSILON>" : Pattern)}\"";
         }
 
     }
 
-    internal sealed class RegexMatchSetIR(
+    internal sealed class RegexMatchSetHIR(
                           string charset
-    ) : RegexIR
+    ) : RegexHIR
     {
 
         public string Charset { get; } = charset;
 
-        internal override RegexIRCommand Command => RegexIRCommand.Search_Set;
+        internal override RegexIRCommand Command => RegexIRCommand.Match_Set;
 
         internal override string AsCommand(
                                  int lineNum
         )
         {
-            return $"[{lineNum}]\t SEARCH_SET \"{Charset}\"";
+            return $"[{lineNum}]\t MATCH_SET \"{Charset}\"";
         }
 
     }
 
-    /// <summary>
-    ///   Configures the failure jump of current position. The failMod ranges 0-2
-    /// </summary>
-    /// <param name="target">
-    ///
-    /// </param>
-    internal sealed class RegexConfigFailIR(
-                          int failMod
-    ) : RegexIR
-    {
-
-        public int FailMod { get; } = failMod & 0x3;
-
-        internal override RegexIRCommand Command => RegexIRCommand.ConfigFail;
-
-        internal override string AsCommand(
-                                 int lineNum
-        )
-        {
-            return $"[{lineNum}]\t CONFIG_FAIL {FailMod}";
-        }
-
-    }
-
-    internal sealed class RegexConditionIR(
+    internal sealed class RegexConditionHIR(
                           int success,
                           int fail
-    ) : RegexIR
+    ) : RegexHIR
     {
 
         public int SuccessCase { get; } = success;
@@ -145,12 +120,33 @@ namespace StgSharp.RegularAnalysis.TextRegex
 
     }
 
-    internal sealed class RegexCountIR(
+    internal sealed class RegexPackHIR(
+                          bool condition,
+                          string name = ""
+    ) : RegexHIR
+    {
+
+        public bool Condition { get; } = condition;
+
+        public string GroupName { get; } = name;
+
+        internal override RegexIRCommand Command => RegexIRCommand.Pack;
+
+        internal override string AsCommand(
+                                 int lineNum
+        )
+        {
+            return $"[{lineNum}]\t PACK {(Condition ? "SUCCESS" : "FAIL")} GROUP: \"{GroupName}\"";
+        }
+
+    }
+
+    internal sealed class RegexCountHIR(
                           int min,
                           int max,
                           bool isGreedy,
                           int operation
-    ) : RegexIR
+    ) : RegexHIR
     {
 
         public bool IsGreedy { get; } = isGreedy;
@@ -172,12 +168,12 @@ namespace StgSharp.RegularAnalysis.TextRegex
 
     }
 
-    internal sealed class RegexCountSeqIR(
+    internal sealed class RegexCountSeqHIR(
                           int min,
                           int max,
                           bool isGreedy,
                           string seq
-    ) : RegexIR
+    ) : RegexHIR
     {
 
         public bool IsGreedy { get; } = isGreedy;
@@ -199,12 +195,12 @@ namespace StgSharp.RegularAnalysis.TextRegex
 
     }
 
-    internal sealed class RegexCountSetIR(
+    internal sealed class RegexCountSetHIR(
                           int min,
                           int max,
                           bool isGreedy,
                           string set
-    ) : RegexIR
+    ) : RegexHIR
     {
 
         public bool IsGreedy { get; } = isGreedy;
@@ -226,9 +222,9 @@ namespace StgSharp.RegularAnalysis.TextRegex
 
     }
 
-    internal sealed class RegexTryIR(
+    internal sealed class RegexTryHIR(
                           int count
-    ) : RegexIR
+    ) : RegexHIR
     {
 
         public int CountOfCommand { get; } = count;
@@ -244,14 +240,14 @@ namespace StgSharp.RegularAnalysis.TextRegex
 
     }
 
-    internal sealed class RegexPopIR(
+    internal sealed class RegexPopHIR(
                           bool isSuccess
-    ) : RegexIR
+    ) : RegexHIR
     {
 
         public bool IsSuccess { get; } = isSuccess;
 
-        internal override RegexIRCommand Command => RegexIRCommand.Try;
+        internal override RegexIRCommand Command => RegexIRCommand.Pop;
 
         internal override string AsCommand(
                                  int lineNum
diff --git a/StgSharp.RegularAnalysis/TextRegex/RegexIRGenerator.cs b/StgSharp.RegularAnalysis/Text/RegexHIRGenerator.cs
similarity index 79%
rename from StgSharp.RegularAnalysis/TextRegex/RegexIRGenerator.cs
rename to StgSharp.RegularAnalysis/Text/RegexHIRGenerator.cs
index e5afb6ee..b1720138 100644
--- a/StgSharp.RegularAnalysis/TextRegex/RegexIRGenerator.cs
+++ b/StgSharp.RegularAnalysis/Text/RegexHIRGenerator.cs
@@ -1,6 +1,6 @@
 //-----------------------------------------------------------------------
 // -----------------------------------------------------------------------
-// file="RegexIRGenerator"
+// file="RegexHIRGenerator"
 // Project: StgSharp
 // AuthorGroup: Nitload
 // Copyright (c) Nitload. All rights reserved.
@@ -32,15 +32,15 @@ using System.Runtime.CompilerServices;
 using System.Text;
 using System.Threading.Tasks;
 
-namespace StgSharp.RegularAnalysis.TextRegex
+namespace StgSharp.RegularAnalysis.Text
 {
-    internal class RegexIRGenerator
+    internal class RegexHIRGenerator
     {
 
         private Node _head = new();
         private Node _tail;
 
-        public RegexIRGenerator()
+        public RegexHIRGenerator()
         {
             _tail = _head;
         }
@@ -52,15 +52,7 @@ namespace StgSharp.RegularAnalysis.TextRegex
                     int fail
         )
         {
-            _tail.Add(new RegexConditionIR(success, fail));
-            Count++;
-        }
-
-        public void EmitConfigFail(
-                    int fail
-        )
-        {
-            _tail.Add(new RegexConfigFailIR(fail));
+            _tail.Add(new RegexConditionHIR(success, fail));
             Count++;
         }
 
@@ -71,7 +63,7 @@ namespace StgSharp.RegularAnalysis.TextRegex
                     int operation
         )
         {
-            _tail.Add(new RegexCountIR(min, max, isGreedy, operation));
+            _tail.Add(new RegexCountHIR(min, max, isGreedy, operation));
             Count++;
         }
 
@@ -82,7 +74,7 @@ namespace StgSharp.RegularAnalysis.TextRegex
                     string seq
         )
         {
-            _tail.Add(new RegexCountSeqIR(min, max, isGreedy, seq));
+            _tail.Add(new RegexCountSeqHIR(min, max, isGreedy, seq));
             Count++;
         }
 
@@ -93,12 +85,12 @@ namespace StgSharp.RegularAnalysis.TextRegex
                     string set
         )
         {
-            _tail.Add(new RegexCountSetIR(min, max, isGreedy, set));
+            _tail.Add(new RegexCountSetHIR(min, max, isGreedy, set));
             Count++;
         }
 
         public void EmitIRStream(
-                    RegexIRGenerator stream
+                    RegexHIRGenerator stream
         )
         {
             if (stream is null || stream.Count == 0) {
@@ -113,7 +105,7 @@ namespace StgSharp.RegularAnalysis.TextRegex
                     string unit
         )
         {
-            _tail.Add(new RegexMatchSeqIR(unit));
+            _tail.Add(new RegexMatchSeqHIR(unit));
             Count++;
         }
 
@@ -121,7 +113,16 @@ namespace StgSharp.RegularAnalysis.TextRegex
                     string unit
         )
         {
-            _tail.Add(new RegexMatchSetIR(unit));
+            _tail.Add(new RegexMatchSetHIR(unit));
+            Count++;
+        }
+
+        public void EmitPack(
+                    bool condition,
+                    string name = ""
+        )
+        {
+            _tail.Add(new RegexPackHIR(condition, name));
             Count++;
         }
 
@@ -129,7 +130,7 @@ namespace StgSharp.RegularAnalysis.TextRegex
                     bool isSuccess
         )
         {
-            _tail.Add(new RegexPopIR(isSuccess));
+            _tail.Add(new RegexPopHIR(isSuccess));
             Count++;
         }
 
@@ -137,13 +138,13 @@ namespace StgSharp.RegularAnalysis.TextRegex
                     int count
         )
         {
-            _tail.Add(new RegexTryIR(count));
+            _tail.Add(new RegexTryHIR(count));
             Count++;
         }
 
-        public List<RegexIR> ToList()
+        public List<RegexHIR> ToList()
         {
-            List<RegexIR> result = new(Count);
+            List<RegexHIR> result = new(Count);
             Node current = _head;
             while (current != null)
             {
@@ -160,11 +161,11 @@ namespace StgSharp.RegularAnalysis.TextRegex
 
             public int Count => IRList.Count;
 
-            public List<RegexIR> IRList { get; } = [];
+            public List<RegexHIR> IRList { get; } = [];
 
             [MethodImpl(MethodImplOptions.AggressiveInlining)]
             public void Add(
-                        RegexIR ir
+                        RegexHIR ir
             )
             {
                 IRList.Add(ir);
diff --git a/StgSharp.RegularAnalysis/TextRegex/RegexInterpreter.Optimize.cs b/StgSharp.RegularAnalysis/Text/RegexInterpreter.ASTOptimize.cs
similarity index 99%
rename from StgSharp.RegularAnalysis/TextRegex/RegexInterpreter.Optimize.cs
rename to StgSharp.RegularAnalysis/Text/RegexInterpreter.ASTOptimize.cs
index 6e7bbb80..36398174 100644
--- a/StgSharp.RegularAnalysis/TextRegex/RegexInterpreter.Optimize.cs
+++ b/StgSharp.RegularAnalysis/Text/RegexInterpreter.ASTOptimize.cs
@@ -37,7 +37,7 @@ using System.Text;
 using System.Threading.Tasks;
 using System.Xml.Linq;
 
-namespace StgSharp.RegularAnalysis.TextRegex
+namespace StgSharp.RegularAnalysis.Text
 {
     public partial class RegexInterpreter
     {
diff --git a/StgSharp.RegularAnalysis/TextRegex/RegexInterpreter.IRGenerate.cs b/StgSharp.RegularAnalysis/Text/RegexInterpreter.HIRGenerate.cs
similarity index 88%
rename from StgSharp.RegularAnalysis/TextRegex/RegexInterpreter.IRGenerate.cs
rename to StgSharp.RegularAnalysis/Text/RegexInterpreter.HIRGenerate.cs
index 7c7e77a2..bc42c472 100644
--- a/StgSharp.RegularAnalysis/TextRegex/RegexInterpreter.IRGenerate.cs
+++ b/StgSharp.RegularAnalysis/Text/RegexInterpreter.HIRGenerate.cs
@@ -1,6 +1,6 @@
 //-----------------------------------------------------------------------
 // -----------------------------------------------------------------------
-// file="RegexInterpreter.IRGenerate"
+// file="RegexInterpreter.HIRGenerate"
 // Project: StgSharp
 // AuthorGroup: Nitload
 // Copyright (c) Nitload. All rights reserved.
@@ -38,13 +38,13 @@ using System.Text;
 using System.Threading.Tasks;
 using System.Xml;
 
-namespace StgSharp.RegularAnalysis.TextRegex
+namespace StgSharp.RegularAnalysis.Text
 {
     public partial class RegexInterpreter
     {
 
-        internal static List<RegexIR> GenerateIR(
-                                      AbstractSyntaxTree<RegexAstNode, RegexElementLabel> tree
+        internal static List<RegexHIR> GenerateIR(
+                                       AbstractSyntaxTree<RegexAstNode, RegexElementLabel> tree
         )
         {
             return GenerateIRRecursion(tree.Root, 0).ToList();
@@ -53,7 +53,7 @@ namespace StgSharp.RegularAnalysis.TextRegex
         [MethodImpl(MethodImplOptions.AggressiveInlining)]
         private static void EmitCount(
                             int curDepth,
-                            RegexIRGenerator ir,
+                            RegexHIRGenerator ir,
                             int min,
                             int max,
                             bool isGreedy,
@@ -66,7 +66,7 @@ namespace StgSharp.RegularAnalysis.TextRegex
             if (isRightOperator)        // seq to be count is complex
             {
                 ir.EmitCount(min, max, isGreedy, 1);
-                RegexIRGenerator rightIR = GenerateIRRecursion(right, curDepth + 1);
+                RegexHIRGenerator rightIR = GenerateIRRecursion(right, curDepth + 1);
                 ir.EmitTry(rightIR.Count);
                 ir.EmitIRStream(rightIR);
             } else
@@ -89,9 +89,9 @@ namespace StgSharp.RegularAnalysis.TextRegex
             }
         }
 
-        private static RegexIRGenerator GenerateIRRecursion(
-                                        RegexAstNode node,
-                                        int depth
+        private static RegexHIRGenerator GenerateIRRecursion(
+                                         RegexAstNode node,
+                                         int depth
         )
         {
 #pragma warning disable IDE0010
@@ -102,8 +102,8 @@ namespace StgSharp.RegularAnalysis.TextRegex
             int curDepth = depth;
             RegexAstNode left = node.Left;
             RegexAstNode right = node.Right;
-            RegexIRGenerator leftIR;
-            RegexIRGenerator rightIR;
+            RegexHIRGenerator leftIR;
+            RegexHIRGenerator rightIR;
 
             Token<RegexElementLabel> token = node.Source;
 
@@ -116,15 +116,16 @@ namespace StgSharp.RegularAnalysis.TextRegex
                     bool is_group = group_source.Length == 1;           // simple group: (pattern)
                     string group_name =
                         group_source.StartsWith(@"(?<", StringComparison.Ordinal) ?
-                        group_source[3..^2] :                   /* named group (?<name>) */
+                        group_source[3..^1] :                   /* named group (?<name>) */
                         string.Empty;                           /* unnamed group (?:) */
 
-                    RegexIRGenerator groupIR = new();
+                    RegexHIRGenerator groupIR = new();
                     rightIR = GenerateIRRecursion(right, curDepth + 1);
                     groupIR.EmitTry(rightIR.Count);
                     groupIR.EmitIRStream(rightIR);
                     groupIR.EmitCondition(2, 1);
-                    /*generate pack ir here*/
+                    groupIR.EmitPop(false);
+                    groupIR.EmitPack(true, group_name);
                     return groupIR;
                 case RegexElementLabel.COUNT:
                     string count_source = token.Value;
@@ -143,14 +144,14 @@ namespace StgSharp.RegularAnalysis.TextRegex
                         _ => throw new InvalidOperationException("Unexpected sequence mask for count operator"),
                     };
                     ReadOnlySpan<char> count_source_span = count_source;
-                    RegexIRGenerator countIR = new();
+                    RegexHIRGenerator countIR = new();
 
                     if (count_source_span[0] == '*')              // infinite count
                     {
-                        EmitCount(curDepth, countIR, 0, int.MaxValue, isGreedy, seq);
+                        EmitCount(curDepth, countIR, 0, -1, isGreedy, seq);
                     } else if (count_source_span[0] == '+')       // at least one
                     {
-                        EmitCount(curDepth, countIR, 1, int.MaxValue, isGreedy, seq);
+                        EmitCount(curDepth, countIR, 1, -1, isGreedy, seq);
                     } else if (count_source_span[0] == '?')       // zero or one
                     {
                         EmitCount(curDepth, countIR, 0, 1, isGreedy, seq);
@@ -159,7 +160,7 @@ namespace StgSharp.RegularAnalysis.TextRegex
                         int comma_pos = count_source_span.IndexOf(',');
                         if (comma_pos == -1)
                         {
-                            if (int.TryParse(count_source[1..(isGreedy ? ^3 : ^2)], out int count))
+                            if (int.TryParse(count_source[1..(isGreedy ? ^2 : ^1)], out int count))
                             {
                                 EmitCount(curDepth, countIR, count, count, isGreedy, seq);
                             } else
@@ -169,7 +170,7 @@ namespace StgSharp.RegularAnalysis.TextRegex
                         } else
                         {
                             if (int.TryParse(count_source[1..comma_pos], out int min) &&
-                                int.TryParse(count_source[(comma_pos + 1)..(isGreedy ? ^3 : ^2)], out int max))
+                                int.TryParse(count_source[(comma_pos + 1)..(isGreedy ? ^1 : ^2)], out int max))
                             {
                                 EmitCount(curDepth, countIR, min, max, isGreedy, seq);
                             } else
@@ -184,7 +185,7 @@ namespace StgSharp.RegularAnalysis.TextRegex
 
                     return countIR;
                 case RegexElementLabel.CONCAT:
-                    RegexIRGenerator concat = new();
+                    RegexHIRGenerator concat = new();
 
                     // process left node
                     if (RegexAstNode.IsNullOrEmpty(left))
@@ -237,13 +238,13 @@ namespace StgSharp.RegularAnalysis.TextRegex
                     }
                     return concat;
                 case RegexElementLabel.ALT:
-                    RegexIRGenerator alt = new();
-                    List<RegexIRGenerator> casesIRList = [];
+                    RegexHIRGenerator alt = new();
+                    List<RegexHIRGenerator> casesIRList = [];
                     RegexAstNode cur_node = right;
                     int ir_count = 0;
                     while (!RegexAstNode.IsNullOrEmpty(cur_node))
                     {
-                        RegexIRGenerator _case = GenerateIRRecursion(cur_node, curDepth + 1);
+                        RegexHIRGenerator _case = GenerateIRRecursion(cur_node, curDepth + 1);
                         casesIRList.Add(_case);
                         ir_count += _case.Count;
                         cur_node = cur_node.Next;
@@ -251,7 +252,7 @@ namespace StgSharp.RegularAnalysis.TextRegex
                     int case_count = casesIRList.Count;
                     while (case_count != 0)
                     {
-                        RegexIRGenerator cur_ir = casesIRList[^case_count];
+                        RegexHIRGenerator cur_ir = casesIRList[^case_count];
                         ir_count -= cur_ir.Count;
                         case_count--;
                         alt.EmitTry(cur_ir.Count);
@@ -264,11 +265,11 @@ namespace StgSharp.RegularAnalysis.TextRegex
                 case RegexElementLabel.UNIT_SPAN:
                     goto case RegexElementLabel.UNIT;
                 case RegexElementLabel.UNIT:
-                    RegexIRGenerator unit_seq = new();
+                    RegexHIRGenerator unit_seq = new();
                     unit_seq.EmitMatchSeq(node.Source.Value);
                     return unit_seq;
                 case RegexElementLabel.UNIT_SET:
-                    RegexIRGenerator unit_set = new();
+                    RegexHIRGenerator unit_set = new();
                     unit_set.EmitMatchSet(node.Source.Value);
                     return unit_set;
                 default:
diff --git a/StgSharp.RegularAnalysis/Text/RegexInterpreter.IRScan.cs b/StgSharp.RegularAnalysis/Text/RegexInterpreter.IRScan.cs
new file mode 100644
index 00000000..61c7b28c
--- /dev/null
+++ b/StgSharp.RegularAnalysis/Text/RegexInterpreter.IRScan.cs
@@ -0,0 +1,69 @@
+//-----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+// file="RegexInterpreter.IRScan"
+// Project: StgSharp
+// AuthorGroup: Nitload
+// Copyright (c) Nitload. All rights reserved.
+//     
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in all
+// copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+//     
+// -----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+
+namespace StgSharp.RegularAnalysis.Text
+{
+    public partial class RegexInterpreter
+    {
+
+        internal static void ScanAndMarkIROptimization(
+                             RegexHIR ir
+        )
+        {
+            switch (ir.Command)
+            {
+                case RegexIRCommand.Match_Seq:
+                    break;
+                case RegexIRCommand.Match_Set:
+                    break;
+                case RegexIRCommand.Pack:
+                    break;
+                case RegexIRCommand.Condition:
+                    break;
+                case RegexIRCommand.Try:
+                    break;
+                case RegexIRCommand.Count:
+                    break;
+                case RegexIRCommand.Count_Seq:
+                    break;
+                case RegexIRCommand.Count_Set:
+                    break;
+                case RegexIRCommand.Pop:
+                    break;
+                default:
+                    break;
+            }
+        }
+
+    }
+}
diff --git a/StgSharp.RegularAnalysis/Text/RegexInterpreter.SourceGenerate.cs b/StgSharp.RegularAnalysis/Text/RegexInterpreter.SourceGenerate.cs
new file mode 100644
index 00000000..67d84307
--- /dev/null
+++ b/StgSharp.RegularAnalysis/Text/RegexInterpreter.SourceGenerate.cs
@@ -0,0 +1,51 @@
+//-----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+// file="RegexInterpreter.SourceGenerate"
+// Project: StgSharp
+// AuthorGroup: Nitload
+// Copyright (c) Nitload. All rights reserved.
+//     
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in all
+// copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+//     
+// -----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+
+namespace StgSharp.RegularAnalysis.Text
+{
+    public partial class RegexInterpreter
+    {
+
+        internal static string GetSourceCode(
+                               IEnumerable<RegexHIR> irs
+        )
+        {
+            StringBuilder sb = new StringBuilder();
+            foreach (RegexHIR ir in irs) {
+                ScanAndMarkIROptimization(ir);
+            }
+            return sb.ToString();
+        }
+
+    }
+}
diff --git a/StgSharp.RegularAnalysis/TextRegex/RegexInterpreter.cs b/StgSharp.RegularAnalysis/Text/RegexInterpreter.cs
similarity index 99%
rename from StgSharp.RegularAnalysis/TextRegex/RegexInterpreter.cs
rename to StgSharp.RegularAnalysis/Text/RegexInterpreter.cs
index 0a5cfbdc..f7f1c02d 100644
--- a/StgSharp.RegularAnalysis/TextRegex/RegexInterpreter.cs
+++ b/StgSharp.RegularAnalysis/Text/RegexInterpreter.cs
@@ -35,7 +35,7 @@ using System.Text;
 
 using System.Threading.Tasks;
 
-namespace StgSharp.RegularAnalysis.TextRegex
+namespace StgSharp.RegularAnalysis.Text
 {
     public partial class RegexInterpreter
     {
diff --git a/StgSharp.RegularAnalysis/Text/RegexLIR.cs b/StgSharp.RegularAnalysis/Text/RegexLIR.cs
new file mode 100644
index 00000000..272de4f3
--- /dev/null
+++ b/StgSharp.RegularAnalysis/Text/RegexLIR.cs
@@ -0,0 +1,60 @@
+//-----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+// file="RegexLIR"
+// Project: StgSharp
+// AuthorGroup: Nitload
+// Copyright (c) Nitload. All rights reserved.
+//     
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in all
+// copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+//     
+// -----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+
+namespace StgSharp.RegularAnalysis.Text
+{
+    internal enum LIRCommand
+    {
+
+        Search = 0,
+        Count = 1,
+        Try = 2,
+
+    }
+
+    internal abstract class RegexLIR
+    {
+
+        public abstract LIRCommand Command { get; }
+
+    }
+
+    internal enum RegexSearchMethod
+    {
+
+        Match,
+        Find,
+        Assert,
+
+    }
+}
diff --git a/StgSharp.RegularAnalysis/TextRegex/RegexTokenParser.cs b/StgSharp.RegularAnalysis/Text/RegexTokenParser.cs
similarity index 98%
rename from StgSharp.RegularAnalysis/TextRegex/RegexTokenParser.cs
rename to StgSharp.RegularAnalysis/Text/RegexTokenParser.cs
index 452a65a1..f4ee54a3 100644
--- a/StgSharp.RegularAnalysis/TextRegex/RegexTokenParser.cs
+++ b/StgSharp.RegularAnalysis/Text/RegexTokenParser.cs
@@ -34,7 +34,7 @@ using System.Linq;
 using System.Text;
 using System.Threading.Tasks;
 
-namespace StgSharp.RegularAnalysis.TextRegex
+namespace StgSharp.RegularAnalysis.Text
 {
     internal class RegexTokenParser : TokenParser<RegexElementLabel, RegexElementLabel>
     {
diff --git a/StgSharp.RegularAnalysis/TextRegex/RegexTokenReader.cs b/StgSharp.RegularAnalysis/Text/RegexTokenReader.cs
similarity index 99%
rename from StgSharp.RegularAnalysis/TextRegex/RegexTokenReader.cs
rename to StgSharp.RegularAnalysis/Text/RegexTokenReader.cs
index d57f3518..953b0eb9 100644
--- a/StgSharp.RegularAnalysis/TextRegex/RegexTokenReader.cs
+++ b/StgSharp.RegularAnalysis/Text/RegexTokenReader.cs
@@ -34,7 +34,7 @@ using System.Security.Permissions;
 using System.Text;
 using System.Threading.Tasks;
 
-namespace StgSharp.RegularAnalysis.TextRegex
+namespace StgSharp.RegularAnalysis.Text
 {
     internal class RegexTokenReader : ITokenReader<RegexElementLabel>
     {
diff --git a/StgSharp.RegularAnalysis/TextRegex/RegexTokenUtil.cs b/StgSharp.RegularAnalysis/Text/RegexTokenUtil.cs
similarity index 97%
rename from StgSharp.RegularAnalysis/TextRegex/RegexTokenUtil.cs
rename to StgSharp.RegularAnalysis/Text/RegexTokenUtil.cs
index 3d51ab3c..1a32cc24 100644
--- a/StgSharp.RegularAnalysis/TextRegex/RegexTokenUtil.cs
+++ b/StgSharp.RegularAnalysis/Text/RegexTokenUtil.cs
@@ -32,7 +32,7 @@ using System.Linq;
 using System.Text;
 using System.Threading.Tasks;
 
-namespace StgSharp.RegularAnalysis.TextRegex
+namespace StgSharp.RegularAnalysis.Text
 {
     internal static class RegexTokenUtil
     {
diff --git a/StgSharp.RegularAnalysis/Text/TextHelper.cs b/StgSharp.RegularAnalysis/Text/TextHelper.cs
new file mode 100644
index 00000000..78d9b4d6
--- /dev/null
+++ b/StgSharp.RegularAnalysis/Text/TextHelper.cs
@@ -0,0 +1,50 @@
+//-----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+// file="TextHelper"
+// Project: StgSharp
+// AuthorGroup: Nitload
+// Copyright (c) Nitload. All rights reserved.
+//     
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in all
+// copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+//     
+// -----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+
+namespace StgSharp.RegularAnalysis.Text
+{
+    public static class TextHelper
+    {
+
+        public static bool TryLastIndexOf(
+                           this ReadOnlySpan<char> span,
+                           ReadOnlySpan<char> value,
+                           out int index
+        )
+        {
+            index = span.LastIndexOf(value);
+            return index != -1;
+        }
+
+    }
+}
diff --git a/StgSharp.RegularAnalysis/Text/TextRegex.cs b/StgSharp.RegularAnalysis/Text/TextRegex.cs
new file mode 100644
index 00000000..b2f4c80d
--- /dev/null
+++ b/StgSharp.RegularAnalysis/Text/TextRegex.cs
@@ -0,0 +1,47 @@
+//-----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+// file="TextRegex"
+// Project: StgSharp
+// AuthorGroup: Nitload
+// Copyright (c) Nitload. All rights reserved.
+//     
+// Permission is hereby granted, free of charge, to any person obtaining a copy
+// of this software and associated documentation files (the "Software"), to deal
+// in the Software without restriction, including without limitation the rights
+// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+// copies of the Software, and to permit persons to whom the Software is
+// furnished to do so, subject to the following conditions:
+//
+// The above copyright notice and this permission notice shall be included in all
+// copies or substantial portions of the Software.
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+// SOFTWARE.
+//     
+// -----------------------------------------------------------------------
+// -----------------------------------------------------------------------
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Runtime.CompilerServices;
+using System.Text;
+using System.Threading.Tasks;
+
+namespace StgSharp.RegularAnalysis.Text
+{
+    public abstract class TextRegex
+    {
+
+        public abstract bool IsMatch(
+                             ReadOnlySpan<char> text
+        );
+
+    }
+
+    public ref struct MatchResult { }
+}
diff --git a/StgSharpDebug/TestRegexAnalyse.cs b/StgSharpDebug/TestRegexAnalyse.cs
index 7d9e80d9..16203514 100644
--- a/StgSharpDebug/TestRegexAnalyse.cs
+++ b/StgSharpDebug/TestRegexAnalyse.cs
@@ -26,7 +26,7 @@
 // -----------------------------------------------------------------------
 // -----------------------------------------------------------------------
 using StgSharp.RegularAnalysis.Abstraction;
-using StgSharp.RegularAnalysis.TextRegex;
+using StgSharp.RegularAnalysis.Text;
 using System;
 using System.Collections.Generic;
 using System.Linq;
@@ -44,7 +44,7 @@ namespace StgSharp
             @"abc[0-9]x|abc[0-9]y|abd[0-9]x|abd[0-9]y|xyz",
             @"get_name|get_value|get_type|set_name|set_value|set_type",
             @"helix|hello|help|hero|hers|her",
-            @"a*.[abc]*(abcd|a(bc)c)+(?<name>abc)"
+            @"a*.[abc]*(abcd|a(bc)c)+(?<name>abc{11,13})"
         ];
 
         public static void Run()
@@ -61,7 +61,7 @@ namespace StgSharp
                     Console.Write(new string('|', enumerator.Depth));
                     Console.WriteLine($"\t{node.Source.Value}\t{node.Source.Flag}");
                 }
-                List<RegexIR> ir = RegexInterpreter.GenerateIR(tree);
+                List<RegexHIR> ir = RegexInterpreter.GenerateIR(tree);
                 Console.WriteLine("-----IR-----");
                 for (int i = 0; i < ir.Count; i++) {
                     Console.WriteLine(ir[i].AsCommand(i));
diff --git a/future/ngra-regex-ir-review.md b/future/ngra-regex-ir-review.md
new file mode 100644
index 00000000..84c2e96a
--- /dev/null
+++ b/future/ngra-regex-ir-review.md
@@ -0,0 +1,1050 @@
+# NGRA 字符串正则 IR 设计补充建议
+
+基于当前 `ngra-regex-ir.md` 的设计，我认为整体方向是成立的：它已经不是“完全低级的执行字节码”，而是保留了 `COUNT`、`COUNT_SEQ`、`COUNT_SET`、`TRY`、`PACK` 等高层结构，因此可以继续承担后续 match-opt 的输入。
+
+不过，为了让规则更容易扩展、避免语义误判，我建议把优化明确拆成三类，并把当前 IR 分成“语义 IR / HIR”和“计划 IR / LIR”。
+
+---
+
+## 1. 总体判断
+
+当前 IR 最大的优点：
+
+1. **足够紧凑**：`COUNT_SEQ`、`COUNT_SET`、`TRY` 等复合指令可以避免过早打碎成 `LoadChar + Compare + Branch`。
+2. **保留 regex 结构**：量词、字符集、确定序列、组打包等信息仍然可见。
+3. **适合继续做 match planning**：后续可以从这些指令中识别 `.*literal`、`[set]*`、bounded repeat、固定长度 block 等结构。
+
+当前 IR 最大的风险：
+
+1. `SEARCH` 这个名字容易误导。当前 HIR 中的 `SEARCH` 实际应当是 **MATCH_AT_CURRENT**，不是允许跳过字符的 `FIND`。
+2. 检查点目前只描述为锚点指针栈，但真正的 backtracking frame 至少还要保存 `pos`、capture watermark、symbol watermark、repeat state 和 continuation。
+3. `COUNT_*` 的语义必须明确是否保留 retry/backtracking，否则 `a*.`、`[abc]*abc` 等场景会被错误优化。
+4. 字符集的语义表示和物理表示要分开：语义上 `[0-9a-z]` 是一个 set，物理上可以用 inline set-region 的多条 range-test 表示，避免每个 set 挂 `List<RangePair>`。
+
+---
+
+## 2. 建议的分层
+
+我建议仍然区分 HIR 和 LIR，不是因为机器必须分，而是因为人写规则时边界更清楚。
+
+```text
+AST
+  ↓
+AST Opt
+  ↓
+HIR Gen
+  ↓
+HIR Analysis
+  ↓
+HIR → LIR / MatchPlan Lowering
+  ↓
+LIR
+  ↓
+SourceGen / RuntimeGen / Interpreter
+```
+
+### HIR：表达 regex 语义
+
+HIR 保留接近正则结构的指令：
+
+```text
+SEARCH_SEQ      // 语义上是当前位置 MATCH_SEQ
+SEARCH_SET      // 语义上是当前位置 MATCH_SET
+SEARCH_DOT      // 或 UNIT_DOT，不应和 literal "." 混淆
+COUNT
+COUNT_SEQ
+COUNT_SET
+TRY
+POP
+PACK
+```
+
+注意：HIR 中可以继续叫 `SEARCH`，但文档里要明确：
+
+```text
+HIR 的 SEARCH_* 不表示向后查找；
+它表示从当前活动指针开始匹配。
+真正允许跳过字符的搜索操作只在 LIR / MatchPlan 阶段引入，命名为 FIND_*。
+```
+
+### LIR：表达匹配计划
+
+LIR 不只是更低级，而是“已经做过策略规划”的 IR：
+
+```text
+MATCH_SEQ
+MATCH_CHAR
+MATCH_DOT
+MATCH_SET
+
+FIND_SEQ
+FIND_CHAR
+FIND_SET
+FIND_LAST_SEQ
+FIND_LAST_SET
+
+CONSUME_WHILE_SET
+CONSUME_UNTIL_CHAR
+CONSUME_UNTIL_SET
+
+COUNT_SET
+COUNT_SEQ
+COUNT_BLOCK
+CHECK_MIN_LENGTH
+CHECK_FIXED_OFFSET
+
+PUSH_BACKTRACK
+POP_BACKTRACK
+COMMIT
+FAIL
+
+CAPTURE_BEGIN
+CAPTURE_END
+PACK
+```
+
+LIR 中应该已经包含：
+
+```text
+LoweredSetId
+SearchMode
+RepeatMode
+RetryMode
+CaptureMode
+EmitHint
+```
+
+---
+
+## 3. 三类优化策略
+
+建议把所有优化分成三类：
+
+```text
+1. Regex algebra rewrite
+   改正则结构。
+
+2. Match plan rewrite
+   改匹配策略。
+
+3. Backend / C# span trick
+   改 C# 代码写法。
+```
+
+这个分类很重要，因为它能直接决定优化应该放在哪个阶段。
+
+---
+
+# 4. Regex algebra rewrite
+
+这类优化应该放在 AST Opt 或 HIR 生成前后。它们改变的是正则结构，而不是 C# 写法。
+
+## 4.1 Literal 合并
+
+```regex
+a b c
+```
+
+变成：
+
+```regex
+abc
+```
+
+HIR：
+
+```text
+SEARCH_SEQ "abc"
+```
+
+不要生成三条 `SEARCH_CHAR`。
+
+---
+
+## 4.2 Concat / Alternation flatten
+
+```regex
+a(?:b(?:c))
+```
+
+变成：
+
+```regex
+abc
+```
+
+```regex
+a|(?:b|c)
+```
+
+变成：
+
+```regex
+a|b|c
+```
+
+---
+
+## 4.3 Alternation factoring / trie
+
+这个你已经放在 AST Opt 做，位置是对的。
+
+```regex
+foobar|foobaz|foozip
+```
+
+变成：
+
+```regex
+foo(?:bar|baz|zip)
+```
+
+或者 trie AST：
+
+```text
+SEQ "foo"
+ALT
+  SEQ "bar"
+  SEQ "baz"
+  SEQ "zip"
+```
+
+这类结构改写越早越好。后续 IR opt 只需要识别这个结构并决定是否 `switch` / first-set dispatch。
+
+---
+
+## 4.4 Bounded repeat 展开
+
+```regex
+a{2,3}
+```
+
+可以变成：
+
+```regex
+aa a?
+```
+
+或：
+
+```text
+SEARCH_SEQ "aa"
+OPTIONAL SEARCH_CHAR 'a'
+```
+
+对于 lazy：
+
+```regex
+a{2,3}?b
+```
+
+要注意尝试顺序。它等价于：
+
+```regex
+aa(?:b|ab)
+```
+
+而不是：
+
+```regex
+aa(?:a|ab)
+```
+
+Greedy 版本：
+
+```regex
+a{2,3}b
+```
+
+如果展开成显式分支，尝试顺序更像：
+
+```regex
+aa(?:ab|b)
+```
+
+语言集合相同，但 backtracking 尝试顺序不同。
+
+---
+
+## 4.5 `a+b` 不建议改成 `a*(ab|b)`
+
+当前文档中写：
+
+```text
+a+b 可以优化为 a*(ab|b)
+```
+
+这个需要修正。
+
+`a+b` 的语言是：
+
+```text
+a^n b, n >= 1
+```
+
+而 `a*(ab|b)` 会额外匹配：
+
+```text
+b
+```
+
+因为第二个分支 `a*b` 可以在 `a*` 取 0 时匹配 `b`。
+
+更安全的结构是：
+
+```regex
+aa*b
+```
+
+或者：
+
+```text
+SEARCH_CHAR 'a'
+COUNT_SEQ 'a' 0..∞
+SEARCH_CHAR 'b'
+```
+
+如果要利用后继 `b` 做策略优化，应在 MatchPlan 阶段处理，而不是做这个不等价的 algebra rewrite。
+
+---
+
+## 4.6 Charset 合并
+
+```regex
+[abc]|[def]
+```
+
+可以变成：
+
+```regex
+[abcdef]
+```
+
+```regex
+[a-f]|[d-z]
+```
+
+可以变成：
+
+```regex
+[a-z]
+```
+
+这属于 AST/HIR 层的字符类代数。
+
+---
+
+## 4.7 Charset 展开为 alternation 要谨慎
+
+文档中写：
+
+```text
+abc[de] 可以展开为 (abcd|abce)
+```
+
+这个是可行的，但不一定总是好。建议规则改成：
+
+```text
+当 charset 很小，并且展开后能形成更长 literal / trie / branch dispatch 时，可以展开；
+否则保留为 MATCH_SET。
+```
+
+例如：
+
+```regex
+abc[de]
+```
+
+既可以是：
+
+```text
+SEARCH_SEQ "abc"
+SEARCH_SET [de]
+```
+
+也可以是：
+
+```text
+ALT
+  SEARCH_SEQ "abcd"
+  SEARCH_SEQ "abce"
+```
+
+前者代码更短，后者可能更适合 required literal 或 trie。需要 code size budget。
+
+---
+
+# 5. Match plan rewrite
+
+这类优化应该放在 HIR Analysis / HIR → LIR lowering。
+
+它们不一定改成另一条正则，而是把匹配过程改成更高效的搜索/消费计划。
+
+---
+
+## 5.1 长度预测应该在 HIR Analysis 做，SourceGen 只负责发代码
+
+当前文档写“长度预测在 source-gen 阶段做”。我建议改成：
+
+```text
+minLength / maxLength 的计算在 HIR Analysis 做；
+SourceGen 只负责把 CHECK_MIN_LENGTH 发成 C#。
+```
+
+例如：
+
+```regex
+abc[0-9]{2}xyz
+```
+
+分析结果：
+
+```text
+MinLength = 8
+```
+
+LIR：
+
+```text
+CHECK_MIN_LENGTH 8
+MATCH_SEQ "abc"
+...
+```
+
+SourceGen：
+
+```csharp
+if ((uint)(end - pos) < 8) goto Fail;
+```
+
+---
+
+## 5.2 `.*` 优化
+
+HIR：
+
+```text
+COUNT_DOT min=0 max=∞ greedy=true
+MATCH_CHAR 'a'
+```
+
+可以 lower 成：
+
+```text
+FIND_LAST_CHAR 'a'
+```
+
+lazy：
+
+```text
+COUNT_DOT min=0 max=∞ greedy=false
+MATCH_CHAR 'a'
+```
+
+lower 成：
+
+```text
+FIND_CHAR 'a'
+```
+
+但要注意：
+
+1. 默认 `.` 不匹配 `\n`，搜索范围应限制在当前行。
+2. Singleline 模式下才可以跨整个 span。
+3. 如果 `.*` 内部或后继结构涉及 capture/backtracking，不能简单“删掉 COUNT”，而应转换为带 retry 的 find plan。
+4. 后继如果是 `MATCH_SEQ "abc"`，则变成 `FIND_SEQ` / `FIND_LAST_SEQ`。
+5. 后继如果是 `MATCH_SET`，则变成 `FIND_SET` / `FIND_LAST_SET`。
+
+---
+
+## 5.3 `[set]*` 优化
+
+文档中写 `[set]*` 优化在 source-gen 阶段做。建议改成：
+
+```text
+是否将 COUNT_SET 变成 CONSUME_WHILE / CONSUME_UNTIL，在 lowering 做；
+具体使用 IndexOfAnyExcept / SearchValues / manual loop，在 SourceGen 做。
+```
+
+例如：
+
+```regex
+[a-f0-9]*
+```
+
+LIR：
+
+```text
+CONSUME_WHILE_SET LowerHex
+```
+
+SourceGen：
+
+```csharp
+int n = span.IndexOfAnyExcept(s_lowerHex);
+```
+
+或者手写 loop / range compare。
+
+---
+
+## 5.4 First-set 搜索
+
+如果非锚定匹配：
+
+```regex
+abc|defg|xyz
+```
+
+first-set 是：
+
+```text
+'a', 'd', 'x'
+```
+
+LIR 可以是：
+
+```text
+FIND_SET { 'a', 'd', 'x' }
+VERIFY_BRANCH
+```
+
+你文档里说“这个不好做，因为 IR 直接把 ALT 打碎成 if-else 串了”。这说明 AST Opt / HIR 里最好保留 ALT region 或者至少保留 branch summary。
+
+即使 AST 已经 trie 化，IR 也应该尽量不要把它完全打碎成不可恢复的 if-else 串。至少给每个 region 计算 summary：
+
+```text
+FirstSet
+MinLength
+RequiredLiteral
+CanBacktrack
+CanCapture
+```
+
+---
+
+## 5.5 Required literal 搜索
+
+例如：
+
+```regex
+.*ERROR:[0-9]+
+```
+
+可以直接规划：
+
+```text
+FIND_SEQ "ERROR:"
+VERIFY_AFTER_ANCHOR
+```
+
+这通常比只做 first-set 更强。
+
+---
+
+## 5.6 Fixed-offset check
+
+```regex
+[a-z]{20}XYZ
+```
+
+可以 lower 成：
+
+```text
+CHECK_MIN_LENGTH 23
+CHECK_AT_OFFSET 20 MATCH_SEQ "XYZ"
+MATCH_SET_REPEAT [a-z] count=20
+MATCH_SEQ "XYZ"
+```
+
+这个属于 match plan，不属于 sourcegen。
+
+---
+
+## 5.7 Auto-atomic / possessive 化
+
+```regex
+[0-9]*[A-Z]
+```
+
+由于：
+
+```text
+[0-9] ∩ [A-Z] = ∅
+```
+
+所以 `[0-9]*` 不需要回退，可以 lower 成：
+
+```text
+COUNT_SET [0-9] greedy atomic
+MATCH_SET [A-Z]
+```
+
+如果后继 first-set 和 repeat set 有交集，就不能这么做。
+
+---
+
+## 5.8 Repeat retry mode
+
+LIR 应该明确 repeat 的 retry 策略：
+
+```text
+RetryMode.None
+RetryMode.BackoffBy1
+RetryMode.BackoffByStride
+RetryMode.SavedPositions
+RetryMode.GeneralBacktrack
+```
+
+例如：
+
+```regex
+(abc[cd])+
+```
+
+每轮固定长度 4，所以可以：
+
+```text
+RetryMode.BackoffByStride stride=4
+```
+
+不需要保存每个内部指令的回溯点。
+
+---
+
+# 6. Backend / C# span trick
+
+这类优化只放 SourceGen。它不再改变 regex 或 match plan。
+
+---
+
+## 6.1 `StartsWith`
+
+LIR：
+
+```text
+MATCH_SEQ "sample"
+```
+
+SourceGen 决定：
+
+```csharp
+slice.StartsWith("sample", StringComparison.Ordinal)
+```
+
+还是短串展开：
+
+```csharp
+slice.Length >= 3 &&
+slice[0] == 'a' &&
+slice[1] == 'b' &&
+slice[2] == 'c'
+```
+
+建议规则：
+
+```text
+长度 1:
+    char compare
+
+长度 2~3:
+    直接 char compare
+
+中长 literal:
+    StartsWith
+
+IgnoreCase:
+    StartsWith(..., OrdinalIgnoreCase)
+```
+
+---
+
+## 6.2 `[0-9]` 不建议写成 `IndexOf("0123456789")` 用于当前位置测试
+
+文档中写：
+
+```text
+[0-9] 展开为 IndexOf("0123456789")
+```
+
+这个要分两种场景：
+
+### 当前位置匹配
+
+如果是：
+
+```text
+MATCH_SET [0-9]
+```
+
+更好是：
+
+```csharp
+(uint)(ch - '0') <= 9
+```
+
+### span 中搜索
+
+如果是：
+
+```text
+FIND_SET [0-9]
+```
+
+可以考虑：
+
+```csharp
+span.IndexOfAnyInRange('0', '9')
+```
+
+或：
+
+```csharp
+span.IndexOfAny(s_digits)
+```
+
+所以规则应写成：
+
+```text
+MATCH_SET [0-9] → unsigned range compare
+FIND_SET [0-9]  → IndexOfAnyInRange / SearchValues
+```
+
+---
+
+## 6.3 SearchValues
+
+对于多字符集合：
+
+```text
+FIND_SET [0-9a-f]
+```
+
+SourceGen 可生成：
+
+```csharp
+private static readonly SearchValues<char> s_lowerHex =
+    SearchValues.Create("0123456789abcdef");
+```
+
+然后：
+
+```csharp
+span.IndexOfAny(s_lowerHex);
+```
+
+当前位置测试不一定要用 `SearchValues`，可能 range compare 或 bitmap 更快。
+
+---
+
+## 6.4 Inline vs helper
+
+复杂 Unicode set：
+
+```text
+MATCH_SET UnicodeWord
+```
+
+SourceGen 决定：
+
+```csharp
+if (!IsUnicodeWord(ch)) goto Fail;
+```
+
+还是：
+
+```csharp
+if (ch < 128)
+{
+    if (!AsciiWordContains(ch)) goto Fail;
+}
+else
+{
+    if (!IsUnicodeWord(ch)) goto Fail;
+}
+```
+
+这个应由代码尺寸、热度、是否 source-gen 决定。
+
+---
+
+## 6.5 Code size budget
+
+SourceGen 应负责最终代码膨胀控制：
+
+```text
+单方法最大语句数
+literal 展开阈值
+bounded repeat 展开阈值
+是否生成 helper
+是否抽 cold path
+SearchValues 字段数量
+```
+
+---
+
+# 7. 字符集的物理表示建议
+
+为了避免每个 set 挂 `List<RangePair>`，可以用线性 IR region 表示同一个字符测试。
+
+例如语义：
+
+```regex
+[0-9a-z]
+```
+
+物理 IR：
+
+```text
+MATCH_SET_BEGIN default=false end=13
+  TEST_RANGE '0' '9' accept=true
+  TEST_RANGE 'a' 'z' accept=true
+MATCH_SET_END
+```
+
+语义：
+
+```regex
+[^0-9a-z]
+```
+
+物理 IR：
+
+```text
+MATCH_SET_BEGIN default=true end=13
+  TEST_RANGE '0' '9' accept=false
+  TEST_RANGE 'a' 'z' accept=false
+MATCH_SET_END
+```
+
+要点是：
+
+```text
+这些 TEST_RANGE 测的是同一个字符；
+整个 MATCH_SET 只推进 pos 一次。
+```
+
+不能把它误解成连续匹配多个字符。
+
+复杂 Unicode set 可以走：
+
+```text
+MATCH_SET_REF setId=42
+```
+
+`setId=42` 指向全局 pooled table，而不是每个节点一个 List。
+
+---
+
+# 8. 检查点语义建议
+
+当前文档把检查点描述为锚点指针栈，这对 capture 起点有用，但不够覆盖 backtracking。
+
+建议把 checkpoint / backtrack frame 定义成：
+
+```text
+BacktrackFrame:
+    pos
+    anchor
+    captureTop
+    symbolTop
+    repeatState
+    continuation
+```
+
+其中：
+
+```text
+pos:
+    活动指针恢复位置
+
+anchor:
+    区间开始位置
+
+captureTop:
+    捕获记录栈回滚点
+
+symbolTop:
+    符号区回滚点
+
+repeatState:
+    当前 COUNT/loop 的计数、起止位置、retry mode
+
+continuation:
+    失败后跳到哪里继续尝试
+```
+
+如果暂时是 tokenizer / no-capture 模式，可以裁掉 captureTop，但 IR 语义最好预留。
+
+---
+
+# 9. 建议修改原文的几个点
+
+## 9.1 “HIT” 应改为 “HIR”
+
+原文：
+
+```text
+将HIT进行进一步优化
+```
+
+应为：
+
+```text
+将 HIR 进行进一步优化
+```
+
+## 9.2 `SEARCH` 的定义应改写
+
+原文说“向后搜索确定长度序列”。建议改为：
+
+```text
+HIR 中 SEARCH_* 表示从当前活动指针开始匹配；
+它不是允许跳过字符的 FIND。
+真正的 FIND_* 在 HIR→LIR 阶段由 match planner 引入。
+```
+
+## 9.3 `a+b → a*(ab|b)` 不等价
+
+建议删除或改成：
+
+```regex
+a+b → aa*b
+```
+
+或者：
+
+```text
+MATCH_CHAR 'a'
+COUNT_SEQ 'a' 0..∞
+MATCH_CHAR 'b'
+```
+
+## 9.4 `[0-9] → IndexOf("0123456789")` 应细分
+
+改成：
+
+```text
+当前位置 MATCH_SET:
+    用 unsigned range compare。
+
+span 搜索 FIND_SET:
+    用 IndexOfAnyInRange / SearchValues / IndexOfAny。
+```
+
+## 9.5 `.*` 优化不要说 “COUNT 消失”
+
+建议改成：
+
+```text
+COUNT_DOT 与后继匹配项合并为 FIND / FIND_LAST 计划。
+```
+
+这样更准确，因为有时还需要保留 retry / range 限制 / newline 限制。
+
+---
+
+# 10. 推荐最终管线
+
+```text
+Tokenize
+  - 区分 Dot 与 EscapedLiteralDot
+  - 区分 raw charset / literal / quantifier / group
+
+AST Build
+  - 建立 regex 语义树
+
+AST Opt
+  - literal merge
+  - concat flatten
+  - alternation factoring / trie
+  - bounded repeat normalization
+  - simple group elimination
+  - charset algebra merge
+
+HIR Gen
+  - SEARCH_* as MATCH_AT_CURRENT
+  - COUNT_*
+  - TRY / POP
+  - PACK
+  - raw set / raw dot
+
+HIR Analysis
+  - minLength / maxLength
+  - nullable
+  - firstSet
+  - lastSet
+  - required literal
+  - fixed length
+  - capture effect
+  - repeat overlap
+
+HIR → LIR / MatchPlan
+  - raw set → lowered set
+  - dot → dot mode
+  - introduce FIND
+  - introduce CONSUME
+  - choose retry mode
+  - choose atomic/possessive
+  - emit check-min-length / fixed-offset checks
+
+LIR
+  - explicit MATCH / FIND / CONSUME
+  - explicit lowered set
+  - explicit backtracking plan
+  - explicit capture plan
+
+SourceGen
+  - StartsWith / SequenceEqual
+  - IndexOf / IndexOfAny / IndexOfAnyExcept
+  - SearchValues
+  - range compare / bitmap
+  - inline/helper split
+  - code size budget
+```
+
+---
+
+# 11. 总结
+
+我建议保留你当前 IR 的核心设计，但把文档里的边界收紧：
+
+```text
+AST Opt:
+    做 Regex algebra rewrite。
+
+HIR:
+    表示从当前位置开始的 regex 语义。
+
+HIR Analysis:
+    提取结构属性。
+
+HIR → LIR:
+    做 Match plan rewrite。
+
+LIR:
+    表示明确的匹配计划。
+
+SourceGen:
+    做 C# span trick。
+```
+
+最重要的改动是：
+
+```text
+SEARCH 在 HIR 中应被定义为 MATCH_AT_CURRENT；
+FIND 只能由后续优化阶段引入。
+```
+
+以及：
+
+```text
+字符集可以用线性 set-region 表示，避免 List 分配；
+但 set-region 内的多条 range-test 必须测试同一个字符，不能变成多个连续 MATCH。
+```
+
+这样你的 IR 会同时满足三个目标：
+
+1. 对人来说可读；
+2. 对优化规则来说容易扩展；
+3. 对 GC 和 sourcegen 来说物理表示足够紧凑。
diff --git a/future/ngra-regex-ir.md b/future/ngra-regex-ir.md
index 7e80e2e9..4d50de5e 100644
--- a/future/ngra-regex-ir.md
+++ b/future/ngra-regex-ir.md
@@ -14,15 +14,62 @@
 
 使用DEFINE指令定义的符号定义在符号区中，符号的类型通过第一次赋值确定，源生成模式下，可以通过第一个赋值指令推导其类型
 
-## 算子
+## HIR算子
 
 1. 搜索算子（SEARCH %str; SEARCH %unicode），向后搜索确定长度序列或者满足规则的字符集，更新活动指针，不更新锚点指针，搜索结果保存在全局上下文中
 2. 搜索算子（SEARCH %set_expression），向后匹配一个满足字符集要求的字符，更新活动指针，不更新锚点指针，搜索结果保存在全局上下文中
-3. 失败配置指令（CINFIG_FAIL %fail），指定当前匹配栈后续如何处理一条指令的失败，可选为：FAST，立刻终止当前层级；CONTINUE，记录这个fail，但是继续操作；REVERSE，反过来，成功了才终止
-4. 条件指令（CONDITION %cmd_true %cmd_fail），比较前一条SEARCH指令是否成功，并指定跳转到指定的位置，如果某种条件不需要跳转，可以写1。如果想要将fail向上级继续抛出，可以%cmd_fail写0.
-5. 跳转执行指令（TRY %begin %end），尝试idx指向的IR块，并记住执行结果数据
-6. 计数指令（COUNT %min %max %is_greedy %cmd），重复执行指令cmd，直到满足参数配置的数量为止
-7. 快速序列计数指令（COUNT_SEQ %min %max %is_greedy %seq），快速搜索一个确定性序列，直到满足参数配置的数量为止
-8. 快速字符集计数指令（COUNT_SET %min %max %is_greedy %set），快速搜索一个字符集，直到满足参数配置的数量为止
-9. 尝试（TRY %count），对自身向后的一段内容在新的匹配栈层级中进行运行，被分层的指令数量由COUNT指令确定
-10. 弹出（POP %is_success），提前终止一个try块，并用is_success表示弹出行为视为匹配成功还是失败
+3. 条件指令（CONDITION %cmd_true %cmd_fail），比较前一条SEARCH指令是否成功，并指定跳转到指定的位置，如果某种条件不需要跳转，可以写1。如果想要将fail向上级继续抛出，可以%cmd_fail写0.
+4. 计数指令（COUNT %min %max %is_greedy %cmd），重复执行指令cmd，直到满足参数配置的数量为止
+5. 快速序列计数指令（COUNT_SEQ %min %max %is_greedy %seq），快速搜索一个确定性序列，直到满足参数配置的数量为止
+6. 快速字符集计数指令（COUNT_SET %min %max %is_greedy %set），快速搜索一个字符集，直到满足参数配置的数量为止
+7. 尝试（TRY %count），对自身向后的一段内容在新的匹配栈层级中进行运行，被分层的指令数量由COUNT指令确定
+8. 弹出（POP %is_success），提前终止一个try块，并用is_success表示弹出行为视为匹配成功还是失败
+9. 打包指令（PACK %is_success %name），检查上一条语句的结果，将匹配结果打包并记录为指定的名字。如果名字为空，则使用默认的序号进行索引
+
+## 优化策略
+
+应当注意的是，将HIT进行进一步优化或者展开的目的：
+
+1. 尽可能降低源生成或者运行时执行中，对多指令上下文的依赖，尽可能保证每一个IR可以单独输出最佳的代码
+2. 减少多余的结构，使得一行指令尽可能处理多的数据
+
+### 语义改写优化
+
+1. 计数器展开
+建议在lower做
+对于形如a+b或者a{2,4}b的正则表达式，可以优化为a*(ab|b)和a{2,3}(ab|b)的写法
+
+2. 确定性计数展开
+建议在lower做
+对于形如a{m,n}的正则表达式，将其展开为a{m}a{0,n-m}的形式
+
+3. charset合并
+建议在ast-opt做
+对于形如[ abc ]|[ def ]的表达式，将其合并为[ abcdef ]
+
+4. charset展开
+在lower做
+对于形如abc[ de ]的表达式，如果字符集容量不超过4，可以站卡为(abcd|abce)形式
+
+### 匹配策略改变
+
+1. 长度预测
+在source-gen阶段做
+对于一个正则表达式，可以计算其所需的最小长度，可以在开始的时候提前进行比较
+
+2. .*优化
+在lower阶段做
+对于COUNT_SET min=0 max=-1 "."指令，也就是.*操作，将后续指令从match转为find
+
+3. [ set ]*优化
+在source-gen阶段做
+对于其他的COUNT_SET min=0 max=-1 "???"指令，使用IndexOf，IndexOfAnyExcept等进行优化
+
+4. 首字母检查
+**这个不好做，因为IR直接把ALT打碎成if-else串了**
+对于(abc|defg|xyz)表达式，可以检查其首字母
+
+### C# span opt trick
+
+1. Span.StartWith
+2. [ 0-9 ]展开为IndexOf("0123456789")
\ No newline at end of file