feat: 添加语法分析基类抽象 (#8)

增加语法分析基类和状态转换接口抽象，为直接生成语法分析器做准备，同时也提前释放一些大对象，降低内存消耗。

Reviewed-on: PostGuard/Canon#8

Canon.Core/Abstractions/GrammarParseBase.cs

@@ -0,0 +1,70 @@
+using Canon.Core.GrammarParser;
+using Canon.Core.LexicalParser;
+
+namespace Canon.Core.Abstractions;
+
+/// <summary>
+/// 语法分析器接口
+/// </summary>
+public abstract class GrammarParserBase
+{
+    public abstract ITransformer BeginTransformer { get; }
+
+    public abstract NonTerminator Begin { get; }
+
+    public SyntaxNode Analyse(IEnumerable<SemanticToken> tokens)
+    {
+        Stack<AnalyseState> stack = [];
+        stack.Push(new AnalyseState(BeginTransformer, new SyntaxNode(SemanticToken.End)));
+
+        using IEnumerator<SemanticToken> enumerator = tokens.GetEnumerator();
+        if (!enumerator.MoveNext())
+        {
+            throw new InvalidOperationException("Input token list is empty");
+        }
+
+        while (true)
+        {
+            AnalyseState top = stack.Peek();
+
+            // 首先尝试进行归约
+            if (top.State.ReduceTable.TryGetValue(enumerator.Current, out ReduceInformation? information))
+            {
+                if (information.Left == Begin)
+                {
+                    // 如果是归约到起始符
+                    // 那么就直接返回不继续进行归约
+                    return top.Node;
+                }
+
+                SyntaxNode newNode = new(information.Left.Type);
+                for (int i = 0; i < information.Length; i++)
+                {
+                    newNode.Children.Add(stack.Pop().Node);
+                }
+
+                stack.Push(new AnalyseState(stack.Peek().State.ShiftTable[information.Left],
+                    newNode));
+                continue;
+            }
+
+            // 如果没有成功归约就进行移进
+            if (top.State.ShiftTable.TryGetValue(enumerator.Current, out ITransformer? next))
+            {
+                stack.Push(new AnalyseState(next, new SyntaxNode(enumerator.Current)));
+                if (enumerator.MoveNext())
+                {
+                    continue;
+                }
+                else
+                {
+                    throw new InvalidOperationException("Run out of token but not accept");
+                }
+            }
+
+            throw new InvalidOperationException("Failed to analyse input grammar");
+        }
+    }
+
+    private record AnalyseState(ITransformer State, SyntaxNode Node);
+}

Canon.Core/Abstractions/ITransformer.cs

@@ -0,0 +1,26 @@
+using Canon.Core.GrammarParser;
+
+namespace Canon.Core.Abstractions;
+
+/// <summary>
+/// 进行归约需要的信息
+/// </summary>
+/// <param name="Length">归约的长度</param>
+/// <param name="Left">归约得到的左部符号</param>
+public record ReduceInformation(int Length, NonTerminator Left);
+
+/// <summary>
+/// 状态的各种迁移信息
+/// </summary>
+public interface ITransformer
+{
+    /// <summary>
+    /// 进行移进的信息
+    /// </summary>
+    public IDictionary<TerminatorBase, ITransformer> ShiftTable { get; }
+
+    /// <summary>
+    /// 进行归约的信息
+    /// </summary>
+    public IDictionary<Terminator, ReduceInformation> ReduceTable { get; }
+}

Canon.Core/GrammarParser/Grammar.cs

@@ -1,94 +1,90 @@
-using Canon.Core.LexicalParser;
+using Canon.Core.Abstractions;
+using Canon.Core.LexicalParser;
 
 namespace Canon.Core.GrammarParser;
 
+/// <summary>
+/// 通过LR分析方法建立的语法
+/// </summary>
 public class Grammar
 {
+    /// <summary>
+    /// 起始符
+    /// </summary>
     public required NonTerminator Begin { get; init; }
 
+    /// <summary>
+    /// 语法中的DFA
+    /// </summary>
+    public required HashSet<LrState> Automation { get; init; }
+
+    /// <summary>
+    /// 起始状态
+    /// </summary>
     public required LrState BeginState { get; init; }
 
-    public SyntaxNode Analyse(IEnumerable<SemanticToken> tokens)
+    public GrammarParserBase ToGrammarParser()
     {
-        Stack<AnalyseState> stack = [];
-        stack.Push(new AnalyseState(BeginState, new SyntaxNode(SemanticToken.End)));
+        Dictionary<LrState, Transformer> transformers = [];
 
-        using IEnumerator<SemanticToken> enumerator = tokens.GetEnumerator();
-        if (!enumerator.MoveNext())
+        foreach (LrState state in Automation)
         {
-            throw new InvalidOperationException("Input token list is empty");
-        }
-
-        while (true)
-        {
-            AnalyseState top = stack.Peek();
-
-            // 尝试进行移进
-            bool acceptFlag = false, reduceFlag = false;
-            foreach (Expression e in top.State.Expressions)
+            ITransformer transformer;
+            if (transformers.TryGetValue(state, out Transformer? oldTransformer))
             {
-                if (e.Pos == e.Right.Count && e.LookAhead == enumerator.Current)
+                transformer = oldTransformer;
+            }
+            else
+            {
+                Transformer newTransformer = new();
+                transformers.Add(state, newTransformer);
+                transformer = newTransformer;
+            }
+
+            // 生成归约的迁移表
+            foreach (Expression expression in state.Expressions)
+            {
+                if (expression.Pos == expression.Right.Count)
                 {
-                    if (e.Left == Begin)
-                    {
-                        acceptFlag = true;
-                    }
-                    else
-                    {
-                        reduceFlag = true;
-                        SyntaxNode newNode = new(e.Left.Type);
-
-                        for (int i = 0; i < e.Right.Count; i++)
-                        {
-                            newNode.Children.Add(stack.Pop().Node);
-                        }
-
-                        stack.Push(new AnalyseState(stack.Peek().State.Transformer[e.Left],
-                            newNode));
-                    }
-                    break;
-                }
-
-                if (e.Right.Count == 0 && e.LookAhead == enumerator.Current)
-                {
-                    // 考虑空产生式的归约
-                    // 显然空产生式是不能accept的
-                    reduceFlag = true;
-                    SyntaxNode newNode = new(e.Left.Type);
-
-                    stack.Push(new AnalyseState(stack.Peek().State.Transformer[e.Left],
-                        newNode));
+                    transformer.ReduceTable.TryAdd(expression.LookAhead, new ReduceInformation(
+                        expression.Right.Count, expression.Left));
                 }
             }
 
-            if (acceptFlag)
+            // 生成移进的迁移表
+            foreach (KeyValuePair<TerminatorBase,LrState> pair in state.Transformer)
             {
-                // 接受文法 退出循环
-                return top.Node;
-            }
-
-            if (reduceFlag)
-            {
-                // 归约
-                continue;
-            }
-
-            // 尝试进行移进
-            if (top.State.Transformer.TryGetValue(enumerator.Current, out LrState? next))
-            {
-                stack.Push(new AnalyseState(next, new SyntaxNode(enumerator.Current)));
-                if (enumerator.MoveNext())
+                ITransformer targetTransformer;
+                if (transformers.TryGetValue(pair.Value, out Transformer? oldTransformer2))
                 {
-                    continue;
+                    targetTransformer = oldTransformer2;
                 }
                 else
                 {
-                    throw new InvalidOperationException("Run out of token but not accept");
+                    Transformer newTransformer = new();
+                    transformers.Add(pair.Value, newTransformer);
+                    targetTransformer = newTransformer;
                 }
+                transformer.ShiftTable.TryAdd(pair.Key, targetTransformer);
             }
-
-            throw new InvalidOperationException("Failed to analyse input grammar");
         }
+
+        return new GrammarParser(transformers[BeginState], Begin);
+    }
+
+    private class GrammarParser(ITransformer beginTransformer, NonTerminator begin) : GrammarParserBase
+    {
+        public override ITransformer BeginTransformer { get; } = beginTransformer;
+        public override NonTerminator Begin { get; } = begin;
+    }
+
+    private class Transformer : ITransformer
+    {
+        public IDictionary<TerminatorBase, ITransformer> ShiftTable { get; }
+            = new Dictionary<TerminatorBase, ITransformer>();
+
+        public IDictionary<Terminator, ReduceInformation> ReduceTable { get; }
+            = new Dictionary<Terminator, ReduceInformation>();
     }
 
     private record AnalyseState(LrState State, SyntaxNode Node);

Canon.Core/GrammarParser/GrammarBuilder.cs

@@ -300,7 +300,7 @@ public class GrammarBuilder
             Automation.UnionWith(addedStates);
         }
 
-        return new Grammar { Begin = Begin, BeginState = beginState };
+        return new Grammar { Begin = Begin, BeginState = beginState, Automation = Automation};
     }
 
     private static bool IsEmptyOnly(List<TerminatorBase> expression)

Canon.Tests/GrammarParserTests/SimpleGrammarTests.cs

@@ -1,4 +1,5 @@
-using Canon.Core.Enums;
+using Canon.Core.Abstractions;
+using Canon.Core.Enums;
 using Canon.Core.GrammarParser;
 using Canon.Core.LexicalParser;
 
@@ -118,6 +119,7 @@ public class SimpleGrammarTests
         };
 
         Grammar grammar = builder.Build();
+        GrammarParserBase parser = grammar.ToGrammarParser();
         // n + n
         List<SemanticToken> tokens =
         [
@@ -142,7 +144,7 @@ public class SimpleGrammarTests
         //   F     n
         //   |
         //   n
-        SyntaxNode root = grammar.Analyse(tokens);
+        SyntaxNode root = parser.Analyse(tokens);
         Assert.Equal(NonTerminatorType.ProgramStruct, root.GetNonTerminatorType());
         Assert.Equal(3, root.Children.Count);
         Assert.Contains(root.Children, node =>
@@ -168,6 +170,8 @@ public class SimpleGrammarTests
         };
 
         Grammar grammar = builder.Build();
+        GrammarParserBase parser = grammar.ToGrammarParser();
+
         // (n + n) * n
         List<SemanticToken> tokens =
         [
@@ -193,7 +197,8 @@ public class SimpleGrammarTests
             SemanticToken.End
         ];
 
-        SyntaxNode root = grammar.Analyse(tokens);
+
+        SyntaxNode root = parser.Analyse(tokens);
         Assert.Equal(18, root.Count());
         Assert.False(root.IsTerminated);
         Assert.Equal(NonTerminatorType.ProgramStruct, root.GetNonTerminatorType());

Canon.Tests/GrammarParserTests/SimpleGrammarWithEmptyTests.cs

@@ -1,4 +1,5 @@
-using Canon.Core.Enums;
+using Canon.Core.Abstractions;
+using Canon.Core.Enums;
 using Canon.Core.GrammarParser;
 using Canon.Core.LexicalParser;
 using Xunit.Abstractions;
@@ -151,6 +152,23 @@ public class SimpleGrammarWithEmptyTests(ITestOutputHelper testOutputHelper)
         _testOutputHelper.WriteLine(state5.ToString());
     }
 
+    [Fact]
+    public void ParserTest()
+    {
+        GrammarBuilder builder = new()
+        {
+            Generators = s_simpleGrammar, Begin = new NonTerminator(NonTerminatorType.StartNonTerminator)
+        };
+
+        Grammar grammar = builder.Build();
+        GrammarParserBase parser = grammar.ToGrammarParser();
+
+        ITransformer transformer1 = parser.BeginTransformer;
+        Assert.Equal(3, transformer1.ShiftTable.Count);
+        Assert.Single(transformer1.ReduceTable);
+        Assert.Contains(new NonTerminator(NonTerminatorType.ProgramStruct),transformer1.ShiftTable);
+    }
+
     [Fact]
     public void AnalyseSingleSentenceTest()
     {
@@ -160,6 +178,7 @@ public class SimpleGrammarWithEmptyTests(ITestOutputHelper testOutputHelper)
         };
 
         Grammar grammar = builder.Build();
+        GrammarParserBase parser = grammar.ToGrammarParser();
 
         List<SemanticToken> tokens =
         [
@@ -168,7 +187,7 @@ public class SimpleGrammarWithEmptyTests(ITestOutputHelper testOutputHelper)
             SemanticToken.End
         ];
 
-        SyntaxNode root = grammar.Analyse(tokens);
+        SyntaxNode root = parser.Analyse(tokens);
 
         Assert.False(root.IsTerminated);
         Assert.Equal(NonTerminatorType.ProgramStruct, root.GetNonTerminatorType());

Canon.Tests/GrammarParserTests/TerminatorTests.cs

@@ -27,11 +27,10 @@ public class TerminatorTests
     public void TerminatorAndKeywordSemanticTokenTest()
     {
         Terminator keywordTerminator = new(KeywordType.Array);
-        LinkedList<char> keywordContent = Utils.GetLinkedList("array [3..9] of integer");
-
-        Assert.True(KeywordSemanticToken.TryParse(0, 0, keywordContent.First!,
-            out KeywordSemanticToken? keywordSemanticToken));
-        Assert.NotNull(keywordSemanticToken);
+        KeywordSemanticToken keywordSemanticToken = new()
+        {
+            LinePos = 0, CharacterPos = 0, KeywordType = KeywordType.Array, LiteralValue = "array"
+        };
         Assert.True(keywordTerminator == keywordSemanticToken);
     }
 
@@ -39,11 +38,10 @@ public class TerminatorTests
     public void TerminatorAndDelimiterSemanticTokenTest()
     {
         Terminator terminator = new(DelimiterType.Period);
-        LinkedList<char> content = Utils.GetLinkedList(".");
-
-        Assert.True(DelimiterSemanticToken.TryParse(0, 0, content.First!,
-            out DelimiterSemanticToken? token));
-        Assert.NotNull(token);
+        DelimiterSemanticToken token = new()
+        {
+            LinePos = 0, CharacterPos = 0, DelimiterType = DelimiterType.Period, LiteralValue = "."
+        };
         Assert.True(token == terminator);
     }
 
@@ -51,44 +49,10 @@ public class TerminatorTests
     public void TerminatorAndOperatorSemanticTokenTest()
     {
         Terminator terminator = new(OperatorType.GreaterEqual);
-        LinkedList<char> content = Utils.GetLinkedList(">=");
-
-        Assert.True(OperatorSemanticToken.TryParse(0, 0, content.First!,
-            out OperatorSemanticToken? token));
-        Assert.NotNull(token);
+        OperatorSemanticToken token = new()
+        {
+            LinePos = 0, CharacterPos = 0, OperatorType = OperatorType.GreaterEqual, LiteralValue = ">="
+        };
         Assert.True(token == terminator);
     }
-
-    [Fact]
-    public void TerminatorAndNumberSemanticTokenTest()
-    {
-        LinkedList<char> content = Utils.GetLinkedList("123");
-
-        Assert.True(NumberSemanticToken.TryParse(0, 0, content.First!,
-            out NumberSemanticToken? token));
-        Assert.NotNull(token);
-        Assert.True(Terminator.NumberTerminator == token);
-    }
-
-    [Fact]
-    public void TerminatorAndCharacterSemanticTokenTest()
-    {
-        LinkedList<char> content = Utils.GetLinkedList("'a'");
-
-        Assert.True(CharacterSemanticToken.TryParse(0, 0, content.First!,
-            out CharacterSemanticToken? token));
-        Assert.NotNull(token);
-        Assert.True(Terminator.CharacterTerminator == token);
-    }
-
-    [Fact]
-    public void TerminatorAndIdentifierSemanticTokenTest()
-    {
-        LinkedList<char> content = Utils.GetLinkedList("gcd");
-
-        Assert.True(IdentifierSemanticToken.TryParse(0, 0, content.First!,
-            out IdentifierSemanticToken? token));
-        Assert.NotNull(token);
-        Assert.True(Terminator.IdentifierTerminator == token);
-    }
 }

Canon.Tests/Utils.cs

@@ -1,4 +1,6 @@
-namespace Canon.Tests;
+using Canon.Core.GrammarParser;
+
+namespace Canon.Tests;
 
 public static class Utils
 {
@@ -13,4 +15,25 @@ public static class Utils
 
         return list;
     }
+
+    /// <summary>
+    /// 验证两棵语法树一致
+    /// </summary>
+    /// <param name="a">一棵语法树</param>
+    /// <param name="b">另一棵语法树</param>
+    public static void CheckSyntaxRoot(SyntaxNode a, SyntaxNode b)
+    {
+        int length = a.Count();
+        Assert.Equal(length, b.Count());
+
+        using IEnumerator<SyntaxNode> aIter = a.GetEnumerator(), bIter = b.GetEnumerator();
+
+        for (int i = 0; i < length; i++)
+        {
+            Assert.True(aIter.MoveNext());
+            Assert.True(bIter.MoveNext());
+
+            Assert.Equal(aIter.Current, bIter.Current);
+        }
+    }
 }