Improved C# lexer performance.

git-svn-id: svn://svn.sharpdevelop.net/sharpdevelop/trunk@907 1ccf3a8d-04fe-1044-b7c0-cef0b8235c61

src/Libraries/NRefactory/Project/NRefactory.csproj

@@ -15,6 +15,12 @@
     <OutputType>Library</OutputType>
     <SignAssembly>true</SignAssembly>
     <AssemblyOriginatorKeyFile>Resources\ICSharpCode.NRefactory.snk</AssemblyOriginatorKeyFile>
+    <AllowUnsafeBlocks>True</AllowUnsafeBlocks>
+    <RegisterForComInterop>False</RegisterForComInterop>
+    <GenerateSerializationAssemblies>Auto</GenerateSerializationAssemblies>
+    <BaseAddress>4194304</BaseAddress>
+    <PlatformTarget>AnyCPU</PlatformTarget>
+    <FileAlignment>4096</FileAlignment>
   </PropertyGroup>
   <PropertyGroup Condition=" '$(Configuration)|$(Platform)' == 'Debug|AnyCPU' ">
     <Optimize>False</Optimize>

src/Libraries/NRefactory/Project/Src/Lexer/AbstractLexer.cs

@@ -37,6 +37,17 @@ namespace ICSharpCode.NRefactory.Parser
 		// used for the original value of strings (with escape sequences).
 		protected StringBuilder originalValue = new StringBuilder();
 		
+		protected bool skipAllComments = false;
+		
+		public bool SkipAllComments {
+			get {
+				return skipAllComments;
+			}
+			set {
+				skipAllComments = value;
+			}
+		}
+		
 		protected int Line {
 			get {
 				return line;
@@ -170,7 +181,7 @@ namespace ICSharpCode.NRefactory.Parser
 			if (curToken == null) {
 				curToken = Next();
 				specialTracker.InformToken(curToken.kind);
-//				Console.WriteLine(ICSharpCode.NRefactory.Parser.CSharp.Tokens.GetTokenString(curToken.kind) + " -- " + curToken.val + "(" + curToken.kind + ")");
+				//Console.WriteLine(ICSharpCode.NRefactory.Parser.CSharp.Tokens.GetTokenString(curToken.kind) + " -- " + curToken.val + "(" + curToken.kind + ")");
 				return curToken;
 			}
 			
@@ -184,7 +195,7 @@ namespace ICSharpCode.NRefactory.Parser
 			}
 			
 			curToken  = curToken.next;
-//			Console.WriteLine(ICSharpCode.NRefactory.Parser.CSharp.Tokens.GetTokenString(curToken.kind) + " -- " + curToken.val + "(" + curToken.kind + ")");
+			//Console.WriteLine(ICSharpCode.NRefactory.Parser.CSharp.Tokens.GetTokenString(curToken.kind) + " -- " + curToken.val + "(" + curToken.kind + ")");
 			return curToken;
 		}
 		
@@ -201,9 +212,17 @@ namespace ICSharpCode.NRefactory.Parser
 			throw new NotSupportedException();
 		}
 		
-		protected bool IsIdentifierPart(char ch)
+		protected bool IsIdentifierPart(int ch)
 		{
-			return Char.IsLetterOrDigit(ch) || ch == '_';
+			// char.IsLetter is slow, so optimize for raw ASCII
+			if (ch < 48)  return false; // 48 = '0'
+			if (ch <= 57) return true;  // 57 = '9'
+			if (ch < 65)  return false; // 65 = 'A'
+			if (ch <= 90) return true;  // 90 = 'Z'
+			if (ch == 95) return true;  // 95 = '_'
+			if (ch < 97)  return false; // 97 = 'a'
+			if (ch <= 122) return true; // 97 = 'z'
+			return char.IsLetter((char)ch); // accept unicode letters
 		}
 		
 		protected bool IsHex(char digit)
@@ -250,6 +269,16 @@ namespace ICSharpCode.NRefactory.Parser
 			return false;
 		}
 		
+		protected void SkipToEOL()
+		{
+			int nextChar;
+			while ((nextChar = reader.Read()) != -1) {
+				if (HandleLineEnd((char)nextChar)) {
+					break;
+				}
+			}
+		}
+		
 		protected string ReadToEOL()
 		{
 			sb.Length = 0;
@@ -259,7 +288,7 @@ namespace ICSharpCode.NRefactory.Parser
 				
 				// Return read string, if EOL is reached
 				if (HandleLineEnd(ch)) {
-					return sb.ToString();;
+					return sb.ToString();
 				}
 				
 				sb.Append(ch);

src/Libraries/NRefactory/Project/Src/Lexer/CSharp/Lexer.cs

@@ -27,6 +27,8 @@ namespace ICSharpCode.NRefactory.Parser.CSharp
 			int nextChar;
 			while ((nextChar = ReaderRead()) != -1) {
 				char ch = (char)nextChar;
+				if (ch == ' ' || ch == '\t')
+					continue;
 				
 				if (Char.IsWhiteSpace(ch)) {
 					HandleLineEnd(ch);
@@ -112,17 +114,17 @@ namespace ICSharpCode.NRefactory.Parser.CSharp
 		
 		string ReadIdent(char ch)
 		{
+			int peek;
 			int curPos     = 1;
 			identBuffer[0] = ch;
-			int peek;
-			while ((peek = ReaderPeek()) != -1 && (Char.IsLetterOrDigit(ch = (char)peek) || ch == '_')) {
+			while (IsIdentifierPart(peek = ReaderPeek())) {
 				ReaderRead();
 				
 				if (curPos < MAX_IDENTIFIER_LENGTH) {
-					identBuffer[curPos++] = ch;
+					identBuffer[curPos++] = (char)peek;
 				} else {
 					errors.Error(Line, Col, String.Format("Identifier too long"));
-					while ((peek = ReaderPeek()) != -1 && (Char.IsLetterOrDigit(ch = (char)peek) || ch == '_')) {
+					while (IsIdentifierPart(ReaderPeek())) {
 						ReaderRead();
 					}
 					break;
@@ -160,7 +162,7 @@ namespace ICSharpCode.NRefactory.Parser.CSharp
 					ReaderRead(); // skip 'x'
 					sb.Length = 0; // Remove '0' from 0x prefix from the stringvalue
 					while (IsHex((char)ReaderPeek())) {
-						sb.Append(Char.ToUpper((char)ReaderRead(), CultureInfo.InvariantCulture));
+						sb.Append((char)ReaderRead());
 					}
 					if (sb.Length == 0) {
 						sb.Append('0'); // dummy value to prevent exception
@@ -176,7 +178,7 @@ namespace ICSharpCode.NRefactory.Parser.CSharp
 					peek = (char)ReaderPeek();
 				}
 				
-				Token nextToken = null; // if we accedently read a 'dot'
+				Token nextToken = null; // if we accidently read a 'dot'
 				if (peek == '.') { // read floating point number
 					ReaderRead();
 					peek = (char)ReaderPeek();
@@ -258,7 +260,7 @@ namespace ICSharpCode.NRefactory.Parser.CSharp
 				}
 				if (isdecimal) {
 					try {
-						return new Token(Tokens.Literal, x, y, stringValue, Decimal.Parse(digit, CultureInfo.InvariantCulture));
+						return new Token(Tokens.Literal, x, y, stringValue, Decimal.Parse(digit, NumberStyles.Any, CultureInfo.InvariantCulture));
 					} catch (Exception) {
 						errors.Error(y, x, String.Format("Can't parse decimal {0}", digit));
 						return new Token(Tokens.Literal, x, y, stringValue, 0m);
@@ -702,8 +704,11 @@ namespace ICSharpCode.NRefactory.Parser.CSharp
 		
 		string ReadCommentToEOL()
 		{
+			if (specialCommentHash == null) {
+				return ReadToEOL();
+			}
 			sb.Length = 0;
-			StringBuilder curWord = specialCommentHash != null ? new StringBuilder() : null;
+			StringBuilder curWord = new StringBuilder();
 			
 			int nextChar;
 			while ((nextChar = ReaderRead()) != -1) {
@@ -714,19 +719,17 @@ namespace ICSharpCode.NRefactory.Parser.CSharp
 				}
 				
 				sb.Append(ch);
-				if (specialCommentHash != null) {
-					if (Char.IsLetter(ch)) {
-						curWord.Append(ch);
-					} else {
-						string tag = curWord.ToString();
-						curWord.Length = 0;
-						if (specialCommentHash.ContainsKey(tag)) {
-							Point p = new Point(Col, Line);
-							string comment = ch + ReadToEOL();
-							tagComments.Add(new TagComment(tag, comment, p, new Point(Col, Line)));
-							sb.Append(comment);
-							break;
-						}
+				if (IsIdentifierPart(nextChar)) {
+					curWord.Append(ch);
+				} else {
+					string tag = curWord.ToString();
+					curWord.Length = 0;
+					if (specialCommentHash.ContainsKey(tag)) {
+						Point p = new Point(Col, Line);
+						string comment = ch + ReadToEOL();
+						tagComments.Add(new TagComment(tag, comment, p, new Point(Col, Line)));
+						sb.Append(comment);
+						break;
 					}
 				}
 			}
@@ -735,32 +738,46 @@ namespace ICSharpCode.NRefactory.Parser.CSharp
 		
 		void ReadSingleLineComment(CommentType commentType)
 		{
-			specialTracker.StartComment(commentType, new Point(Col, Line));
-			specialTracker.AddString(ReadCommentToEOL());
-			specialTracker.FinishComment(new Point(Col, Line));
+			if (skipAllComments) {
+				SkipToEOL();
+			} else {
+				specialTracker.StartComment(commentType, new Point(Col, Line));
+				specialTracker.AddString(ReadCommentToEOL());
+				specialTracker.FinishComment(new Point(Col, Line));
+			}
 		}
 		
 		void ReadMultiLineComment()
 		{
-			specialTracker.StartComment(CommentType.Block, new Point(Col, Line));
 			int nextChar;
-			while ((nextChar = ReaderRead()) != -1) {
-				char ch = (char)nextChar;
-				
-				if (HandleLineEnd(ch)) {
-					specialTracker.AddChar('\n');
-					continue;
+			if (skipAllComments) {
+				while ((nextChar = ReaderRead()) != -1) {
+					char ch = (char)nextChar;
+					if (ch == '*' && ReaderPeek() == '/') {
+						ReaderRead();
+						return;
+					}
 				}
-				
-				// End of multiline comment reached ?
-				if (ch == '*' && ReaderPeek() == '/') {
-					ReaderRead();
-					specialTracker.FinishComment(new Point(Col, Line));
-					return;
+			} else {
+				specialTracker.StartComment(CommentType.Block, new Point(Col, Line));
+				while ((nextChar = ReaderRead()) != -1) {
+					char ch = (char)nextChar;
+					
+					if (HandleLineEnd(ch)) {
+						specialTracker.AddChar('\n');
+						continue;
+					}
+					
+					// End of multiline comment reached ?
+					if (ch == '*' && ReaderPeek() == '/') {
+						ReaderRead();
+						specialTracker.FinishComment(new Point(Col, Line));
+						return;
+					}
+					specialTracker.AddChar(ch);
 				}
-				specialTracker.AddChar(ch);
+				specialTracker.FinishComment(new Point(Col, Line));
 			}
-			specialTracker.FinishComment(new Point(Col, Line));
 			// Reached EOF before end of multiline comment.
 			errors.Error(Line, Col, String.Format("Reached EOF before the end of a multiline comment"));
 		}
@@ -774,16 +791,58 @@ namespace ICSharpCode.NRefactory.Parser.CSharp
 		public override void SkipCurrentBlock()
 		{
 			int braceCount = 0;
-			Token t;
-			while ((t = LookAhead).kind != Tokens.EOF) {
-				if (t.kind == Tokens.OpenCurlyBrace) {
+			while (curToken != null) {
+				if (curToken.kind == Tokens.OpenCurlyBrace) {
 					++braceCount;
-				} else if (t.kind == Tokens.CloseCurlyBrace) {
+				} else if (curToken.kind == Tokens.CloseCurlyBrace) {
 					if (--braceCount < 0)
 						return;
 				}
-				NextToken();
+				lastToken = curToken;
+				curToken = curToken.next;
+			}
+			int nextChar;
+			while ((nextChar = ReaderRead()) != -1) {
+				switch (nextChar) {
+					case '{':
+						braceCount++;
+						break;
+					case '}':
+						if (--braceCount < 0) {
+							curToken = new Token(Tokens.CloseCurlyBrace, Col, Line);
+							return;
+						}
+						break;
+					case '/':
+						int peek = ReaderPeek();
+						if (peek == '/' || peek == '*') {
+							ReadComment();
+						}
+						break;
+					case '#':
+						SkipToEOL();
+						break;
+					case '"':
+						ReadString();
+						break;
+					case '\'':
+						ReadChar();
+						break;
+					case '\r':
+					case '\n':
+						HandleLineEnd((char)nextChar);
+						break;
+					case '@':
+						int next = ReaderRead();
+						if (next == -1) {
+							errors.Error(Line, Col, String.Format("EOF after @"));
+						} else if (next == '"') {
+							ReadVerbatimString();
+						}
+						break;
+				}
 			}
+			curToken = new Token(Tokens.EOF, Col, Line);
 		}
 	}
 }

src/Libraries/NRefactory/Project/Src/Lexer/ILexer.cs

@@ -42,6 +42,15 @@ namespace ICSharpCode.NRefactory.Parser
 			set;
 		}
 		
+		/// <summary>
+		/// Gets/Sets if the lexer should skip adding comments to the special tracker. Set this
+		/// property to true to improve lexing performance.
+		/// </summary>
+		bool SkipAllComments {
+			get;
+			set;
+		}
+		
 		/// <summary>
 		/// Returns the comments that had been read and containing tag key words.
 		/// </summary>

src/Libraries/NRefactory/Project/Src/Lexer/Special/SpecialTracker.cs

@@ -57,7 +57,7 @@ namespace ICSharpCode.NRefactory.Parser
 		{
 			this.currentCommentType = commentType;
 			this.startPosition      = startPosition;
-			this.sb.Length          = 0; 
+			this.sb.Length          = 0;
 		}
 		
 		public void AddChar(char c)

src/Libraries/NRefactory/Project/Src/Parser/AbstractParser.cs

@@ -59,7 +59,6 @@ namespace ICSharpCode.NRefactory.Parser
 			this.errors = lexer.Errors;
 			this.lexer  = lexer;
 			errors.SynErr = new ErrorCodeProc(SynErr);
-			lexer.NextToken();			
 		}
 		
 		public abstract void Parse();

src/Libraries/NRefactory/Project/Src/Parser/CSharp/cs.ATG

@@ -44,6 +44,7 @@ public void Error(string s)
 
 public override Expression ParseExpression()
 {
+	lexer.NextToken();
 	Expression expr;
 	Expr(out expr);
 	return expr;
@@ -648,7 +649,8 @@ PRODUCTIONS
 
 /*--- compilation unit: */
 CS
-(. compilationUnit = new CompilationUnit(); .)
+(. lexer.NextToken(); /* get the first token */
+   compilationUnit = new CompilationUnit(); .)
 =
 	{ UsingDirective }
 	{ IF (IsGlobalAttrTarget()) GlobalAttributeSection }

src/Libraries/NRefactory/Project/Src/Parser/VBNet/VBNET.ATG

@@ -62,6 +62,7 @@ public void Error(string s)
 
 public override Expression ParseExpression()
 {
+	lexer.NextToken();
 	Expression expr;
 	Expr(out expr);
 	return expr;
@@ -476,6 +477,7 @@ PRODUCTIONS
 
 VBNET
 	(.
+		lexer.NextToken(); // get the first token
 		compilationUnit = new CompilationUnit();
 		withStatements = new Stack();
 	.) =

src/Libraries/NRefactory/Test/Lexer/CSharp/CustomLexerTests.cs

@@ -32,6 +32,15 @@ namespace ICSharpCode.NRefactory.Tests.Lexer.CSharp
 			Assert.AreEqual(Tokens.EOF, lexer.NextToken().kind);
 		}
 		
+		[Test]
+		public void TestIdentifier()
+		{
+			ILexer lexer = GenerateLexer(new StringReader("a_Bc05"));
+			Token t = lexer.NextToken();
+			Assert.AreEqual(Tokens.Identifier, t.kind);
+			Assert.AreEqual("a_Bc05", t.val);
+		}
+		
 		[Test]
 		public void TestSkippedEmptyBlock()
 		{
@@ -55,5 +64,38 @@ namespace ICSharpCode.NRefactory.Tests.Lexer.CSharp
 			Assert.AreEqual(Tokens.Plus, lexer.NextToken().kind);
 			Assert.AreEqual(Tokens.EOF, lexer.NextToken().kind);
 		}
+		
+		[Test]
+		public void TestSkippedNonEmptyBlockWithPeek()
+		{
+			ILexer lexer = GenerateLexer(new StringReader("{ TestMethod(\"}\"); // }}}\n" +
+			                                              "while(1) {break;} }+"));
+			Assert.AreEqual(Tokens.OpenCurlyBrace, lexer.NextToken().kind);
+			lexer.NextToken();
+			lexer.StartPeek();
+			lexer.Peek();
+			lexer.Peek();
+			lexer.Peek();
+			lexer.SkipCurrentBlock();
+			Assert.AreEqual(Tokens.CloseCurlyBrace, lexer.LookAhead.kind);
+			Assert.AreEqual(Tokens.Plus, lexer.NextToken().kind);
+			Assert.AreEqual(Tokens.EOF, lexer.NextToken().kind);
+		}
+		
+		[Test]
+		public void TestSkippedEmptyBlockWithPeek()
+		{
+			ILexer lexer = GenerateLexer(new StringReader("{}+"));
+			Assert.AreEqual(Tokens.OpenCurlyBrace, lexer.NextToken().kind);
+			lexer.NextToken();
+			lexer.StartPeek();
+			lexer.Peek();
+			lexer.Peek();
+			lexer.Peek();
+			lexer.SkipCurrentBlock();
+			Assert.AreEqual(Tokens.CloseCurlyBrace, lexer.LookAhead.kind);
+			Assert.AreEqual(Tokens.Plus, lexer.NextToken().kind);
+			Assert.AreEqual(Tokens.EOF, lexer.NextToken().kind);
+		}
 	}
 }

src/Libraries/NRefactory/Test/Lexer/CSharp/NumberLexerTest.cs

@@ -27,6 +27,7 @@ namespace ICSharpCode.NRefactory.Tests.Lexer.CSharp
 			ILexer lexer = GenerateLexer(new StringReader(text));
 			Token t = lexer.NextToken();
 			Assert.AreEqual(Tokens.EOF, lexer.NextToken().kind, "Tokens.EOF");
+			Assert.AreEqual("", lexer.Errors.ErrorOutput);
 			return t;
 		}
 		
@@ -34,6 +35,7 @@ namespace ICSharpCode.NRefactory.Tests.Lexer.CSharp
 		{
 			Token t = GetSingleToken(text);
 			Assert.AreEqual(Tokens.Literal, t.kind, "Tokens.Literal");
+			Assert.AreEqual(text, t.val, "value");
 			Assert.IsNotNull(t.literalValue, "literalValue is null");
 			Assert.AreEqual(val, t.literalValue, "literalValue");
 		}
@@ -90,12 +92,27 @@ namespace ICSharpCode.NRefactory.Tests.Lexer.CSharp
 		public void TestDouble()
 		{
 			CheckToken("1.0", 1.0);
+			CheckToken("1.1", 1.1);
+			CheckToken("1.1e-2", 1.1e-2);
 		}
 		
 		[Test]
 		public void TestFloat()
 		{
+			CheckToken("1f", 1f);
 			CheckToken("1.0f", 1.0f);
+			CheckToken("1.1f", 1.1f);
+			CheckToken("1.1e-2f", 1.1e-2f);
+		}
+		
+		[Test]
+		public void TestDecimal()
+		{
+			CheckToken("1m", 1m);
+			CheckToken("1.0m", 1.0m);
+			CheckToken("1.1m", 1.1m);
+			CheckToken("1.1e-2m", 1.1e-2m);
+			CheckToken("2.0e-5m", 2.0e-5m);
 		}
 	}
 }

src/Libraries/NRefactory/Test/Lexer/VBNet/LiteralsTests.cs

@@ -28,6 +28,7 @@ namespace ICSharpCode.NRefactory.Tests.Lexer.VB
 			Token t = lexer.NextToken();
 			Assert.AreEqual(Tokens.EOL, lexer.NextToken().kind, "Tokens.EOL");
 			Assert.AreEqual(Tokens.EOF, lexer.NextToken().kind, "Tokens.EOF");
+			Assert.AreEqual("", lexer.Errors.ErrorOutput);
 			return t;
 		}