2.6 Pratt Parser (infix)

Author: cedrickchee

ast/ast.go

@@ -205,3 +205,29 @@ func (pe *PrefixExpression) String() string {
 
 	return out.String()
 }
+
+// InfixExpression represents an infix expression node.
+type InfixExpression struct {
+	Token    token.Token // The operator token, e.g. +
+	Left     Expression
+	Operator string
+	Right    Expression
+}
+
+func (oe *InfixExpression) expressionNode() {}
+
+// TokenLiteral prints the literal value of the token associated with this node.
+func (oe *InfixExpression) TokenLiteral() string { return oe.Token.Literal }
+
+// String returns a stringified version of the expression node.
+func (oe *InfixExpression) String() string {
+	var out bytes.Buffer
+
+	out.WriteString("(")
+	out.WriteString(oe.Left.String())
+	out.WriteString(" " + oe.Operator + " ")
+	out.WriteString(oe.Right.String())
+	out.WriteString(")")
+
+	return out.String()
+}

parser/parser.go

@@ -27,6 +27,19 @@ const (
 	CALL            // myFunction(X)
 )
 
+// Precedence table.
+// It associates token types with their precedence.
+var precedences = map[token.TokenType]int{
+	token.EQ:       EQUALS,
+	token.NOT_EQ:   EQUALS,
+	token.LT:       LESSGREATER,
+	token.GT:       LESSGREATER,
+	token.PLUS:     SUM,
+	token.MINUS:    SUM,
+	token.SLASH:    PRODUCT,
+	token.ASTERISK: PRODUCT,
+}
+
 // Pratt parser's idea is the association of parsing functions with token types.
 // Whenever this token type is encountered, the parsing functions are called to
 // parse the appropriate expression and return an AST node that represents it.
@@ -68,6 +81,16 @@ func New(l *lexer.Lexer) *Parser {
 	p.registerPrefix(token.BANG, p.parsePrefixExpression)
 	p.registerPrefix(token.MINUS, p.parsePrefixExpression)
 
+	p.infixParseFns = make(map[token.TokenType]infixParseFn)
+	p.registerInfix(token.PLUS, p.parseInfixExpression)
+	p.registerInfix(token.MINUS, p.parseInfixExpression)
+	p.registerInfix(token.SLASH, p.parseInfixExpression)
+	p.registerInfix(token.ASTERISK, p.parseInfixExpression)
+	p.registerInfix(token.EQ, p.parseInfixExpression)
+	p.registerInfix(token.NOT_EQ, p.parseInfixExpression)
+	p.registerInfix(token.LT, p.parseInfixExpression)
+	p.registerInfix(token.GT, p.parseInfixExpression)
+
 	// Read two tokens, so curToken and peekToken are both set.
 	p.nextToken()
 	p.nextToken()
@@ -192,6 +215,21 @@ func (p *Parser) parseExpression(precedence int) ast.Expression {
 
 	leftExp := prefix()
 
+	// The heart of our Pratt parser.
+	for !p.peekTokenIs(token.SEMICOLON) && precedence < p.peekPrecedence() {
+		// Try to find infixParseFns for the next token.
+		infix := p.infixParseFns[p.peekToken.Type]
+		if infix == nil {
+			return leftExp
+		}
+
+		p.nextToken()
+
+		leftExp = infix(leftExp)
+
+		// Loop until it encounters a token that has a higher precedence.
+	}
+
 	return leftExp
 }
 
@@ -243,6 +281,21 @@ func (p *Parser) parsePrefixExpression() ast.Expression {
 	return expression
 }
 
+func (p *Parser) parseInfixExpression(left ast.Expression) ast.Expression {
+	expression := &ast.InfixExpression{
+		Token:    p.curToken, // the operator of the infix expression
+		Operator: p.curToken.Literal,
+		Left:     left,
+	}
+
+	// Precedence of the operator token.
+	precedence := p.curPrecedence()
+	p.nextToken()
+	expression.Right = p.parseExpression(precedence)
+
+	return expression
+}
+
 // "assertion functions".
 // Enforce the correctness of the order of tokens by checking the type of the
 // next token.
@@ -272,3 +325,21 @@ func (p *Parser) registerPrefix(tokenType token.TokenType, fn prefixParseFn) {
 func (p *Parser) registerInfix(tokenType token.TokenType, fn infixParseFn) {
 	p.infixParseFns[tokenType] = fn
 }
+
+// Returns the precedence associated with the token type of peekToken.
+func (p *Parser) peekPrecedence() int {
+	if p, ok := precedences[p.peekToken.Type]; ok {
+		return p
+	}
+
+	return LOWEST
+}
+
+// Returns the precedence associated with the token type of curToken.
+func (p *Parser) curPrecedence() int {
+	if p, ok := precedences[p.curToken.Type]; ok {
+		return p
+	}
+
+	return LOWEST
+}

parser/parser_test.go

@@ -182,6 +182,140 @@ func TestParsingPrefixExpression(t *testing.T) {
 	}
 }
 
+func TestParsingInfixExpressions(t *testing.T) {
+	// Test parsing infix operators.
+	//  e.g.: `5 + 5;`
+	// As with prefix operator expressions, we can use any expressions to the
+	// left and right of the operator:
+	// <expression> <infix operator> <expression>
+
+	infixTests := []struct {
+		input      string
+		leftValue  int64
+		operator   string
+		rightValue int64
+	}{
+		{"5 + 5;", 5, "+", 5},
+		{"5 - 5;", 5, "-", 5},
+		{"5 * 5;", 5, "*", 5},
+		{"5 / 5;", 5, "/", 5},
+		{"5 > 5;", 5, ">", 5},
+		{"5 < 5;", 5, "<", 5},
+		{"5 == 5;", 5, "==", 5},
+		{"5 != 5;", 5, "!=", 5},
+	}
+
+	for _, tt := range infixTests {
+		l := lexer.New(tt.input)
+		p := New(l)
+		program := p.ParseProgram()
+		checkParserErrors(t, p)
+
+		if len(program.Statements) != 1 {
+			t.Fatalf("program.Statements does not contain %d statements. got=%d\n",
+				1, len(program.Statements))
+		}
+
+		stmt, ok := program.Statements[0].(*ast.ExpressionStatement)
+		if !ok {
+			t.Fatalf("program.Statements[0] is not ast.ExpressionStatement. got=%T",
+				program.Statements[0])
+		}
+
+		exp, ok := stmt.Expression.(*ast.InfixExpression)
+		if !ok {
+			t.Fatalf("exp is not ast.InfixExpression. got=%T", stmt.Expression)
+		}
+
+		if !testIntegerLiteral(t, exp.Left, tt.leftValue) {
+			return
+		}
+
+		if exp.Operator != tt.operator {
+			t.Fatalf("exp.Operator is not '%s'. got=%s",
+				tt.operator, exp.Operator)
+		}
+
+		if !testIntegerLiteral(t, exp.Right, tt.rightValue) {
+			return
+		}
+	}
+}
+
+func TestOperatorPrecedenceParsing(t *testing.T) {
+	// Tests that use multiple operators with different precedences and how the
+	// AST in string form correctly represents this.
+	tests := []struct {
+		input    string
+		expected string
+	}{
+		{
+			"-a * b",
+			"((-a) * b)",
+		},
+		{
+			"!-a",
+			"(!(-a))",
+		},
+		{
+			"a + b + c",
+			"((a + b) + c)",
+		},
+		{
+			"a + b - c",
+			"((a + b) - c)",
+		},
+		{
+			"a * b * c",
+			"((a * b) * c)",
+		},
+		{
+			"a * b / c",
+			"((a * b) / c)",
+		},
+		{
+			"a + b / c",
+			"(a + (b / c))",
+		},
+		{
+			"a + b * c + d / e - f",
+			"(((a + (b * c)) + (d / e)) - f)",
+		},
+		{
+			"3 + 4; -5 * 5",
+			"(3 + 4)((-5) * 5)",
+		},
+		{
+			"5 > 4 == 3 < 4",
+			"((5 > 4) == (3 < 4))",
+		},
+		{
+			"5 < 4 != 3 > 4",
+			"((5 < 4) != (3 > 4))",
+		},
+		{
+			"3 + 4 * 5 == 3 * 1 + 4 * 5",
+			"((3 + (4 * 5)) == ((3 * 1) + (4 * 5)))",
+		},
+		{
+			"3 + 4 * 5 == 3 * 1 + 4 * 5",
+			"((3 + (4 * 5)) == ((3 * 1) + (4 * 5)))",
+		},
+	}
+
+	for _, tt := range tests {
+		l := lexer.New(tt.input)
+		p := New(l)
+		program := p.ParseProgram()
+		checkParserErrors(t, p)
+
+		actual := program.String()
+		if actual != tt.expected {
+			t.Errorf("expected=%q, got=%q", tt.expected, actual)
+		}
+	}
+}
+
 func testIntegerLiteral(t *testing.T, il ast.Expression, value int64) bool {
 	integ, ok := il.(*ast.IntegerLiteral)
 	if !ok {