Parsing identifiers, integer literals and prefix expressions (all as expression statements)

Signed-off-by: jmug <u.g.a.mariano@gmail.com>

pkg/ast/ast.go

@@ -1,9 +1,14 @@
 package ast
 
-import "code.jmug.me/jmug/interpreter-in-go/pkg/token"
+import (
+	"bytes"
+
+	"code.jmug.me/jmug/interpreter-in-go/pkg/token"
+)
 
 type Node interface {
 	TokenLiteral() string
+	String() string
 }
 
 type Statement interface {
@@ -27,6 +32,14 @@ func (p *Program) TokenLiteral() string {
 	return ""
 }
 
+func (p *Program) String() string {
+	var out bytes.Buffer
+	for _, stmt := range p.Statements {
+		out.WriteString(stmt.String())
+	}
+	return out.String()
+}
+
 type LetStatement struct {
 	Token token.Token // TODO: This is a little redundant, figure out if I can get rid of it.
 	Name  *Identifier
@@ -37,6 +50,16 @@ func (ls *LetStatement) statementNode() {}
 func (ls *LetStatement) TokenLiteral() string {
 	return ls.Token.Literal
 }
+func (ls *LetStatement) String() string {
+	var out bytes.Buffer
+	out.WriteString(ls.TokenLiteral() + " ")
+	out.WriteString(ls.Name.String() + " = ")
+	if ls.Value != nil {
+		out.WriteString(ls.Value.String())
+	}
+	out.WriteString(";")
+	return out.String()
+}
 
 type ReturnStatement struct {
 	Token       token.Token // TODO: This is a little redundant, figure out if I can get rid of it.
@@ -47,7 +70,40 @@ func (rs *ReturnStatement) statementNode() {}
 func (rs *ReturnStatement) TokenLiteral() string {
 	return rs.Token.Literal
 }
+func (rs *ReturnStatement) String() string {
+	var out bytes.Buffer
+	out.WriteString(rs.TokenLiteral())
+	if rs.ReturnValue != nil {
+		out.WriteString(" " + rs.ReturnValue.String())
+	}
+	out.WriteString(";")
+	return out.String()
+}
 
+// ExpressionStatement is a simple wrapper of an expression in a statement
+// This is common in scripting languages and allows you to have a source line
+// that is solely an expression, think of the Python REPL and how you can
+// type `1 + 1` and get a result.
+type ExpressionStatement struct {
+	Token      token.Token // The first token in the expression.
+	Expression Expression
+}
+
+func (es *ExpressionStatement) statementNode() {}
+func (es *ExpressionStatement) TokenLiteral() string {
+	return es.Token.Literal
+}
+func (es *ExpressionStatement) String() string {
+	if es.Expression != nil {
+		return es.Expression.String()
+	}
+	return ""
+}
+
+// Identifier is treated as an expression because in certain
+// circumstances they can return values (think `let some = other` where `other`
+// is actually an expression returning a value) and this makes them easier to
+// handle (according to the author).
 type Identifier struct {
 	Token token.Token
 	Value string
@@ -57,3 +113,33 @@ func (i *Identifier) expressionNode() {}
 func (i *Identifier) TokenLiteral() string {
 	return i.Token.Literal
 }
+func (i *Identifier) String() string {
+	return i.Value
+}
+
+type IntegerLiteral struct {
+	Token token.Token
+	Value int64
+}
+
+func (il *IntegerLiteral) expressionNode() {}
+func (il *IntegerLiteral) TokenLiteral() string {
+	return il.Token.Literal
+}
+func (il *IntegerLiteral) String() string {
+	return il.Token.Literal
+}
+
+type PrefixExpression struct {
+	Token    token.Token
+	Operator string
+	Right    Expression
+}
+
+func (pe *PrefixExpression) expressionNode() {}
+func (pe *PrefixExpression) TokenLiteral() string {
+	return pe.Token.Literal
+}
+func (pe *PrefixExpression) String() string {
+	return "(" + pe.Operator + pe.Right.String() + ")"
+}

pkg/ast/ast_test.go

@@ -0,0 +1,29 @@
+package ast
+
+import (
+	"testing"
+
+	"code.jmug.me/jmug/interpreter-in-go/pkg/token"
+)
+
+func TestString(t *testing.T) {
+	program := &Program{
+		Statements: []Statement{
+			&LetStatement{
+				Token: token.Token{Type: token.LET, Literal: "let"},
+				Name: &Identifier{
+					Token: token.Token{Type: token.IDENT, Literal: "myVar"},
+					Value: "myVar",
+				},
+				Value: &Identifier{
+					Token: token.Token{Type: token.IDENT, Literal: "anotherVar"},
+					Value: "anotherVar",
+				},
+			},
+		},
+	}
+
+	if program.String() != "let myVar = anotherVar;" {
+		t.Errorf("program.String() wrong. got=%q", program.String())
+	}
+}

pkg/parser/parser.go

@@ -2,24 +2,38 @@ package parser
 
 import (
 	"fmt"
+	"strconv"
 
 	"code.jmug.me/jmug/interpreter-in-go/pkg/ast"
 	"code.jmug.me/jmug/interpreter-in-go/pkg/lexer"
 	"code.jmug.me/jmug/interpreter-in-go/pkg/token"
 )
 
+type (
+	prefixParseFn func() ast.Expression
+	infixParseFn  func(ast.Expression) ast.Expression
+)
+
 type Parser struct {
-	l         *lexer.Lexer
-	errors    []string
-	curToken  token.Token
-	peekToken token.Token
+	l              *lexer.Lexer
+	errors         []string
+	curToken       token.Token
+	peekToken      token.Token
+	prefixParseFns map[token.TokenType]prefixParseFn
+	infixParseFns  map[token.TokenType]infixParseFn
 }
 
 func New(l *lexer.Lexer) *Parser {
 	p := &Parser{
-		l:      l,
-		errors: []string{},
+		l:              l,
+		errors:         []string{},
+		prefixParseFns: map[token.TokenType]prefixParseFn{},
+		infixParseFns:  map[token.TokenType]infixParseFn{},
 	}
+	p.registerPrefix(token.IDENT, p.parseIdentifier)
+	p.registerPrefix(token.INT, p.parseIntegerLiteral)
+	p.registerPrefix(token.MINUS, p.parsePrefixExpression)
+	p.registerPrefix(token.BANG, p.parsePrefixExpression)
 	// TODO: figure out why this can't be done from `parseProgram`
 	p.nextToken()
 	p.nextToken()
@@ -34,6 +48,9 @@ func (p *Parser) ParseProgram() *ast.Program {
 		if stmt != nil {
 			program.Statements = append(program.Statements, stmt)
 		}
+		// NOTE: For now, this is not only eating the semicolon, it is also
+		// eating every and all tokens until parse statement finds something
+		// it deems valid.
 		p.nextToken()
 	}
 	return program
@@ -46,7 +63,7 @@ func (p *Parser) parseStatement() ast.Statement {
 	case token.RETURN:
 		return p.parseReturnStatement()
 	}
-	return nil
+	return p.parseExpressionStatement()
 }
 
 func (p *Parser) parseLetStatement() ast.Statement {
@@ -59,7 +76,7 @@ func (p *Parser) parseLetStatement() ast.Statement {
 		return nil
 	}
 	// TODO: Skipping until we find the semicolon to avoid parsing the expression.
-	if !p.curTokenIs(token.SEMICOLON) {
+	for !p.curTokenIs(token.SEMICOLON) {
 		p.nextToken()
 	}
 	return stmt
@@ -69,12 +86,61 @@ func (p *Parser) parseReturnStatement() ast.Statement {
 	stmt := &ast.ReturnStatement{Token: p.curToken}
 	p.nextToken()
 	// TODO: Skipping until we find the semicolon to avoid parsing the expression.
-	if !p.curTokenIs(token.SEMICOLON) {
+	for !p.curTokenIs(token.SEMICOLON) {
 		p.nextToken()
 	}
 	return stmt
 }
 
+func (p *Parser) parseExpressionStatement() ast.Statement {
+	stmt := &ast.ExpressionStatement{Token: p.curToken}
+	stmt.Expression = p.parseExpression(LOWEST)
+	// The semicolon is optional for expression statements so they're easier
+	// to type on the REPL. NOTE: It is weird that the last token parsed by
+	// parseExpression does not get consumed.
+	if p.peekTokenIs(token.SEMICOLON) {
+		p.nextToken()
+	}
+	return stmt
+}
+
+func (p *Parser) parseExpression(precedence int) ast.Expression {
+	// TODO: Could this be replaced with an `ok` check?
+	prefix := p.prefixParseFns[p.curToken.Type]
+	if prefix == nil {
+		p.noPrefixParseFnError(p.curToken.Type)
+		return nil
+	}
+	leftExpr := prefix()
+	return leftExpr
+
+}
+
+func (p *Parser) parseIdentifier() ast.Expression {
+	return &ast.Identifier{Token: p.curToken, Value: p.curToken.Literal}
+}
+
+func (p *Parser) parseIntegerLiteral() ast.Expression {
+	exp := &ast.IntegerLiteral{Token: p.curToken}
+	literal, err := strconv.ParseInt(p.curToken.Literal, 0, 64)
+	if err != nil {
+		p.errors = append(p.errors, fmt.Sprintf("could not parse %q as an integer", p.curToken.Literal))
+		return nil
+	}
+	exp.Value = literal
+	return exp
+}
+
+func (p *Parser) parsePrefixExpression() ast.Expression {
+	exp := &ast.PrefixExpression{
+		Token:    p.curToken,
+		Operator: p.curToken.Literal,
+	}
+	p.nextToken()
+	exp.Right = p.parseExpression(PREFIX)
+	return exp
+}
+
 func (p *Parser) curTokenIs(typ token.TokenType) bool {
 	return p.curToken.Type == typ
 }
@@ -114,3 +180,18 @@ func (p *Parser) peekError(typ token.TokenType) {
 		),
 	)
 }
+
+func (p *Parser) noPrefixParseFnError(t token.TokenType) {
+	p.errors = append(
+		p.errors,
+		fmt.Sprintf("no prefix parse function found for %q", t),
+	)
+}
+
+func (p *Parser) registerPrefix(typ token.TokenType, fn prefixParseFn) {
+	p.prefixParseFns[typ] = fn
+}
+
+func (p *Parser) registerInfix(typ token.TokenType, fn infixParseFn) {
+	p.infixParseFns[typ] = fn
+}

pkg/parser/parser_test.go

@@ -1,6 +1,7 @@
 package parser
 
 import (
+	"fmt"
 	"testing"
 
 	"code.jmug.me/jmug/interpreter-in-go/pkg/ast"
@@ -98,6 +99,128 @@ return 993322;
 	}
 }
 
+func TestIdentifierExpressions(t *testing.T) {
+	input := "foobar;"
+	l := lexer.New(input)
+	p := New(l)
+	program := p.ParseProgram()
+	checkParserErrors(t, p)
+	if len(program.Statements) != 1 {
+		t.Fatalf("program has not enough statements. got=%d",
+			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])
+	}
+
+	ident, ok := stmt.Expression.(*ast.Identifier)
+	if !ok {
+		t.Fatalf("exp not *ast.Identifier. got=%T", stmt.Expression)
+	}
+	if ident.Value != "foobar" {
+		t.Errorf("ident.Value not %s. got=%s", "foobar", ident.Value)
+	}
+	if ident.TokenLiteral() != "foobar" {
+		t.Errorf("ident.TokenLiteral not %s. got=%s", "foobar",
+			ident.TokenLiteral())
+	}
+}
+
+func TestIntegerLiteralExpression(t *testing.T) {
+	input := "5;"
+
+	l := lexer.New(input)
+	p := New(l)
+	program := p.ParseProgram()
+	checkParserErrors(t, p)
+
+	if len(program.Statements) != 1 {
+		t.Fatalf("program has not enough statements. got=%d",
+			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])
+	}
+
+	literal, ok := stmt.Expression.(*ast.IntegerLiteral)
+	if !ok {
+		t.Fatalf("exp not *ast.IntegerLiteral. got=%T", stmt.Expression)
+	}
+	if literal.Value != 5 {
+		t.Errorf("literal.Value not %d. got=%d", 5, literal.Value)
+	}
+	if literal.TokenLiteral() != "5" {
+		t.Errorf("literal.TokenLiteral not %s. got=%s", "5",
+			literal.TokenLiteral())
+	}
+}
+
+func TestParsingPrefixExpressions(t *testing.T) {
+	prefixTests := []struct {
+		input        string
+		operator     string
+		integerValue int64
+	}{
+		{"!5;", "!", 5},
+		{"-15;", "-", 15},
+	}
+
+	for _, tt := range prefixTests {
+		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.PrefixExpression)
+		if !ok {
+			t.Fatalf("stmt is not ast.PrefixExpression. got=%T", stmt.Expression)
+		}
+		if exp.Operator != tt.operator {
+			t.Fatalf("exp.Operator is not '%s'. got=%s",
+				tt.operator, exp.Operator)
+		}
+		if !testIntegerLiteral(t, exp.Right, tt.integerValue) {
+			return
+		}
+	}
+}
+
+func testIntegerLiteral(t *testing.T, il ast.Expression, value int64) bool {
+	integ, ok := il.(*ast.IntegerLiteral)
+	if !ok {
+		t.Errorf("il not *ast.IntegerLiteral. got=%T", il)
+		return false
+	}
+
+	if integ.Value != value {
+		t.Errorf("integ.Value not %d. got=%d", value, integ.Value)
+		return false
+	}
+
+	if integ.TokenLiteral() != fmt.Sprintf("%d", value) {
+		t.Errorf("integ.TokenLiteral not %d. got=%s", value,
+			integ.TokenLiteral())
+		return false
+	}
+
+	return true
+}
+
 func checkParserErrors(t *testing.T, p *Parser) {
 	errors := p.Errors()
 	if len(errors) == 0 {

pkg/parser/precedence.go

@@ -0,0 +1,12 @@
+package parser
+
+const (
+	_ int = iota
+	LOWEST
+	EQUALS      // ==
+	LESSGREATER // > or <
+	SUM         // +
+	PRODUCT     // *
+	PREFIX      // -X or !X
+	CALL        // myFunction(X)
+)