interpreter-in-go/pkg/parser/parser.go

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
	prefixParseFns map[token.TokenType]prefixParseFn
	infixParseFns  map[token.TokenType]infixParseFn
}

func New(l *lexer.Lexer) *Parser {
	p := &Parser{
		l:              l,
		errors:         []string{},
		prefixParseFns: map[token.TokenType]prefixParseFn{},
		infixParseFns:  map[token.TokenType]infixParseFn{},
	}
	// Prefix registrations
	p.registerPrefix(token.IDENT, p.parseIdentifier)
	p.registerPrefix(token.INT, p.parseIntegerLiteral)
	p.registerPrefix(token.MINUS, p.parsePrefixExpression)
	p.registerPrefix(token.BANG, p.parsePrefixExpression)
	p.registerPrefix(token.TRUE, p.parseBoolean)
	p.registerPrefix(token.FALSE, p.parseBoolean)
	p.registerPrefix(token.LPAREN, p.parseGroupedExpression)
	p.registerPrefix(token.IF, p.parseIfExpression)
	p.registerPrefix(token.FUNCTION, p.parseFunctionLiteral)
	// Infix registrations
	p.registerInfix(token.PLUS, p.parseInfixExpression)
	p.registerInfix(token.MINUS, p.parseInfixExpression)
	p.registerInfix(token.ASTERISK, p.parseInfixExpression)
	p.registerInfix(token.SLASH, p.parseInfixExpression)
	p.registerInfix(token.GT, p.parseInfixExpression)
	p.registerInfix(token.LT, p.parseInfixExpression)
	p.registerInfix(token.EQ, p.parseInfixExpression)
	p.registerInfix(token.NOT_EQ, p.parseInfixExpression)
	// TODO: figure out why this can't be done from `parseProgram`
	p.nextToken()
	p.nextToken()
	return p
}

func (p *Parser) ParseProgram() *ast.Program {
	program := &ast.Program{}
	program.Statements = []ast.Statement{}
	for !p.curTokenIs(token.EOF) {
		stmt := p.parseStatement()
		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
}

func (p *Parser) parseStatement() ast.Statement {
	switch p.curToken.Type {
	case token.LET:
		return p.parseLetStatement()
	case token.RETURN:
		return p.parseReturnStatement()
	}
	return p.parseExpressionStatement()
}

func (p *Parser) parseBlockStatement() *ast.BlockStatement {
	block := &ast.BlockStatement{Token: p.curToken}
	block.Statements = []ast.Statement{}
	p.nextToken()
	for !p.curTokenIs(token.RBRACE) && !p.curTokenIs(token.EOF) {
		stmt := p.parseStatement()
		if stmt != nil {
			block.Statements = append(block.Statements, stmt)
		}
		// Consume the semicolon.
		p.nextToken()
	}
	return block
}

func (p *Parser) parseLetStatement() ast.Statement {
	stmt := &ast.LetStatement{Token: p.curToken}
	if !p.nextTokenIfPeekIs(token.IDENT) {
		return nil
	}
	stmt.Name = &ast.Identifier{Token: p.curToken, Value: p.curToken.Literal}
	if !p.nextTokenIfPeekIs(token.ASSIGN) {
		return nil
	}
	// TODO: Skipping until we find the semicolon to avoid parsing the expression.
	for !p.curTokenIs(token.SEMICOLON) {
		p.nextToken()
	}
	return stmt
}

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.
	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
	}
	curExpr := prefix()
	for !p.peekTokenIs(token.SEMICOLON) && precedence < p.peekPrecedence() {
		infix := p.infixParseFns[p.peekToken.Type]
		if infix == nil {
			return curExpr
		}
		p.nextToken()
		curExpr = infix(curExpr)
	}
	return curExpr
}

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) parseBoolean() ast.Expression {
	return &ast.Boolean{Token: p.curToken, Value: p.curTokenIs(token.TRUE)}
}

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) parseInfixExpression(left ast.Expression) ast.Expression {
	exp := &ast.InfixExpression{
		Token:    p.curToken,
		Operator: p.curToken.Literal,
		Left:     left,
	}
	precedence := p.curPrecedence()
	p.nextToken()
	exp.Right = p.parseExpression(precedence)
	return exp
}

func (p *Parser) parseGroupedExpression() ast.Expression {
	p.nextToken()
	exp := p.parseExpression(LOWEST)
	if !p.nextTokenIfPeekIs(token.RPAREN) {
		// TODO: Would probably be good to emit an error here?
		return nil
	}
	return exp
}

func (p *Parser) parseIfExpression() ast.Expression {
	exp := &ast.IfExpression{Token: p.curToken}
	if !p.nextTokenIfPeekIs(token.LPAREN) {
		// TODO: Would be good to emit an error here.
		return nil
	}
	p.nextToken()
	exp.Condition = p.parseExpression(LOWEST)
	if !p.nextTokenIfPeekIs(token.RPAREN) {
		// TODO: Would be good to emit an error here.
		return nil
	}
	if !p.nextTokenIfPeekIs(token.LBRACE) {
		// TODO: Would be good to emit an error here.
		return nil
	}
	exp.Consequence = p.parseBlockStatement()
	if p.peekTokenIs(token.ELSE) {
		p.nextToken()
		if !p.nextTokenIfPeekIs(token.LBRACE) {
			// TODO: Would be good to emit an error here.
			return nil
		}
		exp.Alternative = p.parseBlockStatement()
	}
	// We don't consume the RBRACE because it acts as our "end of statement"
	// token, and it's consumed by parseProgram.
	return exp
}

func (p *Parser) parseFunctionLiteral() ast.Expression {
	fn := &ast.FunctionLiteral{Token: p.curToken}
	if !p.nextTokenIfPeekIs(token.LPAREN) {
		// TODO: Would be good to emit an error here.
		return nil
	}
	fn.Parameters = p.parseFunctionParameters()
	if !p.nextTokenIfPeekIs(token.LBRACE) {
		// TODO: Would be good to emit an error here.
		return nil
	}
	fn.Body = p.parseBlockStatement()
	return fn
}

func (p *Parser) parseFunctionParameters() []*ast.Identifier {
	params := []*ast.Identifier{}
	if p.peekTokenIs(token.RPAREN) {
		p.nextToken()
		return params
	}
	// Consume the LPAREN
	p.nextToken()
	params = append(params, &ast.Identifier{Token: p.curToken, Value: p.curToken.Literal})
	for p.peekTokenIs(token.COMMA) {
		// Consume the previous identifier.
		p.nextToken()
		// Consume the comma.
		p.nextToken()
		params = append(params, &ast.Identifier{Token: p.curToken, Value: p.curToken.Literal})
	}
	if !p.nextTokenIfPeekIs(token.RPAREN) {
		// TODO: Would be good to emit an error here.
		return nil
	}
	return params
}

func (p *Parser) curTokenIs(typ token.TokenType) bool {
	return p.curToken.Type == typ
}

func (p *Parser) peekTokenIs(typ token.TokenType) bool {
	return p.peekToken.Type == typ
}

// NOTE: I'll leave the name as-is to avoid deviating from the book (maybe a
// rename at the end?), but I think `nextTokenIfPeek` would be a much better
// name for this.
func (p *Parser) nextTokenIfPeekIs(typ token.TokenType) bool {
	if p.peekTokenIs(typ) {
		p.nextToken()
		return true
	}
	p.peekError(typ)
	return false
}

func (p *Parser) nextToken() {
	p.curToken = p.peekToken
	p.peekToken = p.l.NextToken()
}

func (p *Parser) Errors() []string {
	return p.errors
}

func (p *Parser) peekError(typ token.TokenType) {
	p.errors = append(
		p.errors,
		fmt.Sprintf(
			"expected next token to be %q, got %q instead",
			typ,
			p.peekToken.Type,
		),
	)
}

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
}