jarl/internal/namedconf/lexer.go

// Package namedconf parses and prints the named.conf configuration syntax.
//
// The parser is deliberately generic: a configuration is a list of
// statements, and each statement is a sequence of values (words or quoted
// strings) and blocks, terminated by a semicolon. This is enough to
// represent every BIND 9 statement without knowing its grammar, which is
// checked separately by the options catalog and, ultimately, by
// named-checkconf.
package namedconf

import (
	"fmt"
	"strings"
)

// Pos is a position in a configuration file.
type Pos struct {
	File string
	Line int
}

func (p Pos) String() string {
	if p.File == "" {
		return fmt.Sprintf("line %d", p.Line)
	}
	return fmt.Sprintf("%s:%d", p.File, p.Line)
}

type tokenKind int

const (
	tokEOF tokenKind = iota
	tokWord
	tokString
	tokLBrace
	tokRBrace
	tokSemi
	tokBang
)

func (k tokenKind) String() string {
	switch k {
	case tokEOF:
		return "end of input"
	case tokWord:
		return "word"
	case tokString:
		return "quoted string"
	case tokLBrace:
		return "'{'"
	case tokRBrace:
		return "'}'"
	case tokSemi:
		return "';'"
	case tokBang:
		return "'!'"
	}
	return "?"
}

type token struct {
	kind tokenKind
	text string
	pos  Pos
}

// Error is a syntax error with a position.
type Error struct {
	Pos Pos
	Msg string
}

func (e *Error) Error() string { return e.Pos.String() + ": " + e.Msg }

type lexer struct {
	src  string
	i    int
	line int
	file string
}

func (l *lexer) errorf(format string, args ...any) error {
	return &Error{Pos: Pos{File: l.file, Line: l.line}, Msg: fmt.Sprintf(format, args...)}
}

// next returns the next token, skipping whitespace and all three comment
// styles (/* */, // and #).
func (l *lexer) next() (token, error) {
	for {
		if l.i >= len(l.src) {
			return token{kind: tokEOF, pos: Pos{l.file, l.line}}, nil
		}
		c := l.src[l.i]
		switch {
		case c == '\n':
			l.line++
			l.i++
		case c == ' ' || c == '\t' || c == '\r' || c == '\f' || c == '\v':
			l.i++
		case c == '#':
			l.skipLine()
		case c == '/' && l.peek(1) == '/':
			l.skipLine()
		case c == '/' && l.peek(1) == '*':
			start := l.line
			end := strings.Index(l.src[l.i+2:], "*/")
			if end < 0 {
				l.line = start
				return token{}, l.errorf("unterminated /* comment")
			}
			l.line += strings.Count(l.src[l.i:l.i+2+end], "\n")
			l.i += 2 + end + 2
		default:
			return l.lexToken()
		}
	}
}

func (l *lexer) peek(n int) byte {
	if l.i+n < len(l.src) {
		return l.src[l.i+n]
	}
	return 0
}

func (l *lexer) skipLine() {
	for l.i < len(l.src) && l.src[l.i] != '\n' {
		l.i++
	}
}

func (l *lexer) lexToken() (token, error) {
	pos := Pos{l.file, l.line}
	c := l.src[l.i]
	switch c {
	case '{':
		l.i++
		return token{tokLBrace, "{", pos}, nil
	case '}':
		l.i++
		return token{tokRBrace, "}", pos}, nil
	case ';':
		l.i++
		return token{tokSemi, ";", pos}, nil
	case '!':
		l.i++
		return token{tokBang, "!", pos}, nil
	case '"':
		var sb strings.Builder
		l.i++
		for {
			if l.i >= len(l.src) {
				return token{}, &Error{Pos: pos, Msg: "unterminated quoted string"}
			}
			c := l.src[l.i]
			if c == '\\' && l.i+1 < len(l.src) {
				sb.WriteByte(l.src[l.i+1])
				if l.src[l.i+1] == '\n' {
					l.line++
				}
				l.i += 2
				continue
			}
			if c == '"' {
				l.i++
				break
			}
			if c == '\n' {
				l.line++
			}
			sb.WriteByte(c)
			l.i++
		}
		return token{tokString, sb.String(), pos}, nil
	}
	start := l.i
	for l.i < len(l.src) {
		c := l.src[l.i]
		if c == ' ' || c == '\t' || c == '\r' || c == '\n' || c == '\f' || c == '\v' ||
			c == '{' || c == '}' || c == ';' || c == '"' || c == '!' || c == '#' {
			break
		}
		if c == '/' && (l.peek(1) == '/' || l.peek(1) == '*') {
			break
		}
		l.i++
	}
	return token{tokWord, l.src[start:l.i], pos}, nil
}