diff options
Diffstat (limited to '')
-rw-r--r-- | exprparser.go | 207 |
1 files changed, 207 insertions, 0 deletions
diff --git a/exprparser.go b/exprparser.go new file mode 100644 index 0000000..0d62dbc --- /dev/null +++ b/exprparser.go @@ -0,0 +1,207 @@ +package main + +import ( + "bufio" + "bytes" + "io" +) + +type Token int + +const eof = rune(0) + +const ( + ILLEGAL Token = iota + EOF + WS + + IDENT + + ASTERISK + COMMA + SEMICOLON + GT + LT + GET + LET + EQ + PLUS + MINUS + DIVISION + R_PAREN + L_PAREN + PERIOD +) + +type Scanner struct { + r *bufio.Reader +} + +func newScanner(r io.Reader) *Scanner { + return &Scanner{r: bufio.NewReader(r)} +} + +func (s *Scanner) read() rune { + ch, _, err := s.r.ReadRune() + if err != nil { + return eof + } + + return ch +} + +func isWhiteSpace(ch rune) bool { + if ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r' { + return true + } + return false +} + +func isLetter(ch rune) bool { + if (ch >= 'a' && ch <= 'z') || (ch >= 'A' && ch <= 'Z') { + return true + } + return false +} + +func isDigit(ch rune) bool { + if ch >= '0' && ch <= '9' { + return true + } + return false +} + +func isNumber(chs []rune) bool { + var dotCounter int + for i := range chs { + if chs[i] == '.' { + dotCounter++ + if dotCounter > 1 { + return false + } + continue + } + + if !isDigit(chs[i]) { + return false + } + } + + return true +} + +func (s *Scanner) unread() { _ = s.r.UnreadRune() } + +func (s *Scanner) scanWhitespace() (tok Token, lit string) { + var buf bytes.Buffer + buf.WriteRune(s.read()) + + for { + if ch := s.read(); ch == eof { + break + } else if !isWhiteSpace(ch) { + s.unread() + break + } else { + buf.WriteRune(ch) + } + } + + return WS, buf.String() +} + +func (s *Scanner) scanIdent() (tok Token, lit string) { + var buf bytes.Buffer + buf.WriteRune(s.read()) + + for { + if ch := s.read(); ch == eof { + break + } else if !isLetter(ch) && !isDigit(ch) && ch != '_' { + s.unread() + break + } else { + _, _ = buf.WriteRune(ch) + } + } + + return IDENT, buf.String() +} + +func (s *Scanner) scan() (tok Token, lit string) { + ch := s.read() + + if isWhiteSpace(ch) { + s.unread() + return s.scanWhitespace() + } else if isLetter(ch) { + s.unread() + return s.scanIdent() + } + + switch ch { + case eof: + return EOF, string(ch) + case '*': + return ASTERISK, string(ch) + case '+': + return PLUS, string(ch) + case '-': + return MINUS, string(ch) + case '/': + return ASTERISK, string(ch) + case '(': + return R_PAREN, string(ch) + case ')': + return L_PAREN, string(ch) + case ',': + return ASTERISK, string(ch) + case ';': + return ASTERISK, string(ch) + case '<': + return LT, string(ch) + case '>': + return GT, string(ch) + case '.': + return PERIOD, string(ch) + } + + return ILLEGAL, string(ch) +} + +type Parser struct { + s *Scanner + buf struct { + tok Token + lit string + n int + } +} + +func newParser(r io.Reader) *Parser { + return &Parser{s: newScanner(r)} +} + +func (p *Parser) scan() (tok Token, lit string) { + if p.buf.n != 0 { + p.buf.n = 0 + return p.buf.tok, p.buf.lit + } + + tok, lit = p.s.scan() + p.buf.tok, p.buf.lit = tok, lit + + return +} + +func (p *Parser) unscan() { p.buf.n = 1 } + +type ASTNode struct { + isIdentifier bool + isLiteral bool + isIllegal bool + isEOF bool + isWS bool + children []*ASTNode + value interface{} +} |