Merge pull request #51 from pelletier/pelletier/fix-crlf-support

Fix support for CRLF line ending

.travis.yml

@@ -1,11 +1,9 @@
 language: go
 script: "./test.sh"
-sudo: false
 go:
-    - 1.2.2
     - 1.3.3
     - 1.4.2
-    - 1.5.1
+    - 1.5.3
     - tip
 before_install:
   - go get github.com/axw/gocov/gocov

README.md

@@ -98,7 +98,7 @@ You can run both of them using `./test.sh`.
 
 ## License
 
-Copyright (c) 2013 - 2015 Thomas Pelletier, Eric Anderton
+Copyright (c) 2013 - 2016 Thomas Pelletier, Eric Anderton
 
 Permission is hereby granted, free of charge, to any person obtaining a copy of
 this software and associated documentation files (the "Software"), to deal in

doc.go

@@ -1,7 +1,7 @@
 // Package toml is a TOML markup language parser.
 //
 // This version supports the specification as described in
-// https://github.com/toml-lang/toml/blob/master/versions/toml-v0.2.0.md
+// https://github.com/toml-lang/toml/blob/master/versions/en/toml-v0.4.0.md
 //
 // TOML Parsing
 //

doc_test.go

@@ -69,13 +69,13 @@ func Example_comprehensiveExample() {
 		fmt.Println("User is ", user, ". Password is ", password)
 
 		// show where elements are in the file
-		fmt.Println("User position: %v", configTree.GetPosition("user"))
+		fmt.Printf("User position: %v\n", configTree.GetPosition("user"))
-		fmt.Println("Password position: %v", configTree.GetPosition("password"))
+		fmt.Printf("Password position: %v\n", configTree.GetPosition("password"))
 
 		// use a query to gather elements without walking the tree
 		results, _ := config.Query("$..[user,password]")
 		for ii, item := range results.Values() {
-			fmt.Println("Query result %d: %v", ii, item)
+			fmt.Printf("Query result %d: %v\n", ii, item)
 		}
 	}
 }

example-crlf.toml

@@ -0,0 +1,29 @@
+# This is a TOML document. Boom.
+
+title = "TOML Example"
+
+[owner]
+name = "Tom Preston-Werner"
+organization = "GitHub"
+bio = "GitHub Cofounder & CEO\nLikes tater tots and beer."
+dob = 1979-05-27T07:32:00Z # First class dates? Why not?
+
+[database]
+server = "192.168.1.1"
+ports = [ 8001, 8001, 8002 ]
+connection_max = 5000
+enabled = true
+
+[servers]
+
+  # You can indent as you please. Tabs or spaces. TOML don't care.
+  [servers.alpha]
+  ip = "10.0.0.1"
+  dc = "eqdc10"
+
+  [servers.beta]
+  ip = "10.0.0.2"
+  dc = "eqdc10"
+
+[clients]
+data = [ ["gamma", "delta"], [1, 2] ] # just an update to make sure parsers support it

keysparsing.go

@@ -70,6 +70,9 @@ func parseKey(key string) ([]string, error) {
 	if buffer.Len() > 0 {
 		groups = append(groups, buffer.String())
 	}
+	if len(groups) == 0 {
+		return nil, fmt.Errorf("empty key")
+	}
 	return groups, nil
 }
 

keysparsing_test.go

@@ -42,3 +42,8 @@ func TestDottedKeyBasic(t *testing.T) {
 func TestBaseKeyPound(t *testing.T) {
 	testError(t, "hello#world", "invalid bare character: #")
 }
+
+func TestEmptyKey(t *testing.T) {
+	testError(t, "", "empty key")
+	testError(t, " ", "empty key")
+}

lexer.go

@@ -7,10 +7,11 @@ package toml
 
 import (
 	"fmt"
+	"github.com/pelletier/go-buffruneio"
+	"io"
 	"regexp"
 	"strconv"
 	"strings"
-	"unicode/utf8"
 )
 
 var dateRegexp *regexp.Regexp
@@ -20,47 +21,56 @@ type tomlLexStateFn func() tomlLexStateFn
 
 // Define lexer
 type tomlLexer struct {
-	input  string
+	input         *buffruneio.Reader // Textual source
-	start  int
+	buffer        []rune             // Runes composing the current token
-	pos    int
+	tokens        chan token
-	width  int
+	depth         int
-	tokens chan token
+	line          int
-	depth  int
+	col           int
-	line   int
+	endbufferLine int
-	col    int
+	endbufferCol  int
 }
 
-func (l *tomlLexer) run() {
+// Basic read operations on input
-	for state := l.lexVoid; state != nil; {
+
-		state = state()
+func (l *tomlLexer) read() rune {
+	r, err := l.input.ReadRune()
+	if err != nil {
+		panic(err)
 	}
-	close(l.tokens)
+	if r == '\n' {
+		l.endbufferLine++
+		l.endbufferCol = 1
+	} else {
+		l.endbufferCol++
+	}
+	return r
 }
 
-func (l *tomlLexer) nextStart() {
+func (l *tomlLexer) next() rune {
-	// iterate by runes (utf8 characters)
+	r := l.read()
-	// search for newlines and advance line/col counts
+
-	for i := l.start; i < l.pos; {
+	if r != eof {
-		r, width := utf8.DecodeRuneInString(l.input[i:])
+		l.buffer = append(l.buffer, r)
-		if r == '\n' {
-			l.line++
-			l.col = 1
-		} else {
-			l.col++
-		}
-		i += width
 	}
-	// advance start position to next token
+	return r
-	l.start = l.pos
 }
 
-func (l *tomlLexer) emit(t tokenType) {
+func (l *tomlLexer) ignore() {
-	l.tokens <- token{
+	l.buffer = make([]rune, 0)
-		Position: Position{l.line, l.col},
+	l.line = l.endbufferLine
-		typ:      t,
+	l.col = l.endbufferCol
-		val:      l.input[l.start:l.pos],
+}
+
+func (l *tomlLexer) skip() {
+	l.next()
+	l.ignore()
+}
+
+func (l *tomlLexer) fastForward(n int) {
+	for i := 0; i < n; i++ {
+		l.next()
 	}
-	l.nextStart()
 }
 
 func (l *tomlLexer) emitWithValue(t tokenType, value string) {
@@ -69,27 +79,37 @@ func (l *tomlLexer) emitWithValue(t tokenType, value string) {
 		typ:      t,
 		val:      value,
 	}
-	l.nextStart()
+	l.ignore()
 }
 
-func (l *tomlLexer) next() rune {
+func (l *tomlLexer) emit(t tokenType) {
-	if l.pos >= len(l.input) {
+	l.emitWithValue(t, string(l.buffer))
-		l.width = 0
+}
-		return eof
+
+func (l *tomlLexer) peek() rune {
+	r, err := l.input.ReadRune()
+	if err != nil {
+		panic(err)
 	}
-	var r rune
+	l.input.UnreadRune()
-	r, l.width = utf8.DecodeRuneInString(l.input[l.pos:])
-	l.pos += l.width
 	return r
 }
 
-func (l *tomlLexer) ignore() {
+func (l *tomlLexer) follow(next string) bool {
-	l.nextStart()
+	for _, expectedRune := range next {
+		r, err := l.input.ReadRune()
+		defer l.input.UnreadRune()
+		if err != nil {
+			panic(err)
+		}
+		if expectedRune != r {
+			return false
+		}
+	}
+	return true
 }
 
-func (l *tomlLexer) backup() {
+// Error management
-	l.pos -= l.width
-}
 
 func (l *tomlLexer) errorf(format string, args ...interface{}) tomlLexStateFn {
 	l.tokens <- token{
@@ -100,23 +120,7 @@ func (l *tomlLexer) errorf(format string, args ...interface{}) tomlLexStateFn {
 	return nil
 }
 
-func (l *tomlLexer) peek() rune {
+// State functions
-	r := l.next()
-	l.backup()
-	return r
-}
-
-func (l *tomlLexer) accept(valid string) bool {
-	if strings.IndexRune(valid, l.next()) >= 0 {
-		return true
-	}
-	l.backup()
-	return false
-}
-
-func (l *tomlLexer) follow(next string) bool {
-	return strings.HasPrefix(l.input[l.pos:], next)
-}
 
 func (l *tomlLexer) lexVoid() tomlLexStateFn {
 	for {
@@ -128,10 +132,15 @@ func (l *tomlLexer) lexVoid() tomlLexStateFn {
 			return l.lexComment
 		case '=':
 			return l.lexEqual
+		case '\r':
+			fallthrough
+		case '\n':
+			l.skip()
+			continue
 		}
 
 		if isSpace(next) {
-			l.ignore()
+			l.skip()
 		}
 
 		if l.depth > 0 {
@@ -142,7 +151,8 @@ func (l *tomlLexer) lexVoid() tomlLexStateFn {
 			return l.lexKey
 		}
 
-		if l.next() == eof {
+		if next == eof {
+			l.next()
 			break
 		}
 	}
@@ -177,13 +187,16 @@ func (l *tomlLexer) lexRvalue() tomlLexStateFn {
 			return l.lexLiteralString
 		case ',':
 			return l.lexComma
+		case '\r':
+			fallthrough
 		case '\n':
-			l.ignore()
+			l.skip()
-			l.pos++
 			if l.depth == 0 {
 				return l.lexVoid
 			}
 			return l.lexRvalue
+		case '_':
+			return l.errorf("cannot start number with underscore")
 		}
 
 		if l.follow("true") {
@@ -194,14 +207,20 @@ func (l *tomlLexer) lexRvalue() tomlLexStateFn {
 			return l.lexFalse
 		}
 
-		if isAlphanumeric(next) {
+		if isSpace(next) {
-			return l.lexKey
+			l.skip()
+			continue
 		}
 
-		dateMatch := dateRegexp.FindString(l.input[l.pos:])
+		if next == eof {
+			l.next()
+			break
+		}
+
+		possibleDate := string(l.input.Peek(35))
+		dateMatch := dateRegexp.FindString(possibleDate)
 		if dateMatch != "" {
-			l.ignore()
+			l.fastForward(len(dateMatch))
-			l.pos += len(dateMatch)
 			return l.lexDate
 		}
 
@@ -209,13 +228,10 @@ func (l *tomlLexer) lexRvalue() tomlLexStateFn {
 			return l.lexNumber
 		}
 
-		if isSpace(next) {
+		if isAlphanumeric(next) {
-			l.ignore()
+			return l.lexKey
 		}
 
-		if l.next() == eof {
-			break
-		}
 	}
 
 	l.emit(tokenEOF)
@@ -223,15 +239,13 @@ func (l *tomlLexer) lexRvalue() tomlLexStateFn {
 }
 
 func (l *tomlLexer) lexLeftCurlyBrace() tomlLexStateFn {
-	l.ignore()
+	l.next()
-	l.pos++
 	l.emit(tokenLeftCurlyBrace)
 	return l.lexRvalue
 }
 
 func (l *tomlLexer) lexRightCurlyBrace() tomlLexStateFn {
-	l.ignore()
+	l.next()
-	l.pos++
 	l.emit(tokenRightCurlyBrace)
 	return l.lexRvalue
 }
@@ -242,37 +256,32 @@ func (l *tomlLexer) lexDate() tomlLexStateFn {
 }
 
 func (l *tomlLexer) lexTrue() tomlLexStateFn {
-	l.ignore()
+	l.fastForward(4)
-	l.pos += 4
 	l.emit(tokenTrue)
 	return l.lexRvalue
 }
 
 func (l *tomlLexer) lexFalse() tomlLexStateFn {
-	l.ignore()
+	l.fastForward(5)
-	l.pos += 5
 	l.emit(tokenFalse)
 	return l.lexRvalue
 }
 
 func (l *tomlLexer) lexEqual() tomlLexStateFn {
-	l.ignore()
+	l.next()
-	l.accept("=")
 	l.emit(tokenEqual)
 	return l.lexRvalue
 }
 
 func (l *tomlLexer) lexComma() tomlLexStateFn {
-	l.ignore()
+	l.next()
-	l.accept(",")
 	l.emit(tokenComma)
 	return l.lexRvalue
 }
 
 func (l *tomlLexer) lexKey() tomlLexStateFn {
-	l.ignore()
 	inQuotes := false
-	for r := l.next(); isKeyChar(r) || r == '\n'; r = l.next() {
+	for r := l.peek(); isKeyChar(r) || r == '\n' || r == '\r'; r = l.peek() {
 		if r == '"' {
 			inQuotes = !inQuotes
 		} else if r == '\n' {
@@ -282,46 +291,46 @@ func (l *tomlLexer) lexKey() tomlLexStateFn {
 		} else if !isValidBareChar(r) && !inQuotes {
 			return l.errorf("keys cannot contain %c character", r)
 		}
+		l.next()
 	}
-	l.backup()
 	l.emit(tokenKey)
 	return l.lexVoid
 }
 
 func (l *tomlLexer) lexComment() tomlLexStateFn {
-	for {
+	for next := l.peek(); next != '\n' && next != eof; next = l.peek() {
-		next := l.next()
+		if (next == '\r' && l.follow("\r\n")) {
-		if next == '\n' || next == eof {
 			break
 		}
+		l.next()
 	}
 	l.ignore()
 	return l.lexVoid
 }
 
 func (l *tomlLexer) lexLeftBracket() tomlLexStateFn {
-	l.ignore()
+	l.next()
-	l.pos++
 	l.emit(tokenLeftBracket)
 	return l.lexRvalue
 }
 
 func (l *tomlLexer) lexLiteralString() tomlLexStateFn {
-	l.pos++
+	l.skip()
-	l.ignore()
 	growingString := ""
 
 	// handle special case for triple-quote
 	terminator := "'"
 	if l.follow("''") {
-		l.pos += 2
+		l.skip()
-		l.ignore()
+		l.skip()
 		terminator = "'''"
 
 		// special case: discard leading newline
-		if l.peek() == '\n' {
+		if l.follow("\r\n") {
-			l.pos++
+			l.skip()
-			l.ignore()
+			l.skip()
+		} else if l.peek() == '\n' {
+			l.skip()
 		}
 	}
 
@@ -329,50 +338,51 @@ func (l *tomlLexer) lexLiteralString() tomlLexStateFn {
 	for {
 		if l.follow(terminator) {
 			l.emitWithValue(tokenString, growingString)
-			l.pos += len(terminator)
+			l.fastForward(len(terminator))
 			l.ignore()
 			return l.lexRvalue
 		}
 
-		growingString += string(l.peek())
+		next := l.peek()
-
+		if next == eof {
-		if l.next() == eof {
 			break
 		}
+		growingString += string(l.next())
 	}
 
 	return l.errorf("unclosed string")
 }
 
 func (l *tomlLexer) lexString() tomlLexStateFn {
-	l.pos++
+	l.skip()
-	l.ignore()
 	growingString := ""
 
 	// handle special case for triple-quote
 	terminator := "\""
 	if l.follow("\"\"") {
-		l.pos += 2
+		l.skip()
-		l.ignore()
+		l.skip()
 		terminator = "\"\"\""
 
 		// special case: discard leading newline
-		if l.peek() == '\n' {
+		if l.follow("\r\n") {
-			l.pos++
+			l.skip()
-			l.ignore()
+			l.skip()
+		} else if l.peek() == '\n' {
+			l.skip()
 		}
 	}
 
 	for {
 		if l.follow(terminator) {
 			l.emitWithValue(tokenString, growingString)
-			l.pos += len(terminator)
+			l.fastForward(len(terminator))
 			l.ignore()
 			return l.lexRvalue
 		}
 
 		if l.follow("\\") {
-			l.pos++
+			l.next()
 			switch l.peek() {
 			case '\r':
 				fallthrough
@@ -382,56 +392,60 @@ func (l *tomlLexer) lexString() tomlLexStateFn {
 				fallthrough
 			case ' ':
 				// skip all whitespace chars following backslash
-				l.pos++
 				for strings.ContainsRune("\r\n\t ", l.peek()) {
-					l.pos++
+					l.next()
 				}
-				l.pos--
 			case '"':
 				growingString += "\""
+				l.next()
 			case 'n':
 				growingString += "\n"
+				l.next()
 			case 'b':
 				growingString += "\b"
+				l.next()
 			case 'f':
 				growingString += "\f"
+				l.next()
 			case '/':
 				growingString += "/"
+				l.next()
 			case 't':
 				growingString += "\t"
+				l.next()
 			case 'r':
 				growingString += "\r"
+				l.next()
 			case '\\':
 				growingString += "\\"
+				l.next()
 			case 'u':
-				l.pos++
+				l.next()
 				code := ""
 				for i := 0; i < 4; i++ {
 					c := l.peek()
-					l.pos++
 					if !isHexDigit(c) {
 						return l.errorf("unfinished unicode escape")
 					}
+					l.next()
 					code = code + string(c)
 				}
-				l.pos--
 				intcode, err := strconv.ParseInt(code, 16, 32)
 				if err != nil {
 					return l.errorf("invalid unicode escape: \\u" + code)
 				}
 				growingString += string(rune(intcode))
 			case 'U':
-				l.pos++
+				l.next()
 				code := ""
 				for i := 0; i < 8; i++ {
 					c := l.peek()
-					l.pos++
 					if !isHexDigit(c) {
 						return l.errorf("unfinished unicode escape")
 					}
+					l.next()
 					code = code + string(c)
 				}
-				l.pos--
 				intcode, err := strconv.ParseInt(code, 16, 64)
 				if err != nil {
 					return l.errorf("invalid unicode escape: \\U" + code)
@@ -445,10 +459,11 @@ func (l *tomlLexer) lexString() tomlLexStateFn {
 			if 0x00 <= r && r <= 0x1F {
 				return l.errorf("unescaped control character %U", r)
 			}
+			l.next()
 			growingString += string(r)
 		}
 
-		if l.next() == eof {
+		if l.peek() == eof {
 			break
 		}
 	}
@@ -457,12 +472,11 @@ func (l *tomlLexer) lexString() tomlLexStateFn {
 }
 
 func (l *tomlLexer) lexKeyGroup() tomlLexStateFn {
-	l.ignore()
+	l.next()
-	l.pos++
 
 	if l.peek() == '[' {
 		// token '[[' signifies an array of anonymous key groups
-		l.pos++
+		l.next()
 		l.emit(tokenDoubleLeftBracket)
 		return l.lexInsideKeyGroupArray
 	}
@@ -472,87 +486,85 @@ func (l *tomlLexer) lexKeyGroup() tomlLexStateFn {
 }
 
 func (l *tomlLexer) lexInsideKeyGroupArray() tomlLexStateFn {
-	for {
+	for r := l.peek(); r != eof; r = l.peek() {
-		if l.peek() == ']' {
+		switch r {
-			if l.pos > l.start {
+		case ']':
+			if len(l.buffer) > 0 {
 				l.emit(tokenKeyGroupArray)
 			}
-			l.ignore()
+			l.next()
-			l.pos++
 			if l.peek() != ']' {
-				break // error
+				break
 			}
-			l.pos++
+			l.next()
 			l.emit(tokenDoubleRightBracket)
 			return l.lexVoid
-		} else if l.peek() == '[' {
+		case '[':
 			return l.errorf("group name cannot contain ']'")
-		}
+		default:
-
+			l.next()
-		if l.next() == eof {
-			break
 		}
 	}
 	return l.errorf("unclosed key group array")
 }
 
 func (l *tomlLexer) lexInsideKeyGroup() tomlLexStateFn {
-	for {
+	for r := l.peek(); r != eof; r = l.peek() {
-		if l.peek() == ']' {
+		switch r {
-			if l.pos > l.start {
+		case ']':
+			if len(l.buffer) > 0 {
 				l.emit(tokenKeyGroup)
 			}
-			l.ignore()
+			l.next()
-			l.pos++
 			l.emit(tokenRightBracket)
 			return l.lexVoid
-		} else if l.peek() == '[' {
+		case '[':
 			return l.errorf("group name cannot contain ']'")
-		}
+		default:
-
+			l.next()
-		if l.next() == eof {
-			break
 		}
 	}
 	return l.errorf("unclosed key group")
 }
 
 func (l *tomlLexer) lexRightBracket() tomlLexStateFn {
-	l.ignore()
+	l.next()
-	l.pos++
 	l.emit(tokenRightBracket)
 	return l.lexRvalue
 }
 
 func (l *tomlLexer) lexNumber() tomlLexStateFn {
-	l.ignore()
+	r := l.peek()
-	if !l.accept("+") {
+	if r == '+' || r == '-' {
-		l.accept("-")
+		l.next()
 	}
 	pointSeen := false
 	expSeen := false
 	digitSeen := false
 	for {
-		next := l.next()
+		next := l.peek()
 		if next == '.' {
 			if pointSeen {
 				return l.errorf("cannot have two dots in one float")
 			}
+			l.next()
 			if !isDigit(l.peek()) {
 				return l.errorf("float cannot end with a dot")
 			}
 			pointSeen = true
 		} else if next == 'e' || next == 'E' {
 			expSeen = true
-			if !l.accept("+") {
+			l.next()
-				l.accept("-")
+			r := l.peek()
+			if r == '+' || r == '-' {
+				l.next()
 			}
 		} else if isDigit(next) {
 			digitSeen = true
+			l.next()
 		} else if next == '_' {
-			l.pos++
+			l.next()
 		} else {
-			l.backup()
 			break
 		}
 		if pointSeen && !digitSeen {
@@ -571,17 +583,27 @@ func (l *tomlLexer) lexNumber() tomlLexStateFn {
 	return l.lexRvalue
 }
 
+func (l *tomlLexer) run() {
+	for state := l.lexVoid; state != nil; {
+		state = state()
+	}
+	close(l.tokens)
+}
+
 func init() {
 	dateRegexp = regexp.MustCompile("^\\d{1,4}-\\d{2}-\\d{2}T\\d{2}:\\d{2}:\\d{2}(\\.\\d{1,9})?(Z|[+-]\\d{2}:\\d{2})")
 }
 
 // Entry point
-func lexToml(input string) chan token {
+func lexToml(input io.Reader) chan token {
+	bufferedInput := buffruneio.NewReader(input)
 	l := &tomlLexer{
-		input:  input,
+		input:         bufferedInput,
-		tokens: make(chan token),
+		tokens:        make(chan token),
-		line:   1,
+		line:          1,
-		col:    1,
+		col:           1,
+		endbufferLine: 1,
+		endbufferCol:  1,
 	}
 	go l.run()
 	return l.tokens

lexer_test.go

@@ -1,15 +1,19 @@
 package toml
 
-import "testing"
+import (
+	"strings"
+	"testing"
+)
 
 func testFlow(t *testing.T, input string, expectedFlow []token) {
-	ch := lexToml(input)
+	ch := lexToml(strings.NewReader(input))
 	for _, expected := range expectedFlow {
 		token := <-ch
 		if token != expected {
 			t.Log("While testing: ", input)
 			t.Log("compared (got)", token, "to (expected)", expected)
 			t.Log("\tvalue:", token.val, "<->", expected.val)
+			t.Log("\tvalue as bytes:", []byte(token.val), "<->", []byte(expected.val))
 			t.Log("\ttype:", token.typ.String(), "<->", expected.typ.String())
 			t.Log("\tline:", token.Line, "<->", expected.Line)
 			t.Log("\tcolumn:", token.Col, "<->", expected.Col)
@@ -83,6 +87,19 @@ func TestMultipleKeyGroupsComment(t *testing.T) {
 	})
 }
 
+
+func TestSimpleWindowsCRLF(t *testing.T) {
+	testFlow(t, "a=4\r\nb=2", []token{
+		token{Position{1, 1}, tokenKey, "a"},
+		token{Position{1, 2}, tokenEqual, "="},
+		token{Position{1, 3}, tokenInteger, "4"},
+		token{Position{2, 1}, tokenKey, "b"},
+		token{Position{2, 2}, tokenEqual, "="},
+		token{Position{2, 3}, tokenInteger, "2"},
+		token{Position{2, 4}, tokenEOF, ""},
+	})
+}
+
 func TestBasicKey(t *testing.T) {
 	testFlow(t, "hello", []token{
 		token{Position{1, 1}, tokenKey, "hello"},
@@ -618,3 +635,12 @@ func TestKeyNewline(t *testing.T) {
 		token{Position{1, 1}, tokenError, "keys cannot contain new lines"},
 	})
 }
+
+func TestInvalidFloat(t *testing.T) {
+	testFlow(t, "a=7e1_", []token{
+		token{Position{1, 1}, tokenKey, "a"},
+		token{Position{1, 2}, tokenEqual, "="},
+		token{Position{1, 3}, tokenFloat, "7e1_"},
+		token{Position{1, 7}, tokenEOF, ""},
+	})
+}

match.go

@@ -202,7 +202,7 @@ func (f *matchFilterFn) call(node interface{}, ctx *queryContext) {
 	fn, ok := (*ctx.filters)[f.Name]
 	if !ok {
 		panic(fmt.Sprintf("%s: query context does not have filter '%s'",
-			f.Pos, f.Name))
+			f.Pos.String(), f.Name))
 	}
 	switch castNode := tomlValueCheck(node, ctx).(type) {
 	case *TomlTree:

parser.go

@@ -5,6 +5,7 @@ package toml
 import (
 	"fmt"
 	"reflect"
+	"regexp"
 	"strconv"
 	"strings"
 	"time"
@@ -107,7 +108,7 @@ func (p *tomlParser) parseGroupArray() tomlParserStateFn {
 	var array []*TomlTree
 	if destTree == nil {
 		array = make([]*TomlTree, 0)
-	} else if destTree.([]*TomlTree) != nil {
+	} else if target, ok := destTree.([]*TomlTree); ok && target != nil {
 		array = destTree.([]*TomlTree)
 	} else {
 		p.raiseError(key, "key %s is already assigned and not of type group array", key)
@@ -211,6 +212,16 @@ func (p *tomlParser) parseAssign() tomlParserStateFn {
 	return p.parseStart
 }
 
+var numberUnderscoreInvalidRegexp *regexp.Regexp
+
+func cleanupNumberToken(value string) (string, error) {
+	if numberUnderscoreInvalidRegexp.MatchString(value) {
+		return "", fmt.Errorf("invalid use of _ in number")
+	}
+	cleanedVal := strings.Replace(value, "_", "", -1)
+	return cleanedVal, nil
+}
+
 func (p *tomlParser) parseRvalue() interface{} {
 	tok := p.getToken()
 	if tok == nil || tok.typ == tokenEOF {
@@ -225,14 +236,20 @@ func (p *tomlParser) parseRvalue() interface{} {
 	case tokenFalse:
 		return false
 	case tokenInteger:
-		cleanedVal := strings.Replace(tok.val, "_", "", -1)
+		cleanedVal, err := cleanupNumberToken(tok.val)
+		if err != nil {
+			p.raiseError(tok, "%s", err)
+		}
 		val, err := strconv.ParseInt(cleanedVal, 10, 64)
 		if err != nil {
 			p.raiseError(tok, "%s", err)
 		}
 		return val
 	case tokenFloat:
-		cleanedVal := strings.Replace(tok.val, "_", "", -1)
+		cleanedVal, err := cleanupNumberToken(tok.val)
+		if err != nil {
+			p.raiseError(tok, "%s", err)
+		}
 		val, err := strconv.ParseFloat(cleanedVal, 64)
 		if err != nil {
 			p.raiseError(tok, "%s", err)
@@ -361,3 +378,7 @@ func parseToml(flow chan token) *TomlTree {
 	parser.run()
 	return result
 }
+
+func init() {
+	numberUnderscoreInvalidRegexp = regexp.MustCompile(`([^\d]_|_[^\d]|_$|^_)`)
+}

parser_test.go

@@ -287,7 +287,7 @@ func TestArrayNestedStrings(t *testing.T) {
 
 func TestMissingValue(t *testing.T) {
 	_, err := Load("a = ")
-	if err.Error() != "(1, 4): expecting a value" {
+	if err.Error() != "(1, 5): expecting a value" {
 		t.Error("Bad error message:", err.Error())
 	}
 }
@@ -441,7 +441,7 @@ func TestImplicitDeclarationBefore(t *testing.T) {
 
 func TestFloatsWithoutLeadingZeros(t *testing.T) {
 	_, err := Load("a = .42")
-	if err.Error() != "(1, 4): cannot start float with a dot" {
+	if err.Error() != "(1, 5): cannot start float with a dot" {
 		t.Error("Bad error message:", err.Error())
 	}
 
@@ -494,6 +494,42 @@ func TestParseFile(t *testing.T) {
 	})
 }
 
+func TestParseFileCRLF(t *testing.T) {
+	tree, err := LoadFile("example-crlf.toml")
+
+	assertTree(t, tree, err, map[string]interface{}{
+		"title": "TOML Example",
+		"owner": map[string]interface{}{
+			"name":         "Tom Preston-Werner",
+			"organization": "GitHub",
+			"bio":          "GitHub Cofounder & CEO\nLikes tater tots and beer.",
+			"dob":          time.Date(1979, time.May, 27, 7, 32, 0, 0, time.UTC),
+		},
+		"database": map[string]interface{}{
+			"server":         "192.168.1.1",
+			"ports":          []int64{8001, 8001, 8002},
+			"connection_max": 5000,
+			"enabled":        true,
+		},
+		"servers": map[string]interface{}{
+			"alpha": map[string]interface{}{
+				"ip": "10.0.0.1",
+				"dc": "eqdc10",
+			},
+			"beta": map[string]interface{}{
+				"ip": "10.0.0.2",
+				"dc": "eqdc10",
+			},
+		},
+		"clients": map[string]interface{}{
+			"data": []interface{}{
+				[]string{"gamma", "delta"},
+				[]int64{1, 2},
+			},
+		},
+	})
+}
+
 func TestParseKeyGroupArray(t *testing.T) {
 	tree, err := Load("[[foo.bar]] a = 42\n[[foo.bar]] a = 69")
 	assertTree(t, tree, err, map[string]interface{}{
@@ -626,3 +662,32 @@ func TestDoubleEqual(t *testing.T) {
 		t.Error("Bad error message:", err.Error())
 	}
 }
+
+func TestGroupArrayReassign(t *testing.T) {
+	_, err := Load("[hello]\n[[hello]]")
+	if err.Error() != "(2, 3): key \"hello\" is already assigned and not of type group array" {
+		t.Error("Bad error message:", err.Error())
+	}
+}
+
+func TestInvalidFloatParsing(t *testing.T) {
+	_, err := Load("a=1e_2")
+	if err.Error() != "(1, 3): invalid use of _ in number" {
+		t.Error("Bad error message:", err.Error())
+	}
+
+	_, err = Load("a=1e2_")
+	if err.Error() != "(1, 3): invalid use of _ in number" {
+		t.Error("Bad error message:", err.Error())
+	}
+
+	_, err = Load("a=1__2")
+	if err.Error() != "(1, 3): invalid use of _ in number" {
+		t.Error("Bad error message:", err.Error())
+	}
+
+	_, err = Load("a=_1_2")
+	if err.Error() != "(1, 3): cannot start number with underscore" {
+		t.Error("Bad error message:", err.Error())
+	}
+}

position.go

@@ -6,13 +6,11 @@ import (
 	"fmt"
 )
 
-/*
+// Position of a document element within a TOML document.
-  Position of a document element within a TOML document.
+//
-
+// Line and Col are both 1-indexed positions for the element's line number and
-  Line and Col are both 1-indexed positions for the element's line number and
+// column number, respectively.  Values of zero or less will cause Invalid(),
-  column number, respectively.  Values of zero or less will cause Invalid(),
+// to return true.
-  to return true.
-*/
 type Position struct {
 	Line int // line within the document
 	Col  int // column within the line
@@ -24,7 +22,7 @@ func (p *Position) String() string {
 	return fmt.Sprintf("(%d, %d)", p.Line, p.Col)
 }
 
-// Returns whether or not the position is valid (i.e. with negative or
+// Invalid returns whether or not the position is valid (i.e. with negative or
 // null values)
 func (p *Position) Invalid() bool {
 	return p.Line <= 0 || p.Col <= 0

queryparser.go

@@ -138,7 +138,6 @@ func (p *queryParser) parseMatchExpr() queryParserStateFn {
 		return nil // allow EOF at this stage
 	}
 	return p.parseError(tok, "expected match expression")
-	return nil
 }
 
 func (p *queryParser) parseBracketExpr() queryParserStateFn {

test.sh

@@ -19,6 +19,8 @@ function git_clone() {
   popd
 }
 
+go get github.com/pelletier/go-buffruneio
+
 # get code for BurntSushi TOML validation
 # pinning all to 'HEAD' for version 0.3.x work (TODO: pin to commit hash when tests stabilize)
 git_clone github.com/BurntSushi/toml master HEAD
@@ -66,7 +68,7 @@ else
     echo "Invalid Test TOML for $test:"
     echo "===="
     cat "$invalid_test.toml"
-    
+
     echo "Go-TOML Output for $test:"
     echo "===="
     echo "go-toml Output:"

toml.go

@@ -3,7 +3,8 @@ package toml
 import (
 	"errors"
 	"fmt"
-	"io/ioutil"
+	"io"
+	"os"
 	"runtime"
 	"strconv"
 	"strings"
@@ -360,8 +361,8 @@ func (t *TomlTree) ToString() string {
 	return t.toToml("", "")
 }
 
-// Load creates a TomlTree from a string.
+// LoadReader creates a TomlTree from any io.Reader.
-func Load(content string) (tree *TomlTree, err error) {
+func LoadReader(reader io.Reader) (tree *TomlTree, err error) {
 	defer func() {
 		if r := recover(); r != nil {
 			if _, ok := r.(runtime.Error); ok {
@@ -370,18 +371,21 @@ func Load(content string) (tree *TomlTree, err error) {
 			err = errors.New(r.(string))
 		}
 	}()
-	tree = parseToml(lexToml(content))
+	tree = parseToml(lexToml(reader))
 	return
 }
 
+// Load creates a TomlTree from a string.
+func Load(content string) (tree *TomlTree, err error) {
+	return LoadReader(strings.NewReader(content))
+}
+
 // LoadFile creates a TomlTree from a file.
 func LoadFile(path string) (tree *TomlTree, err error) {
-	buff, ferr := ioutil.ReadFile(path)
+	file, err := os.Open(path)
-	if ferr != nil {
+	if err != nil {
-		err = ferr
+		return nil, err
-	} else {
-		s := string(buff)
-		tree, err = Load(s)
 	}
-	return
+	defer file.Close()
+	return LoadReader(file)
 }

toml_test.go

@@ -65,7 +65,7 @@ func TestTomlQuery(t *testing.T) {
 	}
 
 	if tt, ok := values[0].(*TomlTree); !ok {
-		t.Errorf("Expected type of TomlTree: %T Tv", values[0], values[0])
+		t.Errorf("Expected type of TomlTree: %T", values[0])
 	} else if tt.Get("a") != int64(1) {
 		t.Errorf("Expected 'a' with a value 1: %v", tt.Get("a"))
 	} else if tt.Get("b") != int64(2) {