Fix printf formatting

Generic input

.travis.yml

@@ -1,7 +1,14 @@
 language: go
 script: "./test.sh"
+sudo: false
 go:
-    - 1.1
+    - 1.3.3
-    - 1.2
+    - 1.4.2
-    - 1.3
+    - 1.5.3
     - tip
+before_install:
+  - go get github.com/axw/gocov/gocov
+  - go get github.com/mattn/goveralls
+  - if ! go get code.google.com/p/go.tools/cmd/cover; then go get golang.org/x/tools/cmd/cover; fi
+after_success:
+    - $HOME/gopath/bin/goveralls -service=travis-ci

README.md

@@ -3,10 +3,11 @@
 Go library for the [TOML](https://github.com/mojombo/toml) format.
 
 This library supports TOML version
-[v0.2.0](https://github.com/mojombo/toml/blob/master/versions/toml-v0.2.0.md)
+[v0.4.0](https://github.com/toml-lang/toml/blob/master/versions/en/toml-v0.4.0.md)
 
 [![GoDoc](https://godoc.org/github.com/pelletier/go-toml?status.svg)](http://godoc.org/github.com/pelletier/go-toml)
 [![Build Status](https://travis-ci.org/pelletier/go-toml.svg?branch=master)](https://travis-ci.org/pelletier/go-toml)
+[![Coverage Status](https://coveralls.io/repos/pelletier/go-toml/badge.svg?branch=master&service=github)](https://coveralls.io/github/pelletier/go-toml?branch=master)
 
 ## Features
 
@@ -97,7 +98,7 @@ You can run both of them using `./test.sh`.
 
 ## License
 
-Copyright (c) 2013, 2014 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

cmd/test_program.go

@@ -13,10 +13,12 @@ import (
 func main() {
 	bytes, err := ioutil.ReadAll(os.Stdin)
 	if err != nil {
+		log.Fatalf("Error during TOML read: %s", err)
 		os.Exit(2)
 	}
 	tree, err := toml.Load(string(bytes))
 	if err != nil {
+		log.Fatalf("Error during TOML load: %s", err)
 		os.Exit(1)
 	}
 
@@ -24,6 +26,7 @@ func main() {
 
 	if err := json.NewEncoder(os.Stdout).Encode(typedTree); err != nil {
 		log.Fatalf("Error encoding JSON: %s", err)
+		os.Exit(3)
 	}
 
 	os.Exit(0)

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)
 		}
 	}
 }

keysparsing.go

@@ -0,0 +1,81 @@
+// Parsing keys handling both bare and quoted keys.
+
+package toml
+
+import (
+	"bytes"
+	"fmt"
+	"unicode"
+)
+
+func parseKey(key string) ([]string, error) {
+	groups := []string{}
+	var buffer bytes.Buffer
+	inQuotes := false
+	escapeNext := false
+	ignoreSpace := true
+	expectDot := false
+
+	for _, char := range key {
+		if ignoreSpace {
+			if char == ' ' {
+				continue
+			}
+			ignoreSpace = false
+		}
+		if escapeNext {
+			buffer.WriteRune(char)
+			escapeNext = false
+			continue
+		}
+		switch char {
+		case '\\':
+			escapeNext = true
+			continue
+		case '"':
+			inQuotes = !inQuotes
+			expectDot = false
+		case '.':
+			if inQuotes {
+				buffer.WriteRune(char)
+			} else {
+				groups = append(groups, buffer.String())
+				buffer.Reset()
+				ignoreSpace = true
+				expectDot = false
+			}
+		case ' ':
+			if inQuotes {
+				buffer.WriteRune(char)
+			} else {
+				expectDot = true
+			}
+		default:
+			if !inQuotes && !isValidBareChar(char) {
+				return nil, fmt.Errorf("invalid bare character: %c", char)
+			}
+			if !inQuotes && expectDot {
+				return nil, fmt.Errorf("what?")
+			}
+			buffer.WriteRune(char)
+			expectDot = false
+		}
+	}
+	if inQuotes {
+		return nil, fmt.Errorf("mismatched quotes")
+	}
+	if escapeNext {
+		return nil, fmt.Errorf("unfinished escape sequence")
+	}
+	if buffer.Len() > 0 {
+		groups = append(groups, buffer.String())
+	}
+	if len(groups) == 0 {
+		return nil, fmt.Errorf("empty key")
+	}
+	return groups, nil
+}
+
+func isValidBareChar(r rune) bool {
+	return isAlphanumeric(r) || r == '-' || unicode.IsNumber(r)
+}

keysparsing_test.go

@@ -0,0 +1,49 @@
+package toml
+
+import (
+	"fmt"
+	"testing"
+)
+
+func testResult(t *testing.T, key string, expected []string) {
+	parsed, err := parseKey(key)
+	if err != nil {
+		t.Fatal("Unexpected error:", err)
+	}
+	if len(expected) != len(parsed) {
+		t.Fatal("Expected length", len(expected), "but", len(parsed), "parsed")
+	}
+	for index, expectedKey := range expected {
+		if expectedKey != parsed[index] {
+			t.Fatal("Expected", expectedKey, "at index", index, "but found", parsed[index])
+		}
+	}
+}
+
+func testError(t *testing.T, key string, expectedError string) {
+	_, err := parseKey(key)
+	if fmt.Sprintf("%s", err) != expectedError {
+		t.Fatalf("Expected error \"%s\", but got \"%s\".", expectedError, err)
+	}
+}
+
+func TestBareKeyBasic(t *testing.T) {
+	testResult(t, "test", []string{"test"})
+}
+
+func TestBareKeyDotted(t *testing.T) {
+	testResult(t, "this.is.a.key", []string{"this", "is", "a", "key"})
+}
+
+func TestDottedKeyBasic(t *testing.T) {
+	testResult(t, "\"a.dotted.key\"", []string{"a.dotted.key"})
+}
+
+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 {
@@ -128,21 +132,25 @@ func (l *tomlLexer) lexVoid() tomlLexStateFn {
 			return l.lexComment
 		case '=':
 			return l.lexEqual
+		case '\n':
+			l.skip()
+			continue
 		}
 
 		if isSpace(next) {
-			l.ignore()
+			l.skip()
 		}
 
 		if l.depth > 0 {
 			return l.lexRvalue
 		}
 
-		if isKeyChar(next) {
+		if isKeyStartChar(next) {
 			return l.lexKey
 		}
 
-		if l.next() == eof {
+		if next == eof {
+			l.next()
 			break
 		}
 	}
@@ -158,26 +166,33 @@ func (l *tomlLexer) lexRvalue() tomlLexStateFn {
 		case '.':
 			return l.errorf("cannot start float with a dot")
 		case '=':
-			return l.errorf("cannot have multiple equals for the same key")
+			return l.lexEqual
 		case '[':
 			l.depth++
 			return l.lexLeftBracket
 		case ']':
 			l.depth--
 			return l.lexRightBracket
+		case '{':
+			return l.lexLeftCurlyBrace
+		case '}':
+			return l.lexRightCurlyBrace
 		case '#':
 			return l.lexComment
 		case '"':
 			return l.lexString
+		case '\'':
+			return l.lexLiteralString
 		case ',':
 			return l.lexComma
 		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") {
@@ -188,11 +203,20 @@ func (l *tomlLexer) lexRvalue() tomlLexStateFn {
 			return l.lexFalse
 		}
 
-		if isAlphanumeric(next) {
+		if isSpace(next) {
-			return l.lexKey
+			l.skip()
+			continue
 		}
 
-		if 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.fastForward(len(dateMatch))
 			return l.lexDate
 		}
 
@@ -200,143 +224,233 @@ 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)
 	return nil
 }
 
+func (l *tomlLexer) lexLeftCurlyBrace() tomlLexStateFn {
+	l.next()
+	l.emit(tokenLeftCurlyBrace)
+	return l.lexRvalue
+}
+
+func (l *tomlLexer) lexRightCurlyBrace() tomlLexStateFn {
+	l.next()
+	l.emit(tokenRightCurlyBrace)
+	return l.lexRvalue
+}
+
 func (l *tomlLexer) lexDate() tomlLexStateFn {
-	l.ignore()
-	l.pos += 20 // Fixed size of a date in TOML
 	l.emit(tokenDate)
 	return l.lexRvalue
 }
 
 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 isKeyChar(l.next()) {
+	for r := l.peek(); isKeyChar(r) || r == '\n'; r = l.peek() {
+		if r == '"' {
+			inQuotes = !inQuotes
+		} else if r == '\n' {
+			return l.errorf("keys cannot contain new lines")
+		} else if isSpace(r) && !inQuotes {
+			break
+		} 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()
+		l.next()
-		if next == '\n' || next == eof {
-			break
-		}
 	}
 	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) lexString() tomlLexStateFn {
+func (l *tomlLexer) lexLiteralString() tomlLexStateFn {
-	l.pos++
+	l.skip()
-	l.ignore()
 	growingString := ""
 
+	// handle special case for triple-quote
+	terminator := "'"
+	if l.follow("''") {
+		l.skip()
+		l.skip()
+		terminator = "'''"
+
+		// special case: discard leading newline
+		if l.peek() == '\n' {
+			l.skip()
+		}
+	}
+
+	// find end of string
 	for {
-		if l.peek() == '"' {
+		if l.follow(terminator) {
 			l.emitWithValue(tokenString, growingString)
-			l.pos++
+			l.fastForward(len(terminator))
 			l.ignore()
 			return l.lexRvalue
 		}
 
-		if l.follow("\\\"") {
+		next := l.peek()
-			l.pos++
+		if next == eof {
-			growingString += "\""
+			break
-		} else if l.follow("\\n") {
+		}
-			l.pos++
+		growingString += string(l.next())
-			growingString += "\n"
+	}
-		} else if l.follow("\\b") {
+
-			l.pos++
+	return l.errorf("unclosed string")
-			growingString += "\b"
+}
-		} else if l.follow("\\f") {
+
-			l.pos++
+func (l *tomlLexer) lexString() tomlLexStateFn {
-			growingString += "\f"
+	l.skip()
-		} else if l.follow("\\/") {
+	growingString := ""
-			l.pos++
+
-			growingString += "/"
+	// handle special case for triple-quote
-		} else if l.follow("\\t") {
+	terminator := "\""
-			l.pos++
+	if l.follow("\"\"") {
-			growingString += "\t"
+		l.skip()
-		} else if l.follow("\\r") {
+		l.skip()
-			l.pos++
+		terminator = "\"\"\""
-			growingString += "\r"
+
-		} else if l.follow("\\\\") {
+		// special case: discard leading newline
-			l.pos++
+		if l.peek() == '\n' {
-			growingString += "\\"
+			l.skip()
-		} else if l.follow("\\u") {
+		}
-			l.pos += 2
+	}
-			code := ""
+
-			for i := 0; i < 4; i++ {
+	for {
-				c := l.peek()
+		if l.follow(terminator) {
-				l.pos++
+			l.emitWithValue(tokenString, growingString)
-				if !isHexDigit(c) {
+			l.fastForward(len(terminator))
-					return l.errorf("unfinished unicode escape")
+			l.ignore()
-				}
+			return l.lexRvalue
-				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))
-		} else if l.follow("\\") {
-			l.pos++
-			return l.errorf("invalid escape sequence: \\" + string(l.peek()))
-		} else {
-			growingString += string(l.peek())
 		}
 
-		if l.next() == eof {
+		if l.follow("\\") {
+			l.next()
+			switch l.peek() {
+			case '\r':
+				fallthrough
+			case '\n':
+				fallthrough
+			case '\t':
+				fallthrough
+			case ' ':
+				// skip all whitespace chars following backslash
+				for strings.ContainsRune("\r\n\t ", l.peek()) {
+					l.next()
+				}
+			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.next()
+				code := ""
+				for i := 0; i < 4; i++ {
+					c := l.peek()
+					if !isHexDigit(c) {
+						return l.errorf("unfinished unicode escape")
+					}
+					l.next()
+					code = code + string(c)
+				}
+				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.next()
+				code := ""
+				for i := 0; i < 8; i++ {
+					c := l.peek()
+					if !isHexDigit(c) {
+						return l.errorf("unfinished unicode escape")
+					}
+					l.next()
+					code = code + string(c)
+				}
+				intcode, err := strconv.ParseInt(code, 16, 64)
+				if err != nil {
+					return l.errorf("invalid unicode escape: \\U" + code)
+				}
+				growingString += string(rune(intcode))
+			default:
+				return l.errorf("invalid escape sequence: \\" + string(l.peek()))
+			}
+		} else {
+			r := l.peek()
+			if 0x00 <= r && r <= 0x1F {
+				return l.errorf("unescaped control character %U", r)
+			}
+			l.next()
+			growingString += string(r)
+		}
+
+		if l.peek() == eof {
 			break
 		}
 	}
@@ -345,12 +459,11 @@ 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
 	}
@@ -360,79 +473,85 @@ 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
+			l.next()
+			r := l.peek()
+			if r == '+' || r == '-' {
+				l.next()
+			}
 		} else if isDigit(next) {
 			digitSeen = true
+			l.next()
+		} else if next == '_' {
+			l.next()
 		} else {
-			l.backup()
 			break
 		}
 		if pointSeen && !digitSeen {
@@ -443,7 +562,7 @@ func (l *tomlLexer) lexNumber() tomlLexStateFn {
 	if !digitSeen {
 		return l.errorf("no digit in that number")
 	}
-	if pointSeen {
+	if pointSeen || expSeen {
 		l.emit(tokenFloat)
 	} else {
 		l.emit(tokenInteger)
@@ -451,17 +570,27 @@ 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}Z")
+	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

@@ -1,18 +1,23 @@
 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)
 			t.Log("compared", token, "to", expected)
-			t.Log(token.val, "<->", expected.val)
-			t.Log(token.typ, "<->", expected.typ)
-			t.Log(token.Line, "<->", expected.Line)
-			t.Log(token.Col, "<->", expected.Col)
 			t.FailNow()
 		}
 	}
@@ -39,6 +44,15 @@ func TestValidKeyGroup(t *testing.T) {
 	})
 }
 
+func TestNestedQuotedUnicodeKeyGroup(t *testing.T) {
+	testFlow(t, `[ j . "ʞ" . l ]`, []token{
+		token{Position{1, 1}, tokenLeftBracket, "["},
+		token{Position{1, 2}, tokenKeyGroup, ` j . "ʞ" . l `},
+		token{Position{1, 15}, tokenRightBracket, "]"},
+		token{Position{1, 16}, tokenEOF, ""},
+	})
+}
+
 func TestUnclosedKeyGroup(t *testing.T) {
 	testFlow(t, "[hello world", []token{
 		token{Position{1, 1}, tokenLeftBracket, "["},
@@ -118,19 +132,13 @@ func TestBasicKeyAndEqual(t *testing.T) {
 
 func TestKeyWithSharpAndEqual(t *testing.T) {
 	testFlow(t, "key#name = 5", []token{
-		token{Position{1, 1}, tokenKey, "key#name"},
+		token{Position{1, 1}, tokenError, "keys cannot contain # character"},
-		token{Position{1, 10}, tokenEqual, "="},
-		token{Position{1, 12}, tokenInteger, "5"},
-		token{Position{1, 13}, tokenEOF, ""},
 	})
 }
 
 func TestKeyWithSymbolsAndEqual(t *testing.T) {
-	testFlow(t, "~!@#$^&*()_+-`1234567890[]\\|/?><.,;:' = 5", []token{
+	testFlow(t, "~!@$^&*()_+-`1234567890[]\\|/?><.,;:' = 5", []token{
-		token{Position{1, 1}, tokenKey, "~!@#$^&*()_+-`1234567890[]\\|/?><.,;:'"},
+		token{Position{1, 1}, tokenError, "keys cannot contain ~ character"},
-		token{Position{1, 39}, tokenEqual, "="},
-		token{Position{1, 41}, tokenInteger, "5"},
-		token{Position{1, 42}, tokenEOF, ""},
 	})
 }
 
@@ -276,7 +284,13 @@ func TestKeyEqualArrayBoolsWithComments(t *testing.T) {
 
 func TestDateRegexp(t *testing.T) {
 	if dateRegexp.FindString("1979-05-27T07:32:00Z") == "" {
-		t.Fail()
+		t.Error("basic lexing")
+	}
+	if dateRegexp.FindString("1979-05-27T00:32:00-07:00") == "" {
+		t.Error("offset lexing")
+	}
+	if dateRegexp.FindString("1979-05-27T00:32:00.999999-07:00") == "" {
+		t.Error("nano precision lexing")
 	}
 }
 
@@ -287,6 +301,18 @@ func TestKeyEqualDate(t *testing.T) {
 		token{Position{1, 7}, tokenDate, "1979-05-27T07:32:00Z"},
 		token{Position{1, 27}, tokenEOF, ""},
 	})
+	testFlow(t, "foo = 1979-05-27T00:32:00-07:00", []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 7}, tokenDate, "1979-05-27T00:32:00-07:00"},
+		token{Position{1, 32}, tokenEOF, ""},
+	})
+	testFlow(t, "foo = 1979-05-27T00:32:00.999999-07:00", []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 7}, tokenDate, "1979-05-27T00:32:00.999999-07:00"},
+		token{Position{1, 39}, tokenEOF, ""},
+	})
 }
 
 func TestFloatEndingWithDot(t *testing.T) {
@@ -305,11 +331,48 @@ func TestFloatWithTwoDots(t *testing.T) {
 	})
 }
 
-func TestDoubleEqualKey(t *testing.T) {
+func TestFloatWithExponent1(t *testing.T) {
-	testFlow(t, "foo= = 2", []token{
+	testFlow(t, "a = 5e+22", []token{
-		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 1}, tokenKey, "a"},
-		token{Position{1, 4}, tokenEqual, "="},
+		token{Position{1, 3}, tokenEqual, "="},
-		token{Position{1, 5}, tokenError, "cannot have multiple equals for the same key"},
+		token{Position{1, 5}, tokenFloat, "5e+22"},
+		token{Position{1, 10}, tokenEOF, ""},
+	})
+}
+
+func TestFloatWithExponent2(t *testing.T) {
+	testFlow(t, "a = 5E+22", []token{
+		token{Position{1, 1}, tokenKey, "a"},
+		token{Position{1, 3}, tokenEqual, "="},
+		token{Position{1, 5}, tokenFloat, "5E+22"},
+		token{Position{1, 10}, tokenEOF, ""},
+	})
+}
+
+func TestFloatWithExponent3(t *testing.T) {
+	testFlow(t, "a = -5e+22", []token{
+		token{Position{1, 1}, tokenKey, "a"},
+		token{Position{1, 3}, tokenEqual, "="},
+		token{Position{1, 5}, tokenFloat, "-5e+22"},
+		token{Position{1, 11}, tokenEOF, ""},
+	})
+}
+
+func TestFloatWithExponent4(t *testing.T) {
+	testFlow(t, "a = -5e-22", []token{
+		token{Position{1, 1}, tokenKey, "a"},
+		token{Position{1, 3}, tokenEqual, "="},
+		token{Position{1, 5}, tokenFloat, "-5e-22"},
+		token{Position{1, 11}, tokenEOF, ""},
+	})
+}
+
+func TestFloatWithExponent5(t *testing.T) {
+	testFlow(t, "a = 6.626e-34", []token{
+		token{Position{1, 1}, tokenKey, "a"},
+		token{Position{1, 3}, tokenEqual, "="},
+		token{Position{1, 5}, tokenFloat, "6.626e-34"},
+		token{Position{1, 14}, tokenEOF, ""},
 	})
 }
 
@@ -377,6 +440,34 @@ func TestKeyEqualNumber(t *testing.T) {
 		token{Position{1, 7}, tokenFloat, "-4.2"},
 		token{Position{1, 11}, tokenEOF, ""},
 	})
+
+	testFlow(t, "foo = 1_000", []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 7}, tokenInteger, "1_000"},
+		token{Position{1, 12}, tokenEOF, ""},
+	})
+
+	testFlow(t, "foo = 5_349_221", []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 7}, tokenInteger, "5_349_221"},
+		token{Position{1, 16}, tokenEOF, ""},
+	})
+
+	testFlow(t, "foo = 1_2_3_4_5", []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 7}, tokenInteger, "1_2_3_4_5"},
+		token{Position{1, 16}, tokenEOF, ""},
+	})
+
+	testFlow(t, "flt8 = 9_224_617.445_991_228_313", []token{
+		token{Position{1, 1}, tokenKey, "flt8"},
+		token{Position{1, 6}, tokenEqual, "="},
+		token{Position{1, 8}, tokenFloat, "9_224_617.445_991_228_313"},
+		token{Position{1, 33}, tokenEOF, ""},
+	})
 }
 
 func TestMultiline(t *testing.T) {
@@ -398,6 +489,105 @@ func TestKeyEqualStringUnicodeEscape(t *testing.T) {
 		token{Position{1, 8}, tokenString, "hello ♥"},
 		token{Position{1, 21}, tokenEOF, ""},
 	})
+	testFlow(t, `foo = "hello \U000003B4"`, []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 8}, tokenString, "hello δ"},
+		token{Position{1, 25}, tokenEOF, ""},
+	})
+}
+
+func TestKeyEqualStringNoEscape(t *testing.T) {
+	testFlow(t, "foo = \"hello \u0002\"", []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 8}, tokenError, "unescaped control character U+0002"},
+	})
+	testFlow(t, "foo = \"hello \u001F\"", []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 8}, tokenError, "unescaped control character U+001F"},
+	})
+}
+
+func TestLiteralString(t *testing.T) {
+	testFlow(t, `foo = 'C:\Users\nodejs\templates'`, []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 8}, tokenString, `C:\Users\nodejs\templates`},
+		token{Position{1, 34}, tokenEOF, ""},
+	})
+	testFlow(t, `foo = '\\ServerX\admin$\system32\'`, []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 8}, tokenString, `\\ServerX\admin$\system32\`},
+		token{Position{1, 35}, tokenEOF, ""},
+	})
+	testFlow(t, `foo = 'Tom "Dubs" Preston-Werner'`, []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 8}, tokenString, `Tom "Dubs" Preston-Werner`},
+		token{Position{1, 34}, tokenEOF, ""},
+	})
+	testFlow(t, `foo = '<\i\c*\s*>'`, []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 8}, tokenString, `<\i\c*\s*>`},
+		token{Position{1, 19}, tokenEOF, ""},
+	})
+}
+
+func TestMultilineLiteralString(t *testing.T) {
+	testFlow(t, `foo = '''hello 'literal' world'''`, []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 10}, tokenString, `hello 'literal' world`},
+		token{Position{1, 34}, tokenEOF, ""},
+	})
+
+	testFlow(t, "foo = '''\nhello\n'literal'\nworld'''", []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{2, 1}, tokenString, "hello\n'literal'\nworld"},
+		token{Position{4, 9}, tokenEOF, ""},
+	})
+}
+
+func TestMultilineString(t *testing.T) {
+	testFlow(t, `foo = """hello "literal" world"""`, []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 10}, tokenString, `hello "literal" world`},
+		token{Position{1, 34}, tokenEOF, ""},
+	})
+
+	testFlow(t, "foo = \"\"\"\nhello\\\n\"literal\"\\\nworld\"\"\"", []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{2, 1}, tokenString, "hello\"literal\"world"},
+		token{Position{4, 9}, tokenEOF, ""},
+	})
+
+	testFlow(t, "foo = \"\"\"\\\n    \\\n    \\\n    hello\\\nmultiline\\\nworld\"\"\"", []token{
+		token{Position{1, 1}, tokenKey, "foo"},
+		token{Position{1, 5}, tokenEqual, "="},
+		token{Position{1, 10}, tokenString, "hellomultilineworld"},
+		token{Position{6, 9}, tokenEOF, ""},
+	})
+
+	testFlow(t, "key2 = \"\"\"\nThe quick brown \\\n\n\n  fox jumps over \\\n    the lazy dog.\"\"\"", []token{
+		token{Position{1, 1}, tokenKey, "key2"},
+		token{Position{1, 6}, tokenEqual, "="},
+		token{Position{2, 1}, tokenString, "The quick brown fox jumps over the lazy dog."},
+		token{Position{6, 21}, tokenEOF, ""},
+	})
+
+	testFlow(t, "key2 = \"\"\"\\\n       The quick brown \\\n       fox jumps over \\\n       the lazy dog.\\\n       \"\"\"", []token{
+		token{Position{1, 1}, tokenKey, "key2"},
+		token{Position{1, 6}, tokenEqual, "="},
+		token{Position{1, 11}, tokenString, "The quick brown fox jumps over the lazy dog."},
+		token{Position{5, 11}, tokenEOF, ""},
+	})
 }
 
 func TestUnicodeString(t *testing.T) {
@@ -417,3 +607,27 @@ func TestKeyGroupArray(t *testing.T) {
 		token{Position{1, 8}, tokenEOF, ""},
 	})
 }
+
+func TestQuotedKey(t *testing.T) {
+	testFlow(t, "\"a b\" = 42", []token{
+		token{Position{1, 1}, tokenKey, "\"a b\""},
+		token{Position{1, 7}, tokenEqual, "="},
+		token{Position{1, 9}, tokenInteger, "42"},
+		token{Position{1, 11}, tokenEOF, ""},
+	})
+}
+
+func TestKeyNewline(t *testing.T) {
+	testFlow(t, "a\n= 4", []token{
+		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:

match_test.go

@@ -2,7 +2,6 @@ package toml
 
 import (
 	"fmt"
-	"math"
 	"testing"
 )
 
@@ -110,7 +109,7 @@ func TestPathSliceStart(t *testing.T) {
 	assertPath(t,
 		"$[123:]",
 		buildPath(
-			newMatchSliceFn(123, math.MaxInt64, 1),
+			newMatchSliceFn(123, MaxInt, 1),
 		))
 }
 
@@ -134,7 +133,7 @@ func TestPathSliceStartStep(t *testing.T) {
 	assertPath(t,
 		"$[123::7]",
 		buildPath(
-			newMatchSliceFn(123, math.MaxInt64, 7),
+			newMatchSliceFn(123, MaxInt, 7),
 		))
 }
 
@@ -150,7 +149,7 @@ func TestPathSliceStep(t *testing.T) {
 	assertPath(t,
 		"$[::7]",
 		buildPath(
-			newMatchSliceFn(0, math.MaxInt64, 7),
+			newMatchSliceFn(0, MaxInt, 7),
 		))
 }
 

parser.go

@@ -5,6 +5,7 @@ package toml
 import (
 	"fmt"
 	"reflect"
+	"regexp"
 	"strconv"
 	"strings"
 	"time"
@@ -98,13 +99,16 @@ func (p *tomlParser) parseGroupArray() tomlParserStateFn {
 	}
 
 	// get or create group array element at the indicated part in the path
-	keys := strings.Split(key.val, ".")
+	keys, err := parseKey(key.val)
+	if err != nil {
+		p.raiseError(key, "invalid group array key: %s", err)
+	}
 	p.tree.createSubTree(keys[:len(keys)-1], startToken.Position) // create parent entries
 	destTree := p.tree.GetPath(keys)
 	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)
@@ -153,7 +157,10 @@ func (p *tomlParser) parseGroup() tomlParserStateFn {
 	}
 
 	p.seenGroupKeys = append(p.seenGroupKeys, key.val)
-	keys := strings.Split(key.val, ".")
+	keys, err := parseKey(key.val)
+	if err != nil {
+		p.raiseError(key, "invalid group array key: %s", err)
+	}
 	if err := p.tree.createSubTree(keys, startToken.Position); err != nil {
 		p.raiseError(key, "%s", err)
 	}
@@ -165,6 +172,7 @@ func (p *tomlParser) parseGroup() tomlParserStateFn {
 func (p *tomlParser) parseAssign() tomlParserStateFn {
 	key := p.getToken()
 	p.assume(tokenEqual)
+
 	value := p.parseRvalue()
 	var groupKey []string
 	if len(p.currentGroup) > 0 {
@@ -186,16 +194,34 @@ func (p *tomlParser) parseAssign() tomlParserStateFn {
 	}
 
 	// assign value to the found group
-	localKey := []string{key.val}
+	keyVals, err := parseKey(key.val)
-	finalKey := append(groupKey, key.val)
+	if err != nil {
+		p.raiseError(key, "%s", err)
+	}
+	if len(keyVals) != 1 {
+		p.raiseError(key, "Invalid key")
+	}
+	keyVal := keyVals[0]
+	localKey := []string{keyVal}
+	finalKey := append(groupKey, keyVal)
 	if targetNode.GetPath(localKey) != nil {
 		p.raiseError(key, "The following key was defined twice: %s",
 			strings.Join(finalKey, "."))
 	}
-	targetNode.values[key.val] = &tomlValue{value, key.Position}
+	targetNode.values[keyVal] = &tomlValue{value, key.Position}
 	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 {
@@ -210,25 +236,37 @@ func (p *tomlParser) parseRvalue() interface{} {
 	case tokenFalse:
 		return false
 	case tokenInteger:
-		val, err := strconv.ParseInt(tok.val, 10, 64)
+		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:
-		val, err := strconv.ParseFloat(tok.val, 64)
+		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)
 		}
 		return val
 	case tokenDate:
-		val, err := time.Parse(time.RFC3339, tok.val)
+		val, err := time.ParseInLocation(time.RFC3339Nano, tok.val, time.UTC)
 		if err != nil {
 			p.raiseError(tok, "%s", err)
 		}
 		return val
 	case tokenLeftBracket:
 		return p.parseArray()
+	case tokenLeftCurlyBrace:
+		return p.parseInlineTable()
+	case tokenEqual:
+		p.raiseError(tok, "cannot have multiple equals for the same key")
 	case tokenError:
 		p.raiseError(tok, "%s", tok)
 	}
@@ -238,7 +276,51 @@ func (p *tomlParser) parseRvalue() interface{} {
 	return nil
 }
 
-func (p *tomlParser) parseArray() []interface{} {
+func tokenIsComma(t *token) bool {
+	return t != nil && t.typ == tokenComma
+}
+
+func (p *tomlParser) parseInlineTable() *TomlTree {
+	tree := newTomlTree()
+	var previous *token
+Loop:
+	for {
+		follow := p.peek()
+		if follow == nil || follow.typ == tokenEOF {
+			p.raiseError(follow, "unterminated inline table")
+		}
+		switch follow.typ {
+		case tokenRightCurlyBrace:
+			p.getToken()
+			break Loop
+		case tokenKey:
+			if !tokenIsComma(previous) && previous != nil {
+				p.raiseError(follow, "comma expected between fields in inline table")
+			}
+			key := p.getToken()
+			p.assume(tokenEqual)
+			value := p.parseRvalue()
+			tree.Set(key.val, value)
+		case tokenComma:
+			if previous == nil {
+				p.raiseError(follow, "inline table cannot start with a comma")
+			}
+			if tokenIsComma(previous) {
+				p.raiseError(follow, "need field between two commas in inline table")
+			}
+			p.getToken()
+		default:
+			p.raiseError(follow, "unexpected token type in inline table: %s", follow.typ.String())
+		}
+		previous = follow
+	}
+	if tokenIsComma(previous) {
+		p.raiseError(previous, "trailing comma at the end of inline table")
+	}
+	return tree
+}
+
+func (p *tomlParser) parseArray() interface{} {
 	var array []interface{}
 	arrayType := reflect.TypeOf(nil)
 	for {
@@ -248,7 +330,7 @@ func (p *tomlParser) parseArray() []interface{} {
 		}
 		if follow.typ == tokenRightBracket {
 			p.getToken()
-			return array
+			break
 		}
 		val := p.parseRvalue()
 		if arrayType == nil {
@@ -259,7 +341,7 @@ func (p *tomlParser) parseArray() []interface{} {
 		}
 		array = append(array, val)
 		follow = p.peek()
-		if follow == nil {
+		if follow == nil || follow.typ == tokenEOF {
 			p.raiseError(follow, "unterminated array")
 		}
 		if follow.typ != tokenRightBracket && follow.typ != tokenComma {
@@ -269,6 +351,17 @@ func (p *tomlParser) parseArray() []interface{} {
 			p.getToken()
 		}
 	}
+	// An array of TomlTrees is actually an array of inline
+	// tables, which is a shorthand for a table array. If the
+	// array was not converted from []interface{} to []*TomlTree,
+	// the two notations would not be equivalent.
+	if arrayType == reflect.TypeOf(newTomlTree()) {
+		tomlArray := make([]*TomlTree, len(array))
+		for i, v := range array {
+			tomlArray[i] = v.(*TomlTree)
+		}
+		return tomlArray
+	}
 	return array
 }
 
@@ -285,3 +378,7 @@ func parseToml(flow chan token) *TomlTree {
 	parser.run()
 	return result
 }
+
+func init() {
+	numberUnderscoreInvalidRegexp = regexp.MustCompile(`([^\d]_|_[^\d]|_$|^_)`)
+}

parser_test.go

@@ -51,12 +51,10 @@ func TestSimpleKV(t *testing.T) {
 	})
 }
 
-// NOTE: from the BurntSushi test suite
+func TestNumberInKey(t *testing.T) {
-// NOTE: this test is pure evil due to the embedded '.'
+	tree, err := Load("hello2 = 42")
-func TestSpecialKV(t *testing.T) {
-	tree, err := Load("~!@#$^&*()_+-`1234567890[]\\|/?><.,;: = 1")
 	assertTree(t, tree, err, map[string]interface{}{
-		"~!@#$^&*()_+-`1234567890[]\\|/?><.,;:": int64(1),
+		"hello2": int64(42),
 	})
 }
 
@@ -70,6 +68,44 @@ func TestSimpleNumbers(t *testing.T) {
 	})
 }
 
+func TestNumbersWithUnderscores(t *testing.T) {
+	tree, err := Load("a = 1_000")
+	assertTree(t, tree, err, map[string]interface{}{
+		"a": int64(1000),
+	})
+
+	tree, err = Load("a = 5_349_221")
+	assertTree(t, tree, err, map[string]interface{}{
+		"a": int64(5349221),
+	})
+
+	tree, err = Load("a = 1_2_3_4_5")
+	assertTree(t, tree, err, map[string]interface{}{
+		"a": int64(12345),
+	})
+
+	tree, err = Load("flt8 = 9_224_617.445_991_228_313")
+	assertTree(t, tree, err, map[string]interface{}{
+		"flt8": float64(9224617.445991228313),
+	})
+
+	tree, err = Load("flt9 = 1e1_00")
+	assertTree(t, tree, err, map[string]interface{}{
+		"flt9": float64(1e100),
+	})
+}
+
+func TestFloatsWithExponents(t *testing.T) {
+	tree, err := Load("a = 5e+22\nb = 5E+22\nc = -5e+22\nd = -5e-22\ne = 6.626e-34")
+	assertTree(t, tree, err, map[string]interface{}{
+		"a": float64(5e+22),
+		"b": float64(5E+22),
+		"c": float64(-5e+22),
+		"d": float64(-5e-22),
+		"e": float64(6.626e-34),
+	})
+}
+
 func TestSimpleDate(t *testing.T) {
 	tree, err := Load("a = 1979-05-27T07:32:00Z")
 	assertTree(t, tree, err, map[string]interface{}{
@@ -77,6 +113,20 @@ func TestSimpleDate(t *testing.T) {
 	})
 }
 
+func TestDateOffset(t *testing.T) {
+	tree, err := Load("a = 1979-05-27T00:32:00-07:00")
+	assertTree(t, tree, err, map[string]interface{}{
+		"a": time.Date(1979, time.May, 27, 0, 32, 0, 0, time.FixedZone("", -7*60*60)),
+	})
+}
+
+func TestDateNano(t *testing.T) {
+	tree, err := Load("a = 1979-05-27T00:32:00.999999999-07:00")
+	assertTree(t, tree, err, map[string]interface{}{
+		"a": time.Date(1979, time.May, 27, 0, 32, 0, 999999999, time.FixedZone("", -7*60*60)),
+	})
+}
+
 func TestSimpleString(t *testing.T) {
 	tree, err := Load("a = \"hello world\"")
 	assertTree(t, tree, err, map[string]interface{}{
@@ -84,6 +134,13 @@ func TestSimpleString(t *testing.T) {
 	})
 }
 
+func TestSpaceKey(t *testing.T) {
+	tree, err := Load("\"a b\" = \"hello world\"")
+	assertTree(t, tree, err, map[string]interface{}{
+		"a b": "hello world",
+	})
+}
+
 func TestStringEscapables(t *testing.T) {
 	tree, err := Load("a = \"a \\n b\"")
 	assertTree(t, tree, err, map[string]interface{}{
@@ -127,6 +184,41 @@ func TestNestedKeys(t *testing.T) {
 	})
 }
 
+func TestNestedQuotedUnicodeKeys(t *testing.T) {
+	tree, err := Load("[ j . \"ʞ\" . l ]\nd = 42")
+	assertTree(t, tree, err, map[string]interface{}{
+		"j": map[string]interface{}{
+			"ʞ": map[string]interface{}{
+				"l": map[string]interface{}{
+					"d": int64(42),
+				},
+			},
+		},
+	})
+
+	tree, err = Load("[ g . h . i ]\nd = 42")
+	assertTree(t, tree, err, map[string]interface{}{
+		"g": map[string]interface{}{
+			"h": map[string]interface{}{
+				"i": map[string]interface{}{
+					"d": int64(42),
+				},
+			},
+		},
+	})
+
+	tree, err = Load("[ d.e.f ]\nk = 42")
+	assertTree(t, tree, err, map[string]interface{}{
+		"d": map[string]interface{}{
+			"e": map[string]interface{}{
+				"f": map[string]interface{}{
+					"k": int64(42),
+				},
+			},
+		},
+	})
+}
+
 func TestArrayOne(t *testing.T) {
 	tree, err := Load("a = [1]")
 	assertTree(t, tree, err, map[string]interface{}{
@@ -195,7 +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())
 	}
 }
@@ -205,6 +297,16 @@ func TestUnterminatedArray(t *testing.T) {
 	if err.Error() != "(1, 8): unterminated array" {
 		t.Error("Bad error message:", err.Error())
 	}
+
+	_, err = Load("a = [1")
+	if err.Error() != "(1, 7): unterminated array" {
+		t.Error("Bad error message:", err.Error())
+	}
+
+	_, err = Load("a = [1 2")
+	if err.Error() != "(1, 8): missing comma" {
+		t.Error("Bad error message:", err.Error())
+	}
 }
 
 func TestNewlinesInArrays(t *testing.T) {
@@ -228,6 +330,80 @@ func TestArrayWithExtraCommaComment(t *testing.T) {
 	})
 }
 
+func TestSimpleInlineGroup(t *testing.T) {
+	tree, err := Load("key = {a = 42}")
+	assertTree(t, tree, err, map[string]interface{}{
+		"key": map[string]interface{}{
+			"a": int64(42),
+		},
+	})
+}
+
+func TestDoubleInlineGroup(t *testing.T) {
+	tree, err := Load("key = {a = 42, b = \"foo\"}")
+	assertTree(t, tree, err, map[string]interface{}{
+		"key": map[string]interface{}{
+			"a": int64(42),
+			"b": "foo",
+		},
+	})
+}
+
+func TestExampleInlineGroup(t *testing.T) {
+	tree, err := Load(`name = { first = "Tom", last = "Preston-Werner" }
+point = { x = 1, y = 2 }`)
+	assertTree(t, tree, err, map[string]interface{}{
+		"name": map[string]interface{}{
+			"first": "Tom",
+			"last":  "Preston-Werner",
+		},
+		"point": map[string]interface{}{
+			"x": int64(1),
+			"y": int64(2),
+		},
+	})
+}
+
+func TestExampleInlineGroupInArray(t *testing.T) {
+	tree, err := Load(`points = [{ x = 1, y = 2 }]`)
+	assertTree(t, tree, err, map[string]interface{}{
+		"points": []map[string]interface{}{
+			map[string]interface{}{
+				"x": int64(1),
+				"y": int64(2),
+			},
+		},
+	})
+}
+
+func TestInlineTableUnterminated(t *testing.T) {
+	_, err := Load("foo = {")
+	if err.Error() != "(1, 8): unterminated inline table" {
+		t.Error("Bad error message:", err.Error())
+	}
+}
+
+func TestInlineTableCommaExpected(t *testing.T) {
+	_, err := Load("foo = {hello = 53 test = foo}")
+	if err.Error() != "(1, 19): comma expected between fields in inline table" {
+		t.Error("Bad error message:", err.Error())
+	}
+}
+
+func TestInlineTableCommaStart(t *testing.T) {
+	_, err := Load("foo = {, hello = 53}")
+	if err.Error() != "(1, 8): inline table cannot start with a comma" {
+		t.Error("Bad error message:", err.Error())
+	}
+}
+
+func TestInlineTableDoubleComma(t *testing.T) {
+	_, err := Load("foo = {hello = 53,, foo = 17}")
+	if err.Error() != "(1, 19): need field between two commas in inline table" {
+		t.Error("Bad error message:", err.Error())
+	}
+}
+
 func TestDuplicateGroups(t *testing.T) {
 	_, err := Load("[foo]\na=2\n[foo]b=3")
 	if err.Error() != "(3, 2): duplicated tables" {
@@ -265,7 +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())
 	}
 
@@ -436,3 +612,46 @@ func TestNestedTreePosition(t *testing.T) {
 			"foo.bar.b": Position{3, 1},
 		})
 }
+
+func TestInvalidGroupArray(t *testing.T) {
+	_, err := Load("[key#group]\nanswer = 42")
+	if err == nil {
+		t.Error("Should error")
+	}
+}
+
+func TestDoubleEqual(t *testing.T) {
+	_, err := Load("foo= = 2")
+	if err.Error() != "(1, 6): cannot have multiple equals for the same key" {
+		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

@@ -9,9 +9,10 @@ package toml
 
 import (
 	"fmt"
-	"math"
 )
 
+const MaxInt = int(^uint(0) >> 1)
+
 type queryParser struct {
 	flow         chan token
 	tokensBuffer []token
@@ -137,7 +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 {
@@ -203,7 +203,7 @@ loop: // labeled loop for easy breaking
 
 func (p *queryParser) parseSliceExpr() queryParserStateFn {
 	// init slice to grab all elements
-	start, end, step := 0, math.MaxInt64, 1
+	start, end, step := 0, MaxInt, 1
 
 	// parse optional start
 	tok := p.getToken()

test.sh

@@ -5,15 +5,28 @@ set -e
 # set the path to the present working directory
 export GOPATH=`pwd`
 
-# Vendorize the BurntSushi test suite
+function git_clone() {
-# NOTE: this gets a specific release to avoid versioning issues
+  path=$1
-if [ ! -d 'src/github.com/BurntSushi/toml-test' ]; then
+  branch=$2
-  mkdir -p src/github.com/BurntSushi
+  version=$3
-  git clone https://github.com/BurntSushi/toml-test.git src/github.com/BurntSushi/toml-test
+  if [ ! -d "src/$path" ]; then
-fi
+    mkdir -p src/$path
-pushd src/github.com/BurntSushi/toml-test
+    git clone https://$path.git src/$path
-git reset --hard '0.2.0'  # use the released version, NOT tip
+  fi
-popd
+  pushd src/$path
+  git checkout "$branch"
+  git reset --hard "$version"
+  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
+git_clone github.com/BurntSushi/toml-test master HEAD #was: 0.2.0 HEAD
+
+# build the BurntSushi test application
 go build -o toml-test github.com/BurntSushi/toml-test
 
 # vendorize the current lib for testing
@@ -23,6 +36,42 @@ cp *.go *.toml src/github.com/pelletier/go-toml
 cp cmd/*.go src/github.com/pelletier/go-toml/cmd
 go build -o test_program_bin src/github.com/pelletier/go-toml/cmd/test_program.go
 
-# Run basic unit tests and then the BurntSushi test suite
+# Run basic unit tests
 go test -v github.com/pelletier/go-toml
-./toml-test ./test_program_bin | tee test_out
+
+# run the entire BurntSushi test suite
+if [[ $# -eq 0 ]] ; then
+  echo "Running all BurntSushi tests"
+  ./toml-test ./test_program_bin | tee test_out
+else
+  # run a specific test
+  test=$1
+  test_path='src/github.com/BurntSushi/toml-test/tests'
+  valid_test="$test_path/valid/$test"
+  invalid_test="$test_path/invalid/$test"
+
+  if [ -e "$valid_test.toml" ]; then
+    echo "Valid Test TOML for $test:"
+    echo "===="
+    cat "$valid_test.toml"
+
+    echo "Valid Test JSON for $test:"
+    echo "===="
+    cat "$valid_test.json"
+
+    echo "Go-TOML Output for $test:"
+    echo "===="
+    cat "$valid_test.toml" | ./test_program_bin
+  fi
+
+  if [ -e "$invalid_test.toml" ]; then
+    echo "Invalid Test TOML for $test:"
+    echo "===="
+    cat "$invalid_test.toml"
+
+    echo "Go-TOML Output for $test:"
+    echo "===="
+    echo "go-toml Output:"
+    cat "$invalid_test.toml" | ./test_program_bin
+  fi
+fi

token.go

@@ -26,6 +26,8 @@ const (
 	tokenEqual
 	tokenLeftBracket
 	tokenRightBracket
+	tokenLeftCurlyBrace
+	tokenRightCurlyBrace
 	tokenLeftParen
 	tokenRightParen
 	tokenDoubleLeftBracket
@@ -44,6 +46,7 @@ const (
 )
 
 var tokenTypeNames = []string{
+	"Error",
 	"EOF",
 	"Comment",
 	"Key",
@@ -54,7 +57,9 @@ var tokenTypeNames = []string{
 	"Float",
 	"=",
 	"[",
-	"[",
+	"]",
+	"{",
+	"}",
 	"(",
 	")",
 	"]]",
@@ -117,9 +122,14 @@ func isAlphanumeric(r rune) bool {
 }
 
 func isKeyChar(r rune) bool {
-	// "Keys start with the first non-whitespace character and end with the last
+	// Keys start with the first character that isn't whitespace or [ and end
-	// non-whitespace character before the equals sign."
+	// with the last non-whitespace character before the equals sign. Keys
-	return !(isSpace(r) || r == '\r' || r == '\n' || r == eof || r == '=')
+	// cannot contain a # character."
+	return !(r == '\r' || r == '\n' || r == eof || r == '=')
+}
+
+func isKeyStartChar(r rune) bool {
+	return !(isSpace(r) || r == '\r' || r == '\n' || r == eof || r == '[')
 }
 
 func isDigit(r rune) bool {

toml.go

@@ -3,7 +3,8 @@ package toml
 import (
 	"errors"
 	"fmt"
-	"io/ioutil"
+	"io"
+	"os"
 	"runtime"
 	"strconv"
 	"strings"
@@ -28,6 +29,12 @@ func newTomlTree() *TomlTree {
 	}
 }
 
+func TreeFromMap(m map[string]interface{}) *TomlTree {
+	return &TomlTree{
+		values: m,
+	}
+}
+
 // Has returns a boolean indicating if the given key exists.
 func (t *TomlTree) Has(key string) bool {
 	if key == "" {
@@ -59,7 +66,11 @@ func (t *TomlTree) Get(key string) interface{} {
 	if key == "" {
 		return t
 	}
-	return t.GetPath(strings.Split(key, "."))
+	comps, err := parseKey(key)
+	if err != nil {
+		return nil
+	}
+	return t.GetPath(comps)
 }
 
 // GetPath returns the element in the tree indicated by 'keys'.
@@ -171,7 +182,7 @@ func (t *TomlTree) SetPath(keys []string, value interface{}) {
 		nextTree, exists := subtree.values[intermediateKey]
 		if !exists {
 			nextTree = newTomlTree()
-			subtree.values[intermediateKey] = &nextTree // add new element here
+			subtree.values[intermediateKey] = nextTree // add new element here
 		}
 		switch node := nextTree.(type) {
 		case *TomlTree:
@@ -185,7 +196,21 @@ func (t *TomlTree) SetPath(keys []string, value interface{}) {
 			subtree = node[len(node)-1]
 		}
 	}
-	subtree.values[keys[len(keys)-1]] = value
+
+	var toInsert interface{}
+
+	switch value.(type) {
+	case *TomlTree:
+		toInsert = value
+	case []*TomlTree:
+		toInsert = value
+	case *tomlValue:
+		toInsert = value
+	default:
+		toInsert = &tomlValue{value: value}
+	}
+
+	subtree.values[keys[len(keys)-1]] = toInsert
 }
 
 // createSubTree takes a tree and a key and create the necessary intermediate
@@ -215,8 +240,8 @@ func (t *TomlTree) createSubTree(keys []string, pos Position) error {
 		case *TomlTree:
 			subtree = node
 		default:
-			return fmt.Errorf("unknown type for path %s (%s)",
+			return fmt.Errorf("unknown type for path %s (%s): %T (%#v)",
-				strings.Join(keys, "."), intermediateKey)
+				strings.Join(keys, "."), intermediateKey, nextTree, nextTree)
 		}
 	}
 	return nil
@@ -306,10 +331,17 @@ func (t *TomlTree) toToml(indent, keyspace string) string {
 				result += fmt.Sprintf("\n%s[%s]\n", indent, combinedKey)
 			}
 			result += node.toToml(indent+"  ", combinedKey)
+		case map[string]interface{}:
+			sub := TreeFromMap(node)
+
+			if len(sub.Keys()) > 0 {
+				result += fmt.Sprintf("\n%s[%s]\n", indent, combinedKey)
+			}
+			result += sub.toToml(indent+"  ", combinedKey)
 		case *tomlValue:
 			result += fmt.Sprintf("%s%s = %s\n", indent, k, toTomlValue(node.value, 0))
 		default:
-			panic(fmt.Sprintf("unsupported node type: %v", node))
+			result += fmt.Sprintf("%s%s = %s\n", indent, k, toTomlValue(v, 0))
 		}
 	}
 	return result
@@ -329,8 +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 {
@@ -339,18 +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) {