Implemented a separated lexer.StringBuffer, to not pollute lexer code with string building code. The string builder can provide built strings both as literal as-is string (in TOML: single quotes), or as interpreted strings (in TOML: between double quotes)

lexer/items.go

@@ -17,7 +17,7 @@ const (
 
 // Item represents a lexer item returned from the scanner.
 type Item struct {
-	Type  itemType //Type, e.g. itemNumber, itemSquareBracket
+	Type  itemType //Type, e.g. ItemComment, ItemString
 	Value string   // Value, e.g. "10.42", "["
 }
 

lexer/lexer.go

@@ -7,16 +7,16 @@ import (
 	"unicode/utf8"
 )
 
-// Lexer holds the state of the scanner.
+// Lexer holds the state of the lexer.
 type Lexer struct {
 	input    string       // the scanned input string
-	state    stateFn         // the current state
-	stack    []stateFn       // state stack, for nested parsing
+	state    stateFn      // a function that handles the current state
+	stack    []stateFn    // state function stack, for nested parsing
 	start    int          // start position of the currently scanned item
 	pos      int          // current scanning position in the input
-	width    int             // width of the last rune read
-	strValue strings.Builder // used to build string values
-	items    chan Item       // channel of scanned items
+	width    int          // width of the last rune read, for supporting backup()
+	buffer   StringBuffer // an efficient buffer, used to build string values
+	items    chan Item    // channel of resulting lexer items
 	nextItem Item         // the current item as reached by Next() and retrieved by Get()
 	err      error        // an error message when lexing failed, retrieved by Error()
 }
@@ -69,6 +69,11 @@ func (l *Lexer) Error() error {
 	return l.err
 }
 
+// Get returns the next lexer item, as reached by Next()
+func (l *Lexer) Get() Item {
+	return l.nextItem
+}
+
 // ToArray returns lexer items as an array.
 // When an error occurs during scanning, a partial result will be
 // returned, accompanied by the error that occurred.
@@ -80,11 +85,6 @@ func (l *Lexer) ToArray() ([]Item, error) {
 	return items, l.Error()
 }
 
-// Get returns the next lexer item, as reached by Next()
-func (l *Lexer) Get() Item {
-	return l.nextItem
-}
-
 // pushState adds the state function to its stack.
 // This is used for implementing nested parsing.
 func (l *Lexer) pushState(state stateFn) {
@@ -99,6 +99,7 @@ func (l *Lexer) popState() stateFn {
 	return tail
 }
 
+// TODO niet meer nodig?
 // getAcceptedString returns the string as accepted by the
 // accept* methods so far.
 func (l *Lexer) getAcceptedString() string {
@@ -111,6 +112,7 @@ func (l *Lexer) emit(t itemType, v string) {
 	l.start = l.pos
 }
 
+// TODO niet meer nodig met stringbuilder?
 // ignore skips over the pending input before the current position.
 func (l *Lexer) ignore() {
 	l.start = l.pos
@@ -133,6 +135,7 @@ func (l *Lexer) peek() rune {
 	return r
 }
 
+// TODO nog nodig met stringbuffer?
 // accept consumes the next rune if it's from the valid set of runes.
 func (l *Lexer) accept(runes string) bool {
 	if strings.IndexRune(runes, l.next()) >= 0 {
@@ -150,6 +153,7 @@ func (l *Lexer) upcoming(runes string) bool {
 	return false
 }
 
+// TODO nog nodig met stringbuffer?
 // acceptNot consumes the next rune if it's not from the set of runes.
 func (l *Lexer) acceptNot(runes string) bool {
 	r := l.next()
@@ -198,24 +202,7 @@ func (l *Lexer) skipUntil(runes string) {
 	}
 }
 
-// resetStringBuild initializes a new string builder, used for building
-// string by interpreting input data, e.g. for translating
-// double quoted strings with escape codes into an actual
-// Go string value.
-func (l *Lexer) resetStringBuilder() {
-	l.strValue.Reset()
-}
-
-// addToString adds a rune to the string builder.
-func (l *Lexer) addToString(r rune) {
-	l.strValue.WriteRune(r)
-}
-
-// getString returns the runes in the string builder as a string value.
-func (l *Lexer) getString() string {
-	return l.strValue.String()
-}
-
+// TODO meh... ugly rune.
 var endOfFile rune = -1
 
 // next returns the next rune in the input.

lexer/lexer_test.go

@@ -23,19 +23,24 @@ func TestWhiteSpaceAndNewlines(t *testing.T) {
 	assertSuccessAndCheck(t, "\n", []string{})
 	assertSuccessAndCheck(t, "\n \t\r\n", []string{})
 }
-func TestWhitespacePlusComment(t *testing.T) {
+func TestComments(t *testing.T) {
 	assertSuccessAndCheck(t, "#", []string{`Comment("#")`})
 	assertSuccessAndCheck(t, " \t \t #", []string{`Comment("#")`})
 	assertSuccessAndCheck(t, " \t \t # not empty", []string{`Comment("# not empty")`})
 	assertSuccessAndCheck(t, " \t \t # not empty\r\r\r\n", []string{`Comment("# not empty")`})
 	assertSuccessAndCheck(t, "\n \t\r\n# AAP\r\n", []string{`Comment("# AAP")`})
-	assertSuccessAndCheck(t, "# two lines\n# of comments\n",
+	assertSuccessAndCheck(t,
+		"# two lines\n# of comments\n",
 		[]string{`Comment("# two lines")`, `Comment("# of comments")`})
+	assertSuccessAndCheck(t,
+		`# \tcomment\nwith escape-y chars`,
+		[]string{`Comment("# \\tcomment\\nwith escape-y chars")`})
 }
 
 func TestBareKeyWithoutValue(t *testing.T) {
 	err := "Unexpected end of file (expected an '=' value assignment)"
 	assertFailureAndCheck(t, "a", []string{`Key("a")`}, err)
+	assertFailureAndCheck(t, "_", []string{`Key("_")`}, err)
 	assertFailureAndCheck(t, " a", []string{`Key("a")`}, err)
 	assertFailureAndCheck(t, " a ", []string{`Key("a")`}, err)
 	assertFailureAndCheck(t, "ab", []string{`Key("ab")`}, err)
@@ -95,11 +100,13 @@ func TestBasicStringEscapes(t *testing.T) {
 		`\f`:              "\f",
 		`\r`:              "\r",
 		`\"`:              "\"",
-		`\b\t\n\f\r\"`: "\b\t\n\f\r\"",
+		`\b\t\nhuh\f\r\"`: "\b\t\nhuh\f\r\"",
+		`\u2318`:          "⌘",
+		`\U0001014D`:      "𐅍",
 	} {
 		l := assertSuccess(t, fmt.Sprintf(`x="%s"`, in))
 		if out != l[1].Value {
-			t.Fatalf("Unexpected result when parsing '%s'", in)
+			t.Fatalf("Unexpected result when parsing '%s'\nexpected: %q\nactual: %q", in, out, l[1].Value)
 		}
 	}
 }

lexer/states.go

@@ -22,6 +22,7 @@ const (
 	someQuote      string = singleQuote + doubleQuote
 	bareKey        string = lower + upper + digits + underscore + dash
 	startOfKey     string = bareKey + someQuote
+	quotable       string = `btnfr\"`
 )
 
 func stateKeyValuePair(l *Lexer) stateFn {
@@ -37,16 +38,17 @@ func stateKeyValuePair(l *Lexer) stateFn {
 
 // A '#' hash symbol marks the rest of the line as a comment.
 func stateComment(l *Lexer) stateFn {
-	l.resetStringBuilder()
+	l.buffer.Reset()
 	for {
 		switch {
 		case l.atEndOfFile() || l.accept(newline):
-			l.emit(ItemComment, l.getString())
+			s := l.buffer.AsLiteralString()
+			l.emit(ItemComment, s)
 			return stateKeyValuePair
 		case l.accept(carriageReturn):
 			l.ignore()
 		default:
-			l.addToString(l.next())
+			l.buffer.WriteRune(l.next())
 		}
 	}
 }
@@ -113,24 +115,6 @@ func stateStringValue(l *Lexer) stateFn {
 	return l.unexpectedTokenError("a string value")
 }
 
-func stateBasicStringValue(l *Lexer) stateFn {
-	// Possibly a """ multi-line string start,
-	// possibly the end of an "" empty string.
-	if l.accept(doubleQuote) {
-		// A """ multi-line string.
-		if l.accept(doubleQuote) {
-			l.ignore()
-			return stateMultiLineBasicString
-		}
-		// An "" empty string.
-		l.ignore()
-		l.emit(ItemString, "")
-		return stateKeyValuePair
-	}
-	l.ignore()
-	return stateBasicString
-}
-
 // Basic strings are surrounded by quotation marks. Any Unicode character
 // may be used except those that must be escaped: quotation mark, backslash,
 // and the control characters (U+0000 to U+001F, U+007F).
@@ -152,15 +136,22 @@ func stateBasicStringValue(l *Lexer) stateFn {
 //
 // All other escape sequences not listed above are reserved and,
 // if used, TOML should produce an error.
-
-var basicEscapes = map[rune]rune{
-	'b':  rune(8),
-	't':  rune(9),
-	'n':  rune(10),
-	'f':  rune(12),
-	'r':  rune(13),
-	'"':  rune(34),
-	'\\': rune(92),
+func stateBasicStringValue(l *Lexer) stateFn {
+	// Possibly a """ multi-line string start,
+	// possibly the end of an "" empty string.
+	if l.accept(doubleQuote) {
+		// It's a """ multi-line string.
+		if l.accept(doubleQuote) {
+			l.ignore()
+			return stateMultiLineBasicString
+		}
+		// It's an "" empty string.
+		l.ignore()
+		l.emit(ItemString, "")
+		return stateKeyValuePair
+	}
+	l.ignore()
+	return stateBasicString
 }
 
 func stateParseBasicString(l *Lexer) stateFn {
@@ -171,22 +162,26 @@ func stateParseBasicString(l *Lexer) stateFn {
 		case l.accept(doubleQuote):
 			return l.popState()
 		case l.accept(backslash):
-			r := l.next()
-			if escaped, ok := basicEscapes[r]; ok {
-				l.addToString(escaped)
+			if l.upcoming(quotable) {
+				l.buffer.WriteRune('\\')
+				l.buffer.WriteRune(l.next())
 			} else {
-				return l.errorf("Invalid escape sequence \\%c in string value", r)
+				return l.errorf("Invalid escape sequence \\%c in string value", l.next())
 			}
 		default:
-			l.addToString(l.next())
+			l.buffer.WriteRune(l.next())
 		}
 	}
 }
 
 func stateBasicString(l *Lexer) stateFn {
-	l.resetStringBuilder()
+	l.buffer.Reset()
 	l.pushState(func(l *Lexer) stateFn {
-		l.emit(ItemString, l.getString())
+		s, err := l.buffer.AsInterpretedString()
+		if err != nil {
+			return l.errorf("Invalid data in string: %s", err)
+		}
+		l.emit(ItemString, s)
 		return stateKeyValuePair
 	})
 	return stateParseBasicString

lexer/stringbuf.go

@@ -0,0 +1,62 @@
+package lexer
+
+import (
+	"bytes"
+	"strconv"
+	"strings"
+)
+
+// StringBuffer is a string buffer implementation, which is used by the lexer
+// to efficiently accumulate runes from the input and eventually turn these
+// into a string, either literal or interpreted.
+type StringBuffer struct {
+	buffer bytes.Buffer
+}
+
+// Reset resets the string buffer, in order to build a new string.
+func (b *StringBuffer) Reset() *StringBuffer {
+	b.buffer.Reset()
+	return b
+}
+
+// AddString adds the runes of the input string to the string buffer.
+func (b *StringBuffer) WriteString(s string) *StringBuffer {
+	for _, r := range s {
+		b.WriteRune(r)
+	}
+	return b
+}
+
+// WriteRune adds a single rune to the string buffer.
+func (b *StringBuffer) WriteRune(r rune) *StringBuffer {
+	b.buffer.WriteRune(r)
+	return b
+}
+
+// AsLiteralString returns the string buffer as a literal string.
+// Literal means that no escape sequences are processed.
+func (b *StringBuffer) AsLiteralString() string {
+	return b.buffer.String()
+}
+
+// AsInterpretedString returns the string in its interpreted form.
+// Interpreted means that escape sequences are handled in the way that Go would
+// have, had it been inside double quotes. It translates for example escape
+// sequences like "\n", "\t", \uXXXX" and "\UXXXXXXXX" into their string
+// representations.
+// Since the input might contain invalid escape sequences, this method
+// also returns an error. When an error is returned, the returned string will
+// contain the string as far as it could be interpreted.
+func (b *StringBuffer) AsInterpretedString() (string, error) {
+	var sb strings.Builder
+	tail := b.buffer.String()
+	for len(tail) > 0 {
+		r, _, newtail, err := strconv.UnquoteChar(tail, '"')
+		if err != nil {
+			return sb.String(), err
+		}
+		tail = newtail
+		sb.WriteRune(r)
+	}
+	return sb.String(), nil
+}

lexer/stringbuf_test.go

@@ -0,0 +1,87 @@
+package lexer_test
+
+import "testing"
+import "github.com/mmakaay/toml/lexer"
+
+func TestGeneratingStringDoesNotResetBuffer(t *testing.T) {
+	var b lexer.StringBuffer
+	s1, _ := b.WriteString(`hi\nthere`).AsInterpretedString()
+	s2 := b.AsLiteralString()
+	if s1 != "hi\nthere" {
+		t.Fatalf("Did not get expected string\"X\" for try 1, but %q", s1)
+	}
+	if s2 != "hi\\nthere" {
+		t.Fatalf("Did not get expected string\"X\" for try 2, but %q", s2)
+	}
+}
+
+func TestResetResetsBuffer(t *testing.T) {
+	var b lexer.StringBuffer
+	s := b.WriteRune('X').Reset().AsLiteralString()
+	if s != "" {
+		t.Fatalf("Did not get expected empty string, but %q", s)
+	}
+}
+
+type testCase struct {
+	name          string
+	in            string
+	out           string
+	isSuccessCase bool
+}
+
+const (
+	OK   bool = true
+	FAIL bool = false
+)
+
+func TestAsLiteralString(t *testing.T) {
+	b := lexer.StringBuffer{}
+	for _, c := range []testCase{
+		{"empty string", ``, ``, OK},
+		{"simple string", `Simple string!`, `Simple string!`, OK},
+		{"single quote", `'`, `'`, OK},
+		{"double quote", `"`, `"`, OK},
+		{"escaped single quote", `\'`, `\'`, OK},
+		{"escaped double quote", `\"`, `\"`, OK},
+		{"escape anything", `\x\t\f\n\r\'\"\\`, `\x\t\f\n\r\'\"\\`, OK},
+		{"UTF8 escapes", `\uceb2\U00e0b8bf`, `\uceb2\U00e0b8bf`, OK},
+		{"actual newline", "on\nmultiple\nlines", "on\nmultiple\nlines", OK},
+	} {
+		s := b.Reset().WriteString(c.in).AsLiteralString()
+		if s != c.out {
+			t.Fatalf("[%s] %q -> %q failed: actual result = %q", c.name, c.in, c.out, s)
+		}
+	}
+}
+
+func TestAsInterpretedString(t *testing.T) {
+	b := lexer.StringBuffer{}
+	for _, c := range []testCase{
+		{"empty string", "", "", OK},
+		{"one character", "Simple string!", "Simple string!", OK},
+		{"escaped single quote", `\'`, "", FAIL},
+		{"escaped double quote", `\"`, `"`, OK},
+		{"bare single quote", `'`, "'", OK},
+		{"string in single quotes", `'Hello'`, `'Hello'`, OK},
+		{"string in escaped double quotes", `\"Hello\"`, `"Hello"`, OK},
+		{"escape something", `\t\f\n\r\"\\`, "\t\f\n\r\"\\", OK},
+		{"short UTF8 escapes", `\u2318Wh\u00e9\u00e9!`, `⌘Whéé!`, OK},
+		{"long UTF8 escapes", `\U0001014D \u2318 Wh\u00e9\u00e9!`, `𐅍 ⌘ Whéé!`, OK},
+		{"UTF8 characters", "Ѝюج wut Ж ?", "Ѝюج wut Ж ?", OK},
+		{"example from spec",
+			`I'm a string. \"You can quote me\". Name\tJos\u00E9\nLocation\tSF.`,
+			"I'm a string. \"You can quote me\". Name\tJosé\nLocation\tSF.", OK},
+	} {
+		s, err := b.Reset().WriteString(c.in).AsInterpretedString()
+		if c.isSuccessCase && err != nil {
+			t.Fatalf("[%s] unexpected error for input %q: %s", c.name, c.in, err)
+		}
+		if !c.isSuccessCase && err == nil {
+			t.Fatalf("[%s] expected a failure, but no failure occurred", c.name)
+		}
+		if s != c.out && c.isSuccessCase {
+			t.Fatalf("[%s] %q -> %q failed: actual result = %q", c.name, c.in, c.out, s)
+		}
+	}
+}