Few performance improvements

replaceEscapes() got called for every string, and key.String() gets
called a lot in the parser, so small improvements add up. Also figured
that calling replaceEscapes() for every string isn't really needed.

It's about 20 to 30% faster (depending on the TOML file).

bench_test.go

@@ -52,7 +52,7 @@ func BenchmarkDecode(b *testing.B) {
 	}
 
 	b.Run("large-doc", func(b *testing.B) {
-		d, err := os.ReadFile("testdata/ja-JP.toml")
+		d, err := os.ReadFile("testdata/Cargo.toml")
 		if err != nil {
 			b.Fatal(err)
 		}

decode_test.go

@@ -1290,7 +1290,7 @@ func TestMetaKeys(t *testing.T) {
 }
 
 func TestDecodeParallel(t *testing.T) {
-	doc, err := os.ReadFile("testdata/ja-JP.toml")
+	doc, err := os.ReadFile("testdata/Cargo.toml")
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -1323,3 +1323,21 @@ func errorContains(have error, want string) bool {
 	}
 	return strings.Contains(have.Error(), want)
 }
+
+func BenchmarkEscapes(b *testing.B) {
+	p := new(parser)
+	it := item{}
+	str := strings.Repeat("hello, world!\n", 10)
+	b.ResetTimer()
+	for n := 0; n < b.N; n++ {
+		p.replaceEscapes(it, str)
+	}
+}
+
+func BenchmarkKey(b *testing.B) {
+	k := Key{"cargo-credential-macos-keychain", "version"}
+	b.ResetTimer()
+	for n := 0; n < b.N; n++ {
+		k.String()
+	}
+}

lex.go

@@ -17,6 +17,7 @@ const (
 	itemEOF
 	itemText
 	itemString
+	itemStringEsc
 	itemRawString
 	itemMultilineString
 	itemRawMultilineString
@@ -53,6 +54,7 @@ type lexer struct {
 	state    stateFn
 	items    chan item
 	tomlNext bool
+	esc      bool
 
 	// Allow for backing up up to 4 runes. This is necessary because TOML
 	// contains 3-rune tokens (""" and ''').
@@ -696,7 +698,12 @@ func lexString(lx *lexer) stateFn {
 		return lexStringEscape
 	case r == '"':
 		lx.backup()
-		lx.emit(itemString)
+		if lx.esc {
+			lx.esc = false
+			lx.emit(itemStringEsc)
+		} else {
+			lx.emit(itemString)
+		}
 		lx.next()
 		lx.ignore()
 		return lx.pop()
@@ -746,6 +753,7 @@ func lexMultilineString(lx *lexer) stateFn {
 				lx.backup() /// backup: don't include the """ in the item.
 				lx.backup()
 				lx.backup()
+				lx.esc = false
 				lx.emit(itemMultilineString)
 				lx.next() /// Read over ''' again and discard it.
 				lx.next()
@@ -835,6 +843,7 @@ func lexMultilineStringEscape(lx *lexer) stateFn {
 }
 
 func lexStringEscape(lx *lexer) stateFn {
+	lx.esc = true
 	r := lx.next()
 	switch r {
 	case 'e':
@@ -1199,7 +1208,7 @@ func (itype itemType) String() string {
 		return "EOF"
 	case itemText:
 		return "Text"
-	case itemString, itemRawString, itemMultilineString, itemRawMultilineString:
+	case itemString, itemStringEsc, itemRawString, itemMultilineString, itemRawMultilineString:
 		return "String"
 	case itemBool:
 		return "Bool"

meta.go

@@ -94,21 +94,41 @@ func (md *MetaData) Undecoded() []Key {
 type Key []string
 
 func (k Key) String() string {
-	ss := make([]string, len(k))
-	for i := range k {
-		ss[i] = k.maybeQuoted(i)
+	// This is called quite often, so it's a bit funky to make it faster.
+	var b strings.Builder
+	b.Grow(len(k) * 25)
+outer:
+	for i, kk := range k {
+		if i > 0 {
+			b.WriteByte('.')
+		}
+		if kk == "" {
+			b.WriteString(`""`)
+		} else {
+			for _, r := range kk {
+				// "Inline" isBareKeyChar
+				if !((r >= 'A' && r <= 'Z') || (r >= 'a' && r <= 'z') || (r >= '0' && r <= '9') || r == '_' || r == '-') {
+					b.WriteByte('"')
+					b.WriteString(dblQuotedReplacer.Replace(kk))
+					b.WriteByte('"')
+					continue outer
+				}
+			}
+			b.WriteString(kk)
+		}
 	}
-	return strings.Join(ss, ".")
+	return b.String()
 }
 
 func (k Key) maybeQuoted(i int) string {
 	if k[i] == "" {
 		return `""`
 	}
-	for _, c := range k[i] {
-		if !isBareKeyChar(c, false) {
-			return `"` + dblQuotedReplacer.Replace(k[i]) + `"`
+	for _, r := range k[i] {
+		if (r >= 'A' && r <= 'Z') || (r >= 'a' && r <= 'z') || (r >= '0' && r <= '9') || r == '_' || r == '-' {
+			continue
 		}
+		return `"` + dblQuotedReplacer.Replace(k[i]) + `"`
 	}
 	return k[i]
 }

parse.go

@@ -224,7 +224,7 @@ func (p *parser) keyString(it item) string {
 	switch it.typ {
 	case itemText:
 		return it.val
-	case itemString, itemMultilineString,
+	case itemString, itemStringEsc, itemMultilineString,
 		itemRawString, itemRawMultilineString:
 		s, _ := p.value(it, false)
 		return s.(string)
@@ -244,6 +244,8 @@ var datetimeRepl = strings.NewReplacer(
 func (p *parser) value(it item, parentIsArray bool) (any, tomlType) {
 	switch it.typ {
 	case itemString:
+		return it.val, p.typeOfPrimitive(it)
+	case itemStringEsc:
 		return p.replaceEscapes(it, it.val), p.typeOfPrimitive(it)
 	case itemMultilineString:
 		return p.replaceEscapes(it, p.stripEscapedNewlines(stripFirstNewline(it.val))), p.typeOfPrimitive(it)
@@ -707,8 +709,11 @@ func stripFirstNewline(s string) string {
 // the next newline. After a line-ending backslash, all whitespace is removed
 // until the next non-whitespace character.
 func (p *parser) stripEscapedNewlines(s string) string {
-	var b strings.Builder
-	var i int
+	var (
+		b strings.Builder
+		i int
+	)
+	b.Grow(len(s))
 	for {
 		ix := strings.Index(s[i:], `\`)
 		if ix < 0 {
@@ -738,9 +743,8 @@ func (p *parser) stripEscapedNewlines(s string) string {
 			continue
 		}
 		if !strings.Contains(s[i:j], "\n") {
-			// This is not a line-ending backslash.
-			// (It's a bad escape sequence, but we can let
-			// replaceEscapes catch it.)
+			// This is not a line-ending backslash. (It's a bad escape sequence,
+			// but we can let replaceEscapes catch it.)
 			i++
 			continue
 		}
@@ -751,79 +755,78 @@ func (p *parser) stripEscapedNewlines(s string) string {
 }
 
 func (p *parser) replaceEscapes(it item, str string) string {
-	replaced := make([]rune, 0, len(str))
-	s := []byte(str)
-	r := 0
-	for r < len(s) {
-		if s[r] != '\\' {
-			c, size := utf8.DecodeRune(s[r:])
-			r += size
-			replaced = append(replaced, c)
+	var (
+		b    strings.Builder
+		skip = 0
+	)
+	b.Grow(len(str))
+	for i, c := range str {
+		if skip > 0 {
+			skip--
+			continue
+		}
+		if c != '\\' {
+			b.WriteRune(c)
 			continue
 		}
-		r += 1
-		if r >= len(s) {
+
+		if i >= len(str) {
 			p.bug("Escape sequence at end of string.")
 			return ""
 		}
-		switch s[r] {
+		switch str[i+1] {
 		default:
-			p.bug("Expected valid escape code after \\, but got %q.", s[r])
+			p.bug("Expected valid escape code after \\, but got %q.", str[i+1])
 		case ' ', '\t':
-			p.panicItemf(it, "invalid escape: '\\%c'", s[r])
+			p.panicItemf(it, "invalid escape: '\\%c'", str[i+1])
 		case 'b':
-			replaced = append(replaced, rune(0x0008))
-			r += 1
+			b.WriteByte(0x08)
+			skip = 1
 		case 't':
-			replaced = append(replaced, rune(0x0009))
-			r += 1
+			b.WriteByte(0x09)
+			skip = 1
 		case 'n':
-			replaced = append(replaced, rune(0x000A))
-			r += 1
+			b.WriteByte(0x0a)
+			skip = 1
 		case 'f':
-			replaced = append(replaced, rune(0x000C))
-			r += 1
+			b.WriteByte(0x0c)
+			skip = 1
 		case 'r':
-			replaced = append(replaced, rune(0x000D))
-			r += 1
+			b.WriteByte(0x0d)
+			skip = 1
 		case 'e':
 			if p.tomlNext {
-				replaced = append(replaced, rune(0x001B))
-				r += 1
+				b.WriteByte(0x1b)
+				skip = 1
 			}
 		case '"':
-			replaced = append(replaced, rune(0x0022))
-			r += 1
+			b.WriteByte(0x22)
+			skip = 1
 		case '\\':
-			replaced = append(replaced, rune(0x005C))
-			r += 1
+			b.WriteByte(0x5c)
+			skip = 1
+		// The lexer guarantees the correct number of characters are present;
+		// don't need to check here.
 		case 'x':
 			if p.tomlNext {
-				escaped := p.asciiEscapeToUnicode(it, s[r+1:r+3])
-				replaced = append(replaced, escaped)
-				r += 3
+				escaped := p.asciiEscapeToUnicode(it, str[i+2:i+4])
+				b.WriteRune(escaped)
+				skip = 3
 			}
 		case 'u':
-			// At this point, we know we have a Unicode escape of the form
-			// `uXXXX` at [r, r+5). (Because the lexer guarantees this
-			// for us.)
-			escaped := p.asciiEscapeToUnicode(it, s[r+1:r+5])
-			replaced = append(replaced, escaped)
-			r += 5
+			escaped := p.asciiEscapeToUnicode(it, str[i+2:i+6])
+			b.WriteRune(escaped)
+			skip = 5
 		case 'U':
-			// At this point, we know we have a Unicode escape of the form
-			// `uXXXX` at [r, r+9). (Because the lexer guarantees this
-			// for us.)
-			escaped := p.asciiEscapeToUnicode(it, s[r+1:r+9])
-			replaced = append(replaced, escaped)
-			r += 9
+			escaped := p.asciiEscapeToUnicode(it, str[i+2:i+10])
+			b.WriteRune(escaped)
+			skip = 9
 		}
 	}
-	return string(replaced)
+	return b.String()
 }
 
-func (p *parser) asciiEscapeToUnicode(it item, bs []byte) rune {
-	s := string(bs)
+func (p *parser) asciiEscapeToUnicode(it item, s string) rune {
 	hex, err := strconv.ParseUint(strings.ToLower(s), 16, 32)
 	if err != nil {
 		p.bug("Could not parse '%s' as a hexadecimal number, but the lexer claims it's OK: %s", s, err)