unicode/utf8: table-based algorithm for decoding

This simplifies covering all cases, reducing the number of branches
and making unrolling for simpler functions manageable.
This significantly improves performance of non-ASCII input.

This change will also allow addressing Issue #11733 in an efficient
manner.

RuneCountTenASCIIChars-8             13.7ns ± 4%  13.5ns ± 2%     ~     (p=0.116 n=7+8)
RuneCountTenJapaneseChars-8           153ns ± 3%    74ns ± 2%  -51.42%  (p=0.000 n=8+8)
RuneCountInStringTenASCIIChars-8     13.5ns ± 2%  12.5ns ± 3%   -7.13%  (p=0.000 n=8+7)
RuneCountInStringTenJapaneseChars-8   145ns ± 2%    68ns ± 2%  -53.21%  (p=0.000 n=8+8)
ValidTenASCIIChars-8                 14.1ns ± 3%  12.5ns ± 5%  -11.38%  (p=0.000 n=8+8)
ValidTenJapaneseChars-8               147ns ± 3%    71ns ± 4%  -51.72%  (p=0.000 n=8+8)
ValidStringTenASCIIChars-8           12.5ns ± 3%  12.3ns ± 3%     ~     (p=0.095 n=8+8)
ValidStringTenJapaneseChars-8         146ns ± 4%    70ns ± 2%  -51.62%  (p=0.000 n=8+7)
DecodeASCIIRune-8                    5.91ns ± 2%  4.83ns ± 3%  -18.28%  (p=0.001 n=7+7)
DecodeJapaneseRune-8                 12.2ns ± 7%   8.5ns ± 3%  -29.79%  (p=0.000 n=8+7)
FullASCIIRune-8                      5.95ns ± 3%  4.27ns ± 1%  -28.23%  (p=0.000 n=8+7)
FullJapaneseRune-8                   12.0ns ± 6%   4.3ns ± 3%  -64.39%  (p=0.000 n=8+8)

Change-Id: Iea1d6b0180cbbee1739659a0a38038126beecaca
Reviewed-on: https://go-review.googlesource.com/16940
Reviewed-by: Russ Cox <rsc@golang.org>

2 files changed, 269 insertions, 197 deletions

diff --git a/src/unicode/utf8/utf8.go b/src/unicode/utf8/utf8.go
index 5d29ec080e..bcb0de402e 100644
--- a/src/unicode/utf8/utf8.go
+++ b/src/unicode/utf8/utf8.go
@@ -40,175 +40,106 @@ const (
 	rune1Max = 1<<7 - 1
 	rune2Max = 1<<11 - 1
 	rune3Max = 1<<16 - 1
-)
-
-func decodeRuneInternal(p []byte) (r rune, size int, short bool) {
-	n := len(p)
-	if n < 1 {
-		return RuneError, 0, true
-	}
-	c0 := p[0]
-
-	// 1-byte, 7-bit sequence?
-	if c0 < tx {
-		return rune(c0), 1, false
-	}
-
-	// unexpected continuation byte?
-	if c0 < t2 {
-		return RuneError, 1, false
-	}
-
-	// need first continuation byte
-	if n < 2 {
-		return RuneError, 1, true
-	}
-	c1 := p[1]
-	if c1 < tx || t2 <= c1 {
-		return RuneError, 1, false
-	}
-
-	// 2-byte, 11-bit sequence?
-	if c0 < t3 {
-		r = rune(c0&mask2)<<6 | rune(c1&maskx)
-		if r <= rune1Max {
-			return RuneError, 1, false
-		}
-		return r, 2, false
-	}
-
-	// need second continuation byte
-	if n < 3 {
-		return RuneError, 1, true
-	}
-	c2 := p[2]
-	if c2 < tx || t2 <= c2 {
-		return RuneError, 1, false
-	}
-
-	// 3-byte, 16-bit sequence?
-	if c0 < t4 {
-		r = rune(c0&mask3)<<12 | rune(c1&maskx)<<6 | rune(c2&maskx)
-		if r <= rune2Max {
-			return RuneError, 1, false
-		}
-		if surrogateMin <= r && r <= surrogateMax {
-			return RuneError, 1, false
-		}
-		return r, 3, false
-	}
-
-	// need third continuation byte
-	if n < 4 {
-		return RuneError, 1, true
-	}
-	c3 := p[3]
-	if c3 < tx || t2 <= c3 {
-		return RuneError, 1, false
-	}
 
-	// 4-byte, 21-bit sequence?
-	if c0 < t5 {
-		r = rune(c0&mask4)<<18 | rune(c1&maskx)<<12 | rune(c2&maskx)<<6 | rune(c3&maskx)
-		if r <= rune3Max || MaxRune < r {
-			return RuneError, 1, false
-		}
-		return r, 4, false
-	}
+	// The default lowest and highest continuation byte.
+	locb = 0x80 // 1000 0000
+	hicb = 0xBF // 1011 1111
+
+	// These names of these constants are chosen to give nice alignment in the
+	// table below. The first nibble is an index into acceptRanges or F for
+	// special one-byte cases. The second nibble is the Rune length or the
+	// Status for the special one-byte case.
+	xx = 0xF1 // invalid: size 1
+	as = 0xF0 // ASCII: size 1
+	s1 = 0x02 // accept 0, size 2
+	s2 = 0x13 // accept 1, size 3
+	s3 = 0x03 // accept 0, size 3
+	s4 = 0x23 // accept 2, size 3
+	s5 = 0x34 // accept 3, size 4
+	s6 = 0x04 // accept 0, size 4
+	s7 = 0x44 // accept 4, size 4
+)
 
-	// error
-	return RuneError, 1, false
+// first is information about the first byte in a UTF-8 sequence.
+var first = [256]uint8{
+	//   1   2   3   4   5   6   7   8   9   A   B   C   D   E   F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x00-0x0F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x10-0x1F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x20-0x2F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x30-0x3F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x40-0x4F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x50-0x5F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x60-0x6F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x70-0x7F
+	//   1   2   3   4   5   6   7   8   9   A   B   C   D   E   F
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0x80-0x8F
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0x90-0x9F
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xA0-0xAF
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xB0-0xBF
+	xx, xx, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, // 0xC0-0xCF
+	s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, // 0xD0-0xDF
+	s2, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s4, s3, s3, // 0xE0-0xEF
+	s5, s6, s6, s6, s7, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xF0-0xFF
 }
 
-func decodeRuneInStringInternal(s string) (r rune, size int, short bool) {
-	n := len(s)
-	if n < 1 {
-		return RuneError, 0, true
-	}
-	c0 := s[0]
-
-	// 1-byte, 7-bit sequence?
-	if c0 < tx {
-		return rune(c0), 1, false
-	}
-
-	// unexpected continuation byte?
-	if c0 < t2 {
-		return RuneError, 1, false
-	}
-
-	// need first continuation byte
-	if n < 2 {
-		return RuneError, 1, true
-	}
-	c1 := s[1]
-	if c1 < tx || t2 <= c1 {
-		return RuneError, 1, false
-	}
-
-	// 2-byte, 11-bit sequence?
-	if c0 < t3 {
-		r = rune(c0&mask2)<<6 | rune(c1&maskx)
-		if r <= rune1Max {
-			return RuneError, 1, false
-		}
-		return r, 2, false
-	}
-
-	// need second continuation byte
-	if n < 3 {
-		return RuneError, 1, true
-	}
-	c2 := s[2]
-	if c2 < tx || t2 <= c2 {
-		return RuneError, 1, false
-	}
-
-	// 3-byte, 16-bit sequence?
-	if c0 < t4 {
-		r = rune(c0&mask3)<<12 | rune(c1&maskx)<<6 | rune(c2&maskx)
-		if r <= rune2Max {
-			return RuneError, 1, false
-		}
-		if surrogateMin <= r && r <= surrogateMax {
-			return RuneError, 1, false
-		}
-		return r, 3, false
-	}
-
-	// need third continuation byte
-	if n < 4 {
-		return RuneError, 1, true
-	}
-	c3 := s[3]
-	if c3 < tx || t2 <= c3 {
-		return RuneError, 1, false
-	}
-
-	// 4-byte, 21-bit sequence?
-	if c0 < t5 {
-		r = rune(c0&mask4)<<18 | rune(c1&maskx)<<12 | rune(c2&maskx)<<6 | rune(c3&maskx)
-		if r <= rune3Max || MaxRune < r {
-			return RuneError, 1, false
-		}
-		return r, 4, false
-	}
+// acceptRange gives the range of valid values for the second byte in a UTF-8
+// sequence.
+type acceptRange struct {
+	lo uint8 // lowest value for second byte.
+	hi uint8 // highest value for second byte.
+}
 
-	// error
-	return RuneError, 1, false
+var acceptRanges = [...]acceptRange{
+	0: {locb, hicb},
+	1: {0xA0, hicb},
+	2: {locb, 0x9F},
+	3: {0x90, hicb},
+	4: {locb, 0x8F},
 }
 
 // FullRune reports whether the bytes in p begin with a full UTF-8 encoding of a rune.
 // An invalid encoding is considered a full Rune since it will convert as a width-1 error rune.
 func FullRune(p []byte) bool {
-	_, _, short := decodeRuneInternal(p)
-	return !short
+	n := len(p)
+	if n == 0 {
+		return false
+	}
+	x := first[p[0]]
+	if n >= int(x&7) {
+		return true // ASCII, invalid or valid.
+	}
+	// Must be short or invalid.
+	accept := acceptRanges[x>>4]
+	if n > 1 {
+		if c := p[1]; c < accept.lo || accept.hi < c {
+			return true
+		} else if n > 2 && (p[2] < locb || hicb < p[2]) {
+			return true
+		}
+	}
+	return false
 }
 
 // FullRuneInString is like FullRune but its input is a string.
 func FullRuneInString(s string) bool {
-	_, _, short := decodeRuneInStringInternal(s)
-	return !short
+	n := len(s)
+	if n == 0 {
+		return false
+	}
+	x := first[s[0]]
+	if n >= int(x&7) {
+		return true // ASCII, invalid, or valid.
+	}
+	// Must be short or invalid.
+	accept := acceptRanges[x>>4]
+	if n > 1 {
+		if c := s[1]; c < accept.lo || accept.hi < c {
+			return true
+		} else if n > 2 && (s[2] < locb || hicb < s[2]) {
+			return true
+		}
+	}
+	return false
 }
 
 // DecodeRune unpacks the first UTF-8 encoding in p and returns the rune and
@@ -220,8 +151,43 @@ func FullRuneInString(s string) bool {
 // out of range, or is not the shortest possible UTF-8 encoding for the
 // value. No other validation is performed.
 func DecodeRune(p []byte) (r rune, size int) {
-	r, size, _ = decodeRuneInternal(p)
-	return
+	n := len(p)
+	if n < 1 {
+		return RuneError, 0
+	}
+	p0 := p[0]
+	x := first[p0]
+	if x >= as {
+		// The following code simulates an additional check for x == xx and
+		// handling the ASCII and invalid cases accordingly. This mask-and-or
+		// approach prevents an additional branch.
+		mask := rune(x) << 31 >> 31 // Create 0x0000 or 0xFFFF.
+		return rune(p[0])&^mask | RuneError&mask, 1
+	}
+	sz := x & 7
+	accept := acceptRanges[x>>4]
+	if n < int(sz) {
+		return RuneError, 1
+	}
+	b1 := p[1]
+	if b1 < accept.lo || accept.hi < b1 {
+		return RuneError, 1
+	}
+	if sz == 2 {
+		return rune(p0&mask2)<<6 | rune(b1&maskx), 2
+	}
+	b2 := p[2]
+	if b2 < locb || hicb < b2 {
+		return RuneError, 1
+	}
+	if sz == 3 {
+		return rune(p0&mask3)<<12 | rune(b1&maskx)<<6 | rune(b2&maskx), 3
+	}
+	b3 := p[3]
+	if b3 < locb || hicb < b3 {
+		return RuneError, 1
+	}
+	return rune(p0&mask4)<<18 | rune(b1&maskx)<<12 | rune(b2&maskx)<<6 | rune(b3&maskx), 4
 }
 
 // DecodeRuneInString is like DecodeRune but its input is a string. If s is
@@ -232,8 +198,43 @@ func DecodeRune(p []byte) (r rune, size int) {
 // out of range, or is not the shortest possible UTF-8 encoding for the
 // value. No other validation is performed.
 func DecodeRuneInString(s string) (r rune, size int) {
-	r, size, _ = decodeRuneInStringInternal(s)
-	return
+	n := len(s)
+	if n < 1 {
+		return RuneError, 0
+	}
+	s0 := s[0]
+	x := first[s0]
+	if x >= as {
+		// The following code simulates an additional check for x == xx and
+		// handling the ASCII and invalid cases accordingly. This mask-and-or
+		// approach prevents an additional branch.
+		mask := rune(x) << 31 >> 31 // Create 0x0000 or 0xFFFF.
+		return rune(s[0])&^mask | RuneError&mask, 1
+	}
+	sz := x & 7
+	accept := acceptRanges[x>>4]
+	if n < int(sz) {
+		return RuneError, 1
+	}
+	s1 := s[1]
+	if s1 < accept.lo || accept.hi < s1 {
+		return RuneError, 1
+	}
+	if sz == 2 {
+		return rune(s0&mask2)<<6 | rune(s1&maskx), 2
+	}
+	s2 := s[2]
+	if s2 < locb || hicb < s2 {
+		return RuneError, 1
+	}
+	if sz == 3 {
+		return rune(s0&mask3)<<12 | rune(s1&maskx)<<6 | rune(s2&maskx), 3
+	}
+	s3 := s[3]
+	if s3 < locb || hicb < s3 {
+		return RuneError, 1
+	}
+	return rune(s0&mask4)<<18 | rune(s1&maskx)<<12 | rune(s2&maskx)<<6 | rune(s3&maskx), 4
 }
 
 // DecodeLastRune unpacks the last UTF-8 encoding in p and returns the rune and
@@ -367,73 +368,141 @@ func EncodeRune(p []byte, r rune) int {
 // RuneCount returns the number of runes in p.  Erroneous and short
 // encodings are treated as single runes of width 1 byte.
 func RuneCount(p []byte) int {
-	i := 0
+	np := len(p)
 	var n int
-	for n = 0; i < len(p); n++ {
-		if p[i] < RuneSelf {
+	for i := 0; i < np; {
+		n++
+		c := p[i]
+		if c < RuneSelf {
+			// ASCII fast path
 			i++
-		} else {
-			_, size := DecodeRune(p[i:])
-			i += size
+			continue
+		}
+		x := first[c]
+		if x == xx {
+			i++ // invalid.
+			continue
+		}
+		size := int(x & 7)
+		if i+size > np {
+			i++ // Short or invalid.
+			continue
 		}
+		accept := acceptRanges[x>>4]
+		if c := p[i+1]; c < accept.lo || accept.hi < c {
+			size = 1
+		} else if size == 2 {
+		} else if c := p[i+2]; c < locb || hicb < c {
+			size = 1
+		} else if size == 3 {
+		} else if c := p[i+3]; c < locb || hicb < c {
+			size = 1
+		}
+		i += size
 	}
 	return n
 }
 
 // RuneCountInString is like RuneCount but its input is a string.
 func RuneCountInString(s string) (n int) {
-	for i := 0; i < len(s); {
-		n++
-		if s[i] < RuneSelf {
+	ns := len(s)
+	for i := 0; i < ns; n++ {
+		c := s[i]
+		if c < RuneSelf {
+			// ASCII fast path
 			i++
-		} else {
-			_, size := DecodeRuneInString(s[i:])
-			i += size
+			continue
+		}
+		x := first[c]
+		if x == xx {
+			i++ // invalid.
+			continue
+		}
+		size := int(x & 7)
+		if i+size > ns {
+			i++ // Short or invalid.
+			continue
 		}
+		accept := acceptRanges[x>>4]
+		if c := s[i+1]; c < accept.lo || accept.hi < c {
+			size = 1
+		} else if size == 2 {
+		} else if c := s[i+2]; c < locb || hicb < c {
+			size = 1
+		} else if size == 3 {
+		} else if c := s[i+3]; c < locb || hicb < c {
+			size = 1
+		}
+		i += size
 	}
 	return n
 }
 
-// RuneStart reports whether the byte could be the first byte of
-// an encoded rune.  Second and subsequent bytes always have the top
-// two bits set to 10.
+// RuneStart reports whether the byte could be the first byte of an encoded,
+// possibly invalid rune.  Second and subsequent bytes always have the top two
+// bits set to 10.
 func RuneStart(b byte) bool { return b&0xC0 != 0x80 }
 
 // Valid reports whether p consists entirely of valid UTF-8-encoded runes.
 func Valid(p []byte) bool {
-	i := 0
-	for i < len(p) {
-		if p[i] < RuneSelf {
+	n := len(p)
+	for i := 0; i < n; {
+		pi := p[i]
+		if pi < RuneSelf {
 			i++
-		} else {
-			_, size := DecodeRune(p[i:])
-			if size == 1 {
-				// All valid runes of size 1 (those
-				// below RuneSelf) were handled above.
-				// This must be a RuneError.
-				return false
-			}
-			i += size
+			continue
+		}
+		x := first[pi]
+		if x == xx {
+			return false // Illegal starter byte.
 		}
+		size := int(x & 7)
+		if i+size > n {
+			return false // Short or invalid.
+		}
+		accept := acceptRanges[x>>4]
+		if c := p[i+1]; c < accept.lo || accept.hi < c {
+			return false
+		} else if size == 2 {
+		} else if c := p[i+2]; c < locb || hicb < c {
+			return false
+		} else if size == 3 {
+		} else if c := p[i+3]; c < locb || hicb < c {
+			return false
+		}
+		i += size
 	}
 	return true
 }
 
 // ValidString reports whether s consists entirely of valid UTF-8-encoded runes.
 func ValidString(s string) bool {
-	for i := 0; i < len(s); {
-		if s[i] < RuneSelf {
+	n := len(s)
+	for i := 0; i < n; {
+		si := s[i]
+		if si < RuneSelf {
 			i++
-		} else {
-			_, size := DecodeRuneInString(s[i:])
-			if size == 1 {
-				// All valid runes of size 1 (those
-				// below RuneSelf) were handled above.
-				// This must be a RuneError.
-				return false
-			}
-			i += size
+			continue
+		}
+		x := first[si]
+		if x == xx {
+			return false // Illegal starter byte.
+		}
+		size := int(x & 7)
+		if i+size > n {
+			return false // Short or invalid.
+		}
+		accept := acceptRanges[x>>4]
+		if c := s[i+1]; c < accept.lo || accept.hi < c {
+			return false
+		} else if size == 2 {
+		} else if c := s[i+2]; c < locb || hicb < c {
+			return false
+		} else if size == 3 {
+		} else if c := s[i+3]; c < locb || hicb < c {
+			return false
 		}
+		i += size
 	}
 	return true
 }
diff --git a/src/unicode/utf8/utf8_test.go b/src/unicode/utf8/utf8_test.go
index 36d3959eee..60ce362f09 100644
--- a/src/unicode/utf8/utf8_test.go
+++ b/src/unicode/utf8/utf8_test.go
@@ -300,6 +300,8 @@ var runecounttests = []RuneCountTest{
 	{"☺☻☹", 3},
 	{"1,2,3,4", 7},
 	{"\xe2\x00", 2},
+	{"\xe2\x80", 2},
+	{"a\xe2\x80", 3},
 }
 
 func TestRuneCount(t *testing.T) {
@@ -352,6 +354,7 @@ var validTests = []ValidTest{
 	{"ЖЖ", true},
 	{"брэд-ЛГТМ", true},
 	{"☺☻☹", true},
+	{"aa\xe2", false},
 	{string([]byte{66, 250}), false},
 	{string([]byte{66, 250, 67}), false},
 	{"a\uFFFDb", true},