exp/norm: merged charinfo and decomposition tables. As a result only

one trie lookup per rune is needed. See forminfo.go for a description
of the new format.  Also included leading and trailing canonical
combining class in decomposition information.  This will often avoid
additional trie lookups.

R=r, r
CC=golang-dev
https://golang.org/cl/5616071

src/pkg/exp/norm/composition.go

@@ -102,7 +102,7 @@ func (rb *reorderBuffer) insert(src input, i int, info runeInfo) bool {
 		}
 	}
 	if info.hasDecomposition() {
-		dcomp := rb.f.decompose(src, i)
+		dcomp := info.decomposition()
 		rb.tmpBytes = inputBytes(dcomp)
 		for i := 0; i < len(dcomp); {
 			info = rb.f.info(&rb.tmpBytes, i)

src/pkg/exp/norm/forminfo.go

@@ -6,24 +6,49 @@ package norm
 
 // This file contains Form-specific logic and wrappers for data in tables.go.
 
+// Rune info is stored in a separate trie per composing form. A composing form
+// and its corresponding decomposing form share the same trie.  Each trie maps
+// a rune to a uint16. The values take two forms.  For v >= 0x8000:
+//   bits
+//   0..8:   ccc
+//   9..12:  qcInfo (see below). isYesD is always true (no decompostion).
+//   16:     1
+// For v < 0x8000, the respective rune has a decomposition and v is an index
+// into a byte array of UTF-8 decomposition sequences and additional info and
+// has the form:
+//    <header> <decomp_byte>* [<tccc> [<lccc>]]
+// The header contains the number of bytes in the decomposition (excluding this
+// length byte). The two most significant bits of this lenght byte correspond
+// to bit 2 and 3 of qcIfo (see below).  The byte sequence itself starts at v+1.
+// The byte sequence is followed by a trailing and leading CCC if the values
+// for these are not zero.  The value of v determines which ccc are appended
+// to the sequences.  For v < firstCCC, there are none, for v >= firstCCC,
+// the seqence is followed by a trailing ccc, and for v >= firstLeadingCC
+// there is an additional leading ccc.
+
+const (
+	qcInfoMask      = 0xF  // to clear all but the relevant bits in a qcInfo
+	headerLenMask   = 0x3F // extract the lenght value from the header byte
+	headerFlagsMask = 0xC0 // extract the qcInfo bits from the header byte
+)
+
+// runeInfo is a representation for the data stored in charinfoTrie.
 type runeInfo struct {
 	pos   uint8  // start position in reorderBuffer; used in composition.go
 	size  uint8  // length of UTF-8 encoding of this rune
-	ccc   uint8  // canonical combining class
+	ccc   uint8  // leading canonical combining class (ccc if not decomposition)
+	tccc  uint8  // trailing canonical combining class (ccc if not decomposition)
 	flags qcInfo // quick check flags
+	index uint16
 }
 
 // functions dispatchable per form
 type lookupFunc func(b input, i int) runeInfo
-type decompFunc func(b input, i int) []byte
 
 // formInfo holds Form-specific functions and tables.
 type formInfo struct {
 	form                     Form
-
 	composing, compatibility bool // form type
-
-	decompose decompFunc
 	info                     lookupFunc
 }
 
@@ -38,10 +63,8 @@ func init() {
 		f.form = Form(i)
 		if Form(i) == NFKD || Form(i) == NFKC {
 			f.compatibility = true
-			f.decompose = decomposeNFKC
 			f.info = lookupInfoNFKC
 		} else {
-			f.decompose = decomposeNFC
 			f.info = lookupInfoNFC
 		}
 		if Form(i) == NFC || Form(i) == NFKC {
@@ -76,8 +99,6 @@ func (i runeInfo) boundaryAfter() bool {
 // 
 // When all 4 bits are zero, the character is inert, meaning it is never
 // influenced by normalization.
-//
-// We pack the bits for both NFC/D and NFKC/D in one byte.
 type qcInfo uint8
 
 func (i runeInfo) isYesC() bool { return i.flags&0x4 == 0 }
@@ -91,22 +112,12 @@ func (r runeInfo) isInert() bool {
 	return r.flags&0xf == 0 && r.ccc == 0
 }
 
-// Wrappers for tables.go
-
-// The 16-bit value of the decomposition tries is an index into a byte
-// array of UTF-8 decomposition sequences. The first byte is the number
-// of bytes in the decomposition (excluding this length byte). The actual
-// sequence starts at the offset+1.
-func decomposeNFC(s input, i int) []byte {
-	p := s.decomposeNFC(i)
-	n := decomps[p]
-	p++
-	return decomps[p : p+uint16(n)]
-}
-
-func decomposeNFKC(s input, i int) []byte {
-	p := s.decomposeNFKC(i)
-	n := decomps[p]
+func (r runeInfo) decomposition() []byte {
+	if r.index == 0 {
+		return nil
+	}
+	p := r.index
+	n := decomps[p] & 0x3F
 	p++
 	return decomps[p : p+uint16(n)]
 }
@@ -124,16 +135,40 @@ func combine(a, b rune) rune {
 	return recompMap[key]
 }
 
-// The 16-bit character info has the following bit layout:
-//    0..7   CCC value.
-//    8..11  qcInfo for NFC/NFD
-//   12..15  qcInfo for NFKC/NFKD
 func lookupInfoNFC(b input, i int) runeInfo {
-	v, sz := b.charinfo(i)
-	return runeInfo{size: uint8(sz), ccc: uint8(v), flags: qcInfo(v >> 8)}
+	v, sz := b.charinfoNFC(i)
+	return compInfo(v, sz)
 }
 
 func lookupInfoNFKC(b input, i int) runeInfo {
-	v, sz := b.charinfo(i)
-	return runeInfo{size: uint8(sz), ccc: uint8(v), flags: qcInfo(v >> 12)}
+	v, sz := b.charinfoNFKC(i)
+	return compInfo(v, sz)
+}
+
+// compInfo converts the information contained in v and sz
+// to a runeInfo.  See the comment at the top of the file
+// for more information on the format.
+func compInfo(v uint16, sz int) runeInfo {
+	if v == 0 {
+		return runeInfo{size: uint8(sz)}
+	} else if v >= 0x8000 {
+		return runeInfo{
+			size:  uint8(sz),
+			ccc:   uint8(v),
+			tccc:  uint8(v),
+			flags: qcInfo(v>>8) & qcInfoMask,
+		}
+	}
+	// has decomposition
+	h := decomps[v]
+	f := (qcInfo(h&headerFlagsMask) >> 4) | 0x1
+	ri := runeInfo{size: uint8(sz), flags: f, index: v}
+	if v >= firstCCC {
+		v += uint16(h&headerLenMask) + 1
+		ri.tccc = decomps[v]
+		if v >= firstLeadingCCC {
+			ri.ccc = decomps[v+1]
+		}
+	}
+	return ri
 }

src/pkg/exp/norm/input.go

@@ -11,9 +11,8 @@ type input interface {
 	skipNonStarter(p int) int
 	appendSlice(buf []byte, s, e int) []byte
 	copySlice(buf []byte, s, e int)
-	charinfo(p int) (uint16, int)
-	decomposeNFC(p int) uint16
-	decomposeNFKC(p int) uint16
+	charinfoNFC(p int) (uint16, int)
+	charinfoNFKC(p int) (uint16, int)
 	hangul(p int) rune
 }
 
@@ -42,16 +41,12 @@ func (s inputString) copySlice(buf []byte, b, e int) {
 	copy(buf, s[b:e])
 }
 
-func (s inputString) charinfo(p int) (uint16, int) {
-	return charInfoTrie.lookupString(string(s[p:]))
+func (s inputString) charinfoNFC(p int) (uint16, int) {
+	return nfcTrie.lookupString(string(s[p:]))
 }
 
-func (s inputString) decomposeNFC(p int) uint16 {
-	return nfcDecompTrie.lookupStringUnsafe(string(s[p:]))
-}
-
-func (s inputString) decomposeNFKC(p int) uint16 {
-	return nfkcDecompTrie.lookupStringUnsafe(string(s[p:]))
+func (s inputString) charinfoNFKC(p int) (uint16, int) {
+	return nfkcTrie.lookupString(string(s[p:]))
 }
 
 func (s inputString) hangul(p int) rune {
@@ -84,16 +79,12 @@ func (s inputBytes) copySlice(buf []byte, b, e int) {
 	copy(buf, s[b:e])
 }
 
-func (s inputBytes) charinfo(p int) (uint16, int) {
-	return charInfoTrie.lookup(s[p:])
+func (s inputBytes) charinfoNFC(p int) (uint16, int) {
+	return nfcTrie.lookup(s[p:])
 }
 
-func (s inputBytes) decomposeNFC(p int) uint16 {
-	return nfcDecompTrie.lookupUnsafe(s[p:])
-}
-
-func (s inputBytes) decomposeNFKC(p int) uint16 {
-	return nfkcDecompTrie.lookupUnsafe(s[p:])
+func (s inputBytes) charinfoNFKC(p int) (uint16, int) {
+	return nfkcTrie.lookup(s[p:])
 }
 
 func (s inputBytes) hangul(p int) rune {

src/pkg/exp/norm/maketables.go

@@ -4,6 +4,7 @@
 
 // Normalization table generator.
 // Data read from the web.
+// See forminfo.go for a description of the trie values associated with each rune.
 
 package main
 
@@ -17,6 +18,7 @@ import (
 	"net/http"
 	"os"
 	"regexp"
+	"sort"
 	"strconv"
 	"strings"
 )
@@ -187,18 +189,14 @@ func (f FormInfo) String() string {
 	fmt.Fprintf(buf, "    cmbBackward: %v\n", f.combinesBackward)
 	fmt.Fprintf(buf, "    isOneWay: %v\n", f.isOneWay)
 	fmt.Fprintf(buf, "    inDecomp: %v\n", f.inDecomp)
-	fmt.Fprintf(buf, "    decomposition: %v\n", f.decomp)
-	fmt.Fprintf(buf, "    expandedDecomp: %v\n", f.expandedDecomp)
+	fmt.Fprintf(buf, "    decomposition: %X\n", f.decomp)
+	fmt.Fprintf(buf, "    expandedDecomp: %X\n", f.expandedDecomp)
 
 	return buf.String()
 }
 
 type Decomposition []rune
 
-func (d Decomposition) String() string {
-	return fmt.Sprintf("%.4X", d)
-}
-
 func openReader(file string) (input io.ReadCloser) {
 	if *localFiles {
 		f, err := os.Open(file)
@@ -571,80 +569,121 @@ func makeEntry(f *FormInfo) uint16 {
 	return e
 }
 
-// Bits
-// 0..8:   CCC
-// 9..12:  NF(C|D) qc bits.
-// 13..16: NFK(C|D) qc bits.
-func makeCharInfo(c Char) uint16 {
-	e := makeEntry(&c.forms[FCompatibility])
-	e = e<<4 | makeEntry(&c.forms[FCanonical])
-	e = e<<8 | uint16(c.ccc)
-	return e
+// decompSet keeps track of unique decompositions, grouped by whether
+// the decomposition is followed by a trailing and/or leading CCC.
+type decompSet [4]map[string]bool
+
+func makeDecompSet() decompSet {
+	m := decompSet{}
+	for i, _ := range m {
+		m[i] = make(map[string]bool)
+	}
+	return m
+}
+func (m *decompSet) insert(key int, s string) {
+	m[key][s] = true
 }
 
 func printCharInfoTables() int {
-	// Quick Check + CCC trie.
-	t := newNode()
-	for i, char := range chars {
-		v := makeCharInfo(char)
-		if v != 0 {
-			t.insert(rune(i), v)
+	mkstr := func(r rune, f *FormInfo) (int, string) {
+		d := f.expandedDecomp
+		s := string([]rune(d))
+		if max := 1 << 6; len(s) >= max {
+			const msg = "%U: too many bytes in decomposition: %d >= %d"
+			logger.Fatalf(msg, r, len(s), max)
+		}
+		head := uint8(len(s))
+		if f.quickCheck[MComposed] != QCYes {
+			head |= 0x40
+		}
+		if f.combinesForward {
+			head |= 0x80
+		}
+		s = string([]byte{head}) + s
+
+		lccc := ccc(d[0])
+		tccc := ccc(d[len(d)-1])
+		if tccc < lccc && lccc != 0 {
+			const msg = "%U: lccc (%d) must be <= tcc (%d)"
+			logger.Fatalf(msg, r, lccc, tccc)
+		}
+		index := 0
+		if tccc > 0 || lccc > 0 {
+			s += string([]byte{tccc})
+			index = 1
+			if lccc > 0 {
+				s += string([]byte{lccc})
+				index |= 2
 			}
 		}
-	return t.printTables("charInfo")
-}
+		return index, s
+	}
 
-func printDecompositionTables() int {
-	decompositions := bytes.NewBuffer(make([]byte, 0, 10000))
-	size := 0
-
-	// Map decompositions
-	positionMap := make(map[string]uint16)
+	decompSet := makeDecompSet()
 
 	// Store the uniqued decompositions in a byte buffer,
 	// preceded by their byte length.
 	for _, c := range chars {
-		for f := 0; f < 2; f++ {
-			d := c.forms[f].expandedDecomp
-			s := string([]rune(d))
-			if _, ok := positionMap[s]; !ok {
+		for _, f := range c.forms {
+			if len(f.expandedDecomp) == 0 {
+				continue
+			}
+			if f.combinesBackward {
+				logger.Fatalf("%U: combinesBackward and decompose", c.codePoint)
+			}
+			index, s := mkstr(c.codePoint, &f)
+			decompSet.insert(index, s)
+		}
+	}
+
+	decompositions := bytes.NewBuffer(make([]byte, 0, 10000))
+	size := 0
+	positionMap := make(map[string]uint16)
+	decompositions.WriteString("\000")
+	cname := []string{"firstCCC", "firstLeadingCCC", "", "lastDecomp"}
+	fmt.Println("const (")
+	for i, m := range decompSet {
+		sa := []string{}
+		for s, _ := range m {
+			sa = append(sa, s)
+		}
+		sort.Strings(sa)
+		for _, s := range sa {
 			p := decompositions.Len()
-				decompositions.WriteByte(uint8(len(s)))
 			decompositions.WriteString(s)
 			positionMap[s] = uint16(p)
 		}
+		if cname[i] != "" {
+			fmt.Printf("%s = 0x%X\n", cname[i], decompositions.Len())
 		}
 	}
+	fmt.Println("maxDecomp = 0x8000")
+	fmt.Println(")")
 	b := decompositions.Bytes()
 	printBytes(b, "decomps")
 	size += len(b)
 
-	nfcT := newNode()
-	nfkcT := newNode()
-	for i, c := range chars {
-		d := c.forms[FCanonical].expandedDecomp
+	varnames := []string{"nfc", "nfkc"}
+	for i := 0; i < FNumberOfFormTypes; i++ {
+		trie := newNode()
+		for r, c := range chars {
+			f := c.forms[i]
+			d := f.expandedDecomp
 			if len(d) != 0 {
-			nfcT.insert(rune(i), positionMap[string([]rune(d))])
-			if ccc(c.codePoint) != ccc(d[0]) {
-				// We assume the lead ccc of a decomposition is !=0 in this case.
+				_, key := mkstr(c.codePoint, &f)
+				trie.insert(rune(r), positionMap[key])
+				if c.ccc != ccc(d[0]) {
+					// We assume the lead ccc of a decomposition !=0 in this case.
 					if ccc(d[0]) == 0 {
-					logger.Fatal("Expected differing CCC to be non-zero.")
+						logger.Fatal("Expected leading CCC to be non-zero; ccc is %d", c.ccc)
 					}
 				}
-		}
-		d = c.forms[FCompatibility].expandedDecomp
-		if len(d) != 0 {
-			nfkcT.insert(rune(i), positionMap[string([]rune(d))])
-			if ccc(c.codePoint) != ccc(d[0]) {
-				// We assume the lead ccc of a decomposition is !=0 in this case.
-				if ccc(d[0]) == 0 {
-					logger.Fatal("Expected differing CCC to be non-zero.")
+			} else if v := makeEntry(&f)<<8 | uint16(c.ccc); v != 0 {
+				trie.insert(c.codePoint, 0x8000|v)
 			}
 		}
+		size += trie.printTables(varnames[i])
 	}
-	}
-	size += nfcT.printTables("nfcDecomp")
-	size += nfkcT.printTables("nfkcDecomp")
 	return size
 }
 
@@ -687,15 +726,15 @@ func makeTables() {
 	}
 	list := strings.Split(*tablelist, ",")
 	if *tablelist == "all" {
-		list = []string{"decomp", "recomp", "info"}
+		list = []string{"recomp", "info"}
 	}
 	fmt.Printf(fileHeader, *tablelist, *url)
 
 	fmt.Println("// Version is the Unicode edition from which the tables are derived.")
 	fmt.Printf("const Version = %q\n\n", version())
 
-	if contains(list, "decomp") {
-		size += printDecompositionTables()
+	if contains(list, "info") {
+		size += printCharInfoTables()
 	}
 
 	if contains(list, "recomp") {
@@ -730,9 +769,6 @@ func makeTables() {
 		fmt.Printf("}\n\n")
 	}
 
-	if contains(list, "info") {
-		size += printCharInfoTables()
-	}
 	fmt.Printf("// Total size of tables: %dKB (%d bytes)\n", (size+512)/1024, size)
 }
 
@@ -761,6 +797,11 @@ func verifyComputed() {
 				log.Fatalf("%U: NF*C must be maybe if combinesBackward", i)
 			}
 		}
+		nfc := c.forms[FCanonical]
+		nfkc := c.forms[FCompatibility]
+		if nfc.combinesBackward != nfkc.combinesBackward {
+			logger.Fatalf("%U: Cannot combine combinesBackward\n", c.codePoint)
+		}
 	}
 }
 

src/pkg/exp/norm/normalize.go

@@ -448,7 +448,7 @@ func decomposeToLastBoundary(rb *reorderBuffer, buf []byte) []byte {
 		}
 		// Check that decomposition doesn't result in overflow.
 		if info.hasDecomposition() {
-			dcomp := rb.f.decompose(inputBytes(buf), p-int(info.size))
+			dcomp := info.decomposition()
 			for i := 0; i < len(dcomp); {
 				inf := rb.f.info(inputBytes(dcomp), i)
 				i += int(inf.size)