// getWidthData calls f for every entry for which it is defined.
//
// f may be called multiple times for the same rune. The last call to f is the
// correct value. f is not called for all runes. The default tag type is
// Neutral.
func getWidthData(f func(r rune, tag elem, alt rune)) {
// Set the default values for Unified Ideographs. In line with Annex 11,
// we encode full ranges instead of the defined runes in Unified_Ideograph.
for _, b := range []struct{ lo, hi rune }{
{0x4E00, 0x9FFF}, // the CJK Unified Ideographs block,
{0x3400, 0x4DBF}, // the CJK Unified Ideographs Externsion A block,
{0xF900, 0xFAFF}, // the CJK Compatibility Ideographs block,
{0x20000, 0x2FFFF}, // the Supplementary Ideographic Plane,
{0x30000, 0x3FFFF}, // the Tertiary Ideographic Plane,
} {
for r := b.lo; r <= b.hi; r++ {
f(r, tagWide, 0)
}
}
inverse := map[rune]rune{}
maps := map[string]bool{
"<wide>": true,
"<narrow>": true,
}
// We cannot reuse package norm's decomposition, as we need an unexpanded
// decomposition. We make use of the opportunity to verify that the
// decomposition type is as expected.
ucd.Parse(gen.OpenUCDFile("UnicodeData.txt"), func(p *ucd.Parser) {
r := p.Rune(0)
s := strings.SplitN(p.String(ucd.DecompMapping), " ", 2)
if !maps[s[0]] {
return
}
x, err := strconv.ParseUint(s[1], 16, 32)
if err != nil {
log.Fatalf("Error parsing rune %q", s[1])
}
if inverse[r] != 0 || inverse[rune(x)] != 0 {
log.Fatalf("Circular dependency in mapping between %U and %U", r, x)
}
inverse[r] = rune(x)
inverse[rune(x)] = r
})
// <rune range>;<type>
ucd.Parse(gen.OpenUCDFile("EastAsianWidth.txt"), func(p *ucd.Parser) {
tag, ok := typeMap[p.String(1)]
if !ok {
log.Fatalf("Unknown width type %q", p.String(1))
}
r := p.Rune(0)
alt, ok := inverse[r]
if tag == tagFullwidth || tag == tagHalfwidth && r != wonSign {
tag |= tagNeedsFold
if !ok {
log.Fatalf("Narrow or wide rune %U has no decomposition", r)
}
}
f(r, tag, alt)
})
}
// These tables are hand-extracted from:
// http://www.unicode.org/Public/8.0.0/ucd/extracted/DerivedBidiClass.txt
func visitDefaults(fn func(r rune, c Class)) {
// first write default values for ranges listed above.
visitRunes(fn, AL, []rune{
0x0600, 0x07BF, // Arabic
0x08A0, 0x08FF, // Arabic Extended-A
0xFB50, 0xFDCF, // Arabic Presentation Forms
0xFDF0, 0xFDFF,
0xFE70, 0xFEFF,
0x0001EE00, 0x0001EEFF, // Arabic Mathematical Alpha Symbols
})
visitRunes(fn, R, []rune{
0x0590, 0x05FF, // Hebrew
0x07C0, 0x089F, // Nko et al.
0xFB1D, 0xFB4F,
0x00010800, 0x00010FFF, // Cypriot Syllabary et. al.
0x0001E800, 0x0001EDFF,
0x0001EF00, 0x0001EFFF,
})
visitRunes(fn, ET, []rune{ // European Terminator
0x20A0, 0x20Cf, // Currency symbols
})
rangetable.Visit(unicode.Noncharacter_Code_Point, func(r rune) {
fn(r, BN) // Boundary Neutral
})
ucd.Parse(gen.OpenUCDFile("DerivedCoreProperties.txt"), func(p *ucd.Parser) {
if p.String(1) == "Default_Ignorable_Code_Point" {
fn(p.Rune(0), BN) // Boundary Neutral
}
})
}
func TestName(t *testing.T) {
testtext.SkipIfNotLong(t)
wants := make([]string, 1+unicode.MaxRune)
ucd.Parse(gen.OpenUCDFile("UnicodeData.txt"), func(p *ucd.Parser) {
r, s := p.Rune(0), p.String(ucd.Name)
if s == "" {
return
}
if s[0] == '<' {
const first = ", First>"
if i := strings.Index(s, first); i >= 0 {
s = s[:i] + ">"
}
}
wants[r] = s
})
nErrors := 0
for r, want := range wants {
got := Name(rune(r))
if got != want {
t.Errorf("r=%#08x: got %q, want %q", r, got, want)
nErrors++
if nErrors == 100 {
t.Fatal("too many errors")
}
}
}
}
func TestTables(t *testing.T) {
testtext.SkipIfNotLong(t)
ucd.Parse(gen.OpenUCDFile("BidiBrackets.txt"), func(p *ucd.Parser) {
r1 := p.Rune(0)
want := p.Rune(1)
e, _ := LookupRune(r1)
if got := e.reverseBracket(r1); got != want {
t.Errorf("Reverse(%U) = %U; want %U", r1, got, want)
}
})
done := map[rune]bool{}
test := func(name string, r rune, want string) {
str := string(r)
e, _ := LookupString(str)
if got := labels[e.Class()]; got != want {
t.Errorf("%s:%U: got %s; want %s", name, r, got, want)
}
if e2, sz := LookupRune(r); e != e2 || sz != len(str) {
t.Errorf("LookupRune(%U) = %v, %d; want %v, %d", r, e2, e, sz, len(str))
}
if e2, sz := Lookup([]byte(str)); e != e2 || sz != len(str) {
t.Errorf("Lookup(%U) = %v, %d; want %v, %d", r, e2, e, sz, len(str))
}
done[r] = true
}
// Insert the derived BiDi properties.
ucd.Parse(gen.OpenUCDFile("extracted/DerivedBidiClass.txt"), func(p *ucd.Parser) {
r := p.Rune(0)
test("derived", r, p.String(1))
})
visitDefaults(func(r rune, c Class) {
if !done[r] {
test("default", r, labels[c])
}
})
}
func TestTables(t *testing.T) {
if !*long {
return
}
gen.Init()
trie := newBidiTrie(0)
ucd.Parse(gen.OpenUCDFile("BidiBrackets.txt"), func(p *ucd.Parser) {
r1 := p.Rune(0)
want := p.Rune(1)
e, _ := trie.lookupString(string(r1))
if got := entry(e).reverseBracket(r1); got != want {
t.Errorf("Reverse(%U) = %U; want %U", r1, got, want)
}
})
done := map[rune]bool{}
test := func(name string, r rune, want string) {
e, _ := trie.lookupString(string(r))
if got := labels[entry(e).class(r)]; got != want {
t.Errorf("%s:%U: got %s; want %s", name, r, got, want)
}
done[r] = true
}
// Insert the derived BiDi properties.
ucd.Parse(gen.OpenUCDFile("extracted/DerivedBidiClass.txt"), func(p *ucd.Parser) {
r := p.Rune(0)
test("derived", r, p.String(1))
})
visitDefaults(func(r rune, c class) {
if !done[r] {
test("default", r, labels[c])
}
})
}
func main() {
gen.Init()
// Load data
runes := []rune{}
ucd.Parse(gen.OpenUCDFile("DerivedCoreProperties.txt"), func(p *ucd.Parser) {
if p.String(1) == "Default_Ignorable_Code_Point" {
runes = append(runes, p.Rune(0))
}
})
ucd.Parse(gen.OpenUCDFile("HangulSyllableType.txt"), func(p *ucd.Parser) {
if p.String(1) == "LVT" {
runes = append(runes, p.Rune(0))
}
})
disallowedRunes = rangetable.New(runes...)
assigned = rangetable.Assigned(unicode.Version)
writeTables()
gen.Repackage("gen_trieval.go", "trieval.go", "precis")
}
func main() {
gen.Init()
versions := getVersions()
w := &bytes.Buffer{}
fmt.Fprintf(w, "//go:generate go run gen.go --versions=%s\n\n", strings.Join(versions, ","))
fmt.Fprintf(w, "import \"unicode\"\n\n")
vstr := func(s string) string { return strings.Replace(s, ".", "_", -1) }
fmt.Fprintf(w, "var assigned = map[string]*unicode.RangeTable{\n")
for _, v := range versions {
fmt.Fprintf(w, "\t%q: assigned%s,\n", v, vstr(v))
}
fmt.Fprintf(w, "}\n\n")
var size int
for _, v := range versions {
assigned := []rune{}
r := gen.Open("http://www.unicode.org/Public/", "", v+"/ucd/UnicodeData.txt")
ucd.Parse(r, func(p *ucd.Parser) {
assigned = append(assigned, p.Rune(0))
})
rt := rangetable.New(assigned...)
sz := int(reflect.TypeOf(unicode.RangeTable{}).Size())
sz += int(reflect.TypeOf(unicode.Range16{}).Size()) * len(rt.R16)
sz += int(reflect.TypeOf(unicode.Range32{}).Size()) * len(rt.R32)
fmt.Fprintf(w, "// size %d bytes (%d KiB)\n", sz, sz/1024)
fmt.Fprintf(w, "var assigned%s = ", vstr(v))
print(w, rt)
size += sz
}
fmt.Fprintf(w, "// Total size %d bytes (%d KiB)\n", size, size/1024)
gen.WriteGoFile("tables.go", "rangetable", w.Bytes())
}
func parse() (names []string, counts map[string]int) {
names = make([]string, 1+unicode.MaxRune)
counts = map[string]int{}
ucd.Parse(gen.OpenUCDFile("UnicodeData.txt"), func(p *ucd.Parser) {
r, s := p.Rune(0), p.String(ucd.Name)
if s == "" {
return
}
if s[0] == '<' {
const first = ", First>"
if i := strings.Index(s, first); i >= 0 {
s = s[:i] + ">"
}
}
names[r] = s
counts[s]++
})
return names, counts
}
// TestBidiCharacters performs the tests in BidiCharacterTest.txt.
// See http://www.unicode.org/Public/UCD/latest/ucd/BidiCharacterTest.txt
func TestBidiCharacters(t *testing.T) {
testtext.SkipIfNotLong(t)
ucd.Parse(gen.OpenUCDFile("BidiCharacterTest.txt"), func(p *ucd.Parser) {
var (
types []class
pairTypes []bracketType
pairValues []rune
parLevel level
wantLevel = level(p.Int(2))
wantLevels = p.Strings(3)
wantVisualOrder = p.Strings(4)
)
switch l := p.Int(1); l {
case 0, 1:
parLevel = level(l)
case 2:
parLevel = implicitLevel
default:
// Spec says to ignore unknown parts.
}
trie := newBidiTrie(0)
runes := p.Runes(0)
for _, r := range runes {
// Assign the bracket type.
if d := norm.NFKD.PropertiesString(string(r)).Decomposition(); d != nil {
r = []rune(string(d))[0]
}
e, _ := trie.lookupString(string(r))
entry := entry(e)
// Assign the class for this rune.
types = append(types, entry.class(r))
switch {
case !entry.isBracket():
pairTypes = append(pairTypes, bpNone)
pairValues = append(pairValues, 0)
case entry.isOpen():
pairTypes = append(pairTypes, bpOpen)
pairValues = append(pairValues, r)
default:
pairTypes = append(pairTypes, bpClose)
pairValues = append(pairValues, entry.reverseBracket(r))
}
}
par := newParagraph(types, pairTypes, pairValues, parLevel)
// Test results:
if got := par.embeddingLevel; got != wantLevel {
t.Errorf("%v:level: got %d; want %d", string(runes), got, wantLevel)
}
if *testLevels {
gotLevels := getLevelStrings(types, par.resultLevels)
if got, want := fmt.Sprint(gotLevels), fmt.Sprint(wantLevels); got != want {
t.Errorf("%04X %q:%d: got %v; want %v\nval: %x\npair: %v", runes, string(runes), parLevel, got, want, pairValues, pairTypes)
}
}
order := par.getReordering([]int{len(types)})
order = filterOrder(types, order)
if got, want := fmt.Sprint(order), fmt.Sprint(wantVisualOrder); got != want {
t.Errorf("%04X %q:%d: got %v; want %v\ngot order: %s", runes, string(runes), parLevel, got, want, reorder(runes, order))
}
})
}
func TestTables(t *testing.T) {
testtext.SkipIfNotLong(t)
lookup := func(r rune) info {
v, _ := trie.lookupString(string(r))
return info(v)
}
ucd.Parse(gen.OpenUnicodeFile("idna", "", "IdnaMappingTable.txt"), func(p *ucd.Parser) {
r := p.Rune(0)
x := lookup(r)
if got, want := x.category(), catFromEntry(p); got != want {
t.Errorf("%U:category: got %x; want %x", r, got, want)
}
mapped := false
switch p.String(1) {
case "mapped", "disallowed_STD3_mapped", "deviation":
mapped = true
}
if x.isMapped() != mapped {
t.Errorf("%U:isMapped: got %v; want %v", r, x.isMapped(), mapped)
}
if !mapped {
return
}
want := string(p.Runes(2))
got := string(x.appendMapping(nil, string(r)))
if got != want {
t.Errorf("%U:mapping: got %+q; want %+q", r, got, want)
}
if x.isMapped() {
return
}
wantMark := unicode.In(r, unicode.Mark)
gotMark := x.isModifier()
if gotMark != wantMark {
t.Errorf("IsMark(%U) = %v; want %v", r, gotMark, wantMark)
}
})
ucd.Parse(gen.OpenUCDFile("UnicodeData.txt"), func(p *ucd.Parser) {
r := p.Rune(0)
x := lookup(r)
got := x.isViramaModifier()
const cccVirama = 9
want := p.Int(ucd.CanonicalCombiningClass) == cccVirama
if got != want {
t.Errorf("IsVirama(%U) = %v; want %v", r, got, want)
}
})
ucd.Parse(gen.OpenUCDFile("extracted/DerivedJoiningType.txt"), func(p *ucd.Parser) {
r := p.Rune(0)
x := lookup(r)
if x.isMapped() {
return
}
got := x.joinType()
want := joinType[p.String(1)]
if got != want {
t.Errorf("JoinType(%U) = %x; want %x", r, got, want)
}
})
}
func parse(file string, f func(p *ucd.Parser)) {
ucd.Parse(gen.OpenUCDFile(file), f)
}
func genTables() {
t := triegen.NewTrie("idna")
ucd.Parse(gen.OpenUCDFile("UnicodeData.txt"), func(p *ucd.Parser) {
r := p.Rune(0)
const cccVirama = 9
if p.Int(ucd.CanonicalCombiningClass) == cccVirama {
runes[p.Rune(0)] = viramaModifier
}
switch {
case unicode.In(r, unicode.Mark):
runes[r] |= modifier
}
})
ucd.Parse(gen.OpenUCDFile("extracted/DerivedJoiningType.txt"), func(p *ucd.Parser) {
switch v := p.String(1); v {
case "L", "D", "T", "R":
runes[p.Rune(0)] |= joinType[v] << joinShift
}
})
ucd.Parse(gen.OpenUnicodeFile("idna", "", "IdnaMappingTable.txt"), func(p *ucd.Parser) {
r := p.Rune(0)
// The mappings table explicitly defines surrogates as invalid.
if !utf8.ValidRune(r) {
return
}
cat := catFromEntry(p)
isMapped := cat == mapped || cat == disallowedSTD3Mapped || cat == deviation
if !isMapped {
// Only include additional category information for non-mapped
// runes. The additional information is only used after mapping and
// the bits would clash with mapping information.
// TODO: it would be possible to inline this data and avoid
// additional lookups. This is quite tedious, though, so let's first
// see if we need this.
cat |= category(runes[r])
}
s := string(p.Runes(2))
if s != "" && !isMapped {
log.Fatalf("%U: Mapping with non-mapping category %d", r, cat)
}
t.Insert(r, uint64(makeEntry(r, s))+uint64(cat))
})
w := gen.NewCodeWriter()
defer w.WriteGoFile("tables.go", "idna")
gen.WriteUnicodeVersion(w)
w.WriteVar("mappings", string(mappings))
w.WriteVar("xorData", string(xorData))
sz, err := t.Gen(w, triegen.Compact(&normCompacter{}))
if err != nil {
log.Fatal(err)
}
w.Size += sz
}
func main() {
gen.Init()
// Load data
runes := []rune{}
// PrecisIgnorableProperties: https://tools.ietf.org/html/rfc7564#section-9.13
ucd.Parse(gen.OpenUCDFile("DerivedCoreProperties.txt"), func(p *ucd.Parser) {
if p.String(1) == "Default_Ignorable_Code_Point" {
runes = append(runes, p.Rune(0))
}
})
ucd.Parse(gen.OpenUCDFile("PropList.txt"), func(p *ucd.Parser) {
switch p.String(1) {
case "Noncharacter_Code_Point":
runes = append(runes, p.Rune(0))
}
})
// OldHangulJamo: https://tools.ietf.org/html/rfc5892#section-2.9
ucd.Parse(gen.OpenUCDFile("HangulSyllableType.txt"), func(p *ucd.Parser) {
switch p.String(1) {
case "L", "V", "T":
runes = append(runes, p.Rune(0))
}
})
disallowedRunes = rangetable.New(runes...)
assigned = rangetable.Assigned(unicode.Version)
// Load category data.
runeCategory['l'] = latinSmallL
ucd.Parse(gen.OpenUCDFile("UnicodeData.txt"), func(p *ucd.Parser) {
const cccVirama = 9
if p.Int(ucd.CanonicalCombiningClass) == cccVirama {
setCategory(p.Rune(0), viramaModifier)
}
})
ucd.Parse(gen.OpenUCDFile("Scripts.txt"), func(p *ucd.Parser) {
switch p.String(1) {
case "Greek":
setCategory(p.Rune(0), greek)
case "Hebrew":
setCategory(p.Rune(0), hebrew)
case "Hiragana", "Katakana", "Han":
setCategory(p.Rune(0), japanese)
}
})
// Set the rule categories associated with exceptions. This overrides any
// previously set categories. The original categories are manually
// reintroduced in the categoryTransitions table.
for r, e := range exceptions {
if e.cat != 0 {
runeCategory[r] = e.cat
}
}
cat := map[string]category{
"L": joiningL,
"D": joiningD,
"T": joiningT,
"R": joiningR,
}
ucd.Parse(gen.OpenUCDFile("extracted/DerivedJoiningType.txt"), func(p *ucd.Parser) {
switch v := p.String(1); v {
case "L", "D", "T", "R":
setCategory(p.Rune(0), cat[v])
}
})
writeTables()
gen.Repackage("gen_trieval.go", "trieval.go", "precis")
}
func genTables() {
if numClass > 0x0F {
log.Fatalf("Too many Class constants (%#x > 0x0F).", numClass)
}
w := gen.NewCodeWriter()
defer w.WriteGoFile(*outputFile, "bidi")
gen.WriteUnicodeVersion(w)
t := triegen.NewTrie("bidi")
// Build data about bracket mapping. These bits need to be or-ed with
// any other bits.
orMask := map[rune]uint64{}
xorMap := map[rune]int{}
xorMasks := []rune{0} // First value is no-op.
ucd.Parse(gen.OpenUCDFile("BidiBrackets.txt"), func(p *ucd.Parser) {
r1 := p.Rune(0)
r2 := p.Rune(1)
xor := r1 ^ r2
if _, ok := xorMap[xor]; !ok {
xorMap[xor] = len(xorMasks)
xorMasks = append(xorMasks, xor)
}
entry := uint64(xorMap[xor]) << xorMaskShift
switch p.String(2) {
case "o":
entry |= openMask
case "c", "n":
default:
log.Fatalf("Unknown bracket class %q.", p.String(2))
}
orMask[r1] = entry
})
w.WriteComment(`
xorMasks contains masks to be xor-ed with brackets to get the reverse
version.`)
w.WriteVar("xorMasks", xorMasks)
done := map[rune]bool{}
insert := func(r rune, c class) {
if !done[r] {
t.Insert(r, orMask[r]|uint64(c))
done[r] = true
}
}
// Insert the derived BiDi properties.
ucd.Parse(gen.OpenUCDFile("extracted/DerivedBidiClass.txt"), func(p *ucd.Parser) {
r := p.Rune(0)
class, ok := bidiClass[p.String(1)]
if !ok {
log.Fatalf("%U: Unknown BiDi class %q", r, p.String(1))
}
insert(r, class)
})
visitDefaults(insert)
// TODO: use sparse blocks. This would reduce table size considerably
// from the looks of it.
sz, err := t.Gen(w)
if err != nil {
log.Fatal(err)
}
w.Size += sz
}