in language/gen.go [144:308]
func (b *builder) writeMatchData() {
lm := b.supp.LanguageMatching.LanguageMatches
cldr.MakeSlice(&lm).SelectAnyOf("type", "written_new")
regionHierarchy := map[string][]string{}
for _, g := range b.supp.TerritoryContainment.Group {
regions := strings.Split(g.Contains, " ")
regionHierarchy[g.Type] = append(regionHierarchy[g.Type], regions...)
}
// Regions start at 1, so the slice must be one larger than the number of
// regions.
regionToGroups := make([]uint8, language.NumRegions+1)
idToIndex := map[string]uint8{}
for i, mv := range lm[0].MatchVariable {
if i > 6 {
log.Fatalf("Too many groups: %d", i)
}
idToIndex[mv.Id] = uint8(i + 1)
// TODO: also handle '-'
for _, r := range strings.Split(mv.Value, "+") {
todo := []string{r}
for k := 0; k < len(todo); k++ {
r := todo[k]
regionToGroups[b.regionIndex(r)] |= 1 << uint8(i)
todo = append(todo, regionHierarchy[r]...)
}
}
}
b.w.WriteVar("regionToGroups", regionToGroups)
// maps language id to in- and out-of-group region.
paradigmLocales := [][3]uint16{}
locales := strings.Split(lm[0].ParadigmLocales[0].Locales, " ")
for i := 0; i < len(locales); i += 2 {
x := [3]uint16{}
for j := 0; j < 2; j++ {
pc := strings.SplitN(locales[i+j], "-", 2)
x[0] = b.langIndex(pc[0])
if len(pc) == 2 {
x[1+j] = uint16(b.regionIndex(pc[1]))
}
}
paradigmLocales = append(paradigmLocales, x)
}
b.w.WriteVar("paradigmLocales", paradigmLocales)
b.w.WriteType(mutualIntelligibility{})
b.w.WriteType(scriptIntelligibility{})
b.w.WriteType(regionIntelligibility{})
matchLang := []mutualIntelligibility{}
matchScript := []scriptIntelligibility{}
matchRegion := []regionIntelligibility{}
// Convert the languageMatch entries in lists keyed by desired language.
for _, m := range lm[0].LanguageMatch {
// Different versions of CLDR use different separators.
desired := strings.Replace(m.Desired, "-", "_", -1)
supported := strings.Replace(m.Supported, "-", "_", -1)
d := strings.Split(desired, "_")
s := strings.Split(supported, "_")
if len(d) != len(s) {
log.Fatalf("not supported: desired=%q; supported=%q", desired, supported)
continue
}
distance, _ := strconv.ParseInt(m.Distance, 10, 8)
switch len(d) {
case 2:
if desired == supported && desired == "*_*" {
continue
}
// language-script pair.
matchScript = append(matchScript, scriptIntelligibility{
wantLang: uint16(b.langIndex(d[0])),
haveLang: uint16(b.langIndex(s[0])),
wantScript: uint8(b.scriptIndex(d[1])),
haveScript: uint8(b.scriptIndex(s[1])),
distance: uint8(distance),
})
if m.Oneway != "true" {
matchScript = append(matchScript, scriptIntelligibility{
wantLang: uint16(b.langIndex(s[0])),
haveLang: uint16(b.langIndex(d[0])),
wantScript: uint8(b.scriptIndex(s[1])),
haveScript: uint8(b.scriptIndex(d[1])),
distance: uint8(distance),
})
}
case 1:
if desired == supported && desired == "*" {
continue
}
if distance == 1 {
// nb == no is already handled by macro mapping. Check there
// really is only this case.
if d[0] != "no" || s[0] != "nb" {
log.Fatalf("unhandled equivalence %s == %s", s[0], d[0])
}
continue
}
// TODO: consider dropping oneway field and just doubling the entry.
matchLang = append(matchLang, mutualIntelligibility{
want: uint16(b.langIndex(d[0])),
have: uint16(b.langIndex(s[0])),
distance: uint8(distance),
oneway: m.Oneway == "true",
})
case 3:
if desired == supported && desired == "*_*_*" {
continue
}
if desired != supported {
// This is now supported by CLDR, but only one case, which
// should already be covered by paradigm locales. For instance,
// test case "und, en, en-GU, en-IN, en-GB ; en-ZA ; en-GB" in
// testdata/CLDRLocaleMatcherTest.txt tests this.
if supported != "en_*_GB" {
log.Fatalf("not supported: desired=%q; supported=%q", desired, supported)
}
continue
}
ri := regionIntelligibility{
lang: b.langIndex(d[0]),
distance: uint8(distance),
}
if d[1] != "*" {
ri.script = uint8(b.scriptIndex(d[1]))
}
switch {
case d[2] == "*":
ri.group = 0x80 // not contained in anything
case strings.HasPrefix(d[2], "$!"):
ri.group = 0x80
d[2] = "$" + d[2][len("$!"):]
fallthrough
case strings.HasPrefix(d[2], "$"):
ri.group |= idToIndex[d[2]]
}
matchRegion = append(matchRegion, ri)
default:
log.Fatalf("not supported: desired=%q; supported=%q", desired, supported)
}
}
sort.SliceStable(matchLang, func(i, j int) bool {
return matchLang[i].distance < matchLang[j].distance
})
b.w.WriteComment(`
matchLang holds pairs of langIDs of base languages that are typically
mutually intelligible. Each pair is associated with a confidence and
whether the intelligibility goes one or both ways.`)
b.w.WriteVar("matchLang", matchLang)
b.w.WriteComment(`
matchScript holds pairs of scriptIDs where readers of one script
can typically also read the other. Each is associated with a confidence.`)
sort.SliceStable(matchScript, func(i, j int) bool {
return matchScript[i].distance < matchScript[j].distance
})
b.w.WriteVar("matchScript", matchScript)
sort.SliceStable(matchRegion, func(i, j int) bool {
return matchRegion[i].distance < matchRegion[j].distance
})
b.w.WriteVar("matchRegion", matchRegion)
}