seba-365 - implemented dep
Change-Id: Ia6226d50e7615935a0c8876809a687427ff88c22
diff --git a/vendor/golang.org/x/text/language/lookup.go b/vendor/golang.org/x/text/language/lookup.go
new file mode 100644
index 0000000..1d80ac3
--- /dev/null
+++ b/vendor/golang.org/x/text/language/lookup.go
@@ -0,0 +1,396 @@
+// Copyright 2013 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package language
+
+import (
+ "bytes"
+ "fmt"
+ "sort"
+ "strconv"
+
+ "golang.org/x/text/internal/tag"
+)
+
+// findIndex tries to find the given tag in idx and returns a standardized error
+// if it could not be found.
+func findIndex(idx tag.Index, key []byte, form string) (index int, err error) {
+ if !tag.FixCase(form, key) {
+ return 0, errSyntax
+ }
+ i := idx.Index(key)
+ if i == -1 {
+ return 0, mkErrInvalid(key)
+ }
+ return i, nil
+}
+
+func searchUint(imap []uint16, key uint16) int {
+ return sort.Search(len(imap), func(i int) bool {
+ return imap[i] >= key
+ })
+}
+
+type langID uint16
+
+// getLangID returns the langID of s if s is a canonical subtag
+// or langUnknown if s is not a canonical subtag.
+func getLangID(s []byte) (langID, error) {
+ if len(s) == 2 {
+ return getLangISO2(s)
+ }
+ return getLangISO3(s)
+}
+
+// mapLang returns the mapped langID of id according to mapping m.
+func normLang(id langID) (langID, langAliasType) {
+ k := sort.Search(len(langAliasMap), func(i int) bool {
+ return langAliasMap[i].from >= uint16(id)
+ })
+ if k < len(langAliasMap) && langAliasMap[k].from == uint16(id) {
+ return langID(langAliasMap[k].to), langAliasTypes[k]
+ }
+ return id, langAliasTypeUnknown
+}
+
+// getLangISO2 returns the langID for the given 2-letter ISO language code
+// or unknownLang if this does not exist.
+func getLangISO2(s []byte) (langID, error) {
+ if !tag.FixCase("zz", s) {
+ return 0, errSyntax
+ }
+ if i := lang.Index(s); i != -1 && lang.Elem(i)[3] != 0 {
+ return langID(i), nil
+ }
+ return 0, mkErrInvalid(s)
+}
+
+const base = 'z' - 'a' + 1
+
+func strToInt(s []byte) uint {
+ v := uint(0)
+ for i := 0; i < len(s); i++ {
+ v *= base
+ v += uint(s[i] - 'a')
+ }
+ return v
+}
+
+// converts the given integer to the original ASCII string passed to strToInt.
+// len(s) must match the number of characters obtained.
+func intToStr(v uint, s []byte) {
+ for i := len(s) - 1; i >= 0; i-- {
+ s[i] = byte(v%base) + 'a'
+ v /= base
+ }
+}
+
+// getLangISO3 returns the langID for the given 3-letter ISO language code
+// or unknownLang if this does not exist.
+func getLangISO3(s []byte) (langID, error) {
+ if tag.FixCase("und", s) {
+ // first try to match canonical 3-letter entries
+ for i := lang.Index(s[:2]); i != -1; i = lang.Next(s[:2], i) {
+ if e := lang.Elem(i); e[3] == 0 && e[2] == s[2] {
+ // We treat "und" as special and always translate it to "unspecified".
+ // Note that ZZ and Zzzz are private use and are not treated as
+ // unspecified by default.
+ id := langID(i)
+ if id == nonCanonicalUnd {
+ return 0, nil
+ }
+ return id, nil
+ }
+ }
+ if i := altLangISO3.Index(s); i != -1 {
+ return langID(altLangIndex[altLangISO3.Elem(i)[3]]), nil
+ }
+ n := strToInt(s)
+ if langNoIndex[n/8]&(1<<(n%8)) != 0 {
+ return langID(n) + langNoIndexOffset, nil
+ }
+ // Check for non-canonical uses of ISO3.
+ for i := lang.Index(s[:1]); i != -1; i = lang.Next(s[:1], i) {
+ if e := lang.Elem(i); e[2] == s[1] && e[3] == s[2] {
+ return langID(i), nil
+ }
+ }
+ return 0, mkErrInvalid(s)
+ }
+ return 0, errSyntax
+}
+
+// stringToBuf writes the string to b and returns the number of bytes
+// written. cap(b) must be >= 3.
+func (id langID) stringToBuf(b []byte) int {
+ if id >= langNoIndexOffset {
+ intToStr(uint(id)-langNoIndexOffset, b[:3])
+ return 3
+ } else if id == 0 {
+ return copy(b, "und")
+ }
+ l := lang[id<<2:]
+ if l[3] == 0 {
+ return copy(b, l[:3])
+ }
+ return copy(b, l[:2])
+}
+
+// String returns the BCP 47 representation of the langID.
+// Use b as variable name, instead of id, to ensure the variable
+// used is consistent with that of Base in which this type is embedded.
+func (b langID) String() string {
+ if b == 0 {
+ return "und"
+ } else if b >= langNoIndexOffset {
+ b -= langNoIndexOffset
+ buf := [3]byte{}
+ intToStr(uint(b), buf[:])
+ return string(buf[:])
+ }
+ l := lang.Elem(int(b))
+ if l[3] == 0 {
+ return l[:3]
+ }
+ return l[:2]
+}
+
+// ISO3 returns the ISO 639-3 language code.
+func (b langID) ISO3() string {
+ if b == 0 || b >= langNoIndexOffset {
+ return b.String()
+ }
+ l := lang.Elem(int(b))
+ if l[3] == 0 {
+ return l[:3]
+ } else if l[2] == 0 {
+ return altLangISO3.Elem(int(l[3]))[:3]
+ }
+ // This allocation will only happen for 3-letter ISO codes
+ // that are non-canonical BCP 47 language identifiers.
+ return l[0:1] + l[2:4]
+}
+
+// IsPrivateUse reports whether this language code is reserved for private use.
+func (b langID) IsPrivateUse() bool {
+ return langPrivateStart <= b && b <= langPrivateEnd
+}
+
+type regionID uint16
+
+// getRegionID returns the region id for s if s is a valid 2-letter region code
+// or unknownRegion.
+func getRegionID(s []byte) (regionID, error) {
+ if len(s) == 3 {
+ if isAlpha(s[0]) {
+ return getRegionISO3(s)
+ }
+ if i, err := strconv.ParseUint(string(s), 10, 10); err == nil {
+ return getRegionM49(int(i))
+ }
+ }
+ return getRegionISO2(s)
+}
+
+// getRegionISO2 returns the regionID for the given 2-letter ISO country code
+// or unknownRegion if this does not exist.
+func getRegionISO2(s []byte) (regionID, error) {
+ i, err := findIndex(regionISO, s, "ZZ")
+ if err != nil {
+ return 0, err
+ }
+ return regionID(i) + isoRegionOffset, nil
+}
+
+// getRegionISO3 returns the regionID for the given 3-letter ISO country code
+// or unknownRegion if this does not exist.
+func getRegionISO3(s []byte) (regionID, error) {
+ if tag.FixCase("ZZZ", s) {
+ for i := regionISO.Index(s[:1]); i != -1; i = regionISO.Next(s[:1], i) {
+ if e := regionISO.Elem(i); e[2] == s[1] && e[3] == s[2] {
+ return regionID(i) + isoRegionOffset, nil
+ }
+ }
+ for i := 0; i < len(altRegionISO3); i += 3 {
+ if tag.Compare(altRegionISO3[i:i+3], s) == 0 {
+ return regionID(altRegionIDs[i/3]), nil
+ }
+ }
+ return 0, mkErrInvalid(s)
+ }
+ return 0, errSyntax
+}
+
+func getRegionM49(n int) (regionID, error) {
+ if 0 < n && n <= 999 {
+ const (
+ searchBits = 7
+ regionBits = 9
+ regionMask = 1<<regionBits - 1
+ )
+ idx := n >> searchBits
+ buf := fromM49[m49Index[idx]:m49Index[idx+1]]
+ val := uint16(n) << regionBits // we rely on bits shifting out
+ i := sort.Search(len(buf), func(i int) bool {
+ return buf[i] >= val
+ })
+ if r := fromM49[int(m49Index[idx])+i]; r&^regionMask == val {
+ return regionID(r & regionMask), nil
+ }
+ }
+ var e ValueError
+ fmt.Fprint(bytes.NewBuffer([]byte(e.v[:])), n)
+ return 0, e
+}
+
+// normRegion returns a region if r is deprecated or 0 otherwise.
+// TODO: consider supporting BYS (-> BLR), CSK (-> 200 or CZ), PHI (-> PHL) and AFI (-> DJ).
+// TODO: consider mapping split up regions to new most populous one (like CLDR).
+func normRegion(r regionID) regionID {
+ m := regionOldMap
+ k := sort.Search(len(m), func(i int) bool {
+ return m[i].from >= uint16(r)
+ })
+ if k < len(m) && m[k].from == uint16(r) {
+ return regionID(m[k].to)
+ }
+ return 0
+}
+
+const (
+ iso3166UserAssigned = 1 << iota
+ ccTLD
+ bcp47Region
+)
+
+func (r regionID) typ() byte {
+ return regionTypes[r]
+}
+
+// String returns the BCP 47 representation for the region.
+// It returns "ZZ" for an unspecified region.
+func (r regionID) String() string {
+ if r < isoRegionOffset {
+ if r == 0 {
+ return "ZZ"
+ }
+ return fmt.Sprintf("%03d", r.M49())
+ }
+ r -= isoRegionOffset
+ return regionISO.Elem(int(r))[:2]
+}
+
+// ISO3 returns the 3-letter ISO code of r.
+// Note that not all regions have a 3-letter ISO code.
+// In such cases this method returns "ZZZ".
+func (r regionID) ISO3() string {
+ if r < isoRegionOffset {
+ return "ZZZ"
+ }
+ r -= isoRegionOffset
+ reg := regionISO.Elem(int(r))
+ switch reg[2] {
+ case 0:
+ return altRegionISO3[reg[3]:][:3]
+ case ' ':
+ return "ZZZ"
+ }
+ return reg[0:1] + reg[2:4]
+}
+
+// M49 returns the UN M.49 encoding of r, or 0 if this encoding
+// is not defined for r.
+func (r regionID) M49() int {
+ return int(m49[r])
+}
+
+// IsPrivateUse reports whether r has the ISO 3166 User-assigned status. This
+// may include private-use tags that are assigned by CLDR and used in this
+// implementation. So IsPrivateUse and IsCountry can be simultaneously true.
+func (r regionID) IsPrivateUse() bool {
+ return r.typ()&iso3166UserAssigned != 0
+}
+
+type scriptID uint8
+
+// getScriptID returns the script id for string s. It assumes that s
+// is of the format [A-Z][a-z]{3}.
+func getScriptID(idx tag.Index, s []byte) (scriptID, error) {
+ i, err := findIndex(idx, s, "Zzzz")
+ return scriptID(i), err
+}
+
+// String returns the script code in title case.
+// It returns "Zzzz" for an unspecified script.
+func (s scriptID) String() string {
+ if s == 0 {
+ return "Zzzz"
+ }
+ return script.Elem(int(s))
+}
+
+// IsPrivateUse reports whether this script code is reserved for private use.
+func (s scriptID) IsPrivateUse() bool {
+ return _Qaaa <= s && s <= _Qabx
+}
+
+const (
+ maxAltTaglen = len("en-US-POSIX")
+ maxLen = maxAltTaglen
+)
+
+var (
+ // grandfatheredMap holds a mapping from legacy and grandfathered tags to
+ // their base language or index to more elaborate tag.
+ grandfatheredMap = map[[maxLen]byte]int16{
+ [maxLen]byte{'a', 'r', 't', '-', 'l', 'o', 'j', 'b', 'a', 'n'}: _jbo, // art-lojban
+ [maxLen]byte{'i', '-', 'a', 'm', 'i'}: _ami, // i-ami
+ [maxLen]byte{'i', '-', 'b', 'n', 'n'}: _bnn, // i-bnn
+ [maxLen]byte{'i', '-', 'h', 'a', 'k'}: _hak, // i-hak
+ [maxLen]byte{'i', '-', 'k', 'l', 'i', 'n', 'g', 'o', 'n'}: _tlh, // i-klingon
+ [maxLen]byte{'i', '-', 'l', 'u', 'x'}: _lb, // i-lux
+ [maxLen]byte{'i', '-', 'n', 'a', 'v', 'a', 'j', 'o'}: _nv, // i-navajo
+ [maxLen]byte{'i', '-', 'p', 'w', 'n'}: _pwn, // i-pwn
+ [maxLen]byte{'i', '-', 't', 'a', 'o'}: _tao, // i-tao
+ [maxLen]byte{'i', '-', 't', 'a', 'y'}: _tay, // i-tay
+ [maxLen]byte{'i', '-', 't', 's', 'u'}: _tsu, // i-tsu
+ [maxLen]byte{'n', 'o', '-', 'b', 'o', 'k'}: _nb, // no-bok
+ [maxLen]byte{'n', 'o', '-', 'n', 'y', 'n'}: _nn, // no-nyn
+ [maxLen]byte{'s', 'g', 'n', '-', 'b', 'e', '-', 'f', 'r'}: _sfb, // sgn-BE-FR
+ [maxLen]byte{'s', 'g', 'n', '-', 'b', 'e', '-', 'n', 'l'}: _vgt, // sgn-BE-NL
+ [maxLen]byte{'s', 'g', 'n', '-', 'c', 'h', '-', 'd', 'e'}: _sgg, // sgn-CH-DE
+ [maxLen]byte{'z', 'h', '-', 'g', 'u', 'o', 'y', 'u'}: _cmn, // zh-guoyu
+ [maxLen]byte{'z', 'h', '-', 'h', 'a', 'k', 'k', 'a'}: _hak, // zh-hakka
+ [maxLen]byte{'z', 'h', '-', 'm', 'i', 'n', '-', 'n', 'a', 'n'}: _nan, // zh-min-nan
+ [maxLen]byte{'z', 'h', '-', 'x', 'i', 'a', 'n', 'g'}: _hsn, // zh-xiang
+
+ // Grandfathered tags with no modern replacement will be converted as
+ // follows:
+ [maxLen]byte{'c', 'e', 'l', '-', 'g', 'a', 'u', 'l', 'i', 's', 'h'}: -1, // cel-gaulish
+ [maxLen]byte{'e', 'n', '-', 'g', 'b', '-', 'o', 'e', 'd'}: -2, // en-GB-oed
+ [maxLen]byte{'i', '-', 'd', 'e', 'f', 'a', 'u', 'l', 't'}: -3, // i-default
+ [maxLen]byte{'i', '-', 'e', 'n', 'o', 'c', 'h', 'i', 'a', 'n'}: -4, // i-enochian
+ [maxLen]byte{'i', '-', 'm', 'i', 'n', 'g', 'o'}: -5, // i-mingo
+ [maxLen]byte{'z', 'h', '-', 'm', 'i', 'n'}: -6, // zh-min
+
+ // CLDR-specific tag.
+ [maxLen]byte{'r', 'o', 'o', 't'}: 0, // root
+ [maxLen]byte{'e', 'n', '-', 'u', 's', '-', 'p', 'o', 's', 'i', 'x'}: -7, // en_US_POSIX"
+ }
+
+ altTagIndex = [...]uint8{0, 17, 31, 45, 61, 74, 86, 102}
+
+ altTags = "xtg-x-cel-gaulishen-GB-oxendicten-x-i-defaultund-x-i-enochiansee-x-i-mingonan-x-zh-minen-US-u-va-posix"
+)
+
+func grandfathered(s [maxAltTaglen]byte) (t Tag, ok bool) {
+ if v, ok := grandfatheredMap[s]; ok {
+ if v < 0 {
+ return Make(altTags[altTagIndex[-v-1]:altTagIndex[-v]]), true
+ }
+ t.lang = langID(v)
+ return t, true
+ }
+ return t, false
+}