Initial

Signed-off-by: Alex Ellis (OpenFaaS Ltd) <alexellis2@gmail.com>

vendor/golang.org/x/text/internal/colltab/collelem.go

@@ -0,0 +1,371 @@
+// Copyright 2012 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 colltab
+
+import (
+	"fmt"
+	"unicode"
+)
+
+// Level identifies the collation comparison level.
+// The primary level corresponds to the basic sorting of text.
+// The secondary level corresponds to accents and related linguistic elements.
+// The tertiary level corresponds to casing and related concepts.
+// The quaternary level is derived from the other levels by the
+// various algorithms for handling variable elements.
+type Level int
+
+const (
+	Primary Level = iota
+	Secondary
+	Tertiary
+	Quaternary
+	Identity
+
+	NumLevels
+)
+
+const (
+	defaultSecondary = 0x20
+	defaultTertiary  = 0x2
+	maxTertiary      = 0x1F
+	MaxQuaternary    = 0x1FFFFF // 21 bits.
+)
+
+// Elem is a representation of a collation element. This API provides ways to encode
+// and decode Elems. Implementations of collation tables may use values greater
+// or equal to PrivateUse for their own purposes.  However, these should never be
+// returned by AppendNext.
+type Elem uint32
+
+const (
+	maxCE       Elem = 0xAFFFFFFF
+	PrivateUse       = minContract
+	minContract      = 0xC0000000
+	maxContract      = 0xDFFFFFFF
+	minExpand        = 0xE0000000
+	maxExpand        = 0xEFFFFFFF
+	minDecomp        = 0xF0000000
+)
+
+type ceType int
+
+const (
+	ceNormal           ceType = iota // ceNormal includes implicits (ce == 0)
+	ceContractionIndex               // rune can be a start of a contraction
+	ceExpansionIndex                 // rune expands into a sequence of collation elements
+	ceDecompose                      // rune expands using NFKC decomposition
+)
+
+func (ce Elem) ctype() ceType {
+	if ce <= maxCE {
+		return ceNormal
+	}
+	if ce <= maxContract {
+		return ceContractionIndex
+	} else {
+		if ce <= maxExpand {
+			return ceExpansionIndex
+		}
+		return ceDecompose
+	}
+	panic("should not reach here")
+	return ceType(-1)
+}
+
+// For normal collation elements, we assume that a collation element either has
+// a primary or non-default secondary value, not both.
+// Collation elements with a primary value are of the form
+// 01pppppp pppppppp ppppppp0 ssssssss
+//   - p* is primary collation value
+//   - s* is the secondary collation value
+// 00pppppp pppppppp ppppppps sssttttt, where
+//   - p* is primary collation value
+//   - s* offset of secondary from default value.
+//   - t* is the tertiary collation value
+// 100ttttt cccccccc pppppppp pppppppp
+//   - t* is the tertiar collation value
+//   - c* is the canonical combining class
+//   - p* is the primary collation value
+// Collation elements with a secondary value are of the form
+// 1010cccc ccccssss ssssssss tttttttt, where
+//   - c* is the canonical combining class
+//   - s* is the secondary collation value
+//   - t* is the tertiary collation value
+// 11qqqqqq qqqqqqqq qqqqqqq0 00000000
+//   - q* quaternary value
+const (
+	ceTypeMask              = 0xC0000000
+	ceTypeMaskExt           = 0xE0000000
+	ceIgnoreMask            = 0xF00FFFFF
+	ceType1                 = 0x40000000
+	ceType2                 = 0x00000000
+	ceType3or4              = 0x80000000
+	ceType4                 = 0xA0000000
+	ceTypeQ                 = 0xC0000000
+	Ignore                  = ceType4
+	firstNonPrimary         = 0x80000000
+	lastSpecialPrimary      = 0xA0000000
+	secondaryMask           = 0x80000000
+	hasTertiaryMask         = 0x40000000
+	primaryValueMask        = 0x3FFFFE00
+	maxPrimaryBits          = 21
+	compactPrimaryBits      = 16
+	maxSecondaryBits        = 12
+	maxTertiaryBits         = 8
+	maxCCCBits              = 8
+	maxSecondaryCompactBits = 8
+	maxSecondaryDiffBits    = 4
+	maxTertiaryCompactBits  = 5
+	primaryShift            = 9
+	compactSecondaryShift   = 5
+	minCompactSecondary     = defaultSecondary - 4
+)
+
+func makeImplicitCE(primary int) Elem {
+	return ceType1 | Elem(primary<<primaryShift) | defaultSecondary
+}
+
+// MakeElem returns an Elem for the given values.  It will return an error
+// if the given combination of values is invalid.
+func MakeElem(primary, secondary, tertiary int, ccc uint8) (Elem, error) {
+	if w := primary; w >= 1<<maxPrimaryBits || w < 0 {
+		return 0, fmt.Errorf("makeCE: primary weight out of bounds: %x >= %x", w, 1<<maxPrimaryBits)
+	}
+	if w := secondary; w >= 1<<maxSecondaryBits || w < 0 {
+		return 0, fmt.Errorf("makeCE: secondary weight out of bounds: %x >= %x", w, 1<<maxSecondaryBits)
+	}
+	if w := tertiary; w >= 1<<maxTertiaryBits || w < 0 {
+		return 0, fmt.Errorf("makeCE: tertiary weight out of bounds: %x >= %x", w, 1<<maxTertiaryBits)
+	}
+	ce := Elem(0)
+	if primary != 0 {
+		if ccc != 0 {
+			if primary >= 1<<compactPrimaryBits {
+				return 0, fmt.Errorf("makeCE: primary weight with non-zero CCC out of bounds: %x >= %x", primary, 1<<compactPrimaryBits)
+			}
+			if secondary != defaultSecondary {
+				return 0, fmt.Errorf("makeCE: cannot combine non-default secondary value (%x) with non-zero CCC (%x)", secondary, ccc)
+			}
+			ce = Elem(tertiary << (compactPrimaryBits + maxCCCBits))
+			ce |= Elem(ccc) << compactPrimaryBits
+			ce |= Elem(primary)
+			ce |= ceType3or4
+		} else if tertiary == defaultTertiary {
+			if secondary >= 1<<maxSecondaryCompactBits {
+				return 0, fmt.Errorf("makeCE: secondary weight with non-zero primary out of bounds: %x >= %x", secondary, 1<<maxSecondaryCompactBits)
+			}
+			ce = Elem(primary<<(maxSecondaryCompactBits+1) + secondary)
+			ce |= ceType1
+		} else {
+			d := secondary - defaultSecondary + maxSecondaryDiffBits
+			if d >= 1<<maxSecondaryDiffBits || d < 0 {
+				return 0, fmt.Errorf("makeCE: secondary weight diff out of bounds: %x < 0 || %x > %x", d, d, 1<<maxSecondaryDiffBits)
+			}
+			if tertiary >= 1<<maxTertiaryCompactBits {
+				return 0, fmt.Errorf("makeCE: tertiary weight with non-zero primary out of bounds: %x > %x", tertiary, 1<<maxTertiaryCompactBits)
+			}
+			ce = Elem(primary<<maxSecondaryDiffBits + d)
+			ce = ce<<maxTertiaryCompactBits + Elem(tertiary)
+		}
+	} else {
+		ce = Elem(secondary<<maxTertiaryBits + tertiary)
+		ce += Elem(ccc) << (maxSecondaryBits + maxTertiaryBits)
+		ce |= ceType4
+	}
+	return ce, nil
+}
+
+// MakeQuaternary returns an Elem with the given quaternary value.
+func MakeQuaternary(v int) Elem {
+	return ceTypeQ | Elem(v<<primaryShift)
+}
+
+// Mask sets weights for any level smaller than l to 0.
+// The resulting Elem can be used to test for equality with
+// other Elems to which the same mask has been applied.
+func (ce Elem) Mask(l Level) uint32 {
+	return 0
+}
+
+// CCC returns the canonical combining class associated with the underlying character,
+// if applicable, or 0 otherwise.
+func (ce Elem) CCC() uint8 {
+	if ce&ceType3or4 != 0 {
+		if ce&ceType4 == ceType3or4 {
+			return uint8(ce >> 16)
+		}
+		return uint8(ce >> 20)
+	}
+	return 0
+}
+
+// Primary returns the primary collation weight for ce.
+func (ce Elem) Primary() int {
+	if ce >= firstNonPrimary {
+		if ce > lastSpecialPrimary {
+			return 0
+		}
+		return int(uint16(ce))
+	}
+	return int(ce&primaryValueMask) >> primaryShift
+}
+
+// Secondary returns the secondary collation weight for ce.
+func (ce Elem) Secondary() int {
+	switch ce & ceTypeMask {
+	case ceType1:
+		return int(uint8(ce))
+	case ceType2:
+		return minCompactSecondary + int((ce>>compactSecondaryShift)&0xF)
+	case ceType3or4:
+		if ce < ceType4 {
+			return defaultSecondary
+		}
+		return int(ce>>8) & 0xFFF
+	case ceTypeQ:
+		return 0
+	}
+	panic("should not reach here")
+}
+
+// Tertiary returns the tertiary collation weight for ce.
+func (ce Elem) Tertiary() uint8 {
+	if ce&hasTertiaryMask == 0 {
+		if ce&ceType3or4 == 0 {
+			return uint8(ce & 0x1F)
+		}
+		if ce&ceType4 == ceType4 {
+			return uint8(ce)
+		}
+		return uint8(ce>>24) & 0x1F // type 2
+	} else if ce&ceTypeMask == ceType1 {
+		return defaultTertiary
+	}
+	// ce is a quaternary value.
+	return 0
+}
+
+func (ce Elem) updateTertiary(t uint8) Elem {
+	if ce&ceTypeMask == ceType1 {
+		// convert to type 4
+		nce := ce & primaryValueMask
+		nce |= Elem(uint8(ce)-minCompactSecondary) << compactSecondaryShift
+		ce = nce
+	} else if ce&ceTypeMaskExt == ceType3or4 {
+		ce &= ^Elem(maxTertiary << 24)
+		return ce | (Elem(t) << 24)
+	} else {
+		// type 2 or 4
+		ce &= ^Elem(maxTertiary)
+	}
+	return ce | Elem(t)
+}
+
+// Quaternary returns the quaternary value if explicitly specified,
+// 0 if ce == Ignore, or MaxQuaternary otherwise.
+// Quaternary values are used only for shifted variants.
+func (ce Elem) Quaternary() int {
+	if ce&ceTypeMask == ceTypeQ {
+		return int(ce&primaryValueMask) >> primaryShift
+	} else if ce&ceIgnoreMask == Ignore {
+		return 0
+	}
+	return MaxQuaternary
+}
+
+// Weight returns the collation weight for the given level.
+func (ce Elem) Weight(l Level) int {
+	switch l {
+	case Primary:
+		return ce.Primary()
+	case Secondary:
+		return ce.Secondary()
+	case Tertiary:
+		return int(ce.Tertiary())
+	case Quaternary:
+		return ce.Quaternary()
+	}
+	return 0 // return 0 (ignore) for undefined levels.
+}
+
+// For contractions, collation elements are of the form
+// 110bbbbb bbbbbbbb iiiiiiii iiiinnnn, where
+//   - n* is the size of the first node in the contraction trie.
+//   - i* is the index of the first node in the contraction trie.
+//   - b* is the offset into the contraction collation element table.
+// See contract.go for details on the contraction trie.
+const (
+	maxNBits              = 4
+	maxTrieIndexBits      = 12
+	maxContractOffsetBits = 13
+)
+
+func splitContractIndex(ce Elem) (index, n, offset int) {
+	n = int(ce & (1<<maxNBits - 1))
+	ce >>= maxNBits
+	index = int(ce & (1<<maxTrieIndexBits - 1))
+	ce >>= maxTrieIndexBits
+	offset = int(ce & (1<<maxContractOffsetBits - 1))
+	return
+}
+
+// For expansions, Elems are of the form 11100000 00000000 bbbbbbbb bbbbbbbb,
+// where b* is the index into the expansion sequence table.
+const maxExpandIndexBits = 16
+
+func splitExpandIndex(ce Elem) (index int) {
+	return int(uint16(ce))
+}
+
+// Some runes can be expanded using NFKD decomposition. Instead of storing the full
+// sequence of collation elements, we decompose the rune and lookup the collation
+// elements for each rune in the decomposition and modify the tertiary weights.
+// The Elem, in this case, is of the form 11110000 00000000 wwwwwwww vvvvvvvv, where
+//   - v* is the replacement tertiary weight for the first rune,
+//   - w* is the replacement tertiary weight for the second rune,
+// Tertiary weights of subsequent runes should be replaced with maxTertiary.
+// See https://www.unicode.org/reports/tr10/#Compatibility_Decompositions for more details.
+func splitDecompose(ce Elem) (t1, t2 uint8) {
+	return uint8(ce), uint8(ce >> 8)
+}
+
+const (
+	// These constants were taken from https://www.unicode.org/versions/Unicode6.0.0/ch12.pdf.
+	minUnified       rune = 0x4E00
+	maxUnified            = 0x9FFF
+	minCompatibility      = 0xF900
+	maxCompatibility      = 0xFAFF
+	minRare               = 0x3400
+	maxRare               = 0x4DBF
+)
+const (
+	commonUnifiedOffset = 0x10000
+	rareUnifiedOffset   = 0x20000 // largest rune in common is U+FAFF
+	otherOffset         = 0x50000 // largest rune in rare is U+2FA1D
+	illegalOffset       = otherOffset + int(unicode.MaxRune)
+	maxPrimary          = illegalOffset + 1
+)
+
+// implicitPrimary returns the primary weight for the a rune
+// for which there is no entry for the rune in the collation table.
+// We take a different approach from the one specified in
+// https://unicode.org/reports/tr10/#Implicit_Weights,
+// but preserve the resulting relative ordering of the runes.
+func implicitPrimary(r rune) int {
+	if unicode.Is(unicode.Ideographic, r) {
+		if r >= minUnified && r <= maxUnified {
+			// The most common case for CJK.
+			return int(r) + commonUnifiedOffset
+		}
+		if r >= minCompatibility && r <= maxCompatibility {
+			// This will typically not hit. The DUCET explicitly specifies mappings
+			// for all characters that do not decompose.
+			return int(r) + commonUnifiedOffset
+		}
+		return int(r) + rareUnifiedOffset
+	}
+	return int(r) + otherOffset
+}

vendor/golang.org/x/text/internal/colltab/colltab.go

@@ -0,0 +1,105 @@
+// Copyright 2015 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 colltab contains functionality related to collation tables.
+// It is only to be used by the collate and search packages.
+package colltab // import "golang.org/x/text/internal/colltab"
+
+import (
+	"sort"
+
+	"golang.org/x/text/language"
+)
+
+// MatchLang finds the index of t in tags, using a matching algorithm used for
+// collation and search. tags[0] must be language.Und, the remaining tags should
+// be sorted alphabetically.
+//
+// Language matching for collation and search is different from the matching
+// defined by language.Matcher: the (inferred) base language must be an exact
+// match for the relevant fields. For example, "gsw" should not match "de".
+// Also the parent relation is different, as a parent may have a different
+// script. So usually the parent of zh-Hant is und, whereas for MatchLang it is
+// zh.
+func MatchLang(t language.Tag, tags []language.Tag) int {
+	// Canonicalize the values, including collapsing macro languages.
+	t, _ = language.All.Canonicalize(t)
+
+	base, conf := t.Base()
+	// Estimate the base language, but only use high-confidence values.
+	if conf < language.High {
+		// The root locale supports "search" and "standard". We assume that any
+		// implementation will only use one of both.
+		return 0
+	}
+
+	// Maximize base and script and normalize the tag.
+	if _, s, r := t.Raw(); (r != language.Region{}) {
+		p, _ := language.Raw.Compose(base, s, r)
+		// Taking the parent forces the script to be maximized.
+		p = p.Parent()
+		// Add back region and extensions.
+		t, _ = language.Raw.Compose(p, r, t.Extensions())
+	} else {
+		// Set the maximized base language.
+		t, _ = language.Raw.Compose(base, s, t.Extensions())
+	}
+
+	// Find start index of the language tag.
+	start := 1 + sort.Search(len(tags)-1, func(i int) bool {
+		b, _, _ := tags[i+1].Raw()
+		return base.String() <= b.String()
+	})
+	if start < len(tags) {
+		if b, _, _ := tags[start].Raw(); b != base {
+			return 0
+		}
+	}
+
+	// Besides the base language, script and region, only the collation type and
+	// the custom variant defined in the 'u' extension are used to distinguish a
+	// locale.
+	// Strip all variants and extensions and add back the custom variant.
+	tdef, _ := language.Raw.Compose(t.Raw())
+	tdef, _ = tdef.SetTypeForKey("va", t.TypeForKey("va"))
+
+	// First search for a specialized collation type, if present.
+	try := []language.Tag{tdef}
+	if co := t.TypeForKey("co"); co != "" {
+		tco, _ := tdef.SetTypeForKey("co", co)
+		try = []language.Tag{tco, tdef}
+	}
+
+	for _, tx := range try {
+		for ; tx != language.Und; tx = parent(tx) {
+			for i, t := range tags[start:] {
+				if b, _, _ := t.Raw(); b != base {
+					break
+				}
+				if tx == t {
+					return start + i
+				}
+			}
+		}
+	}
+	return 0
+}
+
+// parent computes the structural parent. This means inheritance may change
+// script. So, unlike the CLDR parent, parent(zh-Hant) == zh.
+func parent(t language.Tag) language.Tag {
+	if t.TypeForKey("va") != "" {
+		t, _ = t.SetTypeForKey("va", "")
+		return t
+	}
+	result := language.Und
+	if b, s, r := t.Raw(); (r != language.Region{}) {
+		result, _ = language.Raw.Compose(b, s, t.Extensions())
+	} else if (s != language.Script{}) {
+		result, _ = language.Raw.Compose(b, t.Extensions())
+	} else if (b != language.Base{}) {
+		result, _ = language.Raw.Compose(t.Extensions())
+	}
+	return result
+}

vendor/golang.org/x/text/internal/colltab/contract.go

@@ -0,0 +1,145 @@
+// Copyright 2012 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 colltab
+
+import "unicode/utf8"
+
+// For a description of ContractTrieSet, see text/collate/build/contract.go.
+
+type ContractTrieSet []struct{ L, H, N, I uint8 }
+
+// ctScanner is used to match a trie to an input sequence.
+// A contraction may match a non-contiguous sequence of bytes in an input string.
+// For example, if there is a contraction for <a, combining_ring>, it should match
+// the sequence <a, combining_cedilla, combining_ring>, as combining_cedilla does
+// not block combining_ring.
+// ctScanner does not automatically skip over non-blocking non-starters, but rather
+// retains the state of the last match and leaves it up to the user to continue
+// the match at the appropriate points.
+type ctScanner struct {
+	states ContractTrieSet
+	s      []byte
+	n      int
+	index  int
+	pindex int
+	done   bool
+}
+
+type ctScannerString struct {
+	states ContractTrieSet
+	s      string
+	n      int
+	index  int
+	pindex int
+	done   bool
+}
+
+func (t ContractTrieSet) scanner(index, n int, b []byte) ctScanner {
+	return ctScanner{s: b, states: t[index:], n: n}
+}
+
+func (t ContractTrieSet) scannerString(index, n int, str string) ctScannerString {
+	return ctScannerString{s: str, states: t[index:], n: n}
+}
+
+// result returns the offset i and bytes consumed p so far.  If no suffix
+// matched, i and p will be 0.
+func (s *ctScanner) result() (i, p int) {
+	return s.index, s.pindex
+}
+
+func (s *ctScannerString) result() (i, p int) {
+	return s.index, s.pindex
+}
+
+const (
+	final   = 0
+	noIndex = 0xFF
+)
+
+// scan matches the longest suffix at the current location in the input
+// and returns the number of bytes consumed.
+func (s *ctScanner) scan(p int) int {
+	pr := p // the p at the rune start
+	str := s.s
+	states, n := s.states, s.n
+	for i := 0; i < n && p < len(str); {
+		e := states[i]
+		c := str[p]
+		// TODO: a significant number of contractions are of a form that
+		// cannot match discontiguous UTF-8 in a normalized string. We could let
+		// a negative value of e.n mean that we can set s.done = true and avoid
+		// the need for additional matches.
+		if c >= e.L {
+			if e.L == c {
+				p++
+				if e.I != noIndex {
+					s.index = int(e.I)
+					s.pindex = p
+				}
+				if e.N != final {
+					i, states, n = 0, states[int(e.H)+n:], int(e.N)
+					if p >= len(str) || utf8.RuneStart(str[p]) {
+						s.states, s.n, pr = states, n, p
+					}
+				} else {
+					s.done = true
+					return p
+				}
+				continue
+			} else if e.N == final && c <= e.H {
+				p++
+				s.done = true
+				s.index = int(c-e.L) + int(e.I)
+				s.pindex = p
+				return p
+			}
+		}
+		i++
+	}
+	return pr
+}
+
+// scan is a verbatim copy of ctScanner.scan.
+func (s *ctScannerString) scan(p int) int {
+	pr := p // the p at the rune start
+	str := s.s
+	states, n := s.states, s.n
+	for i := 0; i < n && p < len(str); {
+		e := states[i]
+		c := str[p]
+		// TODO: a significant number of contractions are of a form that
+		// cannot match discontiguous UTF-8 in a normalized string. We could let
+		// a negative value of e.n mean that we can set s.done = true and avoid
+		// the need for additional matches.
+		if c >= e.L {
+			if e.L == c {
+				p++
+				if e.I != noIndex {
+					s.index = int(e.I)
+					s.pindex = p
+				}
+				if e.N != final {
+					i, states, n = 0, states[int(e.H)+n:], int(e.N)
+					if p >= len(str) || utf8.RuneStart(str[p]) {
+						s.states, s.n, pr = states, n, p
+					}
+				} else {
+					s.done = true
+					return p
+				}
+				continue
+			} else if e.N == final && c <= e.H {
+				p++
+				s.done = true
+				s.index = int(c-e.L) + int(e.I)
+				s.pindex = p
+				return p
+			}
+		}
+		i++
+	}
+	return pr
+}

vendor/golang.org/x/text/internal/colltab/iter.go

@@ -0,0 +1,178 @@
+// Copyright 2015 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 colltab
+
+// An Iter incrementally converts chunks of the input text to collation
+// elements, while ensuring that the collation elements are in normalized order
+// (that is, they are in the order as if the input text were normalized first).
+type Iter struct {
+	Weighter Weighter
+	Elems    []Elem
+	// N is the number of elements in Elems that will not be reordered on
+	// subsequent iterations, N <= len(Elems).
+	N int
+
+	bytes []byte
+	str   string
+	// Because the Elems buffer may contain collation elements that are needed
+	// for look-ahead, we need two positions in the text (bytes or str): one for
+	// the end position in the text for the current iteration and one for the
+	// start of the next call to appendNext.
+	pEnd  int // end position in text corresponding to N.
+	pNext int // pEnd <= pNext.
+}
+
+// Reset sets the position in the current input text to p and discards any
+// results obtained so far.
+func (i *Iter) Reset(p int) {
+	i.Elems = i.Elems[:0]
+	i.N = 0
+	i.pEnd = p
+	i.pNext = p
+}
+
+// Len returns the length of the input text.
+func (i *Iter) Len() int {
+	if i.bytes != nil {
+		return len(i.bytes)
+	}
+	return len(i.str)
+}
+
+// Discard removes the collation elements up to N.
+func (i *Iter) Discard() {
+	// TODO: change this such that only modifiers following starters will have
+	// to be copied.
+	i.Elems = i.Elems[:copy(i.Elems, i.Elems[i.N:])]
+	i.N = 0
+}
+
+// End returns the end position of the input text for which Next has returned
+// results.
+func (i *Iter) End() int {
+	return i.pEnd
+}
+
+// SetInput resets i to input s.
+func (i *Iter) SetInput(s []byte) {
+	i.bytes = s
+	i.str = ""
+	i.Reset(0)
+}
+
+// SetInputString resets i to input s.
+func (i *Iter) SetInputString(s string) {
+	i.str = s
+	i.bytes = nil
+	i.Reset(0)
+}
+
+func (i *Iter) done() bool {
+	return i.pNext >= len(i.str) && i.pNext >= len(i.bytes)
+}
+
+func (i *Iter) appendNext() bool {
+	if i.done() {
+		return false
+	}
+	var sz int
+	if i.bytes == nil {
+		i.Elems, sz = i.Weighter.AppendNextString(i.Elems, i.str[i.pNext:])
+	} else {
+		i.Elems, sz = i.Weighter.AppendNext(i.Elems, i.bytes[i.pNext:])
+	}
+	if sz == 0 {
+		sz = 1
+	}
+	i.pNext += sz
+	return true
+}
+
+// Next appends Elems to the internal array. On each iteration, it will either
+// add starters or modifiers. In the majority of cases, an Elem with a primary
+// value > 0 will have a CCC of 0. The CCC values of collation elements are also
+// used to detect if the input string was not normalized and to adjust the
+// result accordingly.
+func (i *Iter) Next() bool {
+	if i.N == len(i.Elems) && !i.appendNext() {
+		return false
+	}
+
+	// Check if the current segment starts with a starter.
+	prevCCC := i.Elems[len(i.Elems)-1].CCC()
+	if prevCCC == 0 {
+		i.N = len(i.Elems)
+		i.pEnd = i.pNext
+		return true
+	} else if i.Elems[i.N].CCC() == 0 {
+		// set i.N to only cover part of i.Elems for which prevCCC == 0 and
+		// use rest for the next call to next.
+		for i.N++; i.N < len(i.Elems) && i.Elems[i.N].CCC() == 0; i.N++ {
+		}
+		i.pEnd = i.pNext
+		return true
+	}
+
+	// The current (partial) segment starts with modifiers. We need to collect
+	// all successive modifiers to ensure that they are normalized.
+	for {
+		p := len(i.Elems)
+		i.pEnd = i.pNext
+		if !i.appendNext() {
+			break
+		}
+
+		if ccc := i.Elems[p].CCC(); ccc == 0 || len(i.Elems)-i.N > maxCombiningCharacters {
+			// Leave the starter for the next iteration. This ensures that we
+			// do not return sequences of collation elements that cross two
+			// segments.
+			//
+			// TODO: handle large number of combining characters by fully
+			// normalizing the input segment before iteration. This ensures
+			// results are consistent across the text repo.
+			i.N = p
+			return true
+		} else if ccc < prevCCC {
+			i.doNorm(p, ccc) // should be rare, never occurs for NFD and FCC.
+		} else {
+			prevCCC = ccc
+		}
+	}
+
+	done := len(i.Elems) != i.N
+	i.N = len(i.Elems)
+	return done
+}
+
+// nextNoNorm is the same as next, but does not "normalize" the collation
+// elements.
+func (i *Iter) nextNoNorm() bool {
+	// TODO: remove this function. Using this instead of next does not seem
+	// to improve performance in any significant way. We retain this until
+	// later for evaluation purposes.
+	if i.done() {
+		return false
+	}
+	i.appendNext()
+	i.N = len(i.Elems)
+	return true
+}
+
+const maxCombiningCharacters = 30
+
+// doNorm reorders the collation elements in i.Elems.
+// It assumes that blocks of collation elements added with appendNext
+// either start and end with the same CCC or start with CCC == 0.
+// This allows for a single insertion point for the entire block.
+// The correctness of this assumption is verified in builder.go.
+func (i *Iter) doNorm(p int, ccc uint8) {
+	n := len(i.Elems)
+	k := p
+	for p--; p > i.N && ccc < i.Elems[p-1].CCC(); p-- {
+	}
+	i.Elems = append(i.Elems, i.Elems[p:k]...)
+	copy(i.Elems[p:], i.Elems[k:])
+	i.Elems = i.Elems[:n]
+}

vendor/golang.org/x/text/internal/colltab/numeric.go

@@ -0,0 +1,236 @@
+// Copyright 2014 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 colltab
+
+import (
+	"unicode"
+	"unicode/utf8"
+)
+
+// NewNumericWeighter wraps w to replace individual digits to sort based on their
+// numeric value.
+//
+// Weighter w must have a free primary weight after the primary weight for 9.
+// If this is not the case, numeric value will sort at the same primary level
+// as the first primary sorting after 9.
+func NewNumericWeighter(w Weighter) Weighter {
+	getElem := func(s string) Elem {
+		elems, _ := w.AppendNextString(nil, s)
+		return elems[0]
+	}
+	nine := getElem("9")
+
+	// Numbers should order before zero, but the DUCET has no room for this.
+	// TODO: move before zero once we use fractional collation elements.
+	ns, _ := MakeElem(nine.Primary()+1, nine.Secondary(), int(nine.Tertiary()), 0)
+
+	return &numericWeighter{
+		Weighter: w,
+
+		// We assume that w sorts digits of different kinds in order of numeric
+		// value and that the tertiary weight order is preserved.
+		//
+		// TODO: evaluate whether it is worth basing the ranges on the Elem
+		// encoding itself once the move to fractional weights is complete.
+		zero:          getElem("0"),
+		zeroSpecialLo: getElem("０"), // U+FF10 FULLWIDTH DIGIT ZERO
+		zeroSpecialHi: getElem("₀"), // U+2080 SUBSCRIPT ZERO
+		nine:          nine,
+		nineSpecialHi: getElem("₉"), // U+2089 SUBSCRIPT NINE
+		numberStart:   ns,
+	}
+}
+
+// A numericWeighter translates a stream of digits into a stream of weights
+// representing the numeric value.
+type numericWeighter struct {
+	Weighter
+
+	// The Elems below all demarcate boundaries of specific ranges. With the
+	// current element encoding digits are in two ranges: normal (default
+	// tertiary value) and special. For most languages, digits have collation
+	// elements in the normal range.
+	//
+	// Note: the range tests are very specific for the element encoding used by
+	// this implementation. The tests in collate_test.go are designed to fail
+	// if this code is not updated when an encoding has changed.
+
+	zero          Elem // normal digit zero
+	zeroSpecialLo Elem // special digit zero, low tertiary value
+	zeroSpecialHi Elem // special digit zero, high tertiary value
+	nine          Elem // normal digit nine
+	nineSpecialHi Elem // special digit nine
+	numberStart   Elem
+}
+
+// AppendNext calls the namesake of the underlying weigher, but replaces single
+// digits with weights representing their value.
+func (nw *numericWeighter) AppendNext(buf []Elem, s []byte) (ce []Elem, n int) {
+	ce, n = nw.Weighter.AppendNext(buf, s)
+	nc := numberConverter{
+		elems: buf,
+		w:     nw,
+		b:     s,
+	}
+	isZero, ok := nc.checkNextDigit(ce)
+	if !ok {
+		return ce, n
+	}
+	// ce might have been grown already, so take it instead of buf.
+	nc.init(ce, len(buf), isZero)
+	for n < len(s) {
+		ce, sz := nw.Weighter.AppendNext(nc.elems, s[n:])
+		nc.b = s
+		n += sz
+		if !nc.update(ce) {
+			break
+		}
+	}
+	return nc.result(), n
+}
+
+// AppendNextString calls the namesake of the underlying weigher, but replaces
+// single digits with weights representing their value.
+func (nw *numericWeighter) AppendNextString(buf []Elem, s string) (ce []Elem, n int) {
+	ce, n = nw.Weighter.AppendNextString(buf, s)
+	nc := numberConverter{
+		elems: buf,
+		w:     nw,
+		s:     s,
+	}
+	isZero, ok := nc.checkNextDigit(ce)
+	if !ok {
+		return ce, n
+	}
+	nc.init(ce, len(buf), isZero)
+	for n < len(s) {
+		ce, sz := nw.Weighter.AppendNextString(nc.elems, s[n:])
+		nc.s = s
+		n += sz
+		if !nc.update(ce) {
+			break
+		}
+	}
+	return nc.result(), n
+}
+
+type numberConverter struct {
+	w *numericWeighter
+
+	elems    []Elem
+	nDigits  int
+	lenIndex int
+
+	s string // set if the input was of type string
+	b []byte // set if the input was of type []byte
+}
+
+// init completes initialization of a numberConverter and prepares it for adding
+// more digits. elems is assumed to have a digit starting at oldLen.
+func (nc *numberConverter) init(elems []Elem, oldLen int, isZero bool) {
+	// Insert a marker indicating the start of a number and a placeholder
+	// for the number of digits.
+	if isZero {
+		elems = append(elems[:oldLen], nc.w.numberStart, 0)
+	} else {
+		elems = append(elems, 0, 0)
+		copy(elems[oldLen+2:], elems[oldLen:])
+		elems[oldLen] = nc.w.numberStart
+		elems[oldLen+1] = 0
+
+		nc.nDigits = 1
+	}
+	nc.elems = elems
+	nc.lenIndex = oldLen + 1
+}
+
+// checkNextDigit reports whether bufNew adds a single digit relative to the old
+// buffer. If it does, it also reports whether this digit is zero.
+func (nc *numberConverter) checkNextDigit(bufNew []Elem) (isZero, ok bool) {
+	if len(nc.elems) >= len(bufNew) {
+		return false, false
+	}
+	e := bufNew[len(nc.elems)]
+	if e < nc.w.zeroSpecialLo || nc.w.nine < e {
+		// Not a number.
+		return false, false
+	}
+	if e < nc.w.zero {
+		if e > nc.w.nineSpecialHi {
+			// Not a number.
+			return false, false
+		}
+		if !nc.isDigit() {
+			return false, false
+		}
+		isZero = e <= nc.w.zeroSpecialHi
+	} else {
+		// This is the common case if we encounter a digit.
+		isZero = e == nc.w.zero
+	}
+	// Test the remaining added collation elements have a zero primary value.
+	if n := len(bufNew) - len(nc.elems); n > 1 {
+		for i := len(nc.elems) + 1; i < len(bufNew); i++ {
+			if bufNew[i].Primary() != 0 {
+				return false, false
+			}
+		}
+		// In some rare cases, collation elements will encode runes in
+		// unicode.No as a digit. For example Ethiopic digits (U+1369 - U+1371)
+		// are not in Nd. Also some digits that clearly belong in unicode.No,
+		// like U+0C78 TELUGU FRACTION DIGIT ZERO FOR ODD POWERS OF FOUR, have
+		// collation elements indistinguishable from normal digits.
+		// Unfortunately, this means we need to make this check for nearly all
+		// non-Latin digits.
+		//
+		// TODO: check the performance impact and find something better if it is
+		// an issue.
+		if !nc.isDigit() {
+			return false, false
+		}
+	}
+	return isZero, true
+}
+
+func (nc *numberConverter) isDigit() bool {
+	if nc.b != nil {
+		r, _ := utf8.DecodeRune(nc.b)
+		return unicode.In(r, unicode.Nd)
+	}
+	r, _ := utf8.DecodeRuneInString(nc.s)
+	return unicode.In(r, unicode.Nd)
+}
+
+// We currently support a maximum of about 2M digits (the number of primary
+// values). Such numbers will compare correctly against small numbers, but their
+// comparison against other large numbers is undefined.
+//
+// TODO: define a proper fallback, such as comparing large numbers textually or
+// actually allowing numbers of unlimited length.
+//
+// TODO: cap this to a lower number (like 100) and maybe allow a larger number
+// in an option?
+const maxDigits = 1<<maxPrimaryBits - 1
+
+func (nc *numberConverter) update(elems []Elem) bool {
+	isZero, ok := nc.checkNextDigit(elems)
+	if nc.nDigits == 0 && isZero {
+		return true
+	}
+	nc.elems = elems
+	if !ok {
+		return false
+	}
+	nc.nDigits++
+	return nc.nDigits < maxDigits
+}
+
+// result fills in the length element for the digit sequence and returns the
+// completed collation elements.
+func (nc *numberConverter) result() []Elem {
+	e, _ := MakeElem(nc.nDigits, defaultSecondary, defaultTertiary, 0)
+	nc.elems[nc.lenIndex] = e
+	return nc.elems
+}

vendor/golang.org/x/text/internal/colltab/table.go

@@ -0,0 +1,275 @@
+// Copyright 2012 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 colltab
+
+import (
+	"unicode/utf8"
+
+	"golang.org/x/text/unicode/norm"
+)
+
+// Table holds all collation data for a given collation ordering.
+type Table struct {
+	Index Trie // main trie
+
+	// expansion info
+	ExpandElem []uint32
+
+	// contraction info
+	ContractTries  ContractTrieSet
+	ContractElem   []uint32
+	MaxContractLen int
+	VariableTop    uint32
+}
+
+func (t *Table) AppendNext(w []Elem, b []byte) (res []Elem, n int) {
+	return t.appendNext(w, source{bytes: b})
+}
+
+func (t *Table) AppendNextString(w []Elem, s string) (res []Elem, n int) {
+	return t.appendNext(w, source{str: s})
+}
+
+func (t *Table) Start(p int, b []byte) int {
+	// TODO: implement
+	panic("not implemented")
+}
+
+func (t *Table) StartString(p int, s string) int {
+	// TODO: implement
+	panic("not implemented")
+}
+
+func (t *Table) Domain() []string {
+	// TODO: implement
+	panic("not implemented")
+}
+
+func (t *Table) Top() uint32 {
+	return t.VariableTop
+}
+
+type source struct {
+	str   string
+	bytes []byte
+}
+
+func (src *source) lookup(t *Table) (ce Elem, sz int) {
+	if src.bytes == nil {
+		return t.Index.lookupString(src.str)
+	}
+	return t.Index.lookup(src.bytes)
+}
+
+func (src *source) tail(sz int) {
+	if src.bytes == nil {
+		src.str = src.str[sz:]
+	} else {
+		src.bytes = src.bytes[sz:]
+	}
+}
+
+func (src *source) nfd(buf []byte, end int) []byte {
+	if src.bytes == nil {
+		return norm.NFD.AppendString(buf[:0], src.str[:end])
+	}
+	return norm.NFD.Append(buf[:0], src.bytes[:end]...)
+}
+
+func (src *source) rune() (r rune, sz int) {
+	if src.bytes == nil {
+		return utf8.DecodeRuneInString(src.str)
+	}
+	return utf8.DecodeRune(src.bytes)
+}
+
+func (src *source) properties(f norm.Form) norm.Properties {
+	if src.bytes == nil {
+		return f.PropertiesString(src.str)
+	}
+	return f.Properties(src.bytes)
+}
+
+// appendNext appends the weights corresponding to the next rune or
+// contraction in s.  If a contraction is matched to a discontinuous
+// sequence of runes, the weights for the interstitial runes are
+// appended as well.  It returns a new slice that includes the appended
+// weights and the number of bytes consumed from s.
+func (t *Table) appendNext(w []Elem, src source) (res []Elem, n int) {
+	ce, sz := src.lookup(t)
+	tp := ce.ctype()
+	if tp == ceNormal {
+		if ce == 0 {
+			r, _ := src.rune()
+			const (
+				hangulSize  = 3
+				firstHangul = 0xAC00
+				lastHangul  = 0xD7A3
+			)
+			if r >= firstHangul && r <= lastHangul {
+				// TODO: performance can be considerably improved here.
+				n = sz
+				var buf [16]byte // Used for decomposing Hangul.
+				for b := src.nfd(buf[:0], hangulSize); len(b) > 0; b = b[sz:] {
+					ce, sz = t.Index.lookup(b)
+					w = append(w, ce)
+				}
+				return w, n
+			}
+			ce = makeImplicitCE(implicitPrimary(r))
+		}
+		w = append(w, ce)
+	} else if tp == ceExpansionIndex {
+		w = t.appendExpansion(w, ce)
+	} else if tp == ceContractionIndex {
+		n := 0
+		src.tail(sz)
+		if src.bytes == nil {
+			w, n = t.matchContractionString(w, ce, src.str)
+		} else {
+			w, n = t.matchContraction(w, ce, src.bytes)
+		}
+		sz += n
+	} else if tp == ceDecompose {
+		// Decompose using NFKD and replace tertiary weights.
+		t1, t2 := splitDecompose(ce)
+		i := len(w)
+		nfkd := src.properties(norm.NFKD).Decomposition()
+		for p := 0; len(nfkd) > 0; nfkd = nfkd[p:] {
+			w, p = t.appendNext(w, source{bytes: nfkd})
+		}
+		w[i] = w[i].updateTertiary(t1)
+		if i++; i < len(w) {
+			w[i] = w[i].updateTertiary(t2)
+			for i++; i < len(w); i++ {
+				w[i] = w[i].updateTertiary(maxTertiary)
+			}
+		}
+	}
+	return w, sz
+}
+
+func (t *Table) appendExpansion(w []Elem, ce Elem) []Elem {
+	i := splitExpandIndex(ce)
+	n := int(t.ExpandElem[i])
+	i++
+	for _, ce := range t.ExpandElem[i : i+n] {
+		w = append(w, Elem(ce))
+	}
+	return w
+}
+
+func (t *Table) matchContraction(w []Elem, ce Elem, suffix []byte) ([]Elem, int) {
+	index, n, offset := splitContractIndex(ce)
+
+	scan := t.ContractTries.scanner(index, n, suffix)
+	buf := [norm.MaxSegmentSize]byte{}
+	bufp := 0
+	p := scan.scan(0)
+
+	if !scan.done && p < len(suffix) && suffix[p] >= utf8.RuneSelf {
+		// By now we should have filtered most cases.
+		p0 := p
+		bufn := 0
+		rune := norm.NFD.Properties(suffix[p:])
+		p += rune.Size()
+		if rune.LeadCCC() != 0 {
+			prevCC := rune.TrailCCC()
+			// A gap may only occur in the last normalization segment.
+			// This also ensures that len(scan.s) < norm.MaxSegmentSize.
+			if end := norm.NFD.FirstBoundary(suffix[p:]); end != -1 {
+				scan.s = suffix[:p+end]
+			}
+			for p < len(suffix) && !scan.done && suffix[p] >= utf8.RuneSelf {
+				rune = norm.NFD.Properties(suffix[p:])
+				if ccc := rune.LeadCCC(); ccc == 0 || prevCC >= ccc {
+					break
+				}
+				prevCC = rune.TrailCCC()
+				if pp := scan.scan(p); pp != p {
+					// Copy the interstitial runes for later processing.
+					bufn += copy(buf[bufn:], suffix[p0:p])
+					if scan.pindex == pp {
+						bufp = bufn
+					}
+					p, p0 = pp, pp
+				} else {
+					p += rune.Size()
+				}
+			}
+		}
+	}
+	// Append weights for the matched contraction, which may be an expansion.
+	i, n := scan.result()
+	ce = Elem(t.ContractElem[i+offset])
+	if ce.ctype() == ceNormal {
+		w = append(w, ce)
+	} else {
+		w = t.appendExpansion(w, ce)
+	}
+	// Append weights for the runes in the segment not part of the contraction.
+	for b, p := buf[:bufp], 0; len(b) > 0; b = b[p:] {
+		w, p = t.appendNext(w, source{bytes: b})
+	}
+	return w, n
+}
+
+// TODO: unify the two implementations. This is best done after first simplifying
+// the algorithm taking into account the inclusion of both NFC and NFD forms
+// in the table.
+func (t *Table) matchContractionString(w []Elem, ce Elem, suffix string) ([]Elem, int) {
+	index, n, offset := splitContractIndex(ce)
+
+	scan := t.ContractTries.scannerString(index, n, suffix)
+	buf := [norm.MaxSegmentSize]byte{}
+	bufp := 0
+	p := scan.scan(0)
+
+	if !scan.done && p < len(suffix) && suffix[p] >= utf8.RuneSelf {
+		// By now we should have filtered most cases.
+		p0 := p
+		bufn := 0
+		rune := norm.NFD.PropertiesString(suffix[p:])
+		p += rune.Size()
+		if rune.LeadCCC() != 0 {
+			prevCC := rune.TrailCCC()
+			// A gap may only occur in the last normalization segment.
+			// This also ensures that len(scan.s) < norm.MaxSegmentSize.
+			if end := norm.NFD.FirstBoundaryInString(suffix[p:]); end != -1 {
+				scan.s = suffix[:p+end]
+			}
+			for p < len(suffix) && !scan.done && suffix[p] >= utf8.RuneSelf {
+				rune = norm.NFD.PropertiesString(suffix[p:])
+				if ccc := rune.LeadCCC(); ccc == 0 || prevCC >= ccc {
+					break
+				}
+				prevCC = rune.TrailCCC()
+				if pp := scan.scan(p); pp != p {
+					// Copy the interstitial runes for later processing.
+					bufn += copy(buf[bufn:], suffix[p0:p])
+					if scan.pindex == pp {
+						bufp = bufn
+					}
+					p, p0 = pp, pp
+				} else {
+					p += rune.Size()
+				}
+			}
+		}
+	}
+	// Append weights for the matched contraction, which may be an expansion.
+	i, n := scan.result()
+	ce = Elem(t.ContractElem[i+offset])
+	if ce.ctype() == ceNormal {
+		w = append(w, ce)
+	} else {
+		w = t.appendExpansion(w, ce)
+	}
+	// Append weights for the runes in the segment not part of the contraction.
+	for b, p := buf[:bufp], 0; len(b) > 0; b = b[p:] {
+		w, p = t.appendNext(w, source{bytes: b})
+	}
+	return w, n
+}

vendor/golang.org/x/text/internal/colltab/trie.go

@@ -0,0 +1,159 @@
+// Copyright 2012 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.
+
+// The trie in this file is used to associate the first full character in an
+// UTF-8 string to a collation element. All but the last byte in a UTF-8 byte
+// sequence are used to lookup offsets in the index table to be used for the
+// next byte. The last byte is used to index into a table of collation elements.
+// For a full description, see go.text/collate/build/trie.go.
+
+package colltab
+
+const blockSize = 64
+
+type Trie struct {
+	Index0  []uint16 // index for first byte (0xC0-0xFF)
+	Values0 []uint32 // index for first byte (0x00-0x7F)
+	Index   []uint16
+	Values  []uint32
+}
+
+const (
+	t1 = 0x00 // 0000 0000
+	tx = 0x80 // 1000 0000
+	t2 = 0xC0 // 1100 0000
+	t3 = 0xE0 // 1110 0000
+	t4 = 0xF0 // 1111 0000
+	t5 = 0xF8 // 1111 1000
+	t6 = 0xFC // 1111 1100
+	te = 0xFE // 1111 1110
+)
+
+func (t *Trie) lookupValue(n uint16, b byte) Elem {
+	return Elem(t.Values[int(n)<<6+int(b)])
+}
+
+// lookup returns the trie value for the first UTF-8 encoding in s and
+// the width in bytes of this encoding. The size will be 0 if s does not
+// hold enough bytes to complete the encoding. len(s) must be greater than 0.
+func (t *Trie) lookup(s []byte) (v Elem, sz int) {
+	c0 := s[0]
+	switch {
+	case c0 < tx:
+		return Elem(t.Values0[c0]), 1
+	case c0 < t2:
+		return 0, 1
+	case c0 < t3:
+		if len(s) < 2 {
+			return 0, 0
+		}
+		i := t.Index0[c0]
+		c1 := s[1]
+		if c1 < tx || t2 <= c1 {
+			return 0, 1
+		}
+		return t.lookupValue(i, c1), 2
+	case c0 < t4:
+		if len(s) < 3 {
+			return 0, 0
+		}
+		i := t.Index0[c0]
+		c1 := s[1]
+		if c1 < tx || t2 <= c1 {
+			return 0, 1
+		}
+		o := int(i)<<6 + int(c1)
+		i = t.Index[o]
+		c2 := s[2]
+		if c2 < tx || t2 <= c2 {
+			return 0, 2
+		}
+		return t.lookupValue(i, c2), 3
+	case c0 < t5:
+		if len(s) < 4 {
+			return 0, 0
+		}
+		i := t.Index0[c0]
+		c1 := s[1]
+		if c1 < tx || t2 <= c1 {
+			return 0, 1
+		}
+		o := int(i)<<6 + int(c1)
+		i = t.Index[o]
+		c2 := s[2]
+		if c2 < tx || t2 <= c2 {
+			return 0, 2
+		}
+		o = int(i)<<6 + int(c2)
+		i = t.Index[o]
+		c3 := s[3]
+		if c3 < tx || t2 <= c3 {
+			return 0, 3
+		}
+		return t.lookupValue(i, c3), 4
+	}
+	// Illegal rune
+	return 0, 1
+}
+
+// The body of lookupString is a verbatim copy of that of lookup.
+func (t *Trie) lookupString(s string) (v Elem, sz int) {
+	c0 := s[0]
+	switch {
+	case c0 < tx:
+		return Elem(t.Values0[c0]), 1
+	case c0 < t2:
+		return 0, 1
+	case c0 < t3:
+		if len(s) < 2 {
+			return 0, 0
+		}
+		i := t.Index0[c0]
+		c1 := s[1]
+		if c1 < tx || t2 <= c1 {
+			return 0, 1
+		}
+		return t.lookupValue(i, c1), 2
+	case c0 < t4:
+		if len(s) < 3 {
+			return 0, 0
+		}
+		i := t.Index0[c0]
+		c1 := s[1]
+		if c1 < tx || t2 <= c1 {
+			return 0, 1
+		}
+		o := int(i)<<6 + int(c1)
+		i = t.Index[o]
+		c2 := s[2]
+		if c2 < tx || t2 <= c2 {
+			return 0, 2
+		}
+		return t.lookupValue(i, c2), 3
+	case c0 < t5:
+		if len(s) < 4 {
+			return 0, 0
+		}
+		i := t.Index0[c0]
+		c1 := s[1]
+		if c1 < tx || t2 <= c1 {
+			return 0, 1
+		}
+		o := int(i)<<6 + int(c1)
+		i = t.Index[o]
+		c2 := s[2]
+		if c2 < tx || t2 <= c2 {
+			return 0, 2
+		}
+		o = int(i)<<6 + int(c2)
+		i = t.Index[o]
+		c3 := s[3]
+		if c3 < tx || t2 <= c3 {
+			return 0, 3
+		}
+		return t.lookupValue(i, c3), 4
+	}
+	// Illegal rune
+	return 0, 1
+}

vendor/golang.org/x/text/internal/colltab/weighter.go

@@ -0,0 +1,31 @@
+// 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 colltab // import "golang.org/x/text/internal/colltab"
+
+// A Weighter can be used as a source for Collator and Searcher.
+type Weighter interface {
+	// Start finds the start of the segment that includes position p.
+	Start(p int, b []byte) int
+
+	// StartString finds the start of the segment that includes position p.
+	StartString(p int, s string) int
+
+	// AppendNext appends Elems to buf corresponding to the longest match
+	// of a single character or contraction from the start of s.
+	// It returns the new buf and the number of bytes consumed.
+	AppendNext(buf []Elem, s []byte) (ce []Elem, n int)
+
+	// AppendNextString appends Elems to buf corresponding to the longest match
+	// of a single character or contraction from the start of s.
+	// It returns the new buf and the number of bytes consumed.
+	AppendNextString(buf []Elem, s string) (ce []Elem, n int)
+
+	// Domain returns a slice of all single characters and contractions for which
+	// collation elements are defined in this table.
+	Domain() []string
+
+	// Top returns the highest variable primary value.
+	Top() uint32
+}

vendor/golang.org/x/text/internal/gen/code.go

@@ -0,0 +1,375 @@
+// Copyright 2015 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 gen
+
+import (
+	"bytes"
+	"encoding/gob"
+	"fmt"
+	"hash"
+	"hash/fnv"
+	"io"
+	"log"
+	"os"
+	"reflect"
+	"strings"
+	"unicode"
+	"unicode/utf8"
+)
+
+// This file contains utilities for generating code.
+
+// TODO: other write methods like:
+// - slices, maps, types, etc.
+
+// CodeWriter is a utility for writing structured code. It computes the content
+// hash and size of written content. It ensures there are newlines between
+// written code blocks.
+type CodeWriter struct {
+	buf  bytes.Buffer
+	Size int
+	Hash hash.Hash32 // content hash
+	gob  *gob.Encoder
+	// For comments we skip the usual one-line separator if they are followed by
+	// a code block.
+	skipSep bool
+}
+
+func (w *CodeWriter) Write(p []byte) (n int, err error) {
+	return w.buf.Write(p)
+}
+
+// NewCodeWriter returns a new CodeWriter.
+func NewCodeWriter() *CodeWriter {
+	h := fnv.New32()
+	return &CodeWriter{Hash: h, gob: gob.NewEncoder(h)}
+}
+
+// WriteGoFile appends the buffer with the total size of all created structures
+// and writes it as a Go file to the given file with the given package name.
+func (w *CodeWriter) WriteGoFile(filename, pkg string) {
+	f, err := os.Create(filename)
+	if err != nil {
+		log.Fatalf("Could not create file %s: %v", filename, err)
+	}
+	defer f.Close()
+	if _, err = w.WriteGo(f, pkg, ""); err != nil {
+		log.Fatalf("Error writing file %s: %v", filename, err)
+	}
+}
+
+// WriteVersionedGoFile appends the buffer with the total size of all created
+// structures and writes it as a Go file to the given file with the given
+// package name and build tags for the current Unicode version,
+func (w *CodeWriter) WriteVersionedGoFile(filename, pkg string) {
+	tags := buildTags()
+	if tags != "" {
+		pattern := fileToPattern(filename)
+		updateBuildTags(pattern)
+		filename = fmt.Sprintf(pattern, UnicodeVersion())
+	}
+	f, err := os.Create(filename)
+	if err != nil {
+		log.Fatalf("Could not create file %s: %v", filename, err)
+	}
+	defer f.Close()
+	if _, err = w.WriteGo(f, pkg, tags); err != nil {
+		log.Fatalf("Error writing file %s: %v", filename, err)
+	}
+}
+
+// WriteGo appends the buffer with the total size of all created structures and
+// writes it as a Go file to the given writer with the given package name.
+func (w *CodeWriter) WriteGo(out io.Writer, pkg, tags string) (n int, err error) {
+	sz := w.Size
+	if sz > 0 {
+		w.WriteComment("Total table size %d bytes (%dKiB); checksum: %X\n", sz, sz/1024, w.Hash.Sum32())
+	}
+	defer w.buf.Reset()
+	return WriteGo(out, pkg, tags, w.buf.Bytes())
+}
+
+func (w *CodeWriter) printf(f string, x ...interface{}) {
+	fmt.Fprintf(w, f, x...)
+}
+
+func (w *CodeWriter) insertSep() {
+	if w.skipSep {
+		w.skipSep = false
+		return
+	}
+	// Use at least two newlines to ensure a blank space between the previous
+	// block. WriteGoFile will remove extraneous newlines.
+	w.printf("\n\n")
+}
+
+// WriteComment writes a comment block. All line starts are prefixed with "//".
+// Initial empty lines are gobbled. The indentation for the first line is
+// stripped from consecutive lines.
+func (w *CodeWriter) WriteComment(comment string, args ...interface{}) {
+	s := fmt.Sprintf(comment, args...)
+	s = strings.Trim(s, "\n")
+
+	// Use at least two newlines to ensure a blank space between the previous
+	// block. WriteGoFile will remove extraneous newlines.
+	w.printf("\n\n// ")
+	w.skipSep = true
+
+	// strip first indent level.
+	sep := "\n"
+	for ; len(s) > 0 && (s[0] == '\t' || s[0] == ' '); s = s[1:] {
+		sep += s[:1]
+	}
+
+	strings.NewReplacer(sep, "\n// ", "\n", "\n// ").WriteString(w, s)
+
+	w.printf("\n")
+}
+
+func (w *CodeWriter) writeSizeInfo(size int) {
+	w.printf("// Size: %d bytes\n", size)
+}
+
+// WriteConst writes a constant of the given name and value.
+func (w *CodeWriter) WriteConst(name string, x interface{}) {
+	w.insertSep()
+	v := reflect.ValueOf(x)
+
+	switch v.Type().Kind() {
+	case reflect.String:
+		w.printf("const %s %s = ", name, typeName(x))
+		w.WriteString(v.String())
+		w.printf("\n")
+	default:
+		w.printf("const %s = %#v\n", name, x)
+	}
+}
+
+// WriteVar writes a variable of the given name and value.
+func (w *CodeWriter) WriteVar(name string, x interface{}) {
+	w.insertSep()
+	v := reflect.ValueOf(x)
+	oldSize := w.Size
+	sz := int(v.Type().Size())
+	w.Size += sz
+
+	switch v.Type().Kind() {
+	case reflect.String:
+		w.printf("var %s %s = ", name, typeName(x))
+		w.WriteString(v.String())
+	case reflect.Struct:
+		w.gob.Encode(x)
+		fallthrough
+	case reflect.Slice, reflect.Array:
+		w.printf("var %s = ", name)
+		w.writeValue(v)
+		w.writeSizeInfo(w.Size - oldSize)
+	default:
+		w.printf("var %s %s = ", name, typeName(x))
+		w.gob.Encode(x)
+		w.writeValue(v)
+		w.writeSizeInfo(w.Size - oldSize)
+	}
+	w.printf("\n")
+}
+
+func (w *CodeWriter) writeValue(v reflect.Value) {
+	x := v.Interface()
+	switch v.Kind() {
+	case reflect.String:
+		w.WriteString(v.String())
+	case reflect.Array:
+		// Don't double count: callers of WriteArray count on the size being
+		// added, so we need to discount it here.
+		w.Size -= int(v.Type().Size())
+		w.writeSlice(x, true)
+	case reflect.Slice:
+		w.writeSlice(x, false)
+	case reflect.Struct:
+		w.printf("%s{\n", typeName(v.Interface()))
+		t := v.Type()
+		for i := 0; i < v.NumField(); i++ {
+			w.printf("%s: ", t.Field(i).Name)
+			w.writeValue(v.Field(i))
+			w.printf(",\n")
+		}
+		w.printf("}")
+	default:
+		w.printf("%#v", x)
+	}
+}
+
+// WriteString writes a string literal.
+func (w *CodeWriter) WriteString(s string) {
+	io.WriteString(w.Hash, s) // content hash
+	w.Size += len(s)
+
+	const maxInline = 40
+	if len(s) <= maxInline {
+		w.printf("%q", s)
+		return
+	}
+
+	// We will render the string as a multi-line string.
+	const maxWidth = 80 - 4 - len(`"`) - len(`" +`)
+
+	// When starting on its own line, go fmt indents line 2+ an extra level.
+	n, max := maxWidth, maxWidth-4
+
+	// As per https://golang.org/issue/18078, the compiler has trouble
+	// compiling the concatenation of many strings, s0 + s1 + s2 + ... + sN,
+	// for large N. We insert redundant, explicit parentheses to work around
+	// that, lowering the N at any given step: (s0 + s1 + ... + s63) + (s64 +
+	// ... + s127) + etc + (etc + ... + sN).
+	explicitParens, extraComment := len(s) > 128*1024, ""
+	if explicitParens {
+		w.printf(`(`)
+		extraComment = "; the redundant, explicit parens are for https://golang.org/issue/18078"
+	}
+
+	// Print "" +\n, if a string does not start on its own line.
+	b := w.buf.Bytes()
+	if p := len(bytes.TrimRight(b, " \t")); p > 0 && b[p-1] != '\n' {
+		w.printf("\"\" + // Size: %d bytes%s\n", len(s), extraComment)
+		n, max = maxWidth, maxWidth
+	}
+
+	w.printf(`"`)
+
+	for sz, p, nLines := 0, 0, 0; p < len(s); {
+		var r rune
+		r, sz = utf8.DecodeRuneInString(s[p:])
+		out := s[p : p+sz]
+		chars := 1
+		if !unicode.IsPrint(r) || r == utf8.RuneError || r == '"' {
+			switch sz {
+			case 1:
+				out = fmt.Sprintf("\\x%02x", s[p])
+			case 2, 3:
+				out = fmt.Sprintf("\\u%04x", r)
+			case 4:
+				out = fmt.Sprintf("\\U%08x", r)
+			}
+			chars = len(out)
+		} else if r == '\\' {
+			out = "\\" + string(r)
+			chars = 2
+		}
+		if n -= chars; n < 0 {
+			nLines++
+			if explicitParens && nLines&63 == 63 {
+				w.printf("\") + (\"")
+			}
+			w.printf("\" +\n\"")
+			n = max - len(out)
+		}
+		w.printf("%s", out)
+		p += sz
+	}
+	w.printf(`"`)
+	if explicitParens {
+		w.printf(`)`)
+	}
+}
+
+// WriteSlice writes a slice value.
+func (w *CodeWriter) WriteSlice(x interface{}) {
+	w.writeSlice(x, false)
+}
+
+// WriteArray writes an array value.
+func (w *CodeWriter) WriteArray(x interface{}) {
+	w.writeSlice(x, true)
+}
+
+func (w *CodeWriter) writeSlice(x interface{}, isArray bool) {
+	v := reflect.ValueOf(x)
+	w.gob.Encode(v.Len())
+	w.Size += v.Len() * int(v.Type().Elem().Size())
+	name := typeName(x)
+	if isArray {
+		name = fmt.Sprintf("[%d]%s", v.Len(), name[strings.Index(name, "]")+1:])
+	}
+	if isArray {
+		w.printf("%s{\n", name)
+	} else {
+		w.printf("%s{ // %d elements\n", name, v.Len())
+	}
+
+	switch kind := v.Type().Elem().Kind(); kind {
+	case reflect.String:
+		for _, s := range x.([]string) {
+			w.WriteString(s)
+			w.printf(",\n")
+		}
+	case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64,
+		reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64:
+		// nLine and nBlock are the number of elements per line and block.
+		nLine, nBlock, format := 8, 64, "%d,"
+		switch kind {
+		case reflect.Uint8:
+			format = "%#02x,"
+		case reflect.Uint16:
+			format = "%#04x,"
+		case reflect.Uint32:
+			nLine, nBlock, format = 4, 32, "%#08x,"
+		case reflect.Uint, reflect.Uint64:
+			nLine, nBlock, format = 4, 32, "%#016x,"
+		case reflect.Int8:
+			nLine = 16
+		}
+		n := nLine
+		for i := 0; i < v.Len(); i++ {
+			if i%nBlock == 0 && v.Len() > nBlock {
+				w.printf("// Entry %X - %X\n", i, i+nBlock-1)
+			}
+			x := v.Index(i).Interface()
+			w.gob.Encode(x)
+			w.printf(format, x)
+			if n--; n == 0 {
+				n = nLine
+				w.printf("\n")
+			}
+		}
+		w.printf("\n")
+	case reflect.Struct:
+		zero := reflect.Zero(v.Type().Elem()).Interface()
+		for i := 0; i < v.Len(); i++ {
+			x := v.Index(i).Interface()
+			w.gob.EncodeValue(v)
+			if !reflect.DeepEqual(zero, x) {
+				line := fmt.Sprintf("%#v,\n", x)
+				line = line[strings.IndexByte(line, '{'):]
+				w.printf("%d: ", i)
+				w.printf(line)
+			}
+		}
+	case reflect.Array:
+		for i := 0; i < v.Len(); i++ {
+			w.printf("%d: %#v,\n", i, v.Index(i).Interface())
+		}
+	default:
+		panic("gen: slice elem type not supported")
+	}
+	w.printf("}")
+}
+
+// WriteType writes a definition of the type of the given value and returns the
+// type name.
+func (w *CodeWriter) WriteType(x interface{}) string {
+	t := reflect.TypeOf(x)
+	w.printf("type %s struct {\n", t.Name())
+	for i := 0; i < t.NumField(); i++ {
+		w.printf("\t%s %s\n", t.Field(i).Name, t.Field(i).Type)
+	}
+	w.printf("}\n")
+	return t.Name()
+}
+
+// typeName returns the name of the go type of x.
+func typeName(x interface{}) string {
+	t := reflect.ValueOf(x).Type()
+	return strings.Replace(fmt.Sprint(t), "main.", "", 1)
+}

vendor/golang.org/x/text/internal/gen/gen.go

@@ -0,0 +1,347 @@
+// Copyright 2015 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 gen contains common code for the various code generation tools in the
+// text repository. Its usage ensures consistency between tools.
+//
+// This package defines command line flags that are common to most generation
+// tools. The flags allow for specifying specific Unicode and CLDR versions
+// in the public Unicode data repository (https://www.unicode.org/Public).
+//
+// A local Unicode data mirror can be set through the flag -local or the
+// environment variable UNICODE_DIR. The former takes precedence. The local
+// directory should follow the same structure as the public repository.
+//
+// IANA data can also optionally be mirrored by putting it in the iana directory
+// rooted at the top of the local mirror. Beware, though, that IANA data is not
+// versioned. So it is up to the developer to use the right version.
+package gen // import "golang.org/x/text/internal/gen"
+
+import (
+	"bytes"
+	"flag"
+	"fmt"
+	"go/build"
+	"go/format"
+	"io"
+	"io/ioutil"
+	"log"
+	"net/http"
+	"os"
+	"path"
+	"path/filepath"
+	"regexp"
+	"strings"
+	"sync"
+	"unicode"
+
+	"golang.org/x/text/unicode/cldr"
+)
+
+var (
+	url = flag.String("url",
+		"https://www.unicode.org/Public",
+		"URL of Unicode database directory")
+	iana = flag.String("iana",
+		"http://www.iana.org",
+		"URL of the IANA repository")
+	unicodeVersion = flag.String("unicode",
+		getEnv("UNICODE_VERSION", unicode.Version),
+		"unicode version to use")
+	cldrVersion = flag.String("cldr",
+		getEnv("CLDR_VERSION", cldr.Version),
+		"cldr version to use")
+)
+
+func getEnv(name, def string) string {
+	if v := os.Getenv(name); v != "" {
+		return v
+	}
+	return def
+}
+
+// Init performs common initialization for a gen command. It parses the flags
+// and sets up the standard logging parameters.
+func Init() {
+	log.SetPrefix("")
+	log.SetFlags(log.Lshortfile)
+	flag.Parse()
+}
+
+const header = `// Code generated by running "go generate" in golang.org/x/text. DO NOT EDIT.
+
+`
+
+// UnicodeVersion reports the requested Unicode version.
+func UnicodeVersion() string {
+	return *unicodeVersion
+}
+
+// CLDRVersion reports the requested CLDR version.
+func CLDRVersion() string {
+	return *cldrVersion
+}
+
+var tags = []struct{ version, buildTags string }{
+	{"9.0.0", "!go1.10"},
+	{"10.0.0", "go1.10,!go1.13"},
+	{"11.0.0", "go1.13"},
+}
+
+// buildTags reports the build tags used for the current Unicode version.
+func buildTags() string {
+	v := UnicodeVersion()
+	for _, e := range tags {
+		if e.version == v {
+			return e.buildTags
+		}
+	}
+	log.Fatalf("Unknown build tags for Unicode version %q.", v)
+	return ""
+}
+
+// IsLocal reports whether data files are available locally.
+func IsLocal() bool {
+	dir, err := localReadmeFile()
+	if err != nil {
+		return false
+	}
+	if _, err = os.Stat(dir); err != nil {
+		return false
+	}
+	return true
+}
+
+// OpenUCDFile opens the requested UCD file. The file is specified relative to
+// the public Unicode root directory. It will call log.Fatal if there are any
+// errors.
+func OpenUCDFile(file string) io.ReadCloser {
+	return openUnicode(path.Join(*unicodeVersion, "ucd", file))
+}
+
+// OpenCLDRCoreZip opens the CLDR core zip file. It will call log.Fatal if there
+// are any errors.
+func OpenCLDRCoreZip() io.ReadCloser {
+	return OpenUnicodeFile("cldr", *cldrVersion, "core.zip")
+}
+
+// OpenUnicodeFile opens the requested file of the requested category from the
+// root of the Unicode data archive. The file is specified relative to the
+// public Unicode root directory. If version is "", it will use the default
+// Unicode version. It will call log.Fatal if there are any errors.
+func OpenUnicodeFile(category, version, file string) io.ReadCloser {
+	if version == "" {
+		version = UnicodeVersion()
+	}
+	return openUnicode(path.Join(category, version, file))
+}
+
+// OpenIANAFile opens the requested IANA file. The file is specified relative
+// to the IANA root, which is typically either http://www.iana.org or the
+// iana directory in the local mirror. It will call log.Fatal if there are any
+// errors.
+func OpenIANAFile(path string) io.ReadCloser {
+	return Open(*iana, "iana", path)
+}
+
+var (
+	dirMutex sync.Mutex
+	localDir string
+)
+
+const permissions = 0755
+
+func localReadmeFile() (string, error) {
+	p, err := build.Import("golang.org/x/text", "", build.FindOnly)
+	if err != nil {
+		return "", fmt.Errorf("Could not locate package: %v", err)
+	}
+	return filepath.Join(p.Dir, "DATA", "README"), nil
+}
+
+func getLocalDir() string {
+	dirMutex.Lock()
+	defer dirMutex.Unlock()
+
+	readme, err := localReadmeFile()
+	if err != nil {
+		log.Fatal(err)
+	}
+	dir := filepath.Dir(readme)
+	if _, err := os.Stat(readme); err != nil {
+		if err := os.MkdirAll(dir, permissions); err != nil {
+			log.Fatalf("Could not create directory: %v", err)
+		}
+		ioutil.WriteFile(readme, []byte(readmeTxt), permissions)
+	}
+	return dir
+}
+
+const readmeTxt = `Generated by golang.org/x/text/internal/gen. DO NOT EDIT.
+
+This directory contains downloaded files used to generate the various tables
+in the golang.org/x/text subrepo.
+
+Note that the language subtag repo (iana/assignments/language-subtag-registry)
+and all other times in the iana subdirectory are not versioned and will need
+to be periodically manually updated. The easiest way to do this is to remove
+the entire iana directory. This is mostly of concern when updating the language
+package.
+`
+
+// Open opens subdir/path if a local directory is specified and the file exists,
+// where subdir is a directory relative to the local root, or fetches it from
+// urlRoot/path otherwise. It will call log.Fatal if there are any errors.
+func Open(urlRoot, subdir, path string) io.ReadCloser {
+	file := filepath.Join(getLocalDir(), subdir, filepath.FromSlash(path))
+	return open(file, urlRoot, path)
+}
+
+func openUnicode(path string) io.ReadCloser {
+	file := filepath.Join(getLocalDir(), filepath.FromSlash(path))
+	return open(file, *url, path)
+}
+
+// TODO: automatically periodically update non-versioned files.
+
+func open(file, urlRoot, path string) io.ReadCloser {
+	if f, err := os.Open(file); err == nil {
+		return f
+	}
+	r := get(urlRoot, path)
+	defer r.Close()
+	b, err := ioutil.ReadAll(r)
+	if err != nil {
+		log.Fatalf("Could not download file: %v", err)
+	}
+	os.MkdirAll(filepath.Dir(file), permissions)
+	if err := ioutil.WriteFile(file, b, permissions); err != nil {
+		log.Fatalf("Could not create file: %v", err)
+	}
+	return ioutil.NopCloser(bytes.NewReader(b))
+}
+
+func get(root, path string) io.ReadCloser {
+	url := root + "/" + path
+	fmt.Printf("Fetching %s...", url)
+	defer fmt.Println(" done.")
+	resp, err := http.Get(url)
+	if err != nil {
+		log.Fatalf("HTTP GET: %v", err)
+	}
+	if resp.StatusCode != 200 {
+		log.Fatalf("Bad GET status for %q: %q", url, resp.Status)
+	}
+	return resp.Body
+}
+
+// TODO: use Write*Version in all applicable packages.
+
+// WriteUnicodeVersion writes a constant for the Unicode version from which the
+// tables are generated.
+func WriteUnicodeVersion(w io.Writer) {
+	fmt.Fprintf(w, "// UnicodeVersion is the Unicode version from which the tables in this package are derived.\n")
+	fmt.Fprintf(w, "const UnicodeVersion = %q\n\n", UnicodeVersion())
+}
+
+// WriteCLDRVersion writes a constant for the CLDR version from which the
+// tables are generated.
+func WriteCLDRVersion(w io.Writer) {
+	fmt.Fprintf(w, "// CLDRVersion is the CLDR version from which the tables in this package are derived.\n")
+	fmt.Fprintf(w, "const CLDRVersion = %q\n\n", CLDRVersion())
+}
+
+// WriteGoFile prepends a standard file comment and package statement to the
+// given bytes, applies gofmt, and writes them to a file with the given name.
+// It will call log.Fatal if there are any errors.
+func WriteGoFile(filename, pkg string, b []byte) {
+	w, err := os.Create(filename)
+	if err != nil {
+		log.Fatalf("Could not create file %s: %v", filename, err)
+	}
+	defer w.Close()
+	if _, err = WriteGo(w, pkg, "", b); err != nil {
+		log.Fatalf("Error writing file %s: %v", filename, err)
+	}
+}
+
+func fileToPattern(filename string) string {
+	suffix := ".go"
+	if strings.HasSuffix(filename, "_test.go") {
+		suffix = "_test.go"
+	}
+	prefix := filename[:len(filename)-len(suffix)]
+	return fmt.Sprint(prefix, "%s", suffix)
+}
+
+func updateBuildTags(pattern string) {
+	for _, t := range tags {
+		oldFile := fmt.Sprintf(pattern, t.version)
+		b, err := ioutil.ReadFile(oldFile)
+		if err != nil {
+			continue
+		}
+		build := fmt.Sprintf("// +build %s", t.buildTags)
+		b = regexp.MustCompile(`// \+build .*`).ReplaceAll(b, []byte(build))
+		err = ioutil.WriteFile(oldFile, b, 0644)
+		if err != nil {
+			log.Fatal(err)
+		}
+	}
+}
+
+// WriteVersionedGoFile prepends a standard file comment, adds build tags to
+// version the file for the current Unicode version, and package statement to
+// the given bytes, applies gofmt, and writes them to a file with the given
+// name. It will call log.Fatal if there are any errors.
+func WriteVersionedGoFile(filename, pkg string, b []byte) {
+	pattern := fileToPattern(filename)
+	updateBuildTags(pattern)
+	filename = fmt.Sprintf(pattern, UnicodeVersion())
+
+	w, err := os.Create(filename)
+	if err != nil {
+		log.Fatalf("Could not create file %s: %v", filename, err)
+	}
+	defer w.Close()
+	if _, err = WriteGo(w, pkg, buildTags(), b); err != nil {
+		log.Fatalf("Error writing file %s: %v", filename, err)
+	}
+}
+
+// WriteGo prepends a standard file comment and package statement to the given
+// bytes, applies gofmt, and writes them to w.
+func WriteGo(w io.Writer, pkg, tags string, b []byte) (n int, err error) {
+	src := []byte(header)
+	if tags != "" {
+		src = append(src, fmt.Sprintf("// +build %s\n\n", tags)...)
+	}
+	src = append(src, fmt.Sprintf("package %s\n\n", pkg)...)
+	src = append(src, b...)
+	formatted, err := format.Source(src)
+	if err != nil {
+		// Print the generated code even in case of an error so that the
+		// returned error can be meaningfully interpreted.
+		n, _ = w.Write(src)
+		return n, err
+	}
+	return w.Write(formatted)
+}
+
+// Repackage rewrites a Go file from belonging to package main to belonging to
+// the given package.
+func Repackage(inFile, outFile, pkg string) {
+	src, err := ioutil.ReadFile(inFile)
+	if err != nil {
+		log.Fatalf("reading %s: %v", inFile, err)
+	}
+	const toDelete = "package main\n\n"
+	i := bytes.Index(src, []byte(toDelete))
+	if i < 0 {
+		log.Fatalf("Could not find %q in %s.", toDelete, inFile)
+	}
+	w := &bytes.Buffer{}
+	w.Write(src[i+len(toDelete):])
+	WriteGoFile(outFile, pkg, w.Bytes())
+}

vendor/golang.org/x/text/internal/language/common.go

@@ -0,0 +1,16 @@
+// Code generated by running "go generate" in golang.org/x/text. DO NOT EDIT.
+
+package language
+
+// This file contains code common to the maketables.go and the package code.
+
+// AliasType is the type of an alias in AliasMap.
+type AliasType int8
+
+const (
+	Deprecated AliasType = iota
+	Macro
+	Legacy
+
+	AliasTypeUnknown AliasType = -1
+)

vendor/golang.org/x/text/internal/language/compact.go

@@ -0,0 +1,29 @@
+// Copyright 2018 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
+
+// CompactCoreInfo is a compact integer with the three core tags encoded.
+type CompactCoreInfo uint32
+
+// GetCompactCore generates a uint32 value that is guaranteed to be unique for
+// different language, region, and script values.
+func GetCompactCore(t Tag) (cci CompactCoreInfo, ok bool) {
+	if t.LangID > langNoIndexOffset {
+		return 0, false
+	}
+	cci |= CompactCoreInfo(t.LangID) << (8 + 12)
+	cci |= CompactCoreInfo(t.ScriptID) << 12
+	cci |= CompactCoreInfo(t.RegionID)
+	return cci, true
+}
+
+// Tag generates a tag from c.
+func (c CompactCoreInfo) Tag() Tag {
+	return Tag{
+		LangID:   Language(c >> 20),
+		RegionID: Region(c & 0x3ff),
+		ScriptID: Script(c>>12) & 0xff,
+	}
+}

vendor/golang.org/x/text/internal/language/compact/compact.go

@@ -0,0 +1,61 @@
+// Copyright 2018 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 compact defines a compact representation of language tags.
+//
+// Common language tags (at least all for which locale information is defined
+// in CLDR) are assigned a unique index. Each Tag is associated with such an
+// ID for selecting language-related resources (such as translations) as well
+// as one for selecting regional defaults (currency, number formatting, etc.)
+//
+// It may want to export this functionality at some point, but at this point
+// this is only available for use within x/text.
+package compact // import "golang.org/x/text/internal/language/compact"
+
+import (
+	"sort"
+	"strings"
+
+	"golang.org/x/text/internal/language"
+)
+
+// ID is an integer identifying a single tag.
+type ID uint16
+
+func getCoreIndex(t language.Tag) (id ID, ok bool) {
+	cci, ok := language.GetCompactCore(t)
+	if !ok {
+		return 0, false
+	}
+	i := sort.Search(len(coreTags), func(i int) bool {
+		return cci <= coreTags[i]
+	})
+	if i == len(coreTags) || coreTags[i] != cci {
+		return 0, false
+	}
+	return ID(i), true
+}
+
+// Parent returns the ID of the parent or the root ID if id is already the root.
+func (id ID) Parent() ID {
+	return parents[id]
+}
+
+// Tag converts id to an internal language Tag.
+func (id ID) Tag() language.Tag {
+	if int(id) >= len(coreTags) {
+		return specialTags[int(id)-len(coreTags)]
+	}
+	return coreTags[id].Tag()
+}
+
+var specialTags []language.Tag
+
+func init() {
+	tags := strings.Split(specialTagsStr, " ")
+	specialTags = make([]language.Tag, len(tags))
+	for i, t := range tags {
+		specialTags[i] = language.MustParse(t)
+	}
+}

vendor/golang.org/x/text/internal/language/compact/gen.go

@@ -0,0 +1,64 @@
+// 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.
+
+// +build ignore
+
+// Language tag table generator.
+// Data read from the web.
+
+package main
+
+import (
+	"flag"
+	"fmt"
+	"log"
+
+	"golang.org/x/text/internal/gen"
+	"golang.org/x/text/unicode/cldr"
+)
+
+var (
+	test = flag.Bool("test",
+		false,
+		"test existing tables; can be used to compare web data with package data.")
+	outputFile = flag.String("output",
+		"tables.go",
+		"output file for generated tables")
+)
+
+func main() {
+	gen.Init()
+
+	w := gen.NewCodeWriter()
+	defer w.WriteGoFile("tables.go", "compact")
+
+	fmt.Fprintln(w, `import "golang.org/x/text/internal/language"`)
+
+	b := newBuilder(w)
+	gen.WriteCLDRVersion(w)
+
+	b.writeCompactIndex()
+}
+
+type builder struct {
+	w    *gen.CodeWriter
+	data *cldr.CLDR
+	supp *cldr.SupplementalData
+}
+
+func newBuilder(w *gen.CodeWriter) *builder {
+	r := gen.OpenCLDRCoreZip()
+	defer r.Close()
+	d := &cldr.Decoder{}
+	data, err := d.DecodeZip(r)
+	if err != nil {
+		log.Fatal(err)
+	}
+	b := builder{
+		w:    w,
+		data: data,
+		supp: data.Supplemental(),
+	}
+	return &b
+}

vendor/golang.org/x/text/internal/language/compact/gen_index.go

@@ -0,0 +1,113 @@
+// Copyright 2015 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.
+
+// +build ignore
+
+package main
+
+// This file generates derivative tables based on the language package itself.
+
+import (
+	"fmt"
+	"log"
+	"sort"
+	"strings"
+
+	"golang.org/x/text/internal/language"
+)
+
+// Compact indices:
+// Note -va-X variants only apply to localization variants.
+// BCP variants only ever apply to language.
+// The only ambiguity between tags is with regions.
+
+func (b *builder) writeCompactIndex() {
+	// Collect all language tags for which we have any data in CLDR.
+	m := map[language.Tag]bool{}
+	for _, lang := range b.data.Locales() {
+		// We include all locales unconditionally to be consistent with en_US.
+		// We want en_US, even though it has no data associated with it.
+
+		// TODO: put any of the languages for which no data exists at the end
+		// of the index. This allows all components based on ICU to use that
+		// as the cutoff point.
+		// if x := data.RawLDML(lang); false ||
+		// 	x.LocaleDisplayNames != nil ||
+		// 	x.Characters != nil ||
+		// 	x.Delimiters != nil ||
+		// 	x.Measurement != nil ||
+		// 	x.Dates != nil ||
+		// 	x.Numbers != nil ||
+		// 	x.Units != nil ||
+		// 	x.ListPatterns != nil ||
+		// 	x.Collations != nil ||
+		// 	x.Segmentations != nil ||
+		// 	x.Rbnf != nil ||
+		// 	x.Annotations != nil ||
+		// 	x.Metadata != nil {
+
+		// TODO: support POSIX natively, albeit non-standard.
+		tag := language.Make(strings.Replace(lang, "_POSIX", "-u-va-posix", 1))
+		m[tag] = true
+		// }
+	}
+
+	// TODO: plural rules are also defined for the deprecated tags:
+	//    iw mo sh tl
+	// Consider removing these as compact tags.
+
+	// Include locales for plural rules, which uses a different structure.
+	for _, plurals := range b.supp.Plurals {
+		for _, rules := range plurals.PluralRules {
+			for _, lang := range strings.Split(rules.Locales, " ") {
+				m[language.Make(lang)] = true
+			}
+		}
+	}
+
+	var coreTags []language.CompactCoreInfo
+	var special []string
+
+	for t := range m {
+		if x := t.Extensions(); len(x) != 0 && fmt.Sprint(x) != "[u-va-posix]" {
+			log.Fatalf("Unexpected extension %v in %v", x, t)
+		}
+		if len(t.Variants()) == 0 && len(t.Extensions()) == 0 {
+			cci, ok := language.GetCompactCore(t)
+			if !ok {
+				log.Fatalf("Locale for non-basic language %q", t)
+			}
+			coreTags = append(coreTags, cci)
+		} else {
+			special = append(special, t.String())
+		}
+	}
+
+	w := b.w
+
+	sort.Slice(coreTags, func(i, j int) bool { return coreTags[i] < coreTags[j] })
+	sort.Strings(special)
+
+	w.WriteComment(`
+	NumCompactTags is the number of common tags. The maximum tag is
+	NumCompactTags-1.`)
+	w.WriteConst("NumCompactTags", len(m))
+
+	fmt.Fprintln(w, "const (")
+	for i, t := range coreTags {
+		fmt.Fprintf(w, "%s ID = %d\n", ident(t.Tag().String()), i)
+	}
+	for i, t := range special {
+		fmt.Fprintf(w, "%s ID = %d\n", ident(t), i+len(coreTags))
+	}
+	fmt.Fprintln(w, ")")
+
+	w.WriteVar("coreTags", coreTags)
+
+	w.WriteConst("specialTagsStr", strings.Join(special, " "))
+}
+
+func ident(s string) string {
+	return strings.Replace(s, "-", "", -1) + "Index"
+}

vendor/golang.org/x/text/internal/language/compact/gen_parents.go

@@ -0,0 +1,54 @@
+// Copyright 2018 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.
+
+// +build ignore
+
+package main
+
+import (
+	"log"
+
+	"golang.org/x/text/internal/gen"
+	"golang.org/x/text/internal/language"
+	"golang.org/x/text/internal/language/compact"
+	"golang.org/x/text/unicode/cldr"
+)
+
+func main() {
+	r := gen.OpenCLDRCoreZip()
+	defer r.Close()
+
+	d := &cldr.Decoder{}
+	data, err := d.DecodeZip(r)
+	if err != nil {
+		log.Fatalf("DecodeZip: %v", err)
+	}
+
+	w := gen.NewCodeWriter()
+	defer w.WriteGoFile("parents.go", "compact")
+
+	// Create parents table.
+	type ID uint16
+	parents := make([]ID, compact.NumCompactTags)
+	for _, loc := range data.Locales() {
+		tag := language.MustParse(loc)
+		index, ok := compact.FromTag(tag)
+		if !ok {
+			continue
+		}
+		parentIndex := compact.ID(0) // und
+		for p := tag.Parent(); p != language.Und; p = p.Parent() {
+			if x, ok := compact.FromTag(p); ok {
+				parentIndex = x
+				break
+			}
+		}
+		parents[index] = ID(parentIndex)
+	}
+
+	w.WriteComment(`
+	parents maps a compact index of a tag to the compact index of the parent of
+	this tag.`)
+	w.WriteVar("parents", parents)
+}

vendor/golang.org/x/text/internal/language/compact/language.go

@@ -0,0 +1,260 @@
+// 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.
+
+//go:generate go run gen.go gen_index.go -output tables.go
+//go:generate go run gen_parents.go
+
+package compact
+
+// TODO: Remove above NOTE after:
+// - verifying that tables are dropped correctly (most notably matcher tables).
+
+import (
+	"strings"
+
+	"golang.org/x/text/internal/language"
+)
+
+// Tag represents a BCP 47 language tag. It is used to specify an instance of a
+// specific language or locale. All language tag values are guaranteed to be
+// well-formed.
+type Tag struct {
+	// NOTE: exported tags will become part of the public API.
+	language ID
+	locale   ID
+	full     fullTag // always a language.Tag for now.
+}
+
+const _und = 0
+
+type fullTag interface {
+	IsRoot() bool
+	Parent() language.Tag
+}
+
+// Make a compact Tag from a fully specified internal language Tag.
+func Make(t language.Tag) (tag Tag) {
+	if region := t.TypeForKey("rg"); len(region) == 6 && region[2:] == "zzzz" {
+		if r, err := language.ParseRegion(region[:2]); err == nil {
+			tFull := t
+			t, _ = t.SetTypeForKey("rg", "")
+			// TODO: should we not consider "va" for the language tag?
+			var exact1, exact2 bool
+			tag.language, exact1 = FromTag(t)
+			t.RegionID = r
+			tag.locale, exact2 = FromTag(t)
+			if !exact1 || !exact2 {
+				tag.full = tFull
+			}
+			return tag
+		}
+	}
+	lang, ok := FromTag(t)
+	tag.language = lang
+	tag.locale = lang
+	if !ok {
+		tag.full = t
+	}
+	return tag
+}
+
+// Tag returns an internal language Tag version of this tag.
+func (t Tag) Tag() language.Tag {
+	if t.full != nil {
+		return t.full.(language.Tag)
+	}
+	tag := t.language.Tag()
+	if t.language != t.locale {
+		loc := t.locale.Tag()
+		tag, _ = tag.SetTypeForKey("rg", strings.ToLower(loc.RegionID.String())+"zzzz")
+	}
+	return tag
+}
+
+// IsCompact reports whether this tag is fully defined in terms of ID.
+func (t *Tag) IsCompact() bool {
+	return t.full == nil
+}
+
+// MayHaveVariants reports whether a tag may have variants. If it returns false
+// it is guaranteed the tag does not have variants.
+func (t Tag) MayHaveVariants() bool {
+	return t.full != nil || int(t.language) >= len(coreTags)
+}
+
+// MayHaveExtensions reports whether a tag may have extensions. If it returns
+// false it is guaranteed the tag does not have them.
+func (t Tag) MayHaveExtensions() bool {
+	return t.full != nil ||
+		int(t.language) >= len(coreTags) ||
+		t.language != t.locale
+}
+
+// IsRoot returns true if t is equal to language "und".
+func (t Tag) IsRoot() bool {
+	if t.full != nil {
+		return t.full.IsRoot()
+	}
+	return t.language == _und
+}
+
+// Parent returns the CLDR parent of t. In CLDR, missing fields in data for a
+// specific language are substituted with fields from the parent language.
+// The parent for a language may change for newer versions of CLDR.
+func (t Tag) Parent() Tag {
+	if t.full != nil {
+		return Make(t.full.Parent())
+	}
+	if t.language != t.locale {
+		// Simulate stripping -u-rg-xxxxxx
+		return Tag{language: t.language, locale: t.language}
+	}
+	// TODO: use parent lookup table once cycle from internal package is
+	// removed. Probably by internalizing the table and declaring this fast
+	// enough.
+	// lang := compactID(internal.Parent(uint16(t.language)))
+	lang, _ := FromTag(t.language.Tag().Parent())
+	return Tag{language: lang, locale: lang}
+}
+
+// returns token t and the rest of the string.
+func nextToken(s string) (t, tail string) {
+	p := strings.Index(s[1:], "-")
+	if p == -1 {
+		return s[1:], ""
+	}
+	p++
+	return s[1:p], s[p:]
+}
+
+// LanguageID returns an index, where 0 <= index < NumCompactTags, for tags
+// for which data exists in the text repository.The index will change over time
+// and should not be stored in persistent storage. If t does not match a compact
+// index, exact will be false and the compact index will be returned for the
+// first match after repeatedly taking the Parent of t.
+func LanguageID(t Tag) (id ID, exact bool) {
+	return t.language, t.full == nil
+}
+
+// RegionalID returns the ID for the regional variant of this tag. This index is
+// used to indicate region-specific overrides, such as default currency, default
+// calendar and week data, default time cycle, and default measurement system
+// and unit preferences.
+//
+// For instance, the tag en-GB-u-rg-uszzzz specifies British English with US
+// settings for currency, number formatting, etc. The CompactIndex for this tag
+// will be that for en-GB, while the RegionalID will be the one corresponding to
+// en-US.
+func RegionalID(t Tag) (id ID, exact bool) {
+	return t.locale, t.full == nil
+}
+
+// LanguageTag returns t stripped of regional variant indicators.
+//
+// At the moment this means it is stripped of a regional and variant subtag "rg"
+// and "va" in the "u" extension.
+func (t Tag) LanguageTag() Tag {
+	if t.full == nil {
+		return Tag{language: t.language, locale: t.language}
+	}
+	tt := t.Tag()
+	tt.SetTypeForKey("rg", "")
+	tt.SetTypeForKey("va", "")
+	return Make(tt)
+}
+
+// RegionalTag returns the regional variant of the tag.
+//
+// At the moment this means that the region is set from the regional subtag
+// "rg" in the "u" extension.
+func (t Tag) RegionalTag() Tag {
+	rt := Tag{language: t.locale, locale: t.locale}
+	if t.full == nil {
+		return rt
+	}
+	b := language.Builder{}
+	tag := t.Tag()
+	// tag, _ = tag.SetTypeForKey("rg", "")
+	b.SetTag(t.locale.Tag())
+	if v := tag.Variants(); v != "" {
+		for _, v := range strings.Split(v, "-") {
+			b.AddVariant(v)
+		}
+	}
+	for _, e := range tag.Extensions() {
+		b.AddExt(e)
+	}
+	return t
+}
+
+// FromTag reports closest matching ID for an internal language Tag.
+func FromTag(t language.Tag) (id ID, exact bool) {
+	// TODO: perhaps give more frequent tags a lower index.
+	// TODO: we could make the indexes stable. This will excluded some
+	//       possibilities for optimization, so don't do this quite yet.
+	exact = true
+
+	b, s, r := t.Raw()
+	if t.HasString() {
+		if t.IsPrivateUse() {
+			// We have no entries for user-defined tags.
+			return 0, false
+		}
+		hasExtra := false
+		if t.HasVariants() {
+			if t.HasExtensions() {
+				build := language.Builder{}
+				build.SetTag(language.Tag{LangID: b, ScriptID: s, RegionID: r})
+				build.AddVariant(t.Variants())
+				exact = false
+				t = build.Make()
+			}
+			hasExtra = true
+		} else if _, ok := t.Extension('u'); ok {
+			// TODO: va may mean something else. Consider not considering it.
+			// Strip all but the 'va' entry.
+			old := t
+			variant := t.TypeForKey("va")
+			t = language.Tag{LangID: b, ScriptID: s, RegionID: r}
+			if variant != "" {
+				t, _ = t.SetTypeForKey("va", variant)
+				hasExtra = true
+			}
+			exact = old == t
+		} else {
+			exact = false
+		}
+		if hasExtra {
+			// We have some variants.
+			for i, s := range specialTags {
+				if s == t {
+					return ID(i + len(coreTags)), exact
+				}
+			}
+			exact = false
+		}
+	}
+	if x, ok := getCoreIndex(t); ok {
+		return x, exact
+	}
+	exact = false
+	if r != 0 && s == 0 {
+		// Deal with cases where an extra script is inserted for the region.
+		t, _ := t.Maximize()
+		if x, ok := getCoreIndex(t); ok {
+			return x, exact
+		}
+	}
+	for t = t.Parent(); t != root; t = t.Parent() {
+		// No variants specified: just compare core components.
+		// The key has the form lllssrrr, where l, s, and r are nibbles for
+		// respectively the langID, scriptID, and regionID.
+		if x, ok := getCoreIndex(t); ok {
+			return x, exact
+		}
+	}
+	return 0, exact
+}
+
+var root = language.Tag{}

vendor/golang.org/x/text/internal/language/compact/parents.go

@@ -0,0 +1,120 @@
+// Code generated by running "go generate" in golang.org/x/text. DO NOT EDIT.
+
+package compact
+
+// parents maps a compact index of a tag to the compact index of the parent of
+// this tag.
+var parents = []ID{ // 775 elements
+	// Entry 0 - 3F
+	0x0000, 0x0000, 0x0001, 0x0001, 0x0000, 0x0004, 0x0000, 0x0006,
+	0x0000, 0x0008, 0x0000, 0x000a, 0x000a, 0x000a, 0x000a, 0x000a,
+	0x000a, 0x000a, 0x000a, 0x000a, 0x000a, 0x000a, 0x000a, 0x000a,
+	0x000a, 0x000a, 0x000a, 0x000a, 0x000a, 0x000a, 0x000a, 0x000a,
+	0x000a, 0x000a, 0x000a, 0x000a, 0x000a, 0x000a, 0x000a, 0x0000,
+	0x0000, 0x0028, 0x0000, 0x002a, 0x0000, 0x002c, 0x0000, 0x0000,
+	0x002f, 0x002e, 0x002e, 0x0000, 0x0033, 0x0000, 0x0035, 0x0000,
+	0x0037, 0x0000, 0x0039, 0x0000, 0x003b, 0x0000, 0x0000, 0x003e,
+	// Entry 40 - 7F
+	0x0000, 0x0040, 0x0040, 0x0000, 0x0043, 0x0043, 0x0000, 0x0046,
+	0x0000, 0x0048, 0x0000, 0x0000, 0x004b, 0x004a, 0x004a, 0x0000,
+	0x004f, 0x004f, 0x004f, 0x004f, 0x0000, 0x0054, 0x0054, 0x0000,
+	0x0057, 0x0000, 0x0059, 0x0000, 0x005b, 0x0000, 0x005d, 0x005d,
+	0x0000, 0x0060, 0x0000, 0x0062, 0x0000, 0x0064, 0x0000, 0x0066,
+	0x0066, 0x0000, 0x0069, 0x0000, 0x006b, 0x006b, 0x006b, 0x006b,
+	0x006b, 0x006b, 0x006b, 0x0000, 0x0073, 0x0000, 0x0075, 0x0000,
+	0x0077, 0x0000, 0x0000, 0x007a, 0x0000, 0x007c, 0x0000, 0x007e,
+	// Entry 80 - BF
+	0x0000, 0x0080, 0x0080, 0x0000, 0x0083, 0x0083, 0x0000, 0x0086,
+	0x0087, 0x0087, 0x0087, 0x0086, 0x0088, 0x0087, 0x0087, 0x0087,
+	0x0086, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0088,
+	0x0087, 0x0087, 0x0087, 0x0087, 0x0088, 0x0087, 0x0088, 0x0087,
+	0x0087, 0x0088, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087,
+	0x0087, 0x0087, 0x0087, 0x0086, 0x0087, 0x0087, 0x0087, 0x0087,
+	0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087,
+	0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0086, 0x0087, 0x0086,
+	// Entry C0 - FF
+	0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087,
+	0x0088, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087,
+	0x0086, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0088, 0x0087,
+	0x0087, 0x0088, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087,
+	0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0086, 0x0086, 0x0087,
+	0x0087, 0x0086, 0x0087, 0x0087, 0x0087, 0x0087, 0x0087, 0x0000,
+	0x00ef, 0x0000, 0x00f1, 0x00f2, 0x00f2, 0x00f2, 0x00f2, 0x00f2,
+	0x00f2, 0x00f2, 0x00f2, 0x00f2, 0x00f1, 0x00f2, 0x00f1, 0x00f1,
+	// Entry 100 - 13F
+	0x00f2, 0x00f2, 0x00f1, 0x00f2, 0x00f2, 0x00f2, 0x00f2, 0x00f1,
+	0x00f2, 0x00f2, 0x00f2, 0x00f2, 0x00f2, 0x00f2, 0x0000, 0x010e,
+	0x0000, 0x0110, 0x0000, 0x0112, 0x0000, 0x0114, 0x0114, 0x0000,
+	0x0117, 0x0117, 0x0117, 0x0117, 0x0000, 0x011c, 0x0000, 0x011e,
+	0x0000, 0x0120, 0x0120, 0x0000, 0x0123, 0x0123, 0x0123, 0x0123,
+	0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123,
+	0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123,
+	0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123,
+	// Entry 140 - 17F
+	0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123,
+	0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123, 0x0123,
+	0x0123, 0x0123, 0x0000, 0x0152, 0x0000, 0x0154, 0x0000, 0x0156,
+	0x0000, 0x0158, 0x0000, 0x015a, 0x0000, 0x015c, 0x015c, 0x015c,
+	0x0000, 0x0160, 0x0000, 0x0000, 0x0163, 0x0000, 0x0165, 0x0000,
+	0x0167, 0x0167, 0x0167, 0x0000, 0x016b, 0x0000, 0x016d, 0x0000,
+	0x016f, 0x0000, 0x0171, 0x0171, 0x0000, 0x0174, 0x0000, 0x0176,
+	0x0000, 0x0178, 0x0000, 0x017a, 0x0000, 0x017c, 0x0000, 0x017e,
+	// Entry 180 - 1BF
+	0x0000, 0x0000, 0x0000, 0x0182, 0x0000, 0x0184, 0x0184, 0x0184,
+	0x0184, 0x0000, 0x0000, 0x0000, 0x018b, 0x0000, 0x0000, 0x018e,
+	0x0000, 0x0000, 0x0191, 0x0000, 0x0000, 0x0000, 0x0195, 0x0000,
+	0x0197, 0x0000, 0x0000, 0x019a, 0x0000, 0x0000, 0x019d, 0x0000,
+	0x019f, 0x0000, 0x01a1, 0x0000, 0x01a3, 0x0000, 0x01a5, 0x0000,
+	0x01a7, 0x0000, 0x01a9, 0x0000, 0x01ab, 0x0000, 0x01ad, 0x0000,
+	0x01af, 0x0000, 0x01b1, 0x01b1, 0x0000, 0x01b4, 0x0000, 0x01b6,
+	0x0000, 0x01b8, 0x0000, 0x01ba, 0x0000, 0x01bc, 0x0000, 0x0000,
+	// Entry 1C0 - 1FF
+	0x01bf, 0x0000, 0x01c1, 0x0000, 0x01c3, 0x0000, 0x01c5, 0x0000,
+	0x01c7, 0x0000, 0x01c9, 0x0000, 0x01cb, 0x01cb, 0x01cb, 0x01cb,
+	0x0000, 0x01d0, 0x0000, 0x01d2, 0x01d2, 0x0000, 0x01d5, 0x0000,
+	0x01d7, 0x0000, 0x01d9, 0x0000, 0x01db, 0x0000, 0x01dd, 0x0000,
+	0x01df, 0x01df, 0x0000, 0x01e2, 0x0000, 0x01e4, 0x0000, 0x01e6,
+	0x0000, 0x01e8, 0x0000, 0x01ea, 0x0000, 0x01ec, 0x0000, 0x01ee,
+	0x0000, 0x01f0, 0x0000, 0x0000, 0x01f3, 0x0000, 0x01f5, 0x01f5,
+	0x01f5, 0x0000, 0x01f9, 0x0000, 0x01fb, 0x0000, 0x01fd, 0x0000,
+	// Entry 200 - 23F
+	0x01ff, 0x0000, 0x0000, 0x0202, 0x0000, 0x0204, 0x0204, 0x0000,
+	0x0207, 0x0000, 0x0209, 0x0209, 0x0000, 0x020c, 0x020c, 0x0000,
+	0x020f, 0x020f, 0x020f, 0x020f, 0x020f, 0x020f, 0x020f, 0x0000,
+	0x0217, 0x0000, 0x0219, 0x0000, 0x021b, 0x0000, 0x0000, 0x0000,
+	0x0000, 0x0000, 0x0221, 0x0000, 0x0000, 0x0224, 0x0000, 0x0226,
+	0x0226, 0x0000, 0x0229, 0x0000, 0x022b, 0x022b, 0x0000, 0x0000,
+	0x022f, 0x022e, 0x022e, 0x0000, 0x0000, 0x0234, 0x0000, 0x0236,
+	0x0000, 0x0238, 0x0000, 0x0244, 0x023a, 0x0244, 0x0244, 0x0244,
+	// Entry 240 - 27F
+	0x0244, 0x0244, 0x0244, 0x0244, 0x023a, 0x0244, 0x0244, 0x0000,
+	0x0247, 0x0247, 0x0247, 0x0000, 0x024b, 0x0000, 0x024d, 0x0000,
+	0x024f, 0x024f, 0x0000, 0x0252, 0x0000, 0x0254, 0x0254, 0x0254,
+	0x0254, 0x0254, 0x0254, 0x0000, 0x025b, 0x0000, 0x025d, 0x0000,
+	0x025f, 0x0000, 0x0261, 0x0000, 0x0263, 0x0000, 0x0265, 0x0000,
+	0x0000, 0x0268, 0x0268, 0x0268, 0x0000, 0x026c, 0x0000, 0x026e,
+	0x0000, 0x0270, 0x0000, 0x0000, 0x0000, 0x0274, 0x0273, 0x0273,
+	0x0000, 0x0278, 0x0000, 0x027a, 0x0000, 0x027c, 0x0000, 0x0000,
+	// Entry 280 - 2BF
+	0x0000, 0x0000, 0x0281, 0x0000, 0x0000, 0x0284, 0x0000, 0x0286,
+	0x0286, 0x0286, 0x0286, 0x0000, 0x028b, 0x028b, 0x028b, 0x0000,
+	0x028f, 0x028f, 0x028f, 0x028f, 0x028f, 0x0000, 0x0295, 0x0295,
+	0x0295, 0x0295, 0x0000, 0x0000, 0x0000, 0x0000, 0x029d, 0x029d,
+	0x029d, 0x0000, 0x02a1, 0x02a1, 0x02a1, 0x02a1, 0x0000, 0x0000,
+	0x02a7, 0x02a7, 0x02a7, 0x02a7, 0x0000, 0x02ac, 0x0000, 0x02ae,
+	0x02ae, 0x0000, 0x02b1, 0x0000, 0x02b3, 0x0000, 0x02b5, 0x02b5,
+	0x0000, 0x0000, 0x02b9, 0x0000, 0x0000, 0x0000, 0x02bd, 0x0000,
+	// Entry 2C0 - 2FF
+	0x02bf, 0x02bf, 0x0000, 0x0000, 0x02c3, 0x0000, 0x02c5, 0x0000,
+	0x02c7, 0x0000, 0x02c9, 0x0000, 0x02cb, 0x0000, 0x02cd, 0x02cd,
+	0x0000, 0x0000, 0x02d1, 0x0000, 0x02d3, 0x02d0, 0x02d0, 0x0000,
+	0x0000, 0x02d8, 0x02d7, 0x02d7, 0x0000, 0x0000, 0x02dd, 0x0000,
+	0x02df, 0x0000, 0x02e1, 0x0000, 0x0000, 0x02e4, 0x0000, 0x02e6,
+	0x0000, 0x0000, 0x02e9, 0x0000, 0x02eb, 0x0000, 0x02ed, 0x0000,
+	0x02ef, 0x02ef, 0x0000, 0x0000, 0x02f3, 0x02f2, 0x02f2, 0x0000,
+	0x02f7, 0x0000, 0x02f9, 0x02f9, 0x02f9, 0x02f9, 0x02f9, 0x0000,
+	// Entry 300 - 33F
+	0x02ff, 0x0300, 0x02ff, 0x0000, 0x0303, 0x0051, 0x00e6,
+} // Size: 1574 bytes
+
+// Total table size 1574 bytes (1KiB); checksum: 895AAF0B

vendor/golang.org/x/text/internal/language/compact/tags.go

@@ -0,0 +1,91 @@
+// 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 compact
+
+var (
+	und = Tag{}
+
+	Und Tag = Tag{}
+
+	Afrikaans            Tag = Tag{language: afIndex, locale: afIndex}
+	Amharic              Tag = Tag{language: amIndex, locale: amIndex}
+	Arabic               Tag = Tag{language: arIndex, locale: arIndex}
+	ModernStandardArabic Tag = Tag{language: ar001Index, locale: ar001Index}
+	Azerbaijani          Tag = Tag{language: azIndex, locale: azIndex}
+	Bulgarian            Tag = Tag{language: bgIndex, locale: bgIndex}
+	Bengali              Tag = Tag{language: bnIndex, locale: bnIndex}
+	Catalan              Tag = Tag{language: caIndex, locale: caIndex}
+	Czech                Tag = Tag{language: csIndex, locale: csIndex}
+	Danish               Tag = Tag{language: daIndex, locale: daIndex}
+	German               Tag = Tag{language: deIndex, locale: deIndex}
+	Greek                Tag = Tag{language: elIndex, locale: elIndex}
+	English              Tag = Tag{language: enIndex, locale: enIndex}
+	AmericanEnglish      Tag = Tag{language: enUSIndex, locale: enUSIndex}
+	BritishEnglish       Tag = Tag{language: enGBIndex, locale: enGBIndex}
+	Spanish              Tag = Tag{language: esIndex, locale: esIndex}
+	EuropeanSpanish      Tag = Tag{language: esESIndex, locale: esESIndex}
+	LatinAmericanSpanish Tag = Tag{language: es419Index, locale: es419Index}
+	Estonian             Tag = Tag{language: etIndex, locale: etIndex}
+	Persian              Tag = Tag{language: faIndex, locale: faIndex}
+	Finnish              Tag = Tag{language: fiIndex, locale: fiIndex}
+	Filipino             Tag = Tag{language: filIndex, locale: filIndex}
+	French               Tag = Tag{language: frIndex, locale: frIndex}
+	CanadianFrench       Tag = Tag{language: frCAIndex, locale: frCAIndex}
+	Gujarati             Tag = Tag{language: guIndex, locale: guIndex}
+	Hebrew               Tag = Tag{language: heIndex, locale: heIndex}
+	Hindi                Tag = Tag{language: hiIndex, locale: hiIndex}
+	Croatian             Tag = Tag{language: hrIndex, locale: hrIndex}
+	Hungarian            Tag = Tag{language: huIndex, locale: huIndex}
+	Armenian             Tag = Tag{language: hyIndex, locale: hyIndex}
+	Indonesian           Tag = Tag{language: idIndex, locale: idIndex}
+	Icelandic            Tag = Tag{language: isIndex, locale: isIndex}
+	Italian              Tag = Tag{language: itIndex, locale: itIndex}
+	Japanese             Tag = Tag{language: jaIndex, locale: jaIndex}
+	Georgian             Tag = Tag{language: kaIndex, locale: kaIndex}
+	Kazakh               Tag = Tag{language: kkIndex, locale: kkIndex}
+	Khmer                Tag = Tag{language: kmIndex, locale: kmIndex}
+	Kannada              Tag = Tag{language: knIndex, locale: knIndex}
+	Korean               Tag = Tag{language: koIndex, locale: koIndex}
+	Kirghiz              Tag = Tag{language: kyIndex, locale: kyIndex}
+	Lao                  Tag = Tag{language: loIndex, locale: loIndex}
+	Lithuanian           Tag = Tag{language: ltIndex, locale: ltIndex}
+	Latvian              Tag = Tag{language: lvIndex, locale: lvIndex}
+	Macedonian           Tag = Tag{language: mkIndex, locale: mkIndex}
+	Malayalam            Tag = Tag{language: mlIndex, locale: mlIndex}
+	Mongolian            Tag = Tag{language: mnIndex, locale: mnIndex}
+	Marathi              Tag = Tag{language: mrIndex, locale: mrIndex}
+	Malay                Tag = Tag{language: msIndex, locale: msIndex}
+	Burmese              Tag = Tag{language: myIndex, locale: myIndex}
+	Nepali               Tag = Tag{language: neIndex, locale: neIndex}
+	Dutch                Tag = Tag{language: nlIndex, locale: nlIndex}
+	Norwegian            Tag = Tag{language: noIndex, locale: noIndex}
+	Punjabi              Tag = Tag{language: paIndex, locale: paIndex}
+	Polish               Tag = Tag{language: plIndex, locale: plIndex}
+	Portuguese           Tag = Tag{language: ptIndex, locale: ptIndex}
+	BrazilianPortuguese  Tag = Tag{language: ptBRIndex, locale: ptBRIndex}
+	EuropeanPortuguese   Tag = Tag{language: ptPTIndex, locale: ptPTIndex}
+	Romanian             Tag = Tag{language: roIndex, locale: roIndex}
+	Russian              Tag = Tag{language: ruIndex, locale: ruIndex}
+	Sinhala              Tag = Tag{language: siIndex, locale: siIndex}
+	Slovak               Tag = Tag{language: skIndex, locale: skIndex}
+	Slovenian            Tag = Tag{language: slIndex, locale: slIndex}
+	Albanian             Tag = Tag{language: sqIndex, locale: sqIndex}
+	Serbian              Tag = Tag{language: srIndex, locale: srIndex}
+	SerbianLatin         Tag = Tag{language: srLatnIndex, locale: srLatnIndex}
+	Swedish              Tag = Tag{language: svIndex, locale: svIndex}
+	Swahili              Tag = Tag{language: swIndex, locale: swIndex}
+	Tamil                Tag = Tag{language: taIndex, locale: taIndex}
+	Telugu               Tag = Tag{language: teIndex, locale: teIndex}
+	Thai                 Tag = Tag{language: thIndex, locale: thIndex}
+	Turkish              Tag = Tag{language: trIndex, locale: trIndex}
+	Ukrainian            Tag = Tag{language: ukIndex, locale: ukIndex}
+	Urdu                 Tag = Tag{language: urIndex, locale: urIndex}
+	Uzbek                Tag = Tag{language: uzIndex, locale: uzIndex}
+	Vietnamese           Tag = Tag{language: viIndex, locale: viIndex}
+	Chinese              Tag = Tag{language: zhIndex, locale: zhIndex}
+	SimplifiedChinese    Tag = Tag{language: zhHansIndex, locale: zhHansIndex}
+	TraditionalChinese   Tag = Tag{language: zhHantIndex, locale: zhHantIndex}
+	Zulu                 Tag = Tag{language: zuIndex, locale: zuIndex}
+)

vendor/golang.org/x/text/internal/language/compose.go

@@ -0,0 +1,167 @@
+// Copyright 2018 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 (
+	"sort"
+	"strings"
+)
+
+// A Builder allows constructing a Tag from individual components.
+// Its main user is Compose in the top-level language package.
+type Builder struct {
+	Tag Tag
+
+	private    string // the x extension
+	variants   []string
+	extensions []string
+}
+
+// Make returns a new Tag from the current settings.
+func (b *Builder) Make() Tag {
+	t := b.Tag
+
+	if len(b.extensions) > 0 || len(b.variants) > 0 {
+		sort.Sort(sortVariants(b.variants))
+		sort.Strings(b.extensions)
+
+		if b.private != "" {
+			b.extensions = append(b.extensions, b.private)
+		}
+		n := maxCoreSize + tokenLen(b.variants...) + tokenLen(b.extensions...)
+		buf := make([]byte, n)
+		p := t.genCoreBytes(buf)
+		t.pVariant = byte(p)
+		p += appendTokens(buf[p:], b.variants...)
+		t.pExt = uint16(p)
+		p += appendTokens(buf[p:], b.extensions...)
+		t.str = string(buf[:p])
+		// We may not always need to remake the string, but when or when not
+		// to do so is rather tricky.
+		scan := makeScanner(buf[:p])
+		t, _ = parse(&scan, "")
+		return t
+
+	} else if b.private != "" {
+		t.str = b.private
+		t.RemakeString()
+	}
+	return t
+}
+
+// SetTag copies all the settings from a given Tag. Any previously set values
+// are discarded.
+func (b *Builder) SetTag(t Tag) {
+	b.Tag.LangID = t.LangID
+	b.Tag.RegionID = t.RegionID
+	b.Tag.ScriptID = t.ScriptID
+	// TODO: optimize
+	b.variants = b.variants[:0]
+	if variants := t.Variants(); variants != "" {
+		for _, vr := range strings.Split(variants[1:], "-") {
+			b.variants = append(b.variants, vr)
+		}
+	}
+	b.extensions, b.private = b.extensions[:0], ""
+	for _, e := range t.Extensions() {
+		b.AddExt(e)
+	}
+}
+
+// AddExt adds extension e to the tag. e must be a valid extension as returned
+// by Tag.Extension. If the extension already exists, it will be discarded,
+// except for a -u extension, where non-existing key-type pairs will added.
+func (b *Builder) AddExt(e string) {
+	if e[0] == 'x' {
+		if b.private == "" {
+			b.private = e
+		}
+		return
+	}
+	for i, s := range b.extensions {
+		if s[0] == e[0] {
+			if e[0] == 'u' {
+				b.extensions[i] += e[1:]
+			}
+			return
+		}
+	}
+	b.extensions = append(b.extensions, e)
+}
+
+// SetExt sets the extension e to the tag. e must be a valid extension as
+// returned by Tag.Extension. If the extension already exists, it will be
+// overwritten, except for a -u extension, where the individual key-type pairs
+// will be set.
+func (b *Builder) SetExt(e string) {
+	if e[0] == 'x' {
+		b.private = e
+		return
+	}
+	for i, s := range b.extensions {
+		if s[0] == e[0] {
+			if e[0] == 'u' {
+				b.extensions[i] = e + s[1:]
+			} else {
+				b.extensions[i] = e
+			}
+			return
+		}
+	}
+	b.extensions = append(b.extensions, e)
+}
+
+// AddVariant adds any number of variants.
+func (b *Builder) AddVariant(v ...string) {
+	for _, v := range v {
+		if v != "" {
+			b.variants = append(b.variants, v)
+		}
+	}
+}
+
+// ClearVariants removes any variants previously added, including those
+// copied from a Tag in SetTag.
+func (b *Builder) ClearVariants() {
+	b.variants = b.variants[:0]
+}
+
+// ClearExtensions removes any extensions previously added, including those
+// copied from a Tag in SetTag.
+func (b *Builder) ClearExtensions() {
+	b.private = ""
+	b.extensions = b.extensions[:0]
+}
+
+func tokenLen(token ...string) (n int) {
+	for _, t := range token {
+		n += len(t) + 1
+	}
+	return
+}
+
+func appendTokens(b []byte, token ...string) int {
+	p := 0
+	for _, t := range token {
+		b[p] = '-'
+		copy(b[p+1:], t)
+		p += 1 + len(t)
+	}
+	return p
+}
+
+type sortVariants []string
+
+func (s sortVariants) Len() int {
+	return len(s)
+}
+
+func (s sortVariants) Swap(i, j int) {
+	s[j], s[i] = s[i], s[j]
+}
+
+func (s sortVariants) Less(i, j int) bool {
+	return variantIndex[s[i]] < variantIndex[s[j]]
+}

vendor/golang.org/x/text/internal/language/coverage.go

@@ -0,0 +1,28 @@
+// Copyright 2014 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
+
+// BaseLanguages returns the list of all supported base languages. It generates
+// the list by traversing the internal structures.
+func BaseLanguages() []Language {
+	base := make([]Language, 0, NumLanguages)
+	for i := 0; i < langNoIndexOffset; i++ {
+		// We included "und" already for the value 0.
+		if i != nonCanonicalUnd {
+			base = append(base, Language(i))
+		}
+	}
+	i := langNoIndexOffset
+	for _, v := range langNoIndex {
+		for k := 0; k < 8; k++ {
+			if v&1 == 1 {
+				base = append(base, Language(i))
+			}
+			v >>= 1
+			i++
+		}
+	}
+	return base
+}

vendor/golang.org/x/text/internal/language/gen_common.go

@@ -0,0 +1,20 @@
+// Copyright 2014 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.
+
+// +build ignore
+
+package main
+
+// This file contains code common to the maketables.go and the package code.
+
+// AliasType is the type of an alias in AliasMap.
+type AliasType int8
+
+const (
+	Deprecated AliasType = iota
+	Macro
+	Legacy
+
+	AliasTypeUnknown AliasType = -1
+)

vendor/golang.org/x/text/internal/language/language.go

@@ -0,0 +1,596 @@
+// 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.
+
+//go:generate go run gen.go gen_common.go -output tables.go
+
+package language // import "golang.org/x/text/internal/language"
+
+// TODO: Remove above NOTE after:
+// - verifying that tables are dropped correctly (most notably matcher tables).
+
+import (
+	"errors"
+	"fmt"
+	"strings"
+)
+
+const (
+	// maxCoreSize is the maximum size of a BCP 47 tag without variants and
+	// extensions. Equals max lang (3) + script (4) + max reg (3) + 2 dashes.
+	maxCoreSize = 12
+
+	// max99thPercentileSize is a somewhat arbitrary buffer size that presumably
+	// is large enough to hold at least 99% of the BCP 47 tags.
+	max99thPercentileSize = 32
+
+	// maxSimpleUExtensionSize is the maximum size of a -u extension with one
+	// key-type pair. Equals len("-u-") + key (2) + dash + max value (8).
+	maxSimpleUExtensionSize = 14
+)
+
+// Tag represents a BCP 47 language tag. It is used to specify an instance of a
+// specific language or locale. All language tag values are guaranteed to be
+// well-formed. The zero value of Tag is Und.
+type Tag struct {
+	// TODO: the following fields have the form TagTypeID. This name is chosen
+	// to allow refactoring the public package without conflicting with its
+	// Base, Script, and Region methods. Once the transition is fully completed
+	// the ID can be stripped from the name.
+
+	LangID   Language
+	RegionID Region
+	// TODO: we will soon run out of positions for ScriptID. Idea: instead of
+	// storing lang, region, and ScriptID codes, store only the compact index and
+	// have a lookup table from this code to its expansion. This greatly speeds
+	// up table lookup, speed up common variant cases.
+	// This will also immediately free up 3 extra bytes. Also, the pVariant
+	// field can now be moved to the lookup table, as the compact index uniquely
+	// determines the offset of a possible variant.
+	ScriptID Script
+	pVariant byte   // offset in str, includes preceding '-'
+	pExt     uint16 // offset of first extension, includes preceding '-'
+
+	// str is the string representation of the Tag. It will only be used if the
+	// tag has variants or extensions.
+	str string
+}
+
+// Make is a convenience wrapper for Parse that omits the error.
+// In case of an error, a sensible default is returned.
+func Make(s string) Tag {
+	t, _ := Parse(s)
+	return t
+}
+
+// Raw returns the raw base language, script and region, without making an
+// attempt to infer their values.
+// TODO: consider removing
+func (t Tag) Raw() (b Language, s Script, r Region) {
+	return t.LangID, t.ScriptID, t.RegionID
+}
+
+// equalTags compares language, script and region subtags only.
+func (t Tag) equalTags(a Tag) bool {
+	return t.LangID == a.LangID && t.ScriptID == a.ScriptID && t.RegionID == a.RegionID
+}
+
+// IsRoot returns true if t is equal to language "und".
+func (t Tag) IsRoot() bool {
+	if int(t.pVariant) < len(t.str) {
+		return false
+	}
+	return t.equalTags(Und)
+}
+
+// IsPrivateUse reports whether the Tag consists solely of an IsPrivateUse use
+// tag.
+func (t Tag) IsPrivateUse() bool {
+	return t.str != "" && t.pVariant == 0
+}
+
+// RemakeString is used to update t.str in case lang, script or region changed.
+// It is assumed that pExt and pVariant still point to the start of the
+// respective parts.
+func (t *Tag) RemakeString() {
+	if t.str == "" {
+		return
+	}
+	extra := t.str[t.pVariant:]
+	if t.pVariant > 0 {
+		extra = extra[1:]
+	}
+	if t.equalTags(Und) && strings.HasPrefix(extra, "x-") {
+		t.str = extra
+		t.pVariant = 0
+		t.pExt = 0
+		return
+	}
+	var buf [max99thPercentileSize]byte // avoid extra memory allocation in most cases.
+	b := buf[:t.genCoreBytes(buf[:])]
+	if extra != "" {
+		diff := len(b) - int(t.pVariant)
+		b = append(b, '-')
+		b = append(b, extra...)
+		t.pVariant = uint8(int(t.pVariant) + diff)
+		t.pExt = uint16(int(t.pExt) + diff)
+	} else {
+		t.pVariant = uint8(len(b))
+		t.pExt = uint16(len(b))
+	}
+	t.str = string(b)
+}
+
+// genCoreBytes writes a string for the base languages, script and region tags
+// to the given buffer and returns the number of bytes written. It will never
+// write more than maxCoreSize bytes.
+func (t *Tag) genCoreBytes(buf []byte) int {
+	n := t.LangID.StringToBuf(buf[:])
+	if t.ScriptID != 0 {
+		n += copy(buf[n:], "-")
+		n += copy(buf[n:], t.ScriptID.String())
+	}
+	if t.RegionID != 0 {
+		n += copy(buf[n:], "-")
+		n += copy(buf[n:], t.RegionID.String())
+	}
+	return n
+}
+
+// String returns the canonical string representation of the language tag.
+func (t Tag) String() string {
+	if t.str != "" {
+		return t.str
+	}
+	if t.ScriptID == 0 && t.RegionID == 0 {
+		return t.LangID.String()
+	}
+	buf := [maxCoreSize]byte{}
+	return string(buf[:t.genCoreBytes(buf[:])])
+}
+
+// MarshalText implements encoding.TextMarshaler.
+func (t Tag) MarshalText() (text []byte, err error) {
+	if t.str != "" {
+		text = append(text, t.str...)
+	} else if t.ScriptID == 0 && t.RegionID == 0 {
+		text = append(text, t.LangID.String()...)
+	} else {
+		buf := [maxCoreSize]byte{}
+		text = buf[:t.genCoreBytes(buf[:])]
+	}
+	return text, nil
+}
+
+// UnmarshalText implements encoding.TextUnmarshaler.
+func (t *Tag) UnmarshalText(text []byte) error {
+	tag, err := Parse(string(text))
+	*t = tag
+	return err
+}
+
+// Variants returns the part of the tag holding all variants or the empty string
+// if there are no variants defined.
+func (t Tag) Variants() string {
+	if t.pVariant == 0 {
+		return ""
+	}
+	return t.str[t.pVariant:t.pExt]
+}
+
+// VariantOrPrivateUseTags returns variants or private use tags.
+func (t Tag) VariantOrPrivateUseTags() string {
+	if t.pExt > 0 {
+		return t.str[t.pVariant:t.pExt]
+	}
+	return t.str[t.pVariant:]
+}
+
+// HasString reports whether this tag defines more than just the raw
+// components.
+func (t Tag) HasString() bool {
+	return t.str != ""
+}
+
+// Parent returns the CLDR parent of t. In CLDR, missing fields in data for a
+// specific language are substituted with fields from the parent language.
+// The parent for a language may change for newer versions of CLDR.
+func (t Tag) Parent() Tag {
+	if t.str != "" {
+		// Strip the variants and extensions.
+		b, s, r := t.Raw()
+		t = Tag{LangID: b, ScriptID: s, RegionID: r}
+		if t.RegionID == 0 && t.ScriptID != 0 && t.LangID != 0 {
+			base, _ := addTags(Tag{LangID: t.LangID})
+			if base.ScriptID == t.ScriptID {
+				return Tag{LangID: t.LangID}
+			}
+		}
+		return t
+	}
+	if t.LangID != 0 {
+		if t.RegionID != 0 {
+			maxScript := t.ScriptID
+			if maxScript == 0 {
+				max, _ := addTags(t)
+				maxScript = max.ScriptID
+			}
+
+			for i := range parents {
+				if Language(parents[i].lang) == t.LangID && Script(parents[i].maxScript) == maxScript {
+					for _, r := range parents[i].fromRegion {
+						if Region(r) == t.RegionID {
+							return Tag{
+								LangID:   t.LangID,
+								ScriptID: Script(parents[i].script),
+								RegionID: Region(parents[i].toRegion),
+							}
+						}
+					}
+				}
+			}
+
+			// Strip the script if it is the default one.
+			base, _ := addTags(Tag{LangID: t.LangID})
+			if base.ScriptID != maxScript {
+				return Tag{LangID: t.LangID, ScriptID: maxScript}
+			}
+			return Tag{LangID: t.LangID}
+		} else if t.ScriptID != 0 {
+			// The parent for an base-script pair with a non-default script is
+			// "und" instead of the base language.
+			base, _ := addTags(Tag{LangID: t.LangID})
+			if base.ScriptID != t.ScriptID {
+				return Und
+			}
+			return Tag{LangID: t.LangID}
+		}
+	}
+	return Und
+}
+
+// ParseExtension parses s as an extension and returns it on success.
+func ParseExtension(s string) (ext string, err error) {
+	scan := makeScannerString(s)
+	var end int
+	if n := len(scan.token); n != 1 {
+		return "", ErrSyntax
+	}
+	scan.toLower(0, len(scan.b))
+	end = parseExtension(&scan)
+	if end != len(s) {
+		return "", ErrSyntax
+	}
+	return string(scan.b), nil
+}
+
+// HasVariants reports whether t has variants.
+func (t Tag) HasVariants() bool {
+	return uint16(t.pVariant) < t.pExt
+}
+
+// HasExtensions reports whether t has extensions.
+func (t Tag) HasExtensions() bool {
+	return int(t.pExt) < len(t.str)
+}
+
+// Extension returns the extension of type x for tag t. It will return
+// false for ok if t does not have the requested extension. The returned
+// extension will be invalid in this case.
+func (t Tag) Extension(x byte) (ext string, ok bool) {
+	for i := int(t.pExt); i < len(t.str)-1; {
+		var ext string
+		i, ext = getExtension(t.str, i)
+		if ext[0] == x {
+			return ext, true
+		}
+	}
+	return "", false
+}
+
+// Extensions returns all extensions of t.
+func (t Tag) Extensions() []string {
+	e := []string{}
+	for i := int(t.pExt); i < len(t.str)-1; {
+		var ext string
+		i, ext = getExtension(t.str, i)
+		e = append(e, ext)
+	}
+	return e
+}
+
+// TypeForKey returns the type associated with the given key, where key and type
+// are of the allowed values defined for the Unicode locale extension ('u') in
+// https://www.unicode.org/reports/tr35/#Unicode_Language_and_Locale_Identifiers.
+// TypeForKey will traverse the inheritance chain to get the correct value.
+func (t Tag) TypeForKey(key string) string {
+	if start, end, _ := t.findTypeForKey(key); end != start {
+		return t.str[start:end]
+	}
+	return ""
+}
+
+var (
+	errPrivateUse       = errors.New("cannot set a key on a private use tag")
+	errInvalidArguments = errors.New("invalid key or type")
+)
+
+// SetTypeForKey returns a new Tag with the key set to type, where key and type
+// are of the allowed values defined for the Unicode locale extension ('u') in
+// https://www.unicode.org/reports/tr35/#Unicode_Language_and_Locale_Identifiers.
+// An empty value removes an existing pair with the same key.
+func (t Tag) SetTypeForKey(key, value string) (Tag, error) {
+	if t.IsPrivateUse() {
+		return t, errPrivateUse
+	}
+	if len(key) != 2 {
+		return t, errInvalidArguments
+	}
+
+	// Remove the setting if value is "".
+	if value == "" {
+		start, end, _ := t.findTypeForKey(key)
+		if start != end {
+			// Remove key tag and leading '-'.
+			start -= 4
+
+			// Remove a possible empty extension.
+			if (end == len(t.str) || t.str[end+2] == '-') && t.str[start-2] == '-' {
+				start -= 2
+			}
+			if start == int(t.pVariant) && end == len(t.str) {
+				t.str = ""
+				t.pVariant, t.pExt = 0, 0
+			} else {
+				t.str = fmt.Sprintf("%s%s", t.str[:start], t.str[end:])
+			}
+		}
+		return t, nil
+	}
+
+	if len(value) < 3 || len(value) > 8 {
+		return t, errInvalidArguments
+	}
+
+	var (
+		buf    [maxCoreSize + maxSimpleUExtensionSize]byte
+		uStart int // start of the -u extension.
+	)
+
+	// Generate the tag string if needed.
+	if t.str == "" {
+		uStart = t.genCoreBytes(buf[:])
+		buf[uStart] = '-'
+		uStart++
+	}
+
+	// Create new key-type pair and parse it to verify.
+	b := buf[uStart:]
+	copy(b, "u-")
+	copy(b[2:], key)
+	b[4] = '-'
+	b = b[:5+copy(b[5:], value)]
+	scan := makeScanner(b)
+	if parseExtensions(&scan); scan.err != nil {
+		return t, scan.err
+	}
+
+	// Assemble the replacement string.
+	if t.str == "" {
+		t.pVariant, t.pExt = byte(uStart-1), uint16(uStart-1)
+		t.str = string(buf[:uStart+len(b)])
+	} else {
+		s := t.str
+		start, end, hasExt := t.findTypeForKey(key)
+		if start == end {
+			if hasExt {
+				b = b[2:]
+			}
+			t.str = fmt.Sprintf("%s-%s%s", s[:start], b, s[end:])
+		} else {
+			t.str = fmt.Sprintf("%s%s%s", s[:start], value, s[end:])
+		}
+	}
+	return t, nil
+}
+
+// findKeyAndType returns the start and end position for the type corresponding
+// to key or the point at which to insert the key-value pair if the type
+// wasn't found. The hasExt return value reports whether an -u extension was present.
+// Note: the extensions are typically very small and are likely to contain
+// only one key-type pair.
+func (t Tag) findTypeForKey(key string) (start, end int, hasExt bool) {
+	p := int(t.pExt)
+	if len(key) != 2 || p == len(t.str) || p == 0 {
+		return p, p, false
+	}
+	s := t.str
+
+	// Find the correct extension.
+	for p++; s[p] != 'u'; p++ {
+		if s[p] > 'u' {
+			p--
+			return p, p, false
+		}
+		if p = nextExtension(s, p); p == len(s) {
+			return len(s), len(s), false
+		}
+	}
+	// Proceed to the hyphen following the extension name.
+	p++
+
+	// curKey is the key currently being processed.
+	curKey := ""
+
+	// Iterate over keys until we get the end of a section.
+	for {
+		// p points to the hyphen preceding the current token.
+		if p3 := p + 3; s[p3] == '-' {
+			// Found a key.
+			// Check whether we just processed the key that was requested.
+			if curKey == key {
+				return start, p, true
+			}
+			// Set to the next key and continue scanning type tokens.
+			curKey = s[p+1 : p3]
+			if curKey > key {
+				return p, p, true
+			}
+			// Start of the type token sequence.
+			start = p + 4
+			// A type is at least 3 characters long.
+			p += 7 // 4 + 3
+		} else {
+			// Attribute or type, which is at least 3 characters long.
+			p += 4
+		}
+		// p points past the third character of a type or attribute.
+		max := p + 5 // maximum length of token plus hyphen.
+		if len(s) < max {
+			max = len(s)
+		}
+		for ; p < max && s[p] != '-'; p++ {
+		}
+		// Bail if we have exhausted all tokens or if the next token starts
+		// a new extension.
+		if p == len(s) || s[p+2] == '-' {
+			if curKey == key {
+				return start, p, true
+			}
+			return p, p, true
+		}
+	}
+}
+
+// ParseBase parses a 2- or 3-letter ISO 639 code.
+// It returns a ValueError if s is a well-formed but unknown language identifier
+// or another error if another error occurred.
+func ParseBase(s string) (Language, error) {
+	if n := len(s); n < 2 || 3 < n {
+		return 0, ErrSyntax
+	}
+	var buf [3]byte
+	return getLangID(buf[:copy(buf[:], s)])
+}
+
+// ParseScript parses a 4-letter ISO 15924 code.
+// It returns a ValueError if s is a well-formed but unknown script identifier
+// or another error if another error occurred.
+func ParseScript(s string) (Script, error) {
+	if len(s) != 4 {
+		return 0, ErrSyntax
+	}
+	var buf [4]byte
+	return getScriptID(script, buf[:copy(buf[:], s)])
+}
+
+// EncodeM49 returns the Region for the given UN M.49 code.
+// It returns an error if r is not a valid code.
+func EncodeM49(r int) (Region, error) {
+	return getRegionM49(r)
+}
+
+// ParseRegion parses a 2- or 3-letter ISO 3166-1 or a UN M.49 code.
+// It returns a ValueError if s is a well-formed but unknown region identifier
+// or another error if another error occurred.
+func ParseRegion(s string) (Region, error) {
+	if n := len(s); n < 2 || 3 < n {
+		return 0, ErrSyntax
+	}
+	var buf [3]byte
+	return getRegionID(buf[:copy(buf[:], s)])
+}
+
+// IsCountry returns whether this region is a country or autonomous area. This
+// includes non-standard definitions from CLDR.
+func (r Region) IsCountry() bool {
+	if r == 0 || r.IsGroup() || r.IsPrivateUse() && r != _XK {
+		return false
+	}
+	return true
+}
+
+// IsGroup returns whether this region defines a collection of regions. This
+// includes non-standard definitions from CLDR.
+func (r Region) IsGroup() bool {
+	if r == 0 {
+		return false
+	}
+	return int(regionInclusion[r]) < len(regionContainment)
+}
+
+// Contains returns whether Region c is contained by Region r. It returns true
+// if c == r.
+func (r Region) Contains(c Region) bool {
+	if r == c {
+		return true
+	}
+	g := regionInclusion[r]
+	if g >= nRegionGroups {
+		return false
+	}
+	m := regionContainment[g]
+
+	d := regionInclusion[c]
+	b := regionInclusionBits[d]
+
+	// A contained country may belong to multiple disjoint groups. Matching any
+	// of these indicates containment. If the contained region is a group, it
+	// must strictly be a subset.
+	if d >= nRegionGroups {
+		return b&m != 0
+	}
+	return b&^m == 0
+}
+
+var errNoTLD = errors.New("language: region is not a valid ccTLD")
+
+// TLD returns the country code top-level domain (ccTLD). UK is returned for GB.
+// In all other cases it returns either the region itself or an error.
+//
+// This method may return an error for a region for which there exists a
+// canonical form with a ccTLD. To get that ccTLD canonicalize r first. The
+// region will already be canonicalized it was obtained from a Tag that was
+// obtained using any of the default methods.
+func (r Region) TLD() (Region, error) {
+	// See http://en.wikipedia.org/wiki/Country_code_top-level_domain for the
+	// difference between ISO 3166-1 and IANA ccTLD.
+	if r == _GB {
+		r = _UK
+	}
+	if (r.typ() & ccTLD) == 0 {
+		return 0, errNoTLD
+	}
+	return r, nil
+}
+
+// Canonicalize returns the region or a possible replacement if the region is
+// deprecated. It will not return a replacement for deprecated regions that
+// are split into multiple regions.
+func (r Region) Canonicalize() Region {
+	if cr := normRegion(r); cr != 0 {
+		return cr
+	}
+	return r
+}
+
+// Variant represents a registered variant of a language as defined by BCP 47.
+type Variant struct {
+	ID  uint8
+	str string
+}
+
+// ParseVariant parses and returns a Variant. An error is returned if s is not
+// a valid variant.
+func ParseVariant(s string) (Variant, error) {
+	s = strings.ToLower(s)
+	if id, ok := variantIndex[s]; ok {
+		return Variant{id, s}, nil
+	}
+	return Variant{}, NewValueError([]byte(s))
+}
+
+// String returns the string representation of the variant.
+func (v Variant) String() string {
+	return v.str
+}

vendor/golang.org/x/text/internal/language/lookup.go

@@ -0,0 +1,412 @@
+// 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, NewValueError(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 Language 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) (Language, error) {
+	if len(s) == 2 {
+		return getLangISO2(s)
+	}
+	return getLangISO3(s)
+}
+
+// TODO language normalization as well as the AliasMaps could be moved to the
+// higher level package, but it is a bit tricky to separate the generation.
+
+func (id Language) Canonicalize() (Language, AliasType) {
+	return normLang(id)
+}
+
+// mapLang returns the mapped langID of id according to mapping m.
+func normLang(id Language) (Language, AliasType) {
+	k := sort.Search(len(AliasMap), func(i int) bool {
+		return AliasMap[i].From >= uint16(id)
+	})
+	if k < len(AliasMap) && AliasMap[k].From == uint16(id) {
+		return Language(AliasMap[k].To), AliasTypes[k]
+	}
+	return id, AliasTypeUnknown
+}
+
+// getLangISO2 returns the langID for the given 2-letter ISO language code
+// or unknownLang if this does not exist.
+func getLangISO2(s []byte) (Language, error) {
+	if !tag.FixCase("zz", s) {
+		return 0, ErrSyntax
+	}
+	if i := lang.Index(s); i != -1 && lang.Elem(i)[3] != 0 {
+		return Language(i), nil
+	}
+	return 0, NewValueError(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) (Language, 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 := Language(i)
+				if id == nonCanonicalUnd {
+					return 0, nil
+				}
+				return id, nil
+			}
+		}
+		if i := altLangISO3.Index(s); i != -1 {
+			return Language(altLangIndex[altLangISO3.Elem(i)[3]]), nil
+		}
+		n := strToInt(s)
+		if langNoIndex[n/8]&(1<<(n%8)) != 0 {
+			return Language(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 Language(i), nil
+			}
+		}
+		return 0, NewValueError(s)
+	}
+	return 0, ErrSyntax
+}
+
+// StringToBuf writes the string to b and returns the number of bytes
+// written.  cap(b) must be >= 3.
+func (id Language) 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 Language) 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 Language) 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 Language) IsPrivateUse() bool {
+	return langPrivateStart <= b && b <= langPrivateEnd
+}
+
+// SuppressScript returns the script marked as SuppressScript in the IANA
+// language tag repository, or 0 if there is no such script.
+func (b Language) SuppressScript() Script {
+	if b < langNoIndexOffset {
+		return Script(suppressScript[b])
+	}
+	return 0
+}
+
+type Region uint16
+
+// getRegionID returns the region id for s if s is a valid 2-letter region code
+// or unknownRegion.
+func getRegionID(s []byte) (Region, 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) (Region, error) {
+	i, err := findIndex(regionISO, s, "ZZ")
+	if err != nil {
+		return 0, err
+	}
+	return Region(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) (Region, 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 Region(i) + isoRegionOffset, nil
+			}
+		}
+		for i := 0; i < len(altRegionISO3); i += 3 {
+			if tag.Compare(altRegionISO3[i:i+3], s) == 0 {
+				return Region(altRegionIDs[i/3]), nil
+			}
+		}
+		return 0, NewValueError(s)
+	}
+	return 0, ErrSyntax
+}
+
+func getRegionM49(n int) (Region, 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 Region(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 Region) Region {
+	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 Region(m[k].To)
+	}
+	return 0
+}
+
+const (
+	iso3166UserAssigned = 1 << iota
+	ccTLD
+	bcp47Region
+)
+
+func (r Region) typ() byte {
+	return regionTypes[r]
+}
+
+// String returns the BCP 47 representation for the region.
+// It returns "ZZ" for an unspecified region.
+func (r Region) 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 Region) 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 Region) 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 Region) IsPrivateUse() bool {
+	return r.typ()&iso3166UserAssigned != 0
+}
+
+type Script 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) (Script, error) {
+	i, err := findIndex(idx, s, "Zzzz")
+	return Script(i), err
+}
+
+// String returns the script code in title case.
+// It returns "Zzzz" for an unspecified script.
+func (s Script) 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 Script) 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.LangID = Language(v)
+		return t, true
+	}
+	return t, false
+}

vendor/golang.org/x/text/internal/language/match.go

@@ -0,0 +1,226 @@
+// 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 "errors"
+
+type scriptRegionFlags uint8
+
+const (
+	isList = 1 << iota
+	scriptInFrom
+	regionInFrom
+)
+
+func (t *Tag) setUndefinedLang(id Language) {
+	if t.LangID == 0 {
+		t.LangID = id
+	}
+}
+
+func (t *Tag) setUndefinedScript(id Script) {
+	if t.ScriptID == 0 {
+		t.ScriptID = id
+	}
+}
+
+func (t *Tag) setUndefinedRegion(id Region) {
+	if t.RegionID == 0 || t.RegionID.Contains(id) {
+		t.RegionID = id
+	}
+}
+
+// ErrMissingLikelyTagsData indicates no information was available
+// to compute likely values of missing tags.
+var ErrMissingLikelyTagsData = errors.New("missing likely tags data")
+
+// addLikelySubtags sets subtags to their most likely value, given the locale.
+// In most cases this means setting fields for unknown values, but in some
+// cases it may alter a value.  It returns an ErrMissingLikelyTagsData error
+// if the given locale cannot be expanded.
+func (t Tag) addLikelySubtags() (Tag, error) {
+	id, err := addTags(t)
+	if err != nil {
+		return t, err
+	} else if id.equalTags(t) {
+		return t, nil
+	}
+	id.RemakeString()
+	return id, nil
+}
+
+// specializeRegion attempts to specialize a group region.
+func specializeRegion(t *Tag) bool {
+	if i := regionInclusion[t.RegionID]; i < nRegionGroups {
+		x := likelyRegionGroup[i]
+		if Language(x.lang) == t.LangID && Script(x.script) == t.ScriptID {
+			t.RegionID = Region(x.region)
+		}
+		return true
+	}
+	return false
+}
+
+// Maximize returns a new tag with missing tags filled in.
+func (t Tag) Maximize() (Tag, error) {
+	return addTags(t)
+}
+
+func addTags(t Tag) (Tag, error) {
+	// We leave private use identifiers alone.
+	if t.IsPrivateUse() {
+		return t, nil
+	}
+	if t.ScriptID != 0 && t.RegionID != 0 {
+		if t.LangID != 0 {
+			// already fully specified
+			specializeRegion(&t)
+			return t, nil
+		}
+		// Search matches for und-script-region. Note that for these cases
+		// region will never be a group so there is no need to check for this.
+		list := likelyRegion[t.RegionID : t.RegionID+1]
+		if x := list[0]; x.flags&isList != 0 {
+			list = likelyRegionList[x.lang : x.lang+uint16(x.script)]
+		}
+		for _, x := range list {
+			// Deviating from the spec. See match_test.go for details.
+			if Script(x.script) == t.ScriptID {
+				t.setUndefinedLang(Language(x.lang))
+				return t, nil
+			}
+		}
+	}
+	if t.LangID != 0 {
+		// Search matches for lang-script and lang-region, where lang != und.
+		if t.LangID < langNoIndexOffset {
+			x := likelyLang[t.LangID]
+			if x.flags&isList != 0 {
+				list := likelyLangList[x.region : x.region+uint16(x.script)]
+				if t.ScriptID != 0 {
+					for _, x := range list {
+						if Script(x.script) == t.ScriptID && x.flags&scriptInFrom != 0 {
+							t.setUndefinedRegion(Region(x.region))
+							return t, nil
+						}
+					}
+				} else if t.RegionID != 0 {
+					count := 0
+					goodScript := true
+					tt := t
+					for _, x := range list {
+						// We visit all entries for which the script was not
+						// defined, including the ones where the region was not
+						// defined. This allows for proper disambiguation within
+						// regions.
+						if x.flags&scriptInFrom == 0 && t.RegionID.Contains(Region(x.region)) {
+							tt.RegionID = Region(x.region)
+							tt.setUndefinedScript(Script(x.script))
+							goodScript = goodScript && tt.ScriptID == Script(x.script)
+							count++
+						}
+					}
+					if count == 1 {
+						return tt, nil
+					}
+					// Even if we fail to find a unique Region, we might have
+					// an unambiguous script.
+					if goodScript {
+						t.ScriptID = tt.ScriptID
+					}
+				}
+			}
+		}
+	} else {
+		// Search matches for und-script.
+		if t.ScriptID != 0 {
+			x := likelyScript[t.ScriptID]
+			if x.region != 0 {
+				t.setUndefinedRegion(Region(x.region))
+				t.setUndefinedLang(Language(x.lang))
+				return t, nil
+			}
+		}
+		// Search matches for und-region. If und-script-region exists, it would
+		// have been found earlier.
+		if t.RegionID != 0 {
+			if i := regionInclusion[t.RegionID]; i < nRegionGroups {
+				x := likelyRegionGroup[i]
+				if x.region != 0 {
+					t.setUndefinedLang(Language(x.lang))
+					t.setUndefinedScript(Script(x.script))
+					t.RegionID = Region(x.region)
+				}
+			} else {
+				x := likelyRegion[t.RegionID]
+				if x.flags&isList != 0 {
+					x = likelyRegionList[x.lang]
+				}
+				if x.script != 0 && x.flags != scriptInFrom {
+					t.setUndefinedLang(Language(x.lang))
+					t.setUndefinedScript(Script(x.script))
+					return t, nil
+				}
+			}
+		}
+	}
+
+	// Search matches for lang.
+	if t.LangID < langNoIndexOffset {
+		x := likelyLang[t.LangID]
+		if x.flags&isList != 0 {
+			x = likelyLangList[x.region]
+		}
+		if x.region != 0 {
+			t.setUndefinedScript(Script(x.script))
+			t.setUndefinedRegion(Region(x.region))
+		}
+		specializeRegion(&t)
+		if t.LangID == 0 {
+			t.LangID = _en // default language
+		}
+		return t, nil
+	}
+	return t, ErrMissingLikelyTagsData
+}
+
+func (t *Tag) setTagsFrom(id Tag) {
+	t.LangID = id.LangID
+	t.ScriptID = id.ScriptID
+	t.RegionID = id.RegionID
+}
+
+// minimize removes the region or script subtags from t such that
+// t.addLikelySubtags() == t.minimize().addLikelySubtags().
+func (t Tag) minimize() (Tag, error) {
+	t, err := minimizeTags(t)
+	if err != nil {
+		return t, err
+	}
+	t.RemakeString()
+	return t, nil
+}
+
+// minimizeTags mimics the behavior of the ICU 51 C implementation.
+func minimizeTags(t Tag) (Tag, error) {
+	if t.equalTags(Und) {
+		return t, nil
+	}
+	max, err := addTags(t)
+	if err != nil {
+		return t, err
+	}
+	for _, id := range [...]Tag{
+		{LangID: t.LangID},
+		{LangID: t.LangID, RegionID: t.RegionID},
+		{LangID: t.LangID, ScriptID: t.ScriptID},
+	} {
+		if x, err := addTags(id); err == nil && max.equalTags(x) {
+			t.setTagsFrom(id)
+			break
+		}
+	}
+	return t, nil
+}

vendor/golang.org/x/text/internal/language/parse.go

@@ -0,0 +1,594 @@
+// 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"
+	"errors"
+	"fmt"
+	"sort"
+
+	"golang.org/x/text/internal/tag"
+)
+
+// isAlpha returns true if the byte is not a digit.
+// b must be an ASCII letter or digit.
+func isAlpha(b byte) bool {
+	return b > '9'
+}
+
+// isAlphaNum returns true if the string contains only ASCII letters or digits.
+func isAlphaNum(s []byte) bool {
+	for _, c := range s {
+		if !('a' <= c && c <= 'z' || 'A' <= c && c <= 'Z' || '0' <= c && c <= '9') {
+			return false
+		}
+	}
+	return true
+}
+
+// ErrSyntax is returned by any of the parsing functions when the
+// input is not well-formed, according to BCP 47.
+// TODO: return the position at which the syntax error occurred?
+var ErrSyntax = errors.New("language: tag is not well-formed")
+
+// ErrDuplicateKey is returned when a tag contains the same key twice with
+// different values in the -u section.
+var ErrDuplicateKey = errors.New("language: different values for same key in -u extension")
+
+// ValueError is returned by any of the parsing functions when the
+// input is well-formed but the respective subtag is not recognized
+// as a valid value.
+type ValueError struct {
+	v [8]byte
+}
+
+// NewValueError creates a new ValueError.
+func NewValueError(tag []byte) ValueError {
+	var e ValueError
+	copy(e.v[:], tag)
+	return e
+}
+
+func (e ValueError) tag() []byte {
+	n := bytes.IndexByte(e.v[:], 0)
+	if n == -1 {
+		n = 8
+	}
+	return e.v[:n]
+}
+
+// Error implements the error interface.
+func (e ValueError) Error() string {
+	return fmt.Sprintf("language: subtag %q is well-formed but unknown", e.tag())
+}
+
+// Subtag returns the subtag for which the error occurred.
+func (e ValueError) Subtag() string {
+	return string(e.tag())
+}
+
+// scanner is used to scan BCP 47 tokens, which are separated by _ or -.
+type scanner struct {
+	b     []byte
+	bytes [max99thPercentileSize]byte
+	token []byte
+	start int // start position of the current token
+	end   int // end position of the current token
+	next  int // next point for scan
+	err   error
+	done  bool
+}
+
+func makeScannerString(s string) scanner {
+	scan := scanner{}
+	if len(s) <= len(scan.bytes) {
+		scan.b = scan.bytes[:copy(scan.bytes[:], s)]
+	} else {
+		scan.b = []byte(s)
+	}
+	scan.init()
+	return scan
+}
+
+// makeScanner returns a scanner using b as the input buffer.
+// b is not copied and may be modified by the scanner routines.
+func makeScanner(b []byte) scanner {
+	scan := scanner{b: b}
+	scan.init()
+	return scan
+}
+
+func (s *scanner) init() {
+	for i, c := range s.b {
+		if c == '_' {
+			s.b[i] = '-'
+		}
+	}
+	s.scan()
+}
+
+// restToLower converts the string between start and end to lower case.
+func (s *scanner) toLower(start, end int) {
+	for i := start; i < end; i++ {
+		c := s.b[i]
+		if 'A' <= c && c <= 'Z' {
+			s.b[i] += 'a' - 'A'
+		}
+	}
+}
+
+func (s *scanner) setError(e error) {
+	if s.err == nil || (e == ErrSyntax && s.err != ErrSyntax) {
+		s.err = e
+	}
+}
+
+// resizeRange shrinks or grows the array at position oldStart such that
+// a new string of size newSize can fit between oldStart and oldEnd.
+// Sets the scan point to after the resized range.
+func (s *scanner) resizeRange(oldStart, oldEnd, newSize int) {
+	s.start = oldStart
+	if end := oldStart + newSize; end != oldEnd {
+		diff := end - oldEnd
+		if end < cap(s.b) {
+			b := make([]byte, len(s.b)+diff)
+			copy(b, s.b[:oldStart])
+			copy(b[end:], s.b[oldEnd:])
+			s.b = b
+		} else {
+			s.b = append(s.b[end:], s.b[oldEnd:]...)
+		}
+		s.next = end + (s.next - s.end)
+		s.end = end
+	}
+}
+
+// replace replaces the current token with repl.
+func (s *scanner) replace(repl string) {
+	s.resizeRange(s.start, s.end, len(repl))
+	copy(s.b[s.start:], repl)
+}
+
+// gobble removes the current token from the input.
+// Caller must call scan after calling gobble.
+func (s *scanner) gobble(e error) {
+	s.setError(e)
+	if s.start == 0 {
+		s.b = s.b[:+copy(s.b, s.b[s.next:])]
+		s.end = 0
+	} else {
+		s.b = s.b[:s.start-1+copy(s.b[s.start-1:], s.b[s.end:])]
+		s.end = s.start - 1
+	}
+	s.next = s.start
+}
+
+// deleteRange removes the given range from s.b before the current token.
+func (s *scanner) deleteRange(start, end int) {
+	s.b = s.b[:start+copy(s.b[start:], s.b[end:])]
+	diff := end - start
+	s.next -= diff
+	s.start -= diff
+	s.end -= diff
+}
+
+// scan parses the next token of a BCP 47 string.  Tokens that are larger
+// than 8 characters or include non-alphanumeric characters result in an error
+// and are gobbled and removed from the output.
+// It returns the end position of the last token consumed.
+func (s *scanner) scan() (end int) {
+	end = s.end
+	s.token = nil
+	for s.start = s.next; s.next < len(s.b); {
+		i := bytes.IndexByte(s.b[s.next:], '-')
+		if i == -1 {
+			s.end = len(s.b)
+			s.next = len(s.b)
+			i = s.end - s.start
+		} else {
+			s.end = s.next + i
+			s.next = s.end + 1
+		}
+		token := s.b[s.start:s.end]
+		if i < 1 || i > 8 || !isAlphaNum(token) {
+			s.gobble(ErrSyntax)
+			continue
+		}
+		s.token = token
+		return end
+	}
+	if n := len(s.b); n > 0 && s.b[n-1] == '-' {
+		s.setError(ErrSyntax)
+		s.b = s.b[:len(s.b)-1]
+	}
+	s.done = true
+	return end
+}
+
+// acceptMinSize parses multiple tokens of the given size or greater.
+// It returns the end position of the last token consumed.
+func (s *scanner) acceptMinSize(min int) (end int) {
+	end = s.end
+	s.scan()
+	for ; len(s.token) >= min; s.scan() {
+		end = s.end
+	}
+	return end
+}
+
+// Parse parses the given BCP 47 string and returns a valid Tag. If parsing
+// failed it returns an error and any part of the tag that could be parsed.
+// If parsing succeeded but an unknown value was found, it returns
+// ValueError. The Tag returned in this case is just stripped of the unknown
+// value. All other values are preserved. It accepts tags in the BCP 47 format
+// and extensions to this standard defined in
+// https://www.unicode.org/reports/tr35/#Unicode_Language_and_Locale_Identifiers.
+func Parse(s string) (t Tag, err error) {
+	// TODO: consider supporting old-style locale key-value pairs.
+	if s == "" {
+		return Und, ErrSyntax
+	}
+	if len(s) <= maxAltTaglen {
+		b := [maxAltTaglen]byte{}
+		for i, c := range s {
+			// Generating invalid UTF-8 is okay as it won't match.
+			if 'A' <= c && c <= 'Z' {
+				c += 'a' - 'A'
+			} else if c == '_' {
+				c = '-'
+			}
+			b[i] = byte(c)
+		}
+		if t, ok := grandfathered(b); ok {
+			return t, nil
+		}
+	}
+	scan := makeScannerString(s)
+	return parse(&scan, s)
+}
+
+func parse(scan *scanner, s string) (t Tag, err error) {
+	t = Und
+	var end int
+	if n := len(scan.token); n <= 1 {
+		scan.toLower(0, len(scan.b))
+		if n == 0 || scan.token[0] != 'x' {
+			return t, ErrSyntax
+		}
+		end = parseExtensions(scan)
+	} else if n >= 4 {
+		return Und, ErrSyntax
+	} else { // the usual case
+		t, end = parseTag(scan)
+		if n := len(scan.token); n == 1 {
+			t.pExt = uint16(end)
+			end = parseExtensions(scan)
+		} else if end < len(scan.b) {
+			scan.setError(ErrSyntax)
+			scan.b = scan.b[:end]
+		}
+	}
+	if int(t.pVariant) < len(scan.b) {
+		if end < len(s) {
+			s = s[:end]
+		}
+		if len(s) > 0 && tag.Compare(s, scan.b) == 0 {
+			t.str = s
+		} else {
+			t.str = string(scan.b)
+		}
+	} else {
+		t.pVariant, t.pExt = 0, 0
+	}
+	return t, scan.err
+}
+
+// parseTag parses language, script, region and variants.
+// It returns a Tag and the end position in the input that was parsed.
+func parseTag(scan *scanner) (t Tag, end int) {
+	var e error
+	// TODO: set an error if an unknown lang, script or region is encountered.
+	t.LangID, e = getLangID(scan.token)
+	scan.setError(e)
+	scan.replace(t.LangID.String())
+	langStart := scan.start
+	end = scan.scan()
+	for len(scan.token) == 3 && isAlpha(scan.token[0]) {
+		// From http://tools.ietf.org/html/bcp47, <lang>-<extlang> tags are equivalent
+		// to a tag of the form <extlang>.
+		lang, e := getLangID(scan.token)
+		if lang != 0 {
+			t.LangID = lang
+			copy(scan.b[langStart:], lang.String())
+			scan.b[langStart+3] = '-'
+			scan.start = langStart + 4
+		}
+		scan.gobble(e)
+		end = scan.scan()
+	}
+	if len(scan.token) == 4 && isAlpha(scan.token[0]) {
+		t.ScriptID, e = getScriptID(script, scan.token)
+		if t.ScriptID == 0 {
+			scan.gobble(e)
+		}
+		end = scan.scan()
+	}
+	if n := len(scan.token); n >= 2 && n <= 3 {
+		t.RegionID, e = getRegionID(scan.token)
+		if t.RegionID == 0 {
+			scan.gobble(e)
+		} else {
+			scan.replace(t.RegionID.String())
+		}
+		end = scan.scan()
+	}
+	scan.toLower(scan.start, len(scan.b))
+	t.pVariant = byte(end)
+	end = parseVariants(scan, end, t)
+	t.pExt = uint16(end)
+	return t, end
+}
+
+var separator = []byte{'-'}
+
+// parseVariants scans tokens as long as each token is a valid variant string.
+// Duplicate variants are removed.
+func parseVariants(scan *scanner, end int, t Tag) int {
+	start := scan.start
+	varIDBuf := [4]uint8{}
+	variantBuf := [4][]byte{}
+	varID := varIDBuf[:0]
+	variant := variantBuf[:0]
+	last := -1
+	needSort := false
+	for ; len(scan.token) >= 4; scan.scan() {
+		// TODO: measure the impact of needing this conversion and redesign
+		// the data structure if there is an issue.
+		v, ok := variantIndex[string(scan.token)]
+		if !ok {
+			// unknown variant
+			// TODO: allow user-defined variants?
+			scan.gobble(NewValueError(scan.token))
+			continue
+		}
+		varID = append(varID, v)
+		variant = append(variant, scan.token)
+		if !needSort {
+			if last < int(v) {
+				last = int(v)
+			} else {
+				needSort = true
+				// There is no legal combinations of more than 7 variants
+				// (and this is by no means a useful sequence).
+				const maxVariants = 8
+				if len(varID) > maxVariants {
+					break
+				}
+			}
+		}
+		end = scan.end
+	}
+	if needSort {
+		sort.Sort(variantsSort{varID, variant})
+		k, l := 0, -1
+		for i, v := range varID {
+			w := int(v)
+			if l == w {
+				// Remove duplicates.
+				continue
+			}
+			varID[k] = varID[i]
+			variant[k] = variant[i]
+			k++
+			l = w
+		}
+		if str := bytes.Join(variant[:k], separator); len(str) == 0 {
+			end = start - 1
+		} else {
+			scan.resizeRange(start, end, len(str))
+			copy(scan.b[scan.start:], str)
+			end = scan.end
+		}
+	}
+	return end
+}
+
+type variantsSort struct {
+	i []uint8
+	v [][]byte
+}
+
+func (s variantsSort) Len() int {
+	return len(s.i)
+}
+
+func (s variantsSort) Swap(i, j int) {
+	s.i[i], s.i[j] = s.i[j], s.i[i]
+	s.v[i], s.v[j] = s.v[j], s.v[i]
+}
+
+func (s variantsSort) Less(i, j int) bool {
+	return s.i[i] < s.i[j]
+}
+
+type bytesSort struct {
+	b [][]byte
+	n int // first n bytes to compare
+}
+
+func (b bytesSort) Len() int {
+	return len(b.b)
+}
+
+func (b bytesSort) Swap(i, j int) {
+	b.b[i], b.b[j] = b.b[j], b.b[i]
+}
+
+func (b bytesSort) Less(i, j int) bool {
+	for k := 0; k < b.n; k++ {
+		if b.b[i][k] == b.b[j][k] {
+			continue
+		}
+		return b.b[i][k] < b.b[j][k]
+	}
+	return false
+}
+
+// parseExtensions parses and normalizes the extensions in the buffer.
+// It returns the last position of scan.b that is part of any extension.
+// It also trims scan.b to remove excess parts accordingly.
+func parseExtensions(scan *scanner) int {
+	start := scan.start
+	exts := [][]byte{}
+	private := []byte{}
+	end := scan.end
+	for len(scan.token) == 1 {
+		extStart := scan.start
+		ext := scan.token[0]
+		end = parseExtension(scan)
+		extension := scan.b[extStart:end]
+		if len(extension) < 3 || (ext != 'x' && len(extension) < 4) {
+			scan.setError(ErrSyntax)
+			end = extStart
+			continue
+		} else if start == extStart && (ext == 'x' || scan.start == len(scan.b)) {
+			scan.b = scan.b[:end]
+			return end
+		} else if ext == 'x' {
+			private = extension
+			break
+		}
+		exts = append(exts, extension)
+	}
+	sort.Sort(bytesSort{exts, 1})
+	if len(private) > 0 {
+		exts = append(exts, private)
+	}
+	scan.b = scan.b[:start]
+	if len(exts) > 0 {
+		scan.b = append(scan.b, bytes.Join(exts, separator)...)
+	} else if start > 0 {
+		// Strip trailing '-'.
+		scan.b = scan.b[:start-1]
+	}
+	return end
+}
+
+// parseExtension parses a single extension and returns the position of
+// the extension end.
+func parseExtension(scan *scanner) int {
+	start, end := scan.start, scan.end
+	switch scan.token[0] {
+	case 'u':
+		attrStart := end
+		scan.scan()
+		for last := []byte{}; len(scan.token) > 2; scan.scan() {
+			if bytes.Compare(scan.token, last) != -1 {
+				// Attributes are unsorted. Start over from scratch.
+				p := attrStart + 1
+				scan.next = p
+				attrs := [][]byte{}
+				for scan.scan(); len(scan.token) > 2; scan.scan() {
+					attrs = append(attrs, scan.token)
+					end = scan.end
+				}
+				sort.Sort(bytesSort{attrs, 3})
+				copy(scan.b[p:], bytes.Join(attrs, separator))
+				break
+			}
+			last = scan.token
+			end = scan.end
+		}
+		var last, key []byte
+		for attrEnd := end; len(scan.token) == 2; last = key {
+			key = scan.token
+			keyEnd := scan.end
+			end = scan.acceptMinSize(3)
+			// TODO: check key value validity
+			if keyEnd == end || bytes.Compare(key, last) != 1 {
+				// We have an invalid key or the keys are not sorted.
+				// Start scanning keys from scratch and reorder.
+				p := attrEnd + 1
+				scan.next = p
+				keys := [][]byte{}
+				for scan.scan(); len(scan.token) == 2; {
+					keyStart, keyEnd := scan.start, scan.end
+					end = scan.acceptMinSize(3)
+					if keyEnd != end {
+						keys = append(keys, scan.b[keyStart:end])
+					} else {
+						scan.setError(ErrSyntax)
+						end = keyStart
+					}
+				}
+				sort.Stable(bytesSort{keys, 2})
+				if n := len(keys); n > 0 {
+					k := 0
+					for i := 1; i < n; i++ {
+						if !bytes.Equal(keys[k][:2], keys[i][:2]) {
+							k++
+							keys[k] = keys[i]
+						} else if !bytes.Equal(keys[k], keys[i]) {
+							scan.setError(ErrDuplicateKey)
+						}
+					}
+					keys = keys[:k+1]
+				}
+				reordered := bytes.Join(keys, separator)
+				if e := p + len(reordered); e < end {
+					scan.deleteRange(e, end)
+					end = e
+				}
+				copy(scan.b[p:], reordered)
+				break
+			}
+		}
+	case 't':
+		scan.scan()
+		if n := len(scan.token); n >= 2 && n <= 3 && isAlpha(scan.token[1]) {
+			_, end = parseTag(scan)
+			scan.toLower(start, end)
+		}
+		for len(scan.token) == 2 && !isAlpha(scan.token[1]) {
+			end = scan.acceptMinSize(3)
+		}
+	case 'x':
+		end = scan.acceptMinSize(1)
+	default:
+		end = scan.acceptMinSize(2)
+	}
+	return end
+}
+
+// getExtension returns the name, body and end position of the extension.
+func getExtension(s string, p int) (end int, ext string) {
+	if s[p] == '-' {
+		p++
+	}
+	if s[p] == 'x' {
+		return len(s), s[p:]
+	}
+	end = nextExtension(s, p)
+	return end, s[p:end]
+}
+
+// nextExtension finds the next extension within the string, searching
+// for the -<char>- pattern from position p.
+// In the fast majority of cases, language tags will have at most
+// one extension and extensions tend to be small.
+func nextExtension(s string, p int) int {
+	for n := len(s) - 3; p < n; {
+		if s[p] == '-' {
+			if s[p+2] == '-' {
+				return p
+			}
+			p += 3
+		} else {
+			p++
+		}
+	}
+	return len(s)
+}

vendor/golang.org/x/text/internal/language/tags.go

@@ -0,0 +1,48 @@
+// 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
+
+// MustParse is like Parse, but panics if the given BCP 47 tag cannot be parsed.
+// It simplifies safe initialization of Tag values.
+func MustParse(s string) Tag {
+	t, err := Parse(s)
+	if err != nil {
+		panic(err)
+	}
+	return t
+}
+
+// MustParseBase is like ParseBase, but panics if the given base cannot be parsed.
+// It simplifies safe initialization of Base values.
+func MustParseBase(s string) Language {
+	b, err := ParseBase(s)
+	if err != nil {
+		panic(err)
+	}
+	return b
+}
+
+// MustParseScript is like ParseScript, but panics if the given script cannot be
+// parsed. It simplifies safe initialization of Script values.
+func MustParseScript(s string) Script {
+	scr, err := ParseScript(s)
+	if err != nil {
+		panic(err)
+	}
+	return scr
+}
+
+// MustParseRegion is like ParseRegion, but panics if the given region cannot be
+// parsed. It simplifies safe initialization of Region values.
+func MustParseRegion(s string) Region {
+	r, err := ParseRegion(s)
+	if err != nil {
+		panic(err)
+	}
+	return r
+}
+
+// Und is the root language.
+var Und Tag

vendor/golang.org/x/text/internal/tag/tag.go

@@ -0,0 +1,100 @@
+// Copyright 2015 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 tag contains functionality handling tags and related data.
+package tag // import "golang.org/x/text/internal/tag"
+
+import "sort"
+
+// An Index converts tags to a compact numeric value.
+//
+// All elements are of size 4. Tags may be up to 4 bytes long. Excess bytes can
+// be used to store additional information about the tag.
+type Index string
+
+// Elem returns the element data at the given index.
+func (s Index) Elem(x int) string {
+	return string(s[x*4 : x*4+4])
+}
+
+// Index reports the index of the given key or -1 if it could not be found.
+// Only the first len(key) bytes from the start of the 4-byte entries will be
+// considered for the search and the first match in Index will be returned.
+func (s Index) Index(key []byte) int {
+	n := len(key)
+	// search the index of the first entry with an equal or higher value than
+	// key in s.
+	index := sort.Search(len(s)/4, func(i int) bool {
+		return cmp(s[i*4:i*4+n], key) != -1
+	})
+	i := index * 4
+	if cmp(s[i:i+len(key)], key) != 0 {
+		return -1
+	}
+	return index
+}
+
+// Next finds the next occurrence of key after index x, which must have been
+// obtained from a call to Index using the same key. It returns x+1 or -1.
+func (s Index) Next(key []byte, x int) int {
+	if x++; x*4 < len(s) && cmp(s[x*4:x*4+len(key)], key) == 0 {
+		return x
+	}
+	return -1
+}
+
+// cmp returns an integer comparing a and b lexicographically.
+func cmp(a Index, b []byte) int {
+	n := len(a)
+	if len(b) < n {
+		n = len(b)
+	}
+	for i, c := range b[:n] {
+		switch {
+		case a[i] > c:
+			return 1
+		case a[i] < c:
+			return -1
+		}
+	}
+	switch {
+	case len(a) < len(b):
+		return -1
+	case len(a) > len(b):
+		return 1
+	}
+	return 0
+}
+
+// Compare returns an integer comparing a and b lexicographically.
+func Compare(a string, b []byte) int {
+	return cmp(Index(a), b)
+}
+
+// FixCase reformats b to the same pattern of cases as form.
+// If returns false if string b is malformed.
+func FixCase(form string, b []byte) bool {
+	if len(form) != len(b) {
+		return false
+	}
+	for i, c := range b {
+		if form[i] <= 'Z' {
+			if c >= 'a' {
+				c -= 'z' - 'Z'
+			}
+			if c < 'A' || 'Z' < c {
+				return false
+			}
+		} else {
+			if c <= 'Z' {
+				c += 'z' - 'Z'
+			}
+			if c < 'a' || 'z' < c {
+				return false
+			}
+		}
+		b[i] = c
+	}
+	return true
+}

vendor/golang.org/x/text/internal/triegen/compact.go

@@ -0,0 +1,58 @@
+// Copyright 2014 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 triegen
+
+// This file defines Compacter and its implementations.
+
+import "io"
+
+// A Compacter generates an alternative, more space-efficient way to store a
+// trie value block. A trie value block holds all possible values for the last
+// byte of a UTF-8 encoded rune. Excluding ASCII characters, a trie value block
+// always has 64 values, as a UTF-8 encoding ends with a byte in [0x80, 0xC0).
+type Compacter interface {
+	// Size returns whether the Compacter could encode the given block as well
+	// as its size in case it can. len(v) is always 64.
+	Size(v []uint64) (sz int, ok bool)
+
+	// Store stores the block using the Compacter's compression method.
+	// It returns a handle with which the block can be retrieved.
+	// len(v) is always 64.
+	Store(v []uint64) uint32
+
+	// Print writes the data structures associated to the given store to w.
+	Print(w io.Writer) error
+
+	// Handler returns the name of a function that gets called during trie
+	// lookup for blocks generated by the Compacter. The function should be of
+	// the form func (n uint32, b byte) uint64, where n is the index returned by
+	// the Compacter's Store method and b is the last byte of the UTF-8
+	// encoding, where 0x80 <= b < 0xC0, for which to do the lookup in the
+	// block.
+	Handler() string
+}
+
+// simpleCompacter is the default Compacter used by builder. It implements a
+// normal trie block.
+type simpleCompacter builder
+
+func (b *simpleCompacter) Size([]uint64) (sz int, ok bool) {
+	return blockSize * b.ValueSize, true
+}
+
+func (b *simpleCompacter) Store(v []uint64) uint32 {
+	h := uint32(len(b.ValueBlocks) - blockOffset)
+	b.ValueBlocks = append(b.ValueBlocks, v)
+	return h
+}
+
+func (b *simpleCompacter) Print(io.Writer) error {
+	// Structures are printed in print.go.
+	return nil
+}
+
+func (b *simpleCompacter) Handler() string {
+	panic("Handler should be special-cased for this Compacter")
+}

vendor/golang.org/x/text/internal/triegen/print.go

@@ -0,0 +1,251 @@
+// Copyright 2014 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 triegen
+
+import (
+	"bytes"
+	"fmt"
+	"io"
+	"strings"
+	"text/template"
+)
+
+// print writes all the data structures as well as the code necessary to use the
+// trie to w.
+func (b *builder) print(w io.Writer) error {
+	b.Stats.NValueEntries = len(b.ValueBlocks) * blockSize
+	b.Stats.NValueBytes = len(b.ValueBlocks) * blockSize * b.ValueSize
+	b.Stats.NIndexEntries = len(b.IndexBlocks) * blockSize
+	b.Stats.NIndexBytes = len(b.IndexBlocks) * blockSize * b.IndexSize
+	b.Stats.NHandleBytes = len(b.Trie) * 2 * b.IndexSize
+
+	// If we only have one root trie, all starter blocks are at position 0 and
+	// we can access the arrays directly.
+	if len(b.Trie) == 1 {
+		// At this point we cannot refer to the generated tables directly.
+		b.ASCIIBlock = b.Name + "Values"
+		b.StarterBlock = b.Name + "Index"
+	} else {
+		// Otherwise we need to have explicit starter indexes in the trie
+		// structure.
+		b.ASCIIBlock = "t.ascii"
+		b.StarterBlock = "t.utf8Start"
+	}
+
+	b.SourceType = "[]byte"
+	if err := lookupGen.Execute(w, b); err != nil {
+		return err
+	}
+
+	b.SourceType = "string"
+	if err := lookupGen.Execute(w, b); err != nil {
+		return err
+	}
+
+	if err := trieGen.Execute(w, b); err != nil {
+		return err
+	}
+
+	for _, c := range b.Compactions {
+		if err := c.c.Print(w); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func printValues(n int, values []uint64) string {
+	w := &bytes.Buffer{}
+	boff := n * blockSize
+	fmt.Fprintf(w, "\t// Block %#x, offset %#x", n, boff)
+	var newline bool
+	for i, v := range values {
+		if i%6 == 0 {
+			newline = true
+		}
+		if v != 0 {
+			if newline {
+				fmt.Fprintf(w, "\n")
+				newline = false
+			}
+			fmt.Fprintf(w, "\t%#02x:%#04x, ", boff+i, v)
+		}
+	}
+	return w.String()
+}
+
+func printIndex(b *builder, nr int, n *node) string {
+	w := &bytes.Buffer{}
+	boff := nr * blockSize
+	fmt.Fprintf(w, "\t// Block %#x, offset %#x", nr, boff)
+	var newline bool
+	for i, c := range n.children {
+		if i%8 == 0 {
+			newline = true
+		}
+		if c != nil {
+			v := b.Compactions[c.index.compaction].Offset + uint32(c.index.index)
+			if v != 0 {
+				if newline {
+					fmt.Fprintf(w, "\n")
+					newline = false
+				}
+				fmt.Fprintf(w, "\t%#02x:%#02x, ", boff+i, v)
+			}
+		}
+	}
+	return w.String()
+}
+
+var (
+	trieGen = template.Must(template.New("trie").Funcs(template.FuncMap{
+		"printValues": printValues,
+		"printIndex":  printIndex,
+		"title":       strings.Title,
+		"dec":         func(x int) int { return x - 1 },
+		"psize": func(n int) string {
+			return fmt.Sprintf("%d bytes (%.2f KiB)", n, float64(n)/1024)
+		},
+	}).Parse(trieTemplate))
+	lookupGen = template.Must(template.New("lookup").Parse(lookupTemplate))
+)
+
+// TODO: consider the return type of lookup. It could be uint64, even if the
+// internal value type is smaller. We will have to verify this with the
+// performance of unicode/norm, which is very sensitive to such changes.
+const trieTemplate = `{{$b := .}}{{$multi := gt (len .Trie) 1}}
+// {{.Name}}Trie. Total size: {{psize .Size}}. Checksum: {{printf "%08x" .Checksum}}.
+type {{.Name}}Trie struct { {{if $multi}}
+	ascii []{{.ValueType}} // index for ASCII bytes
+	utf8Start  []{{.IndexType}} // index for UTF-8 bytes >= 0xC0
+{{end}}}
+
+func new{{title .Name}}Trie(i int) *{{.Name}}Trie { {{if $multi}}
+	h := {{.Name}}TrieHandles[i]
+	return &{{.Name}}Trie{ {{.Name}}Values[uint32(h.ascii)<<6:], {{.Name}}Index[uint32(h.multi)<<6:] }
+}
+
+type {{.Name}}TrieHandle struct {
+	ascii, multi {{.IndexType}}
+}
+
+// {{.Name}}TrieHandles: {{len .Trie}} handles, {{.Stats.NHandleBytes}} bytes
+var {{.Name}}TrieHandles = [{{len .Trie}}]{{.Name}}TrieHandle{
+{{range .Trie}}	{ {{.ASCIIIndex}}, {{.StarterIndex}} }, // {{printf "%08x" .Checksum}}: {{.Name}}
+{{end}}}{{else}}
+	return &{{.Name}}Trie{}
+}
+{{end}}
+// lookupValue determines the type of block n and looks up the value for b.
+func (t *{{.Name}}Trie) lookupValue(n uint32, b byte) {{.ValueType}}{{$last := dec (len .Compactions)}} {
+	switch { {{range $i, $c := .Compactions}}
+		{{if eq $i $last}}default{{else}}case n < {{$c.Cutoff}}{{end}}:{{if ne $i 0}}
+			n -= {{$c.Offset}}{{end}}
+			return {{print $b.ValueType}}({{$c.Handler}}){{end}}
+	}
+}
+
+// {{.Name}}Values: {{len .ValueBlocks}} blocks, {{.Stats.NValueEntries}} entries, {{.Stats.NValueBytes}} bytes
+// The third block is the zero block.
+var {{.Name}}Values = [{{.Stats.NValueEntries}}]{{.ValueType}} {
+{{range $i, $v := .ValueBlocks}}{{printValues $i $v}}
+{{end}}}
+
+// {{.Name}}Index: {{len .IndexBlocks}} blocks, {{.Stats.NIndexEntries}} entries, {{.Stats.NIndexBytes}} bytes
+// Block 0 is the zero block.
+var {{.Name}}Index = [{{.Stats.NIndexEntries}}]{{.IndexType}} {
+{{range $i, $v := .IndexBlocks}}{{printIndex $b $i $v}}
+{{end}}}
+`
+
+// TODO: consider allowing zero-length strings after evaluating performance with
+// unicode/norm.
+const lookupTemplate = `
+// lookup{{if eq .SourceType "string"}}String{{end}} returns the trie value for the first UTF-8 encoding in s and
+// the width in bytes of this encoding. The size will be 0 if s does not
+// hold enough bytes to complete the encoding. len(s) must be greater than 0.
+func (t *{{.Name}}Trie) lookup{{if eq .SourceType "string"}}String{{end}}(s {{.SourceType}}) (v {{.ValueType}}, sz int) {
+	c0 := s[0]
+	switch {
+	case c0 < 0x80: // is ASCII
+		return {{.ASCIIBlock}}[c0], 1
+	case c0 < 0xC2:
+		return 0, 1  // Illegal UTF-8: not a starter, not ASCII.
+	case c0 < 0xE0: // 2-byte UTF-8
+		if len(s) < 2 {
+			return 0, 0
+		}
+		i := {{.StarterBlock}}[c0]
+		c1 := s[1]
+		if c1 < 0x80 || 0xC0 <= c1 {
+			return 0, 1 // Illegal UTF-8: not a continuation byte.
+		}
+		return t.lookupValue(uint32(i), c1), 2
+	case c0 < 0xF0: // 3-byte UTF-8
+		if len(s) < 3 {
+			return 0, 0
+		}
+		i := {{.StarterBlock}}[c0]
+		c1 := s[1]
+		if c1 < 0x80 || 0xC0 <= c1 {
+			return 0, 1 // Illegal UTF-8: not a continuation byte.
+		}
+		o := uint32(i)<<6 + uint32(c1)
+		i = {{.Name}}Index[o]
+		c2 := s[2]
+		if c2 < 0x80 || 0xC0 <= c2 {
+			return 0, 2 // Illegal UTF-8: not a continuation byte.
+		}
+		return t.lookupValue(uint32(i), c2), 3
+	case c0 < 0xF8: // 4-byte UTF-8
+		if len(s) < 4 {
+			return 0, 0
+		}
+		i := {{.StarterBlock}}[c0]
+		c1 := s[1]
+		if c1 < 0x80 || 0xC0 <= c1 {
+			return 0, 1 // Illegal UTF-8: not a continuation byte.
+		}
+		o := uint32(i)<<6 + uint32(c1)
+		i = {{.Name}}Index[o]
+		c2 := s[2]
+		if c2 < 0x80 || 0xC0 <= c2 {
+			return 0, 2 // Illegal UTF-8: not a continuation byte.
+		}
+		o = uint32(i)<<6 + uint32(c2)
+		i = {{.Name}}Index[o]
+		c3 := s[3]
+		if c3 < 0x80 || 0xC0 <= c3 {
+			return 0, 3 // Illegal UTF-8: not a continuation byte.
+		}
+		return t.lookupValue(uint32(i), c3), 4
+	}
+	// Illegal rune
+	return 0, 1
+}
+
+// lookup{{if eq .SourceType "string"}}String{{end}}Unsafe returns the trie value for the first UTF-8 encoding in s.
+// s must start with a full and valid UTF-8 encoded rune.
+func (t *{{.Name}}Trie) lookup{{if eq .SourceType "string"}}String{{end}}Unsafe(s {{.SourceType}}) {{.ValueType}} {
+	c0 := s[0]
+	if c0 < 0x80 { // is ASCII
+		return {{.ASCIIBlock}}[c0]
+	}
+	i := {{.StarterBlock}}[c0]
+	if c0 < 0xE0 { // 2-byte UTF-8
+		return t.lookupValue(uint32(i), s[1])
+	}
+	i = {{.Name}}Index[uint32(i)<<6+uint32(s[1])]
+	if c0 < 0xF0 { // 3-byte UTF-8
+		return t.lookupValue(uint32(i), s[2])
+	}
+	i = {{.Name}}Index[uint32(i)<<6+uint32(s[2])]
+	if c0 < 0xF8 { // 4-byte UTF-8
+		return t.lookupValue(uint32(i), s[3])
+	}
+	return 0
+}
+`

vendor/golang.org/x/text/internal/triegen/triegen.go

@@ -0,0 +1,494 @@
+// Copyright 2014 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 triegen implements a code generator for a trie for associating
+// unsigned integer values with UTF-8 encoded runes.
+//
+// Many of the go.text packages use tries for storing per-rune information.  A
+// trie is especially useful if many of the runes have the same value. If this
+// is the case, many blocks can be expected to be shared allowing for
+// information on many runes to be stored in little space.
+//
+// As most of the lookups are done directly on []byte slices, the tries use the
+// UTF-8 bytes directly for the lookup. This saves a conversion from UTF-8 to
+// runes and contributes a little bit to better performance. It also naturally
+// provides a fast path for ASCII.
+//
+// Space is also an issue. There are many code points defined in Unicode and as
+// a result tables can get quite large. So every byte counts. The triegen
+// package automatically chooses the smallest integer values to represent the
+// tables. Compacters allow further compression of the trie by allowing for
+// alternative representations of individual trie blocks.
+//
+// triegen allows generating multiple tries as a single structure. This is
+// useful when, for example, one wants to generate tries for several languages
+// that have a lot of values in common. Some existing libraries for
+// internationalization store all per-language data as a dynamically loadable
+// chunk. The go.text packages are designed with the assumption that the user
+// typically wants to compile in support for all supported languages, in line
+// with the approach common to Go to create a single standalone binary. The
+// multi-root trie approach can give significant storage savings in this
+// scenario.
+//
+// triegen generates both tables and code. The code is optimized to use the
+// automatically chosen data types. The following code is generated for a Trie
+// or multiple Tries named "foo":
+//	- type fooTrie
+//		The trie type.
+//
+//	- func newFooTrie(x int) *fooTrie
+//		Trie constructor, where x is the index of the trie passed to Gen.
+//
+//	- func (t *fooTrie) lookup(s []byte) (v uintX, sz int)
+//		The lookup method, where uintX is automatically chosen.
+//
+//	- func lookupString, lookupUnsafe and lookupStringUnsafe
+//		Variants of the above.
+//
+//	- var fooValues and fooIndex and any tables generated by Compacters.
+//		The core trie data.
+//
+//	- var fooTrieHandles
+//		Indexes of starter blocks in case of multiple trie roots.
+//
+// It is recommended that users test the generated trie by checking the returned
+// value for every rune. Such exhaustive tests are possible as the number of
+// runes in Unicode is limited.
+package triegen // import "golang.org/x/text/internal/triegen"
+
+// TODO: Arguably, the internally optimized data types would not have to be
+// exposed in the generated API. We could also investigate not generating the
+// code, but using it through a package. We would have to investigate the impact
+// on performance of making such change, though. For packages like unicode/norm,
+// small changes like this could tank performance.
+
+import (
+	"encoding/binary"
+	"fmt"
+	"hash/crc64"
+	"io"
+	"log"
+	"unicode/utf8"
+)
+
+// builder builds a set of tries for associating values with runes. The set of
+// tries can share common index and value blocks.
+type builder struct {
+	Name string
+
+	// ValueType is the type of the trie values looked up.
+	ValueType string
+
+	// ValueSize is the byte size of the ValueType.
+	ValueSize int
+
+	// IndexType is the type of trie index values used for all UTF-8 bytes of
+	// a rune except the last one.
+	IndexType string
+
+	// IndexSize is the byte size of the IndexType.
+	IndexSize int
+
+	// SourceType is used when generating the lookup functions. If the user
+	// requests StringSupport, all lookup functions will be generated for
+	// string input as well.
+	SourceType string
+
+	Trie []*Trie
+
+	IndexBlocks []*node
+	ValueBlocks [][]uint64
+	Compactions []compaction
+	Checksum    uint64
+
+	ASCIIBlock   string
+	StarterBlock string
+
+	indexBlockIdx map[uint64]int
+	valueBlockIdx map[uint64]nodeIndex
+	asciiBlockIdx map[uint64]int
+
+	// Stats are used to fill out the template.
+	Stats struct {
+		NValueEntries int
+		NValueBytes   int
+		NIndexEntries int
+		NIndexBytes   int
+		NHandleBytes  int
+	}
+
+	err error
+}
+
+// A nodeIndex encodes the index of a node, which is defined by the compaction
+// which stores it and an index within the compaction. For internal nodes, the
+// compaction is always 0.
+type nodeIndex struct {
+	compaction int
+	index      int
+}
+
+// compaction keeps track of stats used for the compaction.
+type compaction struct {
+	c         Compacter
+	blocks    []*node
+	maxHandle uint32
+	totalSize int
+
+	// Used by template-based generator and thus exported.
+	Cutoff  uint32
+	Offset  uint32
+	Handler string
+}
+
+func (b *builder) setError(err error) {
+	if b.err == nil {
+		b.err = err
+	}
+}
+
+// An Option can be passed to Gen.
+type Option func(b *builder) error
+
+// Compact configures the trie generator to use the given Compacter.
+func Compact(c Compacter) Option {
+	return func(b *builder) error {
+		b.Compactions = append(b.Compactions, compaction{
+			c:       c,
+			Handler: c.Handler() + "(n, b)"})
+		return nil
+	}
+}
+
+// Gen writes Go code for a shared trie lookup structure to w for the given
+// Tries. The generated trie type will be called nameTrie. newNameTrie(x) will
+// return the *nameTrie for tries[x]. A value can be looked up by using one of
+// the various lookup methods defined on nameTrie. It returns the table size of
+// the generated trie.
+func Gen(w io.Writer, name string, tries []*Trie, opts ...Option) (sz int, err error) {
+	// The index contains two dummy blocks, followed by the zero block. The zero
+	// block is at offset 0x80, so that the offset for the zero block for
+	// continuation bytes is 0.
+	b := &builder{
+		Name:        name,
+		Trie:        tries,
+		IndexBlocks: []*node{{}, {}, {}},
+		Compactions: []compaction{{
+			Handler: name + "Values[n<<6+uint32(b)]",
+		}},
+		// The 0 key in indexBlockIdx and valueBlockIdx is the hash of the zero
+		// block.
+		indexBlockIdx: map[uint64]int{0: 0},
+		valueBlockIdx: map[uint64]nodeIndex{0: {}},
+		asciiBlockIdx: map[uint64]int{},
+	}
+	b.Compactions[0].c = (*simpleCompacter)(b)
+
+	for _, f := range opts {
+		if err := f(b); err != nil {
+			return 0, err
+		}
+	}
+	b.build()
+	if b.err != nil {
+		return 0, b.err
+	}
+	if err = b.print(w); err != nil {
+		return 0, err
+	}
+	return b.Size(), nil
+}
+
+// A Trie represents a single root node of a trie. A builder may build several
+// overlapping tries at once.
+type Trie struct {
+	root *node
+
+	hiddenTrie
+}
+
+// hiddenTrie contains values we want to be visible to the template generator,
+// but hidden from the API documentation.
+type hiddenTrie struct {
+	Name         string
+	Checksum     uint64
+	ASCIIIndex   int
+	StarterIndex int
+}
+
+// NewTrie returns a new trie root.
+func NewTrie(name string) *Trie {
+	return &Trie{
+		&node{
+			children: make([]*node, blockSize),
+			values:   make([]uint64, utf8.RuneSelf),
+		},
+		hiddenTrie{Name: name},
+	}
+}
+
+// Gen is a convenience wrapper around the Gen func passing t as the only trie
+// and uses the name passed to NewTrie. It returns the size of the generated
+// tables.
+func (t *Trie) Gen(w io.Writer, opts ...Option) (sz int, err error) {
+	return Gen(w, t.Name, []*Trie{t}, opts...)
+}
+
+// node is a node of the intermediate trie structure.
+type node struct {
+	// children holds this node's children. It is always of length 64.
+	// A child node may be nil.
+	children []*node
+
+	// values contains the values of this node. If it is non-nil, this node is
+	// either a root or leaf node:
+	// For root nodes, len(values) == 128 and it maps the bytes in [0x00, 0x7F].
+	// For leaf nodes, len(values) ==  64 and it maps the bytes in [0x80, 0xBF].
+	values []uint64
+
+	index nodeIndex
+}
+
+// Insert associates value with the given rune. Insert will panic if a non-zero
+// value is passed for an invalid rune.
+func (t *Trie) Insert(r rune, value uint64) {
+	if value == 0 {
+		return
+	}
+	s := string(r)
+	if []rune(s)[0] != r && value != 0 {
+		// Note: The UCD tables will always assign what amounts to a zero value
+		// to a surrogate. Allowing a zero value for an illegal rune allows
+		// users to iterate over [0..MaxRune] without having to explicitly
+		// exclude surrogates, which would be tedious.
+		panic(fmt.Sprintf("triegen: non-zero value for invalid rune %U", r))
+	}
+	if len(s) == 1 {
+		// It is a root node value (ASCII).
+		t.root.values[s[0]] = value
+		return
+	}
+
+	n := t.root
+	for ; len(s) > 1; s = s[1:] {
+		if n.children == nil {
+			n.children = make([]*node, blockSize)
+		}
+		p := s[0] % blockSize
+		c := n.children[p]
+		if c == nil {
+			c = &node{}
+			n.children[p] = c
+		}
+		if len(s) > 2 && c.values != nil {
+			log.Fatalf("triegen: insert(%U): found internal node with values", r)
+		}
+		n = c
+	}
+	if n.values == nil {
+		n.values = make([]uint64, blockSize)
+	}
+	if n.children != nil {
+		log.Fatalf("triegen: insert(%U): found leaf node that also has child nodes", r)
+	}
+	n.values[s[0]-0x80] = value
+}
+
+// Size returns the number of bytes the generated trie will take to store. It
+// needs to be exported as it is used in the templates.
+func (b *builder) Size() int {
+	// Index blocks.
+	sz := len(b.IndexBlocks) * blockSize * b.IndexSize
+
+	// Skip the first compaction, which represents the normal value blocks, as
+	// its totalSize does not account for the ASCII blocks, which are managed
+	// separately.
+	sz += len(b.ValueBlocks) * blockSize * b.ValueSize
+	for _, c := range b.Compactions[1:] {
+		sz += c.totalSize
+	}
+
+	// TODO: this computation does not account for the fixed overhead of a using
+	// a compaction, either code or data. As for data, though, the typical
+	// overhead of data is in the order of bytes (2 bytes for cases). Further,
+	// the savings of using a compaction should anyway be substantial for it to
+	// be worth it.
+
+	// For multi-root tries, we also need to account for the handles.
+	if len(b.Trie) > 1 {
+		sz += 2 * b.IndexSize * len(b.Trie)
+	}
+	return sz
+}
+
+func (b *builder) build() {
+	// Compute the sizes of the values.
+	var vmax uint64
+	for _, t := range b.Trie {
+		vmax = maxValue(t.root, vmax)
+	}
+	b.ValueType, b.ValueSize = getIntType(vmax)
+
+	// Compute all block allocations.
+	// TODO: first compute the ASCII blocks for all tries and then the other
+	// nodes. ASCII blocks are more restricted in placement, as they require two
+	// blocks to be placed consecutively. Processing them first may improve
+	// sharing (at least one zero block can be expected to be saved.)
+	for _, t := range b.Trie {
+		b.Checksum += b.buildTrie(t)
+	}
+
+	// Compute the offsets for all the Compacters.
+	offset := uint32(0)
+	for i := range b.Compactions {
+		c := &b.Compactions[i]
+		c.Offset = offset
+		offset += c.maxHandle + 1
+		c.Cutoff = offset
+	}
+
+	// Compute the sizes of indexes.
+	// TODO: different byte positions could have different sizes. So far we have
+	// not found a case where this is beneficial.
+	imax := uint64(b.Compactions[len(b.Compactions)-1].Cutoff)
+	for _, ib := range b.IndexBlocks {
+		if x := uint64(ib.index.index); x > imax {
+			imax = x
+		}
+	}
+	b.IndexType, b.IndexSize = getIntType(imax)
+}
+
+func maxValue(n *node, max uint64) uint64 {
+	if n == nil {
+		return max
+	}
+	for _, c := range n.children {
+		max = maxValue(c, max)
+	}
+	for _, v := range n.values {
+		if max < v {
+			max = v
+		}
+	}
+	return max
+}
+
+func getIntType(v uint64) (string, int) {
+	switch {
+	case v < 1<<8:
+		return "uint8", 1
+	case v < 1<<16:
+		return "uint16", 2
+	case v < 1<<32:
+		return "uint32", 4
+	}
+	return "uint64", 8
+}
+
+const (
+	blockSize = 64
+
+	// Subtract two blocks to offset 0x80, the first continuation byte.
+	blockOffset = 2
+
+	// Subtract three blocks to offset 0xC0, the first non-ASCII starter.
+	rootBlockOffset = 3
+)
+
+var crcTable = crc64.MakeTable(crc64.ISO)
+
+func (b *builder) buildTrie(t *Trie) uint64 {
+	n := t.root
+
+	// Get the ASCII offset. For the first trie, the ASCII block will be at
+	// position 0.
+	hasher := crc64.New(crcTable)
+	binary.Write(hasher, binary.BigEndian, n.values)
+	hash := hasher.Sum64()
+
+	v, ok := b.asciiBlockIdx[hash]
+	if !ok {
+		v = len(b.ValueBlocks)
+		b.asciiBlockIdx[hash] = v
+
+		b.ValueBlocks = append(b.ValueBlocks, n.values[:blockSize], n.values[blockSize:])
+		if v == 0 {
+			// Add the zero block at position 2 so that it will be assigned a
+			// zero reference in the lookup blocks.
+			// TODO: always do this? This would allow us to remove a check from
+			// the trie lookup, but at the expense of extra space. Analyze
+			// performance for unicode/norm.
+			b.ValueBlocks = append(b.ValueBlocks, make([]uint64, blockSize))
+		}
+	}
+	t.ASCIIIndex = v
+
+	// Compute remaining offsets.
+	t.Checksum = b.computeOffsets(n, true)
+	// We already subtracted the normal blockOffset from the index. Subtract the
+	// difference for starter bytes.
+	t.StarterIndex = n.index.index - (rootBlockOffset - blockOffset)
+	return t.Checksum
+}
+
+func (b *builder) computeOffsets(n *node, root bool) uint64 {
+	// For the first trie, the root lookup block will be at position 3, which is
+	// the offset for UTF-8 non-ASCII starter bytes.
+	first := len(b.IndexBlocks) == rootBlockOffset
+	if first {
+		b.IndexBlocks = append(b.IndexBlocks, n)
+	}
+
+	// We special-case the cases where all values recursively are 0. This allows
+	// for the use of a zero block to which all such values can be directed.
+	hash := uint64(0)
+	if n.children != nil || n.values != nil {
+		hasher := crc64.New(crcTable)
+		for _, c := range n.children {
+			var v uint64
+			if c != nil {
+				v = b.computeOffsets(c, false)
+			}
+			binary.Write(hasher, binary.BigEndian, v)
+		}
+		binary.Write(hasher, binary.BigEndian, n.values)
+		hash = hasher.Sum64()
+	}
+
+	if first {
+		b.indexBlockIdx[hash] = rootBlockOffset - blockOffset
+	}
+
+	// Compacters don't apply to internal nodes.
+	if n.children != nil {
+		v, ok := b.indexBlockIdx[hash]
+		if !ok {
+			v = len(b.IndexBlocks) - blockOffset
+			b.IndexBlocks = append(b.IndexBlocks, n)
+			b.indexBlockIdx[hash] = v
+		}
+		n.index = nodeIndex{0, v}
+	} else {
+		h, ok := b.valueBlockIdx[hash]
+		if !ok {
+			bestI, bestSize := 0, blockSize*b.ValueSize
+			for i, c := range b.Compactions[1:] {
+				if sz, ok := c.c.Size(n.values); ok && bestSize > sz {
+					bestI, bestSize = i+1, sz
+				}
+			}
+			c := &b.Compactions[bestI]
+			c.totalSize += bestSize
+			v := c.c.Store(n.values)
+			if c.maxHandle < v {
+				c.maxHandle = v
+			}
+			h = nodeIndex{bestI, int(v)}
+			b.valueBlockIdx[hash] = h
+		}
+		n.index = h
+	}
+	return hash
+}

vendor/golang.org/x/text/internal/ucd/ucd.go

@@ -0,0 +1,371 @@
+// Copyright 2014 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 ucd provides a parser for Unicode Character Database files, the
+// format of which is defined in https://www.unicode.org/reports/tr44/. See
+// https://www.unicode.org/Public/UCD/latest/ucd/ for example files.
+//
+// It currently does not support substitutions of missing fields.
+package ucd // import "golang.org/x/text/internal/ucd"
+
+import (
+	"bufio"
+	"errors"
+	"fmt"
+	"io"
+	"log"
+	"regexp"
+	"strconv"
+	"strings"
+)
+
+// UnicodeData.txt fields.
+const (
+	CodePoint = iota
+	Name
+	GeneralCategory
+	CanonicalCombiningClass
+	BidiClass
+	DecompMapping
+	DecimalValue
+	DigitValue
+	NumericValue
+	BidiMirrored
+	Unicode1Name
+	ISOComment
+	SimpleUppercaseMapping
+	SimpleLowercaseMapping
+	SimpleTitlecaseMapping
+)
+
+// Parse calls f for each entry in the given reader of a UCD file. It will close
+// the reader upon return. It will call log.Fatal if any error occurred.
+//
+// This implements the most common usage pattern of using Parser.
+func Parse(r io.ReadCloser, f func(p *Parser)) {
+	defer r.Close()
+
+	p := New(r)
+	for p.Next() {
+		f(p)
+	}
+	if err := p.Err(); err != nil {
+		r.Close() // os.Exit will cause defers not to be called.
+		log.Fatal(err)
+	}
+}
+
+// An Option is used to configure a Parser.
+type Option func(p *Parser)
+
+func keepRanges(p *Parser) {
+	p.keepRanges = true
+}
+
+var (
+	// KeepRanges prevents the expansion of ranges. The raw ranges can be
+	// obtained by calling Range(0) on the parser.
+	KeepRanges Option = keepRanges
+)
+
+// The Part option register a handler for lines starting with a '@'. The text
+// after a '@' is available as the first field. Comments are handled as usual.
+func Part(f func(p *Parser)) Option {
+	return func(p *Parser) {
+		p.partHandler = f
+	}
+}
+
+// The CommentHandler option passes comments that are on a line by itself to
+// a given handler.
+func CommentHandler(f func(s string)) Option {
+	return func(p *Parser) {
+		p.commentHandler = f
+	}
+}
+
+// A Parser parses Unicode Character Database (UCD) files.
+type Parser struct {
+	scanner *bufio.Scanner
+
+	keepRanges bool // Don't expand rune ranges in field 0.
+
+	err     error
+	comment string
+	field   []string
+	// parsedRange is needed in case Range(0) is called more than once for one
+	// field. In some cases this requires scanning ahead.
+	line                 int
+	parsedRange          bool
+	rangeStart, rangeEnd rune
+
+	partHandler    func(p *Parser)
+	commentHandler func(s string)
+}
+
+func (p *Parser) setError(err error, msg string) {
+	if p.err == nil && err != nil {
+		if msg == "" {
+			p.err = fmt.Errorf("ucd:line:%d: %v", p.line, err)
+		} else {
+			p.err = fmt.Errorf("ucd:line:%d:%s: %v", p.line, msg, err)
+		}
+	}
+}
+
+func (p *Parser) getField(i int) string {
+	if i >= len(p.field) {
+		return ""
+	}
+	return p.field[i]
+}
+
+// Err returns a non-nil error if any error occurred during parsing.
+func (p *Parser) Err() error {
+	return p.err
+}
+
+// New returns a Parser for the given Reader.
+func New(r io.Reader, o ...Option) *Parser {
+	p := &Parser{
+		scanner: bufio.NewScanner(r),
+	}
+	for _, f := range o {
+		f(p)
+	}
+	return p
+}
+
+// Next parses the next line in the file. It returns true if a line was parsed
+// and false if it reached the end of the file.
+func (p *Parser) Next() bool {
+	if !p.keepRanges && p.rangeStart < p.rangeEnd {
+		p.rangeStart++
+		return true
+	}
+	p.comment = ""
+	p.field = p.field[:0]
+	p.parsedRange = false
+
+	for p.scanner.Scan() && p.err == nil {
+		p.line++
+		s := p.scanner.Text()
+		if s == "" {
+			continue
+		}
+		if s[0] == '#' {
+			if p.commentHandler != nil {
+				p.commentHandler(strings.TrimSpace(s[1:]))
+			}
+			continue
+		}
+
+		// Parse line
+		if i := strings.IndexByte(s, '#'); i != -1 {
+			p.comment = strings.TrimSpace(s[i+1:])
+			s = s[:i]
+		}
+		if s[0] == '@' {
+			if p.partHandler != nil {
+				p.field = append(p.field, strings.TrimSpace(s[1:]))
+				p.partHandler(p)
+				p.field = p.field[:0]
+			}
+			p.comment = ""
+			continue
+		}
+		for {
+			i := strings.IndexByte(s, ';')
+			if i == -1 {
+				p.field = append(p.field, strings.TrimSpace(s))
+				break
+			}
+			p.field = append(p.field, strings.TrimSpace(s[:i]))
+			s = s[i+1:]
+		}
+		if !p.keepRanges {
+			p.rangeStart, p.rangeEnd = p.getRange(0)
+		}
+		return true
+	}
+	p.setError(p.scanner.Err(), "scanner failed")
+	return false
+}
+
+func parseRune(b string) (rune, error) {
+	if len(b) > 2 && b[0] == 'U' && b[1] == '+' {
+		b = b[2:]
+	}
+	x, err := strconv.ParseUint(b, 16, 32)
+	return rune(x), err
+}
+
+func (p *Parser) parseRune(s string) rune {
+	x, err := parseRune(s)
+	p.setError(err, "failed to parse rune")
+	return x
+}
+
+// Rune parses and returns field i as a rune.
+func (p *Parser) Rune(i int) rune {
+	if i > 0 || p.keepRanges {
+		return p.parseRune(p.getField(i))
+	}
+	return p.rangeStart
+}
+
+// Runes interprets and returns field i as a sequence of runes.
+func (p *Parser) Runes(i int) (runes []rune) {
+	add := func(s string) {
+		if s = strings.TrimSpace(s); len(s) > 0 {
+			runes = append(runes, p.parseRune(s))
+		}
+	}
+	for b := p.getField(i); ; {
+		i := strings.IndexByte(b, ' ')
+		if i == -1 {
+			add(b)
+			break
+		}
+		add(b[:i])
+		b = b[i+1:]
+	}
+	return
+}
+
+var (
+	errIncorrectLegacyRange = errors.New("ucd: unmatched <* First>")
+
+	// reRange matches one line of a legacy rune range.
+	reRange = regexp.MustCompile("^([0-9A-F]*);<([^,]*), ([^>]*)>(.*)$")
+)
+
+// Range parses and returns field i as a rune range. A range is inclusive at
+// both ends. If the field only has one rune, first and last will be identical.
+// It supports the legacy format for ranges used in UnicodeData.txt.
+func (p *Parser) Range(i int) (first, last rune) {
+	if !p.keepRanges {
+		return p.rangeStart, p.rangeStart
+	}
+	return p.getRange(i)
+}
+
+func (p *Parser) getRange(i int) (first, last rune) {
+	b := p.getField(i)
+	if k := strings.Index(b, ".."); k != -1 {
+		return p.parseRune(b[:k]), p.parseRune(b[k+2:])
+	}
+	// The first field may not be a rune, in which case we may ignore any error
+	// and set the range as 0..0.
+	x, err := parseRune(b)
+	if err != nil {
+		// Disable range parsing henceforth. This ensures that an error will be
+		// returned if the user subsequently will try to parse this field as
+		// a Rune.
+		p.keepRanges = true
+	}
+	// Special case for UnicodeData that was retained for backwards compatibility.
+	if i == 0 && len(p.field) > 1 && strings.HasSuffix(p.field[1], "First>") {
+		if p.parsedRange {
+			return p.rangeStart, p.rangeEnd
+		}
+		mf := reRange.FindStringSubmatch(p.scanner.Text())
+		p.line++
+		if mf == nil || !p.scanner.Scan() {
+			p.setError(errIncorrectLegacyRange, "")
+			return x, x
+		}
+		// Using Bytes would be more efficient here, but Text is a lot easier
+		// and this is not a frequent case.
+		ml := reRange.FindStringSubmatch(p.scanner.Text())
+		if ml == nil || mf[2] != ml[2] || ml[3] != "Last" || mf[4] != ml[4] {
+			p.setError(errIncorrectLegacyRange, "")
+			return x, x
+		}
+		p.rangeStart, p.rangeEnd = x, p.parseRune(p.scanner.Text()[:len(ml[1])])
+		p.parsedRange = true
+		return p.rangeStart, p.rangeEnd
+	}
+	return x, x
+}
+
+// bools recognizes all valid UCD boolean values.
+var bools = map[string]bool{
+	"":      false,
+	"N":     false,
+	"No":    false,
+	"F":     false,
+	"False": false,
+	"Y":     true,
+	"Yes":   true,
+	"T":     true,
+	"True":  true,
+}
+
+// Bool parses and returns field i as a boolean value.
+func (p *Parser) Bool(i int) bool {
+	f := p.getField(i)
+	for s, v := range bools {
+		if f == s {
+			return v
+		}
+	}
+	p.setError(strconv.ErrSyntax, "error parsing bool")
+	return false
+}
+
+// Int parses and returns field i as an integer value.
+func (p *Parser) Int(i int) int {
+	x, err := strconv.ParseInt(string(p.getField(i)), 10, 64)
+	p.setError(err, "error parsing int")
+	return int(x)
+}
+
+// Uint parses and returns field i as an unsigned integer value.
+func (p *Parser) Uint(i int) uint {
+	x, err := strconv.ParseUint(string(p.getField(i)), 10, 64)
+	p.setError(err, "error parsing uint")
+	return uint(x)
+}
+
+// Float parses and returns field i as a decimal value.
+func (p *Parser) Float(i int) float64 {
+	x, err := strconv.ParseFloat(string(p.getField(i)), 64)
+	p.setError(err, "error parsing float")
+	return x
+}
+
+// String parses and returns field i as a string value.
+func (p *Parser) String(i int) string {
+	return string(p.getField(i))
+}
+
+// Strings parses and returns field i as a space-separated list of strings.
+func (p *Parser) Strings(i int) []string {
+	ss := strings.Split(string(p.getField(i)), " ")
+	for i, s := range ss {
+		ss[i] = strings.TrimSpace(s)
+	}
+	return ss
+}
+
+// Comment returns the comments for the current line.
+func (p *Parser) Comment() string {
+	return string(p.comment)
+}
+
+var errUndefinedEnum = errors.New("ucd: undefined enum value")
+
+// Enum interprets and returns field i as a value that must be one of the values
+// in enum.
+func (p *Parser) Enum(i int, enum ...string) string {
+	f := p.getField(i)
+	for _, s := range enum {
+		if f == s {
+			return s
+		}
+	}
+	p.setError(errUndefinedEnum, "error parsing enum")
+	return ""
+}