Added all required dependencies

vendor/golang.org/x/text/LICENSE

@@ -0,0 +1,27 @@
+Copyright (c) 2009 The Go Authors. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+   * Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+   * Redistributions in binary form must reproduce the above
+copyright notice, this list of conditions and the following disclaimer
+in the documentation and/or other materials provided with the
+distribution.
+   * Neither the name of Google Inc. nor the names of its
+contributors may be used to endorse or promote products derived from
+this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

vendor/golang.org/x/text/PATENTS

@@ -0,0 +1,22 @@
+Additional IP Rights Grant (Patents)
+
+"This implementation" means the copyrightable works distributed by
+Google as part of the Go project.
+
+Google hereby grants to You a perpetual, worldwide, non-exclusive,
+no-charge, royalty-free, irrevocable (except as stated in this section)
+patent license to make, have made, use, offer to sell, sell, import,
+transfer and otherwise run, modify and propagate the contents of this
+implementation of Go, where such license applies only to those patent
+claims, both currently owned or controlled by Google and acquired in
+the future, licensable by Google that are necessarily infringed by this
+implementation of Go.  This grant does not include claims that would be
+infringed only as a consequence of further modification of this
+implementation.  If you or your agent or exclusive licensee institute or
+order or agree to the institution of patent litigation against any
+entity (including a cross-claim or counterclaim in a lawsuit) alleging
+that this implementation of Go or any code incorporated within this
+implementation of Go constitutes direct or contributory patent
+infringement, or inducement of patent infringement, then any patent
+rights granted to you under this License for this implementation of Go
+shall terminate as of the date such litigation is filed.

vendor/golang.org/x/text/encoding/charmap/charmap.go

@@ -0,0 +1,209 @@
+// 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 maketables.go
+
+// Package charmap provides simple character encodings such as IBM Code Page 437
+// and Windows 1252.
+package charmap // import "golang.org/x/text/encoding/charmap"
+
+import (
+	"unicode/utf8"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/internal"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/transform"
+)
+
+// These encodings vary only in the way clients should interpret them. Their
+// coded character set is identical and a single implementation can be shared.
+var (
+	// ISO8859_6E is the ISO 8859-6E encoding.
+	ISO8859_6E encoding.Encoding = &iso8859_6E
+
+	// ISO8859_6I is the ISO 8859-6I encoding.
+	ISO8859_6I encoding.Encoding = &iso8859_6I
+
+	// ISO8859_8E is the ISO 8859-8E encoding.
+	ISO8859_8E encoding.Encoding = &iso8859_8E
+
+	// ISO8859_8I is the ISO 8859-8I encoding.
+	ISO8859_8I encoding.Encoding = &iso8859_8I
+
+	iso8859_6E = internal.Encoding{
+		ISO8859_6,
+		"ISO-8859-6E",
+		identifier.ISO88596E,
+	}
+
+	iso8859_6I = internal.Encoding{
+		ISO8859_6,
+		"ISO-8859-6I",
+		identifier.ISO88596I,
+	}
+
+	iso8859_8E = internal.Encoding{
+		ISO8859_8,
+		"ISO-8859-8E",
+		identifier.ISO88598E,
+	}
+
+	iso8859_8I = internal.Encoding{
+		ISO8859_8,
+		"ISO-8859-8I",
+		identifier.ISO88598I,
+	}
+)
+
+// All is a list of all defined encodings in this package.
+var All = listAll
+
+// TODO: implement these encodings, in order of importance.
+// ASCII, ISO8859_1:       Rather common. Close to Windows 1252.
+// ISO8859_9:              Close to Windows 1254.
+
+// utf8Enc holds a rune's UTF-8 encoding in data[:len].
+type utf8Enc struct {
+	len  uint8
+	data [3]byte
+}
+
+// charmap describes an 8-bit character set encoding.
+type charmap struct {
+	// name is the encoding's name.
+	name string
+	// mib is the encoding type of this encoder.
+	mib identifier.MIB
+	// asciiSuperset states whether the encoding is a superset of ASCII.
+	asciiSuperset bool
+	// low is the lower bound of the encoded byte for a non-ASCII rune. If
+	// charmap.asciiSuperset is true then this will be 0x80, otherwise 0x00.
+	low uint8
+	// replacement is the encoded replacement character.
+	replacement byte
+	// decode is the map from encoded byte to UTF-8.
+	decode [256]utf8Enc
+	// encoding is the map from runes to encoded bytes. Each entry is a
+	// uint32: the high 8 bits are the encoded byte and the low 24 bits are
+	// the rune. The table entries are sorted by ascending rune.
+	encode [256]uint32
+}
+
+func (m *charmap) NewDecoder() *encoding.Decoder {
+	return &encoding.Decoder{Transformer: charmapDecoder{charmap: m}}
+}
+
+func (m *charmap) NewEncoder() *encoding.Encoder {
+	return &encoding.Encoder{Transformer: charmapEncoder{charmap: m}}
+}
+
+func (m *charmap) String() string {
+	return m.name
+}
+
+func (m *charmap) ID() (mib identifier.MIB, other string) {
+	return m.mib, ""
+}
+
+// charmapDecoder implements transform.Transformer by decoding to UTF-8.
+type charmapDecoder struct {
+	transform.NopResetter
+	charmap *charmap
+}
+
+func (m charmapDecoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	for i, c := range src {
+		if m.charmap.asciiSuperset && c < utf8.RuneSelf {
+			if nDst >= len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			dst[nDst] = c
+			nDst++
+			nSrc = i + 1
+			continue
+		}
+
+		decode := &m.charmap.decode[c]
+		n := int(decode.len)
+		if nDst+n > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		// It's 15% faster to avoid calling copy for these tiny slices.
+		for j := 0; j < n; j++ {
+			dst[nDst] = decode.data[j]
+			nDst++
+		}
+		nSrc = i + 1
+	}
+	return nDst, nSrc, err
+}
+
+// charmapEncoder implements transform.Transformer by encoding from UTF-8.
+type charmapEncoder struct {
+	transform.NopResetter
+	charmap *charmap
+}
+
+func (m charmapEncoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+loop:
+	for nSrc < len(src) {
+		if nDst >= len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		r = rune(src[nSrc])
+
+		// Decode a 1-byte rune.
+		if r < utf8.RuneSelf {
+			if m.charmap.asciiSuperset {
+				nSrc++
+				dst[nDst] = uint8(r)
+				nDst++
+				continue
+			}
+			size = 1
+
+		} else {
+			// Decode a multi-byte rune.
+			r, size = utf8.DecodeRune(src[nSrc:])
+			if size == 1 {
+				// All valid runes of size 1 (those below utf8.RuneSelf) were
+				// handled above. We have invalid UTF-8 or we haven't seen the
+				// full character yet.
+				if !atEOF && !utf8.FullRune(src[nSrc:]) {
+					err = transform.ErrShortSrc
+				} else {
+					err = internal.RepertoireError(m.charmap.replacement)
+				}
+				break
+			}
+		}
+
+		// Binary search in [low, high) for that rune in the m.charmap.encode table.
+		for low, high := int(m.charmap.low), 0x100; ; {
+			if low >= high {
+				err = internal.RepertoireError(m.charmap.replacement)
+				break loop
+			}
+			mid := (low + high) / 2
+			got := m.charmap.encode[mid]
+			gotRune := rune(got & (1<<24 - 1))
+			if gotRune < r {
+				low = mid + 1
+			} else if gotRune > r {
+				high = mid
+			} else {
+				dst[nDst] = byte(got >> 24)
+				nDst++
+				break
+			}
+		}
+		nSrc += size
+	}
+	return nDst, nSrc, err
+}

vendor/golang.org/x/text/encoding/charmap/maketables.go

@@ -0,0 +1,524 @@
+// 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
+
+package main
+
+import (
+	"bufio"
+	"fmt"
+	"log"
+	"net/http"
+	"sort"
+	"strings"
+	"unicode/utf8"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/internal/gen"
+)
+
+const ascii = "\x00\x01\x02\x03\x04\x05\x06\x07\x08\x09\x0a\x0b\x0c\x0d\x0e\x0f" +
+	"\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1a\x1b\x1c\x1d\x1e\x1f" +
+	` !"#$%&'()*+,-./0123456789:;<=>?` +
+	`@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_` +
+	"`abcdefghijklmnopqrstuvwxyz{|}~\u007f"
+
+var encodings = []struct {
+	name        string
+	mib         string
+	comment     string
+	varName     string
+	replacement byte
+	mapping     string
+}{
+	{
+		"IBM Code Page 437",
+		"PC8CodePage437",
+		"",
+		"CodePage437",
+		encoding.ASCIISub,
+		"http://source.icu-project.org/repos/icu/data/trunk/charset/data/ucm/glibc-IBM437-2.1.2.ucm",
+	},
+	{
+		"IBM Code Page 850",
+		"PC850Multilingual",
+		"",
+		"CodePage850",
+		encoding.ASCIISub,
+		"http://source.icu-project.org/repos/icu/data/trunk/charset/data/ucm/glibc-IBM850-2.1.2.ucm",
+	},
+	{
+		"IBM Code Page 852",
+		"PCp852",
+		"",
+		"CodePage852",
+		encoding.ASCIISub,
+		"http://source.icu-project.org/repos/icu/data/trunk/charset/data/ucm/glibc-IBM852-2.1.2.ucm",
+	},
+	{
+		"IBM Code Page 855",
+		"IBM855",
+		"",
+		"CodePage855",
+		encoding.ASCIISub,
+		"http://source.icu-project.org/repos/icu/data/trunk/charset/data/ucm/glibc-IBM855-2.1.2.ucm",
+	},
+	{
+		"Windows Code Page 858", // PC latin1 with Euro
+		"IBM00858",
+		"",
+		"CodePage858",
+		encoding.ASCIISub,
+		"http://source.icu-project.org/repos/icu/data/trunk/charset/data/ucm/windows-858-2000.ucm",
+	},
+	{
+		"IBM Code Page 860",
+		"IBM860",
+		"",
+		"CodePage860",
+		encoding.ASCIISub,
+		"http://source.icu-project.org/repos/icu/data/trunk/charset/data/ucm/glibc-IBM860-2.1.2.ucm",
+	},
+	{
+		"IBM Code Page 862",
+		"PC862LatinHebrew",
+		"",
+		"CodePage862",
+		encoding.ASCIISub,
+		"http://source.icu-project.org/repos/icu/data/trunk/charset/data/ucm/glibc-IBM862-2.1.2.ucm",
+	},
+	{
+		"IBM Code Page 863",
+		"IBM863",
+		"",
+		"CodePage863",
+		encoding.ASCIISub,
+		"http://source.icu-project.org/repos/icu/data/trunk/charset/data/ucm/glibc-IBM863-2.1.2.ucm",
+	},
+	{
+		"IBM Code Page 865",
+		"IBM865",
+		"",
+		"CodePage865",
+		encoding.ASCIISub,
+		"http://source.icu-project.org/repos/icu/data/trunk/charset/data/ucm/glibc-IBM865-2.1.2.ucm",
+	},
+	{
+		"IBM Code Page 866",
+		"IBM866",
+		"",
+		"CodePage866",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-ibm866.txt",
+	},
+	{
+		"ISO 8859-1",
+		"ISOLatin1",
+		"",
+		"ISO8859_1",
+		encoding.ASCIISub,
+		"http://source.icu-project.org/repos/icu/data/trunk/charset/data/ucm/iso-8859_1-1998.ucm",
+	},
+	{
+		"ISO 8859-2",
+		"ISOLatin2",
+		"",
+		"ISO8859_2",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-2.txt",
+	},
+	{
+		"ISO 8859-3",
+		"ISOLatin3",
+		"",
+		"ISO8859_3",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-3.txt",
+	},
+	{
+		"ISO 8859-4",
+		"ISOLatin4",
+		"",
+		"ISO8859_4",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-4.txt",
+	},
+	{
+		"ISO 8859-5",
+		"ISOLatinCyrillic",
+		"",
+		"ISO8859_5",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-5.txt",
+	},
+	{
+		"ISO 8859-6",
+		"ISOLatinArabic",
+		"",
+		"ISO8859_6,ISO8859_6E,ISO8859_6I",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-6.txt",
+	},
+	{
+		"ISO 8859-7",
+		"ISOLatinGreek",
+		"",
+		"ISO8859_7",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-7.txt",
+	},
+	{
+		"ISO 8859-8",
+		"ISOLatinHebrew",
+		"",
+		"ISO8859_8,ISO8859_8E,ISO8859_8I",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-8.txt",
+	},
+	{
+		"ISO 8859-10",
+		"ISOLatin6",
+		"",
+		"ISO8859_10",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-10.txt",
+	},
+	{
+		"ISO 8859-13",
+		"ISO885913",
+		"",
+		"ISO8859_13",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-13.txt",
+	},
+	{
+		"ISO 8859-14",
+		"ISO885914",
+		"",
+		"ISO8859_14",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-14.txt",
+	},
+	{
+		"ISO 8859-15",
+		"ISO885915",
+		"",
+		"ISO8859_15",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-15.txt",
+	},
+	{
+		"ISO 8859-16",
+		"ISO885916",
+		"",
+		"ISO8859_16",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-iso-8859-16.txt",
+	},
+	{
+		"KOI8-R",
+		"KOI8R",
+		"",
+		"KOI8R",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-koi8-r.txt",
+	},
+	{
+		"KOI8-U",
+		"KOI8U",
+		"",
+		"KOI8U",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-koi8-u.txt",
+	},
+	{
+		"Macintosh",
+		"Macintosh",
+		"",
+		"Macintosh",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-macintosh.txt",
+	},
+	{
+		"Macintosh Cyrillic",
+		"MacintoshCyrillic",
+		"",
+		"MacintoshCyrillic",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-x-mac-cyrillic.txt",
+	},
+	{
+		"Windows 874",
+		"Windows874",
+		"",
+		"Windows874",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-windows-874.txt",
+	},
+	{
+		"Windows 1250",
+		"Windows1250",
+		"",
+		"Windows1250",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-windows-1250.txt",
+	},
+	{
+		"Windows 1251",
+		"Windows1251",
+		"",
+		"Windows1251",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-windows-1251.txt",
+	},
+	{
+		"Windows 1252",
+		"Windows1252",
+		"",
+		"Windows1252",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-windows-1252.txt",
+	},
+	{
+		"Windows 1253",
+		"Windows1253",
+		"",
+		"Windows1253",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-windows-1253.txt",
+	},
+	{
+		"Windows 1254",
+		"Windows1254",
+		"",
+		"Windows1254",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-windows-1254.txt",
+	},
+	{
+		"Windows 1255",
+		"Windows1255",
+		"",
+		"Windows1255",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-windows-1255.txt",
+	},
+	{
+		"Windows 1256",
+		"Windows1256",
+		"",
+		"Windows1256",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-windows-1256.txt",
+	},
+	{
+		"Windows 1257",
+		"Windows1257",
+		"",
+		"Windows1257",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-windows-1257.txt",
+	},
+	{
+		"Windows 1258",
+		"Windows1258",
+		"",
+		"Windows1258",
+		encoding.ASCIISub,
+		"http://encoding.spec.whatwg.org/index-windows-1258.txt",
+	},
+	{
+		"X-User-Defined",
+		"XUserDefined",
+		"It is defined at http://encoding.spec.whatwg.org/#x-user-defined",
+		"XUserDefined",
+		encoding.ASCIISub,
+		ascii +
+			"\uf780\uf781\uf782\uf783\uf784\uf785\uf786\uf787" +
+			"\uf788\uf789\uf78a\uf78b\uf78c\uf78d\uf78e\uf78f" +
+			"\uf790\uf791\uf792\uf793\uf794\uf795\uf796\uf797" +
+			"\uf798\uf799\uf79a\uf79b\uf79c\uf79d\uf79e\uf79f" +
+			"\uf7a0\uf7a1\uf7a2\uf7a3\uf7a4\uf7a5\uf7a6\uf7a7" +
+			"\uf7a8\uf7a9\uf7aa\uf7ab\uf7ac\uf7ad\uf7ae\uf7af" +
+			"\uf7b0\uf7b1\uf7b2\uf7b3\uf7b4\uf7b5\uf7b6\uf7b7" +
+			"\uf7b8\uf7b9\uf7ba\uf7bb\uf7bc\uf7bd\uf7be\uf7bf" +
+			"\uf7c0\uf7c1\uf7c2\uf7c3\uf7c4\uf7c5\uf7c6\uf7c7" +
+			"\uf7c8\uf7c9\uf7ca\uf7cb\uf7cc\uf7cd\uf7ce\uf7cf" +
+			"\uf7d0\uf7d1\uf7d2\uf7d3\uf7d4\uf7d5\uf7d6\uf7d7" +
+			"\uf7d8\uf7d9\uf7da\uf7db\uf7dc\uf7dd\uf7de\uf7df" +
+			"\uf7e0\uf7e1\uf7e2\uf7e3\uf7e4\uf7e5\uf7e6\uf7e7" +
+			"\uf7e8\uf7e9\uf7ea\uf7eb\uf7ec\uf7ed\uf7ee\uf7ef" +
+			"\uf7f0\uf7f1\uf7f2\uf7f3\uf7f4\uf7f5\uf7f6\uf7f7" +
+			"\uf7f8\uf7f9\uf7fa\uf7fb\uf7fc\uf7fd\uf7fe\uf7ff",
+	},
+}
+
+func getWHATWG(url string) string {
+	res, err := http.Get(url)
+	if err != nil {
+		log.Fatalf("%q: Get: %v", url, err)
+	}
+	defer res.Body.Close()
+
+	mapping := make([]rune, 128)
+	for i := range mapping {
+		mapping[i] = '\ufffd'
+	}
+
+	scanner := bufio.NewScanner(res.Body)
+	for scanner.Scan() {
+		s := strings.TrimSpace(scanner.Text())
+		if s == "" || s[0] == '#' {
+			continue
+		}
+		x, y := 0, 0
+		if _, err := fmt.Sscanf(s, "%d\t0x%x", &x, &y); err != nil {
+			log.Fatalf("could not parse %q", s)
+		}
+		if x < 0 || 128 <= x {
+			log.Fatalf("code %d is out of range", x)
+		}
+		if 0x80 <= y && y < 0xa0 {
+			// We diverge from the WHATWG spec by mapping control characters
+			// in the range [0x80, 0xa0) to U+FFFD.
+			continue
+		}
+		mapping[x] = rune(y)
+	}
+	return ascii + string(mapping)
+}
+
+func getUCM(url string) string {
+	res, err := http.Get(url)
+	if err != nil {
+		log.Fatalf("%q: Get: %v", url, err)
+	}
+	defer res.Body.Close()
+
+	mapping := make([]rune, 256)
+	for i := range mapping {
+		mapping[i] = '\ufffd'
+	}
+
+	charsFound := 0
+	scanner := bufio.NewScanner(res.Body)
+	for scanner.Scan() {
+		s := strings.TrimSpace(scanner.Text())
+		if s == "" || s[0] == '#' {
+			continue
+		}
+		var c byte
+		var r rune
+		if _, err := fmt.Sscanf(s, `<U%x> \x%x |0`, &r, &c); err != nil {
+			continue
+		}
+		mapping[c] = r
+		charsFound++
+	}
+
+	if charsFound < 200 {
+		log.Fatalf("%q: only %d characters found (wrong page format?)", url, charsFound)
+	}
+
+	return string(mapping)
+}
+
+func main() {
+	mibs := map[string]bool{}
+	all := []string{}
+
+	w := gen.NewCodeWriter()
+	defer w.WriteGoFile("tables.go", "charmap")
+
+	printf := func(s string, a ...interface{}) { fmt.Fprintf(w, s, a...) }
+
+	printf("import (\n")
+	printf("\t\"golang.org/x/text/encoding\"\n")
+	printf("\t\"golang.org/x/text/encoding/internal/identifier\"\n")
+	printf(")\n\n")
+	for _, e := range encodings {
+		varNames := strings.Split(e.varName, ",")
+		all = append(all, varNames...)
+		varName := varNames[0]
+		switch {
+		case strings.HasPrefix(e.mapping, "http://encoding.spec.whatwg.org/"):
+			e.mapping = getWHATWG(e.mapping)
+		case strings.HasPrefix(e.mapping, "http://source.icu-project.org/repos/icu/data/trunk/charset/data/ucm/"):
+			e.mapping = getUCM(e.mapping)
+		}
+
+		asciiSuperset, low := strings.HasPrefix(e.mapping, ascii), 0x00
+		if asciiSuperset {
+			low = 0x80
+		}
+		lvn := 1
+		if strings.HasPrefix(varName, "ISO") || strings.HasPrefix(varName, "KOI") {
+			lvn = 3
+		}
+		lowerVarName := strings.ToLower(varName[:lvn]) + varName[lvn:]
+		printf("// %s is the %s encoding.\n", varName, e.name)
+		if e.comment != "" {
+			printf("//\n// %s\n", e.comment)
+		}
+		printf("var %s encoding.Encoding = &%s\n\nvar %s = charmap{\nname: %q,\n",
+			varName, lowerVarName, lowerVarName, e.name)
+		if mibs[e.mib] {
+			log.Fatalf("MIB type %q declared multiple times.", e.mib)
+		}
+		printf("mib: identifier.%s,\n", e.mib)
+		printf("asciiSuperset: %t,\n", asciiSuperset)
+		printf("low: 0x%02x,\n", low)
+		printf("replacement: 0x%02x,\n", e.replacement)
+
+		printf("decode: [256]utf8Enc{\n")
+		i, backMapping := 0, map[rune]byte{}
+		for _, c := range e.mapping {
+			if _, ok := backMapping[c]; !ok && c != utf8.RuneError {
+				backMapping[c] = byte(i)
+			}
+			var buf [8]byte
+			n := utf8.EncodeRune(buf[:], c)
+			if n > 3 {
+				panic(fmt.Sprintf("rune %q (%U) is too long", c, c))
+			}
+			printf("{%d,[3]byte{0x%02x,0x%02x,0x%02x}},", n, buf[0], buf[1], buf[2])
+			if i%2 == 1 {
+				printf("\n")
+			}
+			i++
+		}
+		printf("},\n")
+
+		printf("encode: [256]uint32{\n")
+		encode := make([]uint32, 0, 256)
+		for c, i := range backMapping {
+			encode = append(encode, uint32(i)<<24|uint32(c))
+		}
+		sort.Sort(byRune(encode))
+		for len(encode) < cap(encode) {
+			encode = append(encode, encode[len(encode)-1])
+		}
+		for i, enc := range encode {
+			printf("0x%08x,", enc)
+			if i%8 == 7 {
+				printf("\n")
+			}
+		}
+		printf("},\n}\n")
+
+		// Add an estimate of the size of a single charmap{} struct value, which
+		// includes two 256 elem arrays of 4 bytes and some extra fields, which
+		// align to 3 uint64s on 64-bit architectures.
+		w.Size += 2*4*256 + 3*8
+	}
+	// TODO: add proper line breaking.
+	printf("var listAll = []encoding.Encoding{\n%s,\n}\n\n", strings.Join(all, ",\n"))
+}
+
+type byRune []uint32
+
+func (b byRune) Len() int           { return len(b) }
+func (b byRune) Less(i, j int) bool { return b[i]&0xffffff < b[j]&0xffffff }
+func (b byRune) Swap(i, j int)      { b[i], b[j] = b[j], b[i] }

vendor/golang.org/x/text/encoding/encoding.go

@@ -0,0 +1,335 @@
+// 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 encoding defines an interface for character encodings, such as Shift
+// JIS and Windows 1252, that can convert to and from UTF-8.
+//
+// Encoding implementations are provided in other packages, such as
+// golang.org/x/text/encoding/charmap and
+// golang.org/x/text/encoding/japanese.
+package encoding // import "golang.org/x/text/encoding"
+
+import (
+	"errors"
+	"io"
+	"strconv"
+	"unicode/utf8"
+
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/transform"
+)
+
+// TODO:
+// - There seems to be some inconsistency in when decoders return errors
+//   and when not. Also documentation seems to suggest they shouldn't return
+//   errors at all (except for UTF-16).
+// - Encoders seem to rely on or at least benefit from the input being in NFC
+//   normal form. Perhaps add an example how users could prepare their output.
+
+// Encoding is a character set encoding that can be transformed to and from
+// UTF-8.
+type Encoding interface {
+	// NewDecoder returns a Decoder.
+	NewDecoder() *Decoder
+
+	// NewEncoder returns an Encoder.
+	NewEncoder() *Encoder
+}
+
+// A Decoder converts bytes to UTF-8. It implements transform.Transformer.
+//
+// Transforming source bytes that are not of that encoding will not result in an
+// error per se. Each byte that cannot be transcoded will be represented in the
+// output by the UTF-8 encoding of '\uFFFD', the replacement rune.
+type Decoder struct {
+	transform.Transformer
+
+	// This forces external creators of Decoders to use names in struct
+	// initializers, allowing for future extendibility without having to break
+	// code.
+	_ struct{}
+}
+
+// Bytes converts the given encoded bytes to UTF-8. It returns the converted
+// bytes or nil, err if any error occurred.
+func (d *Decoder) Bytes(b []byte) ([]byte, error) {
+	b, _, err := transform.Bytes(d, b)
+	if err != nil {
+		return nil, err
+	}
+	return b, nil
+}
+
+// String converts the given encoded string to UTF-8. It returns the converted
+// string or "", err if any error occurred.
+func (d *Decoder) String(s string) (string, error) {
+	s, _, err := transform.String(d, s)
+	if err != nil {
+		return "", err
+	}
+	return s, nil
+}
+
+// Reader wraps another Reader to decode its bytes.
+//
+// The Decoder may not be used for any other operation as long as the returned
+// Reader is in use.
+func (d *Decoder) Reader(r io.Reader) io.Reader {
+	return transform.NewReader(r, d)
+}
+
+// An Encoder converts bytes from UTF-8. It implements transform.Transformer.
+//
+// Each rune that cannot be transcoded will result in an error. In this case,
+// the transform will consume all source byte up to, not including the offending
+// rune. Transforming source bytes that are not valid UTF-8 will be replaced by
+// `\uFFFD`. To return early with an error instead, use transform.Chain to
+// preprocess the data with a UTF8Validator.
+type Encoder struct {
+	transform.Transformer
+
+	// This forces external creators of Encoders to use names in struct
+	// initializers, allowing for future extendibility without having to break
+	// code.
+	_ struct{}
+}
+
+// Bytes converts bytes from UTF-8. It returns the converted bytes or nil, err if
+// any error occurred.
+func (e *Encoder) Bytes(b []byte) ([]byte, error) {
+	b, _, err := transform.Bytes(e, b)
+	if err != nil {
+		return nil, err
+	}
+	return b, nil
+}
+
+// String converts a string from UTF-8. It returns the converted string or
+// "", err if any error occurred.
+func (e *Encoder) String(s string) (string, error) {
+	s, _, err := transform.String(e, s)
+	if err != nil {
+		return "", err
+	}
+	return s, nil
+}
+
+// Writer wraps another Writer to encode its UTF-8 output.
+//
+// The Encoder may not be used for any other operation as long as the returned
+// Writer is in use.
+func (e *Encoder) Writer(w io.Writer) io.Writer {
+	return transform.NewWriter(w, e)
+}
+
+// ASCIISub is the ASCII substitute character, as recommended by
+// http://unicode.org/reports/tr36/#Text_Comparison
+const ASCIISub = '\x1a'
+
+// Nop is the nop encoding. Its transformed bytes are the same as the source
+// bytes; it does not replace invalid UTF-8 sequences.
+var Nop Encoding = nop{}
+
+type nop struct{}
+
+func (nop) NewDecoder() *Decoder {
+	return &Decoder{Transformer: transform.Nop}
+}
+func (nop) NewEncoder() *Encoder {
+	return &Encoder{Transformer: transform.Nop}
+}
+
+// Replacement is the replacement encoding. Decoding from the replacement
+// encoding yields a single '\uFFFD' replacement rune. Encoding from UTF-8 to
+// the replacement encoding yields the same as the source bytes except that
+// invalid UTF-8 is converted to '\uFFFD'.
+//
+// It is defined at http://encoding.spec.whatwg.org/#replacement
+var Replacement Encoding = replacement{}
+
+type replacement struct{}
+
+func (replacement) NewDecoder() *Decoder {
+	return &Decoder{Transformer: replacementDecoder{}}
+}
+
+func (replacement) NewEncoder() *Encoder {
+	return &Encoder{Transformer: replacementEncoder{}}
+}
+
+func (replacement) ID() (mib identifier.MIB, other string) {
+	return identifier.Replacement, ""
+}
+
+type replacementDecoder struct{ transform.NopResetter }
+
+func (replacementDecoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	if len(dst) < 3 {
+		return 0, 0, transform.ErrShortDst
+	}
+	if atEOF {
+		const fffd = "\ufffd"
+		dst[0] = fffd[0]
+		dst[1] = fffd[1]
+		dst[2] = fffd[2]
+		nDst = 3
+	}
+	return nDst, len(src), nil
+}
+
+type replacementEncoder struct{ transform.NopResetter }
+
+func (replacementEncoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+
+	for ; nSrc < len(src); nSrc += size {
+		r = rune(src[nSrc])
+
+		// Decode a 1-byte rune.
+		if r < utf8.RuneSelf {
+			size = 1
+
+		} else {
+			// Decode a multi-byte rune.
+			r, size = utf8.DecodeRune(src[nSrc:])
+			if size == 1 {
+				// All valid runes of size 1 (those below utf8.RuneSelf) were
+				// handled above. We have invalid UTF-8 or we haven't seen the
+				// full character yet.
+				if !atEOF && !utf8.FullRune(src[nSrc:]) {
+					err = transform.ErrShortSrc
+					break
+				}
+				r = '\ufffd'
+			}
+		}
+
+		if nDst+utf8.RuneLen(r) > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		nDst += utf8.EncodeRune(dst[nDst:], r)
+	}
+	return nDst, nSrc, err
+}
+
+// HTMLEscapeUnsupported wraps encoders to replace source runes outside the
+// repertoire of the destination encoding with HTML escape sequences.
+//
+// This wrapper exists to comply to URL and HTML forms requiring a
+// non-terminating legacy encoder. The produced sequences may lead to data
+// loss as they are indistinguishable from legitimate input. To avoid this
+// issue, use UTF-8 encodings whenever possible.
+func HTMLEscapeUnsupported(e *Encoder) *Encoder {
+	return &Encoder{Transformer: &errorHandler{e, errorToHTML}}
+}
+
+// ReplaceUnsupported wraps encoders to replace source runes outside the
+// repertoire of the destination encoding with an encoding-specific
+// replacement.
+//
+// This wrapper is only provided for backwards compatibility and legacy
+// handling. Its use is strongly discouraged. Use UTF-8 whenever possible.
+func ReplaceUnsupported(e *Encoder) *Encoder {
+	return &Encoder{Transformer: &errorHandler{e, errorToReplacement}}
+}
+
+type errorHandler struct {
+	*Encoder
+	handler func(dst []byte, r rune, err repertoireError) (n int, ok bool)
+}
+
+// TODO: consider making this error public in some form.
+type repertoireError interface {
+	Replacement() byte
+}
+
+func (h errorHandler) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	nDst, nSrc, err = h.Transformer.Transform(dst, src, atEOF)
+	for err != nil {
+		rerr, ok := err.(repertoireError)
+		if !ok {
+			return nDst, nSrc, err
+		}
+		r, sz := utf8.DecodeRune(src[nSrc:])
+		n, ok := h.handler(dst[nDst:], r, rerr)
+		if !ok {
+			return nDst, nSrc, transform.ErrShortDst
+		}
+		err = nil
+		nDst += n
+		if nSrc += sz; nSrc < len(src) {
+			var dn, sn int
+			dn, sn, err = h.Transformer.Transform(dst[nDst:], src[nSrc:], atEOF)
+			nDst += dn
+			nSrc += sn
+		}
+	}
+	return nDst, nSrc, err
+}
+
+func errorToHTML(dst []byte, r rune, err repertoireError) (n int, ok bool) {
+	buf := [8]byte{}
+	b := strconv.AppendUint(buf[:0], uint64(r), 10)
+	if n = len(b) + len("&#;"); n >= len(dst) {
+		return 0, false
+	}
+	dst[0] = '&'
+	dst[1] = '#'
+	dst[copy(dst[2:], b)+2] = ';'
+	return n, true
+}
+
+func errorToReplacement(dst []byte, r rune, err repertoireError) (n int, ok bool) {
+	if len(dst) == 0 {
+		return 0, false
+	}
+	dst[0] = err.Replacement()
+	return 1, true
+}
+
+// ErrInvalidUTF8 means that a transformer encountered invalid UTF-8.
+var ErrInvalidUTF8 = errors.New("encoding: invalid UTF-8")
+
+// UTF8Validator is a transformer that returns ErrInvalidUTF8 on the first
+// input byte that is not valid UTF-8.
+var UTF8Validator transform.Transformer = utf8Validator{}
+
+type utf8Validator struct{ transform.NopResetter }
+
+func (utf8Validator) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	n := len(src)
+	if n > len(dst) {
+		n = len(dst)
+	}
+	for i := 0; i < n; {
+		if c := src[i]; c < utf8.RuneSelf {
+			dst[i] = c
+			i++
+			continue
+		}
+		_, size := utf8.DecodeRune(src[i:])
+		if size == 1 {
+			// All valid runes of size 1 (those below utf8.RuneSelf) were
+			// handled above. We have invalid UTF-8 or we haven't seen the
+			// full character yet.
+			err = ErrInvalidUTF8
+			if !atEOF && !utf8.FullRune(src[i:]) {
+				err = transform.ErrShortSrc
+			}
+			return i, i, err
+		}
+		if i+size > len(dst) {
+			return i, i, transform.ErrShortDst
+		}
+		for ; size > 0; size-- {
+			dst[i] = src[i]
+			i++
+		}
+	}
+	if len(src) > len(dst) {
+		err = transform.ErrShortDst
+	}
+	return n, n, err
+}

vendor/golang.org/x/text/encoding/htmlindex/gen.go

@@ -0,0 +1,167 @@
+// 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
+
+import (
+	"bytes"
+	"encoding/json"
+	"fmt"
+	"log"
+	"strings"
+
+	"golang.org/x/text/internal/gen"
+)
+
+type group struct {
+	Encodings []struct {
+		Labels []string
+		Name   string
+	}
+}
+
+func main() {
+	gen.Init()
+
+	r := gen.Open("http://www.w3.org/TR", "w3", "encoding/indexes/encodings.json")
+	var groups []group
+	if err := json.NewDecoder(r).Decode(&groups); err != nil {
+		log.Fatalf("Error reading encodings.json: %v", err)
+	}
+
+	w := &bytes.Buffer{}
+	fmt.Fprintln(w, "type htmlEncoding byte")
+	fmt.Fprintln(w, "const (")
+	for i, g := range groups {
+		for _, e := range g.Encodings {
+			name := consts[e.Name]
+			if name == "" {
+				log.Fatalf("No const defined for %s.", e.Name)
+			}
+			if i == 0 {
+				fmt.Fprintf(w, "%s htmlEncoding = iota\n", name)
+			} else {
+				fmt.Fprintf(w, "%s\n", name)
+			}
+		}
+	}
+	fmt.Fprintln(w, "numEncodings")
+	fmt.Fprint(w, ")\n\n")
+
+	fmt.Fprintln(w, "var canonical = [numEncodings]string{")
+	for _, g := range groups {
+		for _, e := range g.Encodings {
+			fmt.Fprintf(w, "%q,\n", e.Name)
+		}
+	}
+	fmt.Fprint(w, "}\n\n")
+
+	fmt.Fprintln(w, "var nameMap = map[string]htmlEncoding{")
+	for _, g := range groups {
+		for _, e := range g.Encodings {
+			for _, l := range e.Labels {
+				fmt.Fprintf(w, "%q: %s,\n", l, consts[e.Name])
+			}
+		}
+	}
+	fmt.Fprint(w, "}\n\n")
+
+	var tags []string
+	fmt.Fprintln(w, "var localeMap = []htmlEncoding{")
+	for _, loc := range locales {
+		tags = append(tags, loc.tag)
+		fmt.Fprintf(w, "%s, // %s \n", consts[loc.name], loc.tag)
+	}
+	fmt.Fprint(w, "}\n\n")
+
+	fmt.Fprintf(w, "const locales = %q\n", strings.Join(tags, " "))
+
+	gen.WriteGoFile("tables.go", "htmlindex", w.Bytes())
+}
+
+// consts maps canonical encoding name to internal constant.
+var consts = map[string]string{
+	"utf-8":          "utf8",
+	"ibm866":         "ibm866",
+	"iso-8859-2":     "iso8859_2",
+	"iso-8859-3":     "iso8859_3",
+	"iso-8859-4":     "iso8859_4",
+	"iso-8859-5":     "iso8859_5",
+	"iso-8859-6":     "iso8859_6",
+	"iso-8859-7":     "iso8859_7",
+	"iso-8859-8":     "iso8859_8",
+	"iso-8859-8-i":   "iso8859_8I",
+	"iso-8859-10":    "iso8859_10",
+	"iso-8859-13":    "iso8859_13",
+	"iso-8859-14":    "iso8859_14",
+	"iso-8859-15":    "iso8859_15",
+	"iso-8859-16":    "iso8859_16",
+	"koi8-r":         "koi8r",
+	"koi8-u":         "koi8u",
+	"macintosh":      "macintosh",
+	"windows-874":    "windows874",
+	"windows-1250":   "windows1250",
+	"windows-1251":   "windows1251",
+	"windows-1252":   "windows1252",
+	"windows-1253":   "windows1253",
+	"windows-1254":   "windows1254",
+	"windows-1255":   "windows1255",
+	"windows-1256":   "windows1256",
+	"windows-1257":   "windows1257",
+	"windows-1258":   "windows1258",
+	"x-mac-cyrillic": "macintoshCyrillic",
+	"gbk":            "gbk",
+	"gb18030":        "gb18030",
+	// "hz-gb-2312":     "hzgb2312", // Was removed from WhatWG
+	"big5":           "big5",
+	"euc-jp":         "eucjp",
+	"iso-2022-jp":    "iso2022jp",
+	"shift_jis":      "shiftJIS",
+	"euc-kr":         "euckr",
+	"replacement":    "replacement",
+	"utf-16be":       "utf16be",
+	"utf-16le":       "utf16le",
+	"x-user-defined": "xUserDefined",
+}
+
+// locales is taken from
+// https://html.spec.whatwg.org/multipage/syntax.html#encoding-sniffing-algorithm.
+var locales = []struct{ tag, name string }{
+	{"und", "windows-1252"}, // The default value.
+	{"ar", "windows-1256"},
+	{"ba", "windows-1251"},
+	{"be", "windows-1251"},
+	{"bg", "windows-1251"},
+	{"cs", "windows-1250"},
+	{"el", "iso-8859-7"},
+	{"et", "windows-1257"},
+	{"fa", "windows-1256"},
+	{"he", "windows-1255"},
+	{"hr", "windows-1250"},
+	{"hu", "iso-8859-2"},
+	{"ja", "shift_jis"},
+	{"kk", "windows-1251"},
+	{"ko", "euc-kr"},
+	{"ku", "windows-1254"},
+	{"ky", "windows-1251"},
+	{"lt", "windows-1257"},
+	{"lv", "windows-1257"},
+	{"mk", "windows-1251"},
+	{"pl", "iso-8859-2"},
+	{"ru", "windows-1251"},
+	{"sah", "windows-1251"},
+	{"sk", "windows-1250"},
+	{"sl", "iso-8859-2"},
+	{"sr", "windows-1251"},
+	{"tg", "windows-1251"},
+	{"th", "windows-874"},
+	{"tr", "windows-1254"},
+	{"tt", "windows-1251"},
+	{"uk", "windows-1251"},
+	{"vi", "windows-1258"},
+	{"zh-hans", "gb18030"},
+	{"zh-hant", "big5"},
+}

vendor/golang.org/x/text/encoding/htmlindex/htmlindex.go

@@ -0,0 +1,86 @@
+// 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.
+
+//go:generate go run gen.go
+
+// Package htmlindex maps character set encoding names to Encodings as
+// recommended by the W3C for use in HTML 5. See http://www.w3.org/TR/encoding.
+package htmlindex
+
+// TODO: perhaps have a "bare" version of the index (used by this package) that
+// is not pre-loaded with all encodings. Global variables in encodings prevent
+// the linker from being able to purge unneeded tables. This means that
+// referencing all encodings, as this package does for the default index, links
+// in all encodings unconditionally.
+//
+// This issue can be solved by either solving the linking issue (see
+// https://github.com/golang/go/issues/6330) or refactoring the encoding tables
+// (e.g. moving the tables to internal packages that do not use global
+// variables).
+
+// TODO: allow canonicalizing names
+
+import (
+	"errors"
+	"strings"
+	"sync"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/language"
+)
+
+var (
+	errInvalidName = errors.New("htmlindex: invalid encoding name")
+	errUnknown     = errors.New("htmlindex: unknown Encoding")
+	errUnsupported = errors.New("htmlindex: this encoding is not supported")
+)
+
+var (
+	matcherOnce sync.Once
+	matcher     language.Matcher
+)
+
+// LanguageDefault returns the canonical name of the default encoding for a
+// given language.
+func LanguageDefault(tag language.Tag) string {
+	matcherOnce.Do(func() {
+		tags := []language.Tag{}
+		for _, t := range strings.Split(locales, " ") {
+			tags = append(tags, language.MustParse(t))
+		}
+		matcher = language.NewMatcher(tags)
+	})
+	_, i, _ := matcher.Match(tag)
+	return canonical[localeMap[i]] // Default is Windows-1252.
+}
+
+// Get returns an Encoding for one of the names listed in
+// http://www.w3.org/TR/encoding using the Default Index. Matching is case-
+// insensitive.
+func Get(name string) (encoding.Encoding, error) {
+	x, ok := nameMap[strings.ToLower(strings.TrimSpace(name))]
+	if !ok {
+		return nil, errInvalidName
+	}
+	return encodings[x], nil
+}
+
+// Name reports the canonical name of the given Encoding. It will return
+// an error if e is not associated with a supported encoding scheme.
+func Name(e encoding.Encoding) (string, error) {
+	id, ok := e.(identifier.Interface)
+	if !ok {
+		return "", errUnknown
+	}
+	mib, _ := id.ID()
+	if mib == 0 {
+		return "", errUnknown
+	}
+	v, ok := mibMap[mib]
+	if !ok {
+		return "", errUnsupported
+	}
+	return canonical[v], nil
+}

vendor/golang.org/x/text/encoding/htmlindex/map.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 htmlindex
+
+import (
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/charmap"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/encoding/japanese"
+	"golang.org/x/text/encoding/korean"
+	"golang.org/x/text/encoding/simplifiedchinese"
+	"golang.org/x/text/encoding/traditionalchinese"
+	"golang.org/x/text/encoding/unicode"
+)
+
+// mibMap maps a MIB identifier to an htmlEncoding index.
+var mibMap = map[identifier.MIB]htmlEncoding{
+	identifier.UTF8:              utf8,
+	identifier.UTF16BE:           utf16be,
+	identifier.UTF16LE:           utf16le,
+	identifier.IBM866:            ibm866,
+	identifier.ISOLatin2:         iso8859_2,
+	identifier.ISOLatin3:         iso8859_3,
+	identifier.ISOLatin4:         iso8859_4,
+	identifier.ISOLatinCyrillic:  iso8859_5,
+	identifier.ISOLatinArabic:    iso8859_6,
+	identifier.ISOLatinGreek:     iso8859_7,
+	identifier.ISOLatinHebrew:    iso8859_8,
+	identifier.ISO88598I:         iso8859_8I,
+	identifier.ISOLatin6:         iso8859_10,
+	identifier.ISO885913:         iso8859_13,
+	identifier.ISO885914:         iso8859_14,
+	identifier.ISO885915:         iso8859_15,
+	identifier.ISO885916:         iso8859_16,
+	identifier.KOI8R:             koi8r,
+	identifier.KOI8U:             koi8u,
+	identifier.Macintosh:         macintosh,
+	identifier.MacintoshCyrillic: macintoshCyrillic,
+	identifier.Windows874:        windows874,
+	identifier.Windows1250:       windows1250,
+	identifier.Windows1251:       windows1251,
+	identifier.Windows1252:       windows1252,
+	identifier.Windows1253:       windows1253,
+	identifier.Windows1254:       windows1254,
+	identifier.Windows1255:       windows1255,
+	identifier.Windows1256:       windows1256,
+	identifier.Windows1257:       windows1257,
+	identifier.Windows1258:       windows1258,
+	identifier.XUserDefined:      xUserDefined,
+	identifier.GBK:               gbk,
+	identifier.GB18030:           gb18030,
+	identifier.Big5:              big5,
+	identifier.EUCPkdFmtJapanese: eucjp,
+	identifier.ISO2022JP:         iso2022jp,
+	identifier.ShiftJIS:          shiftJIS,
+	identifier.EUCKR:             euckr,
+	identifier.Replacement:       replacement,
+}
+
+// encodings maps the internal htmlEncoding to an Encoding.
+// TODO: consider using a reusable index in encoding/internal.
+var encodings = [numEncodings]encoding.Encoding{
+	utf8:              unicode.UTF8,
+	ibm866:            charmap.CodePage866,
+	iso8859_2:         charmap.ISO8859_2,
+	iso8859_3:         charmap.ISO8859_3,
+	iso8859_4:         charmap.ISO8859_4,
+	iso8859_5:         charmap.ISO8859_5,
+	iso8859_6:         charmap.ISO8859_6,
+	iso8859_7:         charmap.ISO8859_7,
+	iso8859_8:         charmap.ISO8859_8,
+	iso8859_8I:        charmap.ISO8859_8I,
+	iso8859_10:        charmap.ISO8859_10,
+	iso8859_13:        charmap.ISO8859_13,
+	iso8859_14:        charmap.ISO8859_14,
+	iso8859_15:        charmap.ISO8859_15,
+	iso8859_16:        charmap.ISO8859_16,
+	koi8r:             charmap.KOI8R,
+	koi8u:             charmap.KOI8U,
+	macintosh:         charmap.Macintosh,
+	windows874:        charmap.Windows874,
+	windows1250:       charmap.Windows1250,
+	windows1251:       charmap.Windows1251,
+	windows1252:       charmap.Windows1252,
+	windows1253:       charmap.Windows1253,
+	windows1254:       charmap.Windows1254,
+	windows1255:       charmap.Windows1255,
+	windows1256:       charmap.Windows1256,
+	windows1257:       charmap.Windows1257,
+	windows1258:       charmap.Windows1258,
+	macintoshCyrillic: charmap.MacintoshCyrillic,
+	gbk:               simplifiedchinese.GBK,
+	gb18030:           simplifiedchinese.GB18030,
+	big5:              traditionalchinese.Big5,
+	eucjp:             japanese.EUCJP,
+	iso2022jp:         japanese.ISO2022JP,
+	shiftJIS:          japanese.ShiftJIS,
+	euckr:             korean.EUCKR,
+	replacement:       encoding.Replacement,
+	utf16be:           unicode.UTF16(unicode.BigEndian, unicode.IgnoreBOM),
+	utf16le:           unicode.UTF16(unicode.LittleEndian, unicode.IgnoreBOM),
+	xUserDefined:      charmap.XUserDefined,
+}

vendor/golang.org/x/text/encoding/htmlindex/tables.go

@@ -0,0 +1,352 @@
+// This file was generated by go generate; DO NOT EDIT
+
+package htmlindex
+
+type htmlEncoding byte
+
+const (
+	utf8 htmlEncoding = iota
+	ibm866
+	iso8859_2
+	iso8859_3
+	iso8859_4
+	iso8859_5
+	iso8859_6
+	iso8859_7
+	iso8859_8
+	iso8859_8I
+	iso8859_10
+	iso8859_13
+	iso8859_14
+	iso8859_15
+	iso8859_16
+	koi8r
+	koi8u
+	macintosh
+	windows874
+	windows1250
+	windows1251
+	windows1252
+	windows1253
+	windows1254
+	windows1255
+	windows1256
+	windows1257
+	windows1258
+	macintoshCyrillic
+	gbk
+	gb18030
+	big5
+	eucjp
+	iso2022jp
+	shiftJIS
+	euckr
+	replacement
+	utf16be
+	utf16le
+	xUserDefined
+	numEncodings
+)
+
+var canonical = [numEncodings]string{
+	"utf-8",
+	"ibm866",
+	"iso-8859-2",
+	"iso-8859-3",
+	"iso-8859-4",
+	"iso-8859-5",
+	"iso-8859-6",
+	"iso-8859-7",
+	"iso-8859-8",
+	"iso-8859-8-i",
+	"iso-8859-10",
+	"iso-8859-13",
+	"iso-8859-14",
+	"iso-8859-15",
+	"iso-8859-16",
+	"koi8-r",
+	"koi8-u",
+	"macintosh",
+	"windows-874",
+	"windows-1250",
+	"windows-1251",
+	"windows-1252",
+	"windows-1253",
+	"windows-1254",
+	"windows-1255",
+	"windows-1256",
+	"windows-1257",
+	"windows-1258",
+	"x-mac-cyrillic",
+	"gbk",
+	"gb18030",
+	"big5",
+	"euc-jp",
+	"iso-2022-jp",
+	"shift_jis",
+	"euc-kr",
+	"replacement",
+	"utf-16be",
+	"utf-16le",
+	"x-user-defined",
+}
+
+var nameMap = map[string]htmlEncoding{
+	"unicode-1-1-utf-8":   utf8,
+	"utf-8":               utf8,
+	"utf8":                utf8,
+	"866":                 ibm866,
+	"cp866":               ibm866,
+	"csibm866":            ibm866,
+	"ibm866":              ibm866,
+	"csisolatin2":         iso8859_2,
+	"iso-8859-2":          iso8859_2,
+	"iso-ir-101":          iso8859_2,
+	"iso8859-2":           iso8859_2,
+	"iso88592":            iso8859_2,
+	"iso_8859-2":          iso8859_2,
+	"iso_8859-2:1987":     iso8859_2,
+	"l2":                  iso8859_2,
+	"latin2":              iso8859_2,
+	"csisolatin3":         iso8859_3,
+	"iso-8859-3":          iso8859_3,
+	"iso-ir-109":          iso8859_3,
+	"iso8859-3":           iso8859_3,
+	"iso88593":            iso8859_3,
+	"iso_8859-3":          iso8859_3,
+	"iso_8859-3:1988":     iso8859_3,
+	"l3":                  iso8859_3,
+	"latin3":              iso8859_3,
+	"csisolatin4":         iso8859_4,
+	"iso-8859-4":          iso8859_4,
+	"iso-ir-110":          iso8859_4,
+	"iso8859-4":           iso8859_4,
+	"iso88594":            iso8859_4,
+	"iso_8859-4":          iso8859_4,
+	"iso_8859-4:1988":     iso8859_4,
+	"l4":                  iso8859_4,
+	"latin4":              iso8859_4,
+	"csisolatincyrillic":  iso8859_5,
+	"cyrillic":            iso8859_5,
+	"iso-8859-5":          iso8859_5,
+	"iso-ir-144":          iso8859_5,
+	"iso8859-5":           iso8859_5,
+	"iso88595":            iso8859_5,
+	"iso_8859-5":          iso8859_5,
+	"iso_8859-5:1988":     iso8859_5,
+	"arabic":              iso8859_6,
+	"asmo-708":            iso8859_6,
+	"csiso88596e":         iso8859_6,
+	"csiso88596i":         iso8859_6,
+	"csisolatinarabic":    iso8859_6,
+	"ecma-114":            iso8859_6,
+	"iso-8859-6":          iso8859_6,
+	"iso-8859-6-e":        iso8859_6,
+	"iso-8859-6-i":        iso8859_6,
+	"iso-ir-127":          iso8859_6,
+	"iso8859-6":           iso8859_6,
+	"iso88596":            iso8859_6,
+	"iso_8859-6":          iso8859_6,
+	"iso_8859-6:1987":     iso8859_6,
+	"csisolatingreek":     iso8859_7,
+	"ecma-118":            iso8859_7,
+	"elot_928":            iso8859_7,
+	"greek":               iso8859_7,
+	"greek8":              iso8859_7,
+	"iso-8859-7":          iso8859_7,
+	"iso-ir-126":          iso8859_7,
+	"iso8859-7":           iso8859_7,
+	"iso88597":            iso8859_7,
+	"iso_8859-7":          iso8859_7,
+	"iso_8859-7:1987":     iso8859_7,
+	"sun_eu_greek":        iso8859_7,
+	"csiso88598e":         iso8859_8,
+	"csisolatinhebrew":    iso8859_8,
+	"hebrew":              iso8859_8,
+	"iso-8859-8":          iso8859_8,
+	"iso-8859-8-e":        iso8859_8,
+	"iso-ir-138":          iso8859_8,
+	"iso8859-8":           iso8859_8,
+	"iso88598":            iso8859_8,
+	"iso_8859-8":          iso8859_8,
+	"iso_8859-8:1988":     iso8859_8,
+	"visual":              iso8859_8,
+	"csiso88598i":         iso8859_8I,
+	"iso-8859-8-i":        iso8859_8I,
+	"logical":             iso8859_8I,
+	"csisolatin6":         iso8859_10,
+	"iso-8859-10":         iso8859_10,
+	"iso-ir-157":          iso8859_10,
+	"iso8859-10":          iso8859_10,
+	"iso885910":           iso8859_10,
+	"l6":                  iso8859_10,
+	"latin6":              iso8859_10,
+	"iso-8859-13":         iso8859_13,
+	"iso8859-13":          iso8859_13,
+	"iso885913":           iso8859_13,
+	"iso-8859-14":         iso8859_14,
+	"iso8859-14":          iso8859_14,
+	"iso885914":           iso8859_14,
+	"csisolatin9":         iso8859_15,
+	"iso-8859-15":         iso8859_15,
+	"iso8859-15":          iso8859_15,
+	"iso885915":           iso8859_15,
+	"iso_8859-15":         iso8859_15,
+	"l9":                  iso8859_15,
+	"iso-8859-16":         iso8859_16,
+	"cskoi8r":             koi8r,
+	"koi":                 koi8r,
+	"koi8":                koi8r,
+	"koi8-r":              koi8r,
+	"koi8_r":              koi8r,
+	"koi8-ru":             koi8u,
+	"koi8-u":              koi8u,
+	"csmacintosh":         macintosh,
+	"mac":                 macintosh,
+	"macintosh":           macintosh,
+	"x-mac-roman":         macintosh,
+	"dos-874":             windows874,
+	"iso-8859-11":         windows874,
+	"iso8859-11":          windows874,
+	"iso885911":           windows874,
+	"tis-620":             windows874,
+	"windows-874":         windows874,
+	"cp1250":              windows1250,
+	"windows-1250":        windows1250,
+	"x-cp1250":            windows1250,
+	"cp1251":              windows1251,
+	"windows-1251":        windows1251,
+	"x-cp1251":            windows1251,
+	"ansi_x3.4-1968":      windows1252,
+	"ascii":               windows1252,
+	"cp1252":              windows1252,
+	"cp819":               windows1252,
+	"csisolatin1":         windows1252,
+	"ibm819":              windows1252,
+	"iso-8859-1":          windows1252,
+	"iso-ir-100":          windows1252,
+	"iso8859-1":           windows1252,
+	"iso88591":            windows1252,
+	"iso_8859-1":          windows1252,
+	"iso_8859-1:1987":     windows1252,
+	"l1":                  windows1252,
+	"latin1":              windows1252,
+	"us-ascii":            windows1252,
+	"windows-1252":        windows1252,
+	"x-cp1252":            windows1252,
+	"cp1253":              windows1253,
+	"windows-1253":        windows1253,
+	"x-cp1253":            windows1253,
+	"cp1254":              windows1254,
+	"csisolatin5":         windows1254,
+	"iso-8859-9":          windows1254,
+	"iso-ir-148":          windows1254,
+	"iso8859-9":           windows1254,
+	"iso88599":            windows1254,
+	"iso_8859-9":          windows1254,
+	"iso_8859-9:1989":     windows1254,
+	"l5":                  windows1254,
+	"latin5":              windows1254,
+	"windows-1254":        windows1254,
+	"x-cp1254":            windows1254,
+	"cp1255":              windows1255,
+	"windows-1255":        windows1255,
+	"x-cp1255":            windows1255,
+	"cp1256":              windows1256,
+	"windows-1256":        windows1256,
+	"x-cp1256":            windows1256,
+	"cp1257":              windows1257,
+	"windows-1257":        windows1257,
+	"x-cp1257":            windows1257,
+	"cp1258":              windows1258,
+	"windows-1258":        windows1258,
+	"x-cp1258":            windows1258,
+	"x-mac-cyrillic":      macintoshCyrillic,
+	"x-mac-ukrainian":     macintoshCyrillic,
+	"chinese":             gbk,
+	"csgb2312":            gbk,
+	"csiso58gb231280":     gbk,
+	"gb2312":              gbk,
+	"gb_2312":             gbk,
+	"gb_2312-80":          gbk,
+	"gbk":                 gbk,
+	"iso-ir-58":           gbk,
+	"x-gbk":               gbk,
+	"gb18030":             gb18030,
+	"big5":                big5,
+	"big5-hkscs":          big5,
+	"cn-big5":             big5,
+	"csbig5":              big5,
+	"x-x-big5":            big5,
+	"cseucpkdfmtjapanese": eucjp,
+	"euc-jp":              eucjp,
+	"x-euc-jp":            eucjp,
+	"csiso2022jp":         iso2022jp,
+	"iso-2022-jp":         iso2022jp,
+	"csshiftjis":          shiftJIS,
+	"ms932":               shiftJIS,
+	"ms_kanji":            shiftJIS,
+	"shift-jis":           shiftJIS,
+	"shift_jis":           shiftJIS,
+	"sjis":                shiftJIS,
+	"windows-31j":         shiftJIS,
+	"x-sjis":              shiftJIS,
+	"cseuckr":             euckr,
+	"csksc56011987":       euckr,
+	"euc-kr":              euckr,
+	"iso-ir-149":          euckr,
+	"korean":              euckr,
+	"ks_c_5601-1987":      euckr,
+	"ks_c_5601-1989":      euckr,
+	"ksc5601":             euckr,
+	"ksc_5601":            euckr,
+	"windows-949":         euckr,
+	"csiso2022kr":         replacement,
+	"hz-gb-2312":          replacement,
+	"iso-2022-cn":         replacement,
+	"iso-2022-cn-ext":     replacement,
+	"iso-2022-kr":         replacement,
+	"utf-16be":            utf16be,
+	"utf-16":              utf16le,
+	"utf-16le":            utf16le,
+	"x-user-defined":      xUserDefined,
+}
+
+var localeMap = []htmlEncoding{
+	windows1252, // und
+	windows1256, // ar
+	windows1251, // ba
+	windows1251, // be
+	windows1251, // bg
+	windows1250, // cs
+	iso8859_7,   // el
+	windows1257, // et
+	windows1256, // fa
+	windows1255, // he
+	windows1250, // hr
+	iso8859_2,   // hu
+	shiftJIS,    // ja
+	windows1251, // kk
+	euckr,       // ko
+	windows1254, // ku
+	windows1251, // ky
+	windows1257, // lt
+	windows1257, // lv
+	windows1251, // mk
+	iso8859_2,   // pl
+	windows1251, // ru
+	windows1251, // sah
+	windows1250, // sk
+	iso8859_2,   // sl
+	windows1251, // sr
+	windows1251, // tg
+	windows874,  // th
+	windows1254, // tr
+	windows1251, // tt
+	windows1251, // uk
+	windows1258, // vi
+	gb18030,     // zh-hans
+	big5,        // zh-hant
+}
+
+const locales = "und ar ba be bg cs el et fa he hr hu ja kk ko ku ky lt lv mk pl ru sah sk sl sr tg th tr tt uk vi zh-hans zh-hant"

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

@@ -0,0 +1,137 @@
+// 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
+
+import (
+	"bytes"
+	"encoding/xml"
+	"fmt"
+	"io"
+	"log"
+	"strings"
+
+	"golang.org/x/text/internal/gen"
+)
+
+type registry struct {
+	XMLName  xml.Name `xml:"registry"`
+	Updated  string   `xml:"updated"`
+	Registry []struct {
+		ID     string `xml:"id,attr"`
+		Record []struct {
+			Name string `xml:"name"`
+			Xref []struct {
+				Type string `xml:"type,attr"`
+				Data string `xml:"data,attr"`
+			} `xml:"xref"`
+			Desc struct {
+				Data string `xml:",innerxml"`
+				// Any []struct {
+				// 	Data string `xml:",chardata"`
+				// } `xml:",any"`
+				// Data string `xml:",chardata"`
+			} `xml:"description,"`
+			MIB   string   `xml:"value"`
+			Alias []string `xml:"alias"`
+			MIME  string   `xml:"preferred_alias"`
+		} `xml:"record"`
+	} `xml:"registry"`
+}
+
+func main() {
+	r := gen.OpenIANAFile("assignments/character-sets/character-sets.xml")
+	reg := &registry{}
+	if err := xml.NewDecoder(r).Decode(&reg); err != nil && err != io.EOF {
+		log.Fatalf("Error decoding charset registry: %v", err)
+	}
+	if len(reg.Registry) == 0 || reg.Registry[0].ID != "character-sets-1" {
+		log.Fatalf("Unexpected ID %s", reg.Registry[0].ID)
+	}
+
+	w := &bytes.Buffer{}
+	fmt.Fprintf(w, "const (\n")
+	for _, rec := range reg.Registry[0].Record {
+		constName := ""
+		for _, a := range rec.Alias {
+			if strings.HasPrefix(a, "cs") && strings.IndexByte(a, '-') == -1 {
+				// Some of the constant definitions have comments in them. Strip those.
+				constName = strings.Title(strings.SplitN(a[2:], "\n", 2)[0])
+			}
+		}
+		if constName == "" {
+			switch rec.MIB {
+			case "2085":
+				constName = "HZGB2312" // Not listed as alias for some reason.
+			default:
+				log.Fatalf("No cs alias defined for %s.", rec.MIB)
+			}
+		}
+		if rec.MIME != "" {
+			rec.MIME = fmt.Sprintf(" (MIME: %s)", rec.MIME)
+		}
+		fmt.Fprintf(w, "// %s is the MIB identifier with IANA name %s%s.\n//\n", constName, rec.Name, rec.MIME)
+		if len(rec.Desc.Data) > 0 {
+			fmt.Fprint(w, "// ")
+			d := xml.NewDecoder(strings.NewReader(rec.Desc.Data))
+			inElem := true
+			attr := ""
+			for {
+				t, err := d.Token()
+				if err != nil {
+					if err != io.EOF {
+						log.Fatal(err)
+					}
+					break
+				}
+				switch x := t.(type) {
+				case xml.CharData:
+					attr = "" // Don't need attribute info.
+					a := bytes.Split([]byte(x), []byte("\n"))
+					for i, b := range a {
+						if b = bytes.TrimSpace(b); len(b) != 0 {
+							if !inElem && i > 0 {
+								fmt.Fprint(w, "\n// ")
+							}
+							inElem = false
+							fmt.Fprintf(w, "%s ", string(b))
+						}
+					}
+				case xml.StartElement:
+					if x.Name.Local == "xref" {
+						inElem = true
+						use := false
+						for _, a := range x.Attr {
+							if a.Name.Local == "type" {
+								use = use || a.Value != "person"
+							}
+							if a.Name.Local == "data" && use {
+								attr = a.Value + " "
+							}
+						}
+					}
+				case xml.EndElement:
+					inElem = false
+					fmt.Fprint(w, attr)
+				}
+			}
+			fmt.Fprint(w, "\n")
+		}
+		for _, x := range rec.Xref {
+			switch x.Type {
+			case "rfc":
+				fmt.Fprintf(w, "// Reference: %s\n", strings.ToUpper(x.Data))
+			case "uri":
+				fmt.Fprintf(w, "// Reference: %s\n", x.Data)
+			}
+		}
+		fmt.Fprintf(w, "%s MIB = %s\n", constName, rec.MIB)
+		fmt.Fprintln(w)
+	}
+	fmt.Fprintln(w, ")")
+
+	gen.WriteGoFile("mib.go", "identifier", w.Bytes())
+}

vendor/golang.org/x/text/encoding/internal/identifier/identifier.go

@@ -0,0 +1,81 @@
+// 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.
+
+//go:generate go run gen.go
+
+// Package identifier defines the contract between implementations of Encoding
+// and Index by defining identifiers that uniquely identify standardized coded
+// character sets (CCS) and character encoding schemes (CES), which we will
+// together refer to as encodings, for which Encoding implementations provide
+// converters to and from UTF-8. This package is typically only of concern to
+// implementers of Indexes and Encodings.
+//
+// One part of the identifier is the MIB code, which is defined by IANA and
+// uniquely identifies a CCS or CES. Each code is associated with data that
+// references authorities, official documentation as well as aliases and MIME
+// names.
+//
+// Not all CESs are covered by the IANA registry. The "other" string that is
+// returned by ID can be used to identify other character sets or versions of
+// existing ones.
+//
+// It is recommended that each package that provides a set of Encodings provide
+// the All and Common variables to reference all supported encodings and
+// commonly used subset. This allows Index implementations to include all
+// available encodings without explicitly referencing or knowing about them.
+package identifier
+
+// Note: this package is internal, but could be made public if there is a need
+// for writing third-party Indexes and Encodings.
+
+// References:
+// - http://source.icu-project.org/repos/icu/icu/trunk/source/data/mappings/convrtrs.txt
+// - http://www.iana.org/assignments/character-sets/character-sets.xhtml
+// - http://www.iana.org/assignments/ianacharset-mib/ianacharset-mib
+// - http://www.ietf.org/rfc/rfc2978.txt
+// - http://www.unicode.org/reports/tr22/
+// - http://www.w3.org/TR/encoding/
+// - http://www.w3.org/TR/encoding/indexes/encodings.json
+// - https://encoding.spec.whatwg.org/
+// - https://tools.ietf.org/html/rfc6657#section-5
+
+// Interface can be implemented by Encodings to define the CCS or CES for which
+// it implements conversions.
+type Interface interface {
+	// ID returns an encoding identifier. Exactly one of the mib and other
+	// values should be non-zero.
+	//
+	// In the usual case it is only necessary to indicate the MIB code. The
+	// other string can be used to specify encodings for which there is no MIB,
+	// such as "x-mac-dingbat".
+	//
+	// The other string may only contain the characters a-z, A-Z, 0-9, - and _.
+	ID() (mib MIB, other string)
+
+	// NOTE: the restrictions on the encoding are to allow extending the syntax
+	// with additional information such as versions, vendors and other variants.
+}
+
+// A MIB identifies an encoding. It is derived from the IANA MIB codes and adds
+// some identifiers for some encodings that are not covered by the IANA
+// standard.
+//
+// See http://www.iana.org/assignments/ianacharset-mib.
+type MIB uint16
+
+// These additional MIB types are not defined in IANA. They are added because
+// they are common and defined within the text repo.
+const (
+	// Unofficial marks the start of encodings not registered by IANA.
+	Unofficial MIB = 10000 + iota
+
+	// Replacement is the WhatWG replacement encoding.
+	Replacement
+
+	// XUserDefined is the code for x-user-defined.
+	XUserDefined
+
+	// MacintoshCyrillic is the code for x-mac-cyrillic.
+	MacintoshCyrillic
+)

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

@@ -0,0 +1,75 @@
+// 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 internal contains code that is shared among encoding implementations.
+package internal
+
+import (
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/transform"
+)
+
+// Encoding is an implementation of the Encoding interface that adds the String
+// and ID methods to an existing encoding.
+type Encoding struct {
+	encoding.Encoding
+	Name string
+	MIB  identifier.MIB
+}
+
+// _ verifies that Encoding implements identifier.Interface.
+var _ identifier.Interface = (*Encoding)(nil)
+
+func (e *Encoding) String() string {
+	return e.Name
+}
+
+func (e *Encoding) ID() (mib identifier.MIB, other string) {
+	return e.MIB, ""
+}
+
+// SimpleEncoding is an Encoding that combines two Transformers.
+type SimpleEncoding struct {
+	Decoder transform.Transformer
+	Encoder transform.Transformer
+}
+
+func (e *SimpleEncoding) NewDecoder() *encoding.Decoder {
+	return &encoding.Decoder{Transformer: e.Decoder}
+}
+
+func (e *SimpleEncoding) NewEncoder() *encoding.Encoder {
+	return &encoding.Encoder{Transformer: e.Encoder}
+}
+
+// FuncEncoding is an Encoding that combines two functions returning a new
+// Transformer.
+type FuncEncoding struct {
+	Decoder func() transform.Transformer
+	Encoder func() transform.Transformer
+}
+
+func (e FuncEncoding) NewDecoder() *encoding.Decoder {
+	return &encoding.Decoder{Transformer: e.Decoder()}
+}
+
+func (e FuncEncoding) NewEncoder() *encoding.Encoder {
+	return &encoding.Encoder{Transformer: e.Encoder()}
+}
+
+// A RepertoireError indicates a rune is not in the repertoire of a destination
+// encoding. It is associated with an encoding-specific suggested replacement
+// byte.
+type RepertoireError byte
+
+// Error implements the error interrface.
+func (r RepertoireError) Error() string {
+	return "encoding: rune not supported by encoding."
+}
+
+// Replacement returns the replacement string associated with this error.
+func (r RepertoireError) Replacement() byte { return byte(r) }
+
+var ErrASCIIReplacement = RepertoireError(encoding.ASCIISub)

vendor/golang.org/x/text/encoding/japanese/all.go

@@ -0,0 +1,12 @@
+// 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 japanese
+
+import (
+	"golang.org/x/text/encoding"
+)
+
+// All is a list of all defined encodings in this package.
+var All = []encoding.Encoding{EUCJP, ISO2022JP, ShiftJIS}

vendor/golang.org/x/text/encoding/japanese/eucjp.go

@@ -0,0 +1,211 @@
+// 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 japanese
+
+import (
+	"errors"
+	"unicode/utf8"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/internal"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/transform"
+)
+
+// EUCJP is the EUC-JP encoding.
+var EUCJP encoding.Encoding = &eucJP
+
+var eucJP = internal.Encoding{
+	&internal.SimpleEncoding{eucJPDecoder{}, eucJPEncoder{}},
+	"EUC-JP",
+	identifier.EUCPkdFmtJapanese,
+}
+
+var errInvalidEUCJP = errors.New("japanese: invalid EUC-JP encoding")
+
+type eucJPDecoder struct{ transform.NopResetter }
+
+func (eucJPDecoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+loop:
+	for ; nSrc < len(src); nSrc += size {
+		switch c0 := src[nSrc]; {
+		case c0 < utf8.RuneSelf:
+			r, size = rune(c0), 1
+
+		case c0 == 0x8e:
+			if nSrc+1 >= len(src) {
+				err = transform.ErrShortSrc
+				break loop
+			}
+			c1 := src[nSrc+1]
+			if c1 < 0xa1 || 0xdf < c1 {
+				err = errInvalidEUCJP
+				break loop
+			}
+			r, size = rune(c1)+(0xff61-0xa1), 2
+
+		case c0 == 0x8f:
+			if nSrc+2 >= len(src) {
+				err = transform.ErrShortSrc
+				break loop
+			}
+			c1 := src[nSrc+1]
+			if c1 < 0xa1 || 0xfe < c1 {
+				err = errInvalidEUCJP
+				break loop
+			}
+			c2 := src[nSrc+2]
+			if c2 < 0xa1 || 0xfe < c2 {
+				err = errInvalidEUCJP
+				break loop
+			}
+			r, size = '\ufffd', 3
+			if i := int(c1-0xa1)*94 + int(c2-0xa1); i < len(jis0212Decode) {
+				r = rune(jis0212Decode[i])
+				if r == 0 {
+					r = '\ufffd'
+				}
+			}
+
+		case 0xa1 <= c0 && c0 <= 0xfe:
+			if nSrc+1 >= len(src) {
+				err = transform.ErrShortSrc
+				break loop
+			}
+			c1 := src[nSrc+1]
+			if c1 < 0xa1 || 0xfe < c1 {
+				err = errInvalidEUCJP
+				break loop
+			}
+			r, size = '\ufffd', 2
+			if i := int(c0-0xa1)*94 + int(c1-0xa1); i < len(jis0208Decode) {
+				r = rune(jis0208Decode[i])
+				if r == 0 {
+					r = '\ufffd'
+				}
+			}
+
+		default:
+			err = errInvalidEUCJP
+			break loop
+		}
+
+		if nDst+utf8.RuneLen(r) > len(dst) {
+			err = transform.ErrShortDst
+			break loop
+		}
+		nDst += utf8.EncodeRune(dst[nDst:], r)
+	}
+	if atEOF && err == transform.ErrShortSrc {
+		err = errInvalidEUCJP
+	}
+	return nDst, nSrc, err
+}
+
+type eucJPEncoder struct{ transform.NopResetter }
+
+func (eucJPEncoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+	for ; nSrc < len(src); nSrc += size {
+		r = rune(src[nSrc])
+
+		// Decode a 1-byte rune.
+		if r < utf8.RuneSelf {
+			size = 1
+
+		} else {
+			// Decode a multi-byte rune.
+			r, size = utf8.DecodeRune(src[nSrc:])
+			if size == 1 {
+				// All valid runes of size 1 (those below utf8.RuneSelf) were
+				// handled above. We have invalid UTF-8 or we haven't seen the
+				// full character yet.
+				if !atEOF && !utf8.FullRune(src[nSrc:]) {
+					err = transform.ErrShortSrc
+					break
+				}
+			}
+
+			// func init checks that the switch covers all tables.
+			switch {
+			case encode0Low <= r && r < encode0High:
+				if r = rune(encode0[r-encode0Low]); r != 0 {
+					goto write2or3
+				}
+			case encode1Low <= r && r < encode1High:
+				if r = rune(encode1[r-encode1Low]); r != 0 {
+					goto write2or3
+				}
+			case encode2Low <= r && r < encode2High:
+				if r = rune(encode2[r-encode2Low]); r != 0 {
+					goto write2or3
+				}
+			case encode3Low <= r && r < encode3High:
+				if r = rune(encode3[r-encode3Low]); r != 0 {
+					goto write2or3
+				}
+			case encode4Low <= r && r < encode4High:
+				if r = rune(encode4[r-encode4Low]); r != 0 {
+					goto write2or3
+				}
+			case encode5Low <= r && r < encode5High:
+				if 0xff61 <= r && r < 0xffa0 {
+					goto write2
+				}
+				if r = rune(encode5[r-encode5Low]); r != 0 {
+					goto write2or3
+				}
+			}
+			err = internal.ErrASCIIReplacement
+			break
+		}
+
+		if nDst >= len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst] = uint8(r)
+		nDst++
+		continue
+
+	write2or3:
+		if r>>tableShift == jis0208 {
+			if nDst+2 > len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+		} else {
+			if nDst+3 > len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			dst[nDst] = 0x8f
+			nDst++
+		}
+		dst[nDst+0] = 0xa1 + uint8(r>>codeShift)&codeMask
+		dst[nDst+1] = 0xa1 + uint8(r)&codeMask
+		nDst += 2
+		continue
+
+	write2:
+		if nDst+2 > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst+0] = 0x8e
+		dst[nDst+1] = uint8(r - (0xff61 - 0xa1))
+		nDst += 2
+		continue
+	}
+	return nDst, nSrc, err
+}
+
+func init() {
+	// Check that the hard-coded encode switch covers all tables.
+	if numEncodeTables != 6 {
+		panic("bad numEncodeTables")
+	}
+}

vendor/golang.org/x/text/encoding/japanese/iso2022jp.go

@@ -0,0 +1,296 @@
+// 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 japanese
+
+import (
+	"errors"
+	"unicode/utf8"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/internal"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/transform"
+)
+
+// ISO2022JP is the ISO-2022-JP encoding.
+var ISO2022JP encoding.Encoding = &iso2022JP
+
+var iso2022JP = internal.Encoding{
+	internal.FuncEncoding{iso2022JPNewDecoder, iso2022JPNewEncoder},
+	"ISO-2022-JP",
+	identifier.ISO2022JP,
+}
+
+func iso2022JPNewDecoder() transform.Transformer {
+	return new(iso2022JPDecoder)
+}
+
+func iso2022JPNewEncoder() transform.Transformer {
+	return new(iso2022JPEncoder)
+}
+
+var errInvalidISO2022JP = errors.New("japanese: invalid ISO-2022-JP encoding")
+
+const (
+	asciiState = iota
+	katakanaState
+	jis0208State
+	jis0212State
+)
+
+const asciiEsc = 0x1b
+
+type iso2022JPDecoder int
+
+func (d *iso2022JPDecoder) Reset() {
+	*d = asciiState
+}
+
+func (d *iso2022JPDecoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+loop:
+	for ; nSrc < len(src); nSrc += size {
+		c0 := src[nSrc]
+		if c0 >= utf8.RuneSelf {
+			err = errInvalidISO2022JP
+			break loop
+		}
+
+		if c0 == asciiEsc {
+			if nSrc+2 >= len(src) {
+				err = transform.ErrShortSrc
+				break loop
+			}
+			size = 3
+			c1 := src[nSrc+1]
+			c2 := src[nSrc+2]
+			switch {
+			case c1 == '$' && (c2 == '@' || c2 == 'B'):
+				*d = jis0208State
+				continue
+			case c1 == '$' && c2 == '(':
+				if nSrc+3 >= len(src) {
+					err = transform.ErrShortSrc
+					break loop
+				}
+				size = 4
+				if src[nSrc]+3 == 'D' {
+					*d = jis0212State
+					continue
+				}
+			case c1 == '(' && (c2 == 'B' || c2 == 'J'):
+				*d = asciiState
+				continue
+			case c1 == '(' && c2 == 'I':
+				*d = katakanaState
+				continue
+			}
+			err = errInvalidISO2022JP
+			break loop
+		}
+
+		switch *d {
+		case asciiState:
+			r, size = rune(c0), 1
+
+		case katakanaState:
+			if c0 < 0x21 || 0x60 <= c0 {
+				err = errInvalidISO2022JP
+				break loop
+			}
+			r, size = rune(c0)+(0xff61-0x21), 1
+
+		default:
+			if c0 == 0x0a {
+				*d = asciiState
+				r, size = rune(c0), 1
+				break
+			}
+			if nSrc+1 >= len(src) {
+				err = transform.ErrShortSrc
+				break loop
+			}
+			size = 2
+			c1 := src[nSrc+1]
+			i := int(c0-0x21)*94 + int(c1-0x21)
+			if *d == jis0208State && i < len(jis0208Decode) {
+				r = rune(jis0208Decode[i])
+			} else if *d == jis0212State && i < len(jis0212Decode) {
+				r = rune(jis0212Decode[i])
+			} else {
+				r = '\ufffd'
+				break
+			}
+			if r == 0 {
+				r = '\ufffd'
+			}
+		}
+
+		if nDst+utf8.RuneLen(r) > len(dst) {
+			err = transform.ErrShortDst
+			break loop
+		}
+		nDst += utf8.EncodeRune(dst[nDst:], r)
+	}
+	if atEOF && err == transform.ErrShortSrc {
+		err = errInvalidISO2022JP
+	}
+	return nDst, nSrc, err
+}
+
+type iso2022JPEncoder int
+
+func (e *iso2022JPEncoder) Reset() {
+	*e = asciiState
+}
+
+func (e *iso2022JPEncoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+	for ; nSrc < len(src); nSrc += size {
+		r = rune(src[nSrc])
+
+		// Decode a 1-byte rune.
+		if r < utf8.RuneSelf {
+			size = 1
+
+		} else {
+			// Decode a multi-byte rune.
+			r, size = utf8.DecodeRune(src[nSrc:])
+			if size == 1 {
+				// All valid runes of size 1 (those below utf8.RuneSelf) were
+				// handled above. We have invalid UTF-8 or we haven't seen the
+				// full character yet.
+				if !atEOF && !utf8.FullRune(src[nSrc:]) {
+					err = transform.ErrShortSrc
+					break
+				}
+			}
+
+			// func init checks that the switch covers all tables.
+			//
+			// http://encoding.spec.whatwg.org/#iso-2022-jp says that "the index jis0212
+			// is not used by the iso-2022-jp encoder due to lack of widespread support".
+			//
+			// TODO: do we have to special-case U+00A5 and U+203E, as per
+			// http://encoding.spec.whatwg.org/#iso-2022-jp
+			// Doing so would mean that "\u00a5" would not be preserved
+			// after an encode-decode round trip.
+			switch {
+			case encode0Low <= r && r < encode0High:
+				if r = rune(encode0[r-encode0Low]); r>>tableShift == jis0208 {
+					goto writeJIS
+				}
+			case encode1Low <= r && r < encode1High:
+				if r = rune(encode1[r-encode1Low]); r>>tableShift == jis0208 {
+					goto writeJIS
+				}
+			case encode2Low <= r && r < encode2High:
+				if r = rune(encode2[r-encode2Low]); r>>tableShift == jis0208 {
+					goto writeJIS
+				}
+			case encode3Low <= r && r < encode3High:
+				if r = rune(encode3[r-encode3Low]); r>>tableShift == jis0208 {
+					goto writeJIS
+				}
+			case encode4Low <= r && r < encode4High:
+				if r = rune(encode4[r-encode4Low]); r>>tableShift == jis0208 {
+					goto writeJIS
+				}
+			case encode5Low <= r && r < encode5High:
+				if 0xff61 <= r && r < 0xffa0 {
+					goto writeKatakana
+				}
+				if r = rune(encode5[r-encode5Low]); r>>tableShift == jis0208 {
+					goto writeJIS
+				}
+			}
+
+			// Switch back to ASCII state in case of error so that an ASCII
+			// replacement character can be written in the correct state.
+			if *e != asciiState {
+				if nDst+3 > len(dst) {
+					err = transform.ErrShortDst
+					break
+				}
+				*e = asciiState
+				dst[nDst+0] = asciiEsc
+				dst[nDst+1] = '('
+				dst[nDst+2] = 'B'
+				nDst += 3
+			}
+			err = internal.ErrASCIIReplacement
+			break
+		}
+
+		if *e != asciiState {
+			if nDst+4 > len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			*e = asciiState
+			dst[nDst+0] = asciiEsc
+			dst[nDst+1] = '('
+			dst[nDst+2] = 'B'
+			nDst += 3
+		} else if nDst >= len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst] = uint8(r)
+		nDst++
+		continue
+
+	writeJIS:
+		if *e != jis0208State {
+			if nDst+5 > len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			*e = jis0208State
+			dst[nDst+0] = asciiEsc
+			dst[nDst+1] = '$'
+			dst[nDst+2] = 'B'
+			nDst += 3
+		} else if nDst+2 > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst+0] = 0x21 + uint8(r>>codeShift)&codeMask
+		dst[nDst+1] = 0x21 + uint8(r)&codeMask
+		nDst += 2
+		continue
+
+	writeKatakana:
+		if *e != katakanaState {
+			if nDst+4 > len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			*e = katakanaState
+			dst[nDst+0] = asciiEsc
+			dst[nDst+1] = '('
+			dst[nDst+2] = 'I'
+			nDst += 3
+		} else if nDst >= len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst] = uint8(r - (0xff61 - 0x21))
+		nDst++
+		continue
+	}
+	if atEOF && err == nil && *e != asciiState {
+		if nDst+3 > len(dst) {
+			err = transform.ErrShortDst
+		} else {
+			*e = asciiState
+			dst[nDst+0] = asciiEsc
+			dst[nDst+1] = '('
+			dst[nDst+2] = 'B'
+			nDst += 3
+		}
+	}
+	return nDst, nSrc, err
+}

vendor/golang.org/x/text/encoding/japanese/maketables.go

@@ -0,0 +1,161 @@
+// 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
+
+package main
+
+// This program generates tables.go:
+//	go run maketables.go | gofmt > tables.go
+
+// TODO: Emoji extensions?
+// http://www.unicode.org/faq/emoji_dingbats.html
+// http://www.unicode.org/Public/UNIDATA/EmojiSources.txt
+
+import (
+	"bufio"
+	"fmt"
+	"log"
+	"net/http"
+	"sort"
+	"strings"
+)
+
+type entry struct {
+	jisCode, table int
+}
+
+func main() {
+	fmt.Printf("// generated by go run maketables.go; DO NOT EDIT\n\n")
+	fmt.Printf("// Package japanese provides Japanese encodings such as EUC-JP and Shift JIS.\n")
+	fmt.Printf(`package japanese // import "golang.org/x/text/encoding/japanese"` + "\n\n")
+
+	reverse := [65536]entry{}
+	for i := range reverse {
+		reverse[i].table = -1
+	}
+
+	tables := []struct {
+		url  string
+		name string
+	}{
+		{"http://encoding.spec.whatwg.org/index-jis0208.txt", "0208"},
+		{"http://encoding.spec.whatwg.org/index-jis0212.txt", "0212"},
+	}
+	for i, table := range tables {
+		res, err := http.Get(table.url)
+		if err != nil {
+			log.Fatalf("%q: Get: %v", table.url, err)
+		}
+		defer res.Body.Close()
+
+		mapping := [65536]uint16{}
+
+		scanner := bufio.NewScanner(res.Body)
+		for scanner.Scan() {
+			s := strings.TrimSpace(scanner.Text())
+			if s == "" || s[0] == '#' {
+				continue
+			}
+			x, y := 0, uint16(0)
+			if _, err := fmt.Sscanf(s, "%d 0x%x", &x, &y); err != nil {
+				log.Fatalf("%q: could not parse %q", table.url, s)
+			}
+			if x < 0 || 120*94 <= x {
+				log.Fatalf("%q: JIS code %d is out of range", table.url, x)
+			}
+			mapping[x] = y
+			if reverse[y].table == -1 {
+				reverse[y] = entry{jisCode: x, table: i}
+			}
+		}
+		if err := scanner.Err(); err != nil {
+			log.Fatalf("%q: scanner error: %v", table.url, err)
+		}
+
+		fmt.Printf("// jis%sDecode is the decoding table from JIS %s code to Unicode.\n// It is defined at %s\n",
+			table.name, table.name, table.url)
+		fmt.Printf("var jis%sDecode = [...]uint16{\n", table.name)
+		for i, m := range mapping {
+			if m != 0 {
+				fmt.Printf("\t%d: 0x%04X,\n", i, m)
+			}
+		}
+		fmt.Printf("}\n\n")
+	}
+
+	// Any run of at least separation continuous zero entries in the reverse map will
+	// be a separate encode table.
+	const separation = 1024
+
+	intervals := []interval(nil)
+	low, high := -1, -1
+	for i, v := range reverse {
+		if v.table == -1 {
+			continue
+		}
+		if low < 0 {
+			low = i
+		} else if i-high >= separation {
+			if high >= 0 {
+				intervals = append(intervals, interval{low, high})
+			}
+			low = i
+		}
+		high = i + 1
+	}
+	if high >= 0 {
+		intervals = append(intervals, interval{low, high})
+	}
+	sort.Sort(byDecreasingLength(intervals))
+
+	fmt.Printf("const (\n")
+	fmt.Printf("\tjis0208    = 1\n")
+	fmt.Printf("\tjis0212    = 2\n")
+	fmt.Printf("\tcodeMask   = 0x7f\n")
+	fmt.Printf("\tcodeShift  = 7\n")
+	fmt.Printf("\ttableShift = 14\n")
+	fmt.Printf(")\n\n")
+
+	fmt.Printf("const numEncodeTables = %d\n\n", len(intervals))
+	fmt.Printf("// encodeX are the encoding tables from Unicode to JIS code,\n")
+	fmt.Printf("// sorted by decreasing length.\n")
+	for i, v := range intervals {
+		fmt.Printf("// encode%d: %5d entries for runes in [%5d, %5d).\n", i, v.len(), v.low, v.high)
+	}
+	fmt.Printf("//\n")
+	fmt.Printf("// The high two bits of the value record whether the JIS code comes from the\n")
+	fmt.Printf("// JIS0208 table (high bits == 1) or the JIS0212 table (high bits == 2).\n")
+	fmt.Printf("// The low 14 bits are two 7-bit unsigned integers j1 and j2 that form the\n")
+	fmt.Printf("// JIS code (94*j1 + j2) within that table.\n")
+	fmt.Printf("\n")
+
+	for i, v := range intervals {
+		fmt.Printf("const encode%dLow, encode%dHigh = %d, %d\n\n", i, i, v.low, v.high)
+		fmt.Printf("var encode%d = [...]uint16{\n", i)
+		for j := v.low; j < v.high; j++ {
+			x := reverse[j]
+			if x.table == -1 {
+				continue
+			}
+			fmt.Printf("\t%d - %d: jis%s<<14 | 0x%02X<<7 | 0x%02X,\n",
+				j, v.low, tables[x.table].name, x.jisCode/94, x.jisCode%94)
+		}
+		fmt.Printf("}\n\n")
+	}
+}
+
+// interval is a half-open interval [low, high).
+type interval struct {
+	low, high int
+}
+
+func (i interval) len() int { return i.high - i.low }
+
+// byDecreasingLength sorts intervals by decreasing length.
+type byDecreasingLength []interval
+
+func (b byDecreasingLength) Len() int           { return len(b) }
+func (b byDecreasingLength) Less(i, j int) bool { return b[i].len() > b[j].len() }
+func (b byDecreasingLength) Swap(i, j int)      { b[i], b[j] = b[j], b[i] }

vendor/golang.org/x/text/encoding/japanese/shiftjis.go

@@ -0,0 +1,189 @@
+// 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 japanese
+
+import (
+	"errors"
+	"unicode/utf8"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/internal"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/transform"
+)
+
+// ShiftJIS is the Shift JIS encoding, also known as Code Page 932 and
+// Windows-31J.
+var ShiftJIS encoding.Encoding = &shiftJIS
+
+var shiftJIS = internal.Encoding{
+	&internal.SimpleEncoding{shiftJISDecoder{}, shiftJISEncoder{}},
+	"Shift JIS",
+	identifier.ShiftJIS,
+}
+
+var errInvalidShiftJIS = errors.New("japanese: invalid Shift JIS encoding")
+
+type shiftJISDecoder struct{ transform.NopResetter }
+
+func (shiftJISDecoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+loop:
+	for ; nSrc < len(src); nSrc += size {
+		switch c0 := src[nSrc]; {
+		case c0 < utf8.RuneSelf:
+			r, size = rune(c0), 1
+
+		case 0xa1 <= c0 && c0 < 0xe0:
+			r, size = rune(c0)+(0xff61-0xa1), 1
+
+		case (0x81 <= c0 && c0 < 0xa0) || (0xe0 <= c0 && c0 < 0xfd):
+			if c0 <= 0x9f {
+				c0 -= 0x70
+			} else {
+				c0 -= 0xb0
+			}
+			c0 = 2*c0 - 0x21
+
+			if nSrc+1 >= len(src) {
+				err = transform.ErrShortSrc
+				break loop
+			}
+			c1 := src[nSrc+1]
+			switch {
+			case c1 < 0x40:
+				err = errInvalidShiftJIS
+				break loop
+			case c1 < 0x7f:
+				c0--
+				c1 -= 0x40
+			case c1 == 0x7f:
+				err = errInvalidShiftJIS
+				break loop
+			case c1 < 0x9f:
+				c0--
+				c1 -= 0x41
+			case c1 < 0xfd:
+				c1 -= 0x9f
+			default:
+				err = errInvalidShiftJIS
+				break loop
+			}
+			r, size = '\ufffd', 2
+			if i := int(c0)*94 + int(c1); i < len(jis0208Decode) {
+				r = rune(jis0208Decode[i])
+				if r == 0 {
+					r = '\ufffd'
+				}
+			}
+
+		default:
+			err = errInvalidShiftJIS
+			break loop
+		}
+
+		if nDst+utf8.RuneLen(r) > len(dst) {
+			err = transform.ErrShortDst
+			break loop
+		}
+		nDst += utf8.EncodeRune(dst[nDst:], r)
+	}
+	if atEOF && err == transform.ErrShortSrc {
+		err = errInvalidShiftJIS
+	}
+	return nDst, nSrc, err
+}
+
+type shiftJISEncoder struct{ transform.NopResetter }
+
+func (shiftJISEncoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+loop:
+	for ; nSrc < len(src); nSrc += size {
+		r = rune(src[nSrc])
+
+		// Decode a 1-byte rune.
+		if r < utf8.RuneSelf {
+			size = 1
+
+		} else {
+			// Decode a multi-byte rune.
+			r, size = utf8.DecodeRune(src[nSrc:])
+			if size == 1 {
+				// All valid runes of size 1 (those below utf8.RuneSelf) were
+				// handled above. We have invalid UTF-8 or we haven't seen the
+				// full character yet.
+				if !atEOF && !utf8.FullRune(src[nSrc:]) {
+					err = transform.ErrShortSrc
+					break loop
+				}
+			}
+
+			// func init checks that the switch covers all tables.
+			switch {
+			case encode0Low <= r && r < encode0High:
+				if r = rune(encode0[r-encode0Low]); r>>tableShift == jis0208 {
+					goto write2
+				}
+			case encode1Low <= r && r < encode1High:
+				if r = rune(encode1[r-encode1Low]); r>>tableShift == jis0208 {
+					goto write2
+				}
+			case encode2Low <= r && r < encode2High:
+				if r = rune(encode2[r-encode2Low]); r>>tableShift == jis0208 {
+					goto write2
+				}
+			case encode3Low <= r && r < encode3High:
+				if r = rune(encode3[r-encode3Low]); r>>tableShift == jis0208 {
+					goto write2
+				}
+			case encode4Low <= r && r < encode4High:
+				if r = rune(encode4[r-encode4Low]); r>>tableShift == jis0208 {
+					goto write2
+				}
+			case encode5Low <= r && r < encode5High:
+				if 0xff61 <= r && r < 0xffa0 {
+					r -= 0xff61 - 0xa1
+					goto write1
+				}
+				if r = rune(encode5[r-encode5Low]); r>>tableShift == jis0208 {
+					goto write2
+				}
+			}
+			err = internal.ErrASCIIReplacement
+			break
+		}
+
+	write1:
+		if nDst >= len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst] = uint8(r)
+		nDst++
+		continue
+
+	write2:
+		j1 := uint8(r>>codeShift) & codeMask
+		j2 := uint8(r) & codeMask
+		if nDst+2 > len(dst) {
+			err = transform.ErrShortDst
+			break loop
+		}
+		if j1 <= 61 {
+			dst[nDst+0] = 129 + j1/2
+		} else {
+			dst[nDst+0] = 193 + j1/2
+		}
+		if j1&1 == 0 {
+			dst[nDst+1] = j2 + j2/63 + 64
+		} else {
+			dst[nDst+1] = j2 + 159
+		}
+		nDst += 2
+		continue
+	}
+	return nDst, nSrc, err
+}

vendor/golang.org/x/text/encoding/korean/euckr.go

@@ -0,0 +1,178 @@
+// 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 korean
+
+import (
+	"errors"
+	"unicode/utf8"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/internal"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/transform"
+)
+
+// All is a list of all defined encodings in this package.
+var All = []encoding.Encoding{EUCKR}
+
+// EUCKR is the EUC-KR encoding, also known as Code Page 949.
+var EUCKR encoding.Encoding = &eucKR
+
+var eucKR = internal.Encoding{
+	&internal.SimpleEncoding{eucKRDecoder{}, eucKREncoder{}},
+	"EUC-KR",
+	identifier.EUCKR,
+}
+
+var errInvalidEUCKR = errors.New("korean: invalid EUC-KR encoding")
+
+type eucKRDecoder struct{ transform.NopResetter }
+
+func (eucKRDecoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+loop:
+	for ; nSrc < len(src); nSrc += size {
+		switch c0 := src[nSrc]; {
+		case c0 < utf8.RuneSelf:
+			r, size = rune(c0), 1
+
+		case 0x81 <= c0 && c0 < 0xff:
+			if nSrc+1 >= len(src) {
+				err = transform.ErrShortSrc
+				break loop
+			}
+			c1 := src[nSrc+1]
+			if c0 < 0xc7 {
+				r = 178 * rune(c0-0x81)
+				switch {
+				case 0x41 <= c1 && c1 < 0x5b:
+					r += rune(c1) - (0x41 - 0*26)
+				case 0x61 <= c1 && c1 < 0x7b:
+					r += rune(c1) - (0x61 - 1*26)
+				case 0x81 <= c1 && c1 < 0xff:
+					r += rune(c1) - (0x81 - 2*26)
+				default:
+					err = errInvalidEUCKR
+					break loop
+				}
+			} else if 0xa1 <= c1 && c1 < 0xff {
+				r = 178*(0xc7-0x81) + rune(c0-0xc7)*94 + rune(c1-0xa1)
+			} else {
+				err = errInvalidEUCKR
+				break loop
+			}
+			if int(r) < len(decode) {
+				r = rune(decode[r])
+				if r == 0 {
+					r = '\ufffd'
+				}
+			} else {
+				r = '\ufffd'
+			}
+			size = 2
+
+		default:
+			err = errInvalidEUCKR
+			break loop
+		}
+
+		if nDst+utf8.RuneLen(r) > len(dst) {
+			err = transform.ErrShortDst
+			break loop
+		}
+		nDst += utf8.EncodeRune(dst[nDst:], r)
+	}
+	if atEOF && err == transform.ErrShortSrc {
+		err = errInvalidEUCKR
+	}
+	return nDst, nSrc, err
+}
+
+type eucKREncoder struct{ transform.NopResetter }
+
+func (eucKREncoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+	for ; nSrc < len(src); nSrc += size {
+		r = rune(src[nSrc])
+
+		// Decode a 1-byte rune.
+		if r < utf8.RuneSelf {
+			size = 1
+
+			if nDst >= len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			dst[nDst] = uint8(r)
+			nDst++
+			continue
+
+		} else {
+			// Decode a multi-byte rune.
+			r, size = utf8.DecodeRune(src[nSrc:])
+			if size == 1 {
+				// All valid runes of size 1 (those below utf8.RuneSelf) were
+				// handled above. We have invalid UTF-8 or we haven't seen the
+				// full character yet.
+				if !atEOF && !utf8.FullRune(src[nSrc:]) {
+					err = transform.ErrShortSrc
+					break
+				}
+			}
+
+			// func init checks that the switch covers all tables.
+			switch {
+			case encode0Low <= r && r < encode0High:
+				if r = rune(encode0[r-encode0Low]); r != 0 {
+					goto write2
+				}
+			case encode1Low <= r && r < encode1High:
+				if r = rune(encode1[r-encode1Low]); r != 0 {
+					goto write2
+				}
+			case encode2Low <= r && r < encode2High:
+				if r = rune(encode2[r-encode2Low]); r != 0 {
+					goto write2
+				}
+			case encode3Low <= r && r < encode3High:
+				if r = rune(encode3[r-encode3Low]); r != 0 {
+					goto write2
+				}
+			case encode4Low <= r && r < encode4High:
+				if r = rune(encode4[r-encode4Low]); r != 0 {
+					goto write2
+				}
+			case encode5Low <= r && r < encode5High:
+				if r = rune(encode5[r-encode5Low]); r != 0 {
+					goto write2
+				}
+			case encode6Low <= r && r < encode6High:
+				if r = rune(encode6[r-encode6Low]); r != 0 {
+					goto write2
+				}
+			}
+			err = internal.ErrASCIIReplacement
+			break
+		}
+
+	write2:
+		if nDst+2 > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst+0] = uint8(r >> 8)
+		dst[nDst+1] = uint8(r)
+		nDst += 2
+		continue
+	}
+	return nDst, nSrc, err
+}
+
+func init() {
+	// Check that the hard-coded encode switch covers all tables.
+	if numEncodeTables != 7 {
+		panic("bad numEncodeTables")
+	}
+}

vendor/golang.org/x/text/encoding/korean/maketables.go

@@ -0,0 +1,143 @@
+// 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
+
+package main
+
+// This program generates tables.go:
+//	go run maketables.go | gofmt > tables.go
+
+import (
+	"bufio"
+	"fmt"
+	"log"
+	"net/http"
+	"sort"
+	"strings"
+)
+
+func main() {
+	fmt.Printf("// generated by go run maketables.go; DO NOT EDIT\n\n")
+	fmt.Printf("// Package korean provides Korean encodings such as EUC-KR.\n")
+	fmt.Printf(`package korean // import "golang.org/x/text/encoding/korean"` + "\n\n")
+
+	res, err := http.Get("http://encoding.spec.whatwg.org/index-euc-kr.txt")
+	if err != nil {
+		log.Fatalf("Get: %v", err)
+	}
+	defer res.Body.Close()
+
+	mapping := [65536]uint16{}
+	reverse := [65536]uint16{}
+
+	scanner := bufio.NewScanner(res.Body)
+	for scanner.Scan() {
+		s := strings.TrimSpace(scanner.Text())
+		if s == "" || s[0] == '#' {
+			continue
+		}
+		x, y := uint16(0), uint16(0)
+		if _, err := fmt.Sscanf(s, "%d 0x%x", &x, &y); err != nil {
+			log.Fatalf("could not parse %q", s)
+		}
+		if x < 0 || 178*(0xc7-0x81)+(0xfe-0xc7)*94+(0xff-0xa1) <= x {
+			log.Fatalf("EUC-KR code %d is out of range", x)
+		}
+		mapping[x] = y
+		if reverse[y] == 0 {
+			c0, c1 := uint16(0), uint16(0)
+			if x < 178*(0xc7-0x81) {
+				c0 = uint16(x/178) + 0x81
+				c1 = uint16(x % 178)
+				switch {
+				case c1 < 1*26:
+					c1 += 0x41
+				case c1 < 2*26:
+					c1 += 0x47
+				default:
+					c1 += 0x4d
+				}
+			} else {
+				x -= 178 * (0xc7 - 0x81)
+				c0 = uint16(x/94) + 0xc7
+				c1 = uint16(x%94) + 0xa1
+			}
+			reverse[y] = c0<<8 | c1
+		}
+	}
+	if err := scanner.Err(); err != nil {
+		log.Fatalf("scanner error: %v", err)
+	}
+
+	fmt.Printf("// decode is the decoding table from EUC-KR code to Unicode.\n")
+	fmt.Printf("// It is defined at http://encoding.spec.whatwg.org/index-euc-kr.txt\n")
+	fmt.Printf("var decode = [...]uint16{\n")
+	for i, v := range mapping {
+		if v != 0 {
+			fmt.Printf("\t%d: 0x%04X,\n", i, v)
+		}
+	}
+	fmt.Printf("}\n\n")
+
+	// Any run of at least separation continuous zero entries in the reverse map will
+	// be a separate encode table.
+	const separation = 1024
+
+	intervals := []interval(nil)
+	low, high := -1, -1
+	for i, v := range reverse {
+		if v == 0 {
+			continue
+		}
+		if low < 0 {
+			low = i
+		} else if i-high >= separation {
+			if high >= 0 {
+				intervals = append(intervals, interval{low, high})
+			}
+			low = i
+		}
+		high = i + 1
+	}
+	if high >= 0 {
+		intervals = append(intervals, interval{low, high})
+	}
+	sort.Sort(byDecreasingLength(intervals))
+
+	fmt.Printf("const numEncodeTables = %d\n\n", len(intervals))
+	fmt.Printf("// encodeX are the encoding tables from Unicode to EUC-KR code,\n")
+	fmt.Printf("// sorted by decreasing length.\n")
+	for i, v := range intervals {
+		fmt.Printf("// encode%d: %5d entries for runes in [%5d, %5d).\n", i, v.len(), v.low, v.high)
+	}
+	fmt.Printf("\n")
+
+	for i, v := range intervals {
+		fmt.Printf("const encode%dLow, encode%dHigh = %d, %d\n\n", i, i, v.low, v.high)
+		fmt.Printf("var encode%d = [...]uint16{\n", i)
+		for j := v.low; j < v.high; j++ {
+			x := reverse[j]
+			if x == 0 {
+				continue
+			}
+			fmt.Printf("\t%d-%d: 0x%04X,\n", j, v.low, x)
+		}
+		fmt.Printf("}\n\n")
+	}
+}
+
+// interval is a half-open interval [low, high).
+type interval struct {
+	low, high int
+}
+
+func (i interval) len() int { return i.high - i.low }
+
+// byDecreasingLength sorts intervals by decreasing length.
+type byDecreasingLength []interval
+
+func (b byDecreasingLength) Len() int           { return len(b) }
+func (b byDecreasingLength) Less(i, j int) bool { return b[i].len() > b[j].len() }
+func (b byDecreasingLength) Swap(i, j int)      { b[i], b[j] = b[j], b[i] }

vendor/golang.org/x/text/encoding/simplifiedchinese/all.go

@@ -0,0 +1,12 @@
+// 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 simplifiedchinese
+
+import (
+	"golang.org/x/text/encoding"
+)
+
+// All is a list of all defined encodings in this package.
+var All = []encoding.Encoding{GB18030, GBK, HZGB2312}

vendor/golang.org/x/text/encoding/simplifiedchinese/gbk.go

@@ -0,0 +1,281 @@
+// 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 simplifiedchinese
+
+import (
+	"errors"
+	"unicode/utf8"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/internal"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/transform"
+)
+
+var (
+	// GB18030 is the GB18030 encoding.
+	GB18030 encoding.Encoding = &gbk18030
+	// GBK is the GBK encoding. It encodes an extension of the GB2312 character set
+	// and is also known as Code Page 936.
+	GBK encoding.Encoding = &gbk
+)
+
+var gbk = internal.Encoding{
+	&internal.SimpleEncoding{
+		gbkDecoder{gb18030: false},
+		gbkEncoder{gb18030: false},
+	},
+	"GBK",
+	identifier.GBK,
+}
+
+var gbk18030 = internal.Encoding{
+	&internal.SimpleEncoding{
+		gbkDecoder{gb18030: true},
+		gbkEncoder{gb18030: true},
+	},
+	"GB18030",
+	identifier.GB18030,
+}
+
+var (
+	errInvalidGB18030 = errors.New("simplifiedchinese: invalid GB18030 encoding")
+	errInvalidGBK     = errors.New("simplifiedchinese: invalid GBK encoding")
+)
+
+type gbkDecoder struct {
+	transform.NopResetter
+	gb18030 bool
+}
+
+func (d gbkDecoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+loop:
+	for ; nSrc < len(src); nSrc += size {
+		switch c0 := src[nSrc]; {
+		case c0 < utf8.RuneSelf:
+			r, size = rune(c0), 1
+
+		// Microsoft's Code Page 936 extends GBK 1.0 to encode the euro sign U+20AC
+		// as 0x80. The HTML5 specification at http://encoding.spec.whatwg.org/#gbk
+		// says to treat "gbk" as Code Page 936.
+		case c0 == 0x80:
+			r, size = '€', 1
+
+		case c0 < 0xff:
+			if nSrc+1 >= len(src) {
+				err = transform.ErrShortSrc
+				break loop
+			}
+			c1 := src[nSrc+1]
+			switch {
+			case 0x40 <= c1 && c1 < 0x7f:
+				c1 -= 0x40
+			case 0x80 <= c1 && c1 < 0xff:
+				c1 -= 0x41
+			case d.gb18030 && 0x30 <= c1 && c1 < 0x40:
+				if nSrc+3 >= len(src) {
+					err = transform.ErrShortSrc
+					break loop
+				}
+				c2 := src[nSrc+2]
+				if c2 < 0x81 || 0xff <= c2 {
+					err = errInvalidGB18030
+					break loop
+				}
+				c3 := src[nSrc+3]
+				if c3 < 0x30 || 0x3a <= c3 {
+					err = errInvalidGB18030
+					break loop
+				}
+				size = 4
+				r = ((rune(c0-0x81)*10+rune(c1-0x30))*126+rune(c2-0x81))*10 + rune(c3-0x30)
+				if r < 39420 {
+					i, j := 0, len(gb18030)
+					for i < j {
+						h := i + (j-i)/2
+						if r >= rune(gb18030[h][0]) {
+							i = h + 1
+						} else {
+							j = h
+						}
+					}
+					dec := &gb18030[i-1]
+					r += rune(dec[1]) - rune(dec[0])
+					goto write
+				}
+				r -= 189000
+				if 0 <= r && r < 0x100000 {
+					r += 0x10000
+					goto write
+				}
+				err = errInvalidGB18030
+				break loop
+			default:
+				if d.gb18030 {
+					err = errInvalidGB18030
+				} else {
+					err = errInvalidGBK
+				}
+				break loop
+			}
+			r, size = '\ufffd', 2
+			if i := int(c0-0x81)*190 + int(c1); i < len(decode) {
+				r = rune(decode[i])
+				if r == 0 {
+					r = '\ufffd'
+				}
+			}
+
+		default:
+			if d.gb18030 {
+				err = errInvalidGB18030
+			} else {
+				err = errInvalidGBK
+			}
+			break loop
+		}
+
+	write:
+		if nDst+utf8.RuneLen(r) > len(dst) {
+			err = transform.ErrShortDst
+			break loop
+		}
+		nDst += utf8.EncodeRune(dst[nDst:], r)
+	}
+	if atEOF && err == transform.ErrShortSrc {
+		if d.gb18030 {
+			err = errInvalidGB18030
+		} else {
+			err = errInvalidGBK
+		}
+	}
+	return nDst, nSrc, err
+}
+
+type gbkEncoder struct {
+	transform.NopResetter
+	gb18030 bool
+}
+
+func (e gbkEncoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, r2, size := rune(0), rune(0), 0
+	for ; nSrc < len(src); nSrc += size {
+		r = rune(src[nSrc])
+
+		// Decode a 1-byte rune.
+		if r < utf8.RuneSelf {
+			size = 1
+
+		} else {
+			// Decode a multi-byte rune.
+			r, size = utf8.DecodeRune(src[nSrc:])
+			if size == 1 {
+				// All valid runes of size 1 (those below utf8.RuneSelf) were
+				// handled above. We have invalid UTF-8 or we haven't seen the
+				// full character yet.
+				if !atEOF && !utf8.FullRune(src[nSrc:]) {
+					err = transform.ErrShortSrc
+					break
+				}
+			}
+
+			// func init checks that the switch covers all tables.
+			switch {
+			case encode0Low <= r && r < encode0High:
+				if r2 = rune(encode0[r-encode0Low]); r2 != 0 {
+					goto write2
+				}
+			case encode1Low <= r && r < encode1High:
+				// Microsoft's Code Page 936 extends GBK 1.0 to encode the euro sign U+20AC
+				// as 0x80. The HTML5 specification at http://encoding.spec.whatwg.org/#gbk
+				// says to treat "gbk" as Code Page 936.
+				if r == '€' {
+					r = 0x80
+					goto write1
+				}
+				if r2 = rune(encode1[r-encode1Low]); r2 != 0 {
+					goto write2
+				}
+			case encode2Low <= r && r < encode2High:
+				if r2 = rune(encode2[r-encode2Low]); r2 != 0 {
+					goto write2
+				}
+			case encode3Low <= r && r < encode3High:
+				if r2 = rune(encode3[r-encode3Low]); r2 != 0 {
+					goto write2
+				}
+			case encode4Low <= r && r < encode4High:
+				if r2 = rune(encode4[r-encode4Low]); r2 != 0 {
+					goto write2
+				}
+			}
+
+			if e.gb18030 {
+				if r < 0x10000 {
+					i, j := 0, len(gb18030)
+					for i < j {
+						h := i + (j-i)/2
+						if r >= rune(gb18030[h][1]) {
+							i = h + 1
+						} else {
+							j = h
+						}
+					}
+					dec := &gb18030[i-1]
+					r += rune(dec[0]) - rune(dec[1])
+					goto write4
+				} else if r < 0x110000 {
+					r += 189000 - 0x10000
+					goto write4
+				}
+			}
+			err = internal.ErrASCIIReplacement
+			break
+		}
+
+	write1:
+		if nDst >= len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst] = uint8(r)
+		nDst++
+		continue
+
+	write2:
+		if nDst+2 > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst+0] = uint8(r2 >> 8)
+		dst[nDst+1] = uint8(r2)
+		nDst += 2
+		continue
+
+	write4:
+		if nDst+4 > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst+3] = uint8(r%10 + 0x30)
+		r /= 10
+		dst[nDst+2] = uint8(r%126 + 0x81)
+		r /= 126
+		dst[nDst+1] = uint8(r%10 + 0x30)
+		r /= 10
+		dst[nDst+0] = uint8(r + 0x81)
+		nDst += 4
+		continue
+	}
+	return nDst, nSrc, err
+}
+
+func init() {
+	// Check that the hard-coded encode switch covers all tables.
+	if numEncodeTables != 5 {
+		panic("bad numEncodeTables")
+	}
+}

vendor/golang.org/x/text/encoding/simplifiedchinese/hzgb2312.go

@@ -0,0 +1,240 @@
+// 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 simplifiedchinese
+
+import (
+	"errors"
+	"unicode/utf8"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/internal"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/transform"
+)
+
+// HZGB2312 is the HZ-GB2312 encoding.
+var HZGB2312 encoding.Encoding = &hzGB2312
+
+var hzGB2312 = internal.Encoding{
+	internal.FuncEncoding{hzGB2312NewDecoder, hzGB2312NewEncoder},
+	"HZ-GB2312",
+	identifier.HZGB2312,
+}
+
+func hzGB2312NewDecoder() transform.Transformer {
+	return new(hzGB2312Decoder)
+}
+
+func hzGB2312NewEncoder() transform.Transformer {
+	return new(hzGB2312Encoder)
+}
+
+var errInvalidHZGB2312 = errors.New("simplifiedchinese: invalid HZ-GB2312 encoding")
+
+const (
+	asciiState = iota
+	gbState
+)
+
+type hzGB2312Decoder int
+
+func (d *hzGB2312Decoder) Reset() {
+	*d = asciiState
+}
+
+func (d *hzGB2312Decoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+loop:
+	for ; nSrc < len(src); nSrc += size {
+		c0 := src[nSrc]
+		if c0 >= utf8.RuneSelf {
+			err = errInvalidHZGB2312
+			break loop
+		}
+
+		if c0 == '~' {
+			if nSrc+1 >= len(src) {
+				err = transform.ErrShortSrc
+				break loop
+			}
+			size = 2
+			switch src[nSrc+1] {
+			case '{':
+				*d = gbState
+				continue
+			case '}':
+				*d = asciiState
+				continue
+			case '~':
+				if nDst >= len(dst) {
+					err = transform.ErrShortDst
+					break loop
+				}
+				dst[nDst] = '~'
+				nDst++
+				continue
+			case '\n':
+				continue
+			default:
+				err = errInvalidHZGB2312
+				break loop
+			}
+		}
+
+		if *d == asciiState {
+			r, size = rune(c0), 1
+		} else {
+			if nSrc+1 >= len(src) {
+				err = transform.ErrShortSrc
+				break loop
+			}
+			c1 := src[nSrc+1]
+			if c0 < 0x21 || 0x7e <= c0 || c1 < 0x21 || 0x7f <= c1 {
+				err = errInvalidHZGB2312
+				break loop
+			}
+
+			r, size = '\ufffd', 2
+			if i := int(c0-0x01)*190 + int(c1+0x3f); i < len(decode) {
+				r = rune(decode[i])
+				if r == 0 {
+					r = '\ufffd'
+				}
+			}
+		}
+
+		if nDst+utf8.RuneLen(r) > len(dst) {
+			err = transform.ErrShortDst
+			break loop
+		}
+		nDst += utf8.EncodeRune(dst[nDst:], r)
+	}
+	if atEOF && err == transform.ErrShortSrc {
+		err = errInvalidHZGB2312
+	}
+	return nDst, nSrc, err
+}
+
+type hzGB2312Encoder int
+
+func (d *hzGB2312Encoder) Reset() {
+	*d = asciiState
+}
+
+func (e *hzGB2312Encoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+	for ; nSrc < len(src); nSrc += size {
+		r = rune(src[nSrc])
+
+		// Decode a 1-byte rune.
+		if r < utf8.RuneSelf {
+			size = 1
+			if r == '~' {
+				if nDst+2 > len(dst) {
+					err = transform.ErrShortDst
+					break
+				}
+				dst[nDst+0] = '~'
+				dst[nDst+1] = '~'
+				nDst += 2
+				continue
+			} else if *e != asciiState {
+				if nDst+3 > len(dst) {
+					err = transform.ErrShortDst
+					break
+				}
+				*e = asciiState
+				dst[nDst+0] = '~'
+				dst[nDst+1] = '}'
+				nDst += 2
+			} else if nDst >= len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			dst[nDst] = uint8(r)
+			nDst += 1
+			continue
+
+		}
+
+		// Decode a multi-byte rune.
+		r, size = utf8.DecodeRune(src[nSrc:])
+		if size == 1 {
+			// All valid runes of size 1 (those below utf8.RuneSelf) were
+			// handled above. We have invalid UTF-8 or we haven't seen the
+			// full character yet.
+			if !atEOF && !utf8.FullRune(src[nSrc:]) {
+				err = transform.ErrShortSrc
+				break
+			}
+		}
+
+		// func init checks that the switch covers all tables.
+		switch {
+		case encode0Low <= r && r < encode0High:
+			if r = rune(encode0[r-encode0Low]); r != 0 {
+				goto writeGB
+			}
+		case encode1Low <= r && r < encode1High:
+			if r = rune(encode1[r-encode1Low]); r != 0 {
+				goto writeGB
+			}
+		case encode2Low <= r && r < encode2High:
+			if r = rune(encode2[r-encode2Low]); r != 0 {
+				goto writeGB
+			}
+		case encode3Low <= r && r < encode3High:
+			if r = rune(encode3[r-encode3Low]); r != 0 {
+				goto writeGB
+			}
+		case encode4Low <= r && r < encode4High:
+			if r = rune(encode4[r-encode4Low]); r != 0 {
+				goto writeGB
+			}
+		}
+
+	terminateInASCIIState:
+		// Switch back to ASCII state in case of error so that an ASCII
+		// replacement character can be written in the correct state.
+		if *e != asciiState {
+			if nDst+2 > len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			dst[nDst+0] = '~'
+			dst[nDst+1] = '}'
+			nDst += 2
+		}
+		err = internal.ErrASCIIReplacement
+		break
+
+	writeGB:
+		c0 := uint8(r>>8) - 0x80
+		c1 := uint8(r) - 0x80
+		if c0 < 0x21 || 0x7e <= c0 || c1 < 0x21 || 0x7f <= c1 {
+			goto terminateInASCIIState
+		}
+		if *e == asciiState {
+			if nDst+4 > len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			*e = gbState
+			dst[nDst+0] = '~'
+			dst[nDst+1] = '{'
+			nDst += 2
+		} else if nDst+2 > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst+0] = c0
+		dst[nDst+1] = c1
+		nDst += 2
+		continue
+	}
+	// TODO: should one always terminate in ASCII state to make it safe to
+	// concatenate two HZ-GB2312-encoded strings?
+	return nDst, nSrc, err
+}

vendor/golang.org/x/text/encoding/simplifiedchinese/maketables.go

@@ -0,0 +1,161 @@
+// 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
+
+package main
+
+// This program generates tables.go:
+//	go run maketables.go | gofmt > tables.go
+
+import (
+	"bufio"
+	"fmt"
+	"log"
+	"net/http"
+	"sort"
+	"strings"
+)
+
+func main() {
+	fmt.Printf("// generated by go run maketables.go; DO NOT EDIT\n\n")
+	fmt.Printf("// Package simplifiedchinese provides Simplified Chinese encodings such as GBK.\n")
+	fmt.Printf(`package simplifiedchinese // import "golang.org/x/text/encoding/simplifiedchinese"` + "\n\n")
+
+	printGB18030()
+	printGBK()
+}
+
+func printGB18030() {
+	res, err := http.Get("http://encoding.spec.whatwg.org/index-gb18030.txt")
+	if err != nil {
+		log.Fatalf("Get: %v", err)
+	}
+	defer res.Body.Close()
+
+	fmt.Printf("// gb18030 is the table from http://encoding.spec.whatwg.org/index-gb18030.txt\n")
+	fmt.Printf("var gb18030 = [...][2]uint16{\n")
+	scanner := bufio.NewScanner(res.Body)
+	for scanner.Scan() {
+		s := strings.TrimSpace(scanner.Text())
+		if s == "" || s[0] == '#' {
+			continue
+		}
+		x, y := uint32(0), uint32(0)
+		if _, err := fmt.Sscanf(s, "%d 0x%x", &x, &y); err != nil {
+			log.Fatalf("could not parse %q", s)
+		}
+		if x < 0x10000 && y < 0x10000 {
+			fmt.Printf("\t{0x%04x, 0x%04x},\n", x, y)
+		}
+	}
+	fmt.Printf("}\n\n")
+}
+
+func printGBK() {
+	res, err := http.Get("http://encoding.spec.whatwg.org/index-gbk.txt")
+	if err != nil {
+		log.Fatalf("Get: %v", err)
+	}
+	defer res.Body.Close()
+
+	mapping := [65536]uint16{}
+	reverse := [65536]uint16{}
+
+	scanner := bufio.NewScanner(res.Body)
+	for scanner.Scan() {
+		s := strings.TrimSpace(scanner.Text())
+		if s == "" || s[0] == '#' {
+			continue
+		}
+		x, y := uint16(0), uint16(0)
+		if _, err := fmt.Sscanf(s, "%d 0x%x", &x, &y); err != nil {
+			log.Fatalf("could not parse %q", s)
+		}
+		if x < 0 || 126*190 <= x {
+			log.Fatalf("GBK code %d is out of range", x)
+		}
+		mapping[x] = y
+		if reverse[y] == 0 {
+			c0, c1 := x/190, x%190
+			if c1 >= 0x3f {
+				c1++
+			}
+			reverse[y] = (0x81+c0)<<8 | (0x40 + c1)
+		}
+	}
+	if err := scanner.Err(); err != nil {
+		log.Fatalf("scanner error: %v", err)
+	}
+
+	fmt.Printf("// decode is the decoding table from GBK code to Unicode.\n")
+	fmt.Printf("// It is defined at http://encoding.spec.whatwg.org/index-gbk.txt\n")
+	fmt.Printf("var decode = [...]uint16{\n")
+	for i, v := range mapping {
+		if v != 0 {
+			fmt.Printf("\t%d: 0x%04X,\n", i, v)
+		}
+	}
+	fmt.Printf("}\n\n")
+
+	// Any run of at least separation continuous zero entries in the reverse map will
+	// be a separate encode table.
+	const separation = 1024
+
+	intervals := []interval(nil)
+	low, high := -1, -1
+	for i, v := range reverse {
+		if v == 0 {
+			continue
+		}
+		if low < 0 {
+			low = i
+		} else if i-high >= separation {
+			if high >= 0 {
+				intervals = append(intervals, interval{low, high})
+			}
+			low = i
+		}
+		high = i + 1
+	}
+	if high >= 0 {
+		intervals = append(intervals, interval{low, high})
+	}
+	sort.Sort(byDecreasingLength(intervals))
+
+	fmt.Printf("const numEncodeTables = %d\n\n", len(intervals))
+	fmt.Printf("// encodeX are the encoding tables from Unicode to GBK code,\n")
+	fmt.Printf("// sorted by decreasing length.\n")
+	for i, v := range intervals {
+		fmt.Printf("// encode%d: %5d entries for runes in [%5d, %5d).\n", i, v.len(), v.low, v.high)
+	}
+	fmt.Printf("\n")
+
+	for i, v := range intervals {
+		fmt.Printf("const encode%dLow, encode%dHigh = %d, %d\n\n", i, i, v.low, v.high)
+		fmt.Printf("var encode%d = [...]uint16{\n", i)
+		for j := v.low; j < v.high; j++ {
+			x := reverse[j]
+			if x == 0 {
+				continue
+			}
+			fmt.Printf("\t%d-%d: 0x%04X,\n", j, v.low, x)
+		}
+		fmt.Printf("}\n\n")
+	}
+}
+
+// interval is a half-open interval [low, high).
+type interval struct {
+	low, high int
+}
+
+func (i interval) len() int { return i.high - i.low }
+
+// byDecreasingLength sorts intervals by decreasing length.
+type byDecreasingLength []interval
+
+func (b byDecreasingLength) Len() int           { return len(b) }
+func (b byDecreasingLength) Less(i, j int) bool { return b[i].len() > b[j].len() }
+func (b byDecreasingLength) Swap(i, j int)      { b[i], b[j] = b[j], b[i] }

vendor/golang.org/x/text/encoding/traditionalchinese/big5.go

@@ -0,0 +1,198 @@
+// 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 traditionalchinese
+
+import (
+	"errors"
+	"unicode/utf8"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/internal"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/transform"
+)
+
+// All is a list of all defined encodings in this package.
+var All = []encoding.Encoding{Big5}
+
+// Big5 is the Big5 encoding, also known as Code Page 950.
+var Big5 encoding.Encoding = &big5
+
+var big5 = internal.Encoding{
+	&internal.SimpleEncoding{big5Decoder{}, big5Encoder{}},
+	"Big5",
+	identifier.Big5,
+}
+
+var errInvalidBig5 = errors.New("traditionalchinese: invalid Big5 encoding")
+
+type big5Decoder struct{ transform.NopResetter }
+
+func (big5Decoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size, s := rune(0), 0, ""
+loop:
+	for ; nSrc < len(src); nSrc += size {
+		switch c0 := src[nSrc]; {
+		case c0 < utf8.RuneSelf:
+			r, size = rune(c0), 1
+
+		case 0x81 <= c0 && c0 < 0xff:
+			if nSrc+1 >= len(src) {
+				err = transform.ErrShortSrc
+				break loop
+			}
+			c1 := src[nSrc+1]
+			switch {
+			case 0x40 <= c1 && c1 < 0x7f:
+				c1 -= 0x40
+			case 0xa1 <= c1 && c1 < 0xff:
+				c1 -= 0x62
+			default:
+				err = errInvalidBig5
+				break loop
+			}
+			r, size = '\ufffd', 2
+			if i := int(c0-0x81)*157 + int(c1); i < len(decode) {
+				if 1133 <= i && i < 1167 {
+					// The two-rune special cases for LATIN CAPITAL / SMALL E WITH CIRCUMFLEX
+					// AND MACRON / CARON are from http://encoding.spec.whatwg.org/#big5
+					switch i {
+					case 1133:
+						s = "\u00CA\u0304"
+						goto writeStr
+					case 1135:
+						s = "\u00CA\u030C"
+						goto writeStr
+					case 1164:
+						s = "\u00EA\u0304"
+						goto writeStr
+					case 1166:
+						s = "\u00EA\u030C"
+						goto writeStr
+					}
+				}
+				r = rune(decode[i])
+				if r == 0 {
+					r = '\ufffd'
+				}
+			}
+
+		default:
+			err = errInvalidBig5
+			break loop
+		}
+
+		if nDst+utf8.RuneLen(r) > len(dst) {
+			err = transform.ErrShortDst
+			break loop
+		}
+		nDst += utf8.EncodeRune(dst[nDst:], r)
+		continue loop
+
+	writeStr:
+		if nDst+len(s) > len(dst) {
+			err = transform.ErrShortDst
+			break loop
+		}
+		nDst += copy(dst[nDst:], s)
+		continue loop
+	}
+	if atEOF && err == transform.ErrShortSrc {
+		err = errInvalidBig5
+	}
+	return nDst, nSrc, err
+}
+
+type big5Encoder struct{ transform.NopResetter }
+
+func (big5Encoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	r, size := rune(0), 0
+	for ; nSrc < len(src); nSrc += size {
+		r = rune(src[nSrc])
+
+		// Decode a 1-byte rune.
+		if r < utf8.RuneSelf {
+			size = 1
+			if nDst >= len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			dst[nDst] = uint8(r)
+			nDst++
+			continue
+
+		} else {
+			// Decode a multi-byte rune.
+			r, size = utf8.DecodeRune(src[nSrc:])
+			if size == 1 {
+				// All valid runes of size 1 (those below utf8.RuneSelf) were
+				// handled above. We have invalid UTF-8 or we haven't seen the
+				// full character yet.
+				if !atEOF && !utf8.FullRune(src[nSrc:]) {
+					err = transform.ErrShortSrc
+					break
+				}
+			}
+		}
+
+		if r >= utf8.RuneSelf {
+			// func init checks that the switch covers all tables.
+			switch {
+			case encode0Low <= r && r < encode0High:
+				if r = rune(encode0[r-encode0Low]); r != 0 {
+					goto write2
+				}
+			case encode1Low <= r && r < encode1High:
+				if r = rune(encode1[r-encode1Low]); r != 0 {
+					goto write2
+				}
+			case encode2Low <= r && r < encode2High:
+				if r = rune(encode2[r-encode2Low]); r != 0 {
+					goto write2
+				}
+			case encode3Low <= r && r < encode3High:
+				if r = rune(encode3[r-encode3Low]); r != 0 {
+					goto write2
+				}
+			case encode4Low <= r && r < encode4High:
+				if r = rune(encode4[r-encode4Low]); r != 0 {
+					goto write2
+				}
+			case encode5Low <= r && r < encode5High:
+				if r = rune(encode5[r-encode5Low]); r != 0 {
+					goto write2
+				}
+			case encode6Low <= r && r < encode6High:
+				if r = rune(encode6[r-encode6Low]); r != 0 {
+					goto write2
+				}
+			case encode7Low <= r && r < encode7High:
+				if r = rune(encode7[r-encode7Low]); r != 0 {
+					goto write2
+				}
+			}
+			err = internal.ErrASCIIReplacement
+			break
+		}
+
+	write2:
+		if nDst+2 > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst+0] = uint8(r >> 8)
+		dst[nDst+1] = uint8(r)
+		nDst += 2
+		continue
+	}
+	return nDst, nSrc, err
+}
+
+func init() {
+	// Check that the hard-coded encode switch covers all tables.
+	if numEncodeTables != 8 {
+		panic("bad numEncodeTables")
+	}
+}

vendor/golang.org/x/text/encoding/traditionalchinese/maketables.go

@@ -0,0 +1,140 @@
+// 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
+
+package main
+
+// This program generates tables.go:
+//	go run maketables.go | gofmt > tables.go
+
+import (
+	"bufio"
+	"fmt"
+	"log"
+	"net/http"
+	"sort"
+	"strings"
+)
+
+func main() {
+	fmt.Printf("// generated by go run maketables.go; DO NOT EDIT\n\n")
+	fmt.Printf("// Package traditionalchinese provides Traditional Chinese encodings such as Big5.\n")
+	fmt.Printf(`package traditionalchinese // import "golang.org/x/text/encoding/traditionalchinese"` + "\n\n")
+
+	res, err := http.Get("http://encoding.spec.whatwg.org/index-big5.txt")
+	if err != nil {
+		log.Fatalf("Get: %v", err)
+	}
+	defer res.Body.Close()
+
+	mapping := [65536]uint32{}
+	reverse := [65536 * 4]uint16{}
+
+	scanner := bufio.NewScanner(res.Body)
+	for scanner.Scan() {
+		s := strings.TrimSpace(scanner.Text())
+		if s == "" || s[0] == '#' {
+			continue
+		}
+		x, y := uint16(0), uint32(0)
+		if _, err := fmt.Sscanf(s, "%d 0x%x", &x, &y); err != nil {
+			log.Fatalf("could not parse %q", s)
+		}
+		if x < 0 || 126*157 <= x {
+			log.Fatalf("Big5 code %d is out of range", x)
+		}
+		mapping[x] = y
+
+		// The WHATWG spec http://encoding.spec.whatwg.org/#indexes says that
+		// "The index pointer for code point in index is the first pointer
+		// corresponding to code point in index", which would normally mean
+		// that the code below should be guarded by "if reverse[y] == 0", but
+		// last instead of first seems to match the behavior of
+		// "iconv -f UTF-8 -t BIG5". For example, U+8005 者 occurs twice in
+		// http://encoding.spec.whatwg.org/index-big5.txt, as index 2148
+		// (encoded as "\x8e\xcd") and index 6543 (encoded as "\xaa\xcc")
+		// and "echo 者 | iconv -f UTF-8 -t BIG5 | xxd" gives "\xaa\xcc".
+		c0, c1 := x/157, x%157
+		if c1 < 0x3f {
+			c1 += 0x40
+		} else {
+			c1 += 0x62
+		}
+		reverse[y] = (0x81+c0)<<8 | c1
+	}
+	if err := scanner.Err(); err != nil {
+		log.Fatalf("scanner error: %v", err)
+	}
+
+	fmt.Printf("// decode is the decoding table from Big5 code to Unicode.\n")
+	fmt.Printf("// It is defined at http://encoding.spec.whatwg.org/index-big5.txt\n")
+	fmt.Printf("var decode = [...]uint32{\n")
+	for i, v := range mapping {
+		if v != 0 {
+			fmt.Printf("\t%d: 0x%08X,\n", i, v)
+		}
+	}
+	fmt.Printf("}\n\n")
+
+	// Any run of at least separation continuous zero entries in the reverse map will
+	// be a separate encode table.
+	const separation = 1024
+
+	intervals := []interval(nil)
+	low, high := -1, -1
+	for i, v := range reverse {
+		if v == 0 {
+			continue
+		}
+		if low < 0 {
+			low = i
+		} else if i-high >= separation {
+			if high >= 0 {
+				intervals = append(intervals, interval{low, high})
+			}
+			low = i
+		}
+		high = i + 1
+	}
+	if high >= 0 {
+		intervals = append(intervals, interval{low, high})
+	}
+	sort.Sort(byDecreasingLength(intervals))
+
+	fmt.Printf("const numEncodeTables = %d\n\n", len(intervals))
+	fmt.Printf("// encodeX are the encoding tables from Unicode to Big5 code,\n")
+	fmt.Printf("// sorted by decreasing length.\n")
+	for i, v := range intervals {
+		fmt.Printf("// encode%d: %5d entries for runes in [%6d, %6d).\n", i, v.len(), v.low, v.high)
+	}
+	fmt.Printf("\n")
+
+	for i, v := range intervals {
+		fmt.Printf("const encode%dLow, encode%dHigh = %d, %d\n\n", i, i, v.low, v.high)
+		fmt.Printf("var encode%d = [...]uint16{\n", i)
+		for j := v.low; j < v.high; j++ {
+			x := reverse[j]
+			if x == 0 {
+				continue
+			}
+			fmt.Printf("\t%d-%d: 0x%04X,\n", j, v.low, x)
+		}
+		fmt.Printf("}\n\n")
+	}
+}
+
+// interval is a half-open interval [low, high).
+type interval struct {
+	low, high int
+}
+
+func (i interval) len() int { return i.high - i.low }
+
+// byDecreasingLength sorts intervals by decreasing length.
+type byDecreasingLength []interval
+
+func (b byDecreasingLength) Len() int           { return len(b) }
+func (b byDecreasingLength) Less(i, j int) bool { return b[i].len() > b[j].len() }
+func (b byDecreasingLength) Swap(i, j int)      { b[i], b[j] = b[j], b[i] }

vendor/golang.org/x/text/encoding/unicode/override.go

@@ -0,0 +1,82 @@
+// 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 unicode
+
+import (
+	"golang.org/x/text/transform"
+)
+
+// BOMOverride returns a new decoder transformer that is identical to fallback,
+// except that the presence of a Byte Order Mark at the start of the input
+// causes it to switch to the corresponding Unicode decoding. It will only
+// consider BOMs for UTF-8, UTF-16BE, and UTF-16LE.
+//
+// This differs from using ExpectBOM by allowing a BOM to switch to UTF-8, not
+// just UTF-16 variants, and allowing falling back to any encoding scheme.
+//
+// This technique is recommended by the W3C for use in HTML 5: "For
+// compatibility with deployed content, the byte order mark (also known as BOM)
+// is considered more authoritative than anything else."
+// http://www.w3.org/TR/encoding/#specification-hooks
+//
+// Using BOMOverride is mostly intended for use cases where the first characters
+// of a fallback encoding are known to not be a BOM, for example, for valid HTML
+// and most encodings.
+func BOMOverride(fallback transform.Transformer) transform.Transformer {
+	// TODO: possibly allow a variadic argument of unicode encodings to allow
+	// specifying details of which fallbacks are supported as well as
+	// specifying the details of the implementations. This would also allow for
+	// support for UTF-32, which should not be supported by default.
+	return &bomOverride{fallback: fallback}
+}
+
+type bomOverride struct {
+	fallback transform.Transformer
+	current  transform.Transformer
+}
+
+func (d *bomOverride) Reset() {
+	d.current = nil
+	d.fallback.Reset()
+}
+
+var (
+	// TODO: we could use decode functions here, instead of allocating a new
+	// decoder on every NewDecoder as IgnoreBOM decoders can be stateless.
+	utf16le = UTF16(LittleEndian, IgnoreBOM)
+	utf16be = UTF16(BigEndian, IgnoreBOM)
+)
+
+const utf8BOM = "\ufeff"
+
+func (d *bomOverride) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	if d.current != nil {
+		return d.current.Transform(dst, src, atEOF)
+	}
+	if len(src) < 3 && !atEOF {
+		return 0, 0, transform.ErrShortSrc
+	}
+	d.current = d.fallback
+	bomSize := 0
+	if len(src) >= 2 {
+		if src[0] == 0xFF && src[1] == 0xFE {
+			d.current = utf16le.NewDecoder()
+			bomSize = 2
+		} else if src[0] == 0xFE && src[1] == 0xFF {
+			d.current = utf16be.NewDecoder()
+			bomSize = 2
+		} else if len(src) >= 3 &&
+			src[0] == utf8BOM[0] &&
+			src[1] == utf8BOM[1] &&
+			src[2] == utf8BOM[2] {
+			d.current = transform.Nop
+			bomSize = 3
+		}
+	}
+	if bomSize < len(src) {
+		nDst, nSrc, err = d.current.Transform(dst, src[bomSize:], atEOF)
+	}
+	return nDst, nSrc + bomSize, err
+}

vendor/golang.org/x/text/encoding/unicode/unicode.go

@@ -0,0 +1,434 @@
+// 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 unicode provides Unicode encodings such as UTF-16.
+package unicode // import "golang.org/x/text/encoding/unicode"
+
+import (
+	"errors"
+	"unicode/utf16"
+	"unicode/utf8"
+
+	"golang.org/x/text/encoding"
+	"golang.org/x/text/encoding/internal"
+	"golang.org/x/text/encoding/internal/identifier"
+	"golang.org/x/text/internal/utf8internal"
+	"golang.org/x/text/runes"
+	"golang.org/x/text/transform"
+)
+
+// TODO: I think the Transformers really should return errors on unmatched
+// surrogate pairs and odd numbers of bytes. This is not required by RFC 2781,
+// which leaves it open, but is suggested by WhatWG. It will allow for all error
+// modes as defined by WhatWG: fatal, HTML and Replacement. This would require
+// the introduction of some kind of error type for conveying the erroneous code
+// point.
+
+// UTF8 is the UTF-8 encoding.
+var UTF8 encoding.Encoding = utf8enc
+
+var utf8enc = &internal.Encoding{
+	&internal.SimpleEncoding{utf8Decoder{}, runes.ReplaceIllFormed()},
+	"UTF-8",
+	identifier.UTF8,
+}
+
+type utf8Decoder struct{ transform.NopResetter }
+
+func (utf8Decoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	var pSrc int // point from which to start copy in src
+	var accept utf8internal.AcceptRange
+
+	// The decoder can only make the input larger, not smaller.
+	n := len(src)
+	if len(dst) < n {
+		err = transform.ErrShortDst
+		n = len(dst)
+		atEOF = false
+	}
+	for nSrc < n {
+		c := src[nSrc]
+		if c < utf8.RuneSelf {
+			nSrc++
+			continue
+		}
+		first := utf8internal.First[c]
+		size := int(first & utf8internal.SizeMask)
+		if first == utf8internal.FirstInvalid {
+			goto handleInvalid // invalid starter byte
+		}
+		accept = utf8internal.AcceptRanges[first>>utf8internal.AcceptShift]
+		if nSrc+size > n {
+			if !atEOF {
+				// We may stop earlier than necessary here if the short sequence
+				// has invalid bytes. Not checking for this simplifies the code
+				// and may avoid duplicate computations in certain conditions.
+				if err == nil {
+					err = transform.ErrShortSrc
+				}
+				break
+			}
+			// Determine the maximal subpart of an ill-formed subsequence.
+			switch {
+			case nSrc+1 >= n || src[nSrc+1] < accept.Lo || accept.Hi < src[nSrc+1]:
+				size = 1
+			case nSrc+2 >= n || src[nSrc+2] < utf8internal.LoCB || utf8internal.HiCB < src[nSrc+2]:
+				size = 2
+			default:
+				size = 3 // As we are short, the maximum is 3.
+			}
+			goto handleInvalid
+		}
+		if c = src[nSrc+1]; c < accept.Lo || accept.Hi < c {
+			size = 1
+			goto handleInvalid // invalid continuation byte
+		} else if size == 2 {
+		} else if c = src[nSrc+2]; c < utf8internal.LoCB || utf8internal.HiCB < c {
+			size = 2
+			goto handleInvalid // invalid continuation byte
+		} else if size == 3 {
+		} else if c = src[nSrc+3]; c < utf8internal.LoCB || utf8internal.HiCB < c {
+			size = 3
+			goto handleInvalid // invalid continuation byte
+		}
+		nSrc += size
+		continue
+
+	handleInvalid:
+		// Copy the scanned input so far.
+		nDst += copy(dst[nDst:], src[pSrc:nSrc])
+
+		// Append RuneError to the destination.
+		const runeError = "\ufffd"
+		if nDst+len(runeError) > len(dst) {
+			return nDst, nSrc, transform.ErrShortDst
+		}
+		nDst += copy(dst[nDst:], runeError)
+
+		// Skip the maximal subpart of an ill-formed subsequence according to
+		// the W3C standard way instead of the Go way. This Transform is
+		// probably the only place in the text repo where it is warranted.
+		nSrc += size
+		pSrc = nSrc
+
+		// Recompute the maximum source length.
+		if sz := len(dst) - nDst; sz < len(src)-nSrc {
+			err = transform.ErrShortDst
+			n = nSrc + sz
+			atEOF = false
+		}
+	}
+	return nDst + copy(dst[nDst:], src[pSrc:nSrc]), nSrc, err
+}
+
+// UTF16 returns a UTF-16 Encoding for the given default endianness and byte
+// order mark (BOM) policy.
+//
+// When decoding from UTF-16 to UTF-8, if the BOMPolicy is IgnoreBOM then
+// neither BOMs U+FEFF nor noncharacters U+FFFE in the input stream will affect
+// the endianness used for decoding, and will instead be output as their
+// standard UTF-8 encodings: "\xef\xbb\xbf" and "\xef\xbf\xbe". If the BOMPolicy
+// is UseBOM or ExpectBOM a staring BOM is not written to the UTF-8 output.
+// Instead, it overrides the default endianness e for the remainder of the
+// transformation. Any subsequent BOMs U+FEFF or noncharacters U+FFFE will not
+// affect the endianness used, and will instead be output as their standard
+// UTF-8 encodings. For UseBOM, if there is no starting BOM, it will proceed
+// with the default Endianness. For ExpectBOM, in that case, the transformation
+// will return early with an ErrMissingBOM error.
+//
+// When encoding from UTF-8 to UTF-16, a BOM will be inserted at the start of
+// the output if the BOMPolicy is UseBOM or ExpectBOM. Otherwise, a BOM will not
+// be inserted. The UTF-8 input does not need to contain a BOM.
+//
+// There is no concept of a 'native' endianness. If the UTF-16 data is produced
+// and consumed in a greater context that implies a certain endianness, use
+// IgnoreBOM. Otherwise, use ExpectBOM and always produce and consume a BOM.
+//
+// In the language of http://www.unicode.org/faq/utf_bom.html#bom10, IgnoreBOM
+// corresponds to "Where the precise type of the data stream is known... the
+// BOM should not be used" and ExpectBOM corresponds to "A particular
+// protocol... may require use of the BOM".
+func UTF16(e Endianness, b BOMPolicy) encoding.Encoding {
+	return utf16Encoding{config{e, b}, mibValue[e][b&bomMask]}
+}
+
+// mibValue maps Endianness and BOMPolicy settings to MIB constants. Note that
+// some configurations map to the same MIB identifier. RFC 2781 has requirements
+// and recommendations. Some of the "configurations" are merely recommendations,
+// so multiple configurations could match.
+var mibValue = map[Endianness][numBOMValues]identifier.MIB{
+	BigEndian: [numBOMValues]identifier.MIB{
+		IgnoreBOM: identifier.UTF16BE,
+		UseBOM:    identifier.UTF16, // BigEnding default is preferred by RFC 2781.
+		// TODO: acceptBOM | strictBOM would map to UTF16BE as well.
+	},
+	LittleEndian: [numBOMValues]identifier.MIB{
+		IgnoreBOM: identifier.UTF16LE,
+		UseBOM:    identifier.UTF16, // LittleEndian default is allowed and preferred on Windows.
+		// TODO: acceptBOM | strictBOM would map to UTF16LE as well.
+	},
+	// ExpectBOM is not widely used and has no valid MIB identifier.
+}
+
+// All lists a configuration for each IANA-defined UTF-16 variant.
+var All = []encoding.Encoding{
+	UTF8,
+	UTF16(BigEndian, UseBOM),
+	UTF16(BigEndian, IgnoreBOM),
+	UTF16(LittleEndian, IgnoreBOM),
+}
+
+// BOMPolicy is a UTF-16 encoding's byte order mark policy.
+type BOMPolicy uint8
+
+const (
+	writeBOM   BOMPolicy = 0x01
+	acceptBOM  BOMPolicy = 0x02
+	requireBOM BOMPolicy = 0x04
+	bomMask    BOMPolicy = 0x07
+
+	// HACK: numBOMValues == 8 triggers a bug in the 1.4 compiler (cannot have a
+	// map of an array of length 8 of a type that is also used as a key or value
+	// in another map). See golang.org/issue/11354.
+	// TODO: consider changing this value back to 8 if the use of 1.4.* has
+	// been minimized.
+	numBOMValues = 8 + 1
+
+	// IgnoreBOM means to ignore any byte order marks.
+	IgnoreBOM BOMPolicy = 0
+	// Common and RFC 2781-compliant interpretation for UTF-16BE/LE.
+
+	// UseBOM means that the UTF-16 form may start with a byte order mark, which
+	// will be used to override the default encoding.
+	UseBOM BOMPolicy = writeBOM | acceptBOM
+	// Common and RFC 2781-compliant interpretation for UTF-16.
+
+	// ExpectBOM means that the UTF-16 form must start with a byte order mark,
+	// which will be used to override the default encoding.
+	ExpectBOM BOMPolicy = writeBOM | acceptBOM | requireBOM
+	// Used in Java as Unicode (not to be confused with Java's UTF-16) and
+	// ICU's UTF-16,version=1. Not compliant with RFC 2781.
+
+	// TODO (maybe): strictBOM: BOM must match Endianness. This would allow:
+	// - UTF-16(B|L)E,version=1: writeBOM | acceptBOM | requireBOM | strictBOM
+	//    (UnicodeBig and UnicodeLittle in Java)
+	// - RFC 2781-compliant, but less common interpretation for UTF-16(B|L)E:
+	//    acceptBOM | strictBOM (e.g. assigned to CheckBOM).
+	// This addition would be consistent with supporting ExpectBOM.
+)
+
+// Endianness is a UTF-16 encoding's default endianness.
+type Endianness bool
+
+const (
+	// BigEndian is UTF-16BE.
+	BigEndian Endianness = false
+	// LittleEndian is UTF-16LE.
+	LittleEndian Endianness = true
+)
+
+// ErrMissingBOM means that decoding UTF-16 input with ExpectBOM did not find a
+// starting byte order mark.
+var ErrMissingBOM = errors.New("encoding: missing byte order mark")
+
+type utf16Encoding struct {
+	config
+	mib identifier.MIB
+}
+
+type config struct {
+	endianness Endianness
+	bomPolicy  BOMPolicy
+}
+
+func (u utf16Encoding) NewDecoder() *encoding.Decoder {
+	return &encoding.Decoder{Transformer: &utf16Decoder{
+		initial: u.config,
+		current: u.config,
+	}}
+}
+
+func (u utf16Encoding) NewEncoder() *encoding.Encoder {
+	return &encoding.Encoder{Transformer: &utf16Encoder{
+		endianness:       u.endianness,
+		initialBOMPolicy: u.bomPolicy,
+		currentBOMPolicy: u.bomPolicy,
+	}}
+}
+
+func (u utf16Encoding) ID() (mib identifier.MIB, other string) {
+	return u.mib, ""
+}
+
+func (u utf16Encoding) String() string {
+	e, b := "B", ""
+	if u.endianness == LittleEndian {
+		e = "L"
+	}
+	switch u.bomPolicy {
+	case ExpectBOM:
+		b = "Expect"
+	case UseBOM:
+		b = "Use"
+	case IgnoreBOM:
+		b = "Ignore"
+	}
+	return "UTF-16" + e + "E (" + b + " BOM)"
+}
+
+type utf16Decoder struct {
+	initial config
+	current config
+}
+
+func (u *utf16Decoder) Reset() {
+	u.current = u.initial
+}
+
+func (u *utf16Decoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	if len(src) == 0 {
+		if atEOF && u.current.bomPolicy&requireBOM != 0 {
+			return 0, 0, ErrMissingBOM
+		}
+		return 0, 0, nil
+	}
+	if u.current.bomPolicy&acceptBOM != 0 {
+		if len(src) < 2 {
+			return 0, 0, transform.ErrShortSrc
+		}
+		switch {
+		case src[0] == 0xfe && src[1] == 0xff:
+			u.current.endianness = BigEndian
+			nSrc = 2
+		case src[0] == 0xff && src[1] == 0xfe:
+			u.current.endianness = LittleEndian
+			nSrc = 2
+		default:
+			if u.current.bomPolicy&requireBOM != 0 {
+				return 0, 0, ErrMissingBOM
+			}
+		}
+		u.current.bomPolicy = IgnoreBOM
+	}
+
+	var r rune
+	var dSize, sSize int
+	for nSrc < len(src) {
+		if nSrc+1 < len(src) {
+			x := uint16(src[nSrc+0])<<8 | uint16(src[nSrc+1])
+			if u.current.endianness == LittleEndian {
+				x = x>>8 | x<<8
+			}
+			r, sSize = rune(x), 2
+			if utf16.IsSurrogate(r) {
+				if nSrc+3 < len(src) {
+					x = uint16(src[nSrc+2])<<8 | uint16(src[nSrc+3])
+					if u.current.endianness == LittleEndian {
+						x = x>>8 | x<<8
+					}
+					// Save for next iteration if it is not a high surrogate.
+					if isHighSurrogate(rune(x)) {
+						r, sSize = utf16.DecodeRune(r, rune(x)), 4
+					}
+				} else if !atEOF {
+					err = transform.ErrShortSrc
+					break
+				}
+			}
+			if dSize = utf8.RuneLen(r); dSize < 0 {
+				r, dSize = utf8.RuneError, 3
+			}
+		} else if atEOF {
+			// Single trailing byte.
+			r, dSize, sSize = utf8.RuneError, 3, 1
+		} else {
+			err = transform.ErrShortSrc
+			break
+		}
+		if nDst+dSize > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		nDst += utf8.EncodeRune(dst[nDst:], r)
+		nSrc += sSize
+	}
+	return nDst, nSrc, err
+}
+
+func isHighSurrogate(r rune) bool {
+	return 0xDC00 <= r && r <= 0xDFFF
+}
+
+type utf16Encoder struct {
+	endianness       Endianness
+	initialBOMPolicy BOMPolicy
+	currentBOMPolicy BOMPolicy
+}
+
+func (u *utf16Encoder) Reset() {
+	u.currentBOMPolicy = u.initialBOMPolicy
+}
+
+func (u *utf16Encoder) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	if u.currentBOMPolicy&writeBOM != 0 {
+		if len(dst) < 2 {
+			return 0, 0, transform.ErrShortDst
+		}
+		dst[0], dst[1] = 0xfe, 0xff
+		u.currentBOMPolicy = IgnoreBOM
+		nDst = 2
+	}
+
+	r, size := rune(0), 0
+	for nSrc < len(src) {
+		r = rune(src[nSrc])
+
+		// Decode a 1-byte rune.
+		if r < utf8.RuneSelf {
+			size = 1
+
+		} else {
+			// Decode a multi-byte rune.
+			r, size = utf8.DecodeRune(src[nSrc:])
+			if size == 1 {
+				// All valid runes of size 1 (those below utf8.RuneSelf) were
+				// handled above. We have invalid UTF-8 or we haven't seen the
+				// full character yet.
+				if !atEOF && !utf8.FullRune(src[nSrc:]) {
+					err = transform.ErrShortSrc
+					break
+				}
+			}
+		}
+
+		if r <= 0xffff {
+			if nDst+2 > len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			dst[nDst+0] = uint8(r >> 8)
+			dst[nDst+1] = uint8(r)
+			nDst += 2
+		} else {
+			if nDst+4 > len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			r1, r2 := utf16.EncodeRune(r)
+			dst[nDst+0] = uint8(r1 >> 8)
+			dst[nDst+1] = uint8(r1)
+			dst[nDst+2] = uint8(r2 >> 8)
+			dst[nDst+3] = uint8(r2)
+			nDst += 4
+		}
+		nSrc += size
+	}
+
+	if u.endianness == LittleEndian {
+		for i := 0; i < nDst; i += 2 {
+			dst[i], dst[i+1] = dst[i+1], dst[i]
+		}
+	}
+	return nDst, nSrc, err
+}

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/utf8internal/utf8internal.go

@@ -0,0 +1,87 @@
+// 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 utf8internal contains low-level utf8-related constants, tables, etc.
+// that are used internally by the text package.
+package utf8internal
+
+// The default lowest and highest continuation byte.
+const (
+	LoCB = 0x80 // 1000 0000
+	HiCB = 0xBF // 1011 1111
+)
+
+// Constants related to getting information of first bytes of UTF-8 sequences.
+const (
+	// ASCII identifies a UTF-8 byte as ASCII.
+	ASCII = as
+
+	// FirstInvalid indicates a byte is invalid as a first byte of a UTF-8
+	// sequence.
+	FirstInvalid = xx
+
+	// SizeMask is a mask for the size bits. Use use x&SizeMask to get the size.
+	SizeMask = 7
+
+	// AcceptShift is the right-shift count for the first byte info byte to get
+	// the index into the AcceptRanges table. See AcceptRanges.
+	AcceptShift = 4
+
+	// The names of these constants are chosen to give nice alignment in the
+	// table below. The first nibble is an index into acceptRanges or F for
+	// special one-byte cases. The second nibble is the Rune length or the
+	// Status for the special one-byte case.
+	xx = 0xF1 // invalid: size 1
+	as = 0xF0 // ASCII: size 1
+	s1 = 0x02 // accept 0, size 2
+	s2 = 0x13 // accept 1, size 3
+	s3 = 0x03 // accept 0, size 3
+	s4 = 0x23 // accept 2, size 3
+	s5 = 0x34 // accept 3, size 4
+	s6 = 0x04 // accept 0, size 4
+	s7 = 0x44 // accept 4, size 4
+)
+
+// First is information about the first byte in a UTF-8 sequence.
+var First = [256]uint8{
+	//   1   2   3   4   5   6   7   8   9   A   B   C   D   E   F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x00-0x0F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x10-0x1F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x20-0x2F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x30-0x3F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x40-0x4F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x50-0x5F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x60-0x6F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x70-0x7F
+	//   1   2   3   4   5   6   7   8   9   A   B   C   D   E   F
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0x80-0x8F
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0x90-0x9F
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xA0-0xAF
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xB0-0xBF
+	xx, xx, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, // 0xC0-0xCF
+	s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, // 0xD0-0xDF
+	s2, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s4, s3, s3, // 0xE0-0xEF
+	s5, s6, s6, s6, s7, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xF0-0xFF
+}
+
+// AcceptRange gives the range of valid values for the second byte in a UTF-8
+// sequence for any value for First that is not ASCII or FirstInvalid.
+type AcceptRange struct {
+	Lo uint8 // lowest value for second byte.
+	Hi uint8 // highest value for second byte.
+}
+
+// AcceptRanges is a slice of AcceptRange values. For a given byte sequence b
+//
+//		AcceptRanges[First[b[0]]>>AcceptShift]
+//
+// will give the value of AcceptRange for the multi-byte UTF-8 sequence starting
+// at b[0].
+var AcceptRanges = [...]AcceptRange{
+	0: {LoCB, HiCB},
+	1: {0xA0, HiCB},
+	2: {LoCB, 0x9F},
+	3: {0x90, HiCB},
+	4: {LoCB, 0x8F},
+}

vendor/golang.org/x/text/language/Makefile

@@ -0,0 +1,16 @@
+# 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.
+
+CLEANFILES+=maketables
+
+maketables: maketables.go
+	go build $^
+
+tables:	maketables
+	./maketables > tables.go
+	gofmt -w -s tables.go
+
+# Build (but do not run) maketables during testing,
+# just to make sure it still compiles.
+testshort: maketables

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

@@ -0,0 +1,16 @@
+// This file was generated by go generate; DO NOT EDIT
+
+package language
+
+// This file contains code common to the maketables.go and the package code.
+
+// langAliasType is the type of an alias in langAliasMap.
+type langAliasType int8
+
+const (
+	langDeprecated langAliasType = iota
+	langMacro
+	langLegacy
+
+	langAliasTypeUnknown langAliasType = -1
+)

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

@@ -0,0 +1,197 @@
+// 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
+
+import (
+	"fmt"
+	"sort"
+)
+
+// The Coverage interface is used to define the level of coverage of an
+// internationalization service. Note that not all types are supported by all
+// services. As lists may be generated on the fly, it is recommended that users
+// of a Coverage cache the results.
+type Coverage interface {
+	// Tags returns the list of supported tags.
+	Tags() []Tag
+
+	// BaseLanguages returns the list of supported base languages.
+	BaseLanguages() []Base
+
+	// Scripts returns the list of supported scripts.
+	Scripts() []Script
+
+	// Regions returns the list of supported regions.
+	Regions() []Region
+}
+
+var (
+	// Supported defines a Coverage that lists all supported subtags. Tags
+	// always returns nil.
+	Supported Coverage = allSubtags{}
+)
+
+// TODO:
+// - Support Variants, numbering systems.
+// - CLDR coverage levels.
+// - Set of common tags defined in this package.
+
+type allSubtags struct{}
+
+// Regions returns the list of supported regions. As all regions are in a
+// consecutive range, it simply returns a slice of numbers in increasing order.
+// The "undefined" region is not returned.
+func (s allSubtags) Regions() []Region {
+	reg := make([]Region, numRegions)
+	for i := range reg {
+		reg[i] = Region{regionID(i + 1)}
+	}
+	return reg
+}
+
+// Scripts returns the list of supported scripts. As all scripts are in a
+// consecutive range, it simply returns a slice of numbers in increasing order.
+// The "undefined" script is not returned.
+func (s allSubtags) Scripts() []Script {
+	scr := make([]Script, numScripts)
+	for i := range scr {
+		scr[i] = Script{scriptID(i + 1)}
+	}
+	return scr
+}
+
+// BaseLanguages returns the list of all supported base languages. It generates
+// the list by traversing the internal structures.
+func (s allSubtags) BaseLanguages() []Base {
+	base := make([]Base, 0, numLanguages)
+	for i := 0; i < langNoIndexOffset; i++ {
+		// We included "und" already for the value 0.
+		if i != nonCanonicalUnd {
+			base = append(base, Base{langID(i)})
+		}
+	}
+	i := langNoIndexOffset
+	for _, v := range langNoIndex {
+		for k := 0; k < 8; k++ {
+			if v&1 == 1 {
+				base = append(base, Base{langID(i)})
+			}
+			v >>= 1
+			i++
+		}
+	}
+	return base
+}
+
+// Tags always returns nil.
+func (s allSubtags) Tags() []Tag {
+	return nil
+}
+
+// coverage is used used by NewCoverage which is used as a convenient way for
+// creating Coverage implementations for partially defined data. Very often a
+// package will only need to define a subset of slices. coverage provides a
+// convenient way to do this. Moreover, packages using NewCoverage, instead of
+// their own implementation, will not break if later new slice types are added.
+type coverage struct {
+	tags    func() []Tag
+	bases   func() []Base
+	scripts func() []Script
+	regions func() []Region
+}
+
+func (s *coverage) Tags() []Tag {
+	if s.tags == nil {
+		return nil
+	}
+	return s.tags()
+}
+
+// bases implements sort.Interface and is used to sort base languages.
+type bases []Base
+
+func (b bases) Len() int {
+	return len(b)
+}
+
+func (b bases) Swap(i, j int) {
+	b[i], b[j] = b[j], b[i]
+}
+
+func (b bases) Less(i, j int) bool {
+	return b[i].langID < b[j].langID
+}
+
+// BaseLanguages returns the result from calling s.bases if it is specified or
+// otherwise derives the set of supported base languages from tags.
+func (s *coverage) BaseLanguages() []Base {
+	if s.bases == nil {
+		tags := s.Tags()
+		if len(tags) == 0 {
+			return nil
+		}
+		a := make([]Base, len(tags))
+		for i, t := range tags {
+			a[i] = Base{langID(t.lang)}
+		}
+		sort.Sort(bases(a))
+		k := 0
+		for i := 1; i < len(a); i++ {
+			if a[k] != a[i] {
+				k++
+				a[k] = a[i]
+			}
+		}
+		return a[:k+1]
+	}
+	return s.bases()
+}
+
+func (s *coverage) Scripts() []Script {
+	if s.scripts == nil {
+		return nil
+	}
+	return s.scripts()
+}
+
+func (s *coverage) Regions() []Region {
+	if s.regions == nil {
+		return nil
+	}
+	return s.regions()
+}
+
+// NewCoverage returns a Coverage for the given lists. It is typically used by
+// packages providing internationalization services to define their level of
+// coverage. A list may be of type []T or func() []T, where T is either Tag,
+// Base, Script or Region. The returned Coverage derives the value for Bases
+// from Tags if no func or slice for []Base is specified. For other unspecified
+// types the returned Coverage will return nil for the respective methods.
+func NewCoverage(list ...interface{}) Coverage {
+	s := &coverage{}
+	for _, x := range list {
+		switch v := x.(type) {
+		case func() []Base:
+			s.bases = v
+		case func() []Script:
+			s.scripts = v
+		case func() []Region:
+			s.regions = v
+		case func() []Tag:
+			s.tags = v
+		case []Base:
+			s.bases = func() []Base { return v }
+		case []Script:
+			s.scripts = func() []Script { return v }
+		case []Region:
+			s.regions = func() []Region { return v }
+		case []Tag:
+			s.tags = func() []Tag { return v }
+		default:
+			panic(fmt.Sprintf("language: unsupported set type %T", v))
+		}
+	}
+	return s
+}

vendor/golang.org/x/text/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.
+
+// langAliasType is the type of an alias in langAliasMap.
+type langAliasType int8
+
+const (
+	langDeprecated langAliasType = iota
+	langMacro
+	langLegacy
+
+	langAliasTypeUnknown langAliasType = -1
+)

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

@@ -0,0 +1,162 @@
+// 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 (
+	"bytes"
+	"flag"
+	"fmt"
+	"io/ioutil"
+	"log"
+	"reflect"
+	"sort"
+	"strings"
+
+	"golang.org/x/text/internal/gen"
+	"golang.org/x/text/language"
+	"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.")
+
+	draft = flag.String("draft",
+		"contributed",
+		`Minimal draft requirements (approved, contributed, provisional, unconfirmed).`)
+)
+
+func main() {
+	gen.Init()
+
+	// Read the CLDR zip file.
+	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 func() {
+		buf := &bytes.Buffer{}
+
+		if _, err = w.WriteGo(buf, "language"); err != nil {
+			log.Fatalf("Error formatting file index.go: %v", err)
+		}
+
+		// Since we're generating a table for our own package we need to rewrite
+		// doing the equivalent of go fmt -r 'language.b -> b'. Using
+		// bytes.Replace will do.
+		out := bytes.Replace(buf.Bytes(), []byte("language."), nil, -1)
+		if err := ioutil.WriteFile("index.go", out, 0600); err != nil {
+			log.Fatalf("Could not create file index.go: %v", err)
+		}
+	}()
+
+	m := map[language.Tag]bool{}
+	for _, lang := range 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
+		// }
+	}
+	// Include locales for plural rules, which uses a different structure.
+	for _, plurals := range data.Supplemental().Plurals {
+		for _, rules := range plurals.PluralRules {
+			for _, lang := range strings.Split(rules.Locales, " ") {
+				m[language.Make(lang)] = true
+			}
+		}
+	}
+
+	var core, special []language.Tag
+
+	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 {
+			core = append(core, t)
+		} else {
+			special = append(special, t)
+		}
+	}
+
+	w.WriteComment(`
+	NumCompactTags is the number of common tags. The maximum tag is
+	NumCompactTags-1.`)
+	w.WriteConst("NumCompactTags", len(core)+len(special))
+
+	sort.Sort(byAlpha(special))
+	w.WriteVar("specialTags", special)
+
+	// TODO: order by frequency?
+	sort.Sort(byAlpha(core))
+
+	// Size computations are just an estimate.
+	w.Size += int(reflect.TypeOf(map[uint32]uint16{}).Size())
+	w.Size += len(core) * 6 // size of uint32 and uint16
+
+	fmt.Fprintln(w)
+	fmt.Fprintln(w, "var coreTags = map[uint32]uint16{")
+	fmt.Fprintln(w, "0x0: 0, // und")
+	i := len(special) + 1 // Und and special tags already written.
+	for _, t := range core {
+		if t == language.Und {
+			continue
+		}
+		fmt.Fprint(w.Hash, t, i)
+		b, s, r := t.Raw()
+		fmt.Fprintf(w, "0x%s%s%s: %d, // %s\n",
+			getIndex(b, 3), // 3 is enough as it is guaranteed to be a compact number
+			getIndex(s, 2),
+			getIndex(r, 3),
+			i, t)
+		i++
+	}
+	fmt.Fprintln(w, "}")
+}
+
+// getIndex prints the subtag type and extracts its index of size nibble.
+// If the index is less than n nibbles, the result is prefixed with 0s.
+func getIndex(x interface{}, n int) string {
+	s := fmt.Sprintf("%#v", x) // s is of form Type{typeID: 0x00}
+	s = s[strings.Index(s, "0x")+2 : len(s)-1]
+	return strings.Repeat("0", n-len(s)) + s
+}
+
+type byAlpha []language.Tag
+
+func (a byAlpha) Len() int           { return len(a) }
+func (a byAlpha) Swap(i, j int)      { a[i], a[j] = a[j], a[i] }
+func (a byAlpha) Less(i, j int) bool { return a[i].String() < a[j].String() }

vendor/golang.org/x/text/language/go1_1.go

@@ -0,0 +1,38 @@
+// 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 !go1.2
+
+package language
+
+import "sort"
+
+func sortStable(s sort.Interface) {
+	ss := stableSort{
+		s:   s,
+		pos: make([]int, s.Len()),
+	}
+	for i := range ss.pos {
+		ss.pos[i] = i
+	}
+	sort.Sort(&ss)
+}
+
+type stableSort struct {
+	s   sort.Interface
+	pos []int
+}
+
+func (s *stableSort) Len() int {
+	return len(s.pos)
+}
+
+func (s *stableSort) Less(i, j int) bool {
+	return s.s.Less(i, j) || !s.s.Less(j, i) && s.pos[i] < s.pos[j]
+}
+
+func (s *stableSort) Swap(i, j int) {
+	s.s.Swap(i, j)
+	s.pos[i], s.pos[j] = s.pos[j], s.pos[i]
+}

vendor/golang.org/x/text/language/go1_2.go

@@ -0,0 +1,11 @@
+// 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 go1.2
+
+package language
+
+import "sort"
+
+var sortStable = sort.Stable

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

@@ -0,0 +1,767 @@
+// This file was generated by go generate; DO NOT EDIT
+
+package language
+
+// NumCompactTags is the number of common tags. The maximum tag is
+// NumCompactTags-1.
+const NumCompactTags = 752
+
+var specialTags = []Tag{ // 2 elements
+	0: {lang: 0xd5, region: 0x6d, script: 0x0, pVariant: 0x5, pExt: 0xe, str: "ca-ES-valencia"},
+	1: {lang: 0x134, region: 0x134, script: 0x0, pVariant: 0x5, pExt: 0x5, str: "en-US-u-va-posix"},
+} // Size: 72 bytes
+
+var coreTags = map[uint32]uint16{
+	0x0:        0,   // und
+	0x01500000: 3,   // af
+	0x015000d1: 4,   // af-NA
+	0x01500160: 5,   // af-ZA
+	0x01b00000: 6,   // agq
+	0x01b00051: 7,   // agq-CM
+	0x02000000: 8,   // ak
+	0x0200007f: 9,   // ak-GH
+	0x02600000: 10,  // am
+	0x0260006e: 11,  // am-ET
+	0x03900000: 12,  // ar
+	0x03900001: 13,  // ar-001
+	0x03900022: 14,  // ar-AE
+	0x03900038: 15,  // ar-BH
+	0x03900061: 16,  // ar-DJ
+	0x03900066: 17,  // ar-DZ
+	0x0390006a: 18,  // ar-EG
+	0x0390006b: 19,  // ar-EH
+	0x0390006c: 20,  // ar-ER
+	0x03900096: 21,  // ar-IL
+	0x0390009a: 22,  // ar-IQ
+	0x039000a0: 23,  // ar-JO
+	0x039000a7: 24,  // ar-KM
+	0x039000ab: 25,  // ar-KW
+	0x039000af: 26,  // ar-LB
+	0x039000b8: 27,  // ar-LY
+	0x039000b9: 28,  // ar-MA
+	0x039000c8: 29,  // ar-MR
+	0x039000e0: 30,  // ar-OM
+	0x039000ec: 31,  // ar-PS
+	0x039000f2: 32,  // ar-QA
+	0x03900107: 33,  // ar-SA
+	0x0390010a: 34,  // ar-SD
+	0x03900114: 35,  // ar-SO
+	0x03900116: 36,  // ar-SS
+	0x0390011b: 37,  // ar-SY
+	0x0390011f: 38,  // ar-TD
+	0x03900127: 39,  // ar-TN
+	0x0390015d: 40,  // ar-YE
+	0x03f00000: 41,  // ars
+	0x04200000: 42,  // as
+	0x04200098: 43,  // as-IN
+	0x04300000: 44,  // asa
+	0x0430012e: 45,  // asa-TZ
+	0x04700000: 46,  // ast
+	0x0470006d: 47,  // ast-ES
+	0x05700000: 48,  // az
+	0x0571e000: 49,  // az-Cyrl
+	0x0571e031: 50,  // az-Cyrl-AZ
+	0x05752000: 51,  // az-Latn
+	0x05752031: 52,  // az-Latn-AZ
+	0x05d00000: 53,  // bas
+	0x05d00051: 54,  // bas-CM
+	0x07000000: 55,  // be
+	0x07000046: 56,  // be-BY
+	0x07400000: 57,  // bem
+	0x07400161: 58,  // bem-ZM
+	0x07800000: 59,  // bez
+	0x0780012e: 60,  // bez-TZ
+	0x07d00000: 61,  // bg
+	0x07d00037: 62,  // bg-BG
+	0x08100000: 63,  // bh
+	0x09e00000: 64,  // bm
+	0x09e000c2: 65,  // bm-ML
+	0x0a300000: 66,  // bn
+	0x0a300034: 67,  // bn-BD
+	0x0a300098: 68,  // bn-IN
+	0x0a700000: 69,  // bo
+	0x0a700052: 70,  // bo-CN
+	0x0a700098: 71,  // bo-IN
+	0x0b000000: 72,  // br
+	0x0b000077: 73,  // br-FR
+	0x0b300000: 74,  // brx
+	0x0b300098: 75,  // brx-IN
+	0x0b500000: 76,  // bs
+	0x0b51e000: 77,  // bs-Cyrl
+	0x0b51e032: 78,  // bs-Cyrl-BA
+	0x0b552000: 79,  // bs-Latn
+	0x0b552032: 80,  // bs-Latn-BA
+	0x0d500000: 81,  // ca
+	0x0d500021: 82,  // ca-AD
+	0x0d50006d: 83,  // ca-ES
+	0x0d500077: 84,  // ca-FR
+	0x0d50009d: 85,  // ca-IT
+	0x0da00000: 86,  // ce
+	0x0da00105: 87,  // ce-RU
+	0x0dd00000: 88,  // cgg
+	0x0dd00130: 89,  // cgg-UG
+	0x0e300000: 90,  // chr
+	0x0e300134: 91,  // chr-US
+	0x0e700000: 92,  // ckb
+	0x0e70009a: 93,  // ckb-IQ
+	0x0e70009b: 94,  // ckb-IR
+	0x0f600000: 95,  // cs
+	0x0f60005d: 96,  // cs-CZ
+	0x0fa00000: 97,  // cu
+	0x0fa00105: 98,  // cu-RU
+	0x0fc00000: 99,  // cy
+	0x0fc0007a: 100, // cy-GB
+	0x0fd00000: 101, // da
+	0x0fd00062: 102, // da-DK
+	0x0fd00081: 103, // da-GL
+	0x10400000: 104, // dav
+	0x104000a3: 105, // dav-KE
+	0x10900000: 106, // de
+	0x1090002d: 107, // de-AT
+	0x10900035: 108, // de-BE
+	0x1090004d: 109, // de-CH
+	0x1090005f: 110, // de-DE
+	0x1090009d: 111, // de-IT
+	0x109000b1: 112, // de-LI
+	0x109000b6: 113, // de-LU
+	0x11300000: 114, // dje
+	0x113000d3: 115, // dje-NE
+	0x11b00000: 116, // dsb
+	0x11b0005f: 117, // dsb-DE
+	0x12000000: 118, // dua
+	0x12000051: 119, // dua-CM
+	0x12400000: 120, // dv
+	0x12700000: 121, // dyo
+	0x12700113: 122, // dyo-SN
+	0x12900000: 123, // dz
+	0x12900042: 124, // dz-BT
+	0x12b00000: 125, // ebu
+	0x12b000a3: 126, // ebu-KE
+	0x12c00000: 127, // ee
+	0x12c0007f: 128, // ee-GH
+	0x12c00121: 129, // ee-TG
+	0x13100000: 130, // el
+	0x1310005c: 131, // el-CY
+	0x13100086: 132, // el-GR
+	0x13400000: 133, // en
+	0x13400001: 134, // en-001
+	0x1340001a: 135, // en-150
+	0x13400024: 136, // en-AG
+	0x13400025: 137, // en-AI
+	0x1340002c: 138, // en-AS
+	0x1340002d: 139, // en-AT
+	0x1340002e: 140, // en-AU
+	0x13400033: 141, // en-BB
+	0x13400035: 142, // en-BE
+	0x13400039: 143, // en-BI
+	0x1340003c: 144, // en-BM
+	0x13400041: 145, // en-BS
+	0x13400045: 146, // en-BW
+	0x13400047: 147, // en-BZ
+	0x13400048: 148, // en-CA
+	0x13400049: 149, // en-CC
+	0x1340004d: 150, // en-CH
+	0x1340004f: 151, // en-CK
+	0x13400051: 152, // en-CM
+	0x1340005b: 153, // en-CX
+	0x1340005c: 154, // en-CY
+	0x1340005f: 155, // en-DE
+	0x13400060: 156, // en-DG
+	0x13400062: 157, // en-DK
+	0x13400063: 158, // en-DM
+	0x1340006c: 159, // en-ER
+	0x13400071: 160, // en-FI
+	0x13400072: 161, // en-FJ
+	0x13400073: 162, // en-FK
+	0x13400074: 163, // en-FM
+	0x1340007a: 164, // en-GB
+	0x1340007b: 165, // en-GD
+	0x1340007e: 166, // en-GG
+	0x1340007f: 167, // en-GH
+	0x13400080: 168, // en-GI
+	0x13400082: 169, // en-GM
+	0x13400089: 170, // en-GU
+	0x1340008b: 171, // en-GY
+	0x1340008c: 172, // en-HK
+	0x13400095: 173, // en-IE
+	0x13400096: 174, // en-IL
+	0x13400097: 175, // en-IM
+	0x13400098: 176, // en-IN
+	0x13400099: 177, // en-IO
+	0x1340009e: 178, // en-JE
+	0x1340009f: 179, // en-JM
+	0x134000a3: 180, // en-KE
+	0x134000a6: 181, // en-KI
+	0x134000a8: 182, // en-KN
+	0x134000ac: 183, // en-KY
+	0x134000b0: 184, // en-LC
+	0x134000b3: 185, // en-LR
+	0x134000b4: 186, // en-LS
+	0x134000be: 187, // en-MG
+	0x134000bf: 188, // en-MH
+	0x134000c5: 189, // en-MO
+	0x134000c6: 190, // en-MP
+	0x134000c9: 191, // en-MS
+	0x134000ca: 192, // en-MT
+	0x134000cb: 193, // en-MU
+	0x134000cd: 194, // en-MW
+	0x134000cf: 195, // en-MY
+	0x134000d1: 196, // en-NA
+	0x134000d4: 197, // en-NF
+	0x134000d5: 198, // en-NG
+	0x134000d8: 199, // en-NL
+	0x134000dc: 200, // en-NR
+	0x134000de: 201, // en-NU
+	0x134000df: 202, // en-NZ
+	0x134000e5: 203, // en-PG
+	0x134000e6: 204, // en-PH
+	0x134000e7: 205, // en-PK
+	0x134000ea: 206, // en-PN
+	0x134000eb: 207, // en-PR
+	0x134000ef: 208, // en-PW
+	0x13400106: 209, // en-RW
+	0x13400108: 210, // en-SB
+	0x13400109: 211, // en-SC
+	0x1340010a: 212, // en-SD
+	0x1340010b: 213, // en-SE
+	0x1340010c: 214, // en-SG
+	0x1340010d: 215, // en-SH
+	0x1340010e: 216, // en-SI
+	0x13400111: 217, // en-SL
+	0x13400116: 218, // en-SS
+	0x1340011a: 219, // en-SX
+	0x1340011c: 220, // en-SZ
+	0x1340011e: 221, // en-TC
+	0x13400124: 222, // en-TK
+	0x13400128: 223, // en-TO
+	0x1340012b: 224, // en-TT
+	0x1340012c: 225, // en-TV
+	0x1340012e: 226, // en-TZ
+	0x13400130: 227, // en-UG
+	0x13400132: 228, // en-UM
+	0x13400134: 229, // en-US
+	0x13400138: 230, // en-VC
+	0x1340013b: 231, // en-VG
+	0x1340013c: 232, // en-VI
+	0x1340013e: 233, // en-VU
+	0x13400141: 234, // en-WS
+	0x13400160: 235, // en-ZA
+	0x13400161: 236, // en-ZM
+	0x13400163: 237, // en-ZW
+	0x13700000: 238, // eo
+	0x13700001: 239, // eo-001
+	0x13900000: 240, // es
+	0x1390001e: 241, // es-419
+	0x1390002b: 242, // es-AR
+	0x1390003e: 243, // es-BO
+	0x13900040: 244, // es-BR
+	0x13900050: 245, // es-CL
+	0x13900053: 246, // es-CO
+	0x13900055: 247, // es-CR
+	0x13900058: 248, // es-CU
+	0x13900064: 249, // es-DO
+	0x13900067: 250, // es-EA
+	0x13900068: 251, // es-EC
+	0x1390006d: 252, // es-ES
+	0x13900085: 253, // es-GQ
+	0x13900088: 254, // es-GT
+	0x1390008e: 255, // es-HN
+	0x13900093: 256, // es-IC
+	0x139000ce: 257, // es-MX
+	0x139000d7: 258, // es-NI
+	0x139000e1: 259, // es-PA
+	0x139000e3: 260, // es-PE
+	0x139000e6: 261, // es-PH
+	0x139000eb: 262, // es-PR
+	0x139000f0: 263, // es-PY
+	0x13900119: 264, // es-SV
+	0x13900134: 265, // es-US
+	0x13900135: 266, // es-UY
+	0x1390013a: 267, // es-VE
+	0x13b00000: 268, // et
+	0x13b00069: 269, // et-EE
+	0x14000000: 270, // eu
+	0x1400006d: 271, // eu-ES
+	0x14100000: 272, // ewo
+	0x14100051: 273, // ewo-CM
+	0x14300000: 274, // fa
+	0x14300023: 275, // fa-AF
+	0x1430009b: 276, // fa-IR
+	0x14900000: 277, // ff
+	0x14900051: 278, // ff-CM
+	0x14900083: 279, // ff-GN
+	0x149000c8: 280, // ff-MR
+	0x14900113: 281, // ff-SN
+	0x14c00000: 282, // fi
+	0x14c00071: 283, // fi-FI
+	0x14e00000: 284, // fil
+	0x14e000e6: 285, // fil-PH
+	0x15300000: 286, // fo
+	0x15300062: 287, // fo-DK
+	0x15300075: 288, // fo-FO
+	0x15900000: 289, // fr
+	0x15900035: 290, // fr-BE
+	0x15900036: 291, // fr-BF
+	0x15900039: 292, // fr-BI
+	0x1590003a: 293, // fr-BJ
+	0x1590003b: 294, // fr-BL
+	0x15900048: 295, // fr-CA
+	0x1590004a: 296, // fr-CD
+	0x1590004b: 297, // fr-CF
+	0x1590004c: 298, // fr-CG
+	0x1590004d: 299, // fr-CH
+	0x1590004e: 300, // fr-CI
+	0x15900051: 301, // fr-CM
+	0x15900061: 302, // fr-DJ
+	0x15900066: 303, // fr-DZ
+	0x15900077: 304, // fr-FR
+	0x15900079: 305, // fr-GA
+	0x1590007d: 306, // fr-GF
+	0x15900083: 307, // fr-GN
+	0x15900084: 308, // fr-GP
+	0x15900085: 309, // fr-GQ
+	0x15900090: 310, // fr-HT
+	0x159000a7: 311, // fr-KM
+	0x159000b6: 312, // fr-LU
+	0x159000b9: 313, // fr-MA
+	0x159000ba: 314, // fr-MC
+	0x159000bd: 315, // fr-MF
+	0x159000be: 316, // fr-MG
+	0x159000c2: 317, // fr-ML
+	0x159000c7: 318, // fr-MQ
+	0x159000c8: 319, // fr-MR
+	0x159000cb: 320, // fr-MU
+	0x159000d2: 321, // fr-NC
+	0x159000d3: 322, // fr-NE
+	0x159000e4: 323, // fr-PF
+	0x159000e9: 324, // fr-PM
+	0x15900101: 325, // fr-RE
+	0x15900106: 326, // fr-RW
+	0x15900109: 327, // fr-SC
+	0x15900113: 328, // fr-SN
+	0x1590011b: 329, // fr-SY
+	0x1590011f: 330, // fr-TD
+	0x15900121: 331, // fr-TG
+	0x15900127: 332, // fr-TN
+	0x1590013e: 333, // fr-VU
+	0x1590013f: 334, // fr-WF
+	0x1590015e: 335, // fr-YT
+	0x16400000: 336, // fur
+	0x1640009d: 337, // fur-IT
+	0x16800000: 338, // fy
+	0x168000d8: 339, // fy-NL
+	0x16900000: 340, // ga
+	0x16900095: 341, // ga-IE
+	0x17800000: 342, // gd
+	0x1780007a: 343, // gd-GB
+	0x18a00000: 344, // gl
+	0x18a0006d: 345, // gl-ES
+	0x19c00000: 346, // gsw
+	0x19c0004d: 347, // gsw-CH
+	0x19c00077: 348, // gsw-FR
+	0x19c000b1: 349, // gsw-LI
+	0x19d00000: 350, // gu
+	0x19d00098: 351, // gu-IN
+	0x1a200000: 352, // guw
+	0x1a400000: 353, // guz
+	0x1a4000a3: 354, // guz-KE
+	0x1a500000: 355, // gv
+	0x1a500097: 356, // gv-IM
+	0x1ad00000: 357, // ha
+	0x1ad0007f: 358, // ha-GH
+	0x1ad000d3: 359, // ha-NE
+	0x1ad000d5: 360, // ha-NG
+	0x1b100000: 361, // haw
+	0x1b100134: 362, // haw-US
+	0x1b500000: 363, // he
+	0x1b500096: 364, // he-IL
+	0x1b700000: 365, // hi
+	0x1b700098: 366, // hi-IN
+	0x1ca00000: 367, // hr
+	0x1ca00032: 368, // hr-BA
+	0x1ca0008f: 369, // hr-HR
+	0x1cb00000: 370, // hsb
+	0x1cb0005f: 371, // hsb-DE
+	0x1ce00000: 372, // hu
+	0x1ce00091: 373, // hu-HU
+	0x1d000000: 374, // hy
+	0x1d000027: 375, // hy-AM
+	0x1da00000: 376, // id
+	0x1da00094: 377, // id-ID
+	0x1df00000: 378, // ig
+	0x1df000d5: 379, // ig-NG
+	0x1e200000: 380, // ii
+	0x1e200052: 381, // ii-CN
+	0x1f000000: 382, // is
+	0x1f00009c: 383, // is-IS
+	0x1f100000: 384, // it
+	0x1f10004d: 385, // it-CH
+	0x1f10009d: 386, // it-IT
+	0x1f100112: 387, // it-SM
+	0x1f200000: 388, // iu
+	0x1f800000: 389, // ja
+	0x1f8000a1: 390, // ja-JP
+	0x1fb00000: 391, // jbo
+	0x1ff00000: 392, // jgo
+	0x1ff00051: 393, // jgo-CM
+	0x20200000: 394, // jmc
+	0x2020012e: 395, // jmc-TZ
+	0x20600000: 396, // jv
+	0x20800000: 397, // ka
+	0x2080007c: 398, // ka-GE
+	0x20a00000: 399, // kab
+	0x20a00066: 400, // kab-DZ
+	0x20e00000: 401, // kaj
+	0x20f00000: 402, // kam
+	0x20f000a3: 403, // kam-KE
+	0x21700000: 404, // kcg
+	0x21b00000: 405, // kde
+	0x21b0012e: 406, // kde-TZ
+	0x21f00000: 407, // kea
+	0x21f00059: 408, // kea-CV
+	0x22c00000: 409, // khq
+	0x22c000c2: 410, // khq-ML
+	0x23100000: 411, // ki
+	0x231000a3: 412, // ki-KE
+	0x23a00000: 413, // kk
+	0x23a000ad: 414, // kk-KZ
+	0x23c00000: 415, // kkj
+	0x23c00051: 416, // kkj-CM
+	0x23d00000: 417, // kl
+	0x23d00081: 418, // kl-GL
+	0x23e00000: 419, // kln
+	0x23e000a3: 420, // kln-KE
+	0x24200000: 421, // km
+	0x242000a5: 422, // km-KH
+	0x24900000: 423, // kn
+	0x24900098: 424, // kn-IN
+	0x24b00000: 425, // ko
+	0x24b000a9: 426, // ko-KP
+	0x24b000aa: 427, // ko-KR
+	0x24d00000: 428, // kok
+	0x24d00098: 429, // kok-IN
+	0x26100000: 430, // ks
+	0x26100098: 431, // ks-IN
+	0x26200000: 432, // ksb
+	0x2620012e: 433, // ksb-TZ
+	0x26400000: 434, // ksf
+	0x26400051: 435, // ksf-CM
+	0x26500000: 436, // ksh
+	0x2650005f: 437, // ksh-DE
+	0x26b00000: 438, // ku
+	0x27800000: 439, // kw
+	0x2780007a: 440, // kw-GB
+	0x28100000: 441, // ky
+	0x281000a4: 442, // ky-KG
+	0x28800000: 443, // lag
+	0x2880012e: 444, // lag-TZ
+	0x28c00000: 445, // lb
+	0x28c000b6: 446, // lb-LU
+	0x29a00000: 447, // lg
+	0x29a00130: 448, // lg-UG
+	0x2a600000: 449, // lkt
+	0x2a600134: 450, // lkt-US
+	0x2ac00000: 451, // ln
+	0x2ac00029: 452, // ln-AO
+	0x2ac0004a: 453, // ln-CD
+	0x2ac0004b: 454, // ln-CF
+	0x2ac0004c: 455, // ln-CG
+	0x2af00000: 456, // lo
+	0x2af000ae: 457, // lo-LA
+	0x2b600000: 458, // lrc
+	0x2b60009a: 459, // lrc-IQ
+	0x2b60009b: 460, // lrc-IR
+	0x2b700000: 461, // lt
+	0x2b7000b5: 462, // lt-LT
+	0x2b900000: 463, // lu
+	0x2b90004a: 464, // lu-CD
+	0x2bb00000: 465, // luo
+	0x2bb000a3: 466, // luo-KE
+	0x2bc00000: 467, // luy
+	0x2bc000a3: 468, // luy-KE
+	0x2be00000: 469, // lv
+	0x2be000b7: 470, // lv-LV
+	0x2c800000: 471, // mas
+	0x2c8000a3: 472, // mas-KE
+	0x2c80012e: 473, // mas-TZ
+	0x2e000000: 474, // mer
+	0x2e0000a3: 475, // mer-KE
+	0x2e400000: 476, // mfe
+	0x2e4000cb: 477, // mfe-MU
+	0x2e800000: 478, // mg
+	0x2e8000be: 479, // mg-MG
+	0x2e900000: 480, // mgh
+	0x2e9000d0: 481, // mgh-MZ
+	0x2eb00000: 482, // mgo
+	0x2eb00051: 483, // mgo-CM
+	0x2f600000: 484, // mk
+	0x2f6000c1: 485, // mk-MK
+	0x2fb00000: 486, // ml
+	0x2fb00098: 487, // ml-IN
+	0x30200000: 488, // mn
+	0x302000c4: 489, // mn-MN
+	0x31200000: 490, // mr
+	0x31200098: 491, // mr-IN
+	0x31600000: 492, // ms
+	0x3160003d: 493, // ms-BN
+	0x316000cf: 494, // ms-MY
+	0x3160010c: 495, // ms-SG
+	0x31700000: 496, // mt
+	0x317000ca: 497, // mt-MT
+	0x31c00000: 498, // mua
+	0x31c00051: 499, // mua-CM
+	0x32800000: 500, // my
+	0x328000c3: 501, // my-MM
+	0x33100000: 502, // mzn
+	0x3310009b: 503, // mzn-IR
+	0x33800000: 504, // nah
+	0x33c00000: 505, // naq
+	0x33c000d1: 506, // naq-NA
+	0x33e00000: 507, // nb
+	0x33e000d9: 508, // nb-NO
+	0x33e0010f: 509, // nb-SJ
+	0x34500000: 510, // nd
+	0x34500163: 511, // nd-ZW
+	0x34700000: 512, // nds
+	0x3470005f: 513, // nds-DE
+	0x347000d8: 514, // nds-NL
+	0x34800000: 515, // ne
+	0x34800098: 516, // ne-IN
+	0x348000da: 517, // ne-NP
+	0x35e00000: 518, // nl
+	0x35e0002f: 519, // nl-AW
+	0x35e00035: 520, // nl-BE
+	0x35e0003f: 521, // nl-BQ
+	0x35e0005a: 522, // nl-CW
+	0x35e000d8: 523, // nl-NL
+	0x35e00115: 524, // nl-SR
+	0x35e0011a: 525, // nl-SX
+	0x35f00000: 526, // nmg
+	0x35f00051: 527, // nmg-CM
+	0x36100000: 528, // nn
+	0x361000d9: 529, // nn-NO
+	0x36300000: 530, // nnh
+	0x36300051: 531, // nnh-CM
+	0x36600000: 532, // no
+	0x36c00000: 533, // nqo
+	0x36d00000: 534, // nr
+	0x37100000: 535, // nso
+	0x37700000: 536, // nus
+	0x37700116: 537, // nus-SS
+	0x37e00000: 538, // ny
+	0x38000000: 539, // nyn
+	0x38000130: 540, // nyn-UG
+	0x38700000: 541, // om
+	0x3870006e: 542, // om-ET
+	0x387000a3: 543, // om-KE
+	0x38c00000: 544, // or
+	0x38c00098: 545, // or-IN
+	0x38f00000: 546, // os
+	0x38f0007c: 547, // os-GE
+	0x38f00105: 548, // os-RU
+	0x39400000: 549, // pa
+	0x39405000: 550, // pa-Arab
+	0x394050e7: 551, // pa-Arab-PK
+	0x3942f000: 552, // pa-Guru
+	0x3942f098: 553, // pa-Guru-IN
+	0x39800000: 554, // pap
+	0x3aa00000: 555, // pl
+	0x3aa000e8: 556, // pl-PL
+	0x3b400000: 557, // prg
+	0x3b400001: 558, // prg-001
+	0x3b500000: 559, // ps
+	0x3b500023: 560, // ps-AF
+	0x3b700000: 561, // pt
+	0x3b700029: 562, // pt-AO
+	0x3b700040: 563, // pt-BR
+	0x3b70004d: 564, // pt-CH
+	0x3b700059: 565, // pt-CV
+	0x3b700085: 566, // pt-GQ
+	0x3b70008a: 567, // pt-GW
+	0x3b7000b6: 568, // pt-LU
+	0x3b7000c5: 569, // pt-MO
+	0x3b7000d0: 570, // pt-MZ
+	0x3b7000ed: 571, // pt-PT
+	0x3b700117: 572, // pt-ST
+	0x3b700125: 573, // pt-TL
+	0x3bb00000: 574, // qu
+	0x3bb0003e: 575, // qu-BO
+	0x3bb00068: 576, // qu-EC
+	0x3bb000e3: 577, // qu-PE
+	0x3cb00000: 578, // rm
+	0x3cb0004d: 579, // rm-CH
+	0x3d000000: 580, // rn
+	0x3d000039: 581, // rn-BI
+	0x3d300000: 582, // ro
+	0x3d3000bb: 583, // ro-MD
+	0x3d300103: 584, // ro-RO
+	0x3d500000: 585, // rof
+	0x3d50012e: 586, // rof-TZ
+	0x3d900000: 587, // ru
+	0x3d900046: 588, // ru-BY
+	0x3d9000a4: 589, // ru-KG
+	0x3d9000ad: 590, // ru-KZ
+	0x3d9000bb: 591, // ru-MD
+	0x3d900105: 592, // ru-RU
+	0x3d90012f: 593, // ru-UA
+	0x3dc00000: 594, // rw
+	0x3dc00106: 595, // rw-RW
+	0x3dd00000: 596, // rwk
+	0x3dd0012e: 597, // rwk-TZ
+	0x3e200000: 598, // sah
+	0x3e200105: 599, // sah-RU
+	0x3e300000: 600, // saq
+	0x3e3000a3: 601, // saq-KE
+	0x3e900000: 602, // sbp
+	0x3e90012e: 603, // sbp-TZ
+	0x3f200000: 604, // sdh
+	0x3f300000: 605, // se
+	0x3f300071: 606, // se-FI
+	0x3f3000d9: 607, // se-NO
+	0x3f30010b: 608, // se-SE
+	0x3f500000: 609, // seh
+	0x3f5000d0: 610, // seh-MZ
+	0x3f700000: 611, // ses
+	0x3f7000c2: 612, // ses-ML
+	0x3f800000: 613, // sg
+	0x3f80004b: 614, // sg-CF
+	0x3fe00000: 615, // shi
+	0x3fe52000: 616, // shi-Latn
+	0x3fe520b9: 617, // shi-Latn-MA
+	0x3fed2000: 618, // shi-Tfng
+	0x3fed20b9: 619, // shi-Tfng-MA
+	0x40200000: 620, // si
+	0x402000b2: 621, // si-LK
+	0x40800000: 622, // sk
+	0x40800110: 623, // sk-SK
+	0x40c00000: 624, // sl
+	0x40c0010e: 625, // sl-SI
+	0x41200000: 626, // sma
+	0x41300000: 627, // smi
+	0x41400000: 628, // smj
+	0x41500000: 629, // smn
+	0x41500071: 630, // smn-FI
+	0x41800000: 631, // sms
+	0x41900000: 632, // sn
+	0x41900163: 633, // sn-ZW
+	0x41f00000: 634, // so
+	0x41f00061: 635, // so-DJ
+	0x41f0006e: 636, // so-ET
+	0x41f000a3: 637, // so-KE
+	0x41f00114: 638, // so-SO
+	0x42700000: 639, // sq
+	0x42700026: 640, // sq-AL
+	0x427000c1: 641, // sq-MK
+	0x4270014c: 642, // sq-XK
+	0x42800000: 643, // sr
+	0x4281e000: 644, // sr-Cyrl
+	0x4281e032: 645, // sr-Cyrl-BA
+	0x4281e0bc: 646, // sr-Cyrl-ME
+	0x4281e104: 647, // sr-Cyrl-RS
+	0x4281e14c: 648, // sr-Cyrl-XK
+	0x42852000: 649, // sr-Latn
+	0x42852032: 650, // sr-Latn-BA
+	0x428520bc: 651, // sr-Latn-ME
+	0x42852104: 652, // sr-Latn-RS
+	0x4285214c: 653, // sr-Latn-XK
+	0x42d00000: 654, // ss
+	0x43000000: 655, // ssy
+	0x43100000: 656, // st
+	0x43a00000: 657, // sv
+	0x43a00030: 658, // sv-AX
+	0x43a00071: 659, // sv-FI
+	0x43a0010b: 660, // sv-SE
+	0x43b00000: 661, // sw
+	0x43b0004a: 662, // sw-CD
+	0x43b000a3: 663, // sw-KE
+	0x43b0012e: 664, // sw-TZ
+	0x43b00130: 665, // sw-UG
+	0x44400000: 666, // syr
+	0x44600000: 667, // ta
+	0x44600098: 668, // ta-IN
+	0x446000b2: 669, // ta-LK
+	0x446000cf: 670, // ta-MY
+	0x4460010c: 671, // ta-SG
+	0x45700000: 672, // te
+	0x45700098: 673, // te-IN
+	0x45a00000: 674, // teo
+	0x45a000a3: 675, // teo-KE
+	0x45a00130: 676, // teo-UG
+	0x46100000: 677, // th
+	0x46100122: 678, // th-TH
+	0x46500000: 679, // ti
+	0x4650006c: 680, // ti-ER
+	0x4650006e: 681, // ti-ET
+	0x46700000: 682, // tig
+	0x46c00000: 683, // tk
+	0x46c00126: 684, // tk-TM
+	0x47600000: 685, // tn
+	0x47800000: 686, // to
+	0x47800128: 687, // to-TO
+	0x48000000: 688, // tr
+	0x4800005c: 689, // tr-CY
+	0x4800012a: 690, // tr-TR
+	0x48400000: 691, // ts
+	0x49a00000: 692, // twq
+	0x49a000d3: 693, // twq-NE
+	0x49f00000: 694, // tzm
+	0x49f000b9: 695, // tzm-MA
+	0x4a200000: 696, // ug
+	0x4a200052: 697, // ug-CN
+	0x4a400000: 698, // uk
+	0x4a40012f: 699, // uk-UA
+	0x4aa00000: 700, // ur
+	0x4aa00098: 701, // ur-IN
+	0x4aa000e7: 702, // ur-PK
+	0x4b200000: 703, // uz
+	0x4b205000: 704, // uz-Arab
+	0x4b205023: 705, // uz-Arab-AF
+	0x4b21e000: 706, // uz-Cyrl
+	0x4b21e136: 707, // uz-Cyrl-UZ
+	0x4b252000: 708, // uz-Latn
+	0x4b252136: 709, // uz-Latn-UZ
+	0x4b400000: 710, // vai
+	0x4b452000: 711, // vai-Latn
+	0x4b4520b3: 712, // vai-Latn-LR
+	0x4b4d9000: 713, // vai-Vaii
+	0x4b4d90b3: 714, // vai-Vaii-LR
+	0x4b600000: 715, // ve
+	0x4b900000: 716, // vi
+	0x4b90013d: 717, // vi-VN
+	0x4bf00000: 718, // vo
+	0x4bf00001: 719, // vo-001
+	0x4c200000: 720, // vun
+	0x4c20012e: 721, // vun-TZ
+	0x4c400000: 722, // wa
+	0x4c500000: 723, // wae
+	0x4c50004d: 724, // wae-CH
+	0x4db00000: 725, // wo
+	0x4e800000: 726, // xh
+	0x4f100000: 727, // xog
+	0x4f100130: 728, // xog-UG
+	0x4ff00000: 729, // yav
+	0x4ff00051: 730, // yav-CM
+	0x50800000: 731, // yi
+	0x50800001: 732, // yi-001
+	0x50e00000: 733, // yo
+	0x50e0003a: 734, // yo-BJ
+	0x50e000d5: 735, // yo-NG
+	0x51500000: 736, // yue
+	0x5150008c: 737, // yue-HK
+	0x51e00000: 738, // zgh
+	0x51e000b9: 739, // zgh-MA
+	0x51f00000: 740, // zh
+	0x51f34000: 741, // zh-Hans
+	0x51f34052: 742, // zh-Hans-CN
+	0x51f3408c: 743, // zh-Hans-HK
+	0x51f340c5: 744, // zh-Hans-MO
+	0x51f3410c: 745, // zh-Hans-SG
+	0x51f35000: 746, // zh-Hant
+	0x51f3508c: 747, // zh-Hant-HK
+	0x51f350c5: 748, // zh-Hant-MO
+	0x51f3512d: 749, // zh-Hant-TW
+	0x52400000: 750, // zu
+	0x52400160: 751, // zu-ZA
+}
+
+// Total table size 4580 bytes (4KiB); checksum: A7F72A2A

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

@@ -0,0 +1,975 @@
+// 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 maketables.go gen_common.go -output tables.go
+//go:generate go run gen_index.go
+
+// Package language implements BCP 47 language tags and related functionality.
+//
+// The Tag type, which is used to represent languages, is agnostic to the
+// meaning of its subtags. Tags are not fully canonicalized to preserve
+// information that may be valuable in certain contexts. As a consequence, two
+// different tags may represent identical languages.
+//
+// Initializing language- or locale-specific components usually consists of
+// two steps. The first step is to select a display language based on the
+// preferred languages of the user and the languages supported by an application.
+// The second step is to create the language-specific services based on
+// this selection. Each is discussed in more details below.
+//
+// Matching preferred against supported languages
+//
+// An application may support various languages. This list is typically limited
+// by the languages for which there exists translations of the user interface.
+// Similarly, a user may provide a list of preferred languages which is limited
+// by the languages understood by this user.
+// An application should use a Matcher to find the best supported language based
+// on the user's preferred list.
+// Matchers are aware of the intricacies of equivalence between languages.
+// The default Matcher implementation takes into account things such as
+// deprecated subtags, legacy tags, and mutual intelligibility between scripts
+// and languages.
+//
+// A Matcher for English, Australian English, Danish, and standard Mandarin can
+// be defined as follows:
+//
+//		var matcher = language.NewMatcher([]language.Tag{
+//			language.English,   // The first language is used as fallback.
+// 			language.MustParse("en-AU"),
+//			language.Danish,
+//			language.Chinese,
+//		})
+//
+// The following code selects the best match for someone speaking Spanish and
+// Norwegian:
+//
+// 		preferred := []language.Tag{ language.Spanish, language.Norwegian }
+//		tag, _, _ := matcher.Match(preferred...)
+//
+// In this case, the best match is Danish, as Danish is sufficiently a match to
+// Norwegian to not have to fall back to the default.
+// See ParseAcceptLanguage on how to handle the Accept-Language HTTP header.
+//
+// Selecting language-specific services
+//
+// One should always use the Tag returned by the Matcher to create an instance
+// of any of the language-specific services provided by the text repository.
+// This prevents the mixing of languages, such as having a different language for
+// messages and display names, as well as improper casing or sorting order for
+// the selected language.
+// Using the returned Tag also allows user-defined settings, such as collation
+// order or numbering system to be transparently passed as options.
+//
+// If you have language-specific data in your application, however, it will in
+// most cases suffice to use the index returned by the matcher to identify
+// the user language.
+// The following loop provides an alternative in case this is not sufficient:
+//
+// 		supported := map[language.Tag]data{
+//			language.English:            enData,
+// 			language.MustParse("en-AU"): enAUData,
+//			language.Danish:             daData,
+//			language.Chinese:            zhData,
+// 		}
+//		tag, _, _ := matcher.Match(preferred...)
+//		for ; tag != language.Und; tag = tag.Parent() {
+//			if v, ok := supported[tag]; ok {
+//				return v
+//			}
+//		}
+// 		return enData // should not reach here
+//
+// Repeatedly taking the Parent of the tag returned by Match will eventually
+// match one of the tags used to initialize the Matcher.
+//
+// Canonicalization
+//
+// By default, only legacy and deprecated tags are converted into their
+// canonical equivalent. All other information is preserved. This approach makes
+// the confidence scores more accurate and allows matchers to distinguish
+// between variants that are otherwise lost.
+//
+// As a consequence, two tags that should be treated as identical according to
+// BCP 47 or CLDR, like "en-Latn" and "en", will be represented differently. The
+// Matchers will handle such distinctions, though, and are aware of the
+// equivalence relations. The CanonType type can be used to alter the
+// canonicalization form.
+//
+// References
+//
+// BCP 47 - Tags for Identifying Languages
+// http://tools.ietf.org/html/bcp47
+package language // import "golang.org/x/text/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.
+type Tag struct {
+	lang     langID
+	region   regionID
+	script   scriptID
+	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 {
+	return Default.Make(s)
+}
+
+// Make is a convenience wrapper for c.Parse that omits the error.
+// In case of an error, a sensible default is returned.
+func (c CanonType) Make(s string) Tag {
+	t, _ := c.Parse(s)
+	return t
+}
+
+// Raw returns the raw base language, script and region, without making an
+// attempt to infer their values.
+func (t Tag) Raw() (b Base, s Script, r Region) {
+	return Base{t.lang}, Script{t.script}, Region{t.region}
+}
+
+// equalTags compares language, script and region subtags only.
+func (t Tag) equalTags(a Tag) bool {
+	return t.lang == a.lang && t.script == a.script && t.region == a.region
+}
+
+// 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)
+}
+
+// private reports whether the Tag consists solely of a private use tag.
+func (t Tag) private() bool {
+	return t.str != "" && t.pVariant == 0
+}
+
+// CanonType can be used to enable or disable various types of canonicalization.
+type CanonType int
+
+const (
+	// Replace deprecated base languages with their preferred replacements.
+	DeprecatedBase CanonType = 1 << iota
+	// Replace deprecated scripts with their preferred replacements.
+	DeprecatedScript
+	// Replace deprecated regions with their preferred replacements.
+	DeprecatedRegion
+	// Remove redundant scripts.
+	SuppressScript
+	// Normalize legacy encodings. This includes legacy languages defined in
+	// CLDR as well as bibliographic codes defined in ISO-639.
+	Legacy
+	// Map the dominant language of a macro language group to the macro language
+	// subtag. For example cmn -> zh.
+	Macro
+	// The CLDR flag should be used if full compatibility with CLDR is required.
+	// There are a few cases where language.Tag may differ from CLDR. To follow all
+	// of CLDR's suggestions, use All|CLDR.
+	CLDR
+
+	// Raw can be used to Compose or Parse without Canonicalization.
+	Raw CanonType = 0
+
+	// Replace all deprecated tags with their preferred replacements.
+	Deprecated = DeprecatedBase | DeprecatedScript | DeprecatedRegion
+
+	// All canonicalizations recommended by BCP 47.
+	BCP47 = Deprecated | SuppressScript
+
+	// All canonicalizations.
+	All = BCP47 | Legacy | Macro
+
+	// Default is the canonicalization used by Parse, Make and Compose. To
+	// preserve as much information as possible, canonicalizations that remove
+	// potentially valuable information are not included. The Matcher is
+	// designed to recognize similar tags that would be the same if
+	// they were canonicalized using All.
+	Default = Deprecated | Legacy
+
+	canonLang = DeprecatedBase | Legacy | Macro
+
+	// TODO: LikelyScript, LikelyRegion: suppress similar to ICU.
+)
+
+// canonicalize returns the canonicalized equivalent of the tag and
+// whether there was any change.
+func (t Tag) canonicalize(c CanonType) (Tag, bool) {
+	if c == Raw {
+		return t, false
+	}
+	changed := false
+	if c&SuppressScript != 0 {
+		if t.lang < langNoIndexOffset && uint8(t.script) == suppressScript[t.lang] {
+			t.script = 0
+			changed = true
+		}
+	}
+	if c&canonLang != 0 {
+		for {
+			if l, aliasType := normLang(t.lang); l != t.lang {
+				switch aliasType {
+				case langLegacy:
+					if c&Legacy != 0 {
+						if t.lang == _sh && t.script == 0 {
+							t.script = _Latn
+						}
+						t.lang = l
+						changed = true
+					}
+				case langMacro:
+					if c&Macro != 0 {
+						// We deviate here from CLDR. The mapping "nb" -> "no"
+						// qualifies as a typical Macro language mapping.  However,
+						// for legacy reasons, CLDR maps "no", the macro language
+						// code for Norwegian, to the dominant variant "nb". This
+						// change is currently under consideration for CLDR as well.
+						// See http://unicode.org/cldr/trac/ticket/2698 and also
+						// http://unicode.org/cldr/trac/ticket/1790 for some of the
+						// practical implications. TODO: this check could be removed
+						// if CLDR adopts this change.
+						if c&CLDR == 0 || t.lang != _nb {
+							changed = true
+							t.lang = l
+						}
+					}
+				case langDeprecated:
+					if c&DeprecatedBase != 0 {
+						if t.lang == _mo && t.region == 0 {
+							t.region = _MD
+						}
+						t.lang = l
+						changed = true
+						// Other canonicalization types may still apply.
+						continue
+					}
+				}
+			} else if c&Legacy != 0 && t.lang == _no && c&CLDR != 0 {
+				t.lang = _nb
+				changed = true
+			}
+			break
+		}
+	}
+	if c&DeprecatedScript != 0 {
+		if t.script == _Qaai {
+			changed = true
+			t.script = _Zinh
+		}
+	}
+	if c&DeprecatedRegion != 0 {
+		if r := normRegion(t.region); r != 0 {
+			changed = true
+			t.region = r
+		}
+	}
+	return t, changed
+}
+
+// Canonicalize returns the canonicalized equivalent of the tag.
+func (c CanonType) Canonicalize(t Tag) (Tag, error) {
+	t, changed := t.canonicalize(c)
+	if changed {
+		t.remakeString()
+	}
+	return t, nil
+}
+
+// Confidence indicates the level of certainty for a given return value.
+// For example, Serbian may be written in Cyrillic or Latin script.
+// The confidence level indicates whether a value was explicitly specified,
+// whether it is typically the only possible value, or whether there is
+// an ambiguity.
+type Confidence int
+
+const (
+	No    Confidence = iota // full confidence that there was no match
+	Low                     // most likely value picked out of a set of alternatives
+	High                    // value is generally assumed to be the correct match
+	Exact                   // exact match or explicitly specified value
+)
+
+var confName = []string{"No", "Low", "High", "Exact"}
+
+func (c Confidence) String() string {
+	return confName[c]
+}
+
+// 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.lang.stringToBuf(buf[:])
+	if t.script != 0 {
+		n += copy(buf[n:], "-")
+		n += copy(buf[n:], t.script.String())
+	}
+	if t.region != 0 {
+		n += copy(buf[n:], "-")
+		n += copy(buf[n:], t.region.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.script == 0 && t.region == 0 {
+		return t.lang.String()
+	}
+	buf := [maxCoreSize]byte{}
+	return string(buf[:t.genCoreBytes(buf[:])])
+}
+
+// Base returns the base language of the language tag. If the base language is
+// unspecified, an attempt will be made to infer it from the context.
+// It uses a variant of CLDR's Add Likely Subtags algorithm. This is subject to change.
+func (t Tag) Base() (Base, Confidence) {
+	if t.lang != 0 {
+		return Base{t.lang}, Exact
+	}
+	c := High
+	if t.script == 0 && !(Region{t.region}).IsCountry() {
+		c = Low
+	}
+	if tag, err := addTags(t); err == nil && tag.lang != 0 {
+		return Base{tag.lang}, c
+	}
+	return Base{0}, No
+}
+
+// Script infers the script for the language tag. If it was not explicitly given, it will infer
+// a most likely candidate.
+// If more than one script is commonly used for a language, the most likely one
+// is returned with a low confidence indication. For example, it returns (Cyrl, Low)
+// for Serbian.
+// If a script cannot be inferred (Zzzz, No) is returned. We do not use Zyyy (undetermined)
+// as one would suspect from the IANA registry for BCP 47. In a Unicode context Zyyy marks
+// common characters (like 1, 2, 3, '.', etc.) and is therefore more like multiple scripts.
+// See http://www.unicode.org/reports/tr24/#Values for more details. Zzzz is also used for
+// unknown value in CLDR.  (Zzzz, Exact) is returned if Zzzz was explicitly specified.
+// Note that an inferred script is never guaranteed to be the correct one. Latin is
+// almost exclusively used for Afrikaans, but Arabic has been used for some texts
+// in the past.  Also, the script that is commonly used may change over time.
+// It uses a variant of CLDR's Add Likely Subtags algorithm. This is subject to change.
+func (t Tag) Script() (Script, Confidence) {
+	if t.script != 0 {
+		return Script{t.script}, Exact
+	}
+	sc, c := scriptID(_Zzzz), No
+	if t.lang < langNoIndexOffset {
+		if scr := scriptID(suppressScript[t.lang]); scr != 0 {
+			// Note: it is not always the case that a language with a suppress
+			// script value is only written in one script (e.g. kk, ms, pa).
+			if t.region == 0 {
+				return Script{scriptID(scr)}, High
+			}
+			sc, c = scr, High
+		}
+	}
+	if tag, err := addTags(t); err == nil {
+		if tag.script != sc {
+			sc, c = tag.script, Low
+		}
+	} else {
+		t, _ = (Deprecated | Macro).Canonicalize(t)
+		if tag, err := addTags(t); err == nil && tag.script != sc {
+			sc, c = tag.script, Low
+		}
+	}
+	return Script{sc}, c
+}
+
+// Region returns the region for the language tag. If it was not explicitly given, it will
+// infer a most likely candidate from the context.
+// It uses a variant of CLDR's Add Likely Subtags algorithm. This is subject to change.
+func (t Tag) Region() (Region, Confidence) {
+	if t.region != 0 {
+		return Region{t.region}, Exact
+	}
+	if t, err := addTags(t); err == nil {
+		return Region{t.region}, Low // TODO: differentiate between high and low.
+	}
+	t, _ = (Deprecated | Macro).Canonicalize(t)
+	if tag, err := addTags(t); err == nil {
+		return Region{tag.region}, Low
+	}
+	return Region{_ZZ}, No // TODO: return world instead of undetermined?
+}
+
+// Variant returns the variants specified explicitly for this language tag.
+// or nil if no variant was specified.
+func (t Tag) Variants() []Variant {
+	v := []Variant{}
+	if int(t.pVariant) < int(t.pExt) {
+		for x, str := "", t.str[t.pVariant:t.pExt]; str != ""; {
+			x, str = nextToken(str)
+			v = append(v, Variant{x})
+		}
+	}
+	return v
+}
+
+// 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.
+		t, _ = Raw.Compose(t.Raw())
+		if t.region == 0 && t.script != 0 && t.lang != 0 {
+			base, _ := addTags(Tag{lang: t.lang})
+			if base.script == t.script {
+				return Tag{lang: t.lang}
+			}
+		}
+		return t
+	}
+	if t.lang != 0 {
+		if t.region != 0 {
+			maxScript := t.script
+			if maxScript == 0 {
+				max, _ := addTags(t)
+				maxScript = max.script
+			}
+
+			for i := range parents {
+				if langID(parents[i].lang) == t.lang && scriptID(parents[i].maxScript) == maxScript {
+					for _, r := range parents[i].fromRegion {
+						if regionID(r) == t.region {
+							return Tag{
+								lang:   t.lang,
+								script: scriptID(parents[i].script),
+								region: regionID(parents[i].toRegion),
+							}
+						}
+					}
+				}
+			}
+
+			// Strip the script if it is the default one.
+			base, _ := addTags(Tag{lang: t.lang})
+			if base.script != maxScript {
+				return Tag{lang: t.lang, script: maxScript}
+			}
+			return Tag{lang: t.lang}
+		} else if t.script != 0 {
+			// The parent for an base-script pair with a non-default script is
+			// "und" instead of the base language.
+			base, _ := addTags(Tag{lang: t.lang})
+			if base.script != t.script {
+				return und
+			}
+			return Tag{lang: t.lang}
+		}
+	}
+	return und
+}
+
+// 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:]
+}
+
+// Extension is a single BCP 47 extension.
+type Extension struct {
+	s string
+}
+
+// String returns the string representation of the extension, including the
+// type tag.
+func (e Extension) String() string {
+	return e.s
+}
+
+// ParseExtension parses s as an extension and returns it on success.
+func ParseExtension(s string) (e Extension, err error) {
+	scan := makeScannerString(s)
+	var end int
+	if n := len(scan.token); n != 1 {
+		return Extension{}, errSyntax
+	}
+	scan.toLower(0, len(scan.b))
+	end = parseExtension(&scan)
+	if end != len(s) {
+		return Extension{}, errSyntax
+	}
+	return Extension{string(scan.b)}, nil
+}
+
+// Type returns the one-byte extension type of e. It returns 0 for the zero
+// exception.
+func (e Extension) Type() byte {
+	if e.s == "" {
+		return 0
+	}
+	return e.s[0]
+}
+
+// Tokens returns the list of tokens of e.
+func (e Extension) Tokens() []string {
+	return strings.Split(e.s, "-")
+}
+
+// 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 Extension, 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 Extension{ext}, true
+		}
+	}
+	return Extension{}, false
+}
+
+// Extensions returns all extensions of t.
+func (t Tag) Extensions() []Extension {
+	e := []Extension{}
+	for i := int(t.pExt); i < len(t.str)-1; {
+		var ext string
+		i, ext = getExtension(t.str, i)
+		e = append(e, Extension{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
+// http://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
+// http://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.private() {
+		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
+		}
+	}
+}
+
+// CompactIndex 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. Extensions, except for the
+// 'va' type of the 'u' extension, are ignored. It will return 0, false if no
+// compact tag exists, where 0 is the index for the root language (Und).
+func CompactIndex(t Tag) (index int, ok 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.
+	b, s, r := t.Raw()
+	if len(t.str) > 0 {
+		if strings.HasPrefix(t.str, "x-") {
+			// We have no entries for user-defined tags.
+			return 0, false
+		}
+		if uint16(t.pVariant) != t.pExt {
+			// There are no tags with variants and an u-va type.
+			if t.TypeForKey("va") != "" {
+				return 0, false
+			}
+			t, _ = Raw.Compose(b, s, r, t.Variants())
+		} else if _, ok := t.Extension('u'); ok {
+			// Strip all but the 'va' entry.
+			variant := t.TypeForKey("va")
+			t, _ = Raw.Compose(b, s, r)
+			t, _ = t.SetTypeForKey("va", variant)
+		}
+		if len(t.str) > 0 {
+			// We have some variants.
+			for i, s := range specialTags {
+				if s == t {
+					return i + 1, true
+				}
+			}
+			return 0, false
+		}
+	}
+	// 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.
+	key := uint32(b.langID) << (8 + 12)
+	key |= uint32(s.scriptID) << 12
+	key |= uint32(r.regionID)
+	x, ok := coreTags[key]
+	return int(x), ok
+}
+
+// Base is an ISO 639 language code, used for encoding the base language
+// of a language tag.
+type Base struct {
+	langID
+}
+
+// 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) (Base, error) {
+	if n := len(s); n < 2 || 3 < n {
+		return Base{}, errSyntax
+	}
+	var buf [3]byte
+	l, err := getLangID(buf[:copy(buf[:], s)])
+	return Base{l}, err
+}
+
+// Script is a 4-letter ISO 15924 code for representing scripts.
+// It is idiomatically represented in title case.
+type Script struct {
+	scriptID
+}
+
+// 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 Script{}, errSyntax
+	}
+	var buf [4]byte
+	sc, err := getScriptID(script, buf[:copy(buf[:], s)])
+	return Script{sc}, err
+}
+
+// Region is an ISO 3166-1 or UN M.49 code for representing countries and regions.
+type Region struct {
+	regionID
+}
+
+// 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) {
+	rid, err := getRegionM49(r)
+	return Region{rid}, err
+}
+
+// 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 Region{}, errSyntax
+	}
+	var buf [3]byte
+	r, err := getRegionID(buf[:copy(buf[:], s)])
+	return Region{r}, err
+}
+
+// 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.regionID == 0 || r.IsGroup() || r.IsPrivateUse() && r.regionID != _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.regionID == 0 {
+		return false
+	}
+	return int(regionInclusion[r.regionID]) < 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 {
+	return r.regionID.contains(c.regionID)
+}
+
+func (r regionID) contains(c regionID) 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.regionID == _GB {
+		r = Region{_UK}
+	}
+	if (r.typ() & ccTLD) == 0 {
+		return Region{}, 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.regionID); cr != 0 {
+		return Region{cr}
+	}
+	return r
+}
+
+// Variant represents a registered variant of a language as defined by BCP 47.
+type Variant struct {
+	variant 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 _, ok := variantIndex[s]; ok {
+		return Variant{s}, nil
+	}
+	return Variant{}, mkErrInvalid([]byte(s))
+}
+
+// String returns the string representation of the variant.
+func (v Variant) String() string {
+	return v.variant
+}

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

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

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

@@ -0,0 +1,841 @@
+// 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"
+
+// Matcher is the interface that wraps the Match method.
+//
+// Match returns the best match for any of the given tags, along with
+// a unique index associated with the returned tag and a confidence
+// score.
+type Matcher interface {
+	Match(t ...Tag) (tag Tag, index int, c Confidence)
+}
+
+// Comprehends reports the confidence score for a speaker of a given language
+// to being able to comprehend the written form of an alternative language.
+func Comprehends(speaker, alternative Tag) Confidence {
+	_, _, c := NewMatcher([]Tag{alternative}).Match(speaker)
+	return c
+}
+
+// NewMatcher returns a Matcher that matches an ordered list of preferred tags
+// against a list of supported tags based on written intelligibility, closeness
+// of dialect, equivalence of subtags and various other rules. It is initialized
+// with the list of supported tags. The first element is used as the default
+// value in case no match is found.
+//
+// Its Match method matches the first of the given Tags to reach a certain
+// confidence threshold. The tags passed to Match should therefore be specified
+// in order of preference. Extensions are ignored for matching.
+//
+// The index returned by the Match method corresponds to the index of the
+// matched tag in t, but is augmented with the Unicode extension ('u')of the
+// corresponding preferred tag. This allows user locale options to be passed
+// transparently.
+func NewMatcher(t []Tag) Matcher {
+	return newMatcher(t)
+}
+
+func (m *matcher) Match(want ...Tag) (t Tag, index int, c Confidence) {
+	match, w, c := m.getBest(want...)
+	if match == nil {
+		t = m.default_.tag
+	} else {
+		t, index = match.tag, match.index
+	}
+	// Copy options from the user-provided tag into the result tag. This is hard
+	// to do after the fact, so we do it here.
+	// TODO: consider also adding in variants that are compatible with the
+	// matched language.
+	// TODO: Add back region if it is non-ambiguous? Or create another tag to
+	// preserve the region?
+	if u, ok := w.Extension('u'); ok {
+		t, _ = Raw.Compose(t, u)
+	}
+	return t, index, c
+}
+
+type scriptRegionFlags uint8
+
+const (
+	isList = 1 << iota
+	scriptInFrom
+	regionInFrom
+)
+
+func (t *Tag) setUndefinedLang(id langID) {
+	if t.lang == 0 {
+		t.lang = id
+	}
+}
+
+func (t *Tag) setUndefinedScript(id scriptID) {
+	if t.script == 0 {
+		t.script = id
+	}
+}
+
+func (t *Tag) setUndefinedRegion(id regionID) {
+	if t.region == 0 || t.region.contains(id) {
+		t.region = 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 a 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.region]; i < nRegionGroups {
+		x := likelyRegionGroup[i]
+		if langID(x.lang) == t.lang && scriptID(x.script) == t.script {
+			t.region = regionID(x.region)
+		}
+		return true
+	}
+	return false
+}
+
+func addTags(t Tag) (Tag, error) {
+	// We leave private use identifiers alone.
+	if t.private() {
+		return t, nil
+	}
+	if t.script != 0 && t.region != 0 {
+		if t.lang != 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.region : t.region+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 scriptID(x.script) == t.script {
+				t.setUndefinedLang(langID(x.lang))
+				return t, nil
+			}
+		}
+	}
+	if t.lang != 0 {
+		// Search matches for lang-script and lang-region, where lang != und.
+		if t.lang < langNoIndexOffset {
+			x := likelyLang[t.lang]
+			if x.flags&isList != 0 {
+				list := likelyLangList[x.region : x.region+uint16(x.script)]
+				if t.script != 0 {
+					for _, x := range list {
+						if scriptID(x.script) == t.script && x.flags&scriptInFrom != 0 {
+							t.setUndefinedRegion(regionID(x.region))
+							return t, nil
+						}
+					}
+				} else if t.region != 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.region.contains(regionID(x.region)) {
+							tt.region = regionID(x.region)
+							tt.setUndefinedScript(scriptID(x.script))
+							goodScript = goodScript && tt.script == scriptID(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.script = tt.script
+					}
+				}
+			}
+		}
+	} else {
+		// Search matches for und-script.
+		if t.script != 0 {
+			x := likelyScript[t.script]
+			if x.region != 0 {
+				t.setUndefinedRegion(regionID(x.region))
+				t.setUndefinedLang(langID(x.lang))
+				return t, nil
+			}
+		}
+		// Search matches for und-region. If und-script-region exists, it would
+		// have been found earlier.
+		if t.region != 0 {
+			if i := regionInclusion[t.region]; i < nRegionGroups {
+				x := likelyRegionGroup[i]
+				if x.region != 0 {
+					t.setUndefinedLang(langID(x.lang))
+					t.setUndefinedScript(scriptID(x.script))
+					t.region = regionID(x.region)
+				}
+			} else {
+				x := likelyRegion[t.region]
+				if x.flags&isList != 0 {
+					x = likelyRegionList[x.lang]
+				}
+				if x.script != 0 && x.flags != scriptInFrom {
+					t.setUndefinedLang(langID(x.lang))
+					t.setUndefinedScript(scriptID(x.script))
+					return t, nil
+				}
+			}
+		}
+	}
+
+	// Search matches for lang.
+	if t.lang < langNoIndexOffset {
+		x := likelyLang[t.lang]
+		if x.flags&isList != 0 {
+			x = likelyLangList[x.region]
+		}
+		if x.region != 0 {
+			t.setUndefinedScript(scriptID(x.script))
+			t.setUndefinedRegion(regionID(x.region))
+		}
+		specializeRegion(&t)
+		if t.lang == 0 {
+			t.lang = _en // default language
+		}
+		return t, nil
+	}
+	return t, ErrMissingLikelyTagsData
+}
+
+func (t *Tag) setTagsFrom(id Tag) {
+	t.lang = id.lang
+	t.script = id.script
+	t.region = id.region
+}
+
+// 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{
+		{lang: t.lang},
+		{lang: t.lang, region: t.region},
+		{lang: t.lang, script: t.script},
+	} {
+		if x, err := addTags(id); err == nil && max.equalTags(x) {
+			t.setTagsFrom(id)
+			break
+		}
+	}
+	return t, nil
+}
+
+// Tag Matching
+// CLDR defines an algorithm for finding the best match between two sets of language
+// tags. The basic algorithm defines how to score a possible match and then find
+// the match with the best score
+// (see http://www.unicode.org/reports/tr35/#LanguageMatching).
+// Using scoring has several disadvantages. The scoring obfuscates the importance of
+// the various factors considered, making the algorithm harder to understand. Using
+// scoring also requires the full score to be computed for each pair of tags.
+//
+// We will use a different algorithm which aims to have the following properties:
+// - clarity on the precedence of the various selection factors, and
+// - improved performance by allowing early termination of a comparison.
+//
+// Matching algorithm (overview)
+// Input:
+//   - supported: a set of supported tags
+//   - default:   the default tag to return in case there is no match
+//   - desired:   list of desired tags, ordered by preference, starting with
+//                the most-preferred.
+//
+// Algorithm:
+//   1) Set the best match to the lowest confidence level
+//   2) For each tag in "desired":
+//     a) For each tag in "supported":
+//        1) compute the match between the two tags.
+//        2) if the match is better than the previous best match, replace it
+//           with the new match. (see next section)
+//     b) if the current best match is above a certain threshold, return this
+//        match without proceeding to the next tag in "desired". [See Note 1]
+//   3) If the best match so far is below a certain threshold, return "default".
+//
+// Ranking:
+// We use two phases to determine whether one pair of tags are a better match
+// than another pair of tags. First, we determine a rough confidence level. If the
+// levels are different, the one with the highest confidence wins.
+// Second, if the rough confidence levels are identical, we use a set of tie-breaker
+// rules.
+//
+// The confidence level of matching a pair of tags is determined by finding the
+// lowest confidence level of any matches of the corresponding subtags (the
+// result is deemed as good as its weakest link).
+// We define the following levels:
+//   Exact    - An exact match of a subtag, before adding likely subtags.
+//   MaxExact - An exact match of a subtag, after adding likely subtags.
+//              [See Note 2].
+//   High     - High level of mutual intelligibility between different subtag
+//              variants.
+//   Low      - Low level of mutual intelligibility between different subtag
+//              variants.
+//   No       - No mutual intelligibility.
+//
+// The following levels can occur for each type of subtag:
+//   Base:    Exact, MaxExact, High, Low, No
+//   Script:  Exact, MaxExact [see Note 3], Low, No
+//   Region:  Exact, MaxExact, High
+//   Variant: Exact, High
+//   Private: Exact, No
+//
+// Any result with a confidence level of Low or higher is deemed a possible match.
+// Once a desired tag matches any of the supported tags with a level of MaxExact
+// or higher, the next desired tag is not considered (see Step 2.b).
+// Note that CLDR provides languageMatching data that defines close equivalence
+// classes for base languages, scripts and regions.
+//
+// Tie-breaking
+// If we get the same confidence level for two matches, we apply a sequence of
+// tie-breaking rules. The first that succeeds defines the result. The rules are
+// applied in the following order.
+//   1) Original language was defined and was identical.
+//   2) Original region was defined and was identical.
+//   3) Distance between two maximized regions was the smallest.
+//   4) Original script was defined and was identical.
+//   5) Distance from want tag to have tag using the parent relation [see Note 5.]
+// If there is still no winner after these rules are applied, the first match
+// found wins.
+//
+// Notes:
+// [1] Note that even if we may not have a perfect match, if a match is above a
+//     certain threshold, it is considered a better match than any other match
+//     to a tag later in the list of preferred language tags.
+// [2] In practice, as matching of Exact is done in a separate phase from
+//     matching the other levels, we reuse the Exact level to mean MaxExact in
+//     the second phase. As a consequence, we only need the levels defined by
+//     the Confidence type. The MaxExact confidence level is mapped to High in
+//     the public API.
+// [3] We do not differentiate between maximized script values that were derived
+//     from suppressScript versus most likely tag data. We determined that in
+//     ranking the two, one ranks just after the other. Moreover, the two cannot
+//     occur concurrently. As a consequence, they are identical for practical
+//     purposes.
+// [4] In case of deprecated, macro-equivalents and legacy mappings, we assign
+//     the MaxExact level to allow iw vs he to still be a closer match than
+//     en-AU vs en-US, for example.
+// [5] In CLDR a locale inherits fields that are unspecified for this locale
+//     from its parent. Therefore, if a locale is a parent of another locale,
+//     it is a strong measure for closeness, especially when no other tie
+//     breaker rule applies. One could also argue it is inconsistent, for
+//     example, when pt-AO matches pt (which CLDR equates with pt-BR), even
+//     though its parent is pt-PT according to the inheritance rules.
+//
+// Implementation Details:
+// There are several performance considerations worth pointing out. Most notably,
+// we preprocess as much as possible (within reason) at the time of creation of a
+// matcher. This includes:
+//   - creating a per-language map, which includes data for the raw base language
+//     and its canonicalized variant (if applicable),
+//   - expanding entries for the equivalence classes defined in CLDR's
+//     languageMatch data.
+// The per-language map ensures that typically only a very small number of tags
+// need to be considered. The pre-expansion of canonicalized subtags and
+// equivalence classes reduces the amount of map lookups that need to be done at
+// runtime.
+
+// matcher keeps a set of supported language tags, indexed by language.
+type matcher struct {
+	default_     *haveTag
+	index        map[langID]*matchHeader
+	passSettings bool
+}
+
+// matchHeader has the lists of tags for exact matches and matches based on
+// maximized and canonicalized tags for a given language.
+type matchHeader struct {
+	exact []*haveTag
+	max   []*haveTag
+}
+
+// haveTag holds a supported Tag and its maximized script and region. The maximized
+// or canonicalized language is not stored as it is not needed during matching.
+type haveTag struct {
+	tag Tag
+
+	// index of this tag in the original list of supported tags.
+	index int
+
+	// conf is the maximum confidence that can result from matching this haveTag.
+	// When conf < Exact this means it was inserted after applying a CLDR equivalence rule.
+	conf Confidence
+
+	// Maximized region and script.
+	maxRegion regionID
+	maxScript scriptID
+
+	// altScript may be checked as an alternative match to maxScript. If altScript
+	// matches, the confidence level for this match is Low. Theoretically there
+	// could be multiple alternative scripts. This does not occur in practice.
+	altScript scriptID
+
+	// nextMax is the index of the next haveTag with the same maximized tags.
+	nextMax uint16
+}
+
+func makeHaveTag(tag Tag, index int) (haveTag, langID) {
+	max := tag
+	if tag.lang != 0 {
+		max, _ = max.canonicalize(All)
+		max, _ = addTags(max)
+		max.remakeString()
+	}
+	return haveTag{tag, index, Exact, max.region, max.script, altScript(max.lang, max.script), 0}, max.lang
+}
+
+// altScript returns an alternative script that may match the given script with
+// a low confidence.  At the moment, the langMatch data allows for at most one
+// script to map to another and we rely on this to keep the code simple.
+func altScript(l langID, s scriptID) scriptID {
+	for _, alt := range matchScript {
+		if (alt.lang == 0 || langID(alt.lang) == l) && scriptID(alt.have) == s {
+			return scriptID(alt.want)
+		}
+	}
+	return 0
+}
+
+// addIfNew adds a haveTag to the list of tags only if it is a unique tag.
+// Tags that have the same maximized values are linked by index.
+func (h *matchHeader) addIfNew(n haveTag, exact bool) {
+	// Don't add new exact matches.
+	for _, v := range h.exact {
+		if v.tag.equalsRest(n.tag) {
+			return
+		}
+	}
+	if exact {
+		h.exact = append(h.exact, &n)
+	}
+	// Allow duplicate maximized tags, but create a linked list to allow quickly
+	// comparing the equivalents and bail out.
+	for i, v := range h.max {
+		if v.maxScript == n.maxScript &&
+			v.maxRegion == n.maxRegion &&
+			v.tag.variantOrPrivateTagStr() == n.tag.variantOrPrivateTagStr() {
+			for h.max[i].nextMax != 0 {
+				i = int(h.max[i].nextMax)
+			}
+			h.max[i].nextMax = uint16(len(h.max))
+			break
+		}
+	}
+	h.max = append(h.max, &n)
+}
+
+// header returns the matchHeader for the given language. It creates one if
+// it doesn't already exist.
+func (m *matcher) header(l langID) *matchHeader {
+	if h := m.index[l]; h != nil {
+		return h
+	}
+	h := &matchHeader{}
+	m.index[l] = h
+	return h
+}
+
+// newMatcher builds an index for the given supported tags and returns it as
+// a matcher. It also expands the index by considering various equivalence classes
+// for a given tag.
+func newMatcher(supported []Tag) *matcher {
+	m := &matcher{
+		index: make(map[langID]*matchHeader),
+	}
+	if len(supported) == 0 {
+		m.default_ = &haveTag{}
+		return m
+	}
+	// Add supported languages to the index. Add exact matches first to give
+	// them precedence.
+	for i, tag := range supported {
+		pair, _ := makeHaveTag(tag, i)
+		m.header(tag.lang).addIfNew(pair, true)
+	}
+	m.default_ = m.header(supported[0].lang).exact[0]
+	for i, tag := range supported {
+		pair, max := makeHaveTag(tag, i)
+		if max != tag.lang {
+			m.header(max).addIfNew(pair, false)
+		}
+	}
+
+	// update is used to add indexes in the map for equivalent languages.
+	// If force is true, the update will also apply to derived entries. To
+	// avoid applying a "transitive closure", use false.
+	update := func(want, have uint16, conf Confidence, force bool) {
+		if hh := m.index[langID(have)]; hh != nil {
+			if !force && len(hh.exact) == 0 {
+				return
+			}
+			hw := m.header(langID(want))
+			for _, ht := range hh.max {
+				v := *ht
+				if conf < v.conf {
+					v.conf = conf
+				}
+				v.nextMax = 0 // this value needs to be recomputed
+				if v.altScript != 0 {
+					v.altScript = altScript(langID(want), v.maxScript)
+				}
+				hw.addIfNew(v, conf == Exact && len(hh.exact) > 0)
+			}
+		}
+	}
+
+	// Add entries for languages with mutual intelligibility as defined by CLDR's
+	// languageMatch data.
+	for _, ml := range matchLang {
+		update(ml.want, ml.have, Confidence(ml.conf), false)
+		if !ml.oneway {
+			update(ml.have, ml.want, Confidence(ml.conf), false)
+		}
+	}
+
+	// Add entries for possible canonicalizations. This is an optimization to
+	// ensure that only one map lookup needs to be done at runtime per desired tag.
+	// First we match deprecated equivalents. If they are perfect equivalents
+	// (their canonicalization simply substitutes a different language code, but
+	// nothing else), the match confidence is Exact, otherwise it is High.
+	for i, lm := range langAliasMap {
+		if lm.from == _sh {
+			continue
+		}
+
+		// If deprecated codes match and there is no fiddling with the script or
+		// or region, we consider it an exact match.
+		conf := Exact
+		if langAliasTypes[i] != langMacro {
+			if !isExactEquivalent(langID(lm.from)) {
+				conf = High
+			}
+			update(lm.to, lm.from, conf, true)
+		}
+		update(lm.from, lm.to, conf, true)
+	}
+	return m
+}
+
+// getBest gets the best matching tag in m for any of the given tags, taking into
+// account the order of preference of the given tags.
+func (m *matcher) getBest(want ...Tag) (got *haveTag, orig Tag, c Confidence) {
+	best := bestMatch{}
+	for _, w := range want {
+		var max Tag
+		// Check for exact match first.
+		h := m.index[w.lang]
+		if w.lang != 0 {
+			// Base language is defined.
+			if h == nil {
+				continue
+			}
+			for i := range h.exact {
+				have := h.exact[i]
+				if have.tag.equalsRest(w) {
+					return have, w, Exact
+				}
+			}
+			max, _ = w.canonicalize(Legacy | Deprecated)
+			max, _ = addTags(max)
+		} else {
+			// Base language is not defined.
+			if h != nil {
+				for i := range h.exact {
+					have := h.exact[i]
+					if have.tag.equalsRest(w) {
+						return have, w, Exact
+					}
+				}
+			}
+			if w.script == 0 && w.region == 0 {
+				// We skip all tags matching und for approximate matching, including
+				// private tags.
+				continue
+			}
+			max, _ = addTags(w)
+			if h = m.index[max.lang]; h == nil {
+				continue
+			}
+		}
+		// Check for match based on maximized tag.
+		for i := range h.max {
+			have := h.max[i]
+			best.update(have, w, max.script, max.region)
+			if best.conf == Exact {
+				for have.nextMax != 0 {
+					have = h.max[have.nextMax]
+					best.update(have, w, max.script, max.region)
+				}
+				return best.have, best.want, High
+			}
+		}
+	}
+	if best.conf <= No {
+		if len(want) != 0 {
+			return nil, want[0], No
+		}
+		return nil, Tag{}, No
+	}
+	return best.have, best.want, best.conf
+}
+
+// bestMatch accumulates the best match so far.
+type bestMatch struct {
+	have *haveTag
+	want Tag
+	conf Confidence
+	// Cached results from applying tie-breaking rules.
+	origLang   bool
+	origReg    bool
+	regDist    uint8
+	origScript bool
+	parentDist uint8 // 255 if have is not an ancestor of want tag.
+}
+
+// update updates the existing best match if the new pair is considered to be a
+// better match.
+// To determine if the given pair is a better match, it first computes the rough
+// confidence level. If this surpasses the current match, it will replace it and
+// update the tie-breaker rule cache. If there is a tie, it proceeds with applying
+// a series of tie-breaker rules. If there is no conclusive winner after applying
+// the tie-breaker rules, it leaves the current match as the preferred match.
+func (m *bestMatch) update(have *haveTag, tag Tag, maxScript scriptID, maxRegion regionID) {
+	// Bail if the maximum attainable confidence is below that of the current best match.
+	c := have.conf
+	if c < m.conf {
+		return
+	}
+	if have.maxScript != maxScript {
+		// There is usually very little comprehension between different scripts.
+		// In a few cases there may still be Low comprehension. This possibility is
+		// pre-computed and stored in have.altScript.
+		if Low < m.conf || have.altScript != maxScript {
+			return
+		}
+		c = Low
+	} else if have.maxRegion != maxRegion {
+		// There is usually a small difference between languages across regions.
+		// We use the region distance (below) to disambiguate between equal matches.
+		if High < c {
+			c = High
+		}
+	}
+
+	// We store the results of the computations of the tie-breaker rules along
+	// with the best match. There is no need to do the checks once we determine
+	// we have a winner, but we do still need to do the tie-breaker computations.
+	// We use "beaten" to keep track if we still need to do the checks.
+	beaten := false // true if the new pair defeats the current one.
+	if c != m.conf {
+		if c < m.conf {
+			return
+		}
+		beaten = true
+	}
+
+	// Tie-breaker rules:
+	// We prefer if the pre-maximized language was specified and identical.
+	origLang := have.tag.lang == tag.lang && tag.lang != 0
+	if !beaten && m.origLang != origLang {
+		if m.origLang {
+			return
+		}
+		beaten = true
+	}
+
+	// We prefer if the pre-maximized region was specified and identical.
+	origReg := have.tag.region == tag.region && tag.region != 0
+	if !beaten && m.origReg != origReg {
+		if m.origReg {
+			return
+		}
+		beaten = true
+	}
+
+	// Next we prefer smaller distances between regions, as defined by regionDist.
+	regDist := regionDist(have.maxRegion, maxRegion, tag.lang)
+	if !beaten && m.regDist != regDist {
+		if regDist > m.regDist {
+			return
+		}
+		beaten = true
+	}
+
+	// Next we prefer if the pre-maximized script was specified and identical.
+	origScript := have.tag.script == tag.script && tag.script != 0
+	if !beaten && m.origScript != origScript {
+		if m.origScript {
+			return
+		}
+		beaten = true
+	}
+
+	// Finally we prefer tags which have a closer parent relationship.
+	parentDist := parentDistance(have.tag.region, tag)
+	if !beaten && m.parentDist != parentDist {
+		if parentDist > m.parentDist {
+			return
+		}
+		beaten = true
+	}
+
+	// Update m to the newly found best match.
+	if beaten {
+		m.have = have
+		m.want = tag
+		m.conf = c
+		m.origLang = origLang
+		m.origReg = origReg
+		m.origScript = origScript
+		m.regDist = regDist
+		m.parentDist = parentDist
+	}
+}
+
+// parentDistance returns the number of times Parent must be called before the
+// regions match. It is assumed that it has already been checked that lang and
+// script are identical. If haveRegion does not occur in the ancestor chain of
+// tag, it returns 255.
+func parentDistance(haveRegion regionID, tag Tag) uint8 {
+	p := tag.Parent()
+	d := uint8(1)
+	for haveRegion != p.region {
+		if p.region == 0 {
+			return 255
+		}
+		p = p.Parent()
+		d++
+	}
+	return d
+}
+
+// regionDist wraps regionDistance with some exceptions to the algorithmic distance.
+func regionDist(a, b regionID, lang langID) uint8 {
+	if lang == _en {
+		// Two variants of non-US English are close to each other, regardless of distance.
+		if a != _US && b != _US {
+			return 2
+		}
+	}
+	return uint8(regionDistance(a, b))
+}
+
+// regionDistance computes the distance between two regions based on the
+// distance in the graph of region containments as defined in CLDR. It iterates
+// over increasingly inclusive sets of groups, represented as bit vectors, until
+// the source bit vector has bits in common with the destination vector.
+func regionDistance(a, b regionID) int {
+	if a == b {
+		return 0
+	}
+	p, q := regionInclusion[a], regionInclusion[b]
+	if p < nRegionGroups {
+		p, q = q, p
+	}
+	set := regionInclusionBits
+	if q < nRegionGroups && set[p]&(1<<q) != 0 {
+		return 1
+	}
+	d := 2
+	for goal := set[q]; set[p]&goal == 0; p = regionInclusionNext[p] {
+		d++
+	}
+	return d
+}
+
+func (t Tag) variants() string {
+	if t.pVariant == 0 {
+		return ""
+	}
+	return t.str[t.pVariant:t.pExt]
+}
+
+// variantOrPrivateTagStr returns variants or private use tags.
+func (t Tag) variantOrPrivateTagStr() string {
+	if t.pExt > 0 {
+		return t.str[t.pVariant:t.pExt]
+	}
+	return t.str[t.pVariant:]
+}
+
+// equalsRest compares everything except the language.
+func (a Tag) equalsRest(b Tag) bool {
+	// TODO: don't include extensions in this comparison. To do this efficiently,
+	// though, we should handle private tags separately.
+	return a.script == b.script && a.region == b.region && a.variantOrPrivateTagStr() == b.variantOrPrivateTagStr()
+}
+
+// isExactEquivalent returns true if canonicalizing the language will not alter
+// the script or region of a tag.
+func isExactEquivalent(l langID) bool {
+	for _, o := range notEquivalent {
+		if o == l {
+			return false
+		}
+	}
+	return true
+}
+
+var notEquivalent []langID
+
+func init() {
+	// Create a list of all languages for which canonicalization may alter the
+	// script or region.
+	for _, lm := range langAliasMap {
+		tag := Tag{lang: langID(lm.from)}
+		if tag, _ = tag.canonicalize(All); tag.script != 0 || tag.region != 0 {
+			notEquivalent = append(notEquivalent, langID(lm.from))
+		}
+	}
+}

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

@@ -0,0 +1,859 @@
+// 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"
+	"strconv"
+	"strings"
+
+	"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")
+
+// 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
+}
+
+func mkErrInvalid(s []byte) error {
+	var e ValueError
+	copy(e.v[:], s)
+	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.setError(errSyntax)
+	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
+// http://www.unicode.org/reports/tr35/#Unicode_Language_and_Locale_Identifiers.
+// The resulting tag is canonicalized using the default canonicalization type.
+func Parse(s string) (t Tag, err error) {
+	return Default.Parse(s)
+}
+
+// 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
+// http://www.unicode.org/reports/tr35/#Unicode_Language_and_Locale_Identifiers.
+// The resulting tag is canonicalized using the the canonicalization type c.
+func (c CanonType) 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)
+	t, err = parse(&scan, s)
+	t, changed := t.canonicalize(c)
+	if changed {
+		t.remakeString()
+	}
+	return t, err
+}
+
+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.lang, e = getLangID(scan.token)
+	scan.setError(e)
+	scan.replace(t.lang.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.lang = 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.script, e = getScriptID(script, scan.token)
+		if t.script == 0 {
+			scan.gobble(e)
+		}
+		end = scan.scan()
+	}
+	if n := len(scan.token); n >= 2 && n <= 3 {
+		t.region, e = getRegionID(scan.token)
+		if t.region == 0 {
+			scan.gobble(e)
+		} else {
+			scan.replace(t.region.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(mkErrInvalid(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 [][]byte
+
+func (b bytesSort) Len() int {
+	return len(b)
+}
+
+func (b bytesSort) Swap(i, j int) {
+	b[i], b[j] = b[j], b[i]
+}
+
+func (b bytesSort) Less(i, j int) bool {
+	return bytes.Compare(b[i], b[j]) == -1
+}
+
+// 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))
+	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))
+				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.Sort(bytesSort(keys))
+				reordered := bytes.Join(keys, separator)
+				if e := p + len(reordered); e < end {
+					scan.deleteRange(e, end)
+					end = e
+				}
+				copy(scan.b[p:], bytes.Join(keys, separator))
+				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
+}
+
+// Compose creates a Tag from individual parts, which may be of type Tag, Base,
+// Script, Region, Variant, []Variant, Extension, []Extension or error. If a
+// Base, Script or Region or slice of type Variant or Extension is passed more
+// than once, the latter will overwrite the former. Variants and Extensions are
+// accumulated, but if two extensions of the same type are passed, the latter
+// will replace the former. A Tag overwrites all former values and typically
+// only makes sense as the first argument. The resulting tag is returned after
+// canonicalizing using the Default CanonType. If one or more errors are
+// encountered, one of the errors is returned.
+func Compose(part ...interface{}) (t Tag, err error) {
+	return Default.Compose(part...)
+}
+
+// Compose creates a Tag from individual parts, which may be of type Tag, Base,
+// Script, Region, Variant, []Variant, Extension, []Extension or error. If a
+// Base, Script or Region or slice of type Variant or Extension is passed more
+// than once, the latter will overwrite the former. Variants and Extensions are
+// accumulated, but if two extensions of the same type are passed, the latter
+// will replace the former. A Tag overwrites all former values and typically
+// only makes sense as the first argument. The resulting tag is returned after
+// canonicalizing using CanonType c. If one or more errors are encountered,
+// one of the errors is returned.
+func (c CanonType) Compose(part ...interface{}) (t Tag, err error) {
+	var b builder
+	if err = b.update(part...); err != nil {
+		return und, err
+	}
+	t, _ = b.tag.canonicalize(c)
+
+	if len(b.ext) > 0 || len(b.variant) > 0 {
+		sort.Sort(sortVariant(b.variant))
+		sort.Strings(b.ext)
+		if b.private != "" {
+			b.ext = append(b.ext, b.private)
+		}
+		n := maxCoreSize + tokenLen(b.variant...) + tokenLen(b.ext...)
+		buf := make([]byte, n)
+		p := t.genCoreBytes(buf)
+		t.pVariant = byte(p)
+		p += appendTokens(buf[p:], b.variant...)
+		t.pExt = uint16(p)
+		p += appendTokens(buf[p:], b.ext...)
+		t.str = string(buf[:p])
+	} else if b.private != "" {
+		t.str = b.private
+		t.remakeString()
+	}
+	return
+}
+
+type builder struct {
+	tag Tag
+
+	private string // the x extension
+	ext     []string
+	variant []string
+
+	err error
+}
+
+func (b *builder) addExt(e string) {
+	if e == "" {
+	} else if e[0] == 'x' {
+		b.private = e
+	} else {
+		b.ext = append(b.ext, e)
+	}
+}
+
+var errInvalidArgument = errors.New("invalid Extension or Variant")
+
+func (b *builder) update(part ...interface{}) (err error) {
+	replace := func(l *[]string, s string, eq func(a, b string) bool) bool {
+		if s == "" {
+			b.err = errInvalidArgument
+			return true
+		}
+		for i, v := range *l {
+			if eq(v, s) {
+				(*l)[i] = s
+				return true
+			}
+		}
+		return false
+	}
+	for _, x := range part {
+		switch v := x.(type) {
+		case Tag:
+			b.tag.lang = v.lang
+			b.tag.region = v.region
+			b.tag.script = v.script
+			if v.str != "" {
+				b.variant = nil
+				for x, s := "", v.str[v.pVariant:v.pExt]; s != ""; {
+					x, s = nextToken(s)
+					b.variant = append(b.variant, x)
+				}
+				b.ext, b.private = nil, ""
+				for i, e := int(v.pExt), ""; i < len(v.str); {
+					i, e = getExtension(v.str, i)
+					b.addExt(e)
+				}
+			}
+		case Base:
+			b.tag.lang = v.langID
+		case Script:
+			b.tag.script = v.scriptID
+		case Region:
+			b.tag.region = v.regionID
+		case Variant:
+			if !replace(&b.variant, v.variant, func(a, b string) bool { return a == b }) {
+				b.variant = append(b.variant, v.variant)
+			}
+		case Extension:
+			if !replace(&b.ext, v.s, func(a, b string) bool { return a[0] == b[0] }) {
+				b.addExt(v.s)
+			}
+		case []Variant:
+			b.variant = nil
+			for _, x := range v {
+				b.update(x)
+			}
+		case []Extension:
+			b.ext, b.private = nil, ""
+			for _, e := range v {
+				b.update(e)
+			}
+		// TODO: support parsing of raw strings based on morphology or just extensions?
+		case error:
+			err = v
+		}
+	}
+	return
+}
+
+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 sortVariant []string
+
+func (s sortVariant) Len() int {
+	return len(s)
+}
+
+func (s sortVariant) Swap(i, j int) {
+	s[j], s[i] = s[i], s[j]
+}
+
+func (s sortVariant) Less(i, j int) bool {
+	return variantIndex[s[i]] < variantIndex[s[j]]
+}
+
+func findExt(list []string, x byte) int {
+	for i, e := range list {
+		if e[0] == x {
+			return i
+		}
+	}
+	return -1
+}
+
+// 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)
+}
+
+var errInvalidWeight = errors.New("ParseAcceptLanguage: invalid weight")
+
+// ParseAcceptLanguage parses the contents of a Accept-Language header as
+// defined in http://www.ietf.org/rfc/rfc2616.txt and returns a list of Tags and
+// a list of corresponding quality weights. It is more permissive than RFC 2616
+// and may return non-nil slices even if the input is not valid.
+// The Tags will be sorted by highest weight first and then by first occurrence.
+// Tags with a weight of zero will be dropped. An error will be returned if the
+// input could not be parsed.
+func ParseAcceptLanguage(s string) (tag []Tag, q []float32, err error) {
+	var entry string
+	for s != "" {
+		if entry, s = split(s, ','); entry == "" {
+			continue
+		}
+
+		entry, weight := split(entry, ';')
+
+		// Scan the language.
+		t, err := Parse(entry)
+		if err != nil {
+			id, ok := acceptFallback[entry]
+			if !ok {
+				return nil, nil, err
+			}
+			t = Tag{lang: id}
+		}
+
+		// Scan the optional weight.
+		w := 1.0
+		if weight != "" {
+			weight = consume(weight, 'q')
+			weight = consume(weight, '=')
+			// consume returns the empty string when a token could not be
+			// consumed, resulting in an error for ParseFloat.
+			if w, err = strconv.ParseFloat(weight, 32); err != nil {
+				return nil, nil, errInvalidWeight
+			}
+			// Drop tags with a quality weight of 0.
+			if w <= 0 {
+				continue
+			}
+		}
+
+		tag = append(tag, t)
+		q = append(q, float32(w))
+	}
+	sortStable(&tagSort{tag, q})
+	return tag, q, nil
+}
+
+// consume removes a leading token c from s and returns the result or the empty
+// string if there is no such token.
+func consume(s string, c byte) string {
+	if s == "" || s[0] != c {
+		return ""
+	}
+	return strings.TrimSpace(s[1:])
+}
+
+func split(s string, c byte) (head, tail string) {
+	if i := strings.IndexByte(s, c); i >= 0 {
+		return strings.TrimSpace(s[:i]), strings.TrimSpace(s[i+1:])
+	}
+	return strings.TrimSpace(s), ""
+}
+
+// Add hack mapping to deal with a small number of cases that that occur
+// in Accept-Language (with reasonable frequency).
+var acceptFallback = map[string]langID{
+	"english": _en,
+	"deutsch": _de,
+	"italian": _it,
+	"french":  _fr,
+	"*":       _mul, // defined in the spec to match all languages.
+}
+
+type tagSort struct {
+	tag []Tag
+	q   []float32
+}
+
+func (s *tagSort) Len() int {
+	return len(s.q)
+}
+
+func (s *tagSort) Less(i, j int) bool {
+	return s.q[i] > s.q[j]
+}
+
+func (s *tagSort) Swap(i, j int) {
+	s.tag[i], s.tag[j] = s.tag[j], s.tag[i]
+	s.q[i], s.q[j] = s.q[j], s.q[i]
+}

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

@@ -0,0 +1,143 @@
+// 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
+
+// TODO: Various sets of commonly use tags and regions.
+
+// 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
+}
+
+// MustParse is like Parse, but panics if the given BCP 47 tag cannot be parsed.
+// It simplifies safe initialization of Tag values.
+func (c CanonType) MustParse(s string) Tag {
+	t, err := c.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) Base {
+	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
+}
+
+var (
+	und = Tag{}
+
+	Und Tag = Tag{}
+
+	Afrikaans            Tag = Tag{lang: _af}                //  af
+	Amharic              Tag = Tag{lang: _am}                //  am
+	Arabic               Tag = Tag{lang: _ar}                //  ar
+	ModernStandardArabic Tag = Tag{lang: _ar, region: _001}  //  ar-001
+	Azerbaijani          Tag = Tag{lang: _az}                //  az
+	Bulgarian            Tag = Tag{lang: _bg}                //  bg
+	Bengali              Tag = Tag{lang: _bn}                //  bn
+	Catalan              Tag = Tag{lang: _ca}                //  ca
+	Czech                Tag = Tag{lang: _cs}                //  cs
+	Danish               Tag = Tag{lang: _da}                //  da
+	German               Tag = Tag{lang: _de}                //  de
+	Greek                Tag = Tag{lang: _el}                //  el
+	English              Tag = Tag{lang: _en}                //  en
+	AmericanEnglish      Tag = Tag{lang: _en, region: _US}   //  en-US
+	BritishEnglish       Tag = Tag{lang: _en, region: _GB}   //  en-GB
+	Spanish              Tag = Tag{lang: _es}                //  es
+	EuropeanSpanish      Tag = Tag{lang: _es, region: _ES}   //  es-ES
+	LatinAmericanSpanish Tag = Tag{lang: _es, region: _419}  //  es-419
+	Estonian             Tag = Tag{lang: _et}                //  et
+	Persian              Tag = Tag{lang: _fa}                //  fa
+	Finnish              Tag = Tag{lang: _fi}                //  fi
+	Filipino             Tag = Tag{lang: _fil}               //  fil
+	French               Tag = Tag{lang: _fr}                //  fr
+	CanadianFrench       Tag = Tag{lang: _fr, region: _CA}   //  fr-CA
+	Gujarati             Tag = Tag{lang: _gu}                //  gu
+	Hebrew               Tag = Tag{lang: _he}                //  he
+	Hindi                Tag = Tag{lang: _hi}                //  hi
+	Croatian             Tag = Tag{lang: _hr}                //  hr
+	Hungarian            Tag = Tag{lang: _hu}                //  hu
+	Armenian             Tag = Tag{lang: _hy}                //  hy
+	Indonesian           Tag = Tag{lang: _id}                //  id
+	Icelandic            Tag = Tag{lang: _is}                //  is
+	Italian              Tag = Tag{lang: _it}                //  it
+	Japanese             Tag = Tag{lang: _ja}                //  ja
+	Georgian             Tag = Tag{lang: _ka}                //  ka
+	Kazakh               Tag = Tag{lang: _kk}                //  kk
+	Khmer                Tag = Tag{lang: _km}                //  km
+	Kannada              Tag = Tag{lang: _kn}                //  kn
+	Korean               Tag = Tag{lang: _ko}                //  ko
+	Kirghiz              Tag = Tag{lang: _ky}                //  ky
+	Lao                  Tag = Tag{lang: _lo}                //  lo
+	Lithuanian           Tag = Tag{lang: _lt}                //  lt
+	Latvian              Tag = Tag{lang: _lv}                //  lv
+	Macedonian           Tag = Tag{lang: _mk}                //  mk
+	Malayalam            Tag = Tag{lang: _ml}                //  ml
+	Mongolian            Tag = Tag{lang: _mn}                //  mn
+	Marathi              Tag = Tag{lang: _mr}                //  mr
+	Malay                Tag = Tag{lang: _ms}                //  ms
+	Burmese              Tag = Tag{lang: _my}                //  my
+	Nepali               Tag = Tag{lang: _ne}                //  ne
+	Dutch                Tag = Tag{lang: _nl}                //  nl
+	Norwegian            Tag = Tag{lang: _no}                //  no
+	Punjabi              Tag = Tag{lang: _pa}                //  pa
+	Polish               Tag = Tag{lang: _pl}                //  pl
+	Portuguese           Tag = Tag{lang: _pt}                //  pt
+	BrazilianPortuguese  Tag = Tag{lang: _pt, region: _BR}   //  pt-BR
+	EuropeanPortuguese   Tag = Tag{lang: _pt, region: _PT}   //  pt-PT
+	Romanian             Tag = Tag{lang: _ro}                //  ro
+	Russian              Tag = Tag{lang: _ru}                //  ru
+	Sinhala              Tag = Tag{lang: _si}                //  si
+	Slovak               Tag = Tag{lang: _sk}                //  sk
+	Slovenian            Tag = Tag{lang: _sl}                //  sl
+	Albanian             Tag = Tag{lang: _sq}                //  sq
+	Serbian              Tag = Tag{lang: _sr}                //  sr
+	SerbianLatin         Tag = Tag{lang: _sr, script: _Latn} //  sr-Latn
+	Swedish              Tag = Tag{lang: _sv}                //  sv
+	Swahili              Tag = Tag{lang: _sw}                //  sw
+	Tamil                Tag = Tag{lang: _ta}                //  ta
+	Telugu               Tag = Tag{lang: _te}                //  te
+	Thai                 Tag = Tag{lang: _th}                //  th
+	Turkish              Tag = Tag{lang: _tr}                //  tr
+	Ukrainian            Tag = Tag{lang: _uk}                //  uk
+	Urdu                 Tag = Tag{lang: _ur}                //  ur
+	Uzbek                Tag = Tag{lang: _uz}                //  uz
+	Vietnamese           Tag = Tag{lang: _vi}                //  vi
+	Chinese              Tag = Tag{lang: _zh}                //  zh
+	SimplifiedChinese    Tag = Tag{lang: _zh, script: _Hans} //  zh-Hans
+	TraditionalChinese   Tag = Tag{lang: _zh, script: _Hant} //  zh-Hant
+	Zulu                 Tag = Tag{lang: _zu}                //  zu
+)

vendor/golang.org/x/text/runes/cond.go

@@ -0,0 +1,187 @@
+// 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 runes
+
+import (
+	"unicode/utf8"
+
+	"golang.org/x/text/transform"
+)
+
+// Note: below we pass invalid UTF-8 to the tIn and tNotIn transformers as is.
+// This is done for various reasons:
+// - To retain the semantics of the Nop transformer: if input is passed to a Nop
+//   one would expect it to be unchanged.
+// - It would be very expensive to pass a converted RuneError to a transformer:
+//   a transformer might need more source bytes after RuneError, meaning that
+//   the only way to pass it safely is to create a new buffer and manage the
+//   intermingling of RuneErrors and normal input.
+// - Many transformers leave ill-formed UTF-8 as is, so this is not
+//   inconsistent. Generally ill-formed UTF-8 is only replaced if it is a
+//   logical consequence of the operation (as for Map) or if it otherwise would
+//   pose security concerns (as for Remove).
+// - An alternative would be to return an error on ill-formed UTF-8, but this
+//   would be inconsistent with other operations.
+
+// If returns a transformer that applies tIn to consecutive runes for which
+// s.Contains(r) and tNotIn to consecutive runes for which !s.Contains(r). Reset
+// is called on tIn and tNotIn at the start of each run. A Nop transformer will
+// substitute a nil value passed to tIn or tNotIn. Invalid UTF-8 is translated
+// to RuneError to determine which transformer to apply, but is passed as is to
+// the respective transformer.
+func If(s Set, tIn, tNotIn transform.Transformer) Transformer {
+	if tIn == nil && tNotIn == nil {
+		return Transformer{transform.Nop}
+	}
+	if tIn == nil {
+		tIn = transform.Nop
+	}
+	if tNotIn == nil {
+		tNotIn = transform.Nop
+	}
+	sIn, ok := tIn.(transform.SpanningTransformer)
+	if !ok {
+		sIn = dummySpan{tIn}
+	}
+	sNotIn, ok := tNotIn.(transform.SpanningTransformer)
+	if !ok {
+		sNotIn = dummySpan{tNotIn}
+	}
+
+	a := &cond{
+		tIn:    sIn,
+		tNotIn: sNotIn,
+		f:      s.Contains,
+	}
+	a.Reset()
+	return Transformer{a}
+}
+
+type dummySpan struct{ transform.Transformer }
+
+func (d dummySpan) Span(src []byte, atEOF bool) (n int, err error) {
+	return 0, transform.ErrEndOfSpan
+}
+
+type cond struct {
+	tIn, tNotIn transform.SpanningTransformer
+	f           func(rune) bool
+	check       func(rune) bool               // current check to perform
+	t           transform.SpanningTransformer // current transformer to use
+}
+
+// Reset implements transform.Transformer.
+func (t *cond) Reset() {
+	t.check = t.is
+	t.t = t.tIn
+	t.t.Reset() // notIn will be reset on first usage.
+}
+
+func (t *cond) is(r rune) bool {
+	if t.f(r) {
+		return true
+	}
+	t.check = t.isNot
+	t.t = t.tNotIn
+	t.tNotIn.Reset()
+	return false
+}
+
+func (t *cond) isNot(r rune) bool {
+	if !t.f(r) {
+		return true
+	}
+	t.check = t.is
+	t.t = t.tIn
+	t.tIn.Reset()
+	return false
+}
+
+// This implementation of Span doesn't help all too much, but it needs to be
+// there to satisfy this package's Transformer interface.
+// TODO: there are certainly room for improvements, though. For example, if
+// t.t == transform.Nop (which will a common occurrence) it will save a bundle
+// to special-case that loop.
+func (t *cond) Span(src []byte, atEOF bool) (n int, err error) {
+	p := 0
+	for n < len(src) && err == nil {
+		// Don't process too much at a time as the Spanner that will be
+		// called on this block may terminate early.
+		const maxChunk = 4096
+		max := len(src)
+		if v := n + maxChunk; v < max {
+			max = v
+		}
+		atEnd := false
+		size := 0
+		current := t.t
+		for ; p < max; p += size {
+			r := rune(src[p])
+			if r < utf8.RuneSelf {
+				size = 1
+			} else if r, size = utf8.DecodeRune(src[p:]); size == 1 {
+				if !atEOF && !utf8.FullRune(src[p:]) {
+					err = transform.ErrShortSrc
+					break
+				}
+			}
+			if !t.check(r) {
+				// The next rune will be the start of a new run.
+				atEnd = true
+				break
+			}
+		}
+		n2, err2 := current.Span(src[n:p], atEnd || (atEOF && p == len(src)))
+		n += n2
+		if err2 != nil {
+			return n, err2
+		}
+		// At this point either err != nil or t.check will pass for the rune at p.
+		p = n + size
+	}
+	return n, err
+}
+
+func (t *cond) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	p := 0
+	for nSrc < len(src) && err == nil {
+		// Don't process too much at a time, as the work might be wasted if the
+		// destination buffer isn't large enough to hold the result or a
+		// transform returns an error early.
+		const maxChunk = 4096
+		max := len(src)
+		if n := nSrc + maxChunk; n < len(src) {
+			max = n
+		}
+		atEnd := false
+		size := 0
+		current := t.t
+		for ; p < max; p += size {
+			r := rune(src[p])
+			if r < utf8.RuneSelf {
+				size = 1
+			} else if r, size = utf8.DecodeRune(src[p:]); size == 1 {
+				if !atEOF && !utf8.FullRune(src[p:]) {
+					err = transform.ErrShortSrc
+					break
+				}
+			}
+			if !t.check(r) {
+				// The next rune will be the start of a new run.
+				atEnd = true
+				break
+			}
+		}
+		nDst2, nSrc2, err2 := current.Transform(dst[nDst:], src[nSrc:p], atEnd || (atEOF && p == len(src)))
+		nDst += nDst2
+		nSrc += nSrc2
+		if err2 != nil {
+			return nDst, nSrc, err2
+		}
+		// At this point either err != nil or t.check will pass for the rune at p.
+		p = nSrc + size
+	}
+	return nDst, nSrc, err
+}

vendor/golang.org/x/text/runes/runes.go

@@ -0,0 +1,355 @@
+// 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 runes provide transforms for UTF-8 encoded text.
+package runes // import "golang.org/x/text/runes"
+
+import (
+	"unicode"
+	"unicode/utf8"
+
+	"golang.org/x/text/transform"
+)
+
+// A Set is a collection of runes.
+type Set interface {
+	// Contains returns true if r is contained in the set.
+	Contains(r rune) bool
+}
+
+type setFunc func(rune) bool
+
+func (s setFunc) Contains(r rune) bool {
+	return s(r)
+}
+
+// Note: using funcs here instead of wrapping types result in cleaner
+// documentation and a smaller API.
+
+// In creates a Set with a Contains method that returns true for all runes in
+// the given RangeTable.
+func In(rt *unicode.RangeTable) Set {
+	return setFunc(func(r rune) bool { return unicode.Is(rt, r) })
+}
+
+// In creates a Set with a Contains method that returns true for all runes not
+// in the given RangeTable.
+func NotIn(rt *unicode.RangeTable) Set {
+	return setFunc(func(r rune) bool { return !unicode.Is(rt, r) })
+}
+
+// Predicate creates a Set with a Contains method that returns f(r).
+func Predicate(f func(rune) bool) Set {
+	return setFunc(f)
+}
+
+// Transformer implements the transform.Transformer interface.
+type Transformer struct {
+	t transform.SpanningTransformer
+}
+
+func (t Transformer) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	return t.t.Transform(dst, src, atEOF)
+}
+
+func (t Transformer) Span(b []byte, atEOF bool) (n int, err error) {
+	return t.t.Span(b, atEOF)
+}
+
+func (t Transformer) Reset() { t.t.Reset() }
+
+// Bytes returns a new byte slice with the result of converting b using t.  It
+// calls Reset on t. It returns nil if any error was found. This can only happen
+// if an error-producing Transformer is passed to If.
+func (t Transformer) Bytes(b []byte) []byte {
+	b, _, err := transform.Bytes(t, b)
+	if err != nil {
+		return nil
+	}
+	return b
+}
+
+// String returns a string with the result of converting s using t. It calls
+// Reset on t. It returns the empty string if any error was found. This can only
+// happen if an error-producing Transformer is passed to If.
+func (t Transformer) String(s string) string {
+	s, _, err := transform.String(t, s)
+	if err != nil {
+		return ""
+	}
+	return s
+}
+
+// TODO:
+// - Copy: copying strings and bytes in whole-rune units.
+// - Validation (maybe)
+// - Well-formed-ness (maybe)
+
+const runeErrorString = string(utf8.RuneError)
+
+// Remove returns a Transformer that removes runes r for which s.Contains(r).
+// Illegal input bytes are replaced by RuneError before being passed to f.
+func Remove(s Set) Transformer {
+	if f, ok := s.(setFunc); ok {
+		// This little trick cuts the running time of BenchmarkRemove for sets
+		// created by Predicate roughly in half.
+		// TODO: special-case RangeTables as well.
+		return Transformer{remove(f)}
+	}
+	return Transformer{remove(s.Contains)}
+}
+
+// TODO: remove transform.RemoveFunc.
+
+type remove func(r rune) bool
+
+func (remove) Reset() {}
+
+// Span implements transform.Spanner.
+func (t remove) Span(src []byte, atEOF bool) (n int, err error) {
+	for r, size := rune(0), 0; n < len(src); {
+		if r = rune(src[n]); r < utf8.RuneSelf {
+			size = 1
+		} else if r, size = utf8.DecodeRune(src[n:]); size == 1 {
+			// Invalid rune.
+			if !atEOF && !utf8.FullRune(src[n:]) {
+				err = transform.ErrShortSrc
+			} else {
+				err = transform.ErrEndOfSpan
+			}
+			break
+		}
+		if t(r) {
+			err = transform.ErrEndOfSpan
+			break
+		}
+		n += size
+	}
+	return
+}
+
+// Transform implements transform.Transformer.
+func (t remove) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	for r, size := rune(0), 0; nSrc < len(src); {
+		if r = rune(src[nSrc]); r < utf8.RuneSelf {
+			size = 1
+		} else if r, size = utf8.DecodeRune(src[nSrc:]); size == 1 {
+			// Invalid rune.
+			if !atEOF && !utf8.FullRune(src[nSrc:]) {
+				err = transform.ErrShortSrc
+				break
+			}
+			// We replace illegal bytes with RuneError. Not doing so might
+			// otherwise turn a sequence of invalid UTF-8 into valid UTF-8.
+			// The resulting byte sequence may subsequently contain runes
+			// for which t(r) is true that were passed unnoticed.
+			if !t(utf8.RuneError) {
+				if nDst+3 > len(dst) {
+					err = transform.ErrShortDst
+					break
+				}
+				dst[nDst+0] = runeErrorString[0]
+				dst[nDst+1] = runeErrorString[1]
+				dst[nDst+2] = runeErrorString[2]
+				nDst += 3
+			}
+			nSrc++
+			continue
+		}
+		if t(r) {
+			nSrc += size
+			continue
+		}
+		if nDst+size > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		for i := 0; i < size; i++ {
+			dst[nDst] = src[nSrc]
+			nDst++
+			nSrc++
+		}
+	}
+	return
+}
+
+// Map returns a Transformer that maps the runes in the input using the given
+// mapping. Illegal bytes in the input are converted to utf8.RuneError before
+// being passed to the mapping func.
+func Map(mapping func(rune) rune) Transformer {
+	return Transformer{mapper(mapping)}
+}
+
+type mapper func(rune) rune
+
+func (mapper) Reset() {}
+
+// Span implements transform.Spanner.
+func (t mapper) Span(src []byte, atEOF bool) (n int, err error) {
+	for r, size := rune(0), 0; n < len(src); n += size {
+		if r = rune(src[n]); r < utf8.RuneSelf {
+			size = 1
+		} else if r, size = utf8.DecodeRune(src[n:]); size == 1 {
+			// Invalid rune.
+			if !atEOF && !utf8.FullRune(src[n:]) {
+				err = transform.ErrShortSrc
+			} else {
+				err = transform.ErrEndOfSpan
+			}
+			break
+		}
+		if t(r) != r {
+			err = transform.ErrEndOfSpan
+			break
+		}
+	}
+	return n, err
+}
+
+// Transform implements transform.Transformer.
+func (t mapper) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	var replacement rune
+	var b [utf8.UTFMax]byte
+
+	for r, size := rune(0), 0; nSrc < len(src); {
+		if r = rune(src[nSrc]); r < utf8.RuneSelf {
+			if replacement = t(r); replacement < utf8.RuneSelf {
+				if nDst == len(dst) {
+					err = transform.ErrShortDst
+					break
+				}
+				dst[nDst] = byte(replacement)
+				nDst++
+				nSrc++
+				continue
+			}
+			size = 1
+		} else if r, size = utf8.DecodeRune(src[nSrc:]); size == 1 {
+			// Invalid rune.
+			if !atEOF && !utf8.FullRune(src[nSrc:]) {
+				err = transform.ErrShortSrc
+				break
+			}
+
+			if replacement = t(utf8.RuneError); replacement == utf8.RuneError {
+				if nDst+3 > len(dst) {
+					err = transform.ErrShortDst
+					break
+				}
+				dst[nDst+0] = runeErrorString[0]
+				dst[nDst+1] = runeErrorString[1]
+				dst[nDst+2] = runeErrorString[2]
+				nDst += 3
+				nSrc++
+				continue
+			}
+		} else if replacement = t(r); replacement == r {
+			if nDst+size > len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			for i := 0; i < size; i++ {
+				dst[nDst] = src[nSrc]
+				nDst++
+				nSrc++
+			}
+			continue
+		}
+
+		n := utf8.EncodeRune(b[:], replacement)
+
+		if nDst+n > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		for i := 0; i < n; i++ {
+			dst[nDst] = b[i]
+			nDst++
+		}
+		nSrc += size
+	}
+	return
+}
+
+// ReplaceIllFormed returns a transformer that replaces all input bytes that are
+// not part of a well-formed UTF-8 code sequence with utf8.RuneError.
+func ReplaceIllFormed() Transformer {
+	return Transformer{&replaceIllFormed{}}
+}
+
+type replaceIllFormed struct{ transform.NopResetter }
+
+func (t replaceIllFormed) Span(src []byte, atEOF bool) (n int, err error) {
+	for n < len(src) {
+		// ASCII fast path.
+		if src[n] < utf8.RuneSelf {
+			n++
+			continue
+		}
+
+		r, size := utf8.DecodeRune(src[n:])
+
+		// Look for a valid non-ASCII rune.
+		if r != utf8.RuneError || size != 1 {
+			n += size
+			continue
+		}
+
+		// Look for short source data.
+		if !atEOF && !utf8.FullRune(src[n:]) {
+			err = transform.ErrShortSrc
+			break
+		}
+
+		// We have an invalid rune.
+		err = transform.ErrEndOfSpan
+		break
+	}
+	return n, err
+}
+
+func (t replaceIllFormed) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	for nSrc < len(src) {
+		// ASCII fast path.
+		if r := src[nSrc]; r < utf8.RuneSelf {
+			if nDst == len(dst) {
+				err = transform.ErrShortDst
+				break
+			}
+			dst[nDst] = r
+			nDst++
+			nSrc++
+			continue
+		}
+
+		// Look for a valid non-ASCII rune.
+		if _, size := utf8.DecodeRune(src[nSrc:]); size != 1 {
+			if size != copy(dst[nDst:], src[nSrc:nSrc+size]) {
+				err = transform.ErrShortDst
+				break
+			}
+			nDst += size
+			nSrc += size
+			continue
+		}
+
+		// Look for short source data.
+		if !atEOF && !utf8.FullRune(src[nSrc:]) {
+			err = transform.ErrShortSrc
+			break
+		}
+
+		// We have an invalid rune.
+		if nDst+3 > len(dst) {
+			err = transform.ErrShortDst
+			break
+		}
+		dst[nDst+0] = runeErrorString[0]
+		dst[nDst+1] = runeErrorString[1]
+		dst[nDst+2] = runeErrorString[2]
+		nDst += 3
+		nSrc++
+	}
+	return nDst, nSrc, err
+}

vendor/golang.org/x/text/transform/transform.go

@@ -0,0 +1,705 @@
+// 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 transform provides reader and writer wrappers that transform the
+// bytes passing through as well as various transformations. Example
+// transformations provided by other packages include normalization and
+// conversion between character sets.
+package transform // import "golang.org/x/text/transform"
+
+import (
+	"bytes"
+	"errors"
+	"io"
+	"unicode/utf8"
+)
+
+var (
+	// ErrShortDst means that the destination buffer was too short to
+	// receive all of the transformed bytes.
+	ErrShortDst = errors.New("transform: short destination buffer")
+
+	// ErrShortSrc means that the source buffer has insufficient data to
+	// complete the transformation.
+	ErrShortSrc = errors.New("transform: short source buffer")
+
+	// ErrEndOfSpan means that the input and output (the transformed input)
+	// are not identical.
+	ErrEndOfSpan = errors.New("transform: input and output are not identical")
+
+	// errInconsistentByteCount means that Transform returned success (nil
+	// error) but also returned nSrc inconsistent with the src argument.
+	errInconsistentByteCount = errors.New("transform: inconsistent byte count returned")
+
+	// errShortInternal means that an internal buffer is not large enough
+	// to make progress and the Transform operation must be aborted.
+	errShortInternal = errors.New("transform: short internal buffer")
+)
+
+// Transformer transforms bytes.
+type Transformer interface {
+	// Transform writes to dst the transformed bytes read from src, and
+	// returns the number of dst bytes written and src bytes read. The
+	// atEOF argument tells whether src represents the last bytes of the
+	// input.
+	//
+	// Callers should always process the nDst bytes produced and account
+	// for the nSrc bytes consumed before considering the error err.
+	//
+	// A nil error means that all of the transformed bytes (whether freshly
+	// transformed from src or left over from previous Transform calls)
+	// were written to dst. A nil error can be returned regardless of
+	// whether atEOF is true. If err is nil then nSrc must equal len(src);
+	// the converse is not necessarily true.
+	//
+	// ErrShortDst means that dst was too short to receive all of the
+	// transformed bytes. ErrShortSrc means that src had insufficient data
+	// to complete the transformation. If both conditions apply, then
+	// either error may be returned. Other than the error conditions listed
+	// here, implementations are free to report other errors that arise.
+	Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error)
+
+	// Reset resets the state and allows a Transformer to be reused.
+	Reset()
+}
+
+// SpanningTransformer extends the Transformer interface with a Span method
+// that determines how much of the input already conforms to the Transformer.
+type SpanningTransformer interface {
+	Transformer
+
+	// Span returns a position in src such that transforming src[:n] results in
+	// identical output src[:n] for these bytes. It does not necessarily return
+	// the largest such n. The atEOF argument tells whether src represents the
+	// last bytes of the input.
+	//
+	// Callers should always account for the n bytes consumed before
+	// considering the error err.
+	//
+	// A nil error means that all input bytes are known to be identical to the
+	// output produced by the Transformer. A nil error can be be returned
+	// regardless of whether atEOF is true. If err is nil, then then n must
+	// equal len(src); the converse is not necessarily true.
+	//
+	// ErrEndOfSpan means that the Transformer output may differ from the
+	// input after n bytes. Note that n may be len(src), meaning that the output
+	// would contain additional bytes after otherwise identical output.
+	// ErrShortSrc means that src had insufficient data to determine whether the
+	// remaining bytes would change. Other than the error conditions listed
+	// here, implementations are free to report other errors that arise.
+	//
+	// Calling Span can modify the Transformer state as a side effect. In
+	// effect, it does the transformation just as calling Transform would, only
+	// without copying to a destination buffer and only up to a point it can
+	// determine the input and output bytes are the same. This is obviously more
+	// limited than calling Transform, but can be more efficient in terms of
+	// copying and allocating buffers. Calls to Span and Transform may be
+	// interleaved.
+	Span(src []byte, atEOF bool) (n int, err error)
+}
+
+// NopResetter can be embedded by implementations of Transformer to add a nop
+// Reset method.
+type NopResetter struct{}
+
+// Reset implements the Reset method of the Transformer interface.
+func (NopResetter) Reset() {}
+
+// Reader wraps another io.Reader by transforming the bytes read.
+type Reader struct {
+	r   io.Reader
+	t   Transformer
+	err error
+
+	// dst[dst0:dst1] contains bytes that have been transformed by t but
+	// not yet copied out via Read.
+	dst        []byte
+	dst0, dst1 int
+
+	// src[src0:src1] contains bytes that have been read from r but not
+	// yet transformed through t.
+	src        []byte
+	src0, src1 int
+
+	// transformComplete is whether the transformation is complete,
+	// regardless of whether or not it was successful.
+	transformComplete bool
+}
+
+const defaultBufSize = 4096
+
+// NewReader returns a new Reader that wraps r by transforming the bytes read
+// via t. It calls Reset on t.
+func NewReader(r io.Reader, t Transformer) *Reader {
+	t.Reset()
+	return &Reader{
+		r:   r,
+		t:   t,
+		dst: make([]byte, defaultBufSize),
+		src: make([]byte, defaultBufSize),
+	}
+}
+
+// Read implements the io.Reader interface.
+func (r *Reader) Read(p []byte) (int, error) {
+	n, err := 0, error(nil)
+	for {
+		// Copy out any transformed bytes and return the final error if we are done.
+		if r.dst0 != r.dst1 {
+			n = copy(p, r.dst[r.dst0:r.dst1])
+			r.dst0 += n
+			if r.dst0 == r.dst1 && r.transformComplete {
+				return n, r.err
+			}
+			return n, nil
+		} else if r.transformComplete {
+			return 0, r.err
+		}
+
+		// Try to transform some source bytes, or to flush the transformer if we
+		// are out of source bytes. We do this even if r.r.Read returned an error.
+		// As the io.Reader documentation says, "process the n > 0 bytes returned
+		// before considering the error".
+		if r.src0 != r.src1 || r.err != nil {
+			r.dst0 = 0
+			r.dst1, n, err = r.t.Transform(r.dst, r.src[r.src0:r.src1], r.err == io.EOF)
+			r.src0 += n
+
+			switch {
+			case err == nil:
+				if r.src0 != r.src1 {
+					r.err = errInconsistentByteCount
+				}
+				// The Transform call was successful; we are complete if we
+				// cannot read more bytes into src.
+				r.transformComplete = r.err != nil
+				continue
+			case err == ErrShortDst && (r.dst1 != 0 || n != 0):
+				// Make room in dst by copying out, and try again.
+				continue
+			case err == ErrShortSrc && r.src1-r.src0 != len(r.src) && r.err == nil:
+				// Read more bytes into src via the code below, and try again.
+			default:
+				r.transformComplete = true
+				// The reader error (r.err) takes precedence over the
+				// transformer error (err) unless r.err is nil or io.EOF.
+				if r.err == nil || r.err == io.EOF {
+					r.err = err
+				}
+				continue
+			}
+		}
+
+		// Move any untransformed source bytes to the start of the buffer
+		// and read more bytes.
+		if r.src0 != 0 {
+			r.src0, r.src1 = 0, copy(r.src, r.src[r.src0:r.src1])
+		}
+		n, r.err = r.r.Read(r.src[r.src1:])
+		r.src1 += n
+	}
+}
+
+// TODO: implement ReadByte (and ReadRune??).
+
+// Writer wraps another io.Writer by transforming the bytes read.
+// The user needs to call Close to flush unwritten bytes that may
+// be buffered.
+type Writer struct {
+	w   io.Writer
+	t   Transformer
+	dst []byte
+
+	// src[:n] contains bytes that have not yet passed through t.
+	src []byte
+	n   int
+}
+
+// NewWriter returns a new Writer that wraps w by transforming the bytes written
+// via t. It calls Reset on t.
+func NewWriter(w io.Writer, t Transformer) *Writer {
+	t.Reset()
+	return &Writer{
+		w:   w,
+		t:   t,
+		dst: make([]byte, defaultBufSize),
+		src: make([]byte, defaultBufSize),
+	}
+}
+
+// Write implements the io.Writer interface. If there are not enough
+// bytes available to complete a Transform, the bytes will be buffered
+// for the next write. Call Close to convert the remaining bytes.
+func (w *Writer) Write(data []byte) (n int, err error) {
+	src := data
+	if w.n > 0 {
+		// Append bytes from data to the last remainder.
+		// TODO: limit the amount copied on first try.
+		n = copy(w.src[w.n:], data)
+		w.n += n
+		src = w.src[:w.n]
+	}
+	for {
+		nDst, nSrc, err := w.t.Transform(w.dst, src, false)
+		if _, werr := w.w.Write(w.dst[:nDst]); werr != nil {
+			return n, werr
+		}
+		src = src[nSrc:]
+		if w.n == 0 {
+			n += nSrc
+		} else if len(src) <= n {
+			// Enough bytes from w.src have been consumed. We make src point
+			// to data instead to reduce the copying.
+			w.n = 0
+			n -= len(src)
+			src = data[n:]
+			if n < len(data) && (err == nil || err == ErrShortSrc) {
+				continue
+			}
+		}
+		switch err {
+		case ErrShortDst:
+			// This error is okay as long as we are making progress.
+			if nDst > 0 || nSrc > 0 {
+				continue
+			}
+		case ErrShortSrc:
+			if len(src) < len(w.src) {
+				m := copy(w.src, src)
+				// If w.n > 0, bytes from data were already copied to w.src and n
+				// was already set to the number of bytes consumed.
+				if w.n == 0 {
+					n += m
+				}
+				w.n = m
+				err = nil
+			} else if nDst > 0 || nSrc > 0 {
+				// Not enough buffer to store the remainder. Keep processing as
+				// long as there is progress. Without this case, transforms that
+				// require a lookahead larger than the buffer may result in an
+				// error. This is not something one may expect to be common in
+				// practice, but it may occur when buffers are set to small
+				// sizes during testing.
+				continue
+			}
+		case nil:
+			if w.n > 0 {
+				err = errInconsistentByteCount
+			}
+		}
+		return n, err
+	}
+}
+
+// Close implements the io.Closer interface.
+func (w *Writer) Close() error {
+	src := w.src[:w.n]
+	for {
+		nDst, nSrc, err := w.t.Transform(w.dst, src, true)
+		if _, werr := w.w.Write(w.dst[:nDst]); werr != nil {
+			return werr
+		}
+		if err != ErrShortDst {
+			return err
+		}
+		src = src[nSrc:]
+	}
+}
+
+type nop struct{ NopResetter }
+
+func (nop) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	n := copy(dst, src)
+	if n < len(src) {
+		err = ErrShortDst
+	}
+	return n, n, err
+}
+
+func (nop) Span(src []byte, atEOF bool) (n int, err error) {
+	return len(src), nil
+}
+
+type discard struct{ NopResetter }
+
+func (discard) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	return 0, len(src), nil
+}
+
+var (
+	// Discard is a Transformer for which all Transform calls succeed
+	// by consuming all bytes and writing nothing.
+	Discard Transformer = discard{}
+
+	// Nop is a SpanningTransformer that copies src to dst.
+	Nop SpanningTransformer = nop{}
+)
+
+// chain is a sequence of links. A chain with N Transformers has N+1 links and
+// N+1 buffers. Of those N+1 buffers, the first and last are the src and dst
+// buffers given to chain.Transform and the middle N-1 buffers are intermediate
+// buffers owned by the chain. The i'th link transforms bytes from the i'th
+// buffer chain.link[i].b at read offset chain.link[i].p to the i+1'th buffer
+// chain.link[i+1].b at write offset chain.link[i+1].n, for i in [0, N).
+type chain struct {
+	link []link
+	err  error
+	// errStart is the index at which the error occurred plus 1. Processing
+	// errStart at this level at the next call to Transform. As long as
+	// errStart > 0, chain will not consume any more source bytes.
+	errStart int
+}
+
+func (c *chain) fatalError(errIndex int, err error) {
+	if i := errIndex + 1; i > c.errStart {
+		c.errStart = i
+		c.err = err
+	}
+}
+
+type link struct {
+	t Transformer
+	// b[p:n] holds the bytes to be transformed by t.
+	b []byte
+	p int
+	n int
+}
+
+func (l *link) src() []byte {
+	return l.b[l.p:l.n]
+}
+
+func (l *link) dst() []byte {
+	return l.b[l.n:]
+}
+
+// Chain returns a Transformer that applies t in sequence.
+func Chain(t ...Transformer) Transformer {
+	if len(t) == 0 {
+		return nop{}
+	}
+	c := &chain{link: make([]link, len(t)+1)}
+	for i, tt := range t {
+		c.link[i].t = tt
+	}
+	// Allocate intermediate buffers.
+	b := make([][defaultBufSize]byte, len(t)-1)
+	for i := range b {
+		c.link[i+1].b = b[i][:]
+	}
+	return c
+}
+
+// Reset resets the state of Chain. It calls Reset on all the Transformers.
+func (c *chain) Reset() {
+	for i, l := range c.link {
+		if l.t != nil {
+			l.t.Reset()
+		}
+		c.link[i].p, c.link[i].n = 0, 0
+	}
+}
+
+// TODO: make chain use Span (is going to be fun to implement!)
+
+// Transform applies the transformers of c in sequence.
+func (c *chain) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	// Set up src and dst in the chain.
+	srcL := &c.link[0]
+	dstL := &c.link[len(c.link)-1]
+	srcL.b, srcL.p, srcL.n = src, 0, len(src)
+	dstL.b, dstL.n = dst, 0
+	var lastFull, needProgress bool // for detecting progress
+
+	// i is the index of the next Transformer to apply, for i in [low, high].
+	// low is the lowest index for which c.link[low] may still produce bytes.
+	// high is the highest index for which c.link[high] has a Transformer.
+	// The error returned by Transform determines whether to increase or
+	// decrease i. We try to completely fill a buffer before converting it.
+	for low, i, high := c.errStart, c.errStart, len(c.link)-2; low <= i && i <= high; {
+		in, out := &c.link[i], &c.link[i+1]
+		nDst, nSrc, err0 := in.t.Transform(out.dst(), in.src(), atEOF && low == i)
+		out.n += nDst
+		in.p += nSrc
+		if i > 0 && in.p == in.n {
+			in.p, in.n = 0, 0
+		}
+		needProgress, lastFull = lastFull, false
+		switch err0 {
+		case ErrShortDst:
+			// Process the destination buffer next. Return if we are already
+			// at the high index.
+			if i == high {
+				return dstL.n, srcL.p, ErrShortDst
+			}
+			if out.n != 0 {
+				i++
+				// If the Transformer at the next index is not able to process any
+				// source bytes there is nothing that can be done to make progress
+				// and the bytes will remain unprocessed. lastFull is used to
+				// detect this and break out of the loop with a fatal error.
+				lastFull = true
+				continue
+			}
+			// The destination buffer was too small, but is completely empty.
+			// Return a fatal error as this transformation can never complete.
+			c.fatalError(i, errShortInternal)
+		case ErrShortSrc:
+			if i == 0 {
+				// Save ErrShortSrc in err. All other errors take precedence.
+				err = ErrShortSrc
+				break
+			}
+			// Source bytes were depleted before filling up the destination buffer.
+			// Verify we made some progress, move the remaining bytes to the errStart
+			// and try to get more source bytes.
+			if needProgress && nSrc == 0 || in.n-in.p == len(in.b) {
+				// There were not enough source bytes to proceed while the source
+				// buffer cannot hold any more bytes. Return a fatal error as this
+				// transformation can never complete.
+				c.fatalError(i, errShortInternal)
+				break
+			}
+			// in.b is an internal buffer and we can make progress.
+			in.p, in.n = 0, copy(in.b, in.src())
+			fallthrough
+		case nil:
+			// if i == low, we have depleted the bytes at index i or any lower levels.
+			// In that case we increase low and i. In all other cases we decrease i to
+			// fetch more bytes before proceeding to the next index.
+			if i > low {
+				i--
+				continue
+			}
+		default:
+			c.fatalError(i, err0)
+		}
+		// Exhausted level low or fatal error: increase low and continue
+		// to process the bytes accepted so far.
+		i++
+		low = i
+	}
+
+	// If c.errStart > 0, this means we found a fatal error.  We will clear
+	// all upstream buffers. At this point, no more progress can be made
+	// downstream, as Transform would have bailed while handling ErrShortDst.
+	if c.errStart > 0 {
+		for i := 1; i < c.errStart; i++ {
+			c.link[i].p, c.link[i].n = 0, 0
+		}
+		err, c.errStart, c.err = c.err, 0, nil
+	}
+	return dstL.n, srcL.p, err
+}
+
+// Deprecated: use runes.Remove instead.
+func RemoveFunc(f func(r rune) bool) Transformer {
+	return removeF(f)
+}
+
+type removeF func(r rune) bool
+
+func (removeF) Reset() {}
+
+// Transform implements the Transformer interface.
+func (t removeF) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
+	for r, sz := rune(0), 0; len(src) > 0; src = src[sz:] {
+
+		if r = rune(src[0]); r < utf8.RuneSelf {
+			sz = 1
+		} else {
+			r, sz = utf8.DecodeRune(src)
+
+			if sz == 1 {
+				// Invalid rune.
+				if !atEOF && !utf8.FullRune(src) {
+					err = ErrShortSrc
+					break
+				}
+				// We replace illegal bytes with RuneError. Not doing so might
+				// otherwise turn a sequence of invalid UTF-8 into valid UTF-8.
+				// The resulting byte sequence may subsequently contain runes
+				// for which t(r) is true that were passed unnoticed.
+				if !t(r) {
+					if nDst+3 > len(dst) {
+						err = ErrShortDst
+						break
+					}
+					nDst += copy(dst[nDst:], "\uFFFD")
+				}
+				nSrc++
+				continue
+			}
+		}
+
+		if !t(r) {
+			if nDst+sz > len(dst) {
+				err = ErrShortDst
+				break
+			}
+			nDst += copy(dst[nDst:], src[:sz])
+		}
+		nSrc += sz
+	}
+	return
+}
+
+// grow returns a new []byte that is longer than b, and copies the first n bytes
+// of b to the start of the new slice.
+func grow(b []byte, n int) []byte {
+	m := len(b)
+	if m <= 32 {
+		m = 64
+	} else if m <= 256 {
+		m *= 2
+	} else {
+		m += m >> 1
+	}
+	buf := make([]byte, m)
+	copy(buf, b[:n])
+	return buf
+}
+
+const initialBufSize = 128
+
+// String returns a string with the result of converting s[:n] using t, where
+// n <= len(s). If err == nil, n will be len(s). It calls Reset on t.
+func String(t Transformer, s string) (result string, n int, err error) {
+	t.Reset()
+	if s == "" {
+		// Fast path for the common case for empty input. Results in about a
+		// 86% reduction of running time for BenchmarkStringLowerEmpty.
+		if _, _, err := t.Transform(nil, nil, true); err == nil {
+			return "", 0, nil
+		}
+	}
+
+	// Allocate only once. Note that both dst and src escape when passed to
+	// Transform.
+	buf := [2 * initialBufSize]byte{}
+	dst := buf[:initialBufSize:initialBufSize]
+	src := buf[initialBufSize : 2*initialBufSize]
+
+	// The input string s is transformed in multiple chunks (starting with a
+	// chunk size of initialBufSize). nDst and nSrc are per-chunk (or
+	// per-Transform-call) indexes, pDst and pSrc are overall indexes.
+	nDst, nSrc := 0, 0
+	pDst, pSrc := 0, 0
+
+	// pPrefix is the length of a common prefix: the first pPrefix bytes of the
+	// result will equal the first pPrefix bytes of s. It is not guaranteed to
+	// be the largest such value, but if pPrefix, len(result) and len(s) are
+	// all equal after the final transform (i.e. calling Transform with atEOF
+	// being true returned nil error) then we don't need to allocate a new
+	// result string.
+	pPrefix := 0
+	for {
+		// Invariant: pDst == pPrefix && pSrc == pPrefix.
+
+		n := copy(src, s[pSrc:])
+		nDst, nSrc, err = t.Transform(dst, src[:n], pSrc+n == len(s))
+		pDst += nDst
+		pSrc += nSrc
+
+		// TODO:  let transformers implement an optional Spanner interface, akin
+		// to norm's QuickSpan. This would even allow us to avoid any allocation.
+		if !bytes.Equal(dst[:nDst], src[:nSrc]) {
+			break
+		}
+		pPrefix = pSrc
+		if err == ErrShortDst {
+			// A buffer can only be short if a transformer modifies its input.
+			break
+		} else if err == ErrShortSrc {
+			if nSrc == 0 {
+				// No progress was made.
+				break
+			}
+			// Equal so far and !atEOF, so continue checking.
+		} else if err != nil || pPrefix == len(s) {
+			return string(s[:pPrefix]), pPrefix, err
+		}
+	}
+	// Post-condition: pDst == pPrefix + nDst && pSrc == pPrefix + nSrc.
+
+	// We have transformed the first pSrc bytes of the input s to become pDst
+	// transformed bytes. Those transformed bytes are discontiguous: the first
+	// pPrefix of them equal s[:pPrefix] and the last nDst of them equal
+	// dst[:nDst]. We copy them around, into a new dst buffer if necessary, so
+	// that they become one contiguous slice: dst[:pDst].
+	if pPrefix != 0 {
+		newDst := dst
+		if pDst > len(newDst) {
+			newDst = make([]byte, len(s)+nDst-nSrc)
+		}
+		copy(newDst[pPrefix:pDst], dst[:nDst])
+		copy(newDst[:pPrefix], s[:pPrefix])
+		dst = newDst
+	}
+
+	// Prevent duplicate Transform calls with atEOF being true at the end of
+	// the input. Also return if we have an unrecoverable error.
+	if (err == nil && pSrc == len(s)) ||
+		(err != nil && err != ErrShortDst && err != ErrShortSrc) {
+		return string(dst[:pDst]), pSrc, err
+	}
+
+	// Transform the remaining input, growing dst and src buffers as necessary.
+	for {
+		n := copy(src, s[pSrc:])
+		nDst, nSrc, err := t.Transform(dst[pDst:], src[:n], pSrc+n == len(s))
+		pDst += nDst
+		pSrc += nSrc
+
+		// If we got ErrShortDst or ErrShortSrc, do not grow as long as we can
+		// make progress. This may avoid excessive allocations.
+		if err == ErrShortDst {
+			if nDst == 0 {
+				dst = grow(dst, pDst)
+			}
+		} else if err == ErrShortSrc {
+			if nSrc == 0 {
+				src = grow(src, 0)
+			}
+		} else if err != nil || pSrc == len(s) {
+			return string(dst[:pDst]), pSrc, err
+		}
+	}
+}
+
+// Bytes returns a new byte slice with the result of converting b[:n] using t,
+// where n <= len(b). If err == nil, n will be len(b). It calls Reset on t.
+func Bytes(t Transformer, b []byte) (result []byte, n int, err error) {
+	return doAppend(t, 0, make([]byte, len(b)), b)
+}
+
+// Append appends the result of converting src[:n] using t to dst, where
+// n <= len(src), If err == nil, n will be len(src). It calls Reset on t.
+func Append(t Transformer, dst, src []byte) (result []byte, n int, err error) {
+	if len(dst) == cap(dst) {
+		n := len(src) + len(dst) // It is okay for this to be 0.
+		b := make([]byte, n)
+		dst = b[:copy(b, dst)]
+	}
+	return doAppend(t, len(dst), dst[:cap(dst)], src)
+}
+
+func doAppend(t Transformer, pDst int, dst, src []byte) (result []byte, n int, err error) {
+	t.Reset()
+	pSrc := 0
+	for {
+		nDst, nSrc, err := t.Transform(dst[pDst:], src[pSrc:], true)
+		pDst += nDst
+		pSrc += nSrc
+		if err != ErrShortDst {
+			return dst[:pDst], pSrc, err
+		}
+
+		// Grow the destination buffer, but do not grow as long as we can make
+		// progress. This may avoid excessive allocations.
+		if nDst == 0 {
+			dst = grow(dst, pDst)
+		}
+	}
+}