update go-git to v4.12.0 - fixes #7248 (#7249)

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

@@ -0,0 +1,198 @@
+// 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 gen_trieval.go gen_ranges.go
+
+// Package bidi contains functionality for bidirectional text support.
+//
+// See https://www.unicode.org/reports/tr9.
+//
+// NOTE: UNDER CONSTRUCTION. This API may change in backwards incompatible ways
+// and without notice.
+package bidi // import "golang.org/x/text/unicode/bidi"
+
+// TODO:
+// The following functionality would not be hard to implement, but hinges on
+// the definition of a Segmenter interface. For now this is up to the user.
+// - Iterate over paragraphs
+// - Segmenter to iterate over runs directly from a given text.
+// Also:
+// - Transformer for reordering?
+// - Transformer (validator, really) for Bidi Rule.
+
+// This API tries to avoid dealing with embedding levels for now. Under the hood
+// these will be computed, but the question is to which extent the user should
+// know they exist. We should at some point allow the user to specify an
+// embedding hierarchy, though.
+
+// A Direction indicates the overall flow of text.
+type Direction int
+
+const (
+	// LeftToRight indicates the text contains no right-to-left characters and
+	// that either there are some left-to-right characters or the option
+	// DefaultDirection(LeftToRight) was passed.
+	LeftToRight Direction = iota
+
+	// RightToLeft indicates the text contains no left-to-right characters and
+	// that either there are some right-to-left characters or the option
+	// DefaultDirection(RightToLeft) was passed.
+	RightToLeft
+
+	// Mixed indicates text contains both left-to-right and right-to-left
+	// characters.
+	Mixed
+
+	// Neutral means that text contains no left-to-right and right-to-left
+	// characters and that no default direction has been set.
+	Neutral
+)
+
+type options struct{}
+
+// An Option is an option for Bidi processing.
+type Option func(*options)
+
+// ICU allows the user to define embedding levels. This may be used, for example,
+// to use hierarchical structure of markup languages to define embeddings.
+// The following option may be a way to expose this functionality in this API.
+// // LevelFunc sets a function that associates nesting levels with the given text.
+// // The levels function will be called with monotonically increasing values for p.
+// func LevelFunc(levels func(p int) int) Option {
+// 	panic("unimplemented")
+// }
+
+// DefaultDirection sets the default direction for a Paragraph. The direction is
+// overridden if the text contains directional characters.
+func DefaultDirection(d Direction) Option {
+	panic("unimplemented")
+}
+
+// A Paragraph holds a single Paragraph for Bidi processing.
+type Paragraph struct {
+	// buffers
+}
+
+// SetBytes configures p for the given paragraph text. It replaces text
+// previously set by SetBytes or SetString. If b contains a paragraph separator
+// it will only process the first paragraph and report the number of bytes
+// consumed from b including this separator. Error may be non-nil if options are
+// given.
+func (p *Paragraph) SetBytes(b []byte, opts ...Option) (n int, err error) {
+	panic("unimplemented")
+}
+
+// SetString configures p for the given paragraph text. It replaces text
+// previously set by SetBytes or SetString. If b contains a paragraph separator
+// it will only process the first paragraph and report the number of bytes
+// consumed from b including this separator. Error may be non-nil if options are
+// given.
+func (p *Paragraph) SetString(s string, opts ...Option) (n int, err error) {
+	panic("unimplemented")
+}
+
+// IsLeftToRight reports whether the principle direction of rendering for this
+// paragraphs is left-to-right. If this returns false, the principle direction
+// of rendering is right-to-left.
+func (p *Paragraph) IsLeftToRight() bool {
+	panic("unimplemented")
+}
+
+// Direction returns the direction of the text of this paragraph.
+//
+// The direction may be LeftToRight, RightToLeft, Mixed, or Neutral.
+func (p *Paragraph) Direction() Direction {
+	panic("unimplemented")
+}
+
+// RunAt reports the Run at the given position of the input text.
+//
+// This method can be used for computing line breaks on paragraphs.
+func (p *Paragraph) RunAt(pos int) Run {
+	panic("unimplemented")
+}
+
+// Order computes the visual ordering of all the runs in a Paragraph.
+func (p *Paragraph) Order() (Ordering, error) {
+	panic("unimplemented")
+}
+
+// Line computes the visual ordering of runs for a single line starting and
+// ending at the given positions in the original text.
+func (p *Paragraph) Line(start, end int) (Ordering, error) {
+	panic("unimplemented")
+}
+
+// An Ordering holds the computed visual order of runs of a Paragraph. Calling
+// SetBytes or SetString on the originating Paragraph invalidates an Ordering.
+// The methods of an Ordering should only be called by one goroutine at a time.
+type Ordering struct{}
+
+// Direction reports the directionality of the runs.
+//
+// The direction may be LeftToRight, RightToLeft, Mixed, or Neutral.
+func (o *Ordering) Direction() Direction {
+	panic("unimplemented")
+}
+
+// NumRuns returns the number of runs.
+func (o *Ordering) NumRuns() int {
+	panic("unimplemented")
+}
+
+// Run returns the ith run within the ordering.
+func (o *Ordering) Run(i int) Run {
+	panic("unimplemented")
+}
+
+// TODO: perhaps with options.
+// // Reorder creates a reader that reads the runes in visual order per character.
+// // Modifiers remain after the runes they modify.
+// func (l *Runs) Reorder() io.Reader {
+// 	panic("unimplemented")
+// }
+
+// A Run is a continuous sequence of characters of a single direction.
+type Run struct {
+}
+
+// String returns the text of the run in its original order.
+func (r *Run) String() string {
+	panic("unimplemented")
+}
+
+// Bytes returns the text of the run in its original order.
+func (r *Run) Bytes() []byte {
+	panic("unimplemented")
+}
+
+// TODO: methods for
+// - Display order
+// - headers and footers
+// - bracket replacement.
+
+// Direction reports the direction of the run.
+func (r *Run) Direction() Direction {
+	panic("unimplemented")
+}
+
+// Position of the Run within the text passed to SetBytes or SetString of the
+// originating Paragraph value.
+func (r *Run) Pos() (start, end int) {
+	panic("unimplemented")
+}
+
+// AppendReverse reverses the order of characters of in, appends them to out,
+// and returns the result. Modifiers will still follow the runes they modify.
+// Brackets are replaced with their counterparts.
+func AppendReverse(out, in []byte) []byte {
+	panic("unimplemented")
+}
+
+// ReverseString reverses the order of characters in s and returns a new string.
+// Modifiers will still follow the runes they modify. Brackets are replaced with
+// their counterparts.
+func ReverseString(s string) string {
+	panic("unimplemented")
+}

vendor/golang.org/x/text/unicode/bidi/bracket.go

@@ -0,0 +1,335 @@
+// 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 bidi
+
+import (
+	"container/list"
+	"fmt"
+	"sort"
+)
+
+// This file contains a port of the reference implementation of the
+// Bidi Parentheses Algorithm:
+// https://www.unicode.org/Public/PROGRAMS/BidiReferenceJava/BidiPBAReference.java
+//
+// The implementation in this file covers definitions BD14-BD16 and rule N0
+// of UAX#9.
+//
+// Some preprocessing is done for each rune before data is passed to this
+// algorithm:
+//  - opening and closing brackets are identified
+//  - a bracket pair type, like '(' and ')' is assigned a unique identifier that
+//    is identical for the opening and closing bracket. It is left to do these
+//    mappings.
+//  - The BPA algorithm requires that bracket characters that are canonical
+//    equivalents of each other be able to be substituted for each other.
+//    It is the responsibility of the caller to do this canonicalization.
+//
+// In implementing BD16, this implementation departs slightly from the "logical"
+// algorithm defined in UAX#9. In particular, the stack referenced there
+// supports operations that go beyond a "basic" stack. An equivalent
+// implementation based on a linked list is used here.
+
+// Bidi_Paired_Bracket_Type
+// BD14. An opening paired bracket is a character whose
+// Bidi_Paired_Bracket_Type property value is Open.
+//
+// BD15. A closing paired bracket is a character whose
+// Bidi_Paired_Bracket_Type property value is Close.
+type bracketType byte
+
+const (
+	bpNone bracketType = iota
+	bpOpen
+	bpClose
+)
+
+// bracketPair holds a pair of index values for opening and closing bracket
+// location of a bracket pair.
+type bracketPair struct {
+	opener int
+	closer int
+}
+
+func (b *bracketPair) String() string {
+	return fmt.Sprintf("(%v, %v)", b.opener, b.closer)
+}
+
+// bracketPairs is a slice of bracketPairs with a sort.Interface implementation.
+type bracketPairs []bracketPair
+
+func (b bracketPairs) Len() int           { return len(b) }
+func (b bracketPairs) Swap(i, j int)      { b[i], b[j] = b[j], b[i] }
+func (b bracketPairs) Less(i, j int) bool { return b[i].opener < b[j].opener }
+
+// resolvePairedBrackets runs the paired bracket part of the UBA algorithm.
+//
+// For each rune, it takes the indexes into the original string, the class the
+// bracket type (in pairTypes) and the bracket identifier (pairValues). It also
+// takes the direction type for the start-of-sentence and the embedding level.
+//
+// The identifiers for bracket types are the rune of the canonicalized opening
+// bracket for brackets (open or close) or 0 for runes that are not brackets.
+func resolvePairedBrackets(s *isolatingRunSequence) {
+	p := bracketPairer{
+		sos:              s.sos,
+		openers:          list.New(),
+		codesIsolatedRun: s.types,
+		indexes:          s.indexes,
+	}
+	dirEmbed := L
+	if s.level&1 != 0 {
+		dirEmbed = R
+	}
+	p.locateBrackets(s.p.pairTypes, s.p.pairValues)
+	p.resolveBrackets(dirEmbed, s.p.initialTypes)
+}
+
+type bracketPairer struct {
+	sos Class // direction corresponding to start of sequence
+
+	// The following is a restatement of BD 16 using non-algorithmic language.
+	//
+	// A bracket pair is a pair of characters consisting of an opening
+	// paired bracket and a closing paired bracket such that the
+	// Bidi_Paired_Bracket property value of the former equals the latter,
+	// subject to the following constraints.
+	// - both characters of a pair occur in the same isolating run sequence
+	// - the closing character of a pair follows the opening character
+	// - any bracket character can belong at most to one pair, the earliest possible one
+	// - any bracket character not part of a pair is treated like an ordinary character
+	// - pairs may nest properly, but their spans may not overlap otherwise
+
+	// Bracket characters with canonical decompositions are supposed to be
+	// treated as if they had been normalized, to allow normalized and non-
+	// normalized text to give the same result. In this implementation that step
+	// is pushed out to the caller. The caller has to ensure that the pairValue
+	// slices contain the rune of the opening bracket after normalization for
+	// any opening or closing bracket.
+
+	openers *list.List // list of positions for opening brackets
+
+	// bracket pair positions sorted by location of opening bracket
+	pairPositions bracketPairs
+
+	codesIsolatedRun []Class // directional bidi codes for an isolated run
+	indexes          []int   // array of index values into the original string
+
+}
+
+// matchOpener reports whether characters at given positions form a matching
+// bracket pair.
+func (p *bracketPairer) matchOpener(pairValues []rune, opener, closer int) bool {
+	return pairValues[p.indexes[opener]] == pairValues[p.indexes[closer]]
+}
+
+const maxPairingDepth = 63
+
+// locateBrackets locates matching bracket pairs according to BD16.
+//
+// This implementation uses a linked list instead of a stack, because, while
+// elements are added at the front (like a push) they are not generally removed
+// in atomic 'pop' operations, reducing the benefit of the stack archetype.
+func (p *bracketPairer) locateBrackets(pairTypes []bracketType, pairValues []rune) {
+	// traverse the run
+	// do that explicitly (not in a for-each) so we can record position
+	for i, index := range p.indexes {
+
+		// look at the bracket type for each character
+		if pairTypes[index] == bpNone || p.codesIsolatedRun[i] != ON {
+			// continue scanning
+			continue
+		}
+		switch pairTypes[index] {
+		case bpOpen:
+			// check if maximum pairing depth reached
+			if p.openers.Len() == maxPairingDepth {
+				p.openers.Init()
+				return
+			}
+			// remember opener location, most recent first
+			p.openers.PushFront(i)
+
+		case bpClose:
+			// see if there is a match
+			count := 0
+			for elem := p.openers.Front(); elem != nil; elem = elem.Next() {
+				count++
+				opener := elem.Value.(int)
+				if p.matchOpener(pairValues, opener, i) {
+					// if the opener matches, add nested pair to the ordered list
+					p.pairPositions = append(p.pairPositions, bracketPair{opener, i})
+					// remove up to and including matched opener
+					for ; count > 0; count-- {
+						p.openers.Remove(p.openers.Front())
+					}
+					break
+				}
+			}
+			sort.Sort(p.pairPositions)
+			// if we get here, the closing bracket matched no openers
+			// and gets ignored
+		}
+	}
+}
+
+// Bracket pairs within an isolating run sequence are processed as units so
+// that both the opening and the closing paired bracket in a pair resolve to
+// the same direction.
+//
+// N0. Process bracket pairs in an isolating run sequence sequentially in
+// the logical order of the text positions of the opening paired brackets
+// using the logic given below. Within this scope, bidirectional types EN
+// and AN are treated as R.
+//
+// Identify the bracket pairs in the current isolating run sequence
+// according to BD16. For each bracket-pair element in the list of pairs of
+// text positions:
+//
+// a Inspect the bidirectional types of the characters enclosed within the
+// bracket pair.
+//
+// b If any strong type (either L or R) matching the embedding direction is
+// found, set the type for both brackets in the pair to match the embedding
+// direction.
+//
+// o [ e ] o -> o e e e o
+//
+// o [ o e ] -> o e o e e
+//
+// o [ NI e ] -> o e NI e e
+//
+// c Otherwise, if a strong type (opposite the embedding direction) is
+// found, test for adjacent strong types as follows: 1 First, check
+// backwards before the opening paired bracket until the first strong type
+// (L, R, or sos) is found. If that first preceding strong type is opposite
+// the embedding direction, then set the type for both brackets in the pair
+// to that type. 2 Otherwise, set the type for both brackets in the pair to
+// the embedding direction.
+//
+// o [ o ] e -> o o o o e
+//
+// o [ o NI ] o -> o o o NI o o
+//
+// e [ o ] o -> e e o e o
+//
+// e [ o ] e -> e e o e e
+//
+// e ( o [ o ] NI ) e -> e e o o o o NI e e
+//
+// d Otherwise, do not set the type for the current bracket pair. Note that
+// if the enclosed text contains no strong types the paired brackets will
+// both resolve to the same level when resolved individually using rules N1
+// and N2.
+//
+// e ( NI ) o -> e ( NI ) o
+
+// getStrongTypeN0 maps character's directional code to strong type as required
+// by rule N0.
+//
+// TODO: have separate type for "strong" directionality.
+func (p *bracketPairer) getStrongTypeN0(index int) Class {
+	switch p.codesIsolatedRun[index] {
+	// in the scope of N0, number types are treated as R
+	case EN, AN, AL, R:
+		return R
+	case L:
+		return L
+	default:
+		return ON
+	}
+}
+
+// classifyPairContent reports the strong types contained inside a Bracket Pair,
+// assuming the given embedding direction.
+//
+// It returns ON if no strong type is found. If a single strong type is found,
+// it returns this type. Otherwise it returns the embedding direction.
+//
+// TODO: use separate type for "strong" directionality.
+func (p *bracketPairer) classifyPairContent(loc bracketPair, dirEmbed Class) Class {
+	dirOpposite := ON
+	for i := loc.opener + 1; i < loc.closer; i++ {
+		dir := p.getStrongTypeN0(i)
+		if dir == ON {
+			continue
+		}
+		if dir == dirEmbed {
+			return dir // type matching embedding direction found
+		}
+		dirOpposite = dir
+	}
+	// return ON if no strong type found, or class opposite to dirEmbed
+	return dirOpposite
+}
+
+// classBeforePair determines which strong types are present before a Bracket
+// Pair. Return R or L if strong type found, otherwise ON.
+func (p *bracketPairer) classBeforePair(loc bracketPair) Class {
+	for i := loc.opener - 1; i >= 0; i-- {
+		if dir := p.getStrongTypeN0(i); dir != ON {
+			return dir
+		}
+	}
+	// no strong types found, return sos
+	return p.sos
+}
+
+// assignBracketType implements rule N0 for a single bracket pair.
+func (p *bracketPairer) assignBracketType(loc bracketPair, dirEmbed Class, initialTypes []Class) {
+	// rule "N0, a", inspect contents of pair
+	dirPair := p.classifyPairContent(loc, dirEmbed)
+
+	// dirPair is now L, R, or N (no strong type found)
+
+	// the following logical tests are performed out of order compared to
+	// the statement of the rules but yield the same results
+	if dirPair == ON {
+		return // case "d" - nothing to do
+	}
+
+	if dirPair != dirEmbed {
+		// case "c": strong type found, opposite - check before (c.1)
+		dirPair = p.classBeforePair(loc)
+		if dirPair == dirEmbed || dirPair == ON {
+			// no strong opposite type found before - use embedding (c.2)
+			dirPair = dirEmbed
+		}
+	}
+	// else: case "b", strong type found matching embedding,
+	// no explicit action needed, as dirPair is already set to embedding
+	// direction
+
+	// set the bracket types to the type found
+	p.setBracketsToType(loc, dirPair, initialTypes)
+}
+
+func (p *bracketPairer) setBracketsToType(loc bracketPair, dirPair Class, initialTypes []Class) {
+	p.codesIsolatedRun[loc.opener] = dirPair
+	p.codesIsolatedRun[loc.closer] = dirPair
+
+	for i := loc.opener + 1; i < loc.closer; i++ {
+		index := p.indexes[i]
+		if initialTypes[index] != NSM {
+			break
+		}
+		p.codesIsolatedRun[i] = dirPair
+	}
+
+	for i := loc.closer + 1; i < len(p.indexes); i++ {
+		index := p.indexes[i]
+		if initialTypes[index] != NSM {
+			break
+		}
+		p.codesIsolatedRun[i] = dirPair
+	}
+}
+
+// resolveBrackets implements rule N0 for a list of pairs.
+func (p *bracketPairer) resolveBrackets(dirEmbed Class, initialTypes []Class) {
+	for _, loc := range p.pairPositions {
+		p.assignBracketType(loc, dirEmbed, initialTypes)
+	}
+}

vendor/golang.org/x/text/unicode/bidi/gen.go

@@ -0,0 +1,133 @@
+// 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 (
+	"flag"
+	"log"
+
+	"golang.org/x/text/internal/gen"
+	"golang.org/x/text/internal/triegen"
+	"golang.org/x/text/internal/ucd"
+)
+
+var outputFile = flag.String("out", "tables.go", "output file")
+
+func main() {
+	gen.Init()
+	gen.Repackage("gen_trieval.go", "trieval.go", "bidi")
+	gen.Repackage("gen_ranges.go", "ranges_test.go", "bidi")
+
+	genTables()
+}
+
+// bidiClass names and codes taken from class "bc" in
+// https://www.unicode.org/Public/8.0.0/ucd/PropertyValueAliases.txt
+var bidiClass = map[string]Class{
+	"AL":  AL,  // ArabicLetter
+	"AN":  AN,  // ArabicNumber
+	"B":   B,   // ParagraphSeparator
+	"BN":  BN,  // BoundaryNeutral
+	"CS":  CS,  // CommonSeparator
+	"EN":  EN,  // EuropeanNumber
+	"ES":  ES,  // EuropeanSeparator
+	"ET":  ET,  // EuropeanTerminator
+	"L":   L,   // LeftToRight
+	"NSM": NSM, // NonspacingMark
+	"ON":  ON,  // OtherNeutral
+	"R":   R,   // RightToLeft
+	"S":   S,   // SegmentSeparator
+	"WS":  WS,  // WhiteSpace
+
+	"FSI": Control,
+	"PDF": Control,
+	"PDI": Control,
+	"LRE": Control,
+	"LRI": Control,
+	"LRO": Control,
+	"RLE": Control,
+	"RLI": Control,
+	"RLO": Control,
+}
+
+func genTables() {
+	if numClass > 0x0F {
+		log.Fatalf("Too many Class constants (%#x > 0x0F).", numClass)
+	}
+	w := gen.NewCodeWriter()
+	defer w.WriteVersionedGoFile(*outputFile, "bidi")
+
+	gen.WriteUnicodeVersion(w)
+
+	t := triegen.NewTrie("bidi")
+
+	// Build data about bracket mapping. These bits need to be or-ed with
+	// any other bits.
+	orMask := map[rune]uint64{}
+
+	xorMap := map[rune]int{}
+	xorMasks := []rune{0} // First value is no-op.
+
+	ucd.Parse(gen.OpenUCDFile("BidiBrackets.txt"), func(p *ucd.Parser) {
+		r1 := p.Rune(0)
+		r2 := p.Rune(1)
+		xor := r1 ^ r2
+		if _, ok := xorMap[xor]; !ok {
+			xorMap[xor] = len(xorMasks)
+			xorMasks = append(xorMasks, xor)
+		}
+		entry := uint64(xorMap[xor]) << xorMaskShift
+		switch p.String(2) {
+		case "o":
+			entry |= openMask
+		case "c", "n":
+		default:
+			log.Fatalf("Unknown bracket class %q.", p.String(2))
+		}
+		orMask[r1] = entry
+	})
+
+	w.WriteComment(`
+	xorMasks contains masks to be xor-ed with brackets to get the reverse
+	version.`)
+	w.WriteVar("xorMasks", xorMasks)
+
+	done := map[rune]bool{}
+
+	insert := func(r rune, c Class) {
+		if !done[r] {
+			t.Insert(r, orMask[r]|uint64(c))
+			done[r] = true
+		}
+	}
+
+	// Insert the derived BiDi properties.
+	ucd.Parse(gen.OpenUCDFile("extracted/DerivedBidiClass.txt"), func(p *ucd.Parser) {
+		r := p.Rune(0)
+		class, ok := bidiClass[p.String(1)]
+		if !ok {
+			log.Fatalf("%U: Unknown BiDi class %q", r, p.String(1))
+		}
+		insert(r, class)
+	})
+	visitDefaults(insert)
+
+	// TODO: use sparse blocks. This would reduce table size considerably
+	// from the looks of it.
+
+	sz, err := t.Gen(w)
+	if err != nil {
+		log.Fatal(err)
+	}
+	w.Size += sz
+}
+
+// dummy values to make methods in gen_common compile. The real versions
+// will be generated by this file to tables.go.
+var (
+	xorMasks []rune
+)

vendor/golang.org/x/text/unicode/bidi/gen_ranges.go

@@ -0,0 +1,57 @@
+// 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 (
+	"unicode"
+
+	"golang.org/x/text/internal/gen"
+	"golang.org/x/text/internal/ucd"
+	"golang.org/x/text/unicode/rangetable"
+)
+
+// These tables are hand-extracted from:
+// https://www.unicode.org/Public/8.0.0/ucd/extracted/DerivedBidiClass.txt
+func visitDefaults(fn func(r rune, c Class)) {
+	// first write default values for ranges listed above.
+	visitRunes(fn, AL, []rune{
+		0x0600, 0x07BF, // Arabic
+		0x08A0, 0x08FF, // Arabic Extended-A
+		0xFB50, 0xFDCF, // Arabic Presentation Forms
+		0xFDF0, 0xFDFF,
+		0xFE70, 0xFEFF,
+		0x0001EE00, 0x0001EEFF, // Arabic Mathematical Alpha Symbols
+	})
+	visitRunes(fn, R, []rune{
+		0x0590, 0x05FF, // Hebrew
+		0x07C0, 0x089F, // Nko et al.
+		0xFB1D, 0xFB4F,
+		0x00010800, 0x00010FFF, // Cypriot Syllabary et. al.
+		0x0001E800, 0x0001EDFF,
+		0x0001EF00, 0x0001EFFF,
+	})
+	visitRunes(fn, ET, []rune{ // European Terminator
+		0x20A0, 0x20Cf, // Currency symbols
+	})
+	rangetable.Visit(unicode.Noncharacter_Code_Point, func(r rune) {
+		fn(r, BN) // Boundary Neutral
+	})
+	ucd.Parse(gen.OpenUCDFile("DerivedCoreProperties.txt"), func(p *ucd.Parser) {
+		if p.String(1) == "Default_Ignorable_Code_Point" {
+			fn(p.Rune(0), BN) // Boundary Neutral
+		}
+	})
+}
+
+func visitRunes(fn func(r rune, c Class), c Class, runes []rune) {
+	for i := 0; i < len(runes); i += 2 {
+		lo, hi := runes[i], runes[i+1]
+		for j := lo; j <= hi; j++ {
+			fn(j, c)
+		}
+	}
+}

vendor/golang.org/x/text/unicode/bidi/gen_trieval.go

@@ -0,0 +1,64 @@
+// 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
+
+// Class is the Unicode BiDi class. Each rune has a single class.
+type Class uint
+
+const (
+	L       Class = iota // LeftToRight
+	R                    // RightToLeft
+	EN                   // EuropeanNumber
+	ES                   // EuropeanSeparator
+	ET                   // EuropeanTerminator
+	AN                   // ArabicNumber
+	CS                   // CommonSeparator
+	B                    // ParagraphSeparator
+	S                    // SegmentSeparator
+	WS                   // WhiteSpace
+	ON                   // OtherNeutral
+	BN                   // BoundaryNeutral
+	NSM                  // NonspacingMark
+	AL                   // ArabicLetter
+	Control              // Control LRO - PDI
+
+	numClass
+
+	LRO // LeftToRightOverride
+	RLO // RightToLeftOverride
+	LRE // LeftToRightEmbedding
+	RLE // RightToLeftEmbedding
+	PDF // PopDirectionalFormat
+	LRI // LeftToRightIsolate
+	RLI // RightToLeftIsolate
+	FSI // FirstStrongIsolate
+	PDI // PopDirectionalIsolate
+
+	unknownClass = ^Class(0)
+)
+
+var controlToClass = map[rune]Class{
+	0x202D: LRO, // LeftToRightOverride,
+	0x202E: RLO, // RightToLeftOverride,
+	0x202A: LRE, // LeftToRightEmbedding,
+	0x202B: RLE, // RightToLeftEmbedding,
+	0x202C: PDF, // PopDirectionalFormat,
+	0x2066: LRI, // LeftToRightIsolate,
+	0x2067: RLI, // RightToLeftIsolate,
+	0x2068: FSI, // FirstStrongIsolate,
+	0x2069: PDI, // PopDirectionalIsolate,
+}
+
+// A trie entry has the following bits:
+// 7..5  XOR mask for brackets
+// 4     1: Bracket open, 0: Bracket close
+// 3..0  Class type
+
+const (
+	openMask     = 0x10
+	xorMaskShift = 5
+)

vendor/golang.org/x/text/unicode/bidi/prop.go

@@ -0,0 +1,206 @@
+// Copyright 2016 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 bidi
+
+import "unicode/utf8"
+
+// Properties provides access to BiDi properties of runes.
+type Properties struct {
+	entry uint8
+	last  uint8
+}
+
+var trie = newBidiTrie(0)
+
+// TODO: using this for bidirule reduces the running time by about 5%. Consider
+// if this is worth exposing or if we can find a way to speed up the Class
+// method.
+//
+// // CompactClass is like Class, but maps all of the BiDi control classes
+// // (LRO, RLO, LRE, RLE, PDF, LRI, RLI, FSI, PDI) to the class Control.
+// func (p Properties) CompactClass() Class {
+// 	return Class(p.entry & 0x0F)
+// }
+
+// Class returns the Bidi class for p.
+func (p Properties) Class() Class {
+	c := Class(p.entry & 0x0F)
+	if c == Control {
+		c = controlByteToClass[p.last&0xF]
+	}
+	return c
+}
+
+// IsBracket reports whether the rune is a bracket.
+func (p Properties) IsBracket() bool { return p.entry&0xF0 != 0 }
+
+// IsOpeningBracket reports whether the rune is an opening bracket.
+// IsBracket must return true.
+func (p Properties) IsOpeningBracket() bool { return p.entry&openMask != 0 }
+
+// TODO: find a better API and expose.
+func (p Properties) reverseBracket(r rune) rune {
+	return xorMasks[p.entry>>xorMaskShift] ^ r
+}
+
+var controlByteToClass = [16]Class{
+	0xD: LRO, // U+202D LeftToRightOverride,
+	0xE: RLO, // U+202E RightToLeftOverride,
+	0xA: LRE, // U+202A LeftToRightEmbedding,
+	0xB: RLE, // U+202B RightToLeftEmbedding,
+	0xC: PDF, // U+202C PopDirectionalFormat,
+	0x6: LRI, // U+2066 LeftToRightIsolate,
+	0x7: RLI, // U+2067 RightToLeftIsolate,
+	0x8: FSI, // U+2068 FirstStrongIsolate,
+	0x9: PDI, // U+2069 PopDirectionalIsolate,
+}
+
+// LookupRune returns properties for r.
+func LookupRune(r rune) (p Properties, size int) {
+	var buf [4]byte
+	n := utf8.EncodeRune(buf[:], r)
+	return Lookup(buf[:n])
+}
+
+// TODO: these lookup methods are based on the generated trie code. The returned
+// sizes have slightly different semantics from the generated code, in that it
+// always returns size==1 for an illegal UTF-8 byte (instead of the length
+// of the maximum invalid subsequence). Most Transformers, like unicode/norm,
+// leave invalid UTF-8 untouched, in which case it has performance benefits to
+// do so (without changing the semantics). Bidi requires the semantics used here
+// for the bidirule implementation to be compatible with the Go semantics.
+//  They ultimately should perhaps be adopted by all trie implementations, for
+// convenience sake.
+// This unrolled code also boosts performance of the secure/bidirule package by
+// about 30%.
+// So, to remove this code:
+//   - add option to trie generator to define return type.
+//   - always return 1 byte size for ill-formed UTF-8 runes.
+
+// Lookup returns properties for the first rune in s and the width in bytes of
+// its encoding. The size will be 0 if s does not hold enough bytes to complete
+// the encoding.
+func Lookup(s []byte) (p Properties, sz int) {
+	c0 := s[0]
+	switch {
+	case c0 < 0x80: // is ASCII
+		return Properties{entry: bidiValues[c0]}, 1
+	case c0 < 0xC2:
+		return Properties{}, 1
+	case c0 < 0xE0: // 2-byte UTF-8
+		if len(s) < 2 {
+			return Properties{}, 0
+		}
+		i := bidiIndex[c0]
+		c1 := s[1]
+		if c1 < 0x80 || 0xC0 <= c1 {
+			return Properties{}, 1
+		}
+		return Properties{entry: trie.lookupValue(uint32(i), c1)}, 2
+	case c0 < 0xF0: // 3-byte UTF-8
+		if len(s) < 3 {
+			return Properties{}, 0
+		}
+		i := bidiIndex[c0]
+		c1 := s[1]
+		if c1 < 0x80 || 0xC0 <= c1 {
+			return Properties{}, 1
+		}
+		o := uint32(i)<<6 + uint32(c1)
+		i = bidiIndex[o]
+		c2 := s[2]
+		if c2 < 0x80 || 0xC0 <= c2 {
+			return Properties{}, 1
+		}
+		return Properties{entry: trie.lookupValue(uint32(i), c2), last: c2}, 3
+	case c0 < 0xF8: // 4-byte UTF-8
+		if len(s) < 4 {
+			return Properties{}, 0
+		}
+		i := bidiIndex[c0]
+		c1 := s[1]
+		if c1 < 0x80 || 0xC0 <= c1 {
+			return Properties{}, 1
+		}
+		o := uint32(i)<<6 + uint32(c1)
+		i = bidiIndex[o]
+		c2 := s[2]
+		if c2 < 0x80 || 0xC0 <= c2 {
+			return Properties{}, 1
+		}
+		o = uint32(i)<<6 + uint32(c2)
+		i = bidiIndex[o]
+		c3 := s[3]
+		if c3 < 0x80 || 0xC0 <= c3 {
+			return Properties{}, 1
+		}
+		return Properties{entry: trie.lookupValue(uint32(i), c3)}, 4
+	}
+	// Illegal rune
+	return Properties{}, 1
+}
+
+// LookupString returns properties for the first rune in s and the width in
+// bytes of its encoding. The size will be 0 if s does not hold enough bytes to
+// complete the encoding.
+func LookupString(s string) (p Properties, sz int) {
+	c0 := s[0]
+	switch {
+	case c0 < 0x80: // is ASCII
+		return Properties{entry: bidiValues[c0]}, 1
+	case c0 < 0xC2:
+		return Properties{}, 1
+	case c0 < 0xE0: // 2-byte UTF-8
+		if len(s) < 2 {
+			return Properties{}, 0
+		}
+		i := bidiIndex[c0]
+		c1 := s[1]
+		if c1 < 0x80 || 0xC0 <= c1 {
+			return Properties{}, 1
+		}
+		return Properties{entry: trie.lookupValue(uint32(i), c1)}, 2
+	case c0 < 0xF0: // 3-byte UTF-8
+		if len(s) < 3 {
+			return Properties{}, 0
+		}
+		i := bidiIndex[c0]
+		c1 := s[1]
+		if c1 < 0x80 || 0xC0 <= c1 {
+			return Properties{}, 1
+		}
+		o := uint32(i)<<6 + uint32(c1)
+		i = bidiIndex[o]
+		c2 := s[2]
+		if c2 < 0x80 || 0xC0 <= c2 {
+			return Properties{}, 1
+		}
+		return Properties{entry: trie.lookupValue(uint32(i), c2), last: c2}, 3
+	case c0 < 0xF8: // 4-byte UTF-8
+		if len(s) < 4 {
+			return Properties{}, 0
+		}
+		i := bidiIndex[c0]
+		c1 := s[1]
+		if c1 < 0x80 || 0xC0 <= c1 {
+			return Properties{}, 1
+		}
+		o := uint32(i)<<6 + uint32(c1)
+		i = bidiIndex[o]
+		c2 := s[2]
+		if c2 < 0x80 || 0xC0 <= c2 {
+			return Properties{}, 1
+		}
+		o = uint32(i)<<6 + uint32(c2)
+		i = bidiIndex[o]
+		c3 := s[3]
+		if c3 < 0x80 || 0xC0 <= c3 {
+			return Properties{}, 1
+		}
+		return Properties{entry: trie.lookupValue(uint32(i), c3)}, 4
+	}
+	// Illegal rune
+	return Properties{}, 1
+}

vendor/golang.org/x/text/unicode/bidi/trieval.go

@@ -0,0 +1,60 @@
+// Code generated by running "go generate" in golang.org/x/text. DO NOT EDIT.
+
+package bidi
+
+// Class is the Unicode BiDi class. Each rune has a single class.
+type Class uint
+
+const (
+	L       Class = iota // LeftToRight
+	R                    // RightToLeft
+	EN                   // EuropeanNumber
+	ES                   // EuropeanSeparator
+	ET                   // EuropeanTerminator
+	AN                   // ArabicNumber
+	CS                   // CommonSeparator
+	B                    // ParagraphSeparator
+	S                    // SegmentSeparator
+	WS                   // WhiteSpace
+	ON                   // OtherNeutral
+	BN                   // BoundaryNeutral
+	NSM                  // NonspacingMark
+	AL                   // ArabicLetter
+	Control              // Control LRO - PDI
+
+	numClass
+
+	LRO // LeftToRightOverride
+	RLO // RightToLeftOverride
+	LRE // LeftToRightEmbedding
+	RLE // RightToLeftEmbedding
+	PDF // PopDirectionalFormat
+	LRI // LeftToRightIsolate
+	RLI // RightToLeftIsolate
+	FSI // FirstStrongIsolate
+	PDI // PopDirectionalIsolate
+
+	unknownClass = ^Class(0)
+)
+
+var controlToClass = map[rune]Class{
+	0x202D: LRO, // LeftToRightOverride,
+	0x202E: RLO, // RightToLeftOverride,
+	0x202A: LRE, // LeftToRightEmbedding,
+	0x202B: RLE, // RightToLeftEmbedding,
+	0x202C: PDF, // PopDirectionalFormat,
+	0x2066: LRI, // LeftToRightIsolate,
+	0x2067: RLI, // RightToLeftIsolate,
+	0x2068: FSI, // FirstStrongIsolate,
+	0x2069: PDI, // PopDirectionalIsolate,
+}
+
+// A trie entry has the following bits:
+// 7..5  XOR mask for brackets
+// 4     1: Bracket open, 0: Bracket close
+// 3..0  Class type
+
+const (
+	openMask     = 0x10
+	xorMaskShift = 5
+)

vendor/golang.org/x/text/unicode/norm/composition.go

@@ -407,7 +407,7 @@ func decomposeHangul(buf []byte, r rune) int {
 
 // decomposeHangul algorithmically decomposes a Hangul rune into
 // its Jamo components.
-// See http://unicode.org/reports/tr15/#Hangul for details on decomposing Hangul.
+// See https://unicode.org/reports/tr15/#Hangul for details on decomposing Hangul.
 func (rb *reorderBuffer) decomposeHangul(r rune) {
 	r -= hangulBase
 	x := r % jamoTCount
@@ -420,7 +420,7 @@ func (rb *reorderBuffer) decomposeHangul(r rune) {
 }
 
 // combineHangul algorithmically combines Jamo character components into Hangul.
-// See http://unicode.org/reports/tr15/#Hangul for details on combining Hangul.
+// See https://unicode.org/reports/tr15/#Hangul for details on combining Hangul.
 func (rb *reorderBuffer) combineHangul(s, i, k int) {
 	b := rb.rune[:]
 	bn := rb.nrune
@@ -461,6 +461,10 @@ func (rb *reorderBuffer) combineHangul(s, i, k int) {
 // It should only be used to recompose a single segment, as it will not
 // handle alternations between Hangul and non-Hangul characters correctly.
 func (rb *reorderBuffer) compose() {
+	// Lazily load the map used by the combine func below, but do
+	// it outside of the loop.
+	recompMapOnce.Do(buildRecompMap)
+
 	// UAX #15, section X5 , including Corrigendum #5
 	// "In any character sequence beginning with starter S, a character C is
 	//  blocked from S if and only if there is some character B between S

vendor/golang.org/x/text/unicode/norm/forminfo.go

@@ -4,6 +4,8 @@
 
 package norm
 
+import "encoding/binary"
+
 // This file contains Form-specific logic and wrappers for data in tables.go.
 
 // Rune info is stored in a separate trie per composing form. A composing form
@@ -178,6 +180,17 @@ func (p Properties) TrailCCC() uint8 {
 	return ccc[p.tccc]
 }
 
+func buildRecompMap() {
+	recompMap = make(map[uint32]rune, len(recompMapPacked)/8)
+	var buf [8]byte
+	for i := 0; i < len(recompMapPacked); i += 8 {
+		copy(buf[:], recompMapPacked[i:i+8])
+		key := binary.BigEndian.Uint32(buf[:4])
+		val := binary.BigEndian.Uint32(buf[4:])
+		recompMap[key] = rune(val)
+	}
+}
+
 // Recomposition
 // We use 32-bit keys instead of 64-bit for the two codepoint keys.
 // This clips off the bits of three entries, but we know this will not
@@ -186,8 +199,14 @@ func (p Properties) TrailCCC() uint8 {
 // Note that the recomposition map for NFC and NFKC are identical.
 
 // combine returns the combined rune or 0 if it doesn't exist.
+//
+// The caller is responsible for calling
+// recompMapOnce.Do(buildRecompMap) sometime before this is called.
 func combine(a, b rune) rune {
 	key := uint32(uint16(a))<<16 + uint32(uint16(b))
+	if recompMap == nil {
+		panic("caller error") // see func comment
+	}
 	return recompMap[key]
 }
 

vendor/golang.org/x/text/unicode/norm/iter.go

@@ -128,8 +128,9 @@ func (i *Iter) Next() []byte {
 func nextASCIIBytes(i *Iter) []byte {
 	p := i.p + 1
 	if p >= i.rb.nsrc {
+		p0 := i.p
 		i.setDone()
-		return i.rb.src.bytes[i.p:p]
+		return i.rb.src.bytes[p0:p]
 	}
 	if i.rb.src.bytes[p] < utf8.RuneSelf {
 		p0 := i.p

vendor/golang.org/x/text/unicode/norm/maketables.go

@@ -12,6 +12,7 @@ package main
 
 import (
 	"bytes"
+	"encoding/binary"
 	"flag"
 	"fmt"
 	"io"
@@ -261,7 +262,7 @@ func compactCCC() {
 
 // CompositionExclusions.txt has form:
 // 0958    # ...
-// See http://unicode.org/reports/tr44/ for full explanation
+// See https://unicode.org/reports/tr44/ for full explanation
 func loadCompositionExclusions() {
 	f := gen.OpenUCDFile("CompositionExclusions.txt")
 	defer f.Close()
@@ -735,6 +736,8 @@ func makeTables() {
 			max = n
 		}
 	}
+	fmt.Fprintln(w, `import "sync"`)
+	fmt.Fprintln(w)
 
 	fmt.Fprintln(w, "const (")
 	fmt.Fprintln(w, "\t// Version is the Unicode edition from which the tables are derived.")
@@ -782,16 +785,23 @@ func makeTables() {
 		sz := nrentries * 8
 		size += sz
 		fmt.Fprintf(w, "// recompMap: %d bytes (entries only)\n", sz)
-		fmt.Fprintln(w, "var recompMap = map[uint32]rune{")
+		fmt.Fprintln(w, "var recompMap map[uint32]rune")
+		fmt.Fprintln(w, "var recompMapOnce sync.Once\n")
+		fmt.Fprintln(w, `const recompMapPacked = "" +`)
+		var buf [8]byte
 		for i, c := range chars {
 			f := c.forms[FCanonical]
 			d := f.decomp
 			if !f.isOneWay && len(d) > 0 {
 				key := uint32(uint16(d[0]))<<16 + uint32(uint16(d[1]))
-				fmt.Fprintf(w, "0x%.8X: 0x%.4X,\n", key, i)
+				binary.BigEndian.PutUint32(buf[:4], key)
+				binary.BigEndian.PutUint32(buf[4:], uint32(i))
+				fmt.Fprintf(w, "\t\t%q + // 0x%.8X: 0x%.8X\n", string(buf[:]), key, uint32(i))
 			}
 		}
-		fmt.Fprintf(w, "}\n\n")
+		// hack so we don't have to special case the trailing plus sign
+		fmt.Fprintf(w, `	""`)
+		fmt.Fprintln(w)
 	}
 
 	fmt.Fprintf(w, "// Total size of tables: %dKB (%d bytes)\n", (size+512)/1024, size)
@@ -857,7 +867,7 @@ func verifyComputed() {
 // DerivedNormalizationProps.txt has form:
 // 00C0..00C5    ; NFD_QC; N # ...
 // 0374          ; NFD_QC; N # ...
-// See http://unicode.org/reports/tr44/ for full explanation
+// See https://unicode.org/reports/tr44/ for full explanation
 func testDerived() {
 	f := gen.OpenUCDFile("DerivedNormalizationProps.txt")
 	defer f.Close()

vendor/golang.org/x/text/unicode/norm/normalize.go

@@ -29,8 +29,8 @@ import (
 // proceed independently on both sides:
 //   f(x) == append(f(x[0:n]), f(x[n:])...)
 //
-// References: http://unicode.org/reports/tr15/ and
-// http://unicode.org/notes/tn5/.
+// References: https://unicode.org/reports/tr15/ and
+// https://unicode.org/notes/tn5/.
 type Form int
 
 const (

vendor/golang.org/x/text/unicode/norm/readwriter.go

@@ -60,8 +60,8 @@ func (w *normWriter) Close() error {
 }
 
 // Writer returns a new writer that implements Write(b)
-// by writing f(b) to w.  The returned writer may use an
-// an internal buffer to maintain state across Write calls.
+// by writing f(b) to w. The returned writer may use an
+// internal buffer to maintain state across Write calls.
 // Calling its Close method writes any buffered data to w.
 func (f Form) Writer(w io.Writer) io.WriteCloser {
 	wr := &normWriter{rb: reorderBuffer{}, w: w}

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

@@ -18,7 +18,6 @@ func (Form) Reset() {}
 // Users should either catch ErrShortDst and allow dst to grow or have dst be at
 // least of size MaxTransformChunkSize to be guaranteed of progress.
 func (f Form) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
-	n := 0
 	// Cap the maximum number of src bytes to check.
 	b := src
 	eof := atEOF
@@ -27,13 +26,14 @@ func (f Form) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error)
 		eof = false
 		b = b[:ns]
 	}
-	i, ok := formTable[f].quickSpan(inputBytes(b), n, len(b), eof)
-	n += copy(dst[n:], b[n:i])
+	i, ok := formTable[f].quickSpan(inputBytes(b), 0, len(b), eof)
+	n := copy(dst, b[:i])
 	if !ok {
 		nDst, nSrc, err = f.transform(dst[n:], src[n:], atEOF)
 		return nDst + n, nSrc + n, err
 	}
-	if n < len(src) && !atEOF {
+
+	if err == nil && n < len(src) && !atEOF {
 		err = transform.ErrShortSrc
 	}
 	return n, n, err
@@ -79,7 +79,7 @@ func (f Form) transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error)
 		nSrc += n
 		nDst += n
 		if ok {
-			if n < rb.nsrc && !atEOF {
+			if err == nil && n < rb.nsrc && !atEOF {
 				err = transform.ErrShortSrc
 			}
 			return nDst, nSrc, err