Source file src/regexp/syntax/parse.go

     1  // Copyright 2011 The Go Authors. All rights reserved.
     2  // Use of this source code is governed by a BSD-style
     3  // license that can be found in the LICENSE file.
     4  
     5  package syntax
     6  
     7  import (
     8  	"sort"
     9  	"strings"
    10  	"unicode"
    11  	"unicode/utf8"
    12  )
    13  
    14  // An Error describes a failure to parse a regular expression
    15  // and gives the offending expression.
    16  type Error struct {
    17  	Code ErrorCode
    18  	Expr string
    19  }
    20  
    21  func (e *Error) Error() string {
    22  	return "error parsing regexp: " + e.Code.String() + ": `" + e.Expr + "`"
    23  }
    24  
    25  // An ErrorCode describes a failure to parse a regular expression.
    26  type ErrorCode string
    27  
    28  const (
    29  	// Unexpected error
    30  	ErrInternalError ErrorCode = "regexp/syntax: internal error"
    31  
    32  	// Parse errors
    33  	ErrInvalidCharClass      ErrorCode = "invalid character class"
    34  	ErrInvalidCharRange      ErrorCode = "invalid character class range"
    35  	ErrInvalidEscape         ErrorCode = "invalid escape sequence"
    36  	ErrInvalidNamedCapture   ErrorCode = "invalid named capture"
    37  	ErrInvalidPerlOp         ErrorCode = "invalid or unsupported Perl syntax"
    38  	ErrInvalidRepeatOp       ErrorCode = "invalid nested repetition operator"
    39  	ErrInvalidRepeatSize     ErrorCode = "invalid repeat count"
    40  	ErrInvalidUTF8           ErrorCode = "invalid UTF-8"
    41  	ErrMissingBracket        ErrorCode = "missing closing ]"
    42  	ErrMissingParen          ErrorCode = "missing closing )"
    43  	ErrMissingRepeatArgument ErrorCode = "missing argument to repetition operator"
    44  	ErrTrailingBackslash     ErrorCode = "trailing backslash at end of expression"
    45  	ErrUnexpectedParen       ErrorCode = "unexpected )"
    46  	ErrNestingDepth          ErrorCode = "expression nests too deeply"
    47  	ErrLarge                 ErrorCode = "expression too large"
    48  )
    49  
    50  func (e ErrorCode) String() string {
    51  	return string(e)
    52  }
    53  
    54  // Flags control the behavior of the parser and record information about regexp context.
    55  type Flags uint16
    56  
    57  const (
    58  	FoldCase      Flags = 1 << iota // case-insensitive match
    59  	Literal                         // treat pattern as literal string
    60  	ClassNL                         // allow character classes like [^a-z] and [[:space:]] to match newline
    61  	DotNL                           // allow . to match newline
    62  	OneLine                         // treat ^ and $ as only matching at beginning and end of text
    63  	NonGreedy                       // make repetition operators default to non-greedy
    64  	PerlX                           // allow Perl extensions
    65  	UnicodeGroups                   // allow \p{Han}, \P{Han} for Unicode group and negation
    66  	WasDollar                       // regexp OpEndText was $, not \z
    67  	Simple                          // regexp contains no counted repetition
    68  
    69  	MatchNL = ClassNL | DotNL
    70  
    71  	Perl        = ClassNL | OneLine | PerlX | UnicodeGroups // as close to Perl as possible
    72  	POSIX Flags = 0                                         // POSIX syntax
    73  )
    74  
    75  // Pseudo-ops for parsing stack.
    76  const (
    77  	opLeftParen = opPseudo + iota
    78  	opVerticalBar
    79  )
    80  
    81  // maxHeight is the maximum height of a regexp parse tree.
    82  // It is somewhat arbitrarily chosen, but the idea is to be large enough
    83  // that no one will actually hit in real use but at the same time small enough
    84  // that recursion on the Regexp tree will not hit the 1GB Go stack limit.
    85  // The maximum amount of stack for a single recursive frame is probably
    86  // closer to 1kB, so this could potentially be raised, but it seems unlikely
    87  // that people have regexps nested even this deeply.
    88  // We ran a test on Google's C++ code base and turned up only
    89  // a single use case with depth > 100; it had depth 128.
    90  // Using depth 1000 should be plenty of margin.
    91  // As an optimization, we don't even bother calculating heights
    92  // until we've allocated at least maxHeight Regexp structures.
    93  const maxHeight = 1000
    94  
    95  // maxSize is the maximum size of a compiled regexp in Insts.
    96  // It too is somewhat arbitrarily chosen, but the idea is to be large enough
    97  // to allow significant regexps while at the same time small enough that
    98  // the compiled form will not take up too much memory.
    99  // 128 MB is enough for a 3.3 million Inst structures, which roughly
   100  // corresponds to a 3.3 MB regexp.
   101  const (
   102  	maxSize  = 128 << 20 / instSize
   103  	instSize = 5 * 8 // byte, 2 uint32, slice is 5 64-bit words
   104  )
   105  
   106  // maxRunes is the maximum number of runes allowed in a regexp tree
   107  // counting the runes in all the nodes.
   108  // Ignoring character classes p.numRunes is always less than the length of the regexp.
   109  // Character classes can make it much larger: each \pL adds 1292 runes.
   110  // 128 MB is enough for 32M runes, which is over 26k \pL instances.
   111  // Note that repetitions do not make copies of the rune slices,
   112  // so \pL{1000} is only one rune slice, not 1000.
   113  // We could keep a cache of character classes we've seen,
   114  // so that all the \pL we see use the same rune list,
   115  // but that doesn't remove the problem entirely:
   116  // consider something like [\pL01234][\pL01235][\pL01236]...[\pL^&*()].
   117  // And because the Rune slice is exposed directly in the Regexp,
   118  // there is not an opportunity to change the representation to allow
   119  // partial sharing between different character classes.
   120  // So the limit is the best we can do.
   121  const (
   122  	maxRunes = 128 << 20 / runeSize
   123  	runeSize = 4 // rune is int32
   124  )
   125  
   126  type parser struct {
   127  	flags       Flags     // parse mode flags
   128  	stack       []*Regexp // stack of parsed expressions
   129  	free        *Regexp
   130  	numCap      int // number of capturing groups seen
   131  	wholeRegexp string
   132  	tmpClass    []rune            // temporary char class work space
   133  	numRegexp   int               // number of regexps allocated
   134  	numRunes    int               // number of runes in char classes
   135  	repeats     int64             // product of all repetitions seen
   136  	height      map[*Regexp]int   // regexp height, for height limit check
   137  	size        map[*Regexp]int64 // regexp compiled size, for size limit check
   138  }
   139  
   140  func (p *parser) newRegexp(op Op) *Regexp {
   141  	re := p.free
   142  	if re != nil {
   143  		p.free = re.Sub0[0]
   144  		*re = Regexp{}
   145  	} else {
   146  		re = new(Regexp)
   147  		p.numRegexp++
   148  	}
   149  	re.Op = op
   150  	return re
   151  }
   152  
   153  func (p *parser) reuse(re *Regexp) {
   154  	if p.height != nil {
   155  		delete(p.height, re)
   156  	}
   157  	re.Sub0[0] = p.free
   158  	p.free = re
   159  }
   160  
   161  func (p *parser) checkLimits(re *Regexp) {
   162  	if p.numRunes > maxRunes {
   163  		panic(ErrLarge)
   164  	}
   165  	p.checkSize(re)
   166  	p.checkHeight(re)
   167  }
   168  
   169  func (p *parser) checkSize(re *Regexp) {
   170  	if p.size == nil {
   171  		// We haven't started tracking size yet.
   172  		// Do a relatively cheap check to see if we need to start.
   173  		// Maintain the product of all the repeats we've seen
   174  		// and don't track if the total number of regexp nodes
   175  		// we've seen times the repeat product is in budget.
   176  		if p.repeats == 0 {
   177  			p.repeats = 1
   178  		}
   179  		if re.Op == OpRepeat {
   180  			n := re.Max
   181  			if n == -1 {
   182  				n = re.Min
   183  			}
   184  			if n <= 0 {
   185  				n = 1
   186  			}
   187  			if int64(n) > maxSize/p.repeats {
   188  				p.repeats = maxSize
   189  			} else {
   190  				p.repeats *= int64(n)
   191  			}
   192  		}
   193  		if int64(p.numRegexp) < maxSize/p.repeats {
   194  			return
   195  		}
   196  
   197  		// We need to start tracking size.
   198  		// Make the map and belatedly populate it
   199  		// with info about everything we've constructed so far.
   200  		p.size = make(map[*Regexp]int64)
   201  		for _, re := range p.stack {
   202  			p.checkSize(re)
   203  		}
   204  	}
   205  
   206  	if p.calcSize(re, true) > maxSize {
   207  		panic(ErrLarge)
   208  	}
   209  }
   210  
   211  func (p *parser) calcSize(re *Regexp, force bool) int64 {
   212  	if !force {
   213  		if size, ok := p.size[re]; ok {
   214  			return size
   215  		}
   216  	}
   217  
   218  	var size int64
   219  	switch re.Op {
   220  	case OpLiteral:
   221  		size = int64(len(re.Rune))
   222  	case OpCapture, OpStar:
   223  		// star can be 1+ or 2+; assume 2 pessimistically
   224  		size = 2 + p.calcSize(re.Sub[0], false)
   225  	case OpPlus, OpQuest:
   226  		size = 1 + p.calcSize(re.Sub[0], false)
   227  	case OpConcat:
   228  		for _, sub := range re.Sub {
   229  			size += p.calcSize(sub, false)
   230  		}
   231  	case OpAlternate:
   232  		for _, sub := range re.Sub {
   233  			size += p.calcSize(sub, false)
   234  		}
   235  		if len(re.Sub) > 1 {
   236  			size += int64(len(re.Sub)) - 1
   237  		}
   238  	case OpRepeat:
   239  		sub := p.calcSize(re.Sub[0], false)
   240  		if re.Max == -1 {
   241  			if re.Min == 0 {
   242  				size = 2 + sub // x*
   243  			} else {
   244  				size = 1 + int64(re.Min)*sub // xxx+
   245  			}
   246  			break
   247  		}
   248  		// x{2,5} = xx(x(x(x)?)?)?
   249  		size = int64(re.Max)*sub + int64(re.Max-re.Min)
   250  	}
   251  
   252  	size = max(1, size)
   253  	p.size[re] = size
   254  	return size
   255  }
   256  
   257  func (p *parser) checkHeight(re *Regexp) {
   258  	if p.numRegexp < maxHeight {
   259  		return
   260  	}
   261  	if p.height == nil {
   262  		p.height = make(map[*Regexp]int)
   263  		for _, re := range p.stack {
   264  			p.checkHeight(re)
   265  		}
   266  	}
   267  	if p.calcHeight(re, true) > maxHeight {
   268  		panic(ErrNestingDepth)
   269  	}
   270  }
   271  
   272  func (p *parser) calcHeight(re *Regexp, force bool) int {
   273  	if !force {
   274  		if h, ok := p.height[re]; ok {
   275  			return h
   276  		}
   277  	}
   278  	h := 1
   279  	for _, sub := range re.Sub {
   280  		hsub := p.calcHeight(sub, false)
   281  		if h < 1+hsub {
   282  			h = 1 + hsub
   283  		}
   284  	}
   285  	p.height[re] = h
   286  	return h
   287  }
   288  
   289  // Parse stack manipulation.
   290  
   291  // push pushes the regexp re onto the parse stack and returns the regexp.
   292  func (p *parser) push(re *Regexp) *Regexp {
   293  	p.numRunes += len(re.Rune)
   294  	if re.Op == OpCharClass && len(re.Rune) == 2 && re.Rune[0] == re.Rune[1] {
   295  		// Single rune.
   296  		if p.maybeConcat(re.Rune[0], p.flags&^FoldCase) {
   297  			return nil
   298  		}
   299  		re.Op = OpLiteral
   300  		re.Rune = re.Rune[:1]
   301  		re.Flags = p.flags &^ FoldCase
   302  	} else if re.Op == OpCharClass && len(re.Rune) == 4 &&
   303  		re.Rune[0] == re.Rune[1] && re.Rune[2] == re.Rune[3] &&
   304  		unicode.SimpleFold(re.Rune[0]) == re.Rune[2] &&
   305  		unicode.SimpleFold(re.Rune[2]) == re.Rune[0] ||
   306  		re.Op == OpCharClass && len(re.Rune) == 2 &&
   307  			re.Rune[0]+1 == re.Rune[1] &&
   308  			unicode.SimpleFold(re.Rune[0]) == re.Rune[1] &&
   309  			unicode.SimpleFold(re.Rune[1]) == re.Rune[0] {
   310  		// Case-insensitive rune like [Aa] or [Δδ].
   311  		if p.maybeConcat(re.Rune[0], p.flags|FoldCase) {
   312  			return nil
   313  		}
   314  
   315  		// Rewrite as (case-insensitive) literal.
   316  		re.Op = OpLiteral
   317  		re.Rune = re.Rune[:1]
   318  		re.Flags = p.flags | FoldCase
   319  	} else {
   320  		// Incremental concatenation.
   321  		p.maybeConcat(-1, 0)
   322  	}
   323  
   324  	p.stack = append(p.stack, re)
   325  	p.checkLimits(re)
   326  	return re
   327  }
   328  
   329  // maybeConcat implements incremental concatenation
   330  // of literal runes into string nodes. The parser calls this
   331  // before each push, so only the top fragment of the stack
   332  // might need processing. Since this is called before a push,
   333  // the topmost literal is no longer subject to operators like *
   334  // (Otherwise ab* would turn into (ab)*.)
   335  // If r >= 0 and there's a node left over, maybeConcat uses it
   336  // to push r with the given flags.
   337  // maybeConcat reports whether r was pushed.
   338  func (p *parser) maybeConcat(r rune, flags Flags) bool {
   339  	n := len(p.stack)
   340  	if n < 2 {
   341  		return false
   342  	}
   343  
   344  	re1 := p.stack[n-1]
   345  	re2 := p.stack[n-2]
   346  	if re1.Op != OpLiteral || re2.Op != OpLiteral || re1.Flags&FoldCase != re2.Flags&FoldCase {
   347  		return false
   348  	}
   349  
   350  	// Push re1 into re2.
   351  	re2.Rune = append(re2.Rune, re1.Rune...)
   352  
   353  	// Reuse re1 if possible.
   354  	if r >= 0 {
   355  		re1.Rune = re1.Rune0[:1]
   356  		re1.Rune[0] = r
   357  		re1.Flags = flags
   358  		return true
   359  	}
   360  
   361  	p.stack = p.stack[:n-1]
   362  	p.reuse(re1)
   363  	return false // did not push r
   364  }
   365  
   366  // literal pushes a literal regexp for the rune r on the stack.
   367  func (p *parser) literal(r rune) {
   368  	re := p.newRegexp(OpLiteral)
   369  	re.Flags = p.flags
   370  	if p.flags&FoldCase != 0 {
   371  		r = minFoldRune(r)
   372  	}
   373  	re.Rune0[0] = r
   374  	re.Rune = re.Rune0[:1]
   375  	p.push(re)
   376  }
   377  
   378  // minFoldRune returns the minimum rune fold-equivalent to r.
   379  func minFoldRune(r rune) rune {
   380  	if r < minFold || r > maxFold {
   381  		return r
   382  	}
   383  	m := r
   384  	r0 := r
   385  	for r = unicode.SimpleFold(r); r != r0; r = unicode.SimpleFold(r) {
   386  		m = min(m, r)
   387  	}
   388  	return m
   389  }
   390  
   391  // op pushes a regexp with the given op onto the stack
   392  // and returns that regexp.
   393  func (p *parser) op(op Op) *Regexp {
   394  	re := p.newRegexp(op)
   395  	re.Flags = p.flags
   396  	return p.push(re)
   397  }
   398  
   399  // repeat replaces the top stack element with itself repeated according to op, min, max.
   400  // before is the regexp suffix starting at the repetition operator.
   401  // after is the regexp suffix following after the repetition operator.
   402  // repeat returns an updated 'after' and an error, if any.
   403  func (p *parser) repeat(op Op, min, max int, before, after, lastRepeat string) (string, error) {
   404  	flags := p.flags
   405  	if p.flags&PerlX != 0 {
   406  		if len(after) > 0 && after[0] == '?' {
   407  			after = after[1:]
   408  			flags ^= NonGreedy
   409  		}
   410  		if lastRepeat != "" {
   411  			// In Perl it is not allowed to stack repetition operators:
   412  			// a** is a syntax error, not a doubled star, and a++ means
   413  			// something else entirely, which we don't support!
   414  			return "", &Error{ErrInvalidRepeatOp, lastRepeat[:len(lastRepeat)-len(after)]}
   415  		}
   416  	}
   417  	n := len(p.stack)
   418  	if n == 0 {
   419  		return "", &Error{ErrMissingRepeatArgument, before[:len(before)-len(after)]}
   420  	}
   421  	sub := p.stack[n-1]
   422  	if sub.Op >= opPseudo {
   423  		return "", &Error{ErrMissingRepeatArgument, before[:len(before)-len(after)]}
   424  	}
   425  
   426  	re := p.newRegexp(op)
   427  	re.Min = min
   428  	re.Max = max
   429  	re.Flags = flags
   430  	re.Sub = re.Sub0[:1]
   431  	re.Sub[0] = sub
   432  	p.stack[n-1] = re
   433  	p.checkLimits(re)
   434  
   435  	if op == OpRepeat && (min >= 2 || max >= 2) && !repeatIsValid(re, 1000) {
   436  		return "", &Error{ErrInvalidRepeatSize, before[:len(before)-len(after)]}
   437  	}
   438  
   439  	return after, nil
   440  }
   441  
   442  // repeatIsValid reports whether the repetition re is valid.
   443  // Valid means that the combination of the top-level repetition
   444  // and any inner repetitions does not exceed n copies of the
   445  // innermost thing.
   446  // This function rewalks the regexp tree and is called for every repetition,
   447  // so we have to worry about inducing quadratic behavior in the parser.
   448  // We avoid this by only calling repeatIsValid when min or max >= 2.
   449  // In that case the depth of any >= 2 nesting can only get to 9 without
   450  // triggering a parse error, so each subtree can only be rewalked 9 times.
   451  func repeatIsValid(re *Regexp, n int) bool {
   452  	if re.Op == OpRepeat {
   453  		m := re.Max
   454  		if m == 0 {
   455  			return true
   456  		}
   457  		if m < 0 {
   458  			m = re.Min
   459  		}
   460  		if m > n {
   461  			return false
   462  		}
   463  		if m > 0 {
   464  			n /= m
   465  		}
   466  	}
   467  	for _, sub := range re.Sub {
   468  		if !repeatIsValid(sub, n) {
   469  			return false
   470  		}
   471  	}
   472  	return true
   473  }
   474  
   475  // concat replaces the top of the stack (above the topmost '|' or '(') with its concatenation.
   476  func (p *parser) concat() *Regexp {
   477  	p.maybeConcat(-1, 0)
   478  
   479  	// Scan down to find pseudo-operator | or (.
   480  	i := len(p.stack)
   481  	for i > 0 && p.stack[i-1].Op < opPseudo {
   482  		i--
   483  	}
   484  	subs := p.stack[i:]
   485  	p.stack = p.stack[:i]
   486  
   487  	// Empty concatenation is special case.
   488  	if len(subs) == 0 {
   489  		return p.push(p.newRegexp(OpEmptyMatch))
   490  	}
   491  
   492  	return p.push(p.collapse(subs, OpConcat))
   493  }
   494  
   495  // alternate replaces the top of the stack (above the topmost '(') with its alternation.
   496  func (p *parser) alternate() *Regexp {
   497  	// Scan down to find pseudo-operator (.
   498  	// There are no | above (.
   499  	i := len(p.stack)
   500  	for i > 0 && p.stack[i-1].Op < opPseudo {
   501  		i--
   502  	}
   503  	subs := p.stack[i:]
   504  	p.stack = p.stack[:i]
   505  
   506  	// Make sure top class is clean.
   507  	// All the others already are (see swapVerticalBar).
   508  	if len(subs) > 0 {
   509  		cleanAlt(subs[len(subs)-1])
   510  	}
   511  
   512  	// Empty alternate is special case
   513  	// (shouldn't happen but easy to handle).
   514  	if len(subs) == 0 {
   515  		return p.push(p.newRegexp(OpNoMatch))
   516  	}
   517  
   518  	return p.push(p.collapse(subs, OpAlternate))
   519  }
   520  
   521  // cleanAlt cleans re for eventual inclusion in an alternation.
   522  func cleanAlt(re *Regexp) {
   523  	switch re.Op {
   524  	case OpCharClass:
   525  		re.Rune = cleanClass(&re.Rune)
   526  		if len(re.Rune) == 2 && re.Rune[0] == 0 && re.Rune[1] == unicode.MaxRune {
   527  			re.Rune = nil
   528  			re.Op = OpAnyChar
   529  			return
   530  		}
   531  		if len(re.Rune) == 4 && re.Rune[0] == 0 && re.Rune[1] == '\n'-1 && re.Rune[2] == '\n'+1 && re.Rune[3] == unicode.MaxRune {
   532  			re.Rune = nil
   533  			re.Op = OpAnyCharNotNL
   534  			return
   535  		}
   536  		if cap(re.Rune)-len(re.Rune) > 100 {
   537  			// re.Rune will not grow any more.
   538  			// Make a copy or inline to reclaim storage.
   539  			re.Rune = append(re.Rune0[:0], re.Rune...)
   540  		}
   541  	}
   542  }
   543  
   544  // collapse returns the result of applying op to sub.
   545  // If sub contains op nodes, they all get hoisted up
   546  // so that there is never a concat of a concat or an
   547  // alternate of an alternate.
   548  func (p *parser) collapse(subs []*Regexp, op Op) *Regexp {
   549  	if len(subs) == 1 {
   550  		return subs[0]
   551  	}
   552  	re := p.newRegexp(op)
   553  	re.Sub = re.Sub0[:0]
   554  	for _, sub := range subs {
   555  		if sub.Op == op {
   556  			re.Sub = append(re.Sub, sub.Sub...)
   557  			p.reuse(sub)
   558  		} else {
   559  			re.Sub = append(re.Sub, sub)
   560  		}
   561  	}
   562  	if op == OpAlternate {
   563  		re.Sub = p.factor(re.Sub)
   564  		if len(re.Sub) == 1 {
   565  			old := re
   566  			re = re.Sub[0]
   567  			p.reuse(old)
   568  		}
   569  	}
   570  	return re
   571  }
   572  
   573  // factor factors common prefixes from the alternation list sub.
   574  // It returns a replacement list that reuses the same storage and
   575  // frees (passes to p.reuse) any removed *Regexps.
   576  //
   577  // For example,
   578  //
   579  //	ABC|ABD|AEF|BCX|BCY
   580  //
   581  // simplifies by literal prefix extraction to
   582  //
   583  //	A(B(C|D)|EF)|BC(X|Y)
   584  //
   585  // which simplifies by character class introduction to
   586  //
   587  //	A(B[CD]|EF)|BC[XY]
   588  func (p *parser) factor(sub []*Regexp) []*Regexp {
   589  	if len(sub) < 2 {
   590  		return sub
   591  	}
   592  
   593  	// Round 1: Factor out common literal prefixes.
   594  	var str []rune
   595  	var strflags Flags
   596  	start := 0
   597  	out := sub[:0]
   598  	for i := 0; i <= len(sub); i++ {
   599  		// Invariant: the Regexps that were in sub[0:start] have been
   600  		// used or marked for reuse, and the slice space has been reused
   601  		// for out (len(out) <= start).
   602  		//
   603  		// Invariant: sub[start:i] consists of regexps that all begin
   604  		// with str as modified by strflags.
   605  		var istr []rune
   606  		var iflags Flags
   607  		if i < len(sub) {
   608  			istr, iflags = p.leadingString(sub[i])
   609  			if iflags == strflags {
   610  				same := 0
   611  				for same < len(str) && same < len(istr) && str[same] == istr[same] {
   612  					same++
   613  				}
   614  				if same > 0 {
   615  					// Matches at least one rune in current range.
   616  					// Keep going around.
   617  					str = str[:same]
   618  					continue
   619  				}
   620  			}
   621  		}
   622  
   623  		// Found end of a run with common leading literal string:
   624  		// sub[start:i] all begin with str[0:len(str)], but sub[i]
   625  		// does not even begin with str[0].
   626  		//
   627  		// Factor out common string and append factored expression to out.
   628  		if i == start {
   629  			// Nothing to do - run of length 0.
   630  		} else if i == start+1 {
   631  			// Just one: don't bother factoring.
   632  			out = append(out, sub[start])
   633  		} else {
   634  			// Construct factored form: prefix(suffix1|suffix2|...)
   635  			prefix := p.newRegexp(OpLiteral)
   636  			prefix.Flags = strflags
   637  			prefix.Rune = append(prefix.Rune[:0], str...)
   638  
   639  			for j := start; j < i; j++ {
   640  				sub[j] = p.removeLeadingString(sub[j], len(str))
   641  				p.checkLimits(sub[j])
   642  			}
   643  			suffix := p.collapse(sub[start:i], OpAlternate) // recurse
   644  
   645  			re := p.newRegexp(OpConcat)
   646  			re.Sub = append(re.Sub[:0], prefix, suffix)
   647  			out = append(out, re)
   648  		}
   649  
   650  		// Prepare for next iteration.
   651  		start = i
   652  		str = istr
   653  		strflags = iflags
   654  	}
   655  	sub = out
   656  
   657  	// Round 2: Factor out common simple prefixes,
   658  	// just the first piece of each concatenation.
   659  	// This will be good enough a lot of the time.
   660  	//
   661  	// Complex subexpressions (e.g. involving quantifiers)
   662  	// are not safe to factor because that collapses their
   663  	// distinct paths through the automaton, which affects
   664  	// correctness in some cases.
   665  	start = 0
   666  	out = sub[:0]
   667  	var first *Regexp
   668  	for i := 0; i <= len(sub); i++ {
   669  		// Invariant: the Regexps that were in sub[0:start] have been
   670  		// used or marked for reuse, and the slice space has been reused
   671  		// for out (len(out) <= start).
   672  		//
   673  		// Invariant: sub[start:i] consists of regexps that all begin with ifirst.
   674  		var ifirst *Regexp
   675  		if i < len(sub) {
   676  			ifirst = p.leadingRegexp(sub[i])
   677  			if first != nil && first.Equal(ifirst) &&
   678  				// first must be a character class OR a fixed repeat of a character class.
   679  				(isCharClass(first) || (first.Op == OpRepeat && first.Min == first.Max && isCharClass(first.Sub[0]))) {
   680  				continue
   681  			}
   682  		}
   683  
   684  		// Found end of a run with common leading regexp:
   685  		// sub[start:i] all begin with first but sub[i] does not.
   686  		//
   687  		// Factor out common regexp and append factored expression to out.
   688  		if i == start {
   689  			// Nothing to do - run of length 0.
   690  		} else if i == start+1 {
   691  			// Just one: don't bother factoring.
   692  			out = append(out, sub[start])
   693  		} else {
   694  			// Construct factored form: prefix(suffix1|suffix2|...)
   695  			prefix := first
   696  			for j := start; j < i; j++ {
   697  				reuse := j != start // prefix came from sub[start]
   698  				sub[j] = p.removeLeadingRegexp(sub[j], reuse)
   699  				p.checkLimits(sub[j])
   700  			}
   701  			suffix := p.collapse(sub[start:i], OpAlternate) // recurse
   702  
   703  			re := p.newRegexp(OpConcat)
   704  			re.Sub = append(re.Sub[:0], prefix, suffix)
   705  			out = append(out, re)
   706  		}
   707  
   708  		// Prepare for next iteration.
   709  		start = i
   710  		first = ifirst
   711  	}
   712  	sub = out
   713  
   714  	// Round 3: Collapse runs of single literals into character classes.
   715  	start = 0
   716  	out = sub[:0]
   717  	for i := 0; i <= len(sub); i++ {
   718  		// Invariant: the Regexps that were in sub[0:start] have been
   719  		// used or marked for reuse, and the slice space has been reused
   720  		// for out (len(out) <= start).
   721  		//
   722  		// Invariant: sub[start:i] consists of regexps that are either
   723  		// literal runes or character classes.
   724  		if i < len(sub) && isCharClass(sub[i]) {
   725  			continue
   726  		}
   727  
   728  		// sub[i] is not a char or char class;
   729  		// emit char class for sub[start:i]...
   730  		if i == start {
   731  			// Nothing to do - run of length 0.
   732  		} else if i == start+1 {
   733  			out = append(out, sub[start])
   734  		} else {
   735  			// Make new char class.
   736  			// Start with most complex regexp in sub[start].
   737  			max := start
   738  			for j := start + 1; j < i; j++ {
   739  				if sub[max].Op < sub[j].Op || sub[max].Op == sub[j].Op && len(sub[max].Rune) < len(sub[j].Rune) {
   740  					max = j
   741  				}
   742  			}
   743  			sub[start], sub[max] = sub[max], sub[start]
   744  
   745  			for j := start + 1; j < i; j++ {
   746  				mergeCharClass(sub[start], sub[j])
   747  				p.reuse(sub[j])
   748  			}
   749  			cleanAlt(sub[start])
   750  			out = append(out, sub[start])
   751  		}
   752  
   753  		// ... and then emit sub[i].
   754  		if i < len(sub) {
   755  			out = append(out, sub[i])
   756  		}
   757  		start = i + 1
   758  	}
   759  	sub = out
   760  
   761  	// Round 4: Collapse runs of empty matches into a single empty match.
   762  	start = 0
   763  	out = sub[:0]
   764  	for i := range sub {
   765  		if i+1 < len(sub) && sub[i].Op == OpEmptyMatch && sub[i+1].Op == OpEmptyMatch {
   766  			continue
   767  		}
   768  		out = append(out, sub[i])
   769  	}
   770  	sub = out
   771  
   772  	return sub
   773  }
   774  
   775  // leadingString returns the leading literal string that re begins with.
   776  // The string refers to storage in re or its children.
   777  func (p *parser) leadingString(re *Regexp) ([]rune, Flags) {
   778  	if re.Op == OpConcat && len(re.Sub) > 0 {
   779  		re = re.Sub[0]
   780  	}
   781  	if re.Op != OpLiteral {
   782  		return nil, 0
   783  	}
   784  	return re.Rune, re.Flags & FoldCase
   785  }
   786  
   787  // removeLeadingString removes the first n leading runes
   788  // from the beginning of re. It returns the replacement for re.
   789  func (p *parser) removeLeadingString(re *Regexp, n int) *Regexp {
   790  	if re.Op == OpConcat && len(re.Sub) > 0 {
   791  		// Removing a leading string in a concatenation
   792  		// might simplify the concatenation.
   793  		sub := re.Sub[0]
   794  		sub = p.removeLeadingString(sub, n)
   795  		re.Sub[0] = sub
   796  		if sub.Op == OpEmptyMatch {
   797  			p.reuse(sub)
   798  			switch len(re.Sub) {
   799  			case 0, 1:
   800  				// Impossible but handle.
   801  				re.Op = OpEmptyMatch
   802  				re.Sub = nil
   803  			case 2:
   804  				old := re
   805  				re = re.Sub[1]
   806  				p.reuse(old)
   807  			default:
   808  				copy(re.Sub, re.Sub[1:])
   809  				re.Sub = re.Sub[:len(re.Sub)-1]
   810  			}
   811  		}
   812  		return re
   813  	}
   814  
   815  	if re.Op == OpLiteral {
   816  		re.Rune = re.Rune[:copy(re.Rune, re.Rune[n:])]
   817  		if len(re.Rune) == 0 {
   818  			re.Op = OpEmptyMatch
   819  		}
   820  	}
   821  	return re
   822  }
   823  
   824  // leadingRegexp returns the leading regexp that re begins with.
   825  // The regexp refers to storage in re or its children.
   826  func (p *parser) leadingRegexp(re *Regexp) *Regexp {
   827  	if re.Op == OpEmptyMatch {
   828  		return nil
   829  	}
   830  	if re.Op == OpConcat && len(re.Sub) > 0 {
   831  		sub := re.Sub[0]
   832  		if sub.Op == OpEmptyMatch {
   833  			return nil
   834  		}
   835  		return sub
   836  	}
   837  	return re
   838  }
   839  
   840  // removeLeadingRegexp removes the leading regexp in re.
   841  // It returns the replacement for re.
   842  // If reuse is true, it passes the removed regexp (if no longer needed) to p.reuse.
   843  func (p *parser) removeLeadingRegexp(re *Regexp, reuse bool) *Regexp {
   844  	if re.Op == OpConcat && len(re.Sub) > 0 {
   845  		if reuse {
   846  			p.reuse(re.Sub[0])
   847  		}
   848  		re.Sub = re.Sub[:copy(re.Sub, re.Sub[1:])]
   849  		switch len(re.Sub) {
   850  		case 0:
   851  			re.Op = OpEmptyMatch
   852  			re.Sub = nil
   853  		case 1:
   854  			old := re
   855  			re = re.Sub[0]
   856  			p.reuse(old)
   857  		}
   858  		return re
   859  	}
   860  	if reuse {
   861  		p.reuse(re)
   862  	}
   863  	return p.newRegexp(OpEmptyMatch)
   864  }
   865  
   866  func literalRegexp(s string, flags Flags) *Regexp {
   867  	re := &Regexp{Op: OpLiteral}
   868  	re.Flags = flags
   869  	re.Rune = re.Rune0[:0] // use local storage for small strings
   870  	for _, c := range s {
   871  		if len(re.Rune) >= cap(re.Rune) {
   872  			// string is too long to fit in Rune0.  let Go handle it
   873  			re.Rune = []rune(s)
   874  			break
   875  		}
   876  		re.Rune = append(re.Rune, c)
   877  	}
   878  	return re
   879  }
   880  
   881  // Parsing.
   882  
   883  // Parse parses a regular expression string s, controlled by the specified
   884  // Flags, and returns a regular expression parse tree. The syntax is
   885  // described in the top-level comment.
   886  func Parse(s string, flags Flags) (*Regexp, error) {
   887  	return parse(s, flags)
   888  }
   889  
   890  func parse(s string, flags Flags) (_ *Regexp, err error) {
   891  	defer func() {
   892  		switch r := recover(); r {
   893  		default:
   894  			panic(r)
   895  		case nil:
   896  			// ok
   897  		case ErrLarge: // too big
   898  			err = &Error{Code: ErrLarge, Expr: s}
   899  		case ErrNestingDepth:
   900  			err = &Error{Code: ErrNestingDepth, Expr: s}
   901  		}
   902  	}()
   903  
   904  	if flags&Literal != 0 {
   905  		// Trivial parser for literal string.
   906  		if err := checkUTF8(s); err != nil {
   907  			return nil, err
   908  		}
   909  		return literalRegexp(s, flags), nil
   910  	}
   911  
   912  	// Otherwise, must do real work.
   913  	var (
   914  		p          parser
   915  		c          rune
   916  		op         Op
   917  		lastRepeat string
   918  	)
   919  	p.flags = flags
   920  	p.wholeRegexp = s
   921  	t := s
   922  	for t != "" {
   923  		repeat := ""
   924  	BigSwitch:
   925  		switch t[0] {
   926  		default:
   927  			if c, t, err = nextRune(t); err != nil {
   928  				return nil, err
   929  			}
   930  			p.literal(c)
   931  
   932  		case '(':
   933  			if p.flags&PerlX != 0 && len(t) >= 2 && t[1] == '?' {
   934  				// Flag changes and non-capturing groups.
   935  				if t, err = p.parsePerlFlags(t); err != nil {
   936  					return nil, err
   937  				}
   938  				break
   939  			}
   940  			p.numCap++
   941  			p.op(opLeftParen).Cap = p.numCap
   942  			t = t[1:]
   943  		case '|':
   944  			p.parseVerticalBar()
   945  			t = t[1:]
   946  		case ')':
   947  			if err = p.parseRightParen(); err != nil {
   948  				return nil, err
   949  			}
   950  			t = t[1:]
   951  		case '^':
   952  			if p.flags&OneLine != 0 {
   953  				p.op(OpBeginText)
   954  			} else {
   955  				p.op(OpBeginLine)
   956  			}
   957  			t = t[1:]
   958  		case '$':
   959  			if p.flags&OneLine != 0 {
   960  				p.op(OpEndText).Flags |= WasDollar
   961  			} else {
   962  				p.op(OpEndLine)
   963  			}
   964  			t = t[1:]
   965  		case '.':
   966  			if p.flags&DotNL != 0 {
   967  				p.op(OpAnyChar)
   968  			} else {
   969  				p.op(OpAnyCharNotNL)
   970  			}
   971  			t = t[1:]
   972  		case '[':
   973  			if t, err = p.parseClass(t); err != nil {
   974  				return nil, err
   975  			}
   976  		case '*', '+', '?':
   977  			before := t
   978  			switch t[0] {
   979  			case '*':
   980  				op = OpStar
   981  			case '+':
   982  				op = OpPlus
   983  			case '?':
   984  				op = OpQuest
   985  			}
   986  			after := t[1:]
   987  			if after, err = p.repeat(op, 0, 0, before, after, lastRepeat); err != nil {
   988  				return nil, err
   989  			}
   990  			repeat = before
   991  			t = after
   992  		case '{':
   993  			op = OpRepeat
   994  			before := t
   995  			min, max, after, ok := p.parseRepeat(t)
   996  			if !ok {
   997  				// If the repeat cannot be parsed, { is a literal.
   998  				p.literal('{')
   999  				t = t[1:]
  1000  				break
  1001  			}
  1002  			if min < 0 || min > 1000 || max > 1000 || max >= 0 && min > max {
  1003  				// Numbers were too big, or max is present and min > max.
  1004  				return nil, &Error{ErrInvalidRepeatSize, before[:len(before)-len(after)]}
  1005  			}
  1006  			if after, err = p.repeat(op, min, max, before, after, lastRepeat); err != nil {
  1007  				return nil, err
  1008  			}
  1009  			repeat = before
  1010  			t = after
  1011  		case '\\':
  1012  			if p.flags&PerlX != 0 && len(t) >= 2 {
  1013  				switch t[1] {
  1014  				case 'A':
  1015  					p.op(OpBeginText)
  1016  					t = t[2:]
  1017  					break BigSwitch
  1018  				case 'b':
  1019  					p.op(OpWordBoundary)
  1020  					t = t[2:]
  1021  					break BigSwitch
  1022  				case 'B':
  1023  					p.op(OpNoWordBoundary)
  1024  					t = t[2:]
  1025  					break BigSwitch
  1026  				case 'C':
  1027  					// any byte; not supported
  1028  					return nil, &Error{ErrInvalidEscape, t[:2]}
  1029  				case 'Q':
  1030  					// \Q ... \E: the ... is always literals
  1031  					var lit string
  1032  					lit, t, _ = strings.Cut(t[2:], `\E`)
  1033  					for lit != "" {
  1034  						c, rest, err := nextRune(lit)
  1035  						if err != nil {
  1036  							return nil, err
  1037  						}
  1038  						p.literal(c)
  1039  						lit = rest
  1040  					}
  1041  					break BigSwitch
  1042  				case 'z':
  1043  					p.op(OpEndText)
  1044  					t = t[2:]
  1045  					break BigSwitch
  1046  				}
  1047  			}
  1048  
  1049  			re := p.newRegexp(OpCharClass)
  1050  			re.Flags = p.flags
  1051  
  1052  			// Look for Unicode character group like \p{Han}
  1053  			if len(t) >= 2 && (t[1] == 'p' || t[1] == 'P') {
  1054  				r, rest, err := p.parseUnicodeClass(t, re.Rune0[:0])
  1055  				if err != nil {
  1056  					return nil, err
  1057  				}
  1058  				if r != nil {
  1059  					re.Rune = r
  1060  					t = rest
  1061  					p.push(re)
  1062  					break BigSwitch
  1063  				}
  1064  			}
  1065  
  1066  			// Perl character class escape.
  1067  			if r, rest := p.parsePerlClassEscape(t, re.Rune0[:0]); r != nil {
  1068  				re.Rune = r
  1069  				t = rest
  1070  				p.push(re)
  1071  				break BigSwitch
  1072  			}
  1073  			p.reuse(re)
  1074  
  1075  			// Ordinary single-character escape.
  1076  			if c, t, err = p.parseEscape(t); err != nil {
  1077  				return nil, err
  1078  			}
  1079  			p.literal(c)
  1080  		}
  1081  		lastRepeat = repeat
  1082  	}
  1083  
  1084  	p.concat()
  1085  	if p.swapVerticalBar() {
  1086  		// pop vertical bar
  1087  		p.stack = p.stack[:len(p.stack)-1]
  1088  	}
  1089  	p.alternate()
  1090  
  1091  	n := len(p.stack)
  1092  	if n != 1 {
  1093  		return nil, &Error{ErrMissingParen, s}
  1094  	}
  1095  	return p.stack[0], nil
  1096  }
  1097  
  1098  // parseRepeat parses {min} (max=min) or {min,} (max=-1) or {min,max}.
  1099  // If s is not of that form, it returns ok == false.
  1100  // If s has the right form but the values are too big, it returns min == -1, ok == true.
  1101  func (p *parser) parseRepeat(s string) (min, max int, rest string, ok bool) {
  1102  	if s == "" || s[0] != '{' {
  1103  		return
  1104  	}
  1105  	s = s[1:]
  1106  	var ok1 bool
  1107  	if min, s, ok1 = p.parseInt(s); !ok1 {
  1108  		return
  1109  	}
  1110  	if s == "" {
  1111  		return
  1112  	}
  1113  	if s[0] != ',' {
  1114  		max = min
  1115  	} else {
  1116  		s = s[1:]
  1117  		if s == "" {
  1118  			return
  1119  		}
  1120  		if s[0] == '}' {
  1121  			max = -1
  1122  		} else if max, s, ok1 = p.parseInt(s); !ok1 {
  1123  			return
  1124  		} else if max < 0 {
  1125  			// parseInt found too big a number
  1126  			min = -1
  1127  		}
  1128  	}
  1129  	if s == "" || s[0] != '}' {
  1130  		return
  1131  	}
  1132  	rest = s[1:]
  1133  	ok = true
  1134  	return
  1135  }
  1136  
  1137  // parsePerlFlags parses a Perl flag setting or non-capturing group or both,
  1138  // like (?i) or (?: or (?i:.  It removes the prefix from s and updates the parse state.
  1139  // The caller must have ensured that s begins with "(?".
  1140  func (p *parser) parsePerlFlags(s string) (rest string, err error) {
  1141  	t := s
  1142  
  1143  	// Check for named captures, first introduced in Python's regexp library.
  1144  	// As usual, there are three slightly different syntaxes:
  1145  	//
  1146  	//   (?P<name>expr)   the original, introduced by Python
  1147  	//   (?<name>expr)    the .NET alteration, adopted by Perl 5.10
  1148  	//   (?'name'expr)    another .NET alteration, adopted by Perl 5.10
  1149  	//
  1150  	// Perl 5.10 gave in and implemented the Python version too,
  1151  	// but they claim that the last two are the preferred forms.
  1152  	// PCRE and languages based on it (specifically, PHP and Ruby)
  1153  	// support all three as well. EcmaScript 4 uses only the Python form.
  1154  	//
  1155  	// In both the open source world (via Code Search) and the
  1156  	// Google source tree, (?P<expr>name) and (?<expr>name) are the
  1157  	// dominant forms of named captures and both are supported.
  1158  	startsWithP := len(t) > 4 && t[2] == 'P' && t[3] == '<'
  1159  	startsWithName := len(t) > 3 && t[2] == '<'
  1160  
  1161  	if startsWithP || startsWithName {
  1162  		// position of expr start
  1163  		exprStartPos := 4
  1164  		if startsWithName {
  1165  			exprStartPos = 3
  1166  		}
  1167  
  1168  		// Pull out name.
  1169  		end := strings.IndexRune(t, '>')
  1170  		if end < 0 {
  1171  			if err = checkUTF8(t); err != nil {
  1172  				return "", err
  1173  			}
  1174  			return "", &Error{ErrInvalidNamedCapture, s}
  1175  		}
  1176  
  1177  		capture := t[:end+1]        // "(?P<name>" or "(?<name>"
  1178  		name := t[exprStartPos:end] // "name"
  1179  		if err = checkUTF8(name); err != nil {
  1180  			return "", err
  1181  		}
  1182  		if !isValidCaptureName(name) {
  1183  			return "", &Error{ErrInvalidNamedCapture, capture}
  1184  		}
  1185  
  1186  		// Like ordinary capture, but named.
  1187  		p.numCap++
  1188  		re := p.op(opLeftParen)
  1189  		re.Cap = p.numCap
  1190  		re.Name = name
  1191  		return t[end+1:], nil
  1192  	}
  1193  
  1194  	// Non-capturing group. Might also twiddle Perl flags.
  1195  	var c rune
  1196  	t = t[2:] // skip (?
  1197  	flags := p.flags
  1198  	sign := +1
  1199  	sawFlag := false
  1200  Loop:
  1201  	for t != "" {
  1202  		if c, t, err = nextRune(t); err != nil {
  1203  			return "", err
  1204  		}
  1205  		switch c {
  1206  		default:
  1207  			break Loop
  1208  
  1209  		// Flags.
  1210  		case 'i':
  1211  			flags |= FoldCase
  1212  			sawFlag = true
  1213  		case 'm':
  1214  			flags &^= OneLine
  1215  			sawFlag = true
  1216  		case 's':
  1217  			flags |= DotNL
  1218  			sawFlag = true
  1219  		case 'U':
  1220  			flags |= NonGreedy
  1221  			sawFlag = true
  1222  
  1223  		// Switch to negation.
  1224  		case '-':
  1225  			if sign < 0 {
  1226  				break Loop
  1227  			}
  1228  			sign = -1
  1229  			// Invert flags so that | above turn into &^ and vice versa.
  1230  			// We'll invert flags again before using it below.
  1231  			flags = ^flags
  1232  			sawFlag = false
  1233  
  1234  		// End of flags, starting group or not.
  1235  		case ':', ')':
  1236  			if sign < 0 {
  1237  				if !sawFlag {
  1238  					break Loop
  1239  				}
  1240  				flags = ^flags
  1241  			}
  1242  			if c == ':' {
  1243  				// Open new group
  1244  				p.op(opLeftParen)
  1245  			}
  1246  			p.flags = flags
  1247  			return t, nil
  1248  		}
  1249  	}
  1250  
  1251  	return "", &Error{ErrInvalidPerlOp, s[:len(s)-len(t)]}
  1252  }
  1253  
  1254  // isValidCaptureName reports whether name
  1255  // is a valid capture name: [A-Za-z0-9_]+.
  1256  // PCRE limits names to 32 bytes.
  1257  // Python rejects names starting with digits.
  1258  // We don't enforce either of those.
  1259  func isValidCaptureName(name string) bool {
  1260  	if name == "" {
  1261  		return false
  1262  	}
  1263  	for _, c := range name {
  1264  		if c != '_' && !isalnum(c) {
  1265  			return false
  1266  		}
  1267  	}
  1268  	return true
  1269  }
  1270  
  1271  // parseInt parses a decimal integer.
  1272  func (p *parser) parseInt(s string) (n int, rest string, ok bool) {
  1273  	if s == "" || s[0] < '0' || '9' < s[0] {
  1274  		return
  1275  	}
  1276  	// Disallow leading zeros.
  1277  	if len(s) >= 2 && s[0] == '0' && '0' <= s[1] && s[1] <= '9' {
  1278  		return
  1279  	}
  1280  	t := s
  1281  	for s != "" && '0' <= s[0] && s[0] <= '9' {
  1282  		s = s[1:]
  1283  	}
  1284  	rest = s
  1285  	ok = true
  1286  	// Have digits, compute value.
  1287  	t = t[:len(t)-len(s)]
  1288  	for i := 0; i < len(t); i++ {
  1289  		// Avoid overflow.
  1290  		if n >= 1e8 {
  1291  			n = -1
  1292  			break
  1293  		}
  1294  		n = n*10 + int(t[i]) - '0'
  1295  	}
  1296  	return
  1297  }
  1298  
  1299  // can this be represented as a character class?
  1300  // single-rune literal string, char class, ., and .|\n.
  1301  func isCharClass(re *Regexp) bool {
  1302  	return re.Op == OpLiteral && len(re.Rune) == 1 ||
  1303  		re.Op == OpCharClass ||
  1304  		re.Op == OpAnyCharNotNL ||
  1305  		re.Op == OpAnyChar
  1306  }
  1307  
  1308  // does re match r?
  1309  func matchRune(re *Regexp, r rune) bool {
  1310  	switch re.Op {
  1311  	case OpLiteral:
  1312  		return len(re.Rune) == 1 && re.Rune[0] == r
  1313  	case OpCharClass:
  1314  		for i := 0; i < len(re.Rune); i += 2 {
  1315  			if re.Rune[i] <= r && r <= re.Rune[i+1] {
  1316  				return true
  1317  			}
  1318  		}
  1319  		return false
  1320  	case OpAnyCharNotNL:
  1321  		return r != '\n'
  1322  	case OpAnyChar:
  1323  		return true
  1324  	}
  1325  	return false
  1326  }
  1327  
  1328  // parseVerticalBar handles a | in the input.
  1329  func (p *parser) parseVerticalBar() {
  1330  	p.concat()
  1331  
  1332  	// The concatenation we just parsed is on top of the stack.
  1333  	// If it sits above an opVerticalBar, swap it below
  1334  	// (things below an opVerticalBar become an alternation).
  1335  	// Otherwise, push a new vertical bar.
  1336  	if !p.swapVerticalBar() {
  1337  		p.op(opVerticalBar)
  1338  	}
  1339  }
  1340  
  1341  // mergeCharClass makes dst = dst|src.
  1342  // The caller must ensure that dst.Op >= src.Op,
  1343  // to reduce the amount of copying.
  1344  func mergeCharClass(dst, src *Regexp) {
  1345  	switch dst.Op {
  1346  	case OpAnyChar:
  1347  		// src doesn't add anything.
  1348  	case OpAnyCharNotNL:
  1349  		// src might add \n
  1350  		if matchRune(src, '\n') {
  1351  			dst.Op = OpAnyChar
  1352  		}
  1353  	case OpCharClass:
  1354  		// src is simpler, so either literal or char class
  1355  		if src.Op == OpLiteral {
  1356  			dst.Rune = appendLiteral(dst.Rune, src.Rune[0], src.Flags)
  1357  		} else {
  1358  			dst.Rune = appendClass(dst.Rune, src.Rune)
  1359  		}
  1360  	case OpLiteral:
  1361  		// both literal
  1362  		if src.Rune[0] == dst.Rune[0] && src.Flags == dst.Flags {
  1363  			break
  1364  		}
  1365  		dst.Op = OpCharClass
  1366  		dst.Rune = appendLiteral(dst.Rune[:0], dst.Rune[0], dst.Flags)
  1367  		dst.Rune = appendLiteral(dst.Rune, src.Rune[0], src.Flags)
  1368  	}
  1369  }
  1370  
  1371  // If the top of the stack is an element followed by an opVerticalBar
  1372  // swapVerticalBar swaps the two and returns true.
  1373  // Otherwise it returns false.
  1374  func (p *parser) swapVerticalBar() bool {
  1375  	// If above and below vertical bar are literal or char class,
  1376  	// can merge into a single char class.
  1377  	n := len(p.stack)
  1378  	if n >= 3 && p.stack[n-2].Op == opVerticalBar && isCharClass(p.stack[n-1]) && isCharClass(p.stack[n-3]) {
  1379  		re1 := p.stack[n-1]
  1380  		re3 := p.stack[n-3]
  1381  		// Make re3 the more complex of the two.
  1382  		if re1.Op > re3.Op {
  1383  			re1, re3 = re3, re1
  1384  			p.stack[n-3] = re3
  1385  		}
  1386  		mergeCharClass(re3, re1)
  1387  		p.reuse(re1)
  1388  		p.stack = p.stack[:n-1]
  1389  		return true
  1390  	}
  1391  
  1392  	if n >= 2 {
  1393  		re1 := p.stack[n-1]
  1394  		re2 := p.stack[n-2]
  1395  		if re2.Op == opVerticalBar {
  1396  			if n >= 3 {
  1397  				// Now out of reach.
  1398  				// Clean opportunistically.
  1399  				cleanAlt(p.stack[n-3])
  1400  			}
  1401  			p.stack[n-2] = re1
  1402  			p.stack[n-1] = re2
  1403  			return true
  1404  		}
  1405  	}
  1406  	return false
  1407  }
  1408  
  1409  // parseRightParen handles a ) in the input.
  1410  func (p *parser) parseRightParen() error {
  1411  	p.concat()
  1412  	if p.swapVerticalBar() {
  1413  		// pop vertical bar
  1414  		p.stack = p.stack[:len(p.stack)-1]
  1415  	}
  1416  	p.alternate()
  1417  
  1418  	n := len(p.stack)
  1419  	if n < 2 {
  1420  		return &Error{ErrUnexpectedParen, p.wholeRegexp}
  1421  	}
  1422  	re1 := p.stack[n-1]
  1423  	re2 := p.stack[n-2]
  1424  	p.stack = p.stack[:n-2]
  1425  	if re2.Op != opLeftParen {
  1426  		return &Error{ErrUnexpectedParen, p.wholeRegexp}
  1427  	}
  1428  	// Restore flags at time of paren.
  1429  	p.flags = re2.Flags
  1430  	if re2.Cap == 0 {
  1431  		// Just for grouping.
  1432  		p.push(re1)
  1433  	} else {
  1434  		re2.Op = OpCapture
  1435  		re2.Sub = re2.Sub0[:1]
  1436  		re2.Sub[0] = re1
  1437  		p.push(re2)
  1438  	}
  1439  	return nil
  1440  }
  1441  
  1442  // parseEscape parses an escape sequence at the beginning of s
  1443  // and returns the rune.
  1444  func (p *parser) parseEscape(s string) (r rune, rest string, err error) {
  1445  	t := s[1:]
  1446  	if t == "" {
  1447  		return 0, "", &Error{ErrTrailingBackslash, ""}
  1448  	}
  1449  	c, t, err := nextRune(t)
  1450  	if err != nil {
  1451  		return 0, "", err
  1452  	}
  1453  
  1454  Switch:
  1455  	switch c {
  1456  	default:
  1457  		if c < utf8.RuneSelf && !isalnum(c) {
  1458  			// Escaped non-word characters are always themselves.
  1459  			// PCRE is not quite so rigorous: it accepts things like
  1460  			// \q, but we don't. We once rejected \_, but too many
  1461  			// programs and people insist on using it, so allow \_.
  1462  			return c, t, nil
  1463  		}
  1464  
  1465  	// Octal escapes.
  1466  	case '1', '2', '3', '4', '5', '6', '7':
  1467  		// Single non-zero digit is a backreference; not supported
  1468  		if t == "" || t[0] < '0' || t[0] > '7' {
  1469  			break
  1470  		}
  1471  		fallthrough
  1472  	case '0':
  1473  		// Consume up to three octal digits; already have one.
  1474  		r = c - '0'
  1475  		for i := 1; i < 3; i++ {
  1476  			if t == "" || t[0] < '0' || t[0] > '7' {
  1477  				break
  1478  			}
  1479  			r = r*8 + rune(t[0]) - '0'
  1480  			t = t[1:]
  1481  		}
  1482  		return r, t, nil
  1483  
  1484  	// Hexadecimal escapes.
  1485  	case 'x':
  1486  		if t == "" {
  1487  			break
  1488  		}
  1489  		if c, t, err = nextRune(t); err != nil {
  1490  			return 0, "", err
  1491  		}
  1492  		if c == '{' {
  1493  			// Any number of digits in braces.
  1494  			// Perl accepts any text at all; it ignores all text
  1495  			// after the first non-hex digit. We require only hex digits,
  1496  			// and at least one.
  1497  			nhex := 0
  1498  			r = 0
  1499  			for {
  1500  				if t == "" {
  1501  					break Switch
  1502  				}
  1503  				if c, t, err = nextRune(t); err != nil {
  1504  					return 0, "", err
  1505  				}
  1506  				if c == '}' {
  1507  					break
  1508  				}
  1509  				v := unhex(c)
  1510  				if v < 0 {
  1511  					break Switch
  1512  				}
  1513  				r = r*16 + v
  1514  				if r > unicode.MaxRune {
  1515  					break Switch
  1516  				}
  1517  				nhex++
  1518  			}
  1519  			if nhex == 0 {
  1520  				break Switch
  1521  			}
  1522  			return r, t, nil
  1523  		}
  1524  
  1525  		// Easy case: two hex digits.
  1526  		x := unhex(c)
  1527  		if c, t, err = nextRune(t); err != nil {
  1528  			return 0, "", err
  1529  		}
  1530  		y := unhex(c)
  1531  		if x < 0 || y < 0 {
  1532  			break
  1533  		}
  1534  		return x*16 + y, t, nil
  1535  
  1536  	// C escapes. There is no case 'b', to avoid misparsing
  1537  	// the Perl word-boundary \b as the C backspace \b
  1538  	// when in POSIX mode. In Perl, /\b/ means word-boundary
  1539  	// but /[\b]/ means backspace. We don't support that.
  1540  	// If you want a backspace, embed a literal backspace
  1541  	// character or use \x08.
  1542  	case 'a':
  1543  		return '\a', t, err
  1544  	case 'f':
  1545  		return '\f', t, err
  1546  	case 'n':
  1547  		return '\n', t, err
  1548  	case 'r':
  1549  		return '\r', t, err
  1550  	case 't':
  1551  		return '\t', t, err
  1552  	case 'v':
  1553  		return '\v', t, err
  1554  	}
  1555  	return 0, "", &Error{ErrInvalidEscape, s[:len(s)-len(t)]}
  1556  }
  1557  
  1558  // parseClassChar parses a character class character at the beginning of s
  1559  // and returns it.
  1560  func (p *parser) parseClassChar(s, wholeClass string) (r rune, rest string, err error) {
  1561  	if s == "" {
  1562  		return 0, "", &Error{Code: ErrMissingBracket, Expr: wholeClass}
  1563  	}
  1564  
  1565  	// Allow regular escape sequences even though
  1566  	// many need not be escaped in this context.
  1567  	if s[0] == '\\' {
  1568  		return p.parseEscape(s)
  1569  	}
  1570  
  1571  	return nextRune(s)
  1572  }
  1573  
  1574  type charGroup struct {
  1575  	sign  int
  1576  	class []rune
  1577  }
  1578  
  1579  //go:generate perl make_perl_groups.pl perl_groups.go
  1580  
  1581  // parsePerlClassEscape parses a leading Perl character class escape like \d
  1582  // from the beginning of s. If one is present, it appends the characters to r
  1583  // and returns the new slice r and the remainder of the string.
  1584  func (p *parser) parsePerlClassEscape(s string, r []rune) (out []rune, rest string) {
  1585  	if p.flags&PerlX == 0 || len(s) < 2 || s[0] != '\\' {
  1586  		return
  1587  	}
  1588  	g := perlGroup[s[0:2]]
  1589  	if g.sign == 0 {
  1590  		return
  1591  	}
  1592  	return p.appendGroup(r, g), s[2:]
  1593  }
  1594  
  1595  // parseNamedClass parses a leading POSIX named character class like [:alnum:]
  1596  // from the beginning of s. If one is present, it appends the characters to r
  1597  // and returns the new slice r and the remainder of the string.
  1598  func (p *parser) parseNamedClass(s string, r []rune) (out []rune, rest string, err error) {
  1599  	if len(s) < 2 || s[0] != '[' || s[1] != ':' {
  1600  		return
  1601  	}
  1602  
  1603  	i := strings.Index(s[2:], ":]")
  1604  	if i < 0 {
  1605  		return
  1606  	}
  1607  	i += 2
  1608  	name, s := s[0:i+2], s[i+2:]
  1609  	g := posixGroup[name]
  1610  	if g.sign == 0 {
  1611  		return nil, "", &Error{ErrInvalidCharRange, name}
  1612  	}
  1613  	return p.appendGroup(r, g), s, nil
  1614  }
  1615  
  1616  func (p *parser) appendGroup(r []rune, g charGroup) []rune {
  1617  	if p.flags&FoldCase == 0 {
  1618  		if g.sign < 0 {
  1619  			r = appendNegatedClass(r, g.class)
  1620  		} else {
  1621  			r = appendClass(r, g.class)
  1622  		}
  1623  	} else {
  1624  		tmp := p.tmpClass[:0]
  1625  		tmp = appendFoldedClass(tmp, g.class)
  1626  		p.tmpClass = tmp
  1627  		tmp = cleanClass(&p.tmpClass)
  1628  		if g.sign < 0 {
  1629  			r = appendNegatedClass(r, tmp)
  1630  		} else {
  1631  			r = appendClass(r, tmp)
  1632  		}
  1633  	}
  1634  	return r
  1635  }
  1636  
  1637  var anyTable = &unicode.RangeTable{
  1638  	R16: []unicode.Range16{{Lo: 0, Hi: 1<<16 - 1, Stride: 1}},
  1639  	R32: []unicode.Range32{{Lo: 1 << 16, Hi: unicode.MaxRune, Stride: 1}},
  1640  }
  1641  
  1642  // unicodeTable returns the unicode.RangeTable identified by name
  1643  // and the table of additional fold-equivalent code points.
  1644  func unicodeTable(name string) (*unicode.RangeTable, *unicode.RangeTable) {
  1645  	// Special case: "Any" means any.
  1646  	if name == "Any" {
  1647  		return anyTable, anyTable
  1648  	}
  1649  	if t := unicode.Categories[name]; t != nil {
  1650  		return t, unicode.FoldCategory[name]
  1651  	}
  1652  	if t := unicode.Scripts[name]; t != nil {
  1653  		return t, unicode.FoldScript[name]
  1654  	}
  1655  	return nil, nil
  1656  }
  1657  
  1658  // parseUnicodeClass parses a leading Unicode character class like \p{Han}
  1659  // from the beginning of s. If one is present, it appends the characters to r
  1660  // and returns the new slice r and the remainder of the string.
  1661  func (p *parser) parseUnicodeClass(s string, r []rune) (out []rune, rest string, err error) {
  1662  	if p.flags&UnicodeGroups == 0 || len(s) < 2 || s[0] != '\\' || s[1] != 'p' && s[1] != 'P' {
  1663  		return
  1664  	}
  1665  
  1666  	// Committed to parse or return error.
  1667  	sign := +1
  1668  	if s[1] == 'P' {
  1669  		sign = -1
  1670  	}
  1671  	t := s[2:]
  1672  	c, t, err := nextRune(t)
  1673  	if err != nil {
  1674  		return
  1675  	}
  1676  	var seq, name string
  1677  	if c != '{' {
  1678  		// Single-letter name.
  1679  		seq = s[:len(s)-len(t)]
  1680  		name = seq[2:]
  1681  	} else {
  1682  		// Name is in braces.
  1683  		end := strings.IndexRune(s, '}')
  1684  		if end < 0 {
  1685  			if err = checkUTF8(s); err != nil {
  1686  				return
  1687  			}
  1688  			return nil, "", &Error{ErrInvalidCharRange, s}
  1689  		}
  1690  		seq, t = s[:end+1], s[end+1:]
  1691  		name = s[3:end]
  1692  		if err = checkUTF8(name); err != nil {
  1693  			return
  1694  		}
  1695  	}
  1696  
  1697  	// Group can have leading negation too.  \p{^Han} == \P{Han}, \P{^Han} == \p{Han}.
  1698  	if name != "" && name[0] == '^' {
  1699  		sign = -sign
  1700  		name = name[1:]
  1701  	}
  1702  
  1703  	tab, fold := unicodeTable(name)
  1704  	if tab == nil {
  1705  		return nil, "", &Error{ErrInvalidCharRange, seq}
  1706  	}
  1707  
  1708  	if p.flags&FoldCase == 0 || fold == nil {
  1709  		if sign > 0 {
  1710  			r = appendTable(r, tab)
  1711  		} else {
  1712  			r = appendNegatedTable(r, tab)
  1713  		}
  1714  	} else {
  1715  		// Merge and clean tab and fold in a temporary buffer.
  1716  		// This is necessary for the negative case and just tidy
  1717  		// for the positive case.
  1718  		tmp := p.tmpClass[:0]
  1719  		tmp = appendTable(tmp, tab)
  1720  		tmp = appendTable(tmp, fold)
  1721  		p.tmpClass = tmp
  1722  		tmp = cleanClass(&p.tmpClass)
  1723  		if sign > 0 {
  1724  			r = appendClass(r, tmp)
  1725  		} else {
  1726  			r = appendNegatedClass(r, tmp)
  1727  		}
  1728  	}
  1729  	return r, t, nil
  1730  }
  1731  
  1732  // parseClass parses a character class at the beginning of s
  1733  // and pushes it onto the parse stack.
  1734  func (p *parser) parseClass(s string) (rest string, err error) {
  1735  	t := s[1:] // chop [
  1736  	re := p.newRegexp(OpCharClass)
  1737  	re.Flags = p.flags
  1738  	re.Rune = re.Rune0[:0]
  1739  
  1740  	sign := +1
  1741  	if t != "" && t[0] == '^' {
  1742  		sign = -1
  1743  		t = t[1:]
  1744  
  1745  		// If character class does not match \n, add it here,
  1746  		// so that negation later will do the right thing.
  1747  		if p.flags&ClassNL == 0 {
  1748  			re.Rune = append(re.Rune, '\n', '\n')
  1749  		}
  1750  	}
  1751  
  1752  	class := re.Rune
  1753  	first := true // ] and - are okay as first char in class
  1754  	for t == "" || t[0] != ']' || first {
  1755  		// POSIX: - is only okay unescaped as first or last in class.
  1756  		// Perl: - is okay anywhere.
  1757  		if t != "" && t[0] == '-' && p.flags&PerlX == 0 && !first && (len(t) == 1 || t[1] != ']') {
  1758  			_, size := utf8.DecodeRuneInString(t[1:])
  1759  			return "", &Error{Code: ErrInvalidCharRange, Expr: t[:1+size]}
  1760  		}
  1761  		first = false
  1762  
  1763  		// Look for POSIX [:alnum:] etc.
  1764  		if len(t) > 2 && t[0] == '[' && t[1] == ':' {
  1765  			nclass, nt, err := p.parseNamedClass(t, class)
  1766  			if err != nil {
  1767  				return "", err
  1768  			}
  1769  			if nclass != nil {
  1770  				class, t = nclass, nt
  1771  				continue
  1772  			}
  1773  		}
  1774  
  1775  		// Look for Unicode character group like \p{Han}.
  1776  		nclass, nt, err := p.parseUnicodeClass(t, class)
  1777  		if err != nil {
  1778  			return "", err
  1779  		}
  1780  		if nclass != nil {
  1781  			class, t = nclass, nt
  1782  			continue
  1783  		}
  1784  
  1785  		// Look for Perl character class symbols (extension).
  1786  		if nclass, nt := p.parsePerlClassEscape(t, class); nclass != nil {
  1787  			class, t = nclass, nt
  1788  			continue
  1789  		}
  1790  
  1791  		// Single character or simple range.
  1792  		rng := t
  1793  		var lo, hi rune
  1794  		if lo, t, err = p.parseClassChar(t, s); err != nil {
  1795  			return "", err
  1796  		}
  1797  		hi = lo
  1798  		// [a-] means (a|-) so check for final ].
  1799  		if len(t) >= 2 && t[0] == '-' && t[1] != ']' {
  1800  			t = t[1:]
  1801  			if hi, t, err = p.parseClassChar(t, s); err != nil {
  1802  				return "", err
  1803  			}
  1804  			if hi < lo {
  1805  				rng = rng[:len(rng)-len(t)]
  1806  				return "", &Error{Code: ErrInvalidCharRange, Expr: rng}
  1807  			}
  1808  		}
  1809  		if p.flags&FoldCase == 0 {
  1810  			class = appendRange(class, lo, hi)
  1811  		} else {
  1812  			class = appendFoldedRange(class, lo, hi)
  1813  		}
  1814  	}
  1815  	t = t[1:] // chop ]
  1816  
  1817  	// Use &re.Rune instead of &class to avoid allocation.
  1818  	re.Rune = class
  1819  	class = cleanClass(&re.Rune)
  1820  	if sign < 0 {
  1821  		class = negateClass(class)
  1822  	}
  1823  	re.Rune = class
  1824  	p.push(re)
  1825  	return t, nil
  1826  }
  1827  
  1828  // cleanClass sorts the ranges (pairs of elements of r),
  1829  // merges them, and eliminates duplicates.
  1830  func cleanClass(rp *[]rune) []rune {
  1831  
  1832  	// Sort by lo increasing, hi decreasing to break ties.
  1833  	sort.Sort(ranges{rp})
  1834  
  1835  	r := *rp
  1836  	if len(r) < 2 {
  1837  		return r
  1838  	}
  1839  
  1840  	// Merge abutting, overlapping.
  1841  	w := 2 // write index
  1842  	for i := 2; i < len(r); i += 2 {
  1843  		lo, hi := r[i], r[i+1]
  1844  		if lo <= r[w-1]+1 {
  1845  			// merge with previous range
  1846  			if hi > r[w-1] {
  1847  				r[w-1] = hi
  1848  			}
  1849  			continue
  1850  		}
  1851  		// new disjoint range
  1852  		r[w] = lo
  1853  		r[w+1] = hi
  1854  		w += 2
  1855  	}
  1856  
  1857  	return r[:w]
  1858  }
  1859  
  1860  // inCharClass reports whether r is in the class.
  1861  // It assumes the class has been cleaned by cleanClass.
  1862  func inCharClass(r rune, class []rune) bool {
  1863  	_, ok := sort.Find(len(class)/2, func(i int) int {
  1864  		lo, hi := class[2*i], class[2*i+1]
  1865  		if r > hi {
  1866  			return +1
  1867  		}
  1868  		if r < lo {
  1869  			return -1
  1870  		}
  1871  		return 0
  1872  	})
  1873  	return ok
  1874  }
  1875  
  1876  // appendLiteral returns the result of appending the literal x to the class r.
  1877  func appendLiteral(r []rune, x rune, flags Flags) []rune {
  1878  	if flags&FoldCase != 0 {
  1879  		return appendFoldedRange(r, x, x)
  1880  	}
  1881  	return appendRange(r, x, x)
  1882  }
  1883  
  1884  // appendRange returns the result of appending the range lo-hi to the class r.
  1885  func appendRange(r []rune, lo, hi rune) []rune {
  1886  	// Expand last range or next to last range if it overlaps or abuts.
  1887  	// Checking two ranges helps when appending case-folded
  1888  	// alphabets, so that one range can be expanding A-Z and the
  1889  	// other expanding a-z.
  1890  	n := len(r)
  1891  	for i := 2; i <= 4; i += 2 { // twice, using i=2, i=4
  1892  		if n >= i {
  1893  			rlo, rhi := r[n-i], r[n-i+1]
  1894  			if lo <= rhi+1 && rlo <= hi+1 {
  1895  				if lo < rlo {
  1896  					r[n-i] = lo
  1897  				}
  1898  				if hi > rhi {
  1899  					r[n-i+1] = hi
  1900  				}
  1901  				return r
  1902  			}
  1903  		}
  1904  	}
  1905  
  1906  	return append(r, lo, hi)
  1907  }
  1908  
  1909  const (
  1910  	// minimum and maximum runes involved in folding.
  1911  	// checked during test.
  1912  	minFold = 0x0041
  1913  	maxFold = 0x1e943
  1914  )
  1915  
  1916  // appendFoldedRange returns the result of appending the range lo-hi
  1917  // and its case folding-equivalent runes to the class r.
  1918  func appendFoldedRange(r []rune, lo, hi rune) []rune {
  1919  	// Optimizations.
  1920  	if lo <= minFold && hi >= maxFold {
  1921  		// Range is full: folding can't add more.
  1922  		return appendRange(r, lo, hi)
  1923  	}
  1924  	if hi < minFold || lo > maxFold {
  1925  		// Range is outside folding possibilities.
  1926  		return appendRange(r, lo, hi)
  1927  	}
  1928  	if lo < minFold {
  1929  		// [lo, minFold-1] needs no folding.
  1930  		r = appendRange(r, lo, minFold-1)
  1931  		lo = minFold
  1932  	}
  1933  	if hi > maxFold {
  1934  		// [maxFold+1, hi] needs no folding.
  1935  		r = appendRange(r, maxFold+1, hi)
  1936  		hi = maxFold
  1937  	}
  1938  
  1939  	// Brute force. Depend on appendRange to coalesce ranges on the fly.
  1940  	for c := lo; c <= hi; c++ {
  1941  		r = appendRange(r, c, c)
  1942  		f := unicode.SimpleFold(c)
  1943  		for f != c {
  1944  			r = appendRange(r, f, f)
  1945  			f = unicode.SimpleFold(f)
  1946  		}
  1947  	}
  1948  	return r
  1949  }
  1950  
  1951  // appendClass returns the result of appending the class x to the class r.
  1952  // It assume x is clean.
  1953  func appendClass(r []rune, x []rune) []rune {
  1954  	for i := 0; i < len(x); i += 2 {
  1955  		r = appendRange(r, x[i], x[i+1])
  1956  	}
  1957  	return r
  1958  }
  1959  
  1960  // appendFoldedClass returns the result of appending the case folding of the class x to the class r.
  1961  func appendFoldedClass(r []rune, x []rune) []rune {
  1962  	for i := 0; i < len(x); i += 2 {
  1963  		r = appendFoldedRange(r, x[i], x[i+1])
  1964  	}
  1965  	return r
  1966  }
  1967  
  1968  // appendNegatedClass returns the result of appending the negation of the class x to the class r.
  1969  // It assumes x is clean.
  1970  func appendNegatedClass(r []rune, x []rune) []rune {
  1971  	nextLo := '\u0000'
  1972  	for i := 0; i < len(x); i += 2 {
  1973  		lo, hi := x[i], x[i+1]
  1974  		if nextLo <= lo-1 {
  1975  			r = appendRange(r, nextLo, lo-1)
  1976  		}
  1977  		nextLo = hi + 1
  1978  	}
  1979  	if nextLo <= unicode.MaxRune {
  1980  		r = appendRange(r, nextLo, unicode.MaxRune)
  1981  	}
  1982  	return r
  1983  }
  1984  
  1985  // appendTable returns the result of appending x to the class r.
  1986  func appendTable(r []rune, x *unicode.RangeTable) []rune {
  1987  	for _, xr := range x.R16 {
  1988  		lo, hi, stride := rune(xr.Lo), rune(xr.Hi), rune(xr.Stride)
  1989  		if stride == 1 {
  1990  			r = appendRange(r, lo, hi)
  1991  			continue
  1992  		}
  1993  		for c := lo; c <= hi; c += stride {
  1994  			r = appendRange(r, c, c)
  1995  		}
  1996  	}
  1997  	for _, xr := range x.R32 {
  1998  		lo, hi, stride := rune(xr.Lo), rune(xr.Hi), rune(xr.Stride)
  1999  		if stride == 1 {
  2000  			r = appendRange(r, lo, hi)
  2001  			continue
  2002  		}
  2003  		for c := lo; c <= hi; c += stride {
  2004  			r = appendRange(r, c, c)
  2005  		}
  2006  	}
  2007  	return r
  2008  }
  2009  
  2010  // appendNegatedTable returns the result of appending the negation of x to the class r.
  2011  func appendNegatedTable(r []rune, x *unicode.RangeTable) []rune {
  2012  	nextLo := '\u0000' // lo end of next class to add
  2013  	for _, xr := range x.R16 {
  2014  		lo, hi, stride := rune(xr.Lo), rune(xr.Hi), rune(xr.Stride)
  2015  		if stride == 1 {
  2016  			if nextLo <= lo-1 {
  2017  				r = appendRange(r, nextLo, lo-1)
  2018  			}
  2019  			nextLo = hi + 1
  2020  			continue
  2021  		}
  2022  		for c := lo; c <= hi; c += stride {
  2023  			if nextLo <= c-1 {
  2024  				r = appendRange(r, nextLo, c-1)
  2025  			}
  2026  			nextLo = c + 1
  2027  		}
  2028  	}
  2029  	for _, xr := range x.R32 {
  2030  		lo, hi, stride := rune(xr.Lo), rune(xr.Hi), rune(xr.Stride)
  2031  		if stride == 1 {
  2032  			if nextLo <= lo-1 {
  2033  				r = appendRange(r, nextLo, lo-1)
  2034  			}
  2035  			nextLo = hi + 1
  2036  			continue
  2037  		}
  2038  		for c := lo; c <= hi; c += stride {
  2039  			if nextLo <= c-1 {
  2040  				r = appendRange(r, nextLo, c-1)
  2041  			}
  2042  			nextLo = c + 1
  2043  		}
  2044  	}
  2045  	if nextLo <= unicode.MaxRune {
  2046  		r = appendRange(r, nextLo, unicode.MaxRune)
  2047  	}
  2048  	return r
  2049  }
  2050  
  2051  // negateClass overwrites r and returns r's negation.
  2052  // It assumes the class r is already clean.
  2053  func negateClass(r []rune) []rune {
  2054  	nextLo := '\u0000' // lo end of next class to add
  2055  	w := 0             // write index
  2056  	for i := 0; i < len(r); i += 2 {
  2057  		lo, hi := r[i], r[i+1]
  2058  		if nextLo <= lo-1 {
  2059  			r[w] = nextLo
  2060  			r[w+1] = lo - 1
  2061  			w += 2
  2062  		}
  2063  		nextLo = hi + 1
  2064  	}
  2065  	r = r[:w]
  2066  	if nextLo <= unicode.MaxRune {
  2067  		// It's possible for the negation to have one more
  2068  		// range - this one - than the original class, so use append.
  2069  		r = append(r, nextLo, unicode.MaxRune)
  2070  	}
  2071  	return r
  2072  }
  2073  
  2074  // ranges implements sort.Interface on a []rune.
  2075  // The choice of receiver type definition is strange
  2076  // but avoids an allocation since we already have
  2077  // a *[]rune.
  2078  type ranges struct {
  2079  	p *[]rune
  2080  }
  2081  
  2082  func (ra ranges) Less(i, j int) bool {
  2083  	p := *ra.p
  2084  	i *= 2
  2085  	j *= 2
  2086  	return p[i] < p[j] || p[i] == p[j] && p[i+1] > p[j+1]
  2087  }
  2088  
  2089  func (ra ranges) Len() int {
  2090  	return len(*ra.p) / 2
  2091  }
  2092  
  2093  func (ra ranges) Swap(i, j int) {
  2094  	p := *ra.p
  2095  	i *= 2
  2096  	j *= 2
  2097  	p[i], p[i+1], p[j], p[j+1] = p[j], p[j+1], p[i], p[i+1]
  2098  }
  2099  
  2100  func checkUTF8(s string) error {
  2101  	for s != "" {
  2102  		rune, size := utf8.DecodeRuneInString(s)
  2103  		if rune == utf8.RuneError && size == 1 {
  2104  			return &Error{Code: ErrInvalidUTF8, Expr: s}
  2105  		}
  2106  		s = s[size:]
  2107  	}
  2108  	return nil
  2109  }
  2110  
  2111  func nextRune(s string) (c rune, t string, err error) {
  2112  	c, size := utf8.DecodeRuneInString(s)
  2113  	if c == utf8.RuneError && size == 1 {
  2114  		return 0, "", &Error{Code: ErrInvalidUTF8, Expr: s}
  2115  	}
  2116  	return c, s[size:], nil
  2117  }
  2118  
  2119  func isalnum(c rune) bool {
  2120  	return '0' <= c && c <= '9' || 'A' <= c && c <= 'Z' || 'a' <= c && c <= 'z'
  2121  }
  2122  
  2123  func unhex(c rune) rune {
  2124  	if '0' <= c && c <= '9' {
  2125  		return c - '0'
  2126  	}
  2127  	if 'a' <= c && c <= 'f' {
  2128  		return c - 'a' + 10
  2129  	}
  2130  	if 'A' <= c && c <= 'F' {
  2131  		return c - 'A' + 10
  2132  	}
  2133  	return -1
  2134  }
  2135  

View as plain text