...
Run Format

Source file src/regexp/syntax/parse.go

Documentation: regexp/syntax

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

View as plain text