...
Run Format

Source file src/bytes/bytes.go

Documentation: bytes

     1  // Copyright 2009 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 bytes implements functions for the manipulation of byte slices.
     6  // It is analogous to the facilities of the strings package.
     7  package bytes
     8  
     9  import (
    10  	"internal/bytealg"
    11  	"unicode"
    12  	"unicode/utf8"
    13  )
    14  
    15  func equalPortable(a, b []byte) bool {
    16  	if len(a) != len(b) {
    17  		return false
    18  	}
    19  	for i, c := range a {
    20  		if c != b[i] {
    21  			return false
    22  		}
    23  	}
    24  	return true
    25  }
    26  
    27  // explode splits s into a slice of UTF-8 sequences, one per Unicode code point (still slices of bytes),
    28  // up to a maximum of n byte slices. Invalid UTF-8 sequences are chopped into individual bytes.
    29  func explode(s []byte, n int) [][]byte {
    30  	if n <= 0 {
    31  		n = len(s)
    32  	}
    33  	a := make([][]byte, n)
    34  	var size int
    35  	na := 0
    36  	for len(s) > 0 {
    37  		if na+1 >= n {
    38  			a[na] = s
    39  			na++
    40  			break
    41  		}
    42  		_, size = utf8.DecodeRune(s)
    43  		a[na] = s[0:size:size]
    44  		s = s[size:]
    45  		na++
    46  	}
    47  	return a[0:na]
    48  }
    49  
    50  // Count counts the number of non-overlapping instances of sep in s.
    51  // If sep is an empty slice, Count returns 1 + the number of UTF-8-encoded code points in s.
    52  func Count(s, sep []byte) int {
    53  	// special case
    54  	if len(sep) == 0 {
    55  		return utf8.RuneCount(s) + 1
    56  	}
    57  	if len(sep) == 1 {
    58  		return bytealg.Count(s, sep[0])
    59  	}
    60  	n := 0
    61  	for {
    62  		i := Index(s, sep)
    63  		if i == -1 {
    64  			return n
    65  		}
    66  		n++
    67  		s = s[i+len(sep):]
    68  	}
    69  }
    70  
    71  // Contains reports whether subslice is within b.
    72  func Contains(b, subslice []byte) bool {
    73  	return Index(b, subslice) != -1
    74  }
    75  
    76  // ContainsAny reports whether any of the UTF-8-encoded code points in chars are within b.
    77  func ContainsAny(b []byte, chars string) bool {
    78  	return IndexAny(b, chars) >= 0
    79  }
    80  
    81  // ContainsRune reports whether the rune is contained in the UTF-8-encoded byte slice b.
    82  func ContainsRune(b []byte, r rune) bool {
    83  	return IndexRune(b, r) >= 0
    84  }
    85  
    86  func indexBytePortable(s []byte, c byte) int {
    87  	for i, b := range s {
    88  		if b == c {
    89  			return i
    90  		}
    91  	}
    92  	return -1
    93  }
    94  
    95  // LastIndex returns the index of the last instance of sep in s, or -1 if sep is not present in s.
    96  func LastIndex(s, sep []byte) int {
    97  	n := len(sep)
    98  	if n == 0 {
    99  		return len(s)
   100  	}
   101  	c := sep[0]
   102  	for i := len(s) - n; i >= 0; i-- {
   103  		if s[i] == c && (n == 1 || Equal(s[i:i+n], sep)) {
   104  			return i
   105  		}
   106  	}
   107  	return -1
   108  }
   109  
   110  // LastIndexByte returns the index of the last instance of c in s, or -1 if c is not present in s.
   111  func LastIndexByte(s []byte, c byte) int {
   112  	for i := len(s) - 1; i >= 0; i-- {
   113  		if s[i] == c {
   114  			return i
   115  		}
   116  	}
   117  	return -1
   118  }
   119  
   120  // IndexRune interprets s as a sequence of UTF-8-encoded code points.
   121  // It returns the byte index of the first occurrence in s of the given rune.
   122  // It returns -1 if rune is not present in s.
   123  // If r is utf8.RuneError, it returns the first instance of any
   124  // invalid UTF-8 byte sequence.
   125  func IndexRune(s []byte, r rune) int {
   126  	switch {
   127  	case 0 <= r && r < utf8.RuneSelf:
   128  		return IndexByte(s, byte(r))
   129  	case r == utf8.RuneError:
   130  		for i := 0; i < len(s); {
   131  			r1, n := utf8.DecodeRune(s[i:])
   132  			if r1 == utf8.RuneError {
   133  				return i
   134  			}
   135  			i += n
   136  		}
   137  		return -1
   138  	case !utf8.ValidRune(r):
   139  		return -1
   140  	default:
   141  		var b [utf8.UTFMax]byte
   142  		n := utf8.EncodeRune(b[:], r)
   143  		return Index(s, b[:n])
   144  	}
   145  }
   146  
   147  // IndexAny interprets s as a sequence of UTF-8-encoded Unicode code points.
   148  // It returns the byte index of the first occurrence in s of any of the Unicode
   149  // code points in chars. It returns -1 if chars is empty or if there is no code
   150  // point in common.
   151  func IndexAny(s []byte, chars string) int {
   152  	if chars == "" {
   153  		// Avoid scanning all of s.
   154  		return -1
   155  	}
   156  	if len(s) > 8 {
   157  		if as, isASCII := makeASCIISet(chars); isASCII {
   158  			for i, c := range s {
   159  				if as.contains(c) {
   160  					return i
   161  				}
   162  			}
   163  			return -1
   164  		}
   165  	}
   166  	var width int
   167  	for i := 0; i < len(s); i += width {
   168  		r := rune(s[i])
   169  		if r < utf8.RuneSelf {
   170  			width = 1
   171  		} else {
   172  			r, width = utf8.DecodeRune(s[i:])
   173  		}
   174  		for _, ch := range chars {
   175  			if r == ch {
   176  				return i
   177  			}
   178  		}
   179  	}
   180  	return -1
   181  }
   182  
   183  // LastIndexAny interprets s as a sequence of UTF-8-encoded Unicode code
   184  // points. It returns the byte index of the last occurrence in s of any of
   185  // the Unicode code points in chars. It returns -1 if chars is empty or if
   186  // there is no code point in common.
   187  func LastIndexAny(s []byte, chars string) int {
   188  	if chars == "" {
   189  		// Avoid scanning all of s.
   190  		return -1
   191  	}
   192  	if len(s) > 8 {
   193  		if as, isASCII := makeASCIISet(chars); isASCII {
   194  			for i := len(s) - 1; i >= 0; i-- {
   195  				if as.contains(s[i]) {
   196  					return i
   197  				}
   198  			}
   199  			return -1
   200  		}
   201  	}
   202  	for i := len(s); i > 0; {
   203  		r, size := utf8.DecodeLastRune(s[:i])
   204  		i -= size
   205  		for _, c := range chars {
   206  			if r == c {
   207  				return i
   208  			}
   209  		}
   210  	}
   211  	return -1
   212  }
   213  
   214  // Generic split: splits after each instance of sep,
   215  // including sepSave bytes of sep in the subslices.
   216  func genSplit(s, sep []byte, sepSave, n int) [][]byte {
   217  	if n == 0 {
   218  		return nil
   219  	}
   220  	if len(sep) == 0 {
   221  		return explode(s, n)
   222  	}
   223  	if n < 0 {
   224  		n = Count(s, sep) + 1
   225  	}
   226  
   227  	a := make([][]byte, n)
   228  	n--
   229  	i := 0
   230  	for i < n {
   231  		m := Index(s, sep)
   232  		if m < 0 {
   233  			break
   234  		}
   235  		a[i] = s[: m+sepSave : m+sepSave]
   236  		s = s[m+len(sep):]
   237  		i++
   238  	}
   239  	a[i] = s
   240  	return a[:i+1]
   241  }
   242  
   243  // SplitN slices s into subslices separated by sep and returns a slice of
   244  // the subslices between those separators.
   245  // If sep is empty, SplitN splits after each UTF-8 sequence.
   246  // The count determines the number of subslices to return:
   247  //   n > 0: at most n subslices; the last subslice will be the unsplit remainder.
   248  //   n == 0: the result is nil (zero subslices)
   249  //   n < 0: all subslices
   250  func SplitN(s, sep []byte, n int) [][]byte { return genSplit(s, sep, 0, n) }
   251  
   252  // SplitAfterN slices s into subslices after each instance of sep and
   253  // returns a slice of those subslices.
   254  // If sep is empty, SplitAfterN splits after each UTF-8 sequence.
   255  // The count determines the number of subslices to return:
   256  //   n > 0: at most n subslices; the last subslice will be the unsplit remainder.
   257  //   n == 0: the result is nil (zero subslices)
   258  //   n < 0: all subslices
   259  func SplitAfterN(s, sep []byte, n int) [][]byte {
   260  	return genSplit(s, sep, len(sep), n)
   261  }
   262  
   263  // Split slices s into all subslices separated by sep and returns a slice of
   264  // the subslices between those separators.
   265  // If sep is empty, Split splits after each UTF-8 sequence.
   266  // It is equivalent to SplitN with a count of -1.
   267  func Split(s, sep []byte) [][]byte { return genSplit(s, sep, 0, -1) }
   268  
   269  // SplitAfter slices s into all subslices after each instance of sep and
   270  // returns a slice of those subslices.
   271  // If sep is empty, SplitAfter splits after each UTF-8 sequence.
   272  // It is equivalent to SplitAfterN with a count of -1.
   273  func SplitAfter(s, sep []byte) [][]byte {
   274  	return genSplit(s, sep, len(sep), -1)
   275  }
   276  
   277  var asciiSpace = [256]uint8{'\t': 1, '\n': 1, '\v': 1, '\f': 1, '\r': 1, ' ': 1}
   278  
   279  // Fields interprets s as a sequence of UTF-8-encoded code points.
   280  // It splits the slice s around each instance of one or more consecutive white space
   281  // characters, as defined by unicode.IsSpace, returning a slice of subslices of s or an
   282  // empty slice if s contains only white space.
   283  func Fields(s []byte) [][]byte {
   284  	// First count the fields.
   285  	// This is an exact count if s is ASCII, otherwise it is an approximation.
   286  	n := 0
   287  	wasSpace := 1
   288  	// setBits is used to track which bits are set in the bytes of s.
   289  	setBits := uint8(0)
   290  	for i := 0; i < len(s); i++ {
   291  		r := s[i]
   292  		setBits |= r
   293  		isSpace := int(asciiSpace[r])
   294  		n += wasSpace & ^isSpace
   295  		wasSpace = isSpace
   296  	}
   297  
   298  	if setBits >= utf8.RuneSelf {
   299  		// Some runes in the input slice are not ASCII.
   300  		return FieldsFunc(s, unicode.IsSpace)
   301  	}
   302  
   303  	// ASCII fast path
   304  	a := make([][]byte, n)
   305  	na := 0
   306  	fieldStart := 0
   307  	i := 0
   308  	// Skip spaces in the front of the input.
   309  	for i < len(s) && asciiSpace[s[i]] != 0 {
   310  		i++
   311  	}
   312  	fieldStart = i
   313  	for i < len(s) {
   314  		if asciiSpace[s[i]] == 0 {
   315  			i++
   316  			continue
   317  		}
   318  		a[na] = s[fieldStart:i:i]
   319  		na++
   320  		i++
   321  		// Skip spaces in between fields.
   322  		for i < len(s) && asciiSpace[s[i]] != 0 {
   323  			i++
   324  		}
   325  		fieldStart = i
   326  	}
   327  	if fieldStart < len(s) { // Last field might end at EOF.
   328  		a[na] = s[fieldStart:len(s):len(s)]
   329  	}
   330  	return a
   331  }
   332  
   333  // FieldsFunc interprets s as a sequence of UTF-8-encoded code points.
   334  // It splits the slice s at each run of code points c satisfying f(c) and
   335  // returns a slice of subslices of s. If all code points in s satisfy f(c), or
   336  // len(s) == 0, an empty slice is returned.
   337  // FieldsFunc makes no guarantees about the order in which it calls f(c).
   338  // If f does not return consistent results for a given c, FieldsFunc may crash.
   339  func FieldsFunc(s []byte, f func(rune) bool) [][]byte {
   340  	// A span is used to record a slice of s of the form s[start:end].
   341  	// The start index is inclusive and the end index is exclusive.
   342  	type span struct {
   343  		start int
   344  		end   int
   345  	}
   346  	spans := make([]span, 0, 32)
   347  
   348  	// Find the field start and end indices.
   349  	wasField := false
   350  	fromIndex := 0
   351  	for i := 0; i < len(s); {
   352  		size := 1
   353  		r := rune(s[i])
   354  		if r >= utf8.RuneSelf {
   355  			r, size = utf8.DecodeRune(s[i:])
   356  		}
   357  		if f(r) {
   358  			if wasField {
   359  				spans = append(spans, span{start: fromIndex, end: i})
   360  				wasField = false
   361  			}
   362  		} else {
   363  			if !wasField {
   364  				fromIndex = i
   365  				wasField = true
   366  			}
   367  		}
   368  		i += size
   369  	}
   370  
   371  	// Last field might end at EOF.
   372  	if wasField {
   373  		spans = append(spans, span{fromIndex, len(s)})
   374  	}
   375  
   376  	// Create subslices from recorded field indices.
   377  	a := make([][]byte, len(spans))
   378  	for i, span := range spans {
   379  		a[i] = s[span.start:span.end:span.end]
   380  	}
   381  
   382  	return a
   383  }
   384  
   385  // Join concatenates the elements of s to create a new byte slice. The separator
   386  // sep is placed between elements in the resulting slice.
   387  func Join(s [][]byte, sep []byte) []byte {
   388  	if len(s) == 0 {
   389  		return []byte{}
   390  	}
   391  	if len(s) == 1 {
   392  		// Just return a copy.
   393  		return append([]byte(nil), s[0]...)
   394  	}
   395  	n := len(sep) * (len(s) - 1)
   396  	for _, v := range s {
   397  		n += len(v)
   398  	}
   399  
   400  	b := make([]byte, n)
   401  	bp := copy(b, s[0])
   402  	for _, v := range s[1:] {
   403  		bp += copy(b[bp:], sep)
   404  		bp += copy(b[bp:], v)
   405  	}
   406  	return b
   407  }
   408  
   409  // HasPrefix tests whether the byte slice s begins with prefix.
   410  func HasPrefix(s, prefix []byte) bool {
   411  	return len(s) >= len(prefix) && Equal(s[0:len(prefix)], prefix)
   412  }
   413  
   414  // HasSuffix tests whether the byte slice s ends with suffix.
   415  func HasSuffix(s, suffix []byte) bool {
   416  	return len(s) >= len(suffix) && Equal(s[len(s)-len(suffix):], suffix)
   417  }
   418  
   419  // Map returns a copy of the byte slice s with all its characters modified
   420  // according to the mapping function. If mapping returns a negative value, the character is
   421  // dropped from the byte slice with no replacement. The characters in s and the
   422  // output are interpreted as UTF-8-encoded code points.
   423  func Map(mapping func(r rune) rune, s []byte) []byte {
   424  	// In the worst case, the slice can grow when mapped, making
   425  	// things unpleasant. But it's so rare we barge in assuming it's
   426  	// fine. It could also shrink but that falls out naturally.
   427  	maxbytes := len(s) // length of b
   428  	nbytes := 0        // number of bytes encoded in b
   429  	b := make([]byte, maxbytes)
   430  	for i := 0; i < len(s); {
   431  		wid := 1
   432  		r := rune(s[i])
   433  		if r >= utf8.RuneSelf {
   434  			r, wid = utf8.DecodeRune(s[i:])
   435  		}
   436  		r = mapping(r)
   437  		if r >= 0 {
   438  			rl := utf8.RuneLen(r)
   439  			if rl < 0 {
   440  				rl = len(string(utf8.RuneError))
   441  			}
   442  			if nbytes+rl > maxbytes {
   443  				// Grow the buffer.
   444  				maxbytes = maxbytes*2 + utf8.UTFMax
   445  				nb := make([]byte, maxbytes)
   446  				copy(nb, b[0:nbytes])
   447  				b = nb
   448  			}
   449  			nbytes += utf8.EncodeRune(b[nbytes:maxbytes], r)
   450  		}
   451  		i += wid
   452  	}
   453  	return b[0:nbytes]
   454  }
   455  
   456  // Repeat returns a new byte slice consisting of count copies of b.
   457  //
   458  // It panics if count is negative or if
   459  // the result of (len(b) * count) overflows.
   460  func Repeat(b []byte, count int) []byte {
   461  	// Since we cannot return an error on overflow,
   462  	// we should panic if the repeat will generate
   463  	// an overflow.
   464  	// See Issue golang.org/issue/16237.
   465  	if count < 0 {
   466  		panic("bytes: negative Repeat count")
   467  	} else if count > 0 && len(b)*count/count != len(b) {
   468  		panic("bytes: Repeat count causes overflow")
   469  	}
   470  
   471  	nb := make([]byte, len(b)*count)
   472  	bp := copy(nb, b)
   473  	for bp < len(nb) {
   474  		copy(nb[bp:], nb[:bp])
   475  		bp *= 2
   476  	}
   477  	return nb
   478  }
   479  
   480  // ToUpper treats s as UTF-8-encoded bytes and returns a copy with all the Unicode letters within it mapped to their upper case.
   481  func ToUpper(s []byte) []byte { return Map(unicode.ToUpper, s) }
   482  
   483  // ToLower treats s as UTF-8-encoded bytes and returns a copy with all the Unicode letters mapped to their lower case.
   484  func ToLower(s []byte) []byte { return Map(unicode.ToLower, s) }
   485  
   486  // ToTitle treats s as UTF-8-encoded bytes and returns a copy with all the Unicode letters mapped to their title case.
   487  func ToTitle(s []byte) []byte { return Map(unicode.ToTitle, s) }
   488  
   489  // ToUpperSpecial treats s as UTF-8-encoded bytes and returns a copy with all the Unicode letters mapped to their
   490  // upper case, giving priority to the special casing rules.
   491  func ToUpperSpecial(c unicode.SpecialCase, s []byte) []byte {
   492  	return Map(func(r rune) rune { return c.ToUpper(r) }, s)
   493  }
   494  
   495  // ToLowerSpecial treats s as UTF-8-encoded bytes and returns a copy with all the Unicode letters mapped to their
   496  // lower case, giving priority to the special casing rules.
   497  func ToLowerSpecial(c unicode.SpecialCase, s []byte) []byte {
   498  	return Map(func(r rune) rune { return c.ToLower(r) }, s)
   499  }
   500  
   501  // ToTitleSpecial treats s as UTF-8-encoded bytes and returns a copy with all the Unicode letters mapped to their
   502  // title case, giving priority to the special casing rules.
   503  func ToTitleSpecial(c unicode.SpecialCase, s []byte) []byte {
   504  	return Map(func(r rune) rune { return c.ToTitle(r) }, s)
   505  }
   506  
   507  // isSeparator reports whether the rune could mark a word boundary.
   508  // TODO: update when package unicode captures more of the properties.
   509  func isSeparator(r rune) bool {
   510  	// ASCII alphanumerics and underscore are not separators
   511  	if r <= 0x7F {
   512  		switch {
   513  		case '0' <= r && r <= '9':
   514  			return false
   515  		case 'a' <= r && r <= 'z':
   516  			return false
   517  		case 'A' <= r && r <= 'Z':
   518  			return false
   519  		case r == '_':
   520  			return false
   521  		}
   522  		return true
   523  	}
   524  	// Letters and digits are not separators
   525  	if unicode.IsLetter(r) || unicode.IsDigit(r) {
   526  		return false
   527  	}
   528  	// Otherwise, all we can do for now is treat spaces as separators.
   529  	return unicode.IsSpace(r)
   530  }
   531  
   532  // Title treats s as UTF-8-encoded bytes and returns a copy with all Unicode letters that begin
   533  // words mapped to their title case.
   534  //
   535  // BUG(rsc): The rule Title uses for word boundaries does not handle Unicode punctuation properly.
   536  func Title(s []byte) []byte {
   537  	// Use a closure here to remember state.
   538  	// Hackish but effective. Depends on Map scanning in order and calling
   539  	// the closure once per rune.
   540  	prev := ' '
   541  	return Map(
   542  		func(r rune) rune {
   543  			if isSeparator(prev) {
   544  				prev = r
   545  				return unicode.ToTitle(r)
   546  			}
   547  			prev = r
   548  			return r
   549  		},
   550  		s)
   551  }
   552  
   553  // TrimLeftFunc treats s as UTF-8-encoded bytes and returns a subslice of s by slicing off
   554  // all leading UTF-8-encoded code points c that satisfy f(c).
   555  func TrimLeftFunc(s []byte, f func(r rune) bool) []byte {
   556  	i := indexFunc(s, f, false)
   557  	if i == -1 {
   558  		return nil
   559  	}
   560  	return s[i:]
   561  }
   562  
   563  // TrimRightFunc returns a subslice of s by slicing off all trailing
   564  // UTF-8-encoded code points c that satisfy f(c).
   565  func TrimRightFunc(s []byte, f func(r rune) bool) []byte {
   566  	i := lastIndexFunc(s, f, false)
   567  	if i >= 0 && s[i] >= utf8.RuneSelf {
   568  		_, wid := utf8.DecodeRune(s[i:])
   569  		i += wid
   570  	} else {
   571  		i++
   572  	}
   573  	return s[0:i]
   574  }
   575  
   576  // TrimFunc returns a subslice of s by slicing off all leading and trailing
   577  // UTF-8-encoded code points c that satisfy f(c).
   578  func TrimFunc(s []byte, f func(r rune) bool) []byte {
   579  	return TrimRightFunc(TrimLeftFunc(s, f), f)
   580  }
   581  
   582  // TrimPrefix returns s without the provided leading prefix string.
   583  // If s doesn't start with prefix, s is returned unchanged.
   584  func TrimPrefix(s, prefix []byte) []byte {
   585  	if HasPrefix(s, prefix) {
   586  		return s[len(prefix):]
   587  	}
   588  	return s
   589  }
   590  
   591  // TrimSuffix returns s without the provided trailing suffix string.
   592  // If s doesn't end with suffix, s is returned unchanged.
   593  func TrimSuffix(s, suffix []byte) []byte {
   594  	if HasSuffix(s, suffix) {
   595  		return s[:len(s)-len(suffix)]
   596  	}
   597  	return s
   598  }
   599  
   600  // IndexFunc interprets s as a sequence of UTF-8-encoded code points.
   601  // It returns the byte index in s of the first Unicode
   602  // code point satisfying f(c), or -1 if none do.
   603  func IndexFunc(s []byte, f func(r rune) bool) int {
   604  	return indexFunc(s, f, true)
   605  }
   606  
   607  // LastIndexFunc interprets s as a sequence of UTF-8-encoded code points.
   608  // It returns the byte index in s of the last Unicode
   609  // code point satisfying f(c), or -1 if none do.
   610  func LastIndexFunc(s []byte, f func(r rune) bool) int {
   611  	return lastIndexFunc(s, f, true)
   612  }
   613  
   614  // indexFunc is the same as IndexFunc except that if
   615  // truth==false, the sense of the predicate function is
   616  // inverted.
   617  func indexFunc(s []byte, f func(r rune) bool, truth bool) int {
   618  	start := 0
   619  	for start < len(s) {
   620  		wid := 1
   621  		r := rune(s[start])
   622  		if r >= utf8.RuneSelf {
   623  			r, wid = utf8.DecodeRune(s[start:])
   624  		}
   625  		if f(r) == truth {
   626  			return start
   627  		}
   628  		start += wid
   629  	}
   630  	return -1
   631  }
   632  
   633  // lastIndexFunc is the same as LastIndexFunc except that if
   634  // truth==false, the sense of the predicate function is
   635  // inverted.
   636  func lastIndexFunc(s []byte, f func(r rune) bool, truth bool) int {
   637  	for i := len(s); i > 0; {
   638  		r, size := rune(s[i-1]), 1
   639  		if r >= utf8.RuneSelf {
   640  			r, size = utf8.DecodeLastRune(s[0:i])
   641  		}
   642  		i -= size
   643  		if f(r) == truth {
   644  			return i
   645  		}
   646  	}
   647  	return -1
   648  }
   649  
   650  // asciiSet is a 32-byte value, where each bit represents the presence of a
   651  // given ASCII character in the set. The 128-bits of the lower 16 bytes,
   652  // starting with the least-significant bit of the lowest word to the
   653  // most-significant bit of the highest word, map to the full range of all
   654  // 128 ASCII characters. The 128-bits of the upper 16 bytes will be zeroed,
   655  // ensuring that any non-ASCII character will be reported as not in the set.
   656  type asciiSet [8]uint32
   657  
   658  // makeASCIISet creates a set of ASCII characters and reports whether all
   659  // characters in chars are ASCII.
   660  func makeASCIISet(chars string) (as asciiSet, ok bool) {
   661  	for i := 0; i < len(chars); i++ {
   662  		c := chars[i]
   663  		if c >= utf8.RuneSelf {
   664  			return as, false
   665  		}
   666  		as[c>>5] |= 1 << uint(c&31)
   667  	}
   668  	return as, true
   669  }
   670  
   671  // contains reports whether c is inside the set.
   672  func (as *asciiSet) contains(c byte) bool {
   673  	return (as[c>>5] & (1 << uint(c&31))) != 0
   674  }
   675  
   676  func makeCutsetFunc(cutset string) func(r rune) bool {
   677  	if len(cutset) == 1 && cutset[0] < utf8.RuneSelf {
   678  		return func(r rune) bool {
   679  			return r == rune(cutset[0])
   680  		}
   681  	}
   682  	if as, isASCII := makeASCIISet(cutset); isASCII {
   683  		return func(r rune) bool {
   684  			return r < utf8.RuneSelf && as.contains(byte(r))
   685  		}
   686  	}
   687  	return func(r rune) bool {
   688  		for _, c := range cutset {
   689  			if c == r {
   690  				return true
   691  			}
   692  		}
   693  		return false
   694  	}
   695  }
   696  
   697  // Trim returns a subslice of s by slicing off all leading and
   698  // trailing UTF-8-encoded code points contained in cutset.
   699  func Trim(s []byte, cutset string) []byte {
   700  	return TrimFunc(s, makeCutsetFunc(cutset))
   701  }
   702  
   703  // TrimLeft returns a subslice of s by slicing off all leading
   704  // UTF-8-encoded code points contained in cutset.
   705  func TrimLeft(s []byte, cutset string) []byte {
   706  	return TrimLeftFunc(s, makeCutsetFunc(cutset))
   707  }
   708  
   709  // TrimRight returns a subslice of s by slicing off all trailing
   710  // UTF-8-encoded code points that are contained in cutset.
   711  func TrimRight(s []byte, cutset string) []byte {
   712  	return TrimRightFunc(s, makeCutsetFunc(cutset))
   713  }
   714  
   715  // TrimSpace returns a subslice of s by slicing off all leading and
   716  // trailing white space, as defined by Unicode.
   717  func TrimSpace(s []byte) []byte {
   718  	return TrimFunc(s, unicode.IsSpace)
   719  }
   720  
   721  // Runes interprets s as a sequence of UTF-8-encoded code points.
   722  // It returns a slice of runes (Unicode code points) equivalent to s.
   723  func Runes(s []byte) []rune {
   724  	t := make([]rune, utf8.RuneCount(s))
   725  	i := 0
   726  	for len(s) > 0 {
   727  		r, l := utf8.DecodeRune(s)
   728  		t[i] = r
   729  		i++
   730  		s = s[l:]
   731  	}
   732  	return t
   733  }
   734  
   735  // Replace returns a copy of the slice s with the first n
   736  // non-overlapping instances of old replaced by new.
   737  // If old is empty, it matches at the beginning of the slice
   738  // and after each UTF-8 sequence, yielding up to k+1 replacements
   739  // for a k-rune slice.
   740  // If n < 0, there is no limit on the number of replacements.
   741  func Replace(s, old, new []byte, n int) []byte {
   742  	m := 0
   743  	if n != 0 {
   744  		// Compute number of replacements.
   745  		m = Count(s, old)
   746  	}
   747  	if m == 0 {
   748  		// Just return a copy.
   749  		return append([]byte(nil), s...)
   750  	}
   751  	if n < 0 || m < n {
   752  		n = m
   753  	}
   754  
   755  	// Apply replacements to buffer.
   756  	t := make([]byte, len(s)+n*(len(new)-len(old)))
   757  	w := 0
   758  	start := 0
   759  	for i := 0; i < n; i++ {
   760  		j := start
   761  		if len(old) == 0 {
   762  			if i > 0 {
   763  				_, wid := utf8.DecodeRune(s[start:])
   764  				j += wid
   765  			}
   766  		} else {
   767  			j += Index(s[start:], old)
   768  		}
   769  		w += copy(t[w:], s[start:j])
   770  		w += copy(t[w:], new)
   771  		start = j + len(old)
   772  	}
   773  	w += copy(t[w:], s[start:])
   774  	return t[0:w]
   775  }
   776  
   777  // EqualFold reports whether s and t, interpreted as UTF-8 strings,
   778  // are equal under Unicode case-folding.
   779  func EqualFold(s, t []byte) bool {
   780  	for len(s) != 0 && len(t) != 0 {
   781  		// Extract first rune from each.
   782  		var sr, tr rune
   783  		if s[0] < utf8.RuneSelf {
   784  			sr, s = rune(s[0]), s[1:]
   785  		} else {
   786  			r, size := utf8.DecodeRune(s)
   787  			sr, s = r, s[size:]
   788  		}
   789  		if t[0] < utf8.RuneSelf {
   790  			tr, t = rune(t[0]), t[1:]
   791  		} else {
   792  			r, size := utf8.DecodeRune(t)
   793  			tr, t = r, t[size:]
   794  		}
   795  
   796  		// If they match, keep going; if not, return false.
   797  
   798  		// Easy case.
   799  		if tr == sr {
   800  			continue
   801  		}
   802  
   803  		// Make sr < tr to simplify what follows.
   804  		if tr < sr {
   805  			tr, sr = sr, tr
   806  		}
   807  		// Fast check for ASCII.
   808  		if tr < utf8.RuneSelf {
   809  			// ASCII only, sr/tr must be upper/lower case
   810  			if 'A' <= sr && sr <= 'Z' && tr == sr+'a'-'A' {
   811  				continue
   812  			}
   813  			return false
   814  		}
   815  
   816  		// General case. SimpleFold(x) returns the next equivalent rune > x
   817  		// or wraps around to smaller values.
   818  		r := unicode.SimpleFold(sr)
   819  		for r != sr && r < tr {
   820  			r = unicode.SimpleFold(r)
   821  		}
   822  		if r == tr {
   823  			continue
   824  		}
   825  		return false
   826  	}
   827  
   828  	// One string is empty. Are both?
   829  	return len(s) == len(t)
   830  }
   831  
   832  // Index returns the index of the first instance of sep in s, or -1 if sep is not present in s.
   833  func Index(s, sep []byte) int {
   834  	n := len(sep)
   835  	switch {
   836  	case n == 0:
   837  		return 0
   838  	case n == 1:
   839  		return IndexByte(s, sep[0])
   840  	case n == len(s):
   841  		if Equal(sep, s) {
   842  			return 0
   843  		}
   844  		return -1
   845  	case n > len(s):
   846  		return -1
   847  	case n <= bytealg.MaxLen:
   848  		// Use brute force when s and sep both are small
   849  		if len(s) <= bytealg.MaxBruteForce {
   850  			return bytealg.Index(s, sep)
   851  		}
   852  		c := sep[0]
   853  		i := 0
   854  		t := s[:len(s)-n+1]
   855  		fails := 0
   856  		for i < len(t) {
   857  			if t[i] != c {
   858  				// IndexByte is faster than bytealg.Index, so use it as long as
   859  				// we're not getting lots of false positives.
   860  				o := IndexByte(t[i:], c)
   861  				if o < 0 {
   862  					return -1
   863  				}
   864  				i += o
   865  			}
   866  			if Equal(s[i:i+n], sep) {
   867  				return i
   868  			}
   869  			fails++
   870  			i++
   871  			// Switch to bytealg.Index when IndexByte produces too many false positives.
   872  			if fails > bytealg.Cutover(i) {
   873  				r := bytealg.Index(s[i:], sep)
   874  				if r >= 0 {
   875  					return r + i
   876  				}
   877  				return -1
   878  			}
   879  		}
   880  		return -1
   881  	}
   882  	c := sep[0]
   883  	i := 0
   884  	fails := 0
   885  	t := s[:len(s)-n+1]
   886  	for i < len(t) {
   887  		if t[i] != c {
   888  			o := IndexByte(t[i:], c)
   889  			if o < 0 {
   890  				break
   891  			}
   892  			i += o
   893  		}
   894  		if Equal(s[i:i+n], sep) {
   895  			return i
   896  		}
   897  		i++
   898  		fails++
   899  		if fails >= 4+i>>4 && i < len(t) {
   900  			// Give up on IndexByte, it isn't skipping ahead
   901  			// far enough to be better than Rabin-Karp.
   902  			// Experiments (using IndexPeriodic) suggest
   903  			// the cutover is about 16 byte skips.
   904  			// TODO: if large prefixes of sep are matching
   905  			// we should cutover at even larger average skips,
   906  			// because Equal becomes that much more expensive.
   907  			// This code does not take that effect into account.
   908  			j := indexRabinKarp(s[i:], sep)
   909  			if j < 0 {
   910  				return -1
   911  			}
   912  			return i + j
   913  		}
   914  	}
   915  	return -1
   916  }
   917  
   918  func indexRabinKarp(s, sep []byte) int {
   919  	// Rabin-Karp search
   920  	hashsep, pow := hashStr(sep)
   921  	n := len(sep)
   922  	var h uint32
   923  	for i := 0; i < n; i++ {
   924  		h = h*primeRK + uint32(s[i])
   925  	}
   926  	if h == hashsep && Equal(s[:n], sep) {
   927  		return 0
   928  	}
   929  	for i := n; i < len(s); {
   930  		h *= primeRK
   931  		h += uint32(s[i])
   932  		h -= pow * uint32(s[i-n])
   933  		i++
   934  		if h == hashsep && Equal(s[i-n:i], sep) {
   935  			return i - n
   936  		}
   937  	}
   938  	return -1
   939  }
   940  
   941  // primeRK is the prime base used in Rabin-Karp algorithm.
   942  const primeRK = 16777619
   943  
   944  // hashStr returns the hash and the appropriate multiplicative
   945  // factor for use in Rabin-Karp algorithm.
   946  func hashStr(sep []byte) (uint32, uint32) {
   947  	hash := uint32(0)
   948  	for i := 0; i < len(sep); i++ {
   949  		hash = hash*primeRK + uint32(sep[i])
   950  	}
   951  	var pow, sq uint32 = 1, primeRK
   952  	for i := len(sep); i > 0; i >>= 1 {
   953  		if i&1 != 0 {
   954  			pow *= sq
   955  		}
   956  		sq *= sq
   957  	}
   958  	return hash, pow
   959  }
   960  

View as plain text